You are on page 1of 589

MSC IN ECONOMICS & BUSINESS ADMINISTRATION

APPLIED ECONOMICS AND FINANCE


DEPARTMENT OF ECONOMICS

The relationship between


equity prices and credit
default swap spreads
An empirical analysis

Handed in: 31. August 2009

Written by:

Ida Buus

Charlotte Renneberg J. Nielsen

Thesis supervisor:
Erik Haller Pedersen
The Danish National Bank

COPENHAGEN BUSINESS SCHOOL 2009


Executive summary
This thesis analyses the relationship between equity prices and credit default spreads (CDS). The
CDS is an OTC contract that provides insurance against the risk of default by a particular
company A CDS reflects the credit risk and this credit risk is also present in equity prices as
credit risk influence the value of a firm. The relationship between CDS spreads and equity prices
is interesting to examine since the nature and direction of this relationship can be used when
making economic models, issuing debt and planning arbitrage strategies and the CDS’ played a
role in the depth and increase of the crisis that started in 2008. Furthermore due to the young age
of the CDS market the available research on the subject is limited and our study will contribute
by being the most extensive as far as we know.

We analyse the relationship between equity prices and credit default spreads during the period
2nd January 2004 to 1st May 2009 for 265 firms present in S&P 500 in 2002. We examine the
daily lead-lag relationship in a vector autoregressive model (VAR) or in the case of co-
integration in a vector error correction model (VECM). Additionally we examine the Spearman
Rank correlation, Granger Causality test & Granger-Gonzalo measure.

First, we find that equity prices and CDS spreads are negatively correlated. Second, that equity
prices influence CDS spreads and not the other way around. Third, that the relationship is not
affected by including exogenous variables in the model. Fourth, that the strength of the
relationship increases the higher the credit risk. Fifth, we find that for some sectors the
relationship becomes more pronounced the higher the leverage in the sector.

We examine how an increase in credit risk influence the relationship by splitting our data into
before and after the beginning of the crisis, in rating, and quartiles. We find that the influence of
equity prices on CDS spreads is stronger in the period after the beginning of the crisis than
before. This indicates that the relationship between CDS spreads and equity prices is stronger
under deteriorating market conditions.

Furthermore we examine how the credit rating of the underlying entity affects the relationship.
We do this by splitting our data in investment grade and high yield. The influence of equity
prices on CDS spreads is stronger for the high yield model than the investment grade model.
This indicates that the relationship between CDS spreads and equity prices is stronger the lower
the credit rating of the underlying entity.

2
At last we examine how the size of the CDS spread effects the relationship. We do this by
splitting our data into quartiles. We split the data for index, before and after the beginning of the
crisis, high yield, and investment grade. We find that when comparing the quartiles against each
otherthe influence of equity prices on CDS spreads is strongest for quartile 4, second strongest
for quartile 3, third strongest for quartile 2 and least strongest for quartile 1. Moreover when
comparing the quartile models with their respective overall index we find that quartile 3 and 4
are stronger models than the overall model except for the models for high yield and investment
grade.

The overall conclusion is that there is a negative relationship between CDS spreads and equity
prices and that equity prices lead CDS spreads.

3
Acknowledgement
We would like to express our gratitude and give thanks to the people whom help us complete this
thesis.

BankInvest for kindly giving us access to Bloomberg and thereby enabling us to get high quality
data not otherwise available.

Rasmus Hansen (BankInvest) for continuously inspiration and technical support.

Kirsten Nielsen for editorial support.

Our advisor Erik Haller Pedersen for valuable comments and suggestions.

At last but not least our friends and family for moral support.

4
Table of contents
Executive summary ......................................................................................................................... 2
Acknowledgement ........................................................................................................................... 4
Introduction ................................................................................................................................... 11
1.1 Relation to equity market .................................................................................................... 12
1.2 Objectives of the study ........................................................................................................ 14
1.3 Problem ............................................................................................................................... 15
2 Delimitation ........................................................................................................................... 15
3 Structure ................................................................................................................................ 16
4 Theory ................................................................................................................................... 17
4.1 Efficient markets ................................................................................................................. 17
4.2 Credit Derivatives ............................................................................................................... 18
4.3 Credit default swaps ............................................................................................................ 19
4.3.1 Definition ..................................................................................................................... 19
4.3.2 Valuation of Credit Default Swaps .............................................................................. 22
4.3.3 The CDS market in numbers ........................................................................................ 23
4.4 Equity market ...................................................................................................................... 26
4.4.1 Definition ..................................................................................................................... 26
4.4.2 Valuation models.......................................................................................................... 27
4.5 Merton Model ...................................................................................................................... 28
5 Literature review ................................................................................................................... 30
6 Hypotheses ............................................................................................................................ 32
7 Methodology ......................................................................................................................... 35
7.1 Correlation ........................................................................................................................... 36
7.2 Autocorrelation.................................................................................................................... 36
7.3 Stationarity .......................................................................................................................... 37
7.4 Model selection ................................................................................................................... 39
7.4.1 VAR models ................................................................................................................. 39
7.4.2 VECM model ............................................................................................................... 40
7.5 Granger Causality ................................................................................................................ 41
7.6 Gonzalo and Granger measure ............................................................................................ 42
7.7 Method overview................................................................................................................. 42
8 Data description..................................................................................................................... 43

5
9 Empirical analysis ................................................................................................................. 46
9.1 Index .................................................................................................................................... 47
9.1.1 Model with CDS spreads as dependent variable .......................................................... 49
9.1.2 Model with equity price as dependent variable ............................................................ 50
9.2 Exogenous variables ............................................................................................................ 50
9.2.1 Model with CDS spreads as dependent variable .......................................................... 51
9.2.2 Model with equity price as dependent variable ............................................................ 52
9.2.3 Summary ...................................................................................................................... 53
9.3 Quartiles .............................................................................................................................. 53
9.3.1 Model with CDS spreads as dependent variable .......................................................... 56
9.3.2 Model with equity prices as dependent variable .......................................................... 56
9.3.4 Summary ...................................................................................................................... 57
9.4 Splitted time periods............................................................................................................ 57
9.4.1 2nd January 2004 to 8th August 2007 (Before) ............................................................. 59
9.4.2 9th August 2007 to 1st May 2009 (After) .................................................................... 61
9.5 Time periods split on quartiles ............................................................................................ 63
9.5.1 Before the beginning of the crisis 2nd January 2004 – 8th August 2007 ....................... 65
9.5.2 After the beginning of the crisis 9th August 2007 – 1st May 2009 ............................... 67
9.5.3 Summary ...................................................................................................................... 70
9.6 Rating .................................................................................................................................. 70
9.6.1 Models with CDS spreads as dependent variable ........................................................ 73
9.6.2 Models with equity prices as dependent variable......................................................... 74
9.6.3 Summary ...................................................................................................................... 75
9.7 Rating split on quartiles....................................................................................................... 75
9.7.1 High yield ..................................................................................................................... 78
9.7.2 Investment grade .......................................................................................................... 80
9.8 Sectors ................................................................................................................................. 83
9.8.1 Models with CDS spreads as dependent variable ........................................................ 88
9.8.2 Models with equity prices as dependent variable......................................................... 90
9.8.3 Summary ...................................................................................................................... 90
9.9 Results ................................................................................................................................. 91
9.10 Equity volatility ................................................................................................................. 98
10 Discussion ........................................................................................................................... 101

6
10.1 implications ..................................................................................................................... 101
10.1.1 Economic models ..................................................................................................... 101
10.1.2 New debt issue ......................................................................................................... 102
10.1.3 Investment strategy .................................................................................................. 103
10.2 Critic of results ................................................................................................................ 104
11 Concluding remarks ............................................................................................................ 106
12 Suggested further research ..................................................................................................... 107
Litterature .................................................................................................................................... 109

7
Overview of tables

Table 0: S&P 500 sector distribution

Table 1: Index – descriptive statistics

Table 2: Index - correlation

Table 3: Index - Granger Causality test

Table 4: Index – VAR CDS

Table 5: Index - VAR EQ

Table 6: Exogenous – Granger Causality test

Table 7: Exogenous – VAR EQ

Table 8: Exogenous – VAR CDS

Table 9: Quartile - descriptive statistics

Table 10: Quartile – correlation

Table 11: Quartile – Granger Causality test

Table 12: Quartile – VAR CDS

Table 13: Quartile – VAR EQ

Table 14: Time periods – descriptive statistic mean

Table 15: Time periods – descriptive statistic st.d.

Table 16: Time periods – correlation

Table 17: Before – Granger Causality test

Table 18: Before – VAR CDS

Table 19: Before – VAR EQ

Table 20: After – Granger Causality test

Table 21: After – VAR CDS

Table 22: After – VAR EQ

Table 23: Time period quartiles – descriptive statistic

Table 24: Time period quartiles – correlation

8
Table 25: Before quartiles – Granger Causality test

Table 26: Before quartiles – VAR CDS

Table 27: Before quartiles – VAR EQ

Table 28: After quartiles – Granger Causality test

Table 29: After quartiles – VAR CDS

Table 30: After quartiles – VAR EQ

Table 31: Rating scale and definition

Table 32: Rating – descriptive statistics

Table 33: Rating – correlation

Table 34: Rating – Granger Causality test

Table 35: Rating – VAR CDS

Table 36: Rating – VAR EQ

Table 37: Rating quartile – descriptive statistic CDS

Table 38: Rating quartile – descriptive statistic EQ

Table 39: Rating quartile – correlation

Table 40: High yield quartiles – Granger Causality test

Table 41: High yield quartiles – VAR CDS

Table 42: High yield quartiles – VAR EQ

Table 43: Investment grade quartiles – Granger Causality test

Table 44: Investment grade quartiles – VAR CDS

Table 45: Investment grade quartiles – VAR EQ

Table 46: Sector – descriptive statistics CDS

Table 47: Sector – ranking mean/st.d. CDS

Table 48: Sector – descriptive statistics EQ

Table 49: Sector – ranking mean/st.d. EQ

Table 50: Sector – financial leverage

Table 51: Sector – comparison of correlation, leverage and mean

9
Table 52: Sector – Granger Causality test

Table 53: Sector – VAR CDS

Table 54: Sector – VAR EQ

Table 55: Overview of results

Table 56: Comparison of models

Table 57: Index before quartile raking

Table 58: Hypotheses, decision and argument

Table 59: Volatility - correlation

Table 60: Volatility – Granger Causality test

Table 61: Volatility – VAR/VECM models CDS

Overview of figures

Figure 0: Structure overview

Figure 1: CDS transaction

Figure 2: Credit derivative – pricing relationship

Figure 3: Development in single-name and multi-name CDS

Figure 4: CDS single-name 2008-H2 parted on counterparties

Figure 5: CDS single-name 2008-H2 parted on maturity

Figure 6: Total value of share trading 2008

Figure 7: Method overview

Figure 8: Interaction between CDS and equity index

Figure 9: Development in equity prices and CDS spreads in quartiles

Figure 10: Development in CDS spreads and equity prices

Figure 11: Development in high yield and investment grade CDS spreads and equity prices

Figure 12: Investment strategy 1

Figure 13: Investment strategy 2

10
Introduction
Credit risk1 is a major source of risk for most commercial banks and arises from the possibility
that borrowers, bond issuers, and counterparties may default (Hull:2007:255). It appears in
almost all financial activities and therefore it is important to measure, price and manage
accurately. Credit derivatives can help investors with these issues. Credit derivatives are
financial contracts that transfer the risk between two counterparties without actually transferring
the underlying asset.

One kind of credit derivative is credit default swap (CDS). The buyer of a CDS has a credit risk
exposure in a firm2, typically bonds or loans, and wants to be protected from the risk that the
firm e.g. defaults. The seller of a CDS offers to compensate the buyer of a CDS if the firm e.g.
defaults. To receive this protection the buyer of a CDS makes a series of payments to the seller.
In the event of a e.g. default the seller of a CDS has to fulfil his commitment but if a default does
not happen the seller of a CDS continue to receive the payments from the buyer of a CDS until
the contract expires.

CDS’ were created by JPMorgan Chase in the mid-90’s with the aim to free up capital. At that
time JPMorgan Chase had provided a large amount of loans to corporations and foreign
governments. By federal law they were required to keep huge amounts of capital in reserve in
case any of the loans went bad. They came up with the idea to create the CDS inspired by
hedging for fluctuations in interest rates and commodity prices. By using the CDS they could be
protected if the loans defaulted and at the same time free up capital. (Phillips:2008:1).

As the market matured and developed in size CDS’ became less a hedging instrument and more
a way to speculate for or against the likelihood that particular firms would suffer financial
distress or to take advantage of mispricing in the market. The CDS market was largely exempt
from regulation by the Securities and Exchange Commission (SEC) and the Commodity Future
Trading Commission (CFTC) with the Commodity Futures Modernization Act for 2000, which
also provided Enron’s loophole (The Washington Post:2003). As the derivative market exploded
in popularity Warren Buffet warned publicly that derivatives were “Financial weapons of mass
destructions” (BBC News:2003).

1
Credit risk is defined as risk created by loss associated with the default of the borrower, or the event of credit
rating deterioration (Forte:2006:4).
2
For simplicity we will at this point assume that the buyer of a CDS actual has a risk exposure.

11
Many agreed with this statement when the credit crisis hit the world economy in mid 2008. The
CDS market was accused of causing some of the depth and increased spread of the crisis, and
this became obvious when American International Group Inc. (AIG) crashed (Abbott:2009). AIG
had sold a lot of CDS contracts and never taken into account that the default intensity increased
as it did under the crisis. AIG then became liable to settle a lot of CDS contracts but was not able
to fulfil their commitment. They simply gambled that they never had to pay – but when they
suddenly had to they went broke (Davidson: 2009). The government later saved AIG by a $180
billion rescue plan. The scenario of AIG was not a unique case.

It was possible for AIG and a lot of other firms to issue CDS contracts for such large amounts for
which they did not have the capital in case of a settlement, because the CDS market was not very
transparent or regulated (Gillam:2009). CDS are traded over-the-counter (OTC) and there is no
central reporting mechanism to determine the price of CDS (Morrissey:2008).

At the time of writing the Obama administration has taken steps to a regulation of the CDS
market and proposed a plan to the Congress in August 2009 (Abbott:2009). They proposed that
CDS trading should be supervised, new requirements for trade reporting, clearing of standardised
contracts and capital backing the trades (Brettell:2009). Combined this should heighten the
transparency of the market and limit the counterparty risk. Under the proposal SEC and CFTC
should work together to create industry rules (Abbott:2009).

Even though the rules for the regulation of the market are not yet established initiative to clear all
standardised contracts has been taken. On 6th March 2009 Intercontinental Exchange Inc (ICE)
established ICE trust, the first clearinghouse for CDS’. It is so far also the only one and has
cleared more than $1 trillion in notional volumes in index trades (ICE:2009) and it will expand it
operations to also include single name contracts. Moreover JP Morgan Chase, the largest trader
of OTC derivatives, has given open access to their CDS pricing system in January 2009 by
transferring it to International Swap and Derivatives Association (ISDA) (Pension &
Investment:2009).

All in all a new and more regulated future is set for the CDS market.

1.1 Relation to equity market


CDS spread represent the credit risk in a particular firm and this credit risk is also present in
equities. It is expected that when the credit risk of a firm increases the CDS spread will increase
– i.e. it becomes more expensive to be protected against the risk of default of the firm as the

12
probability of default increases. An increase in credit risk also affects the equity of the firm.
When the credit risk increases the value of the firm is also affected and this will affect the equity
price in a negative way. Intuitively there should be a negative relationship between the CDS
spread and the equity price of a firm. This means that increased credit risk increases the spread
of the CDS and decreases the price of the equity.

The CDS and the equity market differentiate a lot in size, maturity, trade volume and regulation.
Therefore researchers and traders have been interested in analysing how efficient the two
markets are when compared and how the interaction is between them. If inefficiencies exist
arbitrage between the two markets may be possible.

The CDS market has been a free market without regulations until the crisis, which has made it a
very interesting market. A market that has not been regulated in any significant way should self-
regulate according to the “invisible hand” theory by Adam Smith. But does this freedom really
create an efficient market? Other aspects of the CDS market also have to be considered. The
counterparties on the market are only professionals and there are fewer players than on the equity
market. Moreover CDS spreads depend on the probability of default of a particular firm and it
can be argued that some participants are able to estimate this probability better than others due to
more information. For example a bank that works closely with a firm by providing loans, advice
etc. is likely to have more information about the real creditworthiness of this firm than other
financial institutions that do not have any dealings with this specific firm. Therefore there could
be a case of asymmetric information but of course it could also be argued that the two issues are
dealt with in two different parts of the bank and that they might not share the same information
(Hull:2007:309). These are aspects that increase the probabilities of an inefficient market.

On the other hand, the equity market is probably the most watched market in world. Thousands
and thousands of analysts watch its every move. The market reacts fast to new information and
sometimes even before the information is public. The major equity markets are very regulated
and organised making trading as fast and inexpensive as possible. This may heighten the degree
of efficiency but the equity market might also have problems with insider trading – since both
instruments depends a lot on firm specific information.

It is interesting to examine the relationship between the CDS spreads and equity prices because
of the nature and direction of the relationship and because it is a relation that has not been widely
examined. The relationship can be used for example when making economic models, issuing
debt and arbitrage strategies. We will cover this in section 12.

13
1.2 Objectives of the study
The aim of this paper is to examine how equity prices and CDS spreads are related. We find this
interesting since both instruments are affected by the current crisis and the relationship between
them caught our interest. We expect this relationship to be negative and we expect to find that
equity prices influence CDS spreads and not the other way around. Moreover we expect that the
degree of influence increases with the size of the spread and in deteriorating market conditions.
The reasons for this will be discussed in section 6 where we will list the hypotheses for our
empirical work.

Due to the young age of the CDS market the amount of research available within this area is
limited. As far as we know all researchers have found that there exist a negative relationship
between the two markets but the strength of the correlation is not unambiguous. The same is
valid for the direction of price discovery. Though there is a slight tendency in most research
findings that the equity market reacts faster to information and therefore leads the CDS market.
Details of the empirical findings of other researchers will be discussed in section 5.

The limited and ambiguous results of the research give room for further research on the subject.
Our study differentiates from the previously as the time period we examine is longer (5.5 years)
and the number of firms (265) we have in our sample is higher. Moreover we cover a time period
where the CDS market is more matured than in the previous research and this means that we
have been able to get more CDS data than in most of the other research. We have chosen only to
consider a selected sample of S&P 500, which is in the US market.

Besides the higher amount of data our time period also include the recent crisis where the CDS
played a role. This is a unique opportunity to test how or if the relationship change under more
volatile and decreasing conditions. Furthermore we will try to determine how the size of the
CDS spread influence the relationship as not previously done. We will also examine if the
relationship between CDS spreads and equity prices is influenced by the credit quality and sector
of the underlying entity.

We will apply the same techniques as some of the best studies in the field that will be discussed
in section 7. Compare to most of the other research papers we will not only consider the Granger
Causality test but also analyse the actual estimated models. As far as we know there has not been
such an extensive study of this particular subject.

14
1.3 Problem
The main question this thesis seeks to answer is:

What is the relationship between the CDS spreads and the equity prices?

To answer this we will ask the following sub questions:

- How are CDS spreads and equity prices correlated?


- How do equity prices influence CDS spreads?
- How do CDS spreads influence equity prices?
- To which degree are the results robust?
o To which degree are our results influenced by:
 Exogenous variables?
 Market conditions?
 The rating of the underlying entity?
 The size of the CDS spread?
 The level of financial leverage in each sector

2 Delimitation
In this thesis we have decided to analyse the relationship between CDS market and equity market
for the US market where we have chosen 265 companies from S&P 500. An advantage by
choosing to use only US market data is that we consider only one country. Some studies consider
the European market and therefore the analysis will include data from different countries and
these might be subject to different laws and regulations, which can make the studies on two
different countries less comparable and might loose some explanatory power. It means that
country specific issues might affect our results. Moreover the CDS was invented in US and
therefore the market in US might be more matured.

When analysing the relationship between CDS spreads and equity prices we decided to look at
the “big picture” and not consider specific events such as different initiatives a firm can make,
which transfers value between equity holders and creditors. One example could be that a firm
announces an unexpected major share buy-back program or increased cash-dividend financed by
issuing new debt. This will increase the equity price since the equity holders get cash up front,
whereas the credit risk will increase due to higher financial leverage, which increases the CDS
spread.

15
In the analysis in section 9 we consider the relationship between CDS spreads and equity
volatility. It is simply an attempt to further determine the relationship between equity and CDS
spreads and not an attempt to simulate the Merton model as we do not include balance sheet
data.

Moreover, discussions in this paper are based on the assumption that we are only considering the
US market.

3 Structure
The thesis is split in 12 main sections. In section 4 we will describe the theories behind CDS,
equity, efficient markets, and the theoretical link between CDS and equity. Section 5 will go
through other empirical findings in this area. In section 6,7 and 8 we will describe our
hypotheses, the methodology and the data we are using to test them. In section 9 we test our
hypotheses and analyse the results. Section 10 will be a discussion of the implication of our
results. In section 11 we will give concluding remarks and section 12 is a suggestion for further
research.

Figure 0: Structure overview

16
4 Theory

4.1 Efficient markets


As described in section 1.1-1.2 we will examine the relationship between equity prices and CDS
spreads. We will not only consider if there exists a negative relationship between equity prices
and CDS spreads but also if one leads the other in price discovery. In section 1.1 we described
characteristics of the two markets and discussed factors that might increase or decrease the level
of efficiency. But how is an efficient market determined? We will discuss this in this section.

A market is efficient when the security prices fully reflect all available information. This
definition does not take into consideration the costs of trading and acquiring information. It is
more realistic to take this into account and therefore the definition is moderated so prices reflect
information until the marginal cost of acquiring information and transactions cost no longer
exceed the marginal benefit. This means that the investors have incentive to trade the securities
until MC = MR. What generally define a market as efficient or inefficient is whether an investor
is able to consistently earn abnormal return, e.g. that the investor is able to consistently beat the
market and if this is the case the market is inefficient3.

If we find for example that equity prices lead CDS spreads it would mean that an investor would
possibly be able to use this information to consistently beat the market when trading CDS’ as he
would have knowledge about the future movements of CDS spreads by considering the
movements in equity prices. Public available information is available for both the investors that
trade in the CDS market and the investors that trade in the equity market, so if equity prices
reacts before CDS spreads it means that the equity market is faster to incorporate new
information and thereby must be more efficient.

But this argument is only correct if we do not consider insider trading. If we find that equity
prices lead CDS spreads this may be due to a more efficient market but it may also be due to
more insider trading in the equity market. If equity prices react to new information before it
becomes public and CDS spreads react to new information in the instance it becomes public the
CDS market is more efficient than the equity market.

As mentioned we might find in our study that equity prices are leading CDS spreads but due to
the aspect of insider trading we cannot say anything about which market is more efficient. To

3
Fama’s (1970) division of efficiency in weak, semi-strong and strong is widely accepted, but since we do not
conclude on the level of efficiency in this thesis we will not comment on this further.

17
examine which market are efficient event studies is necessary. In event studies it is possible to
determine if one of the markets reacts before or after the announcement and compare the reaction
time as separately events are examined. This is though beyond the scope of this paper.

This section is meant as a clarification for the reader that a possible lead/lag relationship does not
necessarily mean that one market is more efficient than the other. And in this paper we will only
consider the influence of one market on the other and not level of efficiency.

4.2 Credit Derivatives


The purpose of this chapter is to give an introduction to the credit derivative market to broaden
the understanding of the underlying theories behind a credit default swap.

Credit derivatives are defined as:

“A class of financial instruments, the value of which is derived from an underlying market value
driven by the credit risk of private or government entities other than the counterparties to the
credit derivative transaction itself”

(Das:2005:6).

This means that the credit derivative is an instrument to isolate the credit risk from an underlying
instrument, called reference entity (Das:2005:29), and transfer the risk between two
counterparties, which a credit derivatives transaction generally consists of. The payoff from the
contracts depends on the creditworthiness of one or more firms or countries. This relative new
instrument allows firms to trade credit risk much the same way as market risk (Hull:2007:299).

Credit derivatives are Over-The-Counter (OTC) products and this means that they are not traded
as equity on official regulated exchanges but the trading is directly between the two
counterparties. It gives the parties the freedom to specify the terms in the contract by negotiation
and is therefore less standardized. The trades are usually higher due to the relatively high
transaction cost by negotiation a contract the trades are usually higher. The counterparty risk by
OTC trading is also higher than in the exchange-traded market since there are less requirements
and standardizing (Hull:2007:28). The ISDA’s documentation and definitions has become the
accepted market standard used for both trading and structured transactions (Das:2005:113). But
this does not overcome the counterparty risk that OTC trading has. Some products try to
accommodate this by requiring the seller of the protection to make a payment up-front to the
protection buyer. These are called funded credit derivatives and include collateralized debt

18
obligations, bond obligations, and loans obligations (CDO, CBO, and CLO). Not all credit
derivatives are traded as funded and the counterparty risk is still very present when trading credit
default swaps (CDS), credit spread options (CSO) etc. Since the credit crisis and the unfunded
credit derivatives role in this there is greater outside pressure to regulate the market further.
Therefore more credit derivatives are now being cleared through clearinghouses that act as
middlemen in the trade reducing the counterparty risk.

The trading of credit derivatives can be motivated by different reasons and a couple of these are
(Goldman Sachs: 2002:slide 1.07):

• To manage credit risk (by using credit derivatives it is easier to transfer credit risk from
one party to another.)
• To enable users to diversify credit risk (Exposure to countries, market sectors, and types
of financial instruments can selectively be increased or decreased by the involved parties)
• To earn income (Low-cost borrowers with large balance sheets can earn income from
parties who want credit exposure without owning assets.)
• To provide access to exposures that would not otherwise be available (e.g. investors can
gain access to syndicated loans).

4.3 Credit default swaps

4.3.1 Definition
The most popular credit derivative is the credit default swap (CDS), which is a contract that
provides insurance against the risk of default by a particular firm (Hull:2007:299). More
precisely a CDS is a bilateral contract between two parties that agree to isolate and transfer the
credit risk for a reference entity, usually bond and loans, in case of a pre-defined credit event
occurs.

A possible scenario for buying a CDS is that firm A loans 100 millions USD to firm B (reference
entity). But firm A knows it is a risky position and they might lose part of or all 100 million USD
if the reference entity defaults before the loan is repaid. Firm A is not willing to take this risk so
they buy a CDS from firm C. Firm A then becomes the protection buyer and firm C becomes the
protection seller.

19
Figure 1: CDS transaction

Protection buyer
pays spread until
Default before default
year X Protection seller
pays face value
CDS X-year
at default
contract

No default in the Protection buyer


period pays spread

The protection buyer has to pay the periodic payments until the contract expires or the reference
entity defaults (see figure 1). In the case of default the protection seller has to pay the protection
buyer the sum equal to the face value of the debt owed by the reference entity. This means that
the protection buyer obtains the right to sell bonds issued by the reference entity at their face
value when a credit event occurs and the protection seller agrees to buy the bonds for their face
value when a credit event occurs (Hull:2007:299). The transaction of a CDS results in that the
protection buyer has a credit exposure similar to taking a short position while the protection
seller has a credit exposure similar to taking a levered long position in the bond (Jersey:2007:3).

So far we have defined a credit event as default of the reference entity. But this was only for
simplicity according to the International Swaps and Derivatives Association (ISDA) 1999-2003
terms a credit event includes the following (ISDA:2003):

• Bankruptcy
• Failure to pay
o Borrow money, but not accounts payable
o Payment more than $1 million
o After pre-specified grace period
• Obligation default or obligation acceleration
• Repudiation or moratorium
o Mainly applies to sovereigns
• Restructuring
o Reduction or postponement of interest or principal repayments
o Changes in obligation’s seniority causing obligation to become subordinated

20
o Deferral or reduction of loan
o Change in currency or composition of material debt obligation

This also illustrates the flexibility of the CDS that the counterparties can agree to terms that are
very specific for the individual exposure the protection seller has. However, default are the most
popular.

4.3.1.1 Protection buyer – payment of spread


The total amount paid per year to buy protection is known as the CDS spread. Therefore the
credit spread is a measure for the compensation the protection seller gets for the risk of default
on the underlying security compared to the risk free rate. Calculation of the credit spread is as
follows (Das:2005:18):

Credit spread = yield of bond or loan – yield of corresponding risk free security

To find the difference between the yield of a risky security and a risk-free security the coupon
and maturity of the two securities has to be the same. The spread is paid with a frequency that is
decided between the two counterparties and is the cost of protection.

4.3.1.2 Protection seller – payment of face value in case of a credit event


In the case of a credit event the protection seller has to fulfil the CDS contract by paying the
protection buyer the default payment, which is equal to the face value of the debt. There are two
main types of settlement methods: cash settlement and physical delivery settlement
(Das:2005:86). The physical delivery settlement is the most used settlement method and

“Is structured as the payment of an agreed amount by the seller of protection in exchange for
delivery of a defaulted credit asset by the buyer of protection”
(Das:2005:86)

This means that in case of a credit event the protection buyer delivers the bond of the reference
entity to the protection seller in exchange for a cash amount of the face value of the debt. This
form for settlement assumes that the protection buyer is in possession of the debt of the reference
entity.
The other main settlement method is cash settlement, which does not assume the protection
buyer to be in possession of the debt of the reference entity. A cash settlement can take form as a
post default price, where the payment is based on the price of the reference asset following the

21
credit event, or a fixed payment, which is based on a pre-agreed fixed percentage of notional
principal (Das:2005:86).

4.3.1.3 Potential speculation


As described in the previous section the cash settlement of a CDS can happen without the
protection buyer owning any obligations in the reference entity. So even though CDS contracts
can be described as insurance, since it allows debt owners to hedge against credit events – they
differ from insurance as there is no requirement to hold the reference entity asset. This lack of
requirement means that CDS’ can also be used to speculate on changes in credit spreads or
mispricing (Hartmann:2008:4). This means that investors that take long positions in a CDS
without having any exposures to the reference entity have a good upside potential for exploiting
unjustified spreads between the CDS and the bond market. This also means that investors do not
have to worry that much about whether or not the reference entity issues new debt or whether or
not the other debt holders are ready to sell their debt or not (Byström: 2004:3).

4.3.2 Valuation of Credit Default Swaps


In order to better understand the relationship between the CDS market and the equity market it is
important to know how the CDS is priced and which elements affect the price of the CDS.

The valuation of derivative contracts is the cost of hedging. The pricing relationship between
credit derivatives and a cash instrument4 is that a risky asset must reflect the return from a risk-
free asset plus a risk margin. Risk margin is the compensation the investor is getting for the
assumed default risk.

Figure 2: Credit Derivatives - Pricing Relationship

Risk Free Risk margin Loan or Security


Asset + =

Source: Das (2005:461)

4 Cash instruments are financial instruments whose value is determined directly by markets. They can be divided
into securities, which are readily transferable, and other cash instruments such as loans and deposits, where both
borrower and lender have to agree on a transfer.

22
This means that the pricing of a credit default swap should reflect the following factors
(Das:2005:489):

• Default probability of the reference entity


• Expected loss given default, which is the recovery rate on the reference entity or the
deliverable obligations
• Counterparty risk on the credit default swap

These factors are incorporated into the default probability model, where the CDS is split in two
sets of cash flows: A fixed leg and a floating leg. The fixed leg is the spread payments, whereas
the floating leg is the contingent cash flow payable by the seller of protection in case of a credit
event. The value of the credit default swap is equal to the difference between the present values
of the two legs (Troelsen:2008:4).

  
     

 

          1       !  "  


# #

Where:

VCDS = The value of the credit default swap

DFt = Discount factor for t0 to t1 (the chosen risk-free rate)

SPn = Marginal survival probability of reference entity from tn-1 to tn

RR = Recovery rate of the delivered obligation

CDSS = Fee (spread) on credit default swap (bps per annum)

AFt = Accrual factor for tn-1 to tn

SPt = Survival probability of reference entity for t0 to t1

4.3.3 The CDS market in numbers5


In this section we give an overview of the development of the CDS market.

5 All data in this section is from Bank of International Settlement (BIS) 2009

23
The CDS market experienced a steep development as it increased 7 times from 6 trillion USD in
2004 to 42 trillion USD in 2008. In the second half of 2008 the market experienced a contraction
and decreased with 27%. When considering the distribution of CDS’ there was a clear
overweight in the first half of 2004 where single-name CDS’ constituted 80% of the total CDS
market6. Multi-name CDS’ became more widely used and single-name CDS’ part of the total
CDS market had decreased to 62% in second half of 2008, which can be seen in figure 3.

Figure 3:Development in single-name and multi-name CDS’


70
60
Trillions USD

50
40
30
20
10
0

Single-name credit default swaps


Multi-name credit default swaps
Source: BIS 2009

As our data is on single name CDS we will continue the description of the market using only
single name data. From figure 4 it can be seen that almost half of the counterparties in CDS
trading is “Other financial institutions” and 1/3 of the counterparties is banks and security firms.
From this it is very clear that all the trading take place between professionals.

6
The other 20% of the market is multi-name CDS which is CDS contracts where the reference entity is more than
one name as in portfolios or baskets of default swaps or credit default swap indices (BIS:2006:1)

24
Figure 4: CDS single-name 2008-H2 parted on counterparties

Other financial
2% institutions

15%
1% Banks and security
firms

49%
Insurance and
financial guaranty firms
33%

[ SPVs, SPCs, or
SPEs], Hedge
funds, Other residual
financial customers

Source: BIS 2009

From figure 5 it can be seen that 64% of the single name CDS contracts have a maturity over 1
year to 5 years. This is also consistent with the fact that a 5-year contract is considered the most
normal. Contracts with a maturity of 1 year or less is only 7% of the total contracts and this
might be due to the fact that there are quite big transaction cost connected to agreeing on a
contract. 75% of the contracts are on investment grade (not considering the non-rated).

Figure 5: CDS single-name 2008-H2 parted on maturity

7%

29%
Maturity of one year or
less
Maturity over 1 year and
up to 5 years
Maturity over 5 years

64%

Source: BIS 2009

25
4.4 Equity market
The purpose of this section is to give a short introduction to the equity market in order to be able
to better understand elements in the equity price. Equity are traded in a high volume all over the
world and in 2008 the total value of share trading in the world was 184,250 trillion USD. The US
market constituted 38% of the total share trading and NASDAQ and NYSE are the two largest
exchanges in the world. This means that in 2008 the total value of share trading in US was
70,647 trillion USD – which gives a daily turnover of 279 trillion USD (WFE:2008). Even
though the numbers are not 100% comparable it is obvious that the equity market is much higher
than the CDS market.

Figure 6: Total value of share trading 2008

21% US

America - without US

15% Asia - Pacific


62%
Europe - Africa -
Middle East
2%

Source: World Federation of Exchanges 2009

4.4.1 Definition
Equity can be defined as:

“A common stock represent an ownership claim on the earnings and asset of a corporation.
After the holders of debt claims are paid, the management of the firm can either pay out the
remaining earnings to stockholders in the form of dividends or reinvest part or all of the
earnings in the business”

(Elton:2007:17)

This means that it is the ownership claims that are being traded on the exchanges. Equity is also
characterized by having limited liability. This means that the investor in case of a bankruptcy of
the underlying entity is only losing his initial investment and therefore creditors do not have any

26
legal claims on the investor. Moreover an investor gets voting rights by buying a share
(Elton:2007:17).

Equity is not only an investment opportunity but is also used for speculation – the price of equity
is determined by supply and demand and the volatility of equity is high. Therefore equity is
considered as a risky investment. As mentioned before some of the factors affecting the price of
equity are the expectations of risk and future return. It is unlikely that all investors have the same
expectation but if this would be the case there would not be any trading of equity. The price of
equity is the average of all the information and expectations the investors have. Therefore the
degree of efficiency in the market is important because it determines how fast information is
incorporated into the price of equity.

A lot of time and research have been invested into finding an effective way to determine the
future price of equity (Elton:2007:442) and a lot of the trading is driven by the ambition to make
above average return. The attempts have ranged from finding a simple rule for selecting equities
that will perform above average to hypotheses about broad influences affecting equity prices
(Elton:2007:442). If the market is efficient it will not be possible for investors to consistently
pick “winners” but the search for an effective method has and still is occupying thousands of
people.

In general terms the determinants of a common equity price are a function of firm’s earnings,
dividend, risk, cost of money and future growth rate. The problem arises when the factors have
to be specified and implemented into a system that uses these concepts to successfully value or
select equity (Elton:2007:442).

4.4.2 Valuation models7

4.4.2.1 Discounted cash flow model


The discounted cash flow (DCF) model is the model with the most theoretical weight
(Elton:2007:459) and is based on that the value of equity is equal to the present value of the cash
flow that the stockholder expects to receive from it8 (Elton:2007:443).

PV(equity) = PV(expected future dividends)

This can be formalized in the following equation:

7 This section is no way exhaustive but just examples of some of the popular valuation models.
8 In academic literature there is a long history of debate about what should be included in cash flow expected from a
share. We will here assume as Elton (2007:443) that it is the present value of all future dividends.

27
*
% %( %* & * % *
$  & & ) &   &
1 & ' 1 & '( 1 & '* 1 & ' 1 & '*
#

Where P0 is the price today, DIVt is the dividend received at time t, and PH is the price at time H.

The DCF model has high popularity in the investment community but only a small fraction of
analysts use it. The majority of analysts still value equity by applying some sort of price earnings
ratio (P/E) to either present earnings, normalised earnings or forecasted earnings.
(Elton:2007:455)

Another popular form is regression analysis using broad determinants as earnings, growth, risk,
time value of money and dividend policy. This approach needs an estimation of the determinants
and a weighting of these in the model. (Elton:2007:455)

There is a large range of valuation models and so far none of them have turned out to be superior
in the search for above normal performance. The effect of the different models varies also with
the efficiency of the market.

4.5 Merton Model


The purpose of this project is to investigate the link between equity price and CDS spreads and
so far in our study the link has be augmented for based on previously empirical research but
there is also a theoretical background.

A crucial parameter in CDS pricing is the credit risk associated with the underlying entity. This
risk is often deducted from the credit rating made by rating agencies. This is quite infrequently
revised (and also under some critic) so another opportunity is to deduct the credit risk from the
market. Merton (1974) proposed a model that uses equity prices to estimate default probabilities
using equity volatilities, equity quotes and balance sheet information. This model made it
possible for the credit risk to be estimated on a daily basis, and is crucial for CDS pricing.

The basis of the model is that a firm’s equity is an option on the assets of the firm. For simplicity
and to illustrate the model the firm is assumed to have one zero-coupon bond outstanding and the
bond matures at T9.

We defined the variable in the model as:

V0: Value of firm’s assets today

9 The section that describes and deduct the model is based on Hull (2007:269-272)

28
VT: Value of firm’s assets at time T

E0: Value of firm’s equity today

ET: Value of firm’s equity at time T

D: Amount of debt interest and principal due to be repaid at time T

σV: Volatility of assets (assumed constant)

σE: Instantaneous volatility of equity

Two scenarios are defined as a situation where the firm default and one where there is no default
at time T.

Default: If in theory the value of a firm’s assets at time T is smaller than the debt interest and
principal due to be repaid at time T (VT < D) the firm would default and the value of its equity
would be zero.

Non-default: If VT > D the firm should make the debt repayments at time T and the value of
equity would be VT – D (remembering that assets = liabilities + equity), and therefore the firm
would not default.

The value of the firm’s equity at time T extracted from the Merton model is:

ET = max(VT-D, 0)

This can also be seen as a call option on the value of the assets of the firm with a strike price
equal to the repayment required on debt. The Black-Scholes formula for option pricing can then
be used to find the value for equity today.

+$  $ ,    -. ,( 

Where

$/ 12(/
 & 0' & 24 5
 
12 √5

And

29
(    12 √5

The value of debt today can be defined as the value of the firm’s asset subtracted the value of the
firm’s equity (D=V0-E0). The risk neutral probability that the firm will default on debt is N(-d2).

To calculate the probability of default V0 and σV need to be calculated because they are not
directly observable. But if the firm is publicly traded we can observe E0 (using equity prices).

This means that the Black-Scholes equation provides one condition that must be satisfied by V0
and σV and thereafter σE can also be estimated.

From a result of stochastic calculus know as Itô’s Lemma the following is found:

87
17 +$  82 12 $ or 17 +$  , 12 $

This provides another equation that must be satisfied by V0 and σV. The equations provide a pair
of simultaneous equations that can be solved for V0 and σV.

The most important determinant of CDS spreads is the probability that a credit event involving
the underlying references entity occurs. Merton (1974) links this default probability to the equity
market valuation and the volatility of equity return. Therefore there is a theoretical background
to the hypotheses that there exist an empirical link between the equity market quotes and the
CDS spreads.

5 Literature review
In the past decade some empirical work has been done on the relationship between CDS spreads
and equity prices. We will go through the empirical work of other researchers who have dealt
with basically the same research question, as we will examine.

Financial theories as the Merton model and efficient market theories suggest that the equity
market has incorporated the probability of default in the equity price. This is due to the fact that
equity holders are more likely to follow the financial conditions/performance of the firm than
debt holders, since they are residual owners of the firm.

We have already established that in deteriorating conditions the CDS spread will increase and
the equity price decrease. So theoretically there should be a negative relationship between CDS
30
spreads and equity prices. As far as we know no other researcher has found a different
relationship when testing this empirically. Bystrøm (2004) investigated the relationship between
equity prices and the iTraxx10 CDS market and found a negative correlation of about 0.5 when
using Pearson correlation coefficient and Spearman rank correlation. Furthermore, he confirmed
his hypothesis that equity volatility and CDS spreads were positively correlated. Hafer (2008)
analysed the correlation between equity quotes and CDS quotes split on sectors over a period of
6 months from March 1 2007 to August 31, 2009 and found a negative correlation of 0.7 using
the same techniques as Bystrøm. Lake & Apergis (2009) found that equity across European and
US markets were negatively correlated to European CDS spread changes in the period from June
16, 2004 to November 13, 2008 by using error correction (EC) and the multivariable generalized
heteroskedasticity in mean (MVGARCH-M) modelling. The same results were found in the
Japanese market. Kikuchi (2009) found a negative correlation of 0.95 between the iTraxx Japan
and the equity market (TOPIX) using the same technique as Bystrøm. Other Asian studies by
Chang, Fung & Zhang (2008) and Ishikawa & Mezrich (2009) also found a very strong negative
relationship between CDS spreads and equity prices.

In connection with analysing the relationship between equity prices and CDS spreads a part of
the literature is concerned with whether information is embedded at the same time in the CDS
market and the equity market. If information is embedded first into one of the markets, there will
exist a lead/lag relationship between the two markets. A lead/lag relationship indicates
asymmetric information and can be a sign of inefficiencies between the two markets. A lead/lag
relationship between the two markets would not be surprising since they differ in organization,
age, number of traders, and liquidity.

When looking at the empirical findings for testing the relationship between the CDS market and
the equity market the results are mixed.

Longstaff (2003) was among the first to analyse the lead/lag relationship between equity, bonds,
and CDS’ for a sample of 67 single-named CDS firms in the period March 2001 to October 2007
and found no definitive relationship between the three markets.

Lake & Apergis (2009) considered the US and European equity market and found that CDS
spreads leaded equity prices in the period 2004 to 2008 by using MVGARCH-M modelling.
Hartmann (2008) used graphical inspections of US and European data for a period of 6 years

10 iTraxx is a group of international credit derivative indices that are monitored by the International Index Company
(IIC). iTraxx indices cover credit derivative markets in Europe, Asia and Australia

31
(2002-2008) and found that CDS spreads leaded equity prices. Furthermore, he found that his
results were stronger under deteriorating conditions. Zhang (2005) used cumulated changes in
rating-adjusted CDS spreads (CCAS) and cumulative abnormal relative changes of CDS spreads
(CARC) on US data in the period 1997-2003 and found that the CDS market leaded the equity
market. Chan (2008) analysed the relationship between equity prices and CDS spreads for seven
Asian countries in the period 2001-2007 and found by applying VECM, that CDS spreads leaded
equity prices in 5 of the 7 markets, and equity prices only leaded in 1 of the 7 markets.

Norden & Weber (2004) analysed individual equity prices, bond spreads, and CDS spreads of 58
international firms in the period 2000-2002 by using a VAR model approach. They found that
equity prices leaded CDS spreads in 39 of the 58 firms and that CDS spreads leaded equity
prices in 5 of the 58 firms. Forte & Pena (2006) used the VAR model approach to find the
relationship between 52 North American and European non-financial firms in the period
September 12, 2001 to June 25, 2003. They found that equity prices leaded CDS spreads in 24 of
65 cases and CDS spreads leaded equity prices in only 5 of the 65 cases. Fung (2008) examined
the relationship between the equity prices and the CDS spreads for US data from 2001 to 2007.
They found that the relationship depended on the credit quality of the underlying reference
entity. That is, there was mutual feedback of information between equity prices and high yield
CDS spreads, whereas equity prices leaded investment grade CDS spreads. They also found that
volatility of both high yield and investment grade CDS spreads leaded equity volatility, but also
that equity volatility gave some feedback on high yield CDS spreads. Bystrøm (2004) also finds
that information flows from the equity market to the CDS market in the period he was examining
(2004-2006). Realdon (2007) found that CDS spreads and default intensities seem to be driven
by equity prices for five large corporations in the period from January 1, 2003 to June 31, 2006.

As can be seen from the literature review above there is no unambiguous answer to which of the
two markets leads the other. The only thing that seems to hold is that there exist a negative
relationship between CDS spreads and equity prices. Slightly more studies find though that the
equity market leads the CDS market.

6 Hypotheses
In section 4.5 we have established the theory behind CDS spreads and equity prices and the
possible relationship indicated by the Merton model and in section 5 covered prior empirical
findings. We will use a combination of these two to create our hypotheses. Our hypotheses

32
indicate what we expect to find in our analysis. Our methodology section 7 is the link that makes
the testing of the hypotheses possible.

H1: The correlation between CDS spreads and equity prices is negative.

Our hypothesis 1 states that the correlation between CDS spreads and equity prices is negative.
According to the Merton theory and the literature reviews we will explain the reasoning behind
this. If the value of a firm increases due to for example higher earnings than expected the equity
holders will benefit because the equity prices increases. When the value of a firm increases the
distance to default also becomes greater. Since the likelihood of default has decreased the cost of
debt also decreases. Moreover when the likelihood of default is smaller the price of protection
against default also decreases and thereby the spread of the CDS decreases. This inverse
relationship is consistent with results of other empirical studies (see Bystrøm (2004), Hafer
(2008) and Lake et al. (2009))

We expect to find that the equity prices and CDS spreads are negatively correlated.

H2: Equity prices influence CDS spreads and CDS spreads do not influence equity prices.

Our hypothesis 2 states that equity prices influence CDS spreads and CDS spreads do not
influence equity prices and this considers the efficiency of the two markets. In section 4.1 we
stated the theory behind efficient markets and it is important because if the two markets are not
efficient arbitrage is possible. We will not determine to which degree the two markets are
efficient since this usually demand event studies. But instead we will determine if historical
equity prices influence current CDS spreads or/and if historical CDS spreads influence current
equity prices. This is not the same as determining the degree of efficiency because if we for
example find that equity prices leads CDS spreads this may be due to the fact that the equity
market is more efficient than the CDS market. However, it may also be due to the fact that there
is more insider trading in the equity market and it therefore reacts before the CDS market. This
means that the equity market reacts to information before it becomes public while the CDS
market reacts to information when it becomes public – and this does not mean that the CDS
market is less efficient.

The prior empirical studies (see section 5) did not find consistent results when examining this.
But a slight overweight of the studies found that equity prices influences CDS spreads and not
the other way around (Norden & Weber (2004), Forte & Pena (2006), and Fung (2008)). We

33
expect this as well also due to the fact that the CDS market is a young market compared to the
equity market and the size of the CDS market is approximately 4,000 times smaller than the
equity market globally11. Moreover there are thousand and thousand of analysts watching every
move of the equity market so it seems logical that it will react before the CDS market. It could
though be argued that the CDS market has been unregulated until 2009 and the possibility of
insider trading may could be high due to the clientele on the market. But on the other hand there
might as well be insider trading on the equity market. All in all we expect to find that historical
equity prices influence CDS spreads and not the other way around.

H3: The relationship between CDS spreads and equity prices is still strong when including
exogenous variables.

Our hypothesis 3 states that we expect the found relationship between CDS spreads and equity
prices still to exist even though we include exogenous variables into our model. This means that
the previous found relationship is not simply due to the fact that both CDS spreads and equity
prices are influenced by a common 3rd variable. We will include the same variables as exogenous
as Fung (2008).12

H4: When the credit risk increases the strength of the relationship between CDS spreads and
equity prices increases.

H4.a: The relationship between CDS spreads and equity prices is stronger under deteriorating
market conditions.

H4.b: The strength of the relationship between CDS spreads and equity prices becomes more
pronounced the lower the credit quality.

H4.c: The strength of the relationship becomes more pronounced the bigger the CDS spreads of
the underlying index.

Our hypothesis 4 states that the strength of the relationship between CDS spreads and equity
prices increases with an increase in credit risk. The reasoning behind this is that equity bears the
ultimate form of credit risk because it represents the most subordinated claim in the capital
structure of the firm. Thus equity holders are more likely to monitor the performance of the

11The data is not directly comparable so the result is not exact.


12
This is not a perfect test and as with every other model there will always be the risk that the found relationship is
due to omitted variable problem. We are not able to test the model for all possible exogenous variables.

34
company more closely since they have more to loose if a default occurs than creditors who hold
collateral or senior claims on the asset of the firm. The probability of default is higher in:
deteriorating equity market conditions compared to improving equity market conditions, low
credit quality firms compared to high credit quality firms, and in firms with high CDS spreads
compared to firms with low CDS spreads. Investors in equity seem more likely to take action
when the risk of default increases, that is, when the credit risk increases. This means that CDS
spreads should be more responsive to changes in equity prices,, when equity for a low credit
quality firm decreases than when equity for a high credit quality firm decreases. This is also the
finding of Norden & Weber (2007) and Fung et al. (2008).

H5: The strength of the relationship between CDS spreads and equity
equity prices becomes more
pronounced the more debt a sector has.

Our hypothesis 5 states that we argue that the relationship between CDS spreads and equity
prices are more pronounced in sectors with a lot of debt. The reasoning behind this is that
decreasing equity prices cause an increase in financial leverage,
leverage which is the ratio of

. The higher the debt compared to equity the more impact a decrease in equity
prices has on the change in financial leverage. The higher the financial leverage
leve the higher the
probability of default and therefore the higher the CDS spread. This means that investors in
sectors with more debt will be more responsive to changes in equity prices than investors in
sectors with lesser debt, due to that equity represents
represents the most subordinated claim in the capital
structure of the firm.

7 Methodology13
In order to analysee the relationship between equity prices and CDS spreads and to test our
hypotheses it is important to take the type of data we have into account. Since we have time
series data it is important that we use the models that have been designed to time series data and
take advantage of the special properties this type of data have14. Furthermore, time series data
can give some complications that would never happen
ha for cross-sectional
sectional data. Since the order of

13
We test on a 5% significance level
14 Methodology from similar empirical analyse will be used as benchmark, and all econometrics work will be done
in SAS Enterprise Guide 4.1 and SAS 9.1.
9.1 SAS Enterprise Guide 4.1 and SAS 9.1 are a statistical tools
tool generally
used for statistical and econometric analysis such as cross-
cross section, panel analysis, and time series analysis and
forecast

35
observations matters in time series data but not in cross-sectional data, it is mostly this that gives
complications (Studenmund:2006:420).

We will in this section go through the method we will apply to all our data.

7.1 Correlation
The correlation between CDS spreads and equity prices is found by applying Spearman’s Rank
Correlation test. This correlation test is a non-parametric test and does not make any assumptions
about the data and a particular nature of relationship between the two variables (McDoald:2008).
The Spearman’s Rank Correlation Coefficient is defined as

∑ <(
'(  1  6 : =
(  1

where di = difference in the ranks assigned to two different characteristics of the ith individual
and n = number of individuals ranked (Gujarati:2003:406). Rank correlations look at the
similarity of rankings (from the smallest to the largest observation) in two data series, and we
present rank correlation data coefficients because the various data comes from very different
non-normal distributions (Bystrøm:2004:7).

7.2 Autocorrelation
Autocorrelation is defined as:

“correlation between members of series of observations ordered in time”

(Gujarati:2003:442)

This means, that the disturbance term relation to any other observation is influenced by the
disturbance term relating to any observation. Since we are dealing with financial time series data
our data are likely to be autocorrelated, since the time interval between the observations is short
and because we are looking at price indices. We test for autocorrelation by using graphical
inspection, Durbin-Watson test, and Breusch-Godfrey test, since autocorrelation will
underestimate the true standard deviation (Gujarati:2003:455). Since we will use vector
autoregressive models to analyse our data, the presence of autocorrelation only has influence on
the standard deviation and not on our results for analyzing the relationship between CDS spreads
and equity prices.

36
7.3 Stationarity
In time series data a strong relationship between two variables that is not caused by a real
underlying casual relationship can be a problem. This problem is called spurious correlation and
if a regression is made and the variables are spurious correlated it is called a spurious regression.
A spurious regression is a regression that looks correct (i.e. high R2 values F-test rejects R2= 0
and significant t-values) but when taking a closer look there is no relationship between Y and X
what so ever (La Cour:2007). Spurious correlation can be caused by non-stationary time series
data.

A time series is said to be stationary if its mean and variance are constant over time and if the
value of the covariance between the two time periods depends only on the distance between the
two time periods and not the actual time at which the covariance is computed
(Gujarati:2003:797). This means that if mean, variance and autocovariance remain the same no
matter at what time we measure them the data is stationary, otherwise they follow a random walk
and have a unit root.

When the original data is stationary it is said to be stationary at level I(0). If the original data is
not stationary but is transformed stationary by taking first difference – the time series is said to
be stationary at I(1).

A unit root test need to be performed to test if the time series data is stationary. We perform the
Argumeted Dickey-Fuller (ADF) test. The ADF test is an extension of the Dickey-Fuller (DF)
test and to understand the ADF we will briefly discuss the DF test.

The null-hypothesis in the DF test looks like the following (Gujarati:2003:815):

Yt is a random walk: ∆?  1? & @

Yt is a random walk with drift: ∆?  A & 1? & @

Yt is a random walk with drift around stochastic trend: ∆?  A & A( & 1? & @

The ADF test consists of estimating the following regression (Gujarati:2003:817):

∆?  A & A( & 1? &  B< ∆? & D


<#

37
The null hypothesis in all four tests is that δ = 0 , there is a unit root and the time series is
therefore non-stationary15.

The ADF test includes possible serial correlation in the error terms by adding the lagged
difference terms of the regressand (Gujarati:2003:818).

If the results of the unit root tests are showing that the time series data are non-stationary, we
have to transform the non-stationary time series in order to make it stationary. A way to do this is
by taking the first difference of the time series (Gujarati:2003:820). Since taking the first
difference of time series data can throw away some information, it should only be done after
checking for co-integration (Studenmund:2006:439). Co-integration means that despite being
individually non-stationary a linear combination of two or more time series can be stationary,
that is, there is a long-term relationship between the two variables.

In order to check whether or not our data is co-integrated we use Johansen’s methodology for
modelling co-integration (Zivot:2003:445-446). This methodology is a likelihood ratio tests for
the number of co-integrating vectors (r) that should be included in the Vector Autoregressive
model. The hypotheses for the Johansen’s Trace statistic are:

H0(r): r = r0

H1(r0): r > r0

Johansen proposes a sequential testing procedure that consistently determines the number of co-
integration vectors. The way to do the testing is by first test H0 (r0 = 0) against H1 (r0 > 1). If
null-hypothesis is not rejected we conclude that there is no co-integration vectors among the n
variables in our VAR-model. If the null is rejected it is concluded that there is at least one co-
integration vector and continues to test H0 (r0 =1) against H1 (r0 > 1). If this null is not rejected
then it is concluded that there is only one co-integration vector, but if the null is rejected then it
is concluded that there is at least two co-integration vectors. The way to determine whether or
not the null hypothesis should be rejected is by comparing the trace test value with the critical
value. If the trace test value is higher than the critical value the null hypothesis is rejected and
there is at least one co-integration vector (Brocklebank:2003:276).

15 The terms non-stationarity, random walk, and unit root can be treated as synonymous (Gujarati:2003:802)

38
If the time series are co-integrated (Gujarati:2003:823), then there is no need to transform the
time series. That is, if the time series is co-integrated spurious regression can be avoided even
though the dependent variables are non-stationary.

7.4 Model selection


The next step after performing all the tests above is to decide which model to use to analyse the
relationship between the equity market and the CDS market. The choice is between analyzing
our data in a vector autoregressive (VAR) model or in a vector error correction model (VECM).

7.4.1 VAR models


In VAR models several series are modelled in terms of their own past and could look like the
following16:

E  1$ & B E & F G & B( E( & F( G( & )

and

G  H$ & A E & I G & A( E( & I( G( & )

Where each equation contains an error that has zero expected value given past information on y
and z (Wooldridge:2009:649). As Fung (2008) the empirical form of the VAR model we will use
in this study looks like the following:

N N

!  B &  JK +L@ EK &  MK !K & D


K# K#

and

N N

+L@ E  B( &  J(K +L@ EK &  M(K !K & D(


K# K#

Where CDS is the CDS spreads and equity is the equity price.

Before estimating the above equations the maximum lag length, k, has to be decided. If we
include too few lags, it will lead to specification errors, whereas too many lag terms will
16
We omit the part corresponding to the intercept term, since it is irrelevant in subsequent calculations. This also
means that we will not take it into account when analyzing the relationship between CDS spreads and equity prices.

39
consume degrees of freedom and introduce the possibility of multicollinearity
(Gujarati:2003:849). Schwarz Information Criterion (SIC) will be used to decide the number of
lags by choosing the model that gives the lowest values of this criterion. SIC is a stricter measure
than e.g. AIC and punish harsher for including more lags (Gujarati:2003:538).

7.4.2 VECM model


If the time series data is co-integrated the VAR-model will be extended with an error correction
term. Forte and Lovreta (2008) also used the basis of the VECM framework when analyzing the
price discovery relationship in the CDS and stock market. A regression on two co-integrated
variables is an estimate of the equilibrium relationship along with a residual, which is a measure
of the extent to which these variables are out of equilibrium. To estimate the VECM model
economic theory says that the current change in the dependent variable should be affected, not
only by the current change in the independent variable, but also by the extent to which these
variables were out of equilibrium in the preceding period (Studenmund:2006:440). This should
be taken into account when formulating a dynamic relationship between the variables. As Forte
and Lovreta (2008) our empirical form of the VECM we will use in this study looks like the
following:

Δ+L@ E  B & P +L@ E  1$  1 ! 


Q Q

&  AK Δ+L@ EK &  FK Δ!K & D


K# K#

and

Q Q

Δ!  B( & P( +L@ E  1$  1 !  &  A(K Δ+L@ EK &  F(K Δ!K & D(
K# K#

Where ∆CDS and ∆equity is change in CDS spreads and equity price and δ0 and δ1 is the long-
run relationship17. The loadings λ1 and λ2 represent the adjustment coefficients that measure
how quickly CDS spreads and equity price adjust to eliminate the deviations from the long-run
equilibrium (Forte & Lovreta:2008:16-17).

17
The t values and p-values corresponding to the parameters (CDS spreads and equity prices) in vector error
correction term are missing since the parameters have non-Gaussian distributions (SAS 9.1 – help and
documentation)

40
7.5 Granger Causality
The models allow us to test if equity prices Granger cause CDS spreads or if CDS spreads
Granger cause equity prices after controlling for past CDS spreads/equity prices
(Wooldridge:2006:650).

The idea behind the Granger test is:

“…time does not run backward. That is, if event A happens before event B, then it is possible that
A is causing B. However it is not possible that B is causing A. In other words, events in the past
can cause events to happen today. Future events cannot.”

( Gujarati:2003:696)

We can use the Granger Causality test to examine if a change in CDS spreads (equity prices)
cause a change in equity prices (CDS spreads). Then CDS spreads (equity prices) should help
predicting equity prices (CDS spreads) and not vice versa18.

The test is important for our study since it can tell us about the direction of the relationship
between CDS spreads and equity prices. Is equity prices leading CDS spreads or the other way
around or is there no causality between the two markets?

The testing strategy is a two-way causation: CDS spread Granger cause equity prices and equity
prices Granger cause CDS spreads.

The null-hypotheses for this test are as following:

CDS spreads

H0: Equity prices do not Granger cause CDS spreads – CDS spreads are only influenced
by themselves

H1: Equity prices Granger cause CDS spreads

Equity prices

H0: CDS spreads do not Granger cause equity prices - Equity prices are only influenced
by themselves

18 It should be noted that Granger causality does not necessary say anything about real causality. All the variables
might react to some non-modeled factor and if the response of Xt and Yt is staggered in time there will be Granger
causality even though the real causality is different. But it cannot be solved. (Sørensen:2005:2)

41
H1: CDS spreads Granger cause equity prices

We can conclude that one of the variables leads the other if only one of the hypotheses is
rejected.

The Granger Causality test assumes that the error terms are not serial correlated between the two
time series and that the data is stationary – otherwise the test results cannot be trusted.
(Gujarati:2003:697). We check for serial correlation by looking at Durbin Watson-test, which is
also mentioned in section 7.2.

7.6 Gonzalo and Granger measure


If we find that our data is co-integrated and therefore use VECM-modelling we will also
calculate the Gonzalo and Granger measure (GG).

When using VECM modelling adjustment coefficients are estimated λ1 and λ2. Gonzalo and
Granger (1995) proposed a measure of the individual market contributions to price discovery
based on the ratio between the two adjustment coefficients (Forte:2008:18).

P(
RR 
P( P

We define λ1 is the adjustment coefficient for the VECM model with CDS spread as dependent
variable and λ2 is the adjustment coefficient for the VECM model with equity prices as
dependent variable – the lower GG the higher the equity market contribution and the lower the
CDS market contribution to price discovery. Due to this GG can be used to predict the lead/lag
relationship between the CDS and equity market.

7.7 Method overview


So to sum up (figure 7), if our data is non-stationary and co-integrated we use a VECM-model,
GG measure, and Granger Causality test19 and if our data is stationary we use VAR-model and
Granger Causality test.

19
A study made by Toda & Phillips (1993) found that the result of the Granger Causality test can be questioned
when made on co-integrated data if the right number of ranks are not specified. We assume that we have specified
the right number of ranks. If there are differences in the results between the Granger Causality test and the GG test,
we use the result of the GG test.

42
Figure 7: Method overview

stationary VAR Granger causality

data
Granger causality
co-integrated VECM
and GG measure
non-stationary
stationary
VAR -
not cointegrated Granger causality
1.difference

8 Data description
Our initial sample is daily closing equity prices and CDS spreads for firms in S&P 500 from
January 1st 2002 . We started initial with the intention of using a time period from 2002 and forth
but after the data from Bloomberg20 was downloaded and sorted we realized that we did not have
enough CDS spreads data and therefore decided to start our time period 2nd January 2004 where
the amount of CDS spread data was more satisfying.

This means that the data is in the time period: January 2nd 2004 to May 1st 2009.

We chose to take the firms in S&P 500 from January 1st 2002 since the S&P 500 had its closing
low in 2002 after the 2000-2002
2002 bear market. The reason why we do not analyse
analys on the data
from the firms that are in S&P 500 in 2009 is because we would get a survivorship bias for the
firms that survived the credit crisis. Moreover CDS’ are involved in the credit crisis and it is
therefore also interesting to have the firms that cracked under the crisis in the analysis.

As mentioned our data was downloaded from Bloomberg and we downloaded daily closing
prices and spreads.

20
Bloomberg is a provider of financial news and information. It is possible to get trading news, analyst coverage,
and historical and real time price data (Bloomberg.com).

43
The same firm may have different CDS with different maturities but we chose to focus on 5-year
CDS spreads since these are the most liquid and most used. Since our data is true market quotes
we can directly compare equity prices with CDS spreads and they are consistently updated on a
daily basis.

Even though we got the data directly from Bloomberg the sorting process of the data was quite
extensive. Because we considered such a long period the firms in our index did not behave static.
If firms changed their name in the period they were simply kept by their new name. If a firm got
acquired or merged with another firm in our index the two single firms are represented in the
index until the merger/acquisition and after the merger/acquisition the combined firm is
represented. If a firm got acquired or merged with a firm outside our index the firm is
represented in the index until the merger/acquisition and is thereafter removed from our index. If
a firm in our index acquired a firm outside our index the firm will still be represented in our
index. If a firm chose to go private in the period it is represented in the index until the
privatization.

Next step was to sort the data so only the firms with CDS are used in our analysis and we made
sure that the CDS was on the actual firm quoted in the index and not on subsidies. This means
that to every equity price there is a corresponding CDS spread. More over not all our firms had a
CDS when the time period started and they are first let into the index when they got a CDS. So at
all times there are a corresponding amount of CDS spreads and equity prices. A downside of this
is that we do not have a constant number of firms in our index.

Our sorting resulted in a reduction of our initial dataset of 500 firms to 265 firms21. But still a
quite large data set as we have 1,349 daily observations from 265 firms totalling to 357,485
observations22.

Bloomberg quotes CDS prices and not CDS spreads. Therefore it was necessary to download
daily closing rate for LIBOR from Bloomberg, which we chose to use for the risk free rate.
Theoretically the risk-free rate is determined as the yield on a government security. But
according to Hull (Hull:2007:263) derivative traders have a different definition of the risk-free
rate. They use the LIBOR23 -zero curve as the risk-free rate as they think the LIBOR/swap rate
correspond to their opportunity cost of capital. The LIBOR zero-curve is the rate that is used in

21 The number of companies at the time where all companies have a CDS.
22 To see the names of the companies in our index look at appendix 1
23 LIBOR = London InterBank Offer Rate

44
the inter-bank market as it is the rate of interest banks lend to each other at (Das:2005:462). The
LIBOR is usually higher than the treasury rate, since it is not considered to be totally risk-free as
banks are rated lower than countries (AA vs. AAA). All in all we chose to use the LIBOR as the
risk-free rate to create a more realistic picture and subtract the LIBOR from our CDS prices.

The third step is to make the indices we are going to use in our analysis. Like Bystrøm (2005)
and Fung (2008) the indices we create are equally weighted with no re-balancing. The actual
weight of the individual equities varies slightly over time, but we see this investment strategy as
more realistic than a daily re-balancing of the portfolio.

The overall index is created with all the prices and quotes from the firms we have CDS spreads
for. Hereafter we split our data set up in MSCI sectors (Communication, Consumer Cyclical,
Consumer Non-cyclical, Energy, Finance, Industry, Materials, Technology, and Utility)24 and in
rating groups (Investment grade (IG) and high yield (HY)). In the split on ratings we only use the
data for the firms that are rated for comparison reasons, which mean that 261 out of 265 are
included.

We started with an index that consisted of 500 firms and are now left with only 265. This means
that we have lost 47% in our sorting process. To see how this loss of data is distributed over
sectors we compare with how the firms are distributed over sectors in S&P 500 and compare
with our data. Finance, energy, utility and communication have the same placement in the weight
distribution as the 4th, 7th, 8th and 9th largest sectors respectively. In our index consumer non-
cyclical has become the largest represented sector while it was the 3rd largest in S&P 500 leaving
Consumer cyclical as the 2nd largest. Materials has only changed a bit which not can be said for
technology. Technology is in S&P 500 the 2nd largest sector represented but is in our index the
6th largest sector. All in all our index is only 53% of the S&P 500 and this have also changed the
distribution of the sectors represented. We have generally less firms in the individual sectors and
it could be feared that this might give some bias in the data as some indices are constructed with
few firms and therefore they might be more dependent on firm specific development.

The same is valid when considering the rating categories. There are 204 firms represented in
investment grade and only 57 represented in high yield25. This means that there are over 3 times

24 Normally health care is also a part of the sectors but we do not have any observations in this industry due to our
sorting requirements. For a short description of each sector see appendix 1
25
4 companies are not rated and therefore not considered when splitting data on rating.

45
as many firms in the investment grade this is not surprising as we considered S&P 500 firms that
include many of the largest corporations in US.

Table 0: S&P 500 sector distribution


Sectors S&P 500 in 2002 Our data
Sector weight Sector weight compared Sector weight compared
to S&P 500 with our data
Communication 2.6% 1.4% 2.6%
Consumer cyclical 17.4% 11.0% 20.8%
Consumer non-cyclical 15.8% 11.8% 22.3%
Energy 4.8% 3.4% 6.4%
Finance 15.0% 5.6% 10.6%
Industrials 13.6% 8.4% 15.8%
Materials 7.4% 4.0% 7.5%
Technology 16.0% 3.8% 7.2%
Utilities 7.4% 3.6% 6.8%
All sectors 100.0% 53.0% 100.0%

9 Empirical analysis
The aim of this chapter is to test the relationship between equity prices and CDS spreads. In
section 6 we stated 5 hypotheses and we will accept or reject these throughout the analysis. The
hypotheses were created using the theoretical background stated in section 4 and the previous
empirical analyses by other researchers stated in section 5. The methodology we will apply to
our data making the testing of the hypotheses possible is stated in section 7.

We will start by testing the overall relationship between equity prices and CDS spreads at an
index level in section 9.1. After this we will test the relationship more thoroughly at the overall
index level by different robustness tests section 9.2 to 9.5. We make the robustness tests by
including exogenous variables, splitting our data in quartiles, and separating our time period in
before and after the beginning of the credit crisis.

Furthermore we will test whether the relationship is depending on ratings and sector in section
9.6-9.7 and 9.8 respectively.

If the result of the above test indicates that the relationship between equity prices and CDS
spreads is negative and that equity prices influence CDS spreads and not the other way around
we will create a model that also include equity volatility. The Merton model (section 4.5)
indicates that equity prices influence CDS spreads negatively and that equity volatility influence
CDS positively and it could be interesting to test this as well.

46
At last we will consider and compare all the results and conclude on the relationship between
equity prices and CDS spreads.

9.1 Index26
To start with we examine the overall relationship between equity prices and CDS spreads by
examine all our data.

We calculate the standard deviation and mean for the index (table xx). The mean for equity
prices is 42.350 while the mean for CDS spreads is 106.480. The standard deviation for equity
prices is 6.830 while the standard deviation for CDS spreads is 106.40. We find that CDS
spreads are 16 times more volatile than equity prices but besides that we cannot say anything
about the direction of the relationship. To investigate this further we estimate the Spearman Rank
correlation.

Table 1:Index – decriptive


statistics
Index
CDS Equity
mean 106.480 41.350
St. Dev 106.401 6.830

To test if there is a negative relationship between CDS spreads and equity prices and thereby test
H1 we consider the graphic relationship between CDS spreads and equity prices by plotting the
respectively indices against each other. As seen from the figure 8 the relationship is negative
since the slope of the line is negative. When looking at the left side of the figure we see that CDS
spreads are relatively constant, whereas equity prices are increasing. This indicates that CDS
spreads or equity prices have not been able to price the credit risk correctly.

26 This approach is similar to Bystrøm (2005) and Hartmann (2008)

47
Figure 8: Interaction between CDS and equity index
60

50

Equity price (USD)


40

30

20

10

0
0 100 200 300 400 500 600
CDS spread (USD)

The Spearman rank correlation also supports the negative relationship, as we find it to be -
46.7%. This is quite strong and this indicates that we cannot reject H1 since we find the
relationship to be negative. We will further examine this by more sophisticated models.

Table 2: Index - correlation


Correlation
Index -0.467

We will analyse the intertemporal co-movement of CDS spreads and equity prices at index level.
More specifically we aim to explain current equity prices and CDS spreads with a 2-dimensional
VAR model or in the case of co-integration with a VECM model.

We do this to test for a negative relationship between CDS spreads and equity prices (H1) and to
test if equity prices lead CDS spreads (H2). To start with we consider stationarity of the original
data which is rejected by the unit root test (appendix 2, table 1.1, 1.2, 3.1 and 3.2) and likewise
the existence of co-integration is rejected according to Johansen’s co-integration test (appendix
2, table 5.1). The two time series are stationary at 1st difference according to unit root test
(appendix 2, table 1.3, 1.4, 3.3, and 3.4). For these reasons we estimate the relationship with a

48
VAR model where optimal number of lags is found to be two according to Schwartz Information
Criteria27.

According to the Granger Causality test (appendix 2, table 5.3) we cannot reject that equity
prices Granger cause CDS spreads, but can reject that CDS spreads Granger cause equity prices.
Therefore equity prices lead CDS spreads but not the other way around. Since the residuals in the
VAR model are not serial correlated (appendix 2, table 5.5) we can trust this result.

Table 3: Index – Granger Causality test


Equity prices Granger CDS spreads Granger
cause CDS spreads
cause equity prices
Index Yes* No*
* indicates no serial correlation in disturbance terms

9.1.1 Model with CDS spreads as dependent variable


To further examine the relationship between equity prices and CDS spreads we consider the
estimated models. To start with we consider the model with CDS as dependent variable. All the
estimates are significant except 2nd lag for the estimate for CDS spreads. The overall F-test is
also rejected (appendix 2, table 5.4) and the overall explanatory power of the model is 34.17%.

Based on our findings in section 9.1 and our acceptance of H1 we expect to find that equity
prices have a negative influence on CDS spreads. In table 4 we see that the estimates for CDS
spreads are positive while the estimates for equity prices are negative - so when equity prices
increase the CDS spreads decrease, which is in line with our expectations. Furthermore, the
negative influence of equity prices on CDS spreads is more than two times stronger in the 1st lag
than the 2nd lag.

Table 4: Index – VAR CDS


Index CDS as dependent
CDS (t-1) 0.375*

EQ (t-1) -2.260*
CDS (t-2) 0.042
EQ (t-2) -0.835*
2
R 34.17%
* indicates significance at 5% level

27
The SIC-lag estimation is made in SAS 9.1, but results are not enclosed due to the extensiveness of this
optimization.

49
9.1.2 Model with equity price as dependent variable
The picture is different when considering the model with equity as dependent variable. As
mentioned the Granger Causality test indicated that CDS spreads do not influence equity prices.
This is further emphasised when considering the estimates. We find that only the estimates for
equity prices are significantly different from zero – which further confirms that CDS spreads do
not have a significant influence on equity prices. The F test is rejected (appendix 2, table 5.4) but
the R2 is only 1.75. The results of the estimated model are consistent with the results of the
Granger Causality test.

Table 5: Index – VAR EQ


Index EQ as dependent
CDS (t-1) 0.004

EQ (t-1) -0.105*
CDS (t-2) -0.004
EQ (t-2) -0.073*
2
R 1.75%
*indicates significance at 5% level

From the above analysis we find the existence of a negative relationship between CDS spreads
and equity prices as the equity price estimates are negative signed in the VAR model with CDS
as dependent. Furthermore we find that equity prices lead CDS spreads and not the other way
around, which means that we cannot reject H2 that equity prices lead CDS spreads. We continue
to further test the robustness of the results.

9.2 Exogenous variables


To further test the robustness of our results we check for a potential omitted variable problem by
adding additional explanatory variables in our VAR model (H2.c). As Fung (2004) and Collin
Dufresne (2001) the following exogenous variables are included28:

- Changes in 5-year T-bill rate (RATE) (Yahoo finance).


- Slope of the term structure (SLOPE), which is measured by differences in yield for
maturities between 10-year and 2-year treasuries (US Treasury).

28
A potential omitted variable problem can always be argumented for because the perfect variables to predict e.g.
stock prices are not found. But when including exogenous variables further robustness of our model can be checked.
We do not discus the meaning of the single exogenous variable.

50
- The CBOE volatility index (VIX), which is a key measure of market expectations of near
term volatility conveyed by the S&P 500 equity index option price (CBOE)29.

N N

!  B &  JK +L@ EK &  MK !K & S +ST & D
K# K#

N N

+L@ E  B( &  J(K +L@ EK &  M(K !K & S( +ST & D(
K# K#

The Granger Causality test indicates the same relationship as earlier – it cannot be rejected that
equity prices Granger cause CDS spreads and it is rejected that CDS spreads Granger cause
equity prices. But we have problems with serial correlations for both models. The DW statistics
for the model with CDS spread as dependent variable is 2.044 and for the model with equity
prices as dependent variable it is 2.141 (appendix 3, table 7.5). A problem with serial correlation
is not that surprising since we add more variables to the model. We consider the estimated
models.

Table 6: Exogenous - Granger Causality test


Equity prices Granger CDS spreads Granger
cause CDS spreads cause equity prices
Exogenous Yes No
variable
* indicates no serial correlation in disturbance terms

9.2.1 Model with CDS spreads as dependent variable


When considering the VAR model with CDS spread as dependent variable not much has
changed compared with the index model in section 9.1.1. The R2 is only a bit higher at 36.74%
but the signs in the model are the same as well as the significant estimates. The influence from
CDS spreads and equity prices are a bit higher than the index model. Only one of the exogenous
variables is significant (VIX).

29
Since its introduction in 1993 VIX has been considered by many to be a premier barometer of investor sentiment
and market volatility (Fung:2008:16).

51
Table 7: Exogenous – VAR CDS
Exogenous
variables CDS as dependent
VIX (t) -0.254*

TRATE (t) 1.244


SLOPE (t) -0.771
CDS (t-1) 0.372*
EQ (t-1) -2.392*
CDS (t-2) 0.056
EQ (t-2) -0.939*
2
R 36.74%
*indicates significance at 5% level

9.2.2 Model with equity price as dependent variable


The model with equity prices as dependent variable has gained significant more explanatory
power and the R2 has risen from 1.75% to 66.31% compared to the index model in section 9.1.2,
which is consistent with Fungs (2008) experience. We find that all the estimates for the
exogenous variables and the CDS spreads and the overall model are significant. The estimates
for equity prices are not significant. The influence from historical equity prices and CDS spreads
is very low and it seems like the increased R2 is due to the exogenous variables and the
possibility of serial correlation. But this just further confirmes that equity prices are influenced
by other factors than CDS spreads and equity prices.

Table 8: Exogenous – VAR EQ


Exogenous
variables EQ as dependent
VIX (t) 0.195*

TRATE (t) -0.803*


SLOPE (t) 0.538*
CDS (t-1) 0.006*
EQ (t-1) -0.003
CDS (t-2) -0.015*
EQ (t-2) 0.005
R2 66.31%
*indicates significance at 5% level

52
9.2.3 Summary
We have problems with serial correlation in both models but mostly for the model with equity
prices as dependent variable (appendix 3, table 7.5). This means that we cannot trust the Granger
Causality test. But after examining the estimated models the model with CDS spreads as
dependent variable suggest that equity prices influence CDS spreads negatively and this effect is
stronger than any influence from the exogenous variables. The model with equity prices as
dependent variable is more explained by the exogenous variables than CDS spreads. Based on
this we conclude that our initial results are not affected by our robustness test and we can
therefore not reject hypothesis H2.c.

9.3 Quartiles
As stated in hypothesis H3.b CDS spreads from high risk firms should be linked more strongly to
equity prices than CDS spreads from low risk firms. To further test the robustness of our result
and to test hypothesis H3.b we subdivide our data into quartiles, where quartile 1 contains the
25% firms with the lowest CDS spreads, quartile 2 the 25% firms with the second lowest CDS
spreads, quartile 3 contains the 25% of firms with second highest spreads, and quartile 4 contains
the 25% firms with the highest spread.

Table 9: Quartile - descriptive statistics


Quartile 1 Quartile 2 Quartile 3 Quartile 4
CDS Equity CDS Equity CDS Equity CDS Equity
Mean 30.18 43.14 50.83 42.23 97.43 36.98 283.86 43.61
St. Dev 22.1 7.2 41.82 6.47 78.53 6.8 257.48 8.35

We re-calculate the mean and standard deviation for the quartiles. The standard deviation and
mean for CDS spreads increase from quartile 1 to 4. The mean is over 9 times higher in quartile
4 compared with quartile 1. The same is true for the standard deviation, which is almost 12 times
higher in quartile 4 compared with quartile 1. This is not surprising since the quartiles are created
after the size of the average spreads, but we find that the increase from quartile 3 to 4 is
especially steep.

If we consider the mean and standard deviation for equity prices the pattern is not that obvious.
The mean for quartile 1 to 4 is between 36.98-43.16 where the highest mean is in quartile 4 and
the lowest mean is in quartile 3. However, we do find that the mean decreases 15% from quartile
1 to 3 which is in accordance with our expectations – equity prices decrease as CDS spreads

53
increase. The difference in standard deviation between the quartiles is also quite moderate. The
standard deviation is between 6.47-8.35 where the highest is in quartile 4 and lowest is in
quartile 2.

From table 9 we can conclude that CDS spreads in the quartiles are much more volatile than
equity prices. In quartile 1 CDS spreads are 3 times more volatile than equity prices but in
quartile 4 the CDS spreads are 31 times more volatile than equity prices. This pattern is
confirmed graphically (figure 9) where the extreme increase in quartile 4 CDS spreads can be
seen. It can be seen from figure 9 that the equity quartiles actually behave more like what we
expected after mid 2007.

So it will be interesting to see if the relationship between equity prices and CDS spreads holds in
all the quartiles.

Figure 9: Development in equity prices and CDS spreads in quartiles


70 1400

1200 Quartile 1 Qaurtile 2


60
Quartile 3 Quartile 4
50 1000
Equity USD

CDS USD

40 800

30 600

20 400
Quartile 1 Quartile 2
10 200
Quartile 3 Quartile 4
0 0
1-2-04 1-2-05 1-2-06 1-2-07 1-2-08 1-2-09 1-2-04 1-2-05 1-2-06 1-2-07 1-2-08 1-2-09

The above figure compares the development of the equity prices and the CDS spreads in the
quartiles indices.

When looking at the equity prices we see that the development for all quartiles is identical. The
equity prices in all quartiles are slowly increasing until mid 2007, where the equity prices then
decreases. Quartile 3 and 4 experience a steeper decrease than quartile 1 and 2.

The development for CDS spreads are more differentiated. All quartiles have basically stable
CDS spreads until mid 2007, where they increase. As can be seen quartile 4 has the biggest
increase and skyrocketed from mid 2007 until beginning of 2009.

From figure 8 above it seems that the relationship is negative, but it is difficult to see how it
varies for the different quartiles. To answer hypothesis H4.c we will perform some extra

54
econometric work. The Spearman Rank correlation also indicates. The correlation in the
quartiles is in the same range as seen in table 10.

Table 10: Quartile -


correlation
Correlation
Quartile 1 -0.551
Quartile 2 -0.518
Quartile 3 -0.523
Quartile 4 -0.542

To further examine the relationship between the CDS spreads and equity prices in quartile 1 to 4
we firstly consider the stationarity of the original data which is rejected for all quartiles by the
unit root test (appendix 4-7, table 1.1, 1.2, 2.1, 2.2) and likewise is the existence of co-
integration rejected in all the quartiles according to Johansen’s co-integration test (appendix 4-7,
table 3.1). The time series are stationary at 1st difference according to unit root test for all the
quartiles. For these reasons the 4 quartiles are estimated by a VAR model. The optimal numbers
of lags are found to be two in all the 4 models according to Schwartz Information Criteria.

According to the Granger Causality test (table 11) we cannot reject that equity prices Granger
cause CDS spreads, but can reject that CDS spreads Granger cause equity prices for all the
quartiles. Therefore equity prices lead CDS spreads but not the other way around. Most of the
residuals in the VAR models are not serial correlated and it is only the models with CDS as
dependent in quartile 1 and 2 there might be some problems with (DW respectively 2.041 and
2.032 – appendix 4-7, table 3.5). Otherwise the results can be trusted.

Table 11: Quartile – Granger Causality test


Equity prices Granger CDS spreads Granger
cause CDS spreads
cause equity prices
Quartile 1 Yes No*
Quartile 2 Yes No*
Quartile 3 Yes* No*
Quartile 4 Yes* No*
* indicates no serial correlation in disturbance terms

Below we will go through the estimated VAR models for the quartiles.

55
9.3.1 Model with CDS spreads as dependent variable

Table 12: Quartile VAR CDS


CDS Quartile 1 Quartile 2 Quartile 3 Quartile 4
CDS (t-1) 0.282* 0.329* 0.389* 0.321*

EQ (t-1) -0.409* -0.779* -2.395* -5.353*


CDS (t-2) 0.148* 0.001 -0.023 0.066*
EQ (t-2) -0.268* -0.599* -0.844* -1.826*
2
R 24.04% 22.68% 31.79% 28.74%
*indicates significance at 5% level

We start by consider the models with CDS spreads as dependent variable. From table 12 we see
that all estimates are significant, except the estimates for CDS spreads in the 2nd lag in quartile 2
and quartile 3. The F-test for all models is rejected (appendix 4-7, table 3.4). All estimates for
equity prices are negative, and the estimates for CDS spreads are positive except the 2nd lag in
quartile 3. The R2 for the models is in the range from 22.68% to 31.79%. Based on the statement
in hypothesis 3.b we expect quartile 4 to have the highest R2, but in this case quartile 3 has the
highest R2.

As expected we find that the equity price estimates increase in both lags the higher the quartile
number. The equity price estimate in quartile 4 is 13 times higher than the equity price estimate
in quartile 1. Furthermore, the estimates for equity are more influential in the 1st lag than the 2nd
lag.

It does not seem like there is any pattern when we consider the estimates for the CDS spreads,
other than the estimates in the 1st lag have higher influence than the estimates in the 2nd lag.

9.3.2 Model with equity prices as dependent variable


Table 13: Quartile VAR EQ
EQ Quartile 1 Quartile 2 Quartile 3 Quartile 4
CDS (t-1) 0.025* 0.002 0.007 0.001

EQ (t-1) -0.141 -0.112* -0.071* -0.072*


CDS (t-2) -0.002 0.013 -0.008 -0.002
EQ (t-2) -0.104* -0.063* -0.062* -0.043
2
R 3.13% 1.81% 1.47% 0.8%
*indicates significance at 5% level

We further examine the models with equity as dependent variable. We see that basically all the
equity price estimates are significant (except the 2nd lag in quartile 4), whereas only one CDS

56
spread estimate is significant (1st lag in quartile 1). The F-tests for the models are also rejected
(appendix xx, table xx). The R2 for the models are in the range from 0.8% to 3.13%.

We expect the sign for the equity price estimates to be positive, but they are all negative in the
estimated models. Furthermore we expect CDS spread estimates to be negative, but this is only
valid for the CDS spread estimates in the 2nd lag for quartile 1, 3, and 4.

9.3.4 Summary
The Granger Causality test indicated that equity prices Granger cause CDS spreads and not the
other way around. This is further emphasised by looking at the results from the estimated VAR
models. The VAR models with CDS spreads as dependent variable have significant estimates,
high R2 and the influence of the estimates in the dependent variable increases the higher the
quartile number whereas the estimates for CDS spreads in the VAR models with equity prices as
dependent variable are not significant and the R2 is low. These observations indicate that we
cannot reject hypothesis 3.b, and that the strength of the relationship between CDS spreads and
equity prices becomes more pronounced the higher the CDS spreads in the underlying index.

9.4 Splitted time periods


The two time series are splitted in the periods 2nd January 2004 to 8th August 2007 and 9th
August 2007 to 1st May 2009 and the models are re-estimated. The reason behind splitting the
data on 9th August 2007 is due to the fact that 9th August 2007 was the day that BNP Paribas30
announced that it froze the deposits in its three “subprime”-funds, since it was no longer able to
value the underlying assets. Simultaneously the liquidity disappeared on the financial markets in
both Europe and USA. The money market rates skyrocketed and banks were no longer willing to
lend money to each other (Berg&Bech:2009:46)31. The action by BNP Paribas was the last straw
in an already cracked foundation. This was the actual beginning of the financial crisis, which is
also confirmed by figure 10. We consider the period after 9th August 2007 to be a deteriorating
and volatile market, whereas the period before 9th August 2007 is considered to be an improving
and stable market.

30
BNP Paribas is the largest bank in the Euro zone and among the 6th largest banks world wide (BNP Paribas:2009)
31
The actual date of the beginning of the crisis is still discussed and probably impossible to determine to one
specific date, but we will use 9th August 2007

57
Figure 10: Development in CDS spreads and equity prices

600

500

400
USD

300

200
CDS Index EQ Index
100

0
01-02-04 01-02-05 01-02-06 01-02-07 01-02-08 01-02-09

We split the data set in two to test if there is a difference in the relationship between CDS
spreads and equity prices in the two periods (H4.a). As it can be seen from figure 10 the CDS
spreads and equity price are at the same price level from 2004 to mid-2007. But in mid-2007 the
effect of the credit crisis is expressed in CDS spreads as they are skyrocking. The effect in the
equity market is by no means so extreme.

Table 14: Time periods descriptive statistics -


mean
Mean Before After Index
CDS 52.95 217.82 106.40
Equity 42.19 39.61 41.35

The mean for CDS spreads increases 4 times, whereas equity prices decrease 0.90 times. The
standard deviation of the equity index increases 1.7 times from before till after the beginning of
the crisis. But the standard deviation of the CDS spreads increases with 12 times. It is further
confirmed that CDS spreads react more strongly in volatile periods than equity prices.

Table 15: Time period descriptive statistic –


standard deviation
Std Before After Index
CDS 9.84 127.75 106.40
Equity 5.28 9.02 6.83

58
The Spearman Rank correlation (table 16) is at -97.3%, which is 1.4 times higher after the
beginning of the crisis compared with before the beginning of the crisis. Both the correlation for
before and after the beginning of the crisis is higher than the correlation for the whole period.

Table 16: Time periods


-correlation
Correlation
Before -0.716
After -0.973
Index -0.467

This is a quite unique opportunity to stress-test our results by examining if the previous found
relationship holds in a very volatile period.

9.4.1 2nd January 2004 to 8th August 2007 (Before)


To examine the relationship between the CDS spreads and equity prices in the period 2nd January
2004 to 8th August 2007 we firstly consider stationarity of the original data, which is rejected by
the unit roots test (appendix 8, table 1.1, 1.2, 3.1 and 3.2) and likewise the existence of co-
integration is rejected according to Johansen’s co-integration test (appendix 8, table 5.1). The
two time series are stationary at 1st difference according to unit root test (appendix 8, table 1.3,
1.4, 3.3, and 3.4). For these reasons we estimate the relationship with a VAR model where the
optimal number of lags are found to be two according to Schwartz Information Criteria.

According to the Granger Causality test (table 17) we cannot reject that equity prices Granger
cause CDS spreads, and cannot reject that CDS spreads Granger cause equity prices. It seems
that there is a two-way relationship between CDS spreads and equity prices. Since the DW
statistic is 1.91 for CDS spreads we cannot trust this result (appendix 8, table 5.5). It is not a
severe serial correlation but it should be kept in mind. Moreover it should be remarked that the
results of the Granger Causality test for equity are borderline results (appendix 8, table 5.5).

Table 17: Before – Granger Causality test


Equity prices Granger CDS spreads Granger
cause CDS spreads cause equity prices
Before Yes Yes*
* indicates no serial correlation in disturbance terms

59
9.4.1.1 Model with CDS spreads as dependent variable
To further examine the relationship we consider the estimated models. To start with we consider
the model with CDS spreads as dependent. We find that all the estimates in the model are
significant and the F-test is also rejected. The overall explanatory power of the model is 22.37%
(appendix 8, table 5.4).

The estimates have the expected signs, that is, the estimates for equity prices are negative and the
estimates for CDS spreads are positive. The negative influence of equity prices on CDS spreads
is highest in 1st lag as the influence 0.75 times lower in the 2nd lag. The results are corresponding
with the results of the Granger Causality test despite the problems with serial correlation.

Table 18: Before – VAR CDS


Before CDS as dependent
CDS (t-1) 0.232*

EQ (t-1) -0.544*
CDS (t-2) 0.187*
EQ (t-2) -0.413*
2
R 22.37%
*indicates significance at 5% level

9.4.1.2 Model with equity price as dependent variable


But as mentioned the result that CDS spreads Granger cause equity prices is borderline and this
is further confirmed by the estimated model with equity as dependent variable. We find that the
only significant estimates are in the 1st lag of CDS spreads and in the 2nd lag of equity prices.
The overall significance of the model is also rejected but the result is borderline (0.0508 –
appendix 8, table 5.4). The overall explanatory power of the model is only 1.04%.

Table 19: Before – VAR EQ


Before EQ as dependent
CDS (t-1)
-0.033*
EQ (t-1) -0.015
CDS (t-2) -0.009
EQ (t-2) -0.069*
R2 1.04%
*indicates significance at 5% level

60
9.4.1.3 Summary
We can still conclude that equity prices lead CDS spreads even though the results are not as clear
as the results for the overall market index. The results for the model with CDS spreads as
dependent variable are weaker compared with the overall index model as the size of the
estimates for equity prices is lower as well as a lower R2. The Granger Causality test indicates
that a two-way relationship exists between CDS spreads and equity prices. But due to low R2s,
insignificant estimates, acceptance if the F-tests and the fact that the result was borderline we
conclude that the two-way relationship is too weak to be considered important. That is, equity
prices lead CDS spreads.

9.4.2 9th August 2007 to 1st May 2009 (After)


To examine the relationship between CDS spreads and equity prices in the period 9th August
2007 to 1st May 2009 we firstly consider the stationarity. As earlier we find that the original data
is not stationary but we cannot reject the existence of co-integration according to Johansen’s co-
integration test and therefore the relationship is estimated using VECM modelling (appendix 9,
table 1.1, 1.2, 3.1, 3.2, 5.1). The optimal number of lags is found to be 2 according to Schwartz
Information Criteria.

According to the Granger Causality test we cannot reject that equity prices Granger causes CDS
spread but can reject that CDS spreads Granger cause equity prices. The result for equity prices
can be trusted but there are a bit of serial correlation in the model with CDS as dependent
(2.0988 – appendix 9, table 5.7)

Table 20: After – Granger Causality test


Equity prices Granger CDS spreads Granger
cause CDS spreads cause equity prices
After Yes No*32
* indicates no serial correlation in disturbance terms

There are some problems with serial correlation so the Granger causality test cannot be trusted
but the Gonzalo & Granger statistic also indicate that equity prices is leading CDS spreads.

32
This is further emphasised by a GG value of 0.008

61
9.4.2.1 Model with CDS spreads as dependent variable

Table 21: After – VAR CDS


After CDS as dependent
CDS (t-1) 0.413*
EQ (t-1) -2.123*
2
R 38.23%
*indicates significance at 5% level

To further examine the relationship we consider the estimated models. When considering the
VECM model with CDS spread as dependent variablewe find that the two estimated coefficients
are significant. The model seems quite strong as the R2 is 38.23% and it cannot be rejected that
the overall model is significant. Moreover, the estimates have the right signs and the influence
from historical equity prices is more than 5 times stronger than the influence from historical CDS
spreads.

9.4.2.2 Model with equity price as dependent variable

Table 22: After – VAR EQ


After EQ as dependent
CDS (t-1)
0.008
EQ (t-1) -0.136*
R2 22.7%
*indicates significance at 5% level

The estimated model with equity price as dependent variable, further confirms the result from the
Granger and GG test. The model shows that the influence from equity prices is 17 times stronger
than the influence from CDS spreads and only the estimate for equity prices is significant. The
overall explanatory power of the model is quite high at 22.7% but it seems that most of the
explanatory power comes from equity prices. We cannot reject that the overall model is
significant but the result is quite borderline at 0.0418 (appendix 9, table 5.7).

9.4.2.3 Summary
We find stronger evidence for equity prices lead CDS spreads than the other way around. Since
we find that the two time series are co-integrated it means that the two time series are driven by
the same common factor in the long run (Forte:2008:15). This actually suggests that the
relationship between CDS spreads and equity prices is stronger in the more volatile period, that

62
is in the period after the beginning of the crisis. Forte (2008) suggests that the long-term
relationship means that credit risk is priced equally in the equity market and the CDS market in
the long run.

We cannot reject hypothesis 4.a, since the actual estimates have a bigger impact on the CDS
spreads after the beginning of the crisis. Moreover, the R2 is higher e.g. the explanatory power of
the model after the beginning of the crisis is higher than before the beginning of the crisis.

9.5 Time periods split on quartiles


In the period before the crisis we found a weak evidence of a two-way relationship. In order to
analyse this result further, we have split the two time periods into quartiles, where quartile 1
contains the 25% firms with the highest CDS spreads, quartile 2 contains the 25% firms with the
second highest CDS spreads and so forth, to see if we find any indication of a two-way
relationship again.

Table 23: Time periods quartile – descriptive statistics


BEFORE AFTER
Mean St. dev. Mean St. dev.
CDS
Quartile 1 20.017 7.595 51.324 26.878
Quartile 2 30.713 7.629 92.677 51.711
Quartile 3 59.132 13.713 177.073 95.994
Quartile 4 152.635 25.798 556.804 304.206
Equity
Quartile 1 42.246 6.062 45.005 8.846
Quartile 2 41.971 5.246 42.758 8.450
Quartile 3 38.238 4.990 34.354 8.982
Quartile 4 46.935 4.798 36.702 9.805

The mean and standard deviation for the quartiles in the two periods are seen in table 23. As
expected the mean for CDS spreads are increasing the higher the quartile. That is, mean CDS
spread is lowest in quartile 1 and highest in quartile 4. This is true for both before and after the
beginning of the crisis. Before the beginning of the crisis the mean in quartile 4 is 7.5 times
higher than the mean in quartile 1, whereas the mean in quartile 4 is almost 11 times higher than
quartile 1 after the beginning of the crisis. This indicates that the high CDS spreads react more
strongly to deteriorating market conditions.

63
As expected the standard deviation also increases from quartile 1 to quartile 4. Before the
beginning of the crisis the standard deviation for quartile 4 is 3.4 times higher than the standard
deviation in quartile 1, whereas the standard deviation in quartile 4 is more than 11 times higher
than in quartile 1 after the beginning of the crisis. This indicates that the high CDS spreads
become more volatile in deteriorating market conditions. The level of the standard deviation is
on average 7 times higher in the period after the beginning of the crisis than before.

When studying the means and standard deviations for the equity price quartiles we expect to find
that the higher the quartile (i.e. CDS spreads) the lower the equity price. This turns out not to be
the case.

From table 23 we see that the difference in the mean for equity prices is not so high as for CDS
spreads. When looking at the table it is notable that quartile 4 has the highest mean before the
beginning of the crisis, which is the opposite of our expectations. After the beginning of the
crises quartile 4 is the 2nd lowest, which is more in line with our expectations.

We expect the standard deviation to increase from quartile 1 to quartile 4, since higher quartile
(i.e. higher spreads) means more risk and therefore higher standard deviation. Surprisingly the
standard deviation is decreasing from quartile 1 to quartile 4 before the beginning of the crisis.
After the beginning of the crisis the standard deviation behave according to our expectations, that
is, it is increasing from quartile 1 to quartile 4. The level of the standard deviation has on average
increased 1.7 times from before the crisis began till after the crisis began, which is nothing
compared to the standard deviation of CDS spreads.

Table 24: Time periods quartiles


- correlation
Correlation
Before After
Quartile 1 -0.916 -0.929
Quartile 2 -0.883 -0.924
Quartile 3 -0.711 -0.926
Quartile 4 -0.107 -0.978

Before the beginning of the crisis the correlation between CDS spreads and equity prices in
falling the higher the quartile. This could indicate that a mispricing in credit risk is present. The
correlation decreases from -0.711 in quartile 3 to -0.107 in quartile 4, which is notable. !0.7% is
much smaller than what we have found earlier. After the beginning of the crisis the correlations

64
in the 3 first quartiles are basically the same (-0.92), whereas the correlation in quartile 4 is
rather high at -0.978.

From the above analysis we find that CDS spreads react more strongly to changes in market
conditions than equity prices, which is also in line with figure 9 in section 9.3. To further
examine the relationship we consider the estimated models.

9.5.1 Before the beginning of the crisis 2nd January 2004 – 8th August 2007
To examine the relationship between CDS spreads and equity prices in the quartiles before the
beginning of the crisis, we firstly consider the stationarity of the original data, which is rejected
by the unit root test (appendix 10-13, table 1.1, 1.2, 2.1, 2.2) and likewise the existence of co-
integration is rejected according to Johansen’s co-integration test (appendix 10-13, table 3.1).
The time series for all the quartiles are stationary at 1st difference according to unit root test
(appendix 10-13, table 1.3, 1.4, 2.3, 2.4). For these reasons we estimate the relationships with
VAR models. We find that the optimal lags for the 4 models are two according to Schwartz
Information Criteria. For the total period section 9.1 we also found that the models should be
estimated by VAR(2) models.

Table 25 gives the results of the Granger Causality test. We cannot reject that equity prices
Granger cause CDS spreads in all quartiles. We can reject that CDS spreads Granger cause
equity prices in quartile 1 to 3, but cannot reject that CDS spreads Granger cause equity prices in
quartile 4. That is, there seems to be a two-way relationship between CDS spreads and equity
prices in quartile 4.

Table 25: Before quartiles – Granger Causality test


Equity prices Granger CDS spreads Granger
Before cause CDS spreads cause equity prices
Quartile 1 Yes* No*
Quartile 2 Yes No*
Quartile 3 Yes No*
Quartile 4 Yes Yes
* indicates no serial correlation in disturbance terms

The test results for the Granger Causality test can only be trusted if the disturbance terms are not
serial correlated. In table 25 we see that we can trust that equity prices Granger cause CDS
spreads in quartile 1 and that CDS spreads do not Granger cause equity prices in quartile 1 to 3.
The serial correlation is not severe as the DW statistics are between 1.912 and 2.021 (appendix
10-13, table 3.5), but it is still present and should be taken into account when discussing the
relationship between CDS spreads and equity prices.

65
To further examine the relationship we consider the estimated models

9.5.1.1 Models with CDS spreads as dependent variable


We start to examine the models with CDS spreads as dependent variable. From table 26 it can be
seen that all the estimates are significant and the F test for all the models are rejected (appendix
10-13, table 3.4). All the estimates for equity prices are negative and all the estimates for the
CDS spreads are positive, which is according to our expectations.

As expected we find that the equity price estimates increase in both lags the higher the quartile
number. This suggests that equity prices influence on CDS spreads increases the higher the CDS
spread. In quartile 1 and 2 the estimates for equity price in the 2nd lag is more influential than the
estimates in the 1st lag and the opposites is true for quartile 3 and 4, which we have not found
earlier in this paper.

Table 26: Before quartiles - VAR CDS


Before CDS as dependent
Quartile 1 Quartile 2 Quartile 3 Quartile 4
CDS (t-1)
0.135* 0.128* 0.249* 0.142*
EQ (t-1) -0.117* -0.292* -0.862* -1.308*
CDS (t-2) 0.266* 0.218* 0.156* 0.162*
EQ (t-2) -0.185* -0.305* -0.492* -0.929*
R2 16.96% 16.62% 20.92% 12.18%
*indicates significance at 5% level

It does not seem like there is any pattern when we consider the estimates for CDS spreads. In
quartile 1, 2, and 3 the estimates in the 2nd lag are higher than the estimates in the 1st lag. This
means that CDS spreads are more affected by CDS spread than CDS spread.

The R2 for the models is in the range from 12.18% to 20.92%, which is lower than our other
models with CDS as dependent. We can conclude that equity prices affect CDS spreads
negatively, and that the strength of this influence increases the higher the quartile.

9.5.1.2 Models with equity prices as dependent variable


We continue with examining the models with equity priceas dependent variable. From table 27
we see that none of the estimates in quartile 1 and quartile 2 are significant different from zero.
In quartile 3 the only significant value is the estimate for CDS spread in lag 1, and in quartile 4

66
the only significant values are the ones in 2nd lag. The overall F-test (appendix 10-13, table 3.4)
cannot be rejected for quartiles 1-3, but can be rejected for quartile 4.

Table 27: Before quartiles – VAR EQ


Before EQ as dependent
Quartile 1 Quartile 2 Quartile 3 Quartile 4
CDS (t-1)
-0.048 -0.041 -0.019* -0.009
EQ (t-1) -0.029 0.002 0.007 -0.015
CDS (t-2) 0.038 0.019 -0.001 -0.015*
EQ (t-2) -0.060 -0.048 -0.065 -0.070*
R2 0.62% 0.58% 0.73% 0.189%
*indicates significance at 5% level

Since quartile 4 is the only quartile with significant estimates and the only model where the F-
test is rejected, this is the only model we will comment on. We notice that the sign for the equity
price estimate in the 2nd lag is opposite of what we expect, whereas the CDS spread estimate in
the 2nd lag has the expected sign. Since R2 is very low (0.189%) we conclude that the VAR
model in quartile 4 does not explain the relationship between CDS spreads and equity prices.
The R2 for the other models is also low (in the range from 0.58% to 0.73%), which further
indicates that CDS spreads do not have an influence on the equity prices.

9.5.1.3 Summary
To conclude on the above robustness test of the period before the beginning of the crisis we can
conclude that equity prices lead CDS spreads. Based on low R2, insignificant estimates and the
Granger test we also conclude the influences CDS spreads have on equity prices are weak
compared to the other way around, so the two-way relationship we found earlier in this paper
will not be considered further.

9.5.2 After the beginning of the crisis 9th August 2007 – 1st May 2009
To examine the relationship between CDS spreads and equity prices in quartiles after the
beginning of the crisis, we firstly consider the stationarity of the original data. The stationarity is
rejected for all quartiles by the unit root test (appendix 14-17, table 1.1, 1.2, 2.1, 2.2) and
likewise is the existence of co-integration according to Johansen’s co-integration test (appendix
14-17, table 3.1) except for quartile 2. The time series data for the quartile 1, quartile 3, and
quartile 4 are stationary at 1st difference according to the unit root test (appendix 14-17, table 1.3,
1.4, 2.3, 2.4). For theses reasons quartile 1, quartile 3 and quartile 4 will be estimated by VAR

67
models and quartile 2 will be estimated by a VECM. The optimal number of lags is found
according to the Schwartz Information Criteria. Quartile 1 is a VAR(2) model while quartile 3
and quartile 4 are a VAR(1) model. Quartile 2 is a VECM(2) model.

Table 28 gives the results of the Granger Causality test. We cannot reject that equity prices
Granger cause CDS spreads in all quartiles. We can reject that CDS spreads Granger cause
equity prices in quartile 1 and quartile 2, but cannot reject that CDS spreads Granger cause
equity prices in quartile 3 and quartile 4. That is, there seems to be a two-way relationship
between CDS spreads and equity prices in quartile 3 and quartile 4.

Table 28: After quartiles – Granger Causality test


Equity prices Granger CDS spreads Granger
Before cause CDS spreads cause equity prices
Quartile 1 Yes* No
Quartile 2 Yes No
Quartile 3 Yes Yes*
Quartile 4 Yes Yes*
* indicates no serial correlation in disturbance terms

The test results for the Granger Causality test can only be trusted if the disturbance terms are not
serial correlated. In table 28 we see that we can trust that equity prices Granger cause CDS
spreads in quartile 1 and that CDS spreads Granger cause equity prices in quartile 3 and quartile
4. The serial correlation is not severe as the DW statistics are between 2.053 and 2.165 (appendix
14-17, table 3.5), but it is still present and should be taken into account when discussing the
relationship between CDS spreads and equity prices.

To further examine the relationship we consider the estimated models

9.5.2.1 Models with CDS spreads as dependent variable


We start by consider the models with CDS spreads as dependent variable. From table 29 it can
be seen that all the estimates are significant and the F test for all the models are rejected
(appendix 14-17, table 3.4). All the estimates for equity are negative and all the estimates for the
CDS spreads are positive, which is according to our expectations.

As expected we find that the equity price estimates increase in both lags the higher the quartile
number. The estimates for equity prices in quartile 3 and quartile 4 are much higher than the
estimates in quartile 1 and quartile 2 which indicates that the higher the CDS spreads the
stronger impact of equity prices on CDS spreads.

68
Table 29: After quartiles – VAR CDS
After CDS as dependent
Quartile 1 Quartile 2 Quartile 3 Quartile 4
CDS (t-1)
0.282* 0.367* 0.426* 0.400*
EQ (t-1) -0.495* -0.666* -2.908* -6.930*
CDS (t-2) 0.144* - - -
EQ (t-2) -0.305* - - -
R2 25.36% 23.61% 33.41% 31.81%
*indicates significance at 5% level

When considering the estimates for CDS spreads we see that the influence of the CDS spreads
estimates become higher the higher the quartile, but the increase is not as strong as for the
estimates for the equity prices. The equity price estimate in quartile 4 is 17 times higher than the
equity price estimate in quartile 1, whereas the estimate for CDS spreads in quartile 4 is only
1.75 times higher than the estimate in quartile 1 emphasising the increasing influence of equity
prices.

The R2 for the models are in the range from 23.61% to 33.41, which is more in line with our
previous results.

9.5.2.2 Models with equity prices as dependent variable


Next we study the models with equity price as dependent variable. From table 30 we see that it is
only the estimates for the equity prices that are significant. We reject for quartile 1-3 that all the
estimates are equal to zero at the same time even though the result for quartile 2 is a borderline
result as the F test is 0.0421. We cannot reject the F test for quartile 4 but again the result is
borderline as the F test is 0.0516 (appendix 14-17, table 3.4).

Table 30: After quartiles – VAR EQ


After EQ as dependent
Quartile 1 Quartile 2 Quartile 3 Quartile 4
CDS (t-1)
0.024 0.014 0.008 0.002
EQ (t-1) -0.180* -0.135* -0.107* -0.105*
CDS (t-2) 0.0002
EQ (t-2) -0.128*
R2 4.77% 2.27% 1.7% 1.36%
*indicates significance at 5% level

69
Opposite what we have found earlier the value of the equity price estimates decreases the higher
the quartile, and the same is true for the R2. The R2 are though higher than in the previous period
as it is between 1.36-4.77, but still low. Again the results are quite weak and based on the above
analysis it does not seem like CDS spreads are influencing equity prices in any significant way.
All in all we must conclude that the influence of equity price on CDS spreads is more significant
than the other way around.

9.5.3 Summary
From the above analyses we can conclude that equity prices lead CDS spreads and that a two-
way relationship between equity prices and CDS spreads before the beginning of the crisis is to
weak to be considered. When comparing the strength of the estimates we see that the influence
of equity prices on CDS spreads is more pronounced after the beginning of the crisis than before.
That is, the influence from equity prices to CDS spreads is higher the higher the CDS spreads
and the more volatile market conditions.

9.6 Rating
Credit rating is a measure of the general creditworthiness of an obligator or the creditworthiness
of an obligator with respect to a specific debt security (Standard & Poor’s:2006:8). The final
rating of a firm or debt security is a combined picture of the perceived business and financial
risks. Business risk is risk that arises in the firm’s environment and includes among others
country risk, industry characteristics, firm position and technology. Said in other words –
business risk is the obligators competition ability. On the other hand financial risk is risk
associated with the firm’s financial data and includes among others accounting, corporate
governance, financial policies and capital structure (Standard & Poor’s:2006:20). The rating of
firms and debt securities not only illustrate a combined picture of the risks but also makes the
entities comparable. The rating of firms or debt securities is done by rating agencies that are
specialised in this. The 3 major rating agencies are Standard & Poor Corporation, Moody’s
Investor Service and Fitch Ratings. The 3 agencies use different scales as it can be seen from
table 31. It can be seen from the scales that the lower the rating the shorter the distance to
default.

To be able to split our index into investment grade and high yield we downloaded ratings by
S&P, Fitch and Moody’s for all 265 firms – so we had 3 ratings for each firm. We calculated an
average rating for each firm to get a more precise picture using a rating conversion. Based on an

70
average rating we divided our data into investment grade and high yield. As can be seen from
table 31 anything below BBB+/Baa1 is high yield.

Table 31: Rating scale and definitions


S&P Fitch Moody's
Investment grade
Highest and strongest rating, capability to meet financial
AAA AAA Aaa commitment is extremely strong
High creditworthiness and very strong capability to meet
AA+ AA+ Aa1 financial commitment
A+ A+ A1 Strong solvency
BBB+ BBB+ Baa1 Exhibits adequate protection parameters
High yield
Less vulnerable to non payment than other speculative
BB+ BB+ Baa1 issues
Vulnerable to non payment, but has capacity to meet
B B B financial commitment
Currently vulnerable to non payment, and dependent upon
favourable business, financial, and economic conditions to
CCC CCC Caa meet financial commitment
D D D Payment Default
Source: Samson et al:2006:11

To test hypothesis H4.b and to analyse the relationship between CDS spreads and equity prices
more thoroughly we will examine if there is a difference in the relationship when splitting the
data in investment grade and high yield with the corresponding equity prices and CDS spreads
compared with the overall index. We expect a stronger relationship between equity prices and
CDS spreads for the high yield data compared with the investment grade data.

We re-calculate the mean and standard deviation for the time series split in rating. As expected
we find that the mean of high yield CDS spreads is about 5 times higher than the mean of
investment grade CDS spreads. The same is valid for the standard deviation.

When studying the mean and standard deviation for equity prices the difference is not so distinct.
The mean for high yield equity prices is 1.5 times smaller than the mean for investment grade
equity prices. The difference is even smaller when considering the standard deviation where the
standard deviation for high yield equity prices is 1.2 times higher than the standard deviation for
investment grade equity prices.

As usual CDS spreads are more volatile than equity prices and the difference between CDS
spreads and equity is highest in the high yield segment. The standard deviation for high yield
CDS spreads is 35 times higher than the standard deviation for high yield equity prices. While

71
the standard deviation for investment grade CDS spreads are “only” 8 times higher than the
standard deviation for investment grade equity prices.

Table 32: Rating – descriptive statistics


CDS Equity
IG HY IG HY
Mean 58.219 286.740 44.375 29.981
St. dev. 55.556 292.381 7.002 8.416

The picture determined for CDS spread and equity prices is also confirmed when considering
figure 11 where the investment grade CDS spreads are lower than the high yield CDS spreads for
the whole period. The high yield CDS spreads increase steeply mid 2007 and is very affected by
the crisis. This also explains the big difference between the standard deviation for investment
grade CDS and high yield CDS. This result is not surprising since the underlying entity for the
high yield CDS spreads is rated much lower than the investment grade and therefore the distance
to default is smaller.

Table 11: Development in high yield and investment grade CDS spreads and equity prices
1600 70
1400 60
1200 50
1000 40
USD
UDS

800 30
600 20
400 10
200 0
0

EQIG EQHY
CDSIG CDSHY

The estimated correlations (table 33) confirm our expectations that the relationship between high
yield CDS spreads and corresponding equity prices is stronger than the correlation between
investment grade CDS spreads and the corresponding equity prices. The correlation is -66.4% for
high yield and 53.4% for investment grade.

72
Table 33: Rating -
correlation
Correlation
Investment -0.534
grade
High yield -0.664

So far our hypothesis H4.b has only been confirmed and to further examine this we will study
the estimated models.

Both the models for high yield and investment grade are rejected to be stationary at the original
data by the unit root test (appendix 18-19, table 1.1, 1.2, 3.1, and 3.2) and likewise the existence
of co-integration is rejected according to Johansen’s co-integration test (appendix 18-19, table
5.1). Since both high yield and investment grade are stationary at 1st difference according to unit
root test (appendix 18-19, table 1.3, 1.4, 3.3, and 3.4) VAR modelling is used to analyse the
relationship between the CDS spreads and equity prices. The optimal number of lags is found to
be 1 for high yield and 2 for investment grade according to the Schwartz Information Criteria.

Table 34: Rating – Granger Causality test


Equity prices Granger CDS spreads Granger
cause CDS spreads cause equity prices
Investment Yes* No*
grade
High yield Yes No*
* indicates no serial correlation in disturbance terms

The Granger Causality test indicates that it cannot be rejected that equity prices Granger cause
CDS spreads, but it can be rejected that CDS spreads Granger cause equity prices for both high
yield and investment grade. This result indicates that market participants reveal their pricing in
the equity market before they do the same in the CDS market. This result is consistent with the
previous results for index, that equity prices lead CDS spreads. The Granger Causality test would
not reject that investment grade CDS spreads Granger cause investment grade equity prices if it
is tested on a 10% significant level. This means, that there could be a weak two-way interaction
between the two markets on investment grade level.

To further examine the relationship we consider the estimates in the VAR models.

9.6.1 Models with CDS spreads as dependent variable


We start by studying the models with CDS spreads as dependent variable. From table 35 it can
be seen that all the estimates are significant (except CDS in the 2nd lag in investment grade) and

73
the F test for all the models are rejected (appendix 18-19, table 5.4). All the estimates for equity
prices are negative and all the estimates for the CDS spread in the 1st lag are positive, which is
according to our expectations. The estimate for CDS spread in the 2nd lag in investment grade is
negative, but is not significant, so we will not focus more on this estimate.

Table 35: Rating – VAR CDS


CDS as dependent
HY IG
CDS (t-1) 0.416* 0.401*

EQ (t-1) -6.380* -1.263*


CDS (t-2) -0.047
EQ (t-2) -0.497
2
R 27.35% 31.63%
*indicates significance at 5% level

As expected we find that the influence from equity prices on CDS spreads is stronger for the
high yield model than the investment grade model. The influence from the estimated 1st lag
equity prices is 5 times stronger for high yield than investment grade. The R2 for investment
grade is 16% higher than the R2 but the investment grade model does also have a higher number
of lags.

We conclude that the relationship between equity prices and CDS spreads is stronger for high
yield than investment grade.

9.6.2 Models with equity prices as dependent variable


We now consider the models with equity prices as dependent variable. From table 36 it can be
seen that none of the estimates are significant in the high yield index. Furthermore the F-test is
not rejected, so the high yield VAR model is not overall significant, which means, that we will
not analyse this model further.

In the investment grade index all the equity price estimates and 1st lag CDS spreads are
significant. The F-test for the investment grade VAR model is rejected (appendix 18-19, table
5.4) and the R2 is 2.63%. The significant estimates in the models are opposite signed compared
to what we expected which means that the CDS spreads is positive. This is interpreted in the way
that when the CDS spreads increase it will influence equity prices positively.

74
Table 39: Rating – VAR EQ
Equity as dependent
HY IG
CDS (t-1) 0.001 0.020*

EQ (t-1) 0.007 -0.112*


CDS (t-2) -0.011
EQ (t-2) -0.079*
2
R 0.02% 2.63%
*indicates significance at 5% level

The Granger Causality test for investment grade indicated that on a 10% significance level there
might be a two-way relationship between investment grade CDS spreads and equity prices. But
in the estimated model the influence from the significant CDS spread estimate is low and the R2
is also low. Based on this the indications that investment grade CDS spreads influence equity
prices seems questionable. All in all there might exist a weak two-way relationship but the
influence is so low that we will not consider it further. The Granger Causality test for high yield
rejected that CDS spreads Granger cause equity prices and this is further emphasised by the
estimated model.

9.6.3 Summary
From the above analysis we can conclude that the relationship between CDS spreads and equity
prices is stronger for high yield than investment grade. The Granger Causality test and the
estimated models all indicate this.

All in all our results are consistent with what we expected to find and we can accept our
hypothesis H4.b that the relationship between CDS spreads and equity prices is more
pronounced the lower the credit quality.

9.7 Rating split on quartiles


We split the rating indices for high yield and investment grade in quartiles. This is done to test
the robustness of the results found in section 9.6 and to further examine the weak indication of a
two way relationship for investment grade.

75
The indices for investment grade and high yield are split respectively into quartiles where
quartile 1 contains the 25% of firms with the lowest spread, quartile 2 contains the 2nd lowest
25% ect.

We calculate the mean and the standard deviation for each quartile in both the investment grade
and high yield index (table 37 and 38).

Table 37: Rating quartile – descriptive statistics CDS


CDS
Investment Grade High yield
Quartile 1 Quartile 2 Quartile 3 Quartile 4 Quartile 1 Quartile 2 Quartile 3 Quartile 4
Mean 28.59 44.52 65.37 109.75 120.69 197.59 346.27 560.43
St. dev. 22.42 30.67 58.41 106.47 86.70 201.71 248.54 587.21

To start with we consider the CDS spreads and as expected the mean increases the higher the
quartile for both investment grade and high yield. The mean for investment grade is 4 times
higher in quartile 4 than quartile 1 while the mean for high yield is 5 times higher in quartile 4
than 1. The mean for high yield is higher than the mean for investment grade in all quartiles. If
we compare high yield quartile 1 with investment grade quartile 1 and so forth we find that the
high yield mean is 4-5 times higher than investment grade for each quartile.

The standard deviation increases as expected the higher the quartile for both investment grade
and high yield. The standard deviation for investment grade is 5 times higher in quartile 4 than 1
while the standard deviation for high yield is 7 times higher in quartile 4 than 1.

The standard deviation for high yield is also higher than the standard deviation for investment
grade for all quartiles. However, it should be noticed that the standard deviation for investment
grade in quartile 4 is higher than the standard deviation for high yield in quartile 1. All in all the
standard deviation for high yield is 4-6 times higher than the standard deviation for investment
grade.

76
Table 38: Rating quartile – descriptiv statistics EQ
Equity
Investment Grade High yield
Quartile 1 Quartile 2 Quartile 3 Quartile 4 Quartile 1 Quartile 2 Quartile 3 Quartile 4
Mean 49.32 43.89 44.87 37.07 26.80 41.24 24.10 24.93
St. dev. 7.05 6.70 7.76 7.99 5.96 11.17 6.45 9.46

The pattern for equity prices is not as obvious. The mean for investment grade is decreasing the
higher the quartile except for quartile 3 but the difference in the mean between the quartiles is
not as high as in CDS spreads and the highest mean is only 1.3 times higher than the lowest. The
same is the case for high yield where the mean for quartile 1, 3 and 4 is very much at the same
level. Quartile 2 is though higher and almost 2 times higher than the lowest mean in quartile 3.

The standard deviation for investment grade is between 6.70 and 7.99 where the two highest
observations are in quartile 3 and 4. The standard deviation for high yield is between 5.956 -
11.174 where the highest is in quartile 2 which is also the quartile with the highest mean.

We continue to examine the relationship by considering the Spearman rank correlation for high
yield and investment grade for quartiles. We find that the correlation is strongest for high yield in
quartile 3, whereas the correlation is strongest for investment grade in quartile 4. In high yield
there does not seems to be any pattern regarding the strength and number of quartile, whereas the
correlation is increasing the higher the quartile in investment grade. The correlation in
investment grade is lower than high yield for all quartiles except for quartile 4 and this may
indicate a weaker relationship between CDS spreads and equity prices for investment grade
compared with high yield.

Table 39: Rating quartile -


correlation
Correlation
High Investment
Yield grade
Quartile 1 -0.744 -0.344
Quartile 2 -0.628 -0.367
Quartile 3 -0.819 -0.484
Quartile 4 -0.781 -0.831

77
To further examine the relationship between CDS spreads and equity prices we estimate the
relevant models.

9.7.1 High yield


To examine the relationship between the CDS spreads and equity prices in our high yield index
split in quartiles we firstly consider stationarity of the original data which is rejected by the unit
root test (appendix 24-27, table 1.1, 1.2, 2.1 and 2.2) and likewise the existence of co-integration
is rejected according to Johansen’s co-integration test (appendix 24-27, table 3.1), though not for
quartile 3. The two time series are stationary at 1st difference according to unit root test
(appendix 27, table 1.3, 1.4, 2.3, and 2.4) for quartile 1, 2, and 4. For theses reasons we estimate
the relationship with VAR modelling for quartile 1, 2, and 4 and VECM modelling for quartile 3.
The optimal number of lags are found to be two for quartile 2 and quartile 3 and 1 for quartile 2
and quartile 4 according to Schwartz Information Criteria.

Table 40: High yield quartile – Granger Causality test


High yield Equity prices Granger CDS spreads Granger
cause CDS spreads cause equity prices
Quartile 1 Yes* No*
Quartile 2 Yes No*
Quartile 3 Yes No*
Quartile 4 Yes No*
* indicates no serial correlation in disturbance terms

According to the Granger Causality test (table 40) we cannot reject that equity prices Granger
cause CDS spreads, but can reject that CDS spreads Granger cause equity prices. Equity prices
therefore lead CDS spreads but not the other way around. We can trust the results for quartile 1
since the residuals in the VAR and VECM models are not serial correlated (appendix 24-27,
table 3.5). The residuals for the VAR model in quartile 2, 3 and 4 are slightly serial correlated
when testing if equity prices lead CDS spreads, which means, that we cannot quite trust the
result. We will further study the estimated models for high yield.

9.7.1.1 Models with CDS spread as dependent variable


Firstly we examine the estimated model for high yield with CDS spreads as dependent variable.
All the estimates in the models for all quartiles are significant different from zero. The overall
significance of the models cannot be rejected and the R2 are between 14.93% and 22.01%. All
the estimates for CDS spreads are positive while all the estimates for equity prices are negative.

Consistent with our hypothesis we find that the 1st lag equity price estimates increase the higher
the quartile. The influence from equity prices on CDS spreads is almost 3 times higher in quartile

78
4 than quartile 1. Quartile 1 is the only model with a 2nd lag and the influence from equity
decreases as the 2nd lag is 1.7 times smaller than the 1st lag. This is in accordance with our
previous findings.

Table 41: High yield quartiles – VAR CDS


High Yield CDS as dependent
Quartile 1 Quartile 2 Quartile 3 Quartile 4
CDS (t-1) 0.271* 0.376* 0.236* 0.359*

EQ (t-1) -2.520* -2.598* -6.985* -7.446*


CDS (t-2) 0.134* - - -
EQ (t-2) -1.480* - - -
2
R 20.49% 22.01% 15.47% 14.93%
*indicates significance at 5% level

The results from the Granger Causality test are consistent with the estimated models. Based on
the significant estimates and relatively high R2 we can conclude that the negative influence of
equity prices on CDS spreads increases the higher the quartile.

9.7.1.2 Models with equity price as dependent variable


We continue by studying the models with equity price as dependent variable for high yield. From
table 42 we see that none of the estimates in quartile 2 and quartile 3 are significant different
from zero. In quartile 1 the only significant value is 2nd lag of equity prices, and in quartile 4 the
only significant value is 1st lag of equity price. The overall significance of the model is rejected
for all quartiles. The R2 is low for all quartiles and the significant estimates are low.

Table 42: High yield quartiles – VAR EQ


High Yield
EQ as dependent
Quartile 1 Quartile 2 Quartile 3 Quartile 4
CDS (t-1) 0.0003 0.003 0.001 -0.0001

EQ (t-1) -0.011 -0.022 0.014 0.057*


CDS (t-2) -0.003 - - -
EQ (t-2) -0.071* - - -
R2 0.59% 0.2% 0.31% 0.33%
*indicates significance at 5% level

79
The results from the Granger Causality test are consistent with the estimated models. All in all
we must conclude that high yield CDS spreads do not influence equity prices.

9.7.2 Investment grade


We continue by examining the estimated models for investment grade. To examine the
relationship between the investment grade CDS spreads and the corresponding equity prices we
start with examining the stationarity of the original data which is rejected by the unit root test for
all quartiles (appendix 20-23, table 1.1, 1.2, 2.1 and 2.2) and likewise the existence of co-
integration is rejected according to Johansen’s co-integration test for all quartiles (appendix 20-
23, table 3.1). The two time series are stationary at 1st difference according to unit root test
(appendix 43, table 1.3, 1.4, 2.3, and 2.4). So we estimate the relationship with VAR models for
each quartile. The optimal number of lags is found to be two for all quartiles according to
Schwartz Information Criteria.

Table 43: Investment grade quartiles – Granger Causality test


High yield Equity prices Granger CDS spreads Granger
cause CDS spreads cause equity prices
Quartile 1 Yes No*
Quartile 2 Yes Yes*
Quartile 3 Yes* No*
Quartile 4 Yes* Yes*
* indicates no serial correlation in disturbance terms

According to the Granger Causality test (table 5.3) we cannot reject that equity prices Granger
cause CDS spreads. We cannot reject that CDS spreads Granger cause equity prices in quartile 2
and 4 while we reject that CDS spreads Granger cause equity prices for quartile 1 and 3. We can
trust all results except that equity prices Granger cause CDS spreads for quartile 1 and 2. The
residuals in the VAR model for quartile 1 and 2 for CDS are slightly serial correlated. We will
further consider the estimated models for investment grade.

9.7.2.1 Models with CDS spreads as dependent variable


When analysing the estimates of the VAR models with CDS spreads as the dependent variable
(table 44) the sign of the estimates are positive for CDS spreads (except in the 2nd lag in quartile
2) and negative for equity prices. This is consistent with theory. All the estimates are significant
different from zero (except in the 2nd lag in quartile 3).

80
Table 44: Investment grade quartiles – VAR CDS
Investment
grade CDS as dependent
Quartile 1 Quartile 2 Quartile 3 Quartile 4
CDS (t-1) 0.319* 0.212* 0.367* 0.341*

EQ (t-1) -0.341* -0.353* -1.195* -2.912*


CDS (t-2) 0.119* -0.124* 0.011 -0.092*
EQ (t-2) -0.220* -0.628* -0.576* -0.886*
2
R 24.48% 9.18% 29.20% 25.09%
*indicates significance at 5% level

We study the estimated model for investment grade with CDS spreads as the dependent variable.
All the estimates in the models are significant different from zero except 2nd lag CDS spreadsfor
quartile 3. The overall significance of the models cannot be rejected and the R2 are in the range
of 24.48-29.20% except for quartile 2 where the R2 is only 9.18%. All the estimates for equity
prices are negative and almost all the estimates for CDS spreads are positive. However, the 2nd of
CDS spreads for quartile 1 and 3 are negative.

Consistent with our hypothesis we find that the influence of the equity estimates increases the
higher the quartile. The influence from equity prices on CDS spreads is about 8.5 times higher in
quartile 4 than quartile 1. For all quartiles except quartile 2, the 1st lag influence is higher than
the 2nd lag influence.

The results from the Granger Causality test are consistent with the estimated models. Based on
the significant estimates and relatively high R2 we can conclude that the negative influence of
equity prices on CDS spreads increases the higher the quartile. However, the relationship is
weaker in quartile 2 where the R2 is only 9.18% and it is difficult to say what is causing this.
Quartile 2 is also the quartile with the lowest correlation for high yield.

9.7.2.2 Models with equity price as dependent variable


We continue by studying the models with equity price as dependent variable for investment
grade. From table 45 we see that none of the estimates for CDS spreads are significant in quartile
1 and quartile 3. The 2nd lag of CDS spreads are significant in quartile 2 and 1st and 2nd lag are
significant for quartile 4. The overall significance of the models cannot be rejected for all
quartiles. The R2 are low and the influence from the significant CDS estimates are low.

81
Table 45: Investment grade – VAR EQ
Investment
grade EQ as dependent
Quartile 1 Quartile 2 Quartile 3 Quartile 4
CDS (t-1) 0.034 -0.005 0.001 0.009*

EQ (t-1) -0.170* -0.149* -0.088* -0.052


CDS (t-2) 0.004 0.023* 0.006 -0.006*
EQ (t-2) -0.120* -0.097* -0.067* -0.039
2
R 4.33% 3.49% 1.23% 1.40%
*indicates significance at 5% level

The Granger Causality test suggested that CDS spreads Granger cause equity for quartile 2 and
4, which this is also the quartiles where we find significant estimates for CDS spreads. But based
on the above analysis we must conclude that the influence is so low that we can conclude that
CDS spreads do not influence equity prices.

9.7.2.3 Summary
From the above analysis we find quite convincing results that high yield equity prices influence
high yield CDS spreads and both the Granger Causality test and the estimated models indicate
this. The influence increases the higher the quartile and the 1st lag equity prices increase from -
2.520 to -7.446. For the model with equity as dependent variable the Granger causality and
estimated model both indicate that high yield CDS spreads do not influence the corresponding
equity prices.

The influence from equity prices on CDS spreads in the model for investment grade is not as
strong. The negative influence increases the higher the quartile. Further 1st lag equity prices
increase from -0.341 to -2.912 for the high yield segment. For the model with equity as
dependent variable the Granger Causality test indicates that there might be a two-way
relationship for quartile 2 and 4 while it rejects for the quartile 1 and 3 that CDS spreads Granger
cause equity prices. This is confirmed by the estimated models that only quartile 2 and 4 have
significant estimates for CDS spreads. But the estimates are small, the R2 of the models are only
3.49% and 1.40% respectively. Though the F-tests are rejected.

Based on this we can accept our hypothesis 4.b that the relationship between CDS spreads and
equity prices are stronger for high yield than investment grade.

82
9.8 Sectors
To test hypothesis H4 and analyse the relationship between CDS spreads and equity prices more
thoroughly we would like to see if there is a difference in the relationship when splitting our
index data on 9 MSCI sectors33.

We re-calculate the mean and standard deviation for the sectors.


Table 46: Sector – descriptive statistics CDS
CDS Industry
Materials Communication Consumer Consumer, Energy Finance Industrial Technology Utility
Cyclical Non-cyclical
Mean 78.38 117.88 176.19 61.61 91.99 136.03 65.62 134.11 97.17
St. 93.73 116.17 177.9 40.44 64.97 196.97 62.54 175.85 72.62
Dev.

At first we consider the mean and standard deviation for CDS spreads. The sectors with the
highest mean are consumer cyclical and finance with a mean of 176.19 and 136.03 respectively.
Finance and consumer cyclical are also the two sectors with the highest standard deviation of
196.97 and 176.19 respectively. It can be seen from table 46 that the 4 sectors with highest mean
is also the 4 industries with the highest standard deviation. The sectors consumer non-cyclical
and industrial have the lowest means of 61.61 and 65.62 respectively and more over the lowest
standard deviation of 40.44 and 62.54 respectively. Energy, materials and industrial vary in the
rankings for mean and standard deviation, but they stay in the ranking area 6th to 8th.

Table 47: Sector – ranking mean/st.d. CDS

CDS
Highest mean Highest st. dev.
1 Consumer Cyclical Finance
2 Finance Consumer Cyclical
3 Technology Technology
4 Communication Communication
5 Utility Materials
6 Energy Utility
7 Materials Energy
8 Industrial Industrial
Consumer, Non- Consumer, Non-
9 cyclical cyclical

33
Short description of the sectors are given in appendix 1

83
If we consider the mean and standard deviation for equity price the same pattern is not present.
The sectors with the highest mean is energy and industrial with a mean of 50.32 and 48.84
respectively while the sectors with the highest standard deviation are energy and finance with a
standard deviation of 15.8 and 11.8 respectively. The sectors with the lowest means are
communication and consumer cyclical with a mean of 27.62 and 34.78 respectively. The sectors
with the lowest standard deviation are consumer non-cyclical and communication with a
standard deviation of 4.26 and 5.32 respectively. There are not an obvious pattern between
standard deviation and mean but this is probably because the range for both estimates is quite
low.

Table 48: Sector - descriptive statistic EQ


Equity Industry
Materials Communication Consumer Consumer, Energy Finance Industrial Technology Utility
Cyclical Non-cyclical
Mean 48 27.62 34.78 39.01 50.32 45.27 48.84 36.81 37.05
St. Dev. 9.07 5.32 5.66 4.26 15.8 11.8 10.14 7.31 5.96

Table 49: Sector – ranking mean/st.d. EQ

Equity
Highest mean Highest st. dev.
1
Energy Energy
2
Industrial Finance
3
Materials Industrial
4
Finance Materials
Consumer, Non-
5
cyclical Technology
6
Utility Utility
7
Technology Consumer Cyclical
8
Consumer, Cyclical Communication
Consumer, Non-
9
Communication cyclical

We confirm once again that CDS spread is more volatile than equity prices as the standard
deviation of CDS spreads range from 40.44 to196.97 while the standard deviation of equity
prices range from 4.26 to15.8.

84
We want to determine if the amount of debt in a sector will have an impact on the relationship
between CDS spreads and equity prices. We hypothesize that the less debt a sector has the
weaker the relationship between equity prices and CDS spreads (hypothesis H5).

To get a comparable measure of debt we need to take the numbers and sizes of firms in the
different sectors into account. Instead of using the absolute debt for a firm we calculate the
financial leverage for each firm in each sector. Financial leverage, also called gearing, is the use
of debt to increase the expected return on equity and is measured by the ratio of debt to debt plus
equity (Brealey:2006:998). Because we include equity we take into account that a high amount
of debt is only dangerous if the firm does not have enough equity to cover the debt.

Financial leverage = Debt


Debt + Equity

As a proxy for debt we use net debt from the individual firms. Net debt is a measure for a firm’s
overall debt situation and is calculated by the following (Brealey:2006:446):

Net debt = Short-term debt + Long-term debt – cash & cash equivalents

As a proxy for equity we use market capitalization. Market capitalization is the total dollar
market value of all of the firm’s outstanding shares and is calculated by the following:
(Brealey:2006:449)

Market capitalization = # outstanding shares x current market price for 1 share

We use the net debt and market capitalization for each individual firm in the different sectors to
calculate the financial leverage for each firm in each sector. Because we are comparing sectors
we calculate a sector’s average for financial leverage. The average financial leverage for each
sector can be seen in table 50

85
Table 50: Sector – financial leverage
Total
Total net Financial Average Average
In mil. USD Market cap.
debt leverage mkt. Cap net debt
Utility 192,185.57
172,828.89 47% 106,76.98 9,601.6
Financial 726,564.83
3,456,409.10 41% 25,948.74 123,443.18
Communication 260,973.43
188,358.70 31% 37,281.92 26,908.39
Con cyc 833,160.43
388,647.37 25% 15,148.37 7,066.32
Energy 409,116.64
96,986.12 20% 24,065.68 5,705.07
Materials 239,411.44
87,301.92 17% 11,970.57 4,365.1
Industrial 667,397.02
575,837.84 16% 15,890.41 13,710.42
Con non-cyc 1,697,843.81
223,058.46 14% 28,777.01 3,780.65
Technology 643,185.24
16,347.58 3% 33,851.85 860.40
Source: Bloomberg

We find that the sectors with the highest financial leverage are utilities and financial with a
leverage of 47% and 41% respectively. Technology and consumer non-cyclical have the lowest
financial gearing with a leverage of 3% and 14% respectively.

We would expect to find the highest CDS spreads in the sectors with the highest leverage as the
risk of default increases for firms with very high financial leverage, which makes it more
expensive to be protected against default. If we compare the degree of financial leverage with the
CDS spread mean in each sector we find that for some sectors this is true (see table 51).
Financial is the sector with the 2nd highest mean and also the sector with the 2nd highest leverage.
But the sector utilities has the 5th highest mean and the highest leverage. And the sector
technology has the 3rd highest mean but the lowest leverage. All in all the pattern is not
convincing.

86
Table 51: Sector - comparison of correlation, leverage and
mean
Correlation Correlation Leverage Mean
rank rank rank
Technology -0.8686 1 9 3
Finance -0.8478 2 2 2
Utility -0.6311 3 1 5
Industry -0.4635 4 7 8
Communication -0.4622 5 3 4
Energy -0.4494 6 5 6
Consumer, -0.3890 7 4 1
cyclical
Material -0.3672 8 6 7
Consumer, non- -0.3437 9 8 9
cyclical

Based on the calculated financial leverage we will expect the relationship between CDS spreads
and equity prices to be strongest in utility, financial and communication while weakest in
technology, consumer non-cyclical, and industrial. Firstly we will consider the Spearman rank
correlations between CDS spreads and equity prices for each sector (see table 51).

The most correlated sector is technology, which is also the sector with the lowest leverage so this
is not quite what we expected to find. But the 2nd and 3rd most correlated industries are financial
and utility and these are the sectors we found to be most leveraged. But otherwise an obvious
pattern is not present.

Initial we have not found anything that really supports our hypothesis. To further examine this
we estimate the relevant models for the individual sectors.

Data for all 9 industries are rejected to be stationary at the original data by the unit root test
(appendix 28-36, table 1.1, 1.2, 3.1, and 3.2) and likewise the existence of co-integration is
rejected according to Johansen’s co-integration test (appendix 28-36, table 5.1) in 8 of the 9
sectors. Technology is the only sector that is co-integrated, so to analyse the relationship
between CDS spreads and equity prices in the technology sector VECM is used. The other 8
sectors are stationary at 1st difference according to unit root test (appendix 28-36, table 1.3, 1.4,
3.3, and 3.4), so here VAR modelling is used. The optimal number of lags found according to
Schwartz Information Criteria. Most sectors will be estimated by two lags except communication
(1 lag), energy (3) and utility (1).

87
In table 52 the Granger Causality test for the 9 sectors are stated. It cannot be rejected for any of
the 9 sectors that equity prices is Granger causing CDS spreads. We find that it can be rejected
that CDS spreads is Granger causing equity prices for all sectors except for materials. This
suggests that there might exist a two-way relationship between CDS spreads and equity prices
for materials. But there are some problems with serial correlation and the results that equity
prices are Granger causing CDS spreads for communication, consumer cyclical, consumer non-
cyclical, and utility cannot be trusted (appendix 28-36, table 5.5). The result that CDS spreads
are not Granger causing equity prices cannot be trusted for consumer cyclical as well.

Table 52: Sector – Granger causality test


Equity price CDS spreads Granger
interaction term cause equity price
Granger cause CDS interaction term
Up market spreads
Communication Yes No*
Consumer Yes No*
cyclical
Consumer non- Yes No
cyclical
Energy Yes* No*
Financial Yes* No*
Industrial Yes* No*
Materials Yes* Yes*
Technology Yes* 34
No*
Utility Yes No*
* indicates no serial correlation in disturbance terms

We will further examine the relationship by studying the estimated models. Firstly we consider
the models with CDS spreads as dependent variable.

9.8.1 Models with CDS spreads as dependent variable


The 9 models for each sector are all overall significant and also all of the estimates are
significant. The overall explanatory power is on average 18.22% and is ranging from 33.90% for
consumer cyclical to only 3.30% for communication.

The estimates for CDS spreads are all positive in all the models while the estimates for equity
prices are all negative in all the models. For the models that include more than 1 lag we find that
the influence of equity prices decreases the higher the lag.

34
The GG values in technology gives the same result as the Granger Causality test, since GG = 0.0019.

88
Table 53: Sector – VAR CDS
CDS as dependent
Financial Consum Commu- Technolo Materi- Utility Industri Consu- Energy
-er nication -gy al -al mer non-
cyclical cyclical
CDS (t-1) 0.262* 0.379* 0.132* 0.242* 0.339* 0.238* 0.319* 0.250* 0.212*
EQ (t-1) -4.542* -3.365* -2.038* -1.129* -0.856* -0.818* -0.713* -0.621* -0.598*
CDS (t-2) -0.08* 0.136* 0.106* 0.090* 0.147* 0.190*
EQ (t-2) -1.741* -1.276* -0.393* -0.466* -0.342* -0.410*
CDS (t-3) 0.083*
EQ (t-3) -0.296*
2
R 20.20% 33.90% 3.30% 8.14% 22.91% 7.94% 25.26% 17.38% 24.99%
Leverage 2 4 3 9 6 1 7 8 5
rank
Mean rank 2 1 4 3 7 5 8 9 6
*indicates significance at 5% level

We use the 1st lag estimate for equity prices as an indication of the strength of the negative
influence from equity prices on CDS spreads. From the table above the strongest influence is
found for financial, and second strongest for consumer cyclical and so forth. The next lowest row
in table 53 indicates the leverage rank. We see that it is only the technology, utility, and energy
sectors that have a different influence than expected. If instead we examine the mean rank, the
connection seems to be a bit stronger but still not convincing. Furthermore utility and technology
have low R2 compared to the other sectors.

All in all we must conclude that the pattern is not convincing for all sectors but for financial,
consumer cyclical and communication the data seems to show a pattern. It seems like these 3
sectors have some of the highest leverage, standard deviations and means. But regarding our
hypothesis we have not found a connection strong enough to accept it. We have found
indications that for some sectors high leverage is associated with a strong influence from equity
prices on CDS spreads.

89
9.8.2 Models with equity prices as dependent variable
The estimated models35 with equity prices as dependent variable are quite weak as we have also
found earlier. As mentioned the Granger Causality test indicated that CDS spreads do not
influence equity prices for all the sectors except materials. The models for communication,
consumer cyclical, and technology are not overall significant and none of the estimates in the
respective models are significant different from zero. These sectors do also have the 3 lowest R2,
being between 0.11% and 0.33%. Financial and materials are the only sectors that have
significant estimates for CDS spreads for 2nd and 1st lag respectively. Materials is also the only
sector where a two-way relationship is suggested by the Granger Causality test. But the R2 is
only 1.44% and the estimates quite low so if any influence it must be quite low.

Table 54: Sector – VAR EQ


Equity as dependent
Consum- Utility Commu- Consu- Financi Techno Industri Materials Energy
er non- nication mer -al -logy -als
cyclical cyclical
CDS (t-1) -0.010 -0.003 -0.001 -0.001 -0.0002 0.001 0.016 0.017* 0.008
EQ (t-1) -0.106* -0.091* -0.024 -0.007 -0.081* -0.050 -0.081* -0.072* -0.119*
CDS (t-2) -0.0003 0.001 0.004* -0.017 0.003 0.002
EQ (t-2) -0.122* -0.059 -0.038 -0.046 -0.047 -0.105*
CDS (t-3) 0.008
EQ (t-3) 0.0401
R 2
2.25% 0.80% 0.11% 0.33% 1.28% 0.27% 1.28% 1.44% 2.77%
*indicates significance at 5% level

9.8.3 Summary
All in all we must conclude that the estimated models are not convincing enough to accept
hypothesis H5. But we did find that the high leveraged sectors as financial, consumer cyclical
and communication also were the sectors where the estimates for equity prices influenced CDS
spread the most. This indicates that there might be a connection and that there is a stronger
relationship between CDS spreads and equity prices for high leveraged sectors.

35
We use the 1st lag estimate for CDS spreads as an indication of the strength of the negative influence from CDS
spreads on equity prices.

90
9.9 Results
In the analysis section 9.1 to 9.8 we examine how CDS spreads and equity prices are related. At
first we noticed the specific characteristics of CDS spreads. They are extreme volatile and for the
overall index the CDS spreads standard deviation was 16 times higher than for equity prices.
When comparing the data for CDS spreads and equity prices graphically (see graph 10 section
9.4) we noticed how intense the CDS spreads reacted to the crisis compared with the equity
prices. We splitted our data into before the beginning of the crisis and after the beginning of the
crisis to test if the relationship change in a more volatile period that mostly consisted of
deteriorating market conditions compared with a more stable market. To examine how the
creditworthiness of the underlying debt/bond influence the relationship we splitted our data into
high yield and investment grade. For the models mentioned so far we splitted them into quartiles
and reestimated them to see how the size of CDS spreads influenced the results but also to
examine if there were any hidden tendencies. At last we splitted our data into sectors and
compared the estimated models with the financial leverage in each sector.

Throughout our analysis we have estimated 29 VAR/VECM models. For each model we have
analysed the relationship between equity prices and CDS spreads by considering the Granger
causality test and the estimated models with respectively CDS spreads and equity prices as
dependent variable.

For all the models with CDS as dependent variable we found that equity prices are influencing
CDS spreads negatively. This relationship was confirmed by the Granger causality test and the
actual estimates in the models.

For the models with equity prices as dependent variable we found that CDS spreads are not
influencing equity prices. This was also confirmed by the Granger causality test and the actual
estimates in the models. But for the model for investment grade and the model for the period
before the beginning of the crisis there were weak signs of CDS spreads influencing equity
prices. We determined though that when taking the size of the estimates and the R2 into account
the influence was too small to be significant.

Furthermore the negative relationship was also confirmed by the estimated Spearman rank
correlations that ranged from -34.37% to -97.3%36. All in all our results are convincing and

36A correlation of 97.3% seems unrealisticly high when considering empirical data and could be due to that
the data was influence by other common factors. But since our correlation results do not stand alone, but are
held up against our estimated models we will not comment more on this.

91
consistent throughout our robustness tests. We included exogenous variables to test if the results
were due to a third common variable that was not included in the model. But the relationship
seemed stable.

The finding of the negative relationship between equity prices and CDS spreads correspond to
the findings of among others Bystrøm (2004), Hafer (2008), Lake & Apergis (2009) and Chang
et al. (2008). The findings that equity prices influence CDS spreads and that CDS spreads do not
influence is consistent with the findings of Norden & Weber (2004), Forte & Pena (2008) and
Fung (2008). Though Norden & Weber (2004) find that CDS spreads influence equity prices in 5
out of 58 firms and Fore & Pena (2008) find that CDS spreads influence equity prices in 5 out of
65 cases. Furthermore, Fung et al. (2008) find that there is a two-way relationship between
equity prices and CDS spreads in the high yield segment, whereas we find a very weak link
between equity prices and CDS spreads in the investment grade segment, that we though chose
to dismiss due to lack of significance of estimates and low R2. That is, our results are more or
less consistent with the findings of other researchers that have found that equity prices influence
CDS spreads and not the other way around37.

Table 55 gives an overview over our results.

Table 55: Overview of results


Data Result Data Result
Index Equity prices Splitted in quartiles The relationship is
influence CDS stronger the higher
spreads negatively the quartile
Exogenous variables The relationship was
intact
Split in before and The relationship was Splitted in quartiles The relationship is
after the beginning of stronger after the stronger the higher
the credit crisis beginning of the the quartiles and
crisis stronger in after the
beginning of the
crisis
Split in high yield The relationship is Splitted on quartiles The relationship is
and investment grade stronger for high still strongest for
(Rating) yield compared with high yield and the
investment grade strength increases the
higher the quartile
Split in sectors For some sectors the
higher the average
leverage the stronger
the relationship

37
If the reader has an interest in researchers that have found that CDS spreads influence equity prices, we refer to
literature review (section 5).

92
Besides that we consistently found that equity prices influence CDS spreads negatively we also
found that the higher the CDS spreads the stronger the relationship in most cases.

First we splitted our time series into before and after the beginning of the crisis and we found
that the relationship after the beginning of the crisis was stronger than before the crisis. As can
be seen from figure 10 the CDS spreads reacted strongly to the crisis and increased steeply.
Moreover the standard deviation and mean after the beginning of crisis was higher than before
(table 9).

That the relationship becomes stronger the lower the credit rating was confirmed when we
splitted our data into high yield and investment grade. We found that the relationship between
CDS spreads and equity prices was stronger for high yield than investment grade. These findings
are consistent with the findings of Fung (2008).

We tested this further by splitting our models for index, before and after the beginning of the
crisis and high yield and investment grade into quartiles. We found that when we compared the
quartiles the relationship for quartile 2 was stronger than quartile 1 and the relationship for
quartile 3 was stronger than quartile 2 and at last that the relationship for quartile 4 was stronger
than quartile 3. We found this consistently for all the models.

At last we divided our index into sectors and estimated the average financial leverage in each
sector to see if there is any relationship between the amount of debt in a sector and the strength
of the relationship. Based on our previous findings that the relationship becomes stronger the
higher the spread we expected the sector with high financial leverage also to have a stronger
relationship than sectors with low financial leverage. The results were not as clear cut. There
was, however, a relationship between the highest leveraged industries financial, communication
and consumer cyclical as these also had the strongest relationship. But it was difficult to make a
precise conclusion and this could have been due to a coincident. Moreover for example financial
has specific characteristics that we did not go into the depth with and this may also have been the
case for the other sectors.

Until this point we have only compared the splitted data with the other part of the splitedt data.
Now we will try to compare the models with each other. In table 56 we have made a comparison
of the results for the models38.

38 To see a table including the quartiles see appendix 43

93
Table 56: Comparison of models
Ranking
of R- Ranking of 1st lag
Rank 1st lag 2nd lag 3rd lag R-square square estimate
3.6 High yield -6.38 27.35 5 1
2.6 Financial -4.542 -1.741 20.2 10 2
2.5 Consumer, cyclical -3.36513 -1.27632 33.9 3 3
2.1 Index, After -2.12302 38.23 1 5
2.0 Index -2.25953 -0.83542 34.17 2 4
1.5 Investment grade -1.263 -0.497 31.63 4 7
1.1 Industrial -0.713 -0.466 25.26 6 11
1.0 Materials -0.856 -0.393 22.91 8 9
1.0 Communication -2.03795 3.3 14 6
1.0 Energy -0.598 -0.41 -0.296 24.99 7 13
0.9 Index, Before -0.544 -0.413 22.37 9 14
Consumer, non-
0.8 cyclical -0.62065 -0.34168 17.38 11 12
0.7 Technology -1.12926 8.14 12 8
0.6 Utility -0.818 7.94 13 10

The models are sorted after the value of the 1st lag of the estimate for equity prices. But we
cannot only consider the estimates for equity prices, we also have to compare this with the R2
that indicates the overall explanatory power of the models. The problem with only looking at the
estimates can be illustrated by observing the sector communication where the equity estimate is
relatively high but the R2 is only 3.3%.

As can be seen from the table the estimated model for the overall index is actually quite strong.
The estimate for equity prices is the 4th strongest and the R2 is the 2nd strongest. Since this model
includes all our data we will use this as a benchmark when evaluating the other models. This
means that when evaluating the strength we will compare the size of the first lag estimate for the
respective model with the size of the 1st lag estimate of the overall index model and compare the
R2 of the respective model with the R2 of the overall model. We find that the influence from
equity prices on CDS spreads is stronger than for the overall index model in39the models for
financial, high yield, consumer cyclical and index after model

39
It is difficult accurately to compare the models since the number of lags varies and both the estimates and R2
U VWXY ^Y_
should be considered. To get an indication we calculate the following: & _ Every number greater
U VWXZ[\]Y ^Z[\]Y
than 2 indicates that the model is stronger than the overall index model. It is not an exact measure.

94
As mentioned besides financial and consumer cyclical all the other estimated models for sectors
indicate a weaker relationship than the overall model. The sizes of the estimates are smaller for
all compared with the estimate for the overall index and also the R2 is smaller. The weakest
models are for technology and utility where the R2 and the estimates are particular low.

When comparing the models before and after the beginning of the credit crisis we find that the
model for “after” is stronger than the model for “before” and stronger than the overall index. The
estimate for after the crisis is 0.94 times lower than the estimate for the overall index but the R2
is 1.12 times higher than the overall index. Moreover, we found that the after period was co-
integrated which suggest that equity prices and CDS spreads are driven by the same common
factor in the long run suggesting a stronger relationship between them.

When comparing the models for rating we find that the model for high yield is stronger than the
model for investment grade but the model for high yield is also stronger than the model for the
overall index. The 1st lag estimate for high yield is -6.38, which is almost 3 times higher than the
corresponding estimate for the overall index model. The R2 is 1.2 times higher for the overall
index compared with the high yield. But when taking both the estimate and the R2 into
consideration we find that the model for high yield is stronger than the overall index. The
opposite is true for the model for investment grade.

The conclusions made throughout the analysis are still consistent when taking the comparison
with the overall index into account.

We will now consider the models we splitted into quartiles. When comparing the quartiles we
found that the higher the quartile the stronger the influence. We will now compare the models for
the quartiles with the overall models. This means that for the quartiles on index we will compare
them with the overall index model and for the quartiles on high yield we will compare them with
the overall model for high yield and so on.

For index and the indices before and after the beginning of the crisis we find that quartile 3 and 4
are stronger models than the overall model while quartile 1 and 2 are weaker models (see table
47 for example and appendix 44 for all the tables).

95
Table 57: Index before quartile ranking
Rank 1st lag 2nd lag R2
2.9 Index, Before - Q4 -1.308 -0.929 12.18
2.5 Index, Before - Q3 -0.862 -0.492 20.92
2.0 Index, Before -0.544 -0.413 22.37
1.3 Index, Before - Q2 -0.292 -0.305 16.62
1.0 Index, Before -Q1 -0.117 -0.185 16.96

The same picture is not valid for investment grade and high yield when we compare the quartiles
with the overall models. We find that the overall model for high yield is stronger than all the
quartiles while for investment grade only quartile 4 is stronger than the overall model for
investment grade.

We find that in for the models for quartile 3 and 4 the influence is stronger than the overall
model and for quartile 1 and 2 the influence is weaker. This again suggests that the level of the
CDS spread effects the strength of the influence. It is worth noticing that the R2 is lower for all
the quartiles model compared with the overall model. But the picture is not valid for the models
for high yield and investment grade. In these models only quartile 4 is stronger than the overall
model. It is worth noticing that the result that the overall model is so strong is mainly due to
much higher R2 for the overall models. When the data is splitted in quartiles it looses explanatory
power and all the R2 for quartiles are lower than for the overall models. This is consistent for all
the estimated quartiles models.

We are now at the point in the study where we will consider our stated hypotheses in section 6
again. In table 58 we have listed the hypotheses our decision and argument for the decision. As
can be seen we can accept all our hypotheses. H5 is accepted with certain conditions and the
acceptance of H6 is inconclusive.

96
Table 58: Hypothesis, decision and arguments
Hypothesis Decision Argument
H1: The correlation Accept - For all 29 models with CDS spread as
between CDS spreads dependent the estimates for equity prices
and equity prices is were negative signed.
negative. - The Spearman rank correlations were
negative for all models.
H2: Equity prices Accept - For all 29 models with CDS spread as
influence CDS spreads dependent variable the equity estimates
and CDS spreads do not were significant, the R2 relatively high and
influence equity prices. the F-tests were rejected.
- The Granger causality test indicated for all
models that equity prices influence CDS
spreads.
- 27 of the 29 models with equity price as
dependent variablehad insignificant CDS
estimates and low R2 and the Granger
causality test also indicated that CDS
spreads do not influence equity prices.
- 2 of the 29 had significant estimates for
CDS spreads but the influence and the R2
were too low to have a real impact.
H3: The relationship Accept - When including exogenous variables in the
between CDS spreads model we still found that equity prices
and equity prices is still influence CDS spreads and that CDS
strong when including spreads did not influence equity prices.
exogenous variables.
H4.a: The relationship Accept - The influence from equity prices on CDS
between CDS spreads spreads was stronger after the beginning of
and equity prices is the crisis than before the beginning of the
stronger under crisis.
deteriorating market
conditions.
H4.b: The strength of the Accept - The influence from equity prices on CDS
relationship between spreads was stronger for high yield model
CDS spreads and equity than the investment grade model and the
prices becomes more overall model.
pronounced the lower the
credit quality.
H4.c: The strength of the Accept, when - The influence of equity on CDS spreads
relationship becomes comparing was higher the higher the quartile. And
more pronounced the quartile with when comparing with the overall model
higher the CDS spreads quartile. quartile 3 and 4 were stronger than the
of the underlying index. overall model and quartile 1 and 2 weaker.
Except for rating where only quartile 4 was
stronger for investment grade while not of
the quartiles were stronger for high yield.
H5: The strength of the Inconclusive - We found that the relationship was
relationship between stronger for the sectors financial,
CDS spreads and equity communication and consumer cyclical.
prices becomes more This indicated a relationship but the results
pronounced the more were not unambiguous.
debt an sector has.

When testing the validity of our hypothesis we test at the same time if the underlying rationale
for the hypothesis is valid.

97
By accepting H1 we accept that when equity prices increase the distance to default decreases,
which makes the price of protection smaller, i.e. the CDS spread decreases. The acceptance of
H2 indicates that the higher amount of analysts watching the equity market compared to the CDS
market influence that the price discovery takes place in the equity prices before CDS spreads.
But whether or not this is due to a higher form of efficiency or insider trading is beyond the
scope of our paper. When accepting H3 we accept that the 3 exogenous variables in our model
do influence the relationship. Moreover by accepting H4 we accept that investors are more likely
to react when they are in situations where the probability of default increases. That is, in
deteriorating market conditions, investment in low credit quality, or when investing in CDS with
high spreads. Since we did not find an unambiguous answer to H5 we cannot accept the rationale
that investors who has invested in sectors with a high debt will be more responsive to changes in
equity prices and CDS spreads than investors who have invested in sectors with lesser debt.

9.10 Equity volatility


In section 9.1 to 9.8 we find convincing evidence that CDS spreads and equity prices are
negatively correlated and that equity prices influence CDS spreads and not the other way around.
We accepted our hypothesis 1 and 2 and due to this we will briefly examine how equity volatility
affects CDS spreads.

In section 4.5 we described the Merton model, where equity volatility is included as a variable
that affect the CDS spreads. Supposedly equity volatility should influence CDS spreads
positively. This seems logical since the volatility also is an indicator of how risky equity is. The
risk of equity portrait the risk of the specific firm and as the risk increases the distance to default
decreases. Therefore when the equity volatility increase the CDS spreads should increase as well.

Since we have found the suggested relationship between the equity prices and CDS spreads to
hold we will include equity volatility as a 3rd variable in our model. This time we will only
examine the model with CDS spread as dependent variable as we have now established that CDS
spreads do not influence equity prices. Moreover we are not interested in how equity volatility
affects equity prices.

To create the variable equity volatility we need to determine how to estimate equity volatility.
Volatility is the standard deviation and since standard deviation needs more than one observation
to be calculated we estimate equity volatility as the rolling standard deviation. As Bystrøm
(2004) we estimate the rolling standard deviation for different lengths and we consider 1 month,
2 months, 3 months, 6 months and 1 year.

98
To get a preliminary indication of the results we estimate the Spearman Rank correlation. As
expected all the correlations are positive. Bystrøm (2004) found the strongest correlation for 3
months (Bystrøm:2005:8) which is normally also the length used by traders, but we find it to be
highest for 1 year. This is quite surprising as it means that CDS spreads are more influenced by a
standard deviation that covers a whole year and the correlation is decreasing from 1 month to 6
months so we would expect 1 year to be the lowest. Besides the high estimate in 1 year we get
that 1 month and 2 months is also quite high. To further examine this we consider the estimated
models.

Table 59: Volatility -


correlation
Correlation
1 month 0.5493
2 months 0.5296
3 months 0.4733
6 months 0.3977
1 year 0.6010
None of the volatility estimates are stationary and based on previously findings we know that
either are equity prices or CDS spreads. But we cannot reject the existence of co-integration
according to Johansen’s Co-integration tests except for the model including 1 year volatility
(appendix 37-42, table 3.1). The following models are therefore estimated by VECM modelling
except for the model including 1 year volatility where we use a VAR model.

As earlier the Granger Causality test indicates that equity prices Granger cause CDS spreads and
not the other way around. But equity volatility is also Granger causing CDS spreads. The results
are the same for all the 5 models and both the Granger Causality test and the Granger Gonzales
test show the same relationship: equity prices Granger cause CDS spreads and equity volatility
Granger cause CDS spreads.

Table 60: Volatility – Granger Causality test


Investment Equity prices Granger Equity volatility Granger
cause CDS spreads cause CDS spreads (GG)
grade
(GG)
1 month Yes* Yes*
2 month Yes* Yes*
3 month Yes* Yes*
6 month Yes* Yes*
40 Yes* Yes*
1 year
* indicates no serial correlation in disturbance terms

40
This is the result from the Granger Causality test.

99
We consider the estimated VECM models and VAR model in table 61. All the estimates are
significant different from zero except 2nd lag of the CDS estimate. The overall F-test is rejected
for all the models. As expected the equity price estimate influence CDS spreads negatively and
the CDS spreads estimates are positive in all lags and for all models. The equity volatility is
positive for all the VECM models in the 1st lag. In the 2nd lag and for the 1 year model the
estimate is negative which is not in accordance with what we expected. The influence decreases
as the number of lags increases for all the estimates except for 2 months equity.

Table 61: Volatility – VAR/VECM CDS


1 month 2 month 3 month 6 month 1 year
CDS (t-1) 0.425*
EQ (t-1) -2.226*
Time (t-1) -13.843*
D_CDS (t-1) 0.417* 0.373* 0.359* 0.362*
D_EQ (t-1) -1.971* -1.904* -1.990* -1.781*
D_Time (t-1) 4.295* 15.329* 18.047* 57.519*
D_CDS (t-2) 0.048 0.032 0.037
D_EQ (t-2) -0.949* -0.859* -0.820*
D_Time (t-2) -17.594* -13.895* -50.233*
R2 33.53% 36.02% 35.70% 37.42% 33.70%

If we consider equity prices (as previously) we find that the influence of equity prices is higher
than the influence of CDS spreads. On an average the influence is about 5 times higher in the 1st
lag and 22 times higher in the 2nd lag. However, this is not surprising.

We find that equity volatility influences CDS spreads more than equity prices. The estimates for
equity volatility are higher than the estimates for equity prices in all the models. The influence of
equity volatility increases the higher the interval is for the rolling standard deviation except for 1
year. In 1 month the influence is 2 times stronger than the influence from equity prices in the 1st
lag while the influence is 32 times stronger for 6 months in the 1st lag. Moreover the estimates
for equity volatility do not decrease as much as the estimates for equity prices in the 2nd lag.

In section 9.1 we found that for the overall index model the R2 was 34.17%. When including
equity volatility into the model we get an R2 that range from 33.53% to 37.42% so it seems like
the model has gained explanatory power. The influence from equity volatility on CDS spreads is
highest when the rolling standard deviation covers 6 months, but a time period between 2 months
and 6 months seem to estimate good models.

It is interesting that the 2nd lag of equity volatility is negative especially because the sizes of the
estimates are quite big. If we consider the total influence from equity volatility some of the

100
positive strong influence in the 1st lag estimates is off set by the strong negatively influence from
the 2nd lag estimates. If we for example study the model for 6 months were the influence from
equity volatility is strongest the combined influence from equity volatility is about 7.3 (57.519-
50.233) this is, however, still stronger than the influence from equity prices (-1.781-
0.820=2.601)41.

All in all we find that equity volatility influences CDS spreads positively in 1st lag except for 1
year and negatively in the 2nd lag. Thereby the results are not as clear cut as we had hoped for,
but they still indicate that equity volatility might influence positively due to positive 1st lag and
the Spearman Rank correlations. But regardless of the signs of the estimates we conclude that
equity volatility does influence CDS spreads based on the estimated models and the Granger
Causality test.

However, our findings reveal that further research is needed to draw a definite conclusion, and
we must leave this for future research.

10 Discussion

10.1 implications
In this section we want to discus the implications of our results and elements to be taken into
consideration before using our findings.

In this section we assume that there is a negative relationship between CDS spreads and bond
prices based on findings from other researchers. It should be kept in mind that the relationship
between CDS spreads and bond prices are still being analysed. Researchers such as Norden &
Weber (2007) find that there is a negative relationship between CDS spreads and bond prices.
That is, when CDS spreads increases (decrease) bond prices decrease (increase). This is due to
the fact that investors want to be compensated for the higher (lower) risk by a higher (lower)
yield.

10.1.1 Economic models


We have found that there is an inverse relationship between CDS spreads and equity prices. That
is, we found that an increase (decrease) in equity prices causes a decrease (increase) in CDS

41
This is though not theoretically correct when we consider the model as a whole. The value to insert in the model
for 1st lag equity volatility would not be the same value to insert for 2nd lag equity volatility so the example is just a
simplification.

101
spreads. The inverse relationship can be used when making economic models, which take the
development in credit risk into consideration. Based on our research we know that equity prices
influence CDS spreads and therefore it can be included as a variable when estimating credit risk.
As an example could be the Merton model and testing for its validity on empirical data.

10.1.2 New debt issue


Additionally, we also found that equity prices lead CDS spreads. That is, news that influences
the credit risk of a firm is incorporated into the equity price before it is incorporated into the
CDS spreads. The knowledge of the existence of a lead-lag relationship can be used when
issuing new debt.

The higher the credit risk the more investors expect to be compensated, when investing in
corporate bonds, that is, the higher yield they expect to get. Therefore corporate bonds sell at
lower prices and higher yields than government bonds due to the presence of credit risk
(Brealey:2006:250). When a firm wants to issue new debt, it wants to do it as cheap as possible
and therefore they want to issue a bond with as low a yield as possible.

As mentioned in section 4.3 CDS spreads are conceived as credit risk. Due to this fact the
finance department of a firm can with advantage take the development of the equity prices into
account in the decision of when to issue the new debt.

When equity prices increase (decrease), CDS spreads/credit risk will decrease (increase). The
lower (higher) the credit risk, the lower (higher) yield the investors expect in compensation and
the higher (lower) the price of the bond. If the equity price of the company is increasing then the
issuer should wait until the lag effect has occurred, since the return that bondholders demand for
investing in this bond is depending on the credit risk.

On the other hand if the finance department observes a decrease in equity price they should issue
the new debt before the lag effect occurs, in order to get the cheapest funding as possible.

If a firm needs to issue new debt as fast as possible it can be difficult to wait until the lag effect
has occurred. Furthermore, if the equity price is extreme volatile, it is a gamble to decide when
the best time is to issue new debt. But the main point is, when an increase (decrease) in equity
prices causes a decrease (increase) in CDS spreads, it makes it less (more) expensive for a firm
to fund itself in the credit market.

102
ye on equity prices the development in credit risk can be monitored.
Moreover, by keeping an eye
As an example if a bank have Pepcico Inc. in its book, that is own a Pepsico Inc. bond, the
bank’s credit risk depends on the credit risk of Pepsico Inc. among others. The bank can
therefore by keeping an eye on the equity price get a feeling of the development of their credit
risk.

10.1.3 Investment strategy


We started our thesis with defining the
the three forms of efficient markets. We did that in order to
make the reader aware of the consequences by lack of efficiency. If there are differences in
efficiencies between two markets it is possible to exploit this flaw and hereby make money.

As mentioned
ed before, we found that equity prices lead CDS spreads. That is, we found that price
discovery regarding credit risk is being expressed in the equity prices before the CDS spreads. In
the investment world credit investors can exploit this flaw to make investment
investment decision. We
found that there only exists a one-way
one way relationship, which means equity investors cannot really
use this discovery in their investment decisions. An example could be credit investors that hold
protection on CDS for Pepsico Inc. and observe
observe that the equity price of Pepsico Inc. is
increasing. This means that they should sell protection on CDS (that is, go long in the credit),
and when the lag effect has occurred they should buy protection on CDS (that is, go short in the
credit) and through that makee risk free money.

Figure 11: Investment strategy 1

•Credit
Credit investor needs protection
Day 1 Buys CDS at 500 bp
•Buys

•Equity
Equity price increases
. •Credit
Credit investor knows that the CDS spread will react to this increase with a 2 days lag.

•Credit
Credit investor sells his CDS at 500 bp
Day 2

•The
The CDS spread reacts to the increase in equity and decreases
.

• The credit investor buys the CDS again but now to the adjusted price at 300 bp. He earned
Day 3 200 bp in profit

On the other hand, if credit investors observe a decrease of equity prices they should buy
protection on CDS (that is, go short in the credit) and when the lag effect has occurred they

103
should sell the protection on CDS (that is, go long in the credit). Of course the profitability is
depending on the transaction cost.

Figure 12: Investment strategy 2

•The
The equity price decreases
•The
The arbitrage strategist knows that when the equity price decreases the CDS spread
will increase with 2 days lag
Day 1 • He buys a CDS at 500 bp

• 2 days adjustment time - the CDS reacts to the decrease in equity price
.
• The CDS spread has adjusted to the decrease in equity and is increased to 700 bp
• The arbitrage strategist sells his CDS at 700 bp
Day 4 •He
He earned 200 bp in profit

The lead-lag
lag relationship between two markets can only be used to make money if not all
investors are aware of the flaw. As an example if Pepsico’s equity price increases, then credit
investors wantss to be the protection seller of CDS on Pepsico. If all investors want to sell CDS
for Pepsico at the same time, then the supply will exceed demand driving the price down. When
it then is time to buy the CDS for Pepsico the demand will exceed the supply driving the price
up. This will make the transaction cost for trying to exploit the lead-lag
lead lag relationship too high,
and arbitrage will not
ot be possible.

Furthermore, we have not analysed the precise lead-lag


lead lag relationship or the actual strength of the
relationship, which of course have an impact on the above suggestions for applying our findings.
Moreover, our R2 values are in the range from 3.3% to 38.23%, which means that there are other
factors than equity prices that influence the CDS spreads and the investor should also take the
influence of these into account.

10.2 Critic of results


The above use of our findings is depending on the correctness
correctness of our results. As mentioned in
section 5 there have been other studies on the relationship between CDS spreads and equity
prices and the findings are not unambiguous. Our findings are the opposite of Lake & Apergis
(2009) and Hartmann (2008) but we
we join among others Norden & Weber (2004), Forte & Pena

104
(2008) and Fung (2008) in their findings that equity prices lead CDS spreads. Our study has been
far more extensive than theirs, however, our findings can be discussed.

First of all, we only look at the US market, which does not mean that we would get the same
results if making a similar study on the European market. There could be some country specific
issues that could influence the relationship. The possibility of our findings for the US market to
be the same in the European market is present, since both markets are characterized as developed
markets, whereas there could be some issues if trying to apply our findings in developing
markets like Latin America or Asia.

Furthermore, we have analysed a section of S&P 500 (that is we took the S&P 500 from the
beginning of 2002 as the basis for our study sample. From that we narrowed our sample down to
only containing the companies with proper data and a corresponding CDS). That is, we have
analysed the most traded companies in the US market. An argument could be made that when
companies are traded a lot the efficiency must be high, since these companies are being analysed
all the time. This means, that the equity market should be efficient whereas it is more
questionable that the CDS’ for the same companies will be scrutinised at the same level. This
could have an influence on our results. By this we mean, that since companies outside S&P 500
may not get analysed as much so the lead-lag relationship between CDS spreads and equity
prices could be different.

We have only checked the relationship on indices in order to avoid firm specific issues. It could
be that firm specific issues could change the relationship.

Moreover, the specific time period we have chosen to analyse can have an impact on the lead-lag
relationship. The time period we are analysing 2nd January 2004 to 1st May 2009 is before the
CDS market became regulated. This means that applying our result to a later time period the
results may not be the same.

At last we have encountered the same problems as other analysts trying to fit empirical data to
statistical models. It is a generalization and there might be tendencies in our data that we did not
find throughout our testing. All in all we have made an effort to apply the right statistical
methods and include the right data to insure results of a high quality. Moreover we have
robustness tested our data and the results have been consistent throughout this so we conclude
that the results can be trusted for the US market for S&P 500 companies.

105
11 Concluding remarks
The main question this thesis seeks to answer is, what is the relationship between CDS spreads
and equity markets. We find this question interesting, since even though CDS and equity are two
different financial products, they are both influenced by credit risk.

A CDS is a bilateral contract between two parties that agree to isolate and transfer the credit risk
for a reference entity in case of a pre-defined credit event occurs. The higher the probability of
the credit event, the higher the credit risk and the higher the CDS spread. CDS spread represents
the credit risk of a particular firm, and this credit risk is also represented in equities.

Equity is the ownership claim on the earnings and assets of a firm, but it also bears the ultimate
form of credit risk because it represents the most subordinated claim in the capital structure of
the firm. When credit risk increases the value of the firm is also affected and this will influence
the equity prices in a negative way. So intuitively a negative relationship should exist between
the two markets. Furthermore, if a lead-lag relationship exists between the two markets, there is
possibility of arbitrage between the two markets.

Our data is expanding over the period 2nd January 2004 to 1st May 2009 consisting of 265 firms.
The firms in our sample are basically the firms in S&P 500 from beginning of 2002 that have a
corresponding CDS. From these data we made equally weighted indices to analyse our research
question. We used Spearman Rank correlation coefficient, Granger Causality test & Gonzalo-
Granger (GG) measure and VECM/VAR- models to analyse the relationship between CDS
spreads and equity prices.

The correlation between CDS spreads and equity prices is negative

Our analysis showed that there is an inverse relationship between equity prices and CDS spreads.
All 29 models with CDS spread as dependent variable the estimates for equity prices were
negative and the Spearman Rank correlation was negative.

Equity prices influence CDS spreads and CDS spreads do not influence equity prices

For all 29 models with CDS spreads as dependent variable the equity estimates were significant,
the R2s were relatively high and the F-tests were rejected. Furthermore in all models the Granger
Causality test and the GG measure indicated that equity prices influence CDS spreads. 27 of the
29 models with equity as dependent variable had significant CDS estimates and low R2s and the

106
Granger Causality test and GG measure also indicated that CDS spreads do not influence equity
prices. In 2 of the 29 models there were significant estimates for CDS spreads, but the influence
and the R2s were too low to have an impact.

The robustness tests of our results that equity prices influence CDS spreads and not the other
way around further confirmed the validity of our results. Neither the influence of exogenous
variables nor splitting our data in before & after the beginning of the crisis, in quartiles, on rating
and on sectors influence the findings that there is a negative relationship between CDS spreads
and equity prices and that equity prices influences CDS spreads and not the other way around.
However we find that equity prices have a stronger influence on CDS spreads after the beginning
of the crisis than before the beginning of the crisis. Moreover the influence from equity prices on
CDS spreads is more pronounced the lower the credit quality and the higher the CDS spread on
the underlying index. Furthermore we did not find an unambiguous answer to the question if the
strength of the relationship between CDS spread and equity prices becomes more pronounced the
more debt a sector has.

We found that equity prices influence CDS spreads and the relationship is negative which is in
accordance with the Merton model. Therefore we found it interesting to test if the suggested
relationship for equity volatility in the Merton model also was valid for our data. We expanded
our index model by including equity volatility. The Spearman Rank correlation was positive but
the results of the estimated models were ambiguous.

From our analysis we conclude that CDS spreads and equity prices are negative correlated and
that equity prices influences CDS spreads and not the other way around. These findings can be
used in economic models, when issuing new debt and making investment decisions.

12 Suggested further research


During our work with this thesis some interesting questions have come to mind, which we found
were beyond the scope of this paper to answer, but could be relevant to investigate further.

When looking at the same research questions that we are investigating it would be interesting to
examine if the relationship between CDS spreads and equity prices changes if the same study is
made in another market, such as the European market.

Moreover it could be interesting to repeat this study when the regulation of the CDS market is
put into force and see if the relationship between CDS spreads and equity prices will change

107
under the new regulation. Another aspect of the crisis is that due to the collapse of Bear Stearns,
Merrill Lynch and Lehman Brothers the top 5 institutions are now providing 88% of the total
notional amount bought and sold, and is therefore a reflection of the increasing concentration and
counterparty risk within the CDS market (Reuters3:2009). All in all to test if the “new market”
changes our findings.

Furthermore to further examine the lead-lag relationship event studies could be a way to
facilitate this. It could be interesting to examine individual firms and follow how and when
equity prices and CDS spreads react to announcements. When analysing this, it will be possible
to detect the efficiency in each market and the degree of insider trading. Furthermore it will then
be possible to check for the influence that e.g. unexpected cash dividends financed by debt have
on the relationship between CDS spreads and equity prices.

Moreover it could be interesting to analyse the correctness of the pricing of CDS’ before the
beginning of the crisis. As seen in figure 9 we see that the CDS spreads skyrocketed after the
beginning of the crisis, whereas the corresponding decrease in equity prices was moderate. This
raises the question if the CDS’ were correctly priced before the beginning of the crisis as the
relationship between equity prices and CDS spreads was stronger after the beginning of the
crises.

Since the Merton model is the theoretical link between CDS spreads and equity prices it could be
interesting to test the application of this model on real data.

All in all due to the young age and continued development of the CDS market there are still
many unexamined subjects.

108
Litterature

Abid, Fathi & Naifar, Nader “The determinants of credit default swap rates: An explanatory
study”, 2005, University of Sfax

Abbot, Charles “White house dividends CDS regulatory jurisdiction” 2009, Reuters

Bank for International settlement “Guidelines for semi-annual credit default swaps statistics –
June” 2006

BBC News 2003: http://news.bbc.co.uk/2/hi/business/2817995.stm

BNP Paribas 2009: www.bank.bnpparibas.com

Brealey, Meyers, & Allen ”Corporate finance” 2006, 8th International Edition, McGraw-Hill

Brettel, Karen “CDS participants want federal laws to trump states” 2009, Reuters

Brocklebank, John Clare & Dickey, David A. ”SAS for forecasting” 2003, 2nd edition, SAS

Bystrøm, Hans ”Credit default swaps and equity prices: The iTraxx CDS index market” 2004,
Lund University

Chan, Kam C., Fung, Hung-Gay & Zhang, Gaiyan ”On the relationship between Asian
sovereign credit default swap markets and equity market”, 2008, University of Missouri – St.
Louis & Western Kentucky University

Chorafas, Dimitris N. ”Credit derivatives & the management of risk”, 2000, 1st edition, New
York Institute of Finance

109
CBO: www.cboe.com/micro/vix/historical.aspx

Das, Satyajit ”Credit derivatives – CDO’s and structured credit products” 2005, 1st edition,
John Wiley and Sons

Davidson “AIG takes step in winding down money losing unit” 2009, Reuters

Eirikdottir, Asta Björk & Petursdottir, Brynja Kolbrun “The influence of the financial crisis in
2007 on the relationship between the credit derivative market and capital market – The
evolvement of the CDS spreads during the crisis”, 2008, Aarhus Business School & University
of Aarhus

Elton, Edwin J., Gruber, Martin J., Brown, Stephen J. & Goetzmann, William N. ”Modern
portfolio theory and investment analysis” 2007, 7th edition, Wiley

Forte, Santiago & Lovreta, Lidija “Credit risk discovery in the stock and CDS market: Who,
when and why leads?” 2008, Universitat Ramon Llull

Forte, Santiago & Pena, Juan Ignacio “Credit spreads: An empirical analysis on the
informational content of stocks, bonds, and CDS”, 2008, Universidad Carlos III de Madrid &
ESADE Business School

Fung, Hung-Gay, Sierra, Gregory E., Yau, Jot & Zhang, Gaiyan “Are the U.S. stock market and
credit default swap market related? Evidence from the CDX Indices” 2008, University of
Missouri – St. Louis, Federal Reserve Bank of Richmond & Seattle University

Gillian, Carey “US treasury: Derivatives need oversight, not bans” 2009, Reuters

Goldman Sachs “Credit derivatives Intensive”, 2006, ACF Consultants Ltd.

110
Gonzalo, Jesus & Granger, Clive “Estimation of common long-memory components in co-
integrated systems”, 1995, Boston University & University of California

Greatrex, Caitlin A. “The credit default swap market determinants”, 2008, Fordham University

Gujarati, Damodar N. ”Basic Econometrics”, 2003, 4th International edition, McGraw-Hill

Hafer, Shane “The relationship between CDS spreads and equities market and volatility with
respect to credit events for single-name CDS within CDX.NA.IG Index” 2008

Hartmann, Stefan “Global Quantitative – Buying and selling on credit”, 2008, Royal Bank of

Scotland

Hull, John C. ”Risk Management and Financial Institutions”, 2007, 1st International edition,
Pearson Education

ICE : www.ice.com

Ishikawa, Yasushi & Mezrich, Joseph J. “CDS-equity arbitrage strategy”, 2009, Nomura

Jersey, Ira, Makedon, Alex & Lee, David “Credit derivative handbook”, 2007, Credit Suisse

Kikuchi, Masatoshi & Uomoto, Tohihiro “Credit and equity markets”, 2009, Bank of America
Merrill Lynch

LaCour, Lisbeth “Lecture notes – Fall 2007”, 2007, Copenhagen Business School

111
Lake, Andreas & Apergis, Nicolas “Credit default swaps and stock prices: Further evidence
within and across markets from mean and volatility transmission with a MVGARCH-M model
and newer data”, 2009, University of Pireaeus

Longstaff, Francis A., Mithal, Sanjay & Neis, Eric “Corporate yield spreads, default risk or
liquidity? New evidence from the credit default swap market”, 2004, Anderson School of UCLA
and Deutsche Bank

Merton, Robert C. “On the pricing of corporate debt: The risk structure of interest rates” 1974,
American Finance Association

McDonald 2009: http://udel.edu/-mcdonald/statspearman.html

Morrissey, Janet “Credit default swaps: The next crisis?” 2008, Times

Mortensen, Allan ”Essays on the pricing of corporate bonds and credit derivatives”, 2005
Copenhagen Business School

Norden, Lars & Weber, Martin “The co-movement of credit default swap and stock markets: an
empirical analysis”, 2007, University of Mannheim

Pension & Invest 2009:


http://www.pionline.com/apps/pbcs.dll/article?AID=/20090129/DAILYREG/901299987&Assig
nSessionID=273359031797577&AssignSessionID=273359031797577

Philips, Matthew “The monster that ate Wall Street – How credit default swaps – an insurance
against bad loans – turned from a smart bet into a killer” 2008, Newsweek

Realdon, Marc “Credit default swap rates and stock prices”, 2007, University of York

112
Reuters 2009 -1:
http://www.reuters.com/article/americasDealsNews/idUSTRE57A5D620090811

Reuters 2009 -2: http://www.reuters.com/article/GCA-CreditCrisis/idUSTRE56T02P20090730

Reuters 2009 -3: http://www.reuters.com/article/etfNews/idUSN1948326520090820

SAS Institute ”SAS/Ets 9.1 – User’s guide” 2004, 4 volume set, SAS publishing staff

Taylor, Francesca ”Mastering Derivatives Markets”, 2000, 2nd edition, Financial Times

Standard & Poor’: Samson, Solomon B., Sprinzen, Scott, Dubois-Pelerin, Emmanuel & Pfeil,
Kenneth C. “Corporate Ratings Criteria”, 2009, Standard & Poor’s

Shafi, Summarah Shaheen & Ahmad, Sahar Zia ”Syntestiske CDO’er – som instrument til
styring og afdækning af kredit risiko”, 2007, Copenhagen Business School

Sørensen, Bent E. “Granger Causality” 2005

Toda, Hiro & Phillips, Peter C.B. “Fully modified least squares and vector autoregression”,
1993, Econometria

Troelsen, Brian & Lundtoft, Carsten “Credit default swaps – en analyse af forskelle og
anvendelsesmuligheder for tre model tilgange” 2008, Handelshøjskolen i Århus

US Treasury: www.ustreas.gov

Washington Post 2003:


http://pqasb.pqarchiver.com/washingtonpost/access/433656031.html?dids=433656031:43365603
1&FMT=ABS&FMTS=ABS:FT&date=OCT+30%2C+2003&author=Peter+Behr&pub=The+W
ashington+Post&desc=Senate+Democrats+Attempting+to+Close+%27Enron+Loophole%27&p
qatl=google
113
Wooldridge, Jeffrey M. ”Introductory econometrics – a modern approach”, 2006, 4th edition
South-Western

Yahoo Finance: www.finance.yahoo.com

Zhang, Gaiyan “Intra-industry credit contagion – Evidence from the credit default swap market
and the stock market”, 2005, University of California

114
Table of contents - Appendices
Appendix 1 – NAMES IN INDEX & SECTOR DESCRIPTION .................................................................... 3
Appendix 2 - INDEX .............................................................................................................................................. 13
Appendix 3 – EXOGENE VARIABLES ............................................................................................................. 27
Appendix 4 – INDEX QUARTILE 1 .................................................................................................................. 48
Appendix 5 – INDEX QUARTILE 2 .................................................................................................................. 58
Appendix 6 – INDEX QUARTILE 3 .................................................................................................................. 68
Appendix 7 – INDEX QUARTILE 4 .................................................................................................................. 78
Appendix 8 - INDEX 02.01.2004 – 08.08.2007 (BEFORE) ................................................................... 88
Appendix 9 - INDEX 09.08.2007 – 01.05.2009 (AFTER) .................................................................... 102
Appendix 10 – Before QUARTILE 1 ............................................................................................................. 116
Appendix 11 – Before QUARTILE 2 ............................................................................................................. 126
Appendix 12 – Before QUARTILE 3 ............................................................................................................. 136
Appendix 13 – Before QUARTILE 4 ............................................................................................................. 146
Appendix 14 – After QUARTILE 1 ................................................................................................................ 156
Appendix 15 – After Quartile 2 ...................................................................................................................... 166
Appendix 16 – After QUARTILE 3 ................................................................................................................ 176
Appendix 17 – After QUARTILE 4 ................................................................................................................ 186
Appendix 18 – INVESTMENT GRADE ......................................................................................................... 196
Appendix 19 – HIGH YIELD ............................................................................................................................. 209
Appendix 20 – Investment grade QUARTILE 1 ...................................................................................... 222
Appendix 21 – Investment grade QUARTILE 2 ...................................................................................... 231
Appendix 22 – Investment grade QUARTILE 3 ...................................................................................... 242
Appendix 23 – Investment grade QUARTILE 4 ...................................................................................... 252
Appendix 24 – High yield QUARTILE 1 ...................................................................................................... 262
Appendix 25 – High yield QUARTILE 2 ...................................................................................................... 272
Appendix 26 – High yield QUARTILE 3 ...................................................................................................... 282
Appendix 27 – High yield QUARTILE 4 ...................................................................................................... 292
Appendix 28 - COMMUNICATION ................................................................................................................ 302
Appendix 29 – CONSUMER, CYCLICAL....................................................................................................... 315
Appendix 30 – CONSUMER, NON-CYCLICAL ........................................................................................... 328
Appendix 31 - ENERGY ..................................................................................................................................... 341

1
Appendix 32 - FINANCE.................................................................................................................................... 354
Appendix 33 - INDUSTRIALS .......................................................................................................................... 367
Appendix 34 – MATERIALS ............................................................................................................................. 380
Appendix 35 - TECHNOLOGY ......................................................................................................................... 393
Appendix 36 - UTILITIES ................................................................................................................................. 406
Appendix 37 – 1 WEEK ..................................................................................................................................... 419
APPENDIX 38 - 1 MONTH ................................................................................................................................ 425
APPENDIX 39 - 2 MONTHS .............................................................................................................................. 435
APPENDIX 40 - 3 MONTHS .............................................................................................................................. 445
APPENDIX 41 - 6 MONTHS .............................................................................................................................. 455
APPENDIX 41 - 1 YEAR ..................................................................................................................................... 465
Appendix 43 – OVERVIEW OF RESULTS ................................................................................................... 474
Appendix 44 – QUARTILE RANKING .......................................................................................................... 475

2
Appendix 1 – NAMES IN INDEX & SECTOR DESCRIPTION
NAMES SECTOR RATING
3M CO Industrial Investment grade
Consumer, Non-
ABBOTT LABORATORIES cyclical Investment grade
ACE LTD Financial Investment grade
ADVANCED MICRO DEVICES Technology High yield
Consumer, Non-
AETNA INC cyclical Investment grade
AIR PRODUCTS & CHEMICALS INC Basic Materials Investment grade
Consumer, Non-
ALBERTSON'S LLC cyclical High yield
ALCAN INC Basic Materials Investment grade
ALCOA INC Basic Materials Investment grade
Consumer, Non-
ALLERGAN INC cyclical Investment grade
ALLSTATE CORP Financial Investment grade
ALLTEL CORP Communications Investment grade
AMBAC FINANCIAL GROUP INC Financial High yield
AMERICAN ELECTRIC POWER Utilities Investment grade
AMERICAN EXPRESS CO Financial Investment grade
AMERICAN INTERNATIONAL
GROUP Financial Investment grade
Consumer, Non-
AMERISOURCEBERGEN CORP cyclical Investment grade
Consumer, Non-
AMGEN INC cyclical Investment grade
AMR CORP Consumer, Cyclical High yield
ANADARKO PETROLEUM CORP Energy Investment grade
Consumer, Non-
ANHEUSER-BUSCH COS INC. cyclical Investment grade
AOL TIME WARNER Communications Investment grade
AON CORP Financial Investment grade
APACHE CORP Energy Investment grade
APPLIED MATERIALS INC Technology Investment grade
Consumer, Non-
ARCHER-DANIELS-MIDLAND CO cyclical Investment grade
ASHLAND INC Basic Materials High yield
AT&T INC Communications Investment grade
AUTOZONE INC Consumer, Cyclical Investment grade
Consumer, Non-
AVERY DENNISON CORP cyclical Investment grade
Consumer, Non-
AVON PRODUCTS INC cyclical Investment grade
BAKER HUGHES INC Energy Investment grade
BALL CORP Industrial High yield
BANK OF AMERICA CORP Financial Investment grade
BARRICK GOLD CORP Basic Materials Investment grade

3
Consumer, Non-
BAXTER INTERNATIONAL INC cyclical Investment grade
Consumer, Non-
BECTON DICKINSON AND CO cyclical Investment grade
BELLSOUTH CORP Communications Investment grade
BEST BUY CO INC Consumer, Cyclical Investment grade
BLACK & DECKER CORP Industrial Investment grade
BOEING CO Industrial Investment grade
Consumer, Non-
BOSTON SCIENTIFIC CORP cyclical High yield
Consumer, Non-
BRISTOL-MYERS SQUIBB CO cyclical Investment grade
BRUNSWICK CORP Consumer, Cyclical High yield
BURLINGTON NORTHERN SANTA
FE Industrial Investment grade
Consumer, Non-
CAMPBELL SOUP CO cyclical Investment grade
CAPITAL ONE FINANCIAL CORP Financial Investment grade
Consumer, Non-
CARDINAL HEALTH INC cyclical Investment grade
CARNIVAL CORP Consumer, Cyclical Investment grade
CATERPILLAR INC Industrial Investment grade
CENTEX CORP Consumer, Cyclical High yield
CENTURYTEL INC Communications High yield
CHESAPEAKE ENERGY CORP Energy High yield
CHEVRON CORP Energy Investment grade
CHUBB CORP Financial Investment grade
Consumer, Non-
CIGNA CORP cyclical Investment grade
CINERGY CORP Utilities Investment grade
CISCO SYSTEMS INC Communications Investment grade
CIT GROUP INC Financial High yield
CITIGROUP INC Financial Investment grade
CMS ENERGY CORP Utilities Investment grade
Consumer, Non-
COCA-COLA CO/THE cyclical Investment grade
Consumer, Non-
COCA-COLA ENTERPRISES cyclical Investment grade
Consumer, Non-
COLGATE-PALMOLIVE CO cyclical Investment grade
COMCAST CORP-CL A Communications Investment grade
COMPUTER SCIENCES CORP Technology Investment grade
Consumer, Non-
CONAGRA FOODS INC cyclical Investment grade
CONOCOPHILLIPS CO Energy Investment grade
CONSOLIDATED EDISON INC Utilities Investment grade
CONSTELLATION ENERGY GROUP Utilities Investment grade
Consumer, Non-
CONVERGYS CORP cyclical High yield
COOPER INDUSTRIES LTD-CL A Industrial Investment grade
COOPER TIRE & RUBBER Consumer, Cyclical High yield
CORNING INC Communications Investment grade

4
COSTCO WHOLESALE CORP Consumer, Cyclical Investment grade
CRANE CO Industrial Investment grade
CSX CORP Industrial Investment grade
CUMMINS INC Industrial Investment grade
CVS CAREMARK CORP Consumer, Cyclical Investment grade
DANAHER CORP Industrial Investment grade
DARDEN RESTAURANTS INC Consumer, Cyclical Investment grade
DEERE & CO Industrial Investment grade
DELL INC Technology Investment grade
Consumer, Non-
DELUXE CORP cyclical High yield
DEVON ENERGY CORPORATION Energy Investment grade
DOMINION RESOURCES INC/VA Utilities Investment grade
DOVER CORP Industrial Investment grade
DOW CHEMICAL Basic Materials Investment grade
DTE ENERGY COMPANY Utilities Investment grade
DU PONT (E.I.) DE NEMOURS Basic Materials Investment grade
DUKE ENERGY CORP Utilities Investment grade
DYNEGY INC-CL A Utilities Not rated
EASTMAN CHEMICAL COMPANY Basic Materials Investment grade
EASTMAN KODAK CO Industrial High yield
EATON CORP Industrial Investment grade
EL PASO CORP Energy High yield
ELECTRONIC DATA SYSTEMS CORP Technology Investment grade
Consumer, Non-
ELI LILLY & CO cyclical Investment grade
EMC CORP/MASS Technology Investment grade
EMERSON ELECTRIC CO Industrial Investment grade
Consumer, Non-
EQUIFAX INC cyclical Investment grade
EXELON CORP Utilities Investment grade
FEDEX CORP Industrial Investment grade
FIRST DATA CORP Technology High yield
FIRSTENERGY CORP Utilities Investment grade
FLUOR CORP Industrial Investment grade
FORD MOTOR CO Consumer, Cyclical High yield
Consumer, Non-
FORTUNE BRANDS INC cyclical Investment grade
FREEPORT-MCMORAN COPPER Basic Materials High yield
GAP INC/THE Consumer, Cyclical High yield
GENERAL DYNAMICS CORP Industrial Investment grade
GENERAL ELECTRIC CO Industrial Investment grade
Consumer, Non-
GENERAL MILLS INC cyclical Investment grade
Consumer, Non-
GENZYME CORP cyclical Investment grade
GEORGIA-PACIFIC LLC Basic Materials High yield
GOODRICH CORP Industrial Investment grade
GOODYEAR TIRE & RUBBER CO Consumer, Cyclical High yield
Consumer, Non-
H&R BLOCK INC cyclical Investment grade

5
HALLIBURTON CO Energy Investment grade
HARRAH'S ENTERTAINMENT INC Consumer, Cyclical High yield
HARTFORD FINANCIAL SVCS GRP Financial Investment grade
HASBRO INC Consumer, Cyclical Investment grade
Consumer, Non-
HCA INC cyclical High yield
Consumer, Non-
HEALTH MGMT ASSOCIATES INC-A cyclical High yield
Consumer, Non-
HERSHEY CO/THE cyclical Investment grade
HESS CORP Energy Investment grade
HEWLETT-PACKARD CO Technology Investment grade
HILTON HOTELS CORP Consumer, Cyclical Not rated
Consumer, Non-
HJ HEINZ CO cyclical Investment grade
HOME DEPOT INC Consumer, Cyclical Investment grade
Consumer, Non-
HUMANA INC cyclical Investment grade
INTEL CORP Technology Investment grade
INTERNATIONAL PAPER CO Basic Materials Investment grade
INTERPUBLIC GROUP OF COS INC Communications High yield
INTL BUSINESS MACHINES CORP Technology Investment grade
INTL GAME TECHNOLOGY Consumer, Cyclical Investment grade
INTUIT INC Technology Investment grade
ITT CORP Industrial Investment grade
J.C. PENNEY CO INC Consumer, Cyclical High yield
JABIL CIRCUIT INC Industrial High yield
Consumer, Non-
JOHNSON & JOHNSON cyclical Investment grade
JOHNSON CONTROLS INC Consumer, Cyclical Investment grade
JONES APPAREL GROUP INC Consumer, Cyclical High yield
JPMORGAN CHASE & CO Financial Investment grade
KB HOME Consumer, Cyclical High yield
Consumer, Non-
KELLOGG CO cyclical Investment grade
KERR-MCGEE CORP Energy Investment grade
Consumer, Non-
KIMBERLY-CLARK CORP cyclical Investment grade
KOHLS CORP Consumer, Cyclical Investment grade
Consumer, Non-
KROGER CO cyclical Investment grade
LEGGETT & PLATT INC Industrial Investment grade
LEXMARK INTERNATIONAL INC-A Technology Investment grade
LIMITED BRANDS INC Consumer, Cyclical High yield
LINCOLN NATIONAL CORP Financial Investment grade
LIZ CLAIBORNE INC Consumer, Cyclical High yield
LOCKHEED MARTIN CORP Industrial Investment grade
LOEWS CORP Financial Investment grade
LOUISIANA-PACIFIC CORP Industrial High yield
LOWE'S COS INC Consumer, Cyclical Investment grade
MACY'S INC Consumer, Cyclical High yield

6
MARATHON OIL CORP Energy Investment grade
MARRIOTT INTERNATIONAL-CL A Consumer, Cyclical Investment grade
MARSH & MCLENNAN COS Financial Investment grade
MASCO CORP Industrial High yield
MATTEL INC Consumer, Cyclical Investment grade
MBIA INC Financial High yield
MCDONALD'S CORP Consumer, Cyclical Investment grade
Consumer, Non-
MCKESSON CORP cyclical Investment grade
MEADWESTVACO CORP Basic Materials Investment grade
Consumer, Non-
MEDTRONIC INC cyclical Investment grade
Consumer, Non-
MERCK & CO. INC. cyclical Investment grade
MERRILL LYNCH & CO INC Financial Investment grade
METLIFE INC Financial Investment grade
MGIC INVESTMENT CORP Financial High yield
MIRANT CORP Utilities High yield
MORGAN STANLEY Financial Investment grade
MOTOROLA INC Communications Investment grade
NATIONAL CITY CORP Financial Investment grade
NAVISTAR INTERNATIONAL Consumer, Cyclical High yield
NEW YORK TIMES CO -CL A Communications High yield
NEWELL RUBBERMAID INC Consumer, Cyclical Investment grade
NEWMONT MINING CORP Basic Materials Investment grade
NIKE INC -CL B Consumer, Cyclical Investment grade
NISOURCE INC Utilities Investment grade
NOBLE ENERGY INC Energy Investment grade
NORDSTROM INC Consumer, Cyclical Investment grade
NORFOLK SOUTHERN CORP Industrial Investment grade
NORTHROP GRUMMAN CORP Industrial Investment grade
NUCOR CORP Basic Materials Investment grade
OCCIDENTAL PETROLEUM CORP Energy Investment grade
OFFICE DEPOT INC Consumer, Cyclical High yield
OMNICOM GROUP Communications Investment grade
PACTIV CORPORATION Industrial Investment grade
Consumer, Non-
PEPSI BOTTLING GROUP INC cyclical Investment grade
Consumer, Non-
PEPSICO INC cyclical Investment grade
Consumer, Non-
PFIZER INC cyclical Investment grade
Consumer, Non-
PHILIP MORRIS INTERNATIONAL cyclical Investment grade
PITNEY BOWES INC Technology Investment grade
PPG INDUSTRIES INC Basic Materials Investment grade
PRAXAIR INC Basic Materials Investment grade
Consumer, Non-
PROCTER & GAMBLE CO/THE cyclical Investment grade
PROGRESS ENERGY INC Utilities Investment grade
PULTE HOMES INC Consumer, Cyclical High yield

7
QWEST COMMUNICATIONS INTL Communications High yield
RADIOSHACK CORP Consumer, Cyclical High yield
RAYTHEON COMPANY Industrial Investment grade
ROCKWELL COLLINS INC. Industrial Investment grade
ROHM AND HAAS CO Basic Materials Investment grade
Consumer, Non-
RR DONNELLEY & SONS CO cyclical Investment grade
RYDER SYSTEM INC Industrial Investment grade
SABRE HOLDINGS CORP-CL A Consumer, Cyclical High yield
Consumer, Non-
SAFEWAY INC cyclical Investment grade
Consumer, Non-
SARA LEE CORP cyclical Investment grade
Consumer, Non-
SCHERING-PLOUGH CORP cyclical Investment grade
SEALED AIR CORP Industrial High yield
SEMPRA ENERGY Utilities Investment grade
SHERWIN-WILLIAMS CO/THE Basic Materials Investment grade
SLM CORP Financial Investment grade
SOUTHERN CO Utilities Investment grade
SOUTHWEST AIRLINES CO Consumer, Cyclical Investment grade
SPRINT NEXTEL CORP Communications High yield
STANLEY WORKS/THE Industrial Investment grade
STAPLES INC Consumer, Cyclical Investment grade
STARWOOD HOTELS & RESORTS Consumer, Cyclical High yield
SUN MICROSYSTEMS INC Technology High yield
SUNOCO INC Energy Investment grade
Consumer, Non-
SUPERVALU INC cyclical High yield
Consumer, Non-
SYSCO CORP cyclical Investment grade
TARGET CORP Consumer, Cyclical Investment grade
TECO ENERGY INC Utilities Investment grade
TEMPLE-INLAND INC Industrial High yield
Consumer, Non-
TENET HEALTHCARE CORP cyclical High yield
TEXTRON INC Industrial Investment grade
THOMAS & BETTS CORP Industrial Investment grade
TORCHMARK CORP Financial Investment grade
TOYS "R" US INC Consumer, Cyclical High yield
TRANSOCEAN LTD Energy Not rated
TRW AUTOMOTIVE HOLDINGS
CORP Consumer, Cyclical Not rated
Consumer, Non-
UNILEVER N V -NY SHARES cyclical Investment grade
UNION PACIFIC CORP Industrial Investment grade
UNISYS CORP Technology High yield
UNITED STATES STEEL CORP Basic Materials High yield
UNITED TECHNOLOGIES CORP Industrial Investment grade
Consumer, Non-
UNITEDHEALTH GROUP INC cyclical Investment grade
UNUM GROUP Financial Investment grade

8
Consumer, Non-
UST INC cyclical Investment grade
VERIZON COMMUNICATIONS INC Communications Investment grade
VF CORP Consumer, Cyclical Investment grade
VIACOM INC-CLASS B Communications Investment grade
WACHOVIA CORP Financial Investment grade
WAL-MART STORES INC Consumer, Cyclical Investment grade
WALT DISNEY CO/THE Communications Investment grade
WASTE MANAGEMENT INC Industrial Investment grade
Consumer, Non-
WATSON PHARMACEUTICALS INC cyclical High yield
Consumer, Non-
WELLPOINT INC cyclical Investment grade
WELLS FARGO & CO Financial Investment grade
WEYERHAEUSER CO Basic Materials Investment grade
WHIRLPOOL CORP Consumer, Cyclical Investment grade
WILLIAMS COS INC Energy Investment grade
WORTHINGTON INDUSTRIES Industrial Investment grade
Consumer, Non-
WYETH cyclical Investment grade
XCEL ENERGY INC Utilities Investment grade
XEROX CORP Technology Investment grade
XL CAPITAL LTD -CLASS A Financial Investment grade
YUM! BRANDS INC Consumer, Cyclical Investment grade

SECTOR DESCRIPTION

Consumer Cyclical
Advertising
Apparel Retail
Apparel, Accessories & Luxury Goods
Auto Parts & Equipment
Automobile Manufacturers
Automotive Retail
Broadcasting
Cable & Satellite
Casinos & Gaming
Computer & Electronics Retail
Consumer Electronics
Department Stores
Distributors
Education Services
Footwear
General Merchandise Stores
Home Furnishings
Home Improvement Retail
Homebuilding
Homefurnishing Retail
Hotels, Resorts & Cruise Lines
Household Appliances

9
Housewares & Specialties
Internet Retail
Leisure Products
Motorcycle Manufacturers
Movies & Entertainment
Photographic Products
Publishing & Printing
Restaurants
Specialized Consumer Services
Specialty Stores
Tires & Rubber

Consumer Non-Cyclical
Agricultural Products
Brewers
Distillers & Vintners
Drug Retail
Food Distributors
Food Retail
Household Products
HyperMarkets & Super Centers
Packaged Foods & Meats
Personal Products
Soft Drinks
Tobacco
Energy
Coal & Consumable Fuels
Integrated Oil & Gas
Oil & Gas Drilling
Oil & Gas Equipment & Services
Oil & Gas Exploration & Production
Oil & Gas Refining & Marketing
Oil & Gas Storage & Transportation

Financials
Asset Management & Custody Banks
Consumer Finance
Diversified REITs
Diversified Banks
Industrial REITs
Insurance Brokers
Investment Banking & Brokerage
Life & Health Insurance
Multi-line Insurance
Multi-Sector Holdings
Office REITs
Other Diversified Financial Services
Property & Casualty Insurance
Real Estate Services
Regional Banks
Residential REITs

10
Retail REITs
Specialized Finance
Specialized REITs
Thrifts & Mortgage Finance

Health Care
Biotechnology
Health Care Distributors
Health Care Equipment
Health Care Facilities
Health Care Services
Health Care Supplies
Health Care Technology
Life Sciences Tools & Services
Managed Health Care
Pharmaceuticals

Industrials
Aerospace & Defense
Air Freight & Logistics
Airlines
Building Products
Commercial Printing
Construction & Engineering
Construction & Farm Machinery & Heavy Trucks
Diversified Support Services
Electrical Components & Equipment
Environmental & Facilities Services
Human Resource & Employment Services
Industrial Conglomerates
Industrial Machinery
Office Services & Supplies
Railroads
Research & Consulting Services
Trading Companies & Distributors
Trucking

Technology
Application Software
Communications Equipment
Computer Hardware
Computer Storage & Peripherals
Data Processing & Outsourced Services
Electronic Components
Electronic Equipment & Instruments
Electronic Manufacturing Services
Home Entertainment Software
Internet Software & Services
IT Consulting & Other Services
Office Electronics
Semiconductor Equipment

11
Semiconductors
Systems Software

Materials
Aluminum
Construction Materials
Diversified Chemicals
Diversified Metals & Mining
Fertilizers & Agricultural Chemicals
Forest Products
Gold
Industrial Gases
Metal & Glass Containers
Paper Packaging
Paper Products
Specialty Chemicals
Steel

Communication
Integrated Telecommunication Services
Wireless Telecommunication Services
Utilities
Electric Utilities
Gas Utilities
Independent Power Producers & Energy Traders
Multi-Utilities

12
Appendix 2 - INDEX
INDEX (CDS)

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 11312.695 1.00000 | |********************| 0

1 11291.190 0.99810 | .|********************| 0.027227

2 11263.927 0.99569 | . |********************| 0.047098

3 11232.201 0.99288 | . |********************| 0.060729

4 11197.517 0.98982 | . |********************| 0.071761

5 11161.541 0.98664 | . |********************| 0.081254

6 11123.203 0.98325 | . |********************| 0.089696

7 11082.896 0.97969 | . |********************| 0.097359

8 11041.691 0.97604 | . |********************| 0.104411

9 11000.505 0.97240 | . |******************* | 0.110969

10 10959.327 0.96876 | . |******************* | 0.117115

11 10917.523 0.96507 | . |******************* | 0.122912

12 10873.723 0.96120 | . |******************* | 0.128406

13 10828.449 0.95719 | . |******************* | 0.133634

14 10781.828 0.95307 | . |******************* | 0.138623

15 10733.913 0.94884 | . |******************* | 0.143398

16 10683.389 0.94437 | . |******************* | 0.147979

17 10630.790 0.93972 | . |******************* | 0.152381

18 10577.156 0.93498 | . |******************* | 0.156618

19 10522.507 0.93015 | . |******************* | 0.160702

20 10467.283 0.92527 | . |******************* | 0.164645

21 10412.153 0.92040 | . |****************** | 0.168455

22 10358.705 0.91567 | . |****************** | 0.172143

13
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

23 10306.599 0.91106 | . |****************** | 0.175716

24 10255.183 0.90652 | . |****************** | 0.179184

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 1.7268 0.9800 2.17 0.9932

1 1.0988 0.9275 0.82 0.8884

2 1.1379 0.9327 0.86 0.8953

Single Mean 0 1.0650 0.9900 0.94 0.9961 2.69 0.3796

1 -0.1688 0.9485 -0.09 0.9488 0.78 0.8709

2 -0.1121 0.9517 -0.06 0.9518 0.82 0.8618

Trend 0 -1.1192 0.9868 -0.73 0.9699 2.66 0.6421

1 -3.2723 0.9251 -1.27 0.8931 1.60 0.8574

2 -3.1436 0.9309 -1.24 0.9002 1.57 0.8628

14
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 10.632948 1.00000 | |********************| 0

1 5.051047 0.47504 | .|********** | 0.027237

2 2.253406 0.21193 | .|**** | 0.032812

3 1.483638 0.13953 | .|*** | 0.033812

4 1.085903 0.10213 | .|** | 0.034237

5 1.405905 0.13222 | .|*** | 0.034462

6 0.702715 0.06609 | .|* | 0.034837

7 0.266877 0.02510 | .|* | 0.034929

8 0.328332 0.03088 | .|* | 0.034943

9 0.925592 0.08705 | .|** | 0.034963

10 1.031483 0.09701 | .|** | 0.035123

11 0.552260 0.05194 | .|* | 0.035322

12 0.286075 0.02690 | .|* | 0.035378

13 -0.208355 -.01960 | .|. | 0.035393

14 -0.264833 -.02491 | .|. | 0.035401

15 0.668175 0.06284 | .|* | 0.035414

16 0.880643 0.08282 | .|** | 0.035497

17 1.012160 0.09519 | .|** | 0.035640

18 0.293166 0.02757 | .|* | 0.035828

19 -0.330771 -.03111 | *|. | 0.035844

20 -0.397939 -.03743 | *|. | 0.035864

21 -0.757544 -.07124 | *|. | 0.035893

22 -0.476802 -.04484 | *|. | 0.035998

23 -0.068611 -.00645 | .|. | 0.036039

24 0.400809 0.03770 | .|* | 0.036040

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

15
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -703.857 0.0001 -21.80 <.0001

1 -723.767 0.0001 -18.92 <.0001

2 -619.290 0.0001 -15.72 <.0001

Single Mean 0 -706.227 0.0001 -21.84 <.0001 238.52 0.0010

1 -728.241 0.0001 -18.97 <.0001 179.95 0.0010

2 -625.006 0.0001 -15.76 <.0001 124.25 0.0010

Trend 0 -708.725 0.0001 -21.88 <.0001 239.41 0.0010

1 -733.036 0.0001 -19.02 <.0001 180.83 0.0010

2 -631.283 0.0001 -15.81 <.0001 124.98 0.0010

16
AUTOCORRELATION

Graph 2.1– Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0009

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 1347.4987 <.0001

AR(2) 1347.7435 <.0001

AR(3) 1347.7444 <.0001

AR(4) 1347.7469 <.0001

17
Graph 2.2 – 1. difference

Table 2.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 1.0485

Table 2.4– 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 304.6171 <.0001

AR(2) 304.8681 <.0001

AR(3) 308.5867 <.0001

AR(4) 309.0580 <.0001

18
INDEX (EQUITY)

STATIONARITY

Table 3.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.999530 1.00000 | |********************| 0

1 0.993348 0.99382 | .|********************| 0.027227

2 0.988667 0.98913 | . |********************| 0.046964

3 0.985205 0.98567 | . |********************| 0.060466

4 0.981183 0.98164 | . |********************| 0.071390

5 0.977553 0.97801 | . |********************| 0.080778

6 0.974304 0.97476 | . |******************* | 0.089125

7 0.971509 0.97197 | . |******************* | 0.096705

8 0.968950 0.96941 | . |******************* | 0.103695

9 0.966057 0.96651 | . |******************* | 0.110208

10 0.963221 0.96367 | . |******************* | 0.116322

11 0.960318 0.96077 | . |******************* | 0.122097

12 0.957141 0.95759 | . |******************* | 0.127578

13 0.953725 0.95417 | . |******************* | 0.132799

14 0.950369 0.95082 | . |******************* | 0.137788

15 0.947251 0.94770 | . |******************* | 0.142568

16 0.944667 0.94511 | . |******************* | 0.147164

17 0.941419 0.94186 | . |******************* | 0.151597

18 0.937762 0.93820 | . |******************* | 0.155874

19 0.934771 0.93521 | . |******************* | 0.160006

20 0.931888 0.93233 | . |******************* | 0.164008

21 0.928717 0.92915 | . |******************* | 0.167890

22 0.926275 0.92671 | . |******************* | 0.171660

23 0.923543 0.92398 | . |****************** | 0.175329

24 0.920837 0.92127 | . |****************** | 0.178903

19
Table 3.2 – Original Data
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -7.7794 0.0538 -2.16 0.0301

1 -6.0439 0.0907 -1.94 0.0506

2 -4.8352 0.1309 -1.79 0.0704

Single Mean 0 -7.7799 0.2291 -2.15 0.2230 2.36 0.4653

1 -6.0450 0.3439 -1.94 0.3152 1.92 0.5780

2 -4.8369 0.4513 -1.79 0.3866 1.66 0.6447

Trend 0 -9.3308 0.4835 -1.84 0.6843 2.42 0.6923

1 -6.0098 0.7437 -1.37 0.8693 1.87 0.8021

2 -3.6312 0.9072 -0.95 0.9482 1.70 0.8378

20
Table 3.1 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.244547 1.00000 | |********************| 0

1 -0.024803 -.10142 | **|. | 0.027237

2 -0.017205 -.07035 | *|. | 0.027515

3 0.017225 0.07044 | .|* | 0.027649

4 -0.0023169 -.00947 | .|. | 0.027781

5 -0.0020052 -.00820 | .|. | 0.027784

6 -0.0030006 -.01227 | .|. | 0.027786

7 -0.0039573 -.01618 | .|. | 0.027790

8 0.015257 0.06239 | .|* | 0.027797

9 0.00046374 0.00190 | .|. | 0.027900

10 0.0041198 0.01685 | .|. | 0.027900

11 0.0083180 0.03401 | .|* | 0.027908

12 0.0027066 0.01107 | .|. | 0.027939

13 0.0072001 0.02944 | .|* | 0.027942

14 -0.0030945 -.01265 | .|. | 0.027965

15 -0.021559 -.08816 | **|. | 0.027969

16 0.017914 0.07326 | .|* | 0.028175

17 0.013730 0.05615 | .|* | 0.028316

18 -0.015347 -.06276 | *|. | 0.028398

19 0.0021885 0.00895 | .|. | 0.028501

20 0.0042333 0.01731 | .|. | 0.028503

21 -0.012173 -.04978 | *|. | 0.028511

22 0.0090403 0.03697 | .|* | 0.028575

23 0.00097680 0.00399 | .|. | 0.028610

24 -0.0043117 -.01763 | .|. | 0.028611

21
Table 3.4 – 1. Difference
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1483.51 0.0001 -40.62 <.0001

1 -1745.23 0.0001 -29.52 <.0001

2 -1460.82 0.0001 -21.72 <.0001

Single Mean 0 -1483.63 0.0001 -40.61 <.0001 824.46 0.0010

1 -1745.77 0.0001 -29.51 <.0001 435.50 0.0010

2 -1461.83 0.0001 -21.72 <.0001 235.87 0.0010

Trend 0 -1487.33 0.0001 -40.70 <.0001 828.31 0.0010

1 -1761.77 0.0001 -29.63 <.0001 439.07 0.0010

2 -1490.36 0.0001 -21.84 <.0001 238.54 0.0010

22
AUTOCORRELATION

Graph 4.1 – Original data

Table 4.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0052

Table 4.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 1340.9748 <.0001

AR(2) 1341.0454 <.0001

AR(3) 1341.0868 <.0001

AR(4) 1341.1067 <.0001

23
Graph 4.2 – 1. difference

Table 4.4 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 2.2024

Table 4.5 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 13.8680 0.0002

AR(2) 22.7271 <.0001

AR(3) 26.7906 <.0001

AR(4) 26.7950 <.0001

24
MODEL

CO-INTEGRATION

Table 5.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0042 6.0197 15.34
1 1 0.0003 0.3718 3.84

VAR-MODEL

Table 5.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 - 0.0952 0.07248 -1.31 0.1892 1
AR1_1_1 0.37478 0.02755 13.60 0.0001 CDS (t-1)
AR1_1_2 - 2.25953 0.14986 -15.08 0.0001 EQ (t-1)
AR2_1_1 0.04223 0.02570 1.64 0.1006 CDS (t-2)
AR2_1_2 - 0.83542 0.15966 -5.23 0.0001 EQ (t-2)
EQ 1 diff Constant2 0.00542 0.01343 0.40 0.6863 1
AR1_2_1 0.00425 0.00510 0.83 0.4048 CDS (t-1)
AR1_2_2 - 0.10474 0.02776 -3.77 0.0002 EQ (t-1)
AR2_2_1 - 0.00385 0.00476 -0.81 0.4183 CDS (t-2)
AR2_2_2 - 0.07330 0.02958 -2.48 0.0133 EQ (t-2)

GRANGER CAUSALITY TEST

Table 5.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 235.37 < 0.0001
2 2 0.91 0.6347

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

25
UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 5.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.3417 2.65234 174.00 < 0.0001
EQ 1diff 0.0175 0.49137 5.99 < 0.0001

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 5.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 2.01470 9999.99 < 0.0001 155.82 < 0.0001
EQ 1diff 1.99136 1447.86 < 0.0001 30.72 < 0.0001

26
Appendix 3 – EXOGENE VARIABLES
CBOE VOLATILITY INDEX (VIX)

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 147.650 1.00000 | |********************| 0

1 145.679 0.98665 | .|********************| 0.027227

2 144.245 0.97694 | . |********************| 0.046739

3 143.373 0.97103 | . |******************* | 0.059996

4 142.244 0.96338 | . |******************* | 0.070693

5 141.479 0.95820 | . |******************* | 0.079834

6 140.509 0.95164 | . |******************* | 0.087947

7 139.621 0.94562 | . |******************* | 0.095275

8 139.031 0.94163 | . |******************* | 0.101995

9 138.245 0.93630 | . |******************* | 0.108248

10 137.525 0.93142 | . |******************* | 0.114093

11 136.590 0.92509 | . |******************* | 0.119597

12 135.461 0.91745 | . |****************** | 0.124789

13 134.655 0.91199 | . |****************** | 0.129693

14 133.744 0.90581 | . |****************** | 0.134363

15 132.818 0.89955 | . |****************** | 0.138816

16 131.852 0.89300 | . |****************** | 0.143071

17 130.894 0.88652 | . |****************** | 0.147145

18 129.760 0.87883 | . |****************** | 0.151053

19 128.585 0.87088 | . |***************** | 0.154797

20 127.279 0.86203 | . |***************** | 0.158387

21 125.965 0.85313 | . |***************** | 0.161827

22 124.831 0.84545 | . |***************** | 0.165128

27
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

23 123.844 0.83877 | . |***************** | 0.168306

24 122.959 0.83277 | . |***************** | 0.171377

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -3.8518 0.1776 -1.26 0.1928

1 -2.6266 0.2659 -0.99 0.2876

2 -1.5666 0.3859 -0.70 0.4130

Single Mean 0 -16.3979 0.0281 -2.80 0.0589 3.95 0.0893

1 -12.1178 0.0808 -2.38 0.1480 2.87 0.3335

2 -8.3386 0.2006 -1.94 0.3130 1.95 0.5711

Trend 0 -29.9809 0.0080 -3.94 0.0109 7.83 0.0106

1 -22.8432 0.0383 -3.43 0.0481 5.95 0.0663

2 -16.3995 0.1415 -2.91 0.1601 4.33 0.3056

28
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 3.766279 1.00000 | |********************| 0

1 -0.552903 -.14680 | ***|. | 0.027237

2 -0.558776 -.14836 | ***|. | 0.027818

3 0.234799 0.06234 | .|* | 0.028398

4 -0.369187 -.09802 | **|. | 0.028500

5 0.222950 0.05920 | .|* | 0.028749

6 -0.084789 -.02251 | .|. | 0.028839

7 -0.308664 -.08195 | **|. | 0.028852

8 0.206039 0.05471 | .|* | 0.029024

9 -0.090688 -.02408 | .|. | 0.029101

10 0.273132 0.07252 | .|* | 0.029116

11 0.173172 0.04598 | .|* | 0.029249

12 -0.326779 -.08676 | **|. | 0.029303

13 0.086211 0.02289 | .|. | 0.029493

14 0.027498 0.00730 | .|. | 0.029506

15 0.014509 0.00385 | .|. | 0.029507

16 -0.027127 -.00720 | .|. | 0.029508

17 0.171170 0.04545 | .|* | 0.029509

18 0.056488 0.01500 | .|. | 0.029561

19 0.103978 0.02761 | .|* | 0.029566

20 0.0046781 0.00124 | .|. | 0.029586

21 -0.200699 -.05329 | *|. | 0.029586

22 -0.163292 -.04336 | *|. | 0.029657

23 -0.049875 -.01324 | .|. | 0.029704

24 -0.021121 -.00561 | .|. | 0.029708

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

29
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1544.73 0.0001 -42.53 <.0001

1 -2192.48 0.0001 -33.09 <.0001

2 -2100.86 0.0001 -24.26 <.0001

Single Mean 0 -1544.80 0.0001 -42.52 <.0001 903.84 0.0010

1 -2192.94 0.0001 -33.08 <.0001 547.06 0.0010

2 -2102.14 0.0001 -24.26 <.0001 294.19 0.0010

Trend 0 -1544.94 0.0001 -42.50 <.0001 903.33 0.0010

1 -2193.71 0.0001 -33.07 <.0001 546.82 0.0010

2 -2103.93 0.0001 -24.25 <.0001 294.08 0.0010

AUTOCORRELATION

Graph 2.1 – Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0255

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

30
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1314.7966 <.0001

AR(2) 1315.4766 <.0001

AR(3) 1316.4369 <.0001

AR(4) 1316.4452 <.0001

Graph 2.3 – 1. difference

Table 2.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 2.2932

Table 2.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

31
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 29.0595 <.0001

AR(2) 68.8431 <.0001

AR(3) 69.0227 <.0001

AR(4) 86.6205 <.0001

32
5-YEAR T-BILL RATE

STATIONARITY

Table 3.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.855836 1.00000 | |********************| 0

1 0.852251 0.99581 | .|********************| 0.027227

2 0.848860 0.99185 | . |********************| 0.047026

3 0.845884 0.98837 | . |********************| 0.060580

4 0.842679 0.98463 | . |********************| 0.071542

5 0.839482 0.98089 | . |********************| 0.080967

6 0.836226 0.97709 | . |********************| 0.089343

7 0.832981 0.97330 | . |******************* | 0.096941

8 0.829569 0.96931 | . |******************* | 0.103933

9 0.826298 0.96549 | . |******************* | 0.110431

10 0.822943 0.96157 | . |******************* | 0.116520

11 0.819390 0.95741 | . |******************* | 0.122261

12 0.815820 0.95324 | . |******************* | 0.127698

13 0.811831 0.94858 | . |******************* | 0.132868

14 0.807696 0.94375 | . |******************* | 0.137797

15 0.803976 0.93940 | . |******************* | 0.142508

16 0.800211 0.93500 | . |******************* | 0.147027

17 0.796574 0.93076 | . |******************* | 0.151370

18 0.792680 0.92621 | . |******************* | 0.155555

19 0.788808 0.92168 | . |****************** | 0.159591

20 0.784707 0.91689 | . |****************** | 0.163489

21 0.780496 0.91197 | . |****************** | 0.167257

22 0.776428 0.90722 | . |****************** | 0.170903

23 0.772392 0.90250 | . |****************** | 0.174437

24 0.768272 0.89769 | . |****************** | 0.177865

33
Table 3.2 – Original Data
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.4570 0.5795 -0.69 0.4187

1 -0.4415 0.5830 -0.69 0.4166

2 -0.3829 0.5960 -0.66 0.4303

Single Mean 0 -2.1911 0.7559 -0.79 0.8210 0.45 0.9646

1 -1.8828 0.7926 -0.71 0.8431 0.40 0.9755

2 -1.3196 0.8550 -0.54 0.8799 0.30 0.9900

Trend 0 -4.0811 0.8818 -1.38 0.8659 1.99 0.7788

1 -3.7214 0.9024 -1.31 0.8841 1.98 0.7810

2 -3.1578 0.9303 -1.23 0.9039 2.25 0.7262

34
Table 3.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.0048568 1.00000 | |********************| 0

1 -0.0002039 -.04198 | *|. | 0.027237

2 -0.0004528 -.09323 | **|. | 0.027285

3 0.00010471 0.02156 | .|. | 0.027520

4 -0.0000989 -.02036 | .|. | 0.027533

5 0.00011284 0.02323 | .|. | 0.027544

6 -0.0000781 -.01607 | .|. | 0.027558

7 0.00017111 0.03523 | .|* | 0.027565

8 -0.0002120 -.04365 | *|. | 0.027599

9 5.45905E-7 0.00011 | .|. | 0.027650

10 0.00032602 0.06713 | .|* | 0.027650

11 -0.0001244 -.02562 | *|. | 0.027770

12 0.00033929 0.06986 | .|* | 0.027788

13 0.00016077 0.03310 | .|* | 0.027918

14 -0.0002173 -.04475 | *|. | 0.027947

15 -1.646E-6 -.00034 | .|. | 0.028000

16 -0.0000739 -.01522 | .|. | 0.028000

17 0.00028972 0.05965 | .|* | 0.028006

18 -3.1303E-6 -.00064 | .|. | 0.028100

19 0.00004855 0.01000 | .|. | 0.028100

20 -0.0000230 -.00473 | .|. | 0.028103

21 -0.0001165 -.02398 | .|. | 0.028104

22 0.00002614 0.00538 | .|. | 0.028119

23 0.00017892 0.03684 | .|* | 0.028120

24 -0.0001239 -.02551 | *|. | 0.028155

Table 3.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

35
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1403.26 0.0001 -38.25 <.0001

1 -1696.12 0.0001 -29.13 <.0001

2 -1622.54 0.0001 -22.49 <.0001

Single Mean 0 -1403.54 0.0001 -38.25 <.0001 731.43 0.0010

1 -1697.19 0.0001 -29.13 <.0001 424.20 0.0010

2 -1624.78 0.0001 -22.49 <.0001 252.86 0.0010

Trend 0 -1405.70 0.0001 -38.29 <.0001 733.22 0.0010

1 -1706.57 0.0001 -29.20 <.0001 426.32 0.0010

2 -1645.72 0.0001 -22.57 <.0001 254.74 0.0010

AUTOCORRELATION

Graph 4.1 – Original data

Table 4.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0057

Table 4.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

36
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1341.1670 <.0001

AR(2) 1341.1799 <.0001

AR(3) 1341.2363 <.0001

AR(4) 1341.2384 <.0001

Graph 4.2 – 1. difference

Table 4.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 2.0839

Table 4.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

37
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 2.3752 0.1233

AR(2) 14.5617 0.0007

AR(3) 14.8020 0.0020

AR(4) 15.8466 0.0032

38
SLOPE OF TERM STRUCTURE

STATIONARITY

Table 5.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.720257 1.00000 | |********************| 0

1 0.717911 0.99674 | .|********************| 0.027227

2 0.715454 0.99333 | . |********************| 0.047056

3 0.713039 0.98998 | . |********************| 0.060639

4 0.710725 0.98677 | . |********************| 0.071625

5 0.708443 0.98360 | . |********************| 0.081078

6 0.706198 0.98048 | . |********************| 0.089488

7 0.704096 0.97756 | . |********************| 0.097125

8 0.702016 0.97468 | . |******************* | 0.104164

9 0.700004 0.97188 | . |******************* | 0.110718

10 0.698029 0.96914 | . |******************* | 0.116871

11 0.695978 0.96629 | . |******************* | 0.122684

12 0.693904 0.96341 | . |******************* | 0.128202

13 0.691750 0.96042 | . |******************* | 0.133461

14 0.689654 0.95751 | . |******************* | 0.138490

15 0.687511 0.95454 | . |******************* | 0.143313

16 0.685349 0.95153 | . |******************* | 0.147951

17 0.683003 0.94828 | . |******************* | 0.152420

18 0.680581 0.94491 | . |******************* | 0.156732

19 0.678325 0.94178 | . |******************* | 0.160900

20 0.676034 0.93860 | . |******************* | 0.164935

21 0.673647 0.93529 | . |******************* | 0.168849

22 0.671053 0.93169 | . |******************* | 0.172646

23 0.668419 0.92803 | . |******************* | 0.176334

24 0.665820 0.92442 | . |****************** | 0.179918

39
Table 5.2– Original Data
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.9054 0.4896 -0.78 0.3802

1 -1.0598 0.4628 -0.84 0.3545

2 -1.0708 0.4610 -0.83 0.3585

Single Mean 0 -1.7601 0.8068 -1.03 0.7435 0.54 0.9417

1 -2.0694 0.7705 -1.11 0.7129 0.63 0.9140

2 -2.1121 0.7654 -1.11 0.7133 0.62 0.9148

Trend 0 -1.8991 0.9727 -1.12 0.9247 6.72 0.0370

1 -2.1962 0.9651 -1.20 0.9101 6.01 0.0630

2 -2.2285 0.9642 -1.20 0.9099 5.82 0.0738

40
Table 5.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.0015550 1.00000 | |********************| 0

1 0.00012929 0.08314 | .|** | 0.027237

2 0.00004041 0.02599 | .|* | 0.027424

3 3.38519E-6 0.00218 | .|. | 0.027443

4 0.00003142 0.02021 | .|. | 0.027443

5 -0.0000397 -.02554 | *|. | 0.027454

6 -0.0000905 -.05818 | *|. | 0.027471

7 -1.8168E-6 -.00117 | .|. | 0.027563

8 1.1555E-6 0.00074 | .|. | 0.027563

9 0.00003788 0.02436 | .|. | 0.027563

10 0.00001340 0.00862 | .|. | 0.027579

11 0.00006302 0.04053 | .|* | 0.027581

12 0.00001376 0.00885 | .|. | 0.027625

13 -4.1149E-6 -.00265 | .|. | 0.027627

14 0.00001956 0.01258 | .|. | 0.027627

15 0.00003424 0.02202 | .|. | 0.027631

16 0.00005857 0.03767 | .|* | 0.027644

17 0.00008565 0.05508 | .|* | 0.027682

18 -0.0000366 -.02354 | .|. | 0.027764

19 0.00010694 0.06877 | .|* | 0.027778

20 0.00016064 0.10331 | .|** | 0.027904

21 0.00012103 0.07784 | .|** | 0.028187

22 0.00005078 0.03266 | .|* | 0.028346

23 -0.0000247 -.01591 | .|. | 0.028374

24 9.3807E-6 0.00603 | .|. | 0.028380

Table 5.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

41
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1234.91 0.0001 -33.74 <.0001

1 -1186.91 0.0001 -24.31 <.0001

2 -1190.39 0.0001 -20.22 <.0001

Single Mean 0 -1234.92 0.0001 -33.73 <.0001 568.85 0.0010

1 -1186.95 0.0001 -24.30 <.0001 295.16 0.0010

2 -1190.47 0.0001 -20.21 <.0001 204.25 0.0010

Trend 0 -1246.08 0.0001 -34.01 <.0001 578.21 0.0010

1 -1215.54 0.0001 -24.59 <.0001 302.28 0.0010

2 -1244.63 0.0001 -20.53 <.0001 210.72 0.0010

AUTOCORRELATION

Graph 6.1 – Original data

Table 6.1 – Original data

Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0022

Table 6.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

42
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1342.8525 <.0001

AR(2) 1342.8660 <.0001

AR(3) 1342.8664 <.0001

AR(4) 1342.8666 <.0001

Graph 6.2 – 1. difference

Table 6.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 1.8328

43
Table 6.4 – 1. difference
Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 9.3258 0.0023

AR(2) 9.8266 0.0073

AR(3) 9.8291 0.0201

AR(4) 10.3633 0.0347

MODEL

CO-INTEGRATION

Table 7.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0042 6.0197 15.34
1 1 0.0003 0.3718 3.84

44
VAR-MODEL

Table 7.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 -0.08735 0.07114 -1.23 0.2197 1
XL0_1_1 -0.25408 0.03914 -6.49 0.0001 VIX(t)
XL0_1_2 1.24396 1.06906 1.16 0.2448 TRATE(t)
XL0_1_3 -0.77067 1.82220 -0.42 0.6724 SLOPE(t)
AR1_1_1 0.37176 0.02708 13.73 0.0001 CDS (t-1)
AR1_1_2 -2.39199 0.14845 -16.11 0.0001 EQ (t-1)
AR2_1_1 0.05620 0.02529 2.22 0.0265 CDS (t-2)
AR2_1_2 -0.93920 0.15738 -5.97 0.0001 EQ (t-2)
EQ 1 diff Constant2 -0.00050 0.00787 -0.06 0.9495 1
XL0_2_1 0.19530 0.00433 45.08 0.0001 VIX(t)
XL0_2_2 -0.80278 0.11832 -6.78 0.0001 TRATE(t)
XL0_2_3 0.53786 0.20168 2.67 0.0077 SLOPE(t)
AR1_2_1 0.00639 0.00300 2.13 0.0333 CDS (t-1)
AR1_2_2 -0.00311 0.01643 -0.19 0.8499 EQ (t-1)
AR2_2_1 -0.01451 0.00280 -5.18 0.0001 CDS (t-2)
AR2_2_2 0.00482 0.01742 0.28 0.7819 EQ (t-2)

45
GRANGER CAUSALITY TEST

Table 7.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 8 253.00 < .0001
2 6 197.89 < .0001
3 8 36.12 < .0001
4 6 35.99 < .0001
5 2 235.37 < .0001
6 2 0.91 0.6347
7 2 27.26 < .0001
8 2 6.54 0.0380
9 2 0.73 0.6951
10 2 195.41 < .0001
11 2 21.46 < .0001
12 2 7.95 0.0187

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff VIX1diff TRATE1diff SLOPE1diff
Test 2: Group 1 Variable: CDS 1 diff
Group 2 Variable: VIX1diff TRATE1diff SLOPE1diff
Test 3: Group 1 Variable: EQ 1diff
Group 2 Variable: CDS 1diff VIX1diff TRATE1diff SLOPE1diff
Test 4: Group 1 Variable: EQ 1diff
Group 2 Variable: VIX1diff TRATE1diff SLOPE1diff
Test 5: Group 1 Variable: CDS 1diff
Group 2 Variable: EQ 1diff
Test 6: Group 1 Variable: EQ 1diff
Group 2 Variable: CDS1diff
Test 7: Group 1 Variable: EQ1diff
Group 2 Variable: VIX1diff
Test 8: Group 1 Variable: EQ1diff
Group 2 Variable: TRATE1diff
Test 9: Group 1 Variable: EQ1diff
Group 2 Variable: SLOPE1diff
Test 10: Group 1 Variable: CDS1diff
Group 2 Variable: VIX1diff
Test 11: Group 1 Variable: CDS1diff
Group 2 Variable: TRATE1diff
Test 12: Group 1 Variable: CDS1diff
Group 2 Variable: SLOPE1diff

46
UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 7.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.3674 2.60282 111.03 < .0001
EQ 1diff 0.6631 0.28808 376.16 < .0001

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 7.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 2.04381 9999.99 < .0001 115.12 < .0001
EQ 1diff 2.14053 1183.28 < .0001 75.17 < .0001

47
Appendix 4 – INDEX QUARTILE 1
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 487.841 1.00000 | |********************| 0

1 486.937 0.99815 | .|********************| 0.027227

2 485.710 0.99563 | . |********************| 0.047100

3 484.243 0.99263 | . |********************| 0.060729

4 482.626 0.98931 | . |********************| 0.071755

5 480.993 0.98596 | . |********************| 0.081240

6 479.274 0.98244 | . |********************| 0.089672

7 477.507 0.97882 | . |********************| 0.097325

8 475.692 0.97510 | . |********************| 0.104367

9 473.848 0.97132 | . |******************* | 0.110915

10 471.970 0.96747 | . |******************* | 0.117051

11 470.037 0.96350 | . |******************* | 0.122836

12 468.014 0.95936 | . |******************* | 0.128316

13 465.936 0.95510 | . |******************* | 0.133527

14 463.843 0.95081 | . |******************* | 0.138499

15 461.698 0.94641 | . |******************* | 0.143256

16 459.477 0.94186 | . |******************* | 0.147818

17 457.190 0.93717 | . |******************* | 0.152202

18 454.851 0.93238 | . |******************* | 0.156421

19 452.491 0.92754 | . |******************* | 0.160488

20 450.113 0.92266 | . |****************** | 0.164414

21 447.770 0.91786 | . |****************** | 0.168208

48
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

22 445.527 0.91326 | . |****************** | 0.171881

23 443.326 0.90875 | . |****************** | 0.175441

24 441.160 0.90431 | . |****************** | 0.178896

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.9897 0.9112 1.15 0.9355

1 0.4921 0.8053 0.37 0.7920

2 0.2556 0.7446 0.17 0.7366

Single Mean 0 0.3447 0.9722 0.24 0.9748 0.80 0.8653

1 -1.1035 0.8765 -0.49 0.8899 0.46 0.9609

2 -1.7870 0.8037 -0.71 0.8420 0.52 0.9458

Trend 0 -1.3734 0.9831 -0.81 0.9635 2.01 0.7735

1 -3.2588 0.9257 -1.26 0.8963 1.48 0.8824

2 -4.1250 0.8791 -1.42 0.8545 1.52 0.8732

49
Table 1.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.772498 1.00000 | |********************| 0

1 0.307236 0.39772 | .|******** | 0.027237

2 0.195739 0.25338 | .|***** | 0.031249

3 0.109234 0.14140 | .|*** | 0.032738

4 0.0068702 0.00889 | .|. | 0.033188

5 0.040710 0.05270 | .|* | 0.033190

6 0.015233 0.01972 | .|. | 0.033252

7 0.024525 0.03175 | .|* | 0.033261

8 0.016204 0.02098 | .|. | 0.033283

9 0.046833 0.06063 | .|* | 0.033293

10 0.080371 0.10404 | .|** | 0.033375

11 0.056075 0.07259 | .|* | 0.033614

12 0.0023241 0.00301 | .|. | 0.033730

13 -0.022018 -.02850 | *|. | 0.033731

14 -0.0073153 -.00947 | .|. | 0.033749

15 0.049955 0.06467 | .|* | 0.033751

16 0.053134 0.06878 | .|* | 0.033842

17 0.050843 0.06582 | .|* | 0.033946

18 0.010021 0.01297 | .|. | 0.034040

19 0.010322 0.01336 | .|. | 0.034044

20 -0.045009 -.05826 | *|. | 0.034048

21 -0.075963 -.09833 | **|. | 0.034122

22 -0.031616 -.04093 | *|. | 0.034331

23 -0.016852 -.02181 | .|. | 0.034368

24 0.045126 0.05842 | .|* | 0.0343

50
Table 1.4 – 1. Difference
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -810.151 0.0001 -24.06 <.0001

1 -643.525 0.0001 -17.91 <.0001

2 -628.007 0.0001 -15.91 <.0001

Single Mean 0 -811.108 0.0001 -24.07 <.0001 289.66 0.0010

1 -644.919 0.0001 -17.92 <.0001 160.60 0.0010

2 -630.029 0.0001 -15.92 <.0001 126.73 0.0010

Trend 0 -813.113 0.0001 -24.10 <.0001 290.42 0.0010

1 -647.762 0.0001 -17.95 <.0001 161.11 0.0010

2 -634.339 0.0001 -15.95 <.0001 127.26 0.0010

51
EQUITY

STATIONARITY

Table 2.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 51.807088 1.00000 | |********************| 0

1 51.607412 0.99615 | .|********************| 0.027227

2 51.445070 0.99301 | . |********************| 0.047037

3 51.309426 0.99039 | . |********************| 0.060617

4 51.144684 0.98721 | . |********************| 0.071614

5 50.983886 0.98411 | . |********************| 0.081077

6 50.826616 0.98107 | . |********************| 0.089495

7 50.667581 0.97800 | . |********************| 0.097141

8 50.510637 0.97498 | . |******************* | 0.104185

9 50.338238 0.97165 | . |******************* | 0.110742

10 50.169792 0.96840 | . |******************* | 0.116891

11 50.004112 0.96520 | . |******************* | 0.122694

12 49.831349 0.96186 | . |******************* | 0.128199

13 49.650716 0.95838 | . |******************* | 0.133442

14 49.458948 0.95468 | . |******************* | 0.138450

15 49.276272 0.95115 | . |******************* | 0.143247

16 49.108181 0.94790 | . |******************* | 0.147854

17 48.921946 0.94431 | . |******************* | 0.152292

18 48.724164 0.94049 | . |******************* | 0.156573

19 48.546710 0.93707 | . |******************* | 0.160706

20 48.364833 0.93356 | . |******************* | 0.164707

21 48.169356 0.92978 | . |******************* | 0.168583

22 47.990963 0.92634 | . |******************* | 0.172343

23 47.801711 0.92269 | . |****************** | 0.175995

52
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 47.610694 0.91900 | . |****************** | 0.179545

Table 2.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.1093 0.6581 -0.25 0.5975

1 -0.0928 0.6619 -0.24 0.6004

2 -0.0786 0.6651 -0.23 0.6048

Single Mean 0 -3.4903 0.5980 -1.30 0.6326 0.84 0.8550

1 -2.6084 0.7049 -1.10 0.7161 0.61 0.9181

2 -2.0317 0.7750 -0.96 0.7681 0.47 0.9599

Trend 0 -2.0719 0.9685 -0.71 0.9717 1.58 0.8619

1 -1.0401 0.9879 -0.40 0.9873 1.76 0.8255

2 -0.3434 0.9943 -0.15 0.9940 2.12 0.7527

53
Table 2.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.277040 1.00000 | |********************| 0

1 -0.036145 -.13047 | ***|. | 0.027237

2 -0.026308 -.09496 | **|. | 0.027696

3 0.024305 0.08773 | .|** | 0.027937

4 -0.0016870 -.00609 | .|. | 0.028141

5 -0.0030630 -.01106 | .|. | 0.028142

6 -0.0008476 -.00306 | .|. | 0.028145

7 -0.0053859 -.01944 | .|. | 0.028145

8 0.019296 0.06965 | .|* | 0.028155

9 -0.0078143 -.02821 | *|. | 0.028283

10 0.0075814 0.02737 | .|* | 0.028303

11 0.0053409 0.01928 | .|. | 0.028323

12 0.0036034 0.01301 | .|. | 0.028333

13 0.0094548 0.03413 | .|* | 0.028337

14 -0.0057296 -.02068 | .|. | 0.028368

15 -0.022478 -.08114 | **|. | 0.028379

16 0.017017 0.06142 | .|* | 0.028550

17 0.015712 0.05671 | .|* | 0.028648

18 -0.022095 -.07975 | **|. | 0.028731

19 0.0037307 0.01347 | .|. | 0.028895

20 0.0082480 0.02977 | .|* | 0.028900

21 -0.018674 -.06741 | *|. | 0.028923

22 0.0082810 0.02989 | .|* | 0.029039

23 0.0068175 0.02461 | .|. | 0.029062

24 -0.0087783 -.03169 | *|. | 0.029077

54
Table 2.4 – 1. Difference
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1522.75 0.0001 -41.84 <.0001

1 -1913.12 0.0001 -30.91 <.0001

2 -1564.45 0.0001 -22.21 <.0001

Single Mean 0 -1522.75 0.0001 -41.82 <.0001 874.57 0.0010

1 -1913.13 0.0001 -30.90 <.0001 477.31 0.0010

2 -1564.48 0.0001 -22.20 <.0001 246.41 0.0010

Trend 0 -1525.74 0.0001 -41.90 <.0001 877.78 0.0010

1 -1927.90 0.0001 -31.00 <.0001 480.56 0.0010

2 -1591.15 0.0001 -22.31 <.0001 248.77 0.0010

55
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0049 7.8734 15.34
1 1 0.0009 1.2181 3.84

VAR-MODEL

Table 3.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.01669 0.02095 0.8 0.4258 1
AR1_1_1 0.2817 0.02693 10.46 0.0001 CDS (t-1)
AR1_1_2 -0.40939 0.04065 -10.07 0.0001 EQ (t-1)
AR2_1_1 0.14763 0.02616 5.64 0.0001 CDS (t-2)
AR2_1_2 -0.26813 0.04141 -6.48 0.0001 EQ (t-2)
EQ 1 diff Constant2 -0.00166 0.01417 -0.12 0.9068 1
AR1_2_1 0.02454 0.01821 1.35 0.1779 CDS (t-1)
AR1_2_2 -0.14057 0.02749 -5.11 0.0001 EQ (t-1)
AR2_2_1 -0.00243 0.01769 -0.14 0.8908 CDS (t-2)
AR2_2_2 -0.10397 0.028 -3.71 0.0002 EQ (t-2)

GRANGER CAUSALITY TEST

Table 3.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 125.97 <0.0001
2 2 2.01 0.3857

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

56
UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.2404 1.76789 106.11 <.0001
EQ 1diff 0.0313 0.51931 10.82 <.0001

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 2.04137 9999.99 <.0001 47.42 <.0001
EQ 1diff 1.98496 2423.68 <.0001 50.23 <.0001

57
Appendix 5 – INDEX QUARTILE 2
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 1747.380 1.00000 | |********************| 0

1 1743.298 0.99766 | .|********************| 0.027227

2 1737.725 0.99447 | . |********************| 0.047085

3 1731.436 0.99088 | . |********************| 0.060689

4 1724.827 0.98709 | . |********************| 0.071686

5 1718.282 0.98335 | . |********************| 0.081138

6 1711.565 0.97950 | . |********************| 0.089538

7 1704.754 0.97561 | . |********************| 0.097157

8 1697.606 0.97152 | . |******************* | 0.104166

9 1690.526 0.96746 | . |******************* | 0.110679

10 1683.443 0.96341 | . |******************* | 0.116780

11 1676.059 0.95918 | . |******************* | 0.122530

12 1668.295 0.95474 | . |******************* | 0.127975

13 1660.445 0.95025 | . |******************* | 0.133151

14 1652.588 0.94575 | . |******************* | 0.138086

15 1644.437 0.94109 | . |******************* | 0.142807

16 1635.872 0.93619 | . |******************* | 0.147333

17 1627.011 0.93111 | . |******************* | 0.151678

18 1617.962 0.92594 | . |******************* | 0.155858

19 1608.919 0.92076 | . |****************** | 0.159884

20 1599.814 0.91555 | . |****************** | 0.163767

21 1590.779 0.91038 | . |****************** | 0.167519

22 1582.112 0.90542 | . |****************** | 0.171147

58
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

23 1573.597 0.90055 | . |****************** | 0.174661

24 1565.242 0.89577 | . |****************** | 0.178070

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 1.2456 0.9455 1.20 0.9416

1 0.4815 0.8026 0.30 0.7732

2 0.5477 0.8192 0.35 0.7860

Single Mean 0 0.3169 0.9712 0.19 0.9723 0.99 0.8176

1 -1.5389 0.8316 -0.61 0.8654 0.59 0.9253

2 -1.4198 0.8445 -0.58 0.8734 0.60 0.9231

Trend 0 -1.9099 0.9725 -0.93 0.9512 1.61 0.8548

1 -4.7989 0.8350 -1.52 0.8216 1.64 0.8484

2 -4.5712 0.8505 -1.48 0.8368 1.58 0.8615

59
Table 1.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 3.451462 1.00000 | |********************| 0

1 1.388682 0.40235 | .|******** | 0.027237

2 0.497149 0.14404 | .|*** | 0.031337

3 0.130639 0.03785 | .|* | 0.031825

4 -0.070437 -.02041 | .|. | 0.031858

5 -0.045129 -.01308 | .|. | 0.031868

6 -0.115866 -.03357 | *|. | 0.031872

7 0.213033 0.06172 | .|* | 0.031898

8 -0.044576 -.01292 | .|. | 0.031986

9 0.163090 0.04725 | .|* | 0.031990

10 0.448833 0.13004 | .|*** | 0.032042

11 0.119061 0.03450 | .|* | 0.032431

12 -0.152535 -.04419 | *|. | 0.032458

13 -0.231334 -.06703 | *|. | 0.032503

14 -0.065488 -.01897 | .|. | 0.032605

15 0.216767 0.06280 | .|* | 0.032613

16 0.203559 0.05898 | .|* | 0.032703

17 0.184253 0.05338 | .|* | 0.032782

18 -0.113668 -.03293 | *|. | 0.032846

19 -0.062587 -.01813 | .|. | 0.032871

20 -0.104976 -.03041 | *|. | 0.032878

21 -0.179136 -.05190 | *|. | 0.032899

22 -0.060802 -.01762 | .|. | 0.032960

23 -0.013162 -.00381 | .|. | 0.032967

24 0.139917 0.04054 | .|* | 0.032967

Table 1.4 – 1. Difference

60
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -803.502 0.0001 -23.91 <.0001

1 -835.912 0.0001 -20.40 <.0001

2 -867.347 0.0001 -18.04 <.0001

Single Mean 0 -804.667 0.0001 -23.92 <.0001 286.18 0.0010

1 -838.286 0.0001 -20.42 <.0001 208.55 0.0010

2 -871.462 0.0001 -18.07 <.0001 163.21 0.0010

Trend 0 -806.099 0.0001 -23.94 <.0001 286.66 0.0010

1 -841.167 0.0001 -20.45 <.0001 209.04 0.0010

2 -876.766 0.0001 -18.10 <.0001 163.75 0.0010

61
EQUITY

STATIONARITY

Table 2.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 41.830086 1.00000 | |********************| 0

1 41.635294 0.99534 | .|********************| 0.027227

2 41.469759 0.99139 | . |********************| 0.047012

3 41.320370 0.98781 | . |********************| 0.060558

4 41.146891 0.98367 | . |********************| 0.071512

5 40.978243 0.97964 | . |********************| 0.080923

6 40.817039 0.97578 | . |********************| 0.089282

7 40.653424 0.97187 | . |******************* | 0.096865

8 40.495265 0.96809 | . |******************* | 0.103842

9 40.322806 0.96397 | . |******************* | 0.110330

10 40.144567 0.95971 | . |******************* | 0.116406

11 39.970628 0.95555 | . |******************* | 0.122131

12 39.786625 0.95115 | . |******************* | 0.127552

13 39.599867 0.94668 | . |******************* | 0.132706

14 39.403549 0.94199 | . |******************* | 0.137621

15 39.218775 0.93757 | . |******************* | 0.142320

16 39.049548 0.93353 | . |******************* | 0.146828

17 38.860025 0.92900 | . |******************* | 0.151163

18 38.662212 0.92427 | . |****************** | 0.155338

19 38.482271 0.91997 | . |****************** | 0.159363

20 38.298194 0.91557 | . |****************** | 0.163252

21 38.104342 0.91093 | . |****************** | 0.167015

22 37.925771 0.90666 | . |****************** | 0.170658

23 37.733818 0.90207 | . |****************** | 0.174192

24 37.547004 0.89761 | . |****************** | 0.177622

62
Table 2.2 – Original Data
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.1334 0.6526 -0.29 0.5802

1 -0.1163 0.6565 -0.28 0.5838

2 -0.1058 0.6589 -0.28 0.5863

Single Mean 0 -4.1575 0.5220 -1.39 0.5912 0.96 0.8243

1 -3.2987 0.6207 -1.22 0.6698 0.75 0.8805

2 -2.8168 0.6793 -1.11 0.7129 0.63 0.9139

Trend 0 -3.1124 0.9323 -1.00 0.9428 1.68 0.8410

1 -2.1754 0.9657 -0.77 0.9667 1.74 0.8286

2 -1.6398 0.9783 -0.62 0.9771 1.86 0.8044

63
Table 2.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.278924 1.00000 | |********************| 0

1 -0.029002 -.10398 | **|. | 0.027237

2 -0.016501 -.05916 | *|. | 0.027530

3 0.019528 0.07001 | .|* | 0.027624

4 -0.0038686 -.01387 | .|. | 0.027755

5 -0.0063284 -.02269 | .|. | 0.027760

6 -0.0014621 -.00524 | .|. | 0.027774

7 -0.0062103 -.02227 | .|. | 0.027775

8 0.015943 0.05716 | .|* | 0.027788

9 0.0020174 0.00723 | .|. | 0.027875

10 0.0049988 0.01792 | .|. | 0.027876

11 0.0084844 0.03042 | .|* | 0.027885

12 -0.0012208 -.00438 | .|. | 0.027910

13 0.0088700 0.03180 | .|* | 0.027910

14 -0.0085014 -.03048 | *|. | 0.027937

15 -0.025642 -.09193 | **|. | 0.027962

16 0.017937 0.06431 | .|* | 0.028185

17 0.013097 0.04695 | .|* | 0.028294

18 -0.017944 -.06433 | *|. | 0.028351

19 0.0028112 0.01008 | .|. | 0.028459

20 0.0033589 0.01204 | .|. | 0.028462

21 -0.012371 -.04435 | *|. | 0.028466

22 0.011658 0.04179 | .|* | 0.028517

23 0.00012880 0.00046 | .|. | 0.028562

24 -0.0033936 -.01217 | .|. | 0.028562

Table 2.4 – 1. Difference

64
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1487.05 0.0001 -40.72 <.0001

1 -1712.17 0.0001 -29.24 <.0001

2 -1428.20 0.0001 -21.57 <.0001

Single Mean 0 -1487.06 0.0001 -40.71 <.0001 828.61 0.0010

1 -1712.21 0.0001 -29.23 <.0001 427.10 0.0010

2 -1428.27 0.0001 -21.56 <.0001 232.41 0.0010

Trend 0 -1489.67 0.0001 -40.77 <.0001 831.22 0.0010

1 -1723.28 0.0001 -29.31 <.0001 429.51 0.0010

2 -1447.82 0.0001 -21.65 <.0001 234.25 0.0010

MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0036 5.3292 15.34
1 1 0.0003 0.441 3.84

65
VAR-MODEL

Table 3.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.04554 0.04469 1.02 0.3084 1
AR1_1_1 0.32866 0.02743 11.98 0.0001 CDS (t-1)
AR1_1_2 -0.77904 0.08684 -8.97 0.0001 EQ (t-1)
AR2_1_1 0.0012 0.02648 0.05 0.964 CDS (t-2)
AR2_1_2 -0.5991 0.08819 -6.79 0.0001 EQ (t-2)
EQ 1 diff Constant2 -0.0027 0.01432 -0.19 0.8505 1
AR1_2_1 0.00213 0.00879 0.24 0.8087 CDS (t-1)
AR1_2_2 -0.11207 0.02782 -4.03 0.0001 EQ (t-1)
AR2_2_1 0.01301 0.00848 1.53 0.1254 CDS (t-2)
AR2_2_2 -0.06256 0.02825 -2.21 0.027 EQ (t-2)

GRANGER CAUSALITY TEST

Table 3.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 111.71 <.0001
2 2 3.21 0.2011

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.2268 1.63777 98.33 <.0001
EQ 1diff 0.0181 0.52468 6.18 <.0001

66
UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 2.03194 9999.99 <.0001 57.76 <.0001
EQ 1diff 1.99186 1905.64 <.0001 34.22 <.0001

67
Appendix 6 – INDEX QUARTILE 3
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 6161.916 1.00000 | |********************| 0

1 6150.988 0.99823 | .|********************| 0.027227

2 6134.790 0.99560 | . |********************| 0.047102

3 6116.168 0.99258 | . |********************| 0.060730

4 6096.115 0.98932 | . |********************| 0.071755

5 6075.565 0.98599 | . |********************| 0.081240

6 6053.601 0.98242 | . |********************| 0.089673

7 6030.547 0.97868 | . |********************| 0.097325

8 6007.370 0.97492 | . |******************* | 0.104366

9 5984.693 0.97124 | . |******************* | 0.110911

10 5961.822 0.96753 | . |******************* | 0.117046

11 5937.754 0.96362 | . |******************* | 0.122832

12 5912.101 0.95946 | . |******************* | 0.128314

13 5885.383 0.95512 | . |******************* | 0.133526

14 5858.410 0.95074 | . |******************* | 0.138498

15 5830.921 0.94628 | . |******************* | 0.143254

16 5802.046 0.94160 | . |******************* | 0.147815

17 5771.816 0.93669 | . |******************* | 0.152197

18 5741.068 0.93170 | . |******************* | 0.156412

19 5710.346 0.92672 | . |******************* | 0.160473

20 5679.415 0.92170 | . |****************** | 0.164392

21 5648.686 0.91671 | . |****************** | 0.168180

68
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

22 5618.677 0.91184 | . |****************** | 0.171844

23 5589.241 0.90706 | . |****************** | 0.175394

24 5560.034 0.90232 | . |****************** | 0.178837

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.8530 0.8871 0.88 0.8981

1 0.0501 0.6947 0.03 0.6933

2 0.2538 0.7441 0.17 0.7365

Single Mean 0 -0.1318 0.9506 -0.08 0.9492 0.72 0.8882

1 -2.1759 0.7577 -0.86 0.8007 0.64 0.9090

2 -1.7166 0.8118 -0.73 0.8371 0.60 0.9231

Trend 0 -2.3554 0.9604 -1.17 0.9145 1.52 0.8734

1 -5.6427 0.7725 -1.73 0.7392 1.77 0.8232

2 -4.8710 0.8299 -1.60 0.7916 1.60 0.8565

69
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 10.980745 1.00000 | |********************| 0

1 4.943245 0.45017 | .|********* | 0.027237

2 1.568950 0.14288 | .|*** | 0.032288

3 0.980727 0.08931 | .|** | 0.032754

4 0.414541 0.03775 | .|* | 0.032934

5 0.736206 0.06705 | .|* | 0.032966

6 0.368956 0.03360 | .|* | 0.033067

7 -0.344732 -.03139 | *|. | 0.033092

8 -0.428562 -.03903 | *|. | 0.033114

9 0.592933 0.05400 | .|* | 0.033148

10 1.501306 0.13672 | .|*** | 0.033214

11 0.819516 0.07463 | .|* | 0.033629

12 0.208910 0.01903 | .|. | 0.033751

13 -0.661025 -.06020 | *|. | 0.033759

14 -0.525018 -.04781 | *|. | 0.033839

15 0.417138 0.03799 | .|* | 0.033889

16 0.696176 0.06340 | .|* | 0.033920

17 0.491038 0.04472 | .|* | 0.034008

18 -0.271042 -.02468 | .|. | 0.034052

19 -0.184341 -.01679 | .|. | 0.034065

20 -0.260203 -.02370 | .|. | 0.034071

21 -0.245026 -.02231 | .|. | 0.034083

22 -0.182714 -.01664 | .|. | 0.034094

23 -0.047814 -.00435 | .|. | 0.034100

24 0.529006 0.04818 | .|* | 0.034101

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

70
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -739.345 0.0001 -22.56 <.0001

1 -857.000 0.0001 -20.65 <.0001

2 -716.238 0.0001 -16.75 <.0001

Single Mean 0 -740.148 0.0001 -22.56 <.0001 254.57 0.0010

1 -858.834 0.0001 -20.66 <.0001 213.45 0.0010

2 -718.604 0.0001 -16.76 <.0001 140.43 0.0010

Trend 0 -741.040 0.0001 -22.57 <.0001 254.76 0.0010

1 -860.869 0.0001 -20.67 <.0001 213.72 0.0010

2 -721.295 0.0001 -16.77 <.0001 140.66 0.0010

71
EQUITY

STATIONARITY

Table 2.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 46.238368 1.00000 | |********************| 0

1 46.060748 0.99616 | .|********************| 0.027227

2 45.896805 0.99261 | . |********************| 0.047037

3 45.746703 0.98937 | . |********************| 0.060607

4 45.580544 0.98577 | . |********************| 0.071585

5 45.417202 0.98224 | . |********************| 0.081026

6 45.256484 0.97876 | . |********************| 0.089418

7 45.089886 0.97516 | . |********************| 0.097035

8 44.922913 0.97155 | . |******************* | 0.104047

9 44.742909 0.96766 | . |******************* | 0.110567

10 44.557434 0.96365 | . |******************* | 0.116676

11 44.379741 0.95980 | . |******************* | 0.122434

12 44.189627 0.95569 | . |******************* | 0.127890

13 43.990113 0.95138 | . |******************* | 0.133079

14 43.784257 0.94692 | . |******************* | 0.138029

15 43.584484 0.94260 | . |******************* | 0.142763

16 43.389427 0.93839 | . |******************* | 0.147304

17 43.175923 0.93377 | . |******************* | 0.151671

18 42.954042 0.92897 | . |******************* | 0.155874

19 42.741529 0.92437 | . |****************** | 0.159926

20 42.526218 0.91972 | . |****************** | 0.163838

21 42.298224 0.91479 | . |****************** | 0.167622

22 42.076705 0.91000 | . |****************** | 0.171283

72
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

23 41.844116 0.90497 | . |****************** | 0.174830

24 41.620640 0.90013 | . |****************** | 0.178268

Table 2.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.2782 0.6197 -0.62 0.4487

1 -0.2700 0.6216 -0.65 0.4366

2 -0.2581 0.6243 -0.67 0.4278

Single Mean 0 -1.3754 0.8492 -0.55 0.8780 0.29 0.9900

1 -0.9227 0.8933 -0.40 0.9069 0.25 0.9900

2 -0.5375 0.9246 -0.25 0.9294 0.23 0.9900

Trend 0 -1.4939 0.9810 -0.60 0.9784 1.82 0.8125

1 -1.0295 0.9880 -0.45 0.9858 1.96 0.7837

2 -0.6347 0.9921 -0.30 0.9907 2.23 0.7295

73
Table 2.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.210840 1.00000 | |********************| 0

1 -0.015696 -.07445 | *|. | 0.027237

2 -0.014873 -.07054 | *|. | 0.027387

3 0.010715 0.05082 | .|* | 0.027522

4 -0.0007388 -.00350 | .|. | 0.027591

5 -0.0003609 -.00171 | .|. | 0.027592

6 0.00088089 0.00418 | .|. | 0.027592

7 -0.0031969 -.01516 | .|. | 0.027592

8 0.013149 0.06237 | .|* | 0.027598

9 0.00022184 0.00105 | .|. | 0.027703

10 0.0023432 0.01111 | .|. | 0.027703

11 0.0094430 0.04479 | .|* | 0.027706

12 0.0043277 0.02053 | .|. | 0.027760

13 0.0043163 0.02047 | .|. | 0.027771

14 -0.0012247 -.00581 | .|. | 0.027782

15 -0.014996 -.07113 | *|. | 0.027783

16 0.016879 0.08005 | .|** | 0.027918

17 0.012878 0.06108 | .|* | 0.028088

18 -0.0091542 -.04342 | *|. | 0.028186

19 0.00055910 0.00265 | .|. | 0.028235

20 0.0034836 0.01652 | .|. | 0.028236

21 -0.010097 -.04789 | *|. | 0.028243

22 0.0093670 0.04443 | .|* | 0.028303

23 -0.0012766 -.00605 | .|. | 0.028355

24 -0.0016588 -.00787 | .|. | 0.028356

Table 2.4 – 1. Difference

74
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1446.99 0.0001 -39.52 <.0001

1 -1684.71 0.0001 -29.00 <.0001

2 -1481.66 0.0001 -21.82 <.0001

Single Mean 0 -1447.31 0.0001 -39.51 <.0001 780.67 0.0010

1 -1686.04 0.0001 -29.00 <.0001 420.45 0.0010

2 -1484.22 0.0001 -21.82 <.0001 238.07 0.0010

Trend 0 -1450.86 0.0001 -39.60 <.0001 784.12 0.0010

1 -1700.71 0.0001 -29.11 <.0001 423.69 0.0010

2 -1511.97 0.0001 -21.94 <.0001 240.62 0.0010

MODEL

75
CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0049 7.5076 15.34
1 1 0.0007 0.8945 3.84

VAR-MODEL

Table 3.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.04732 0.07484 0.63 0.5273 1
AR1_1_1 0.38875 0.02744 14.17 0.0001 CDS (t-1)
AR1_1_2 -2.39542 0.16573 -14.45 0.0001 EQ (t-1)
AR2_1_1 -0.02298 0.02561 -0.9 0.3696 CDS (t-2)
AR2_1_2 -0.84414 0.17646 -4.78 0.0001 EQ (t-2)
EQ 1 diff Constant2 -0.00761 0.01247 -0.61 0.5419 1
AR1_2_1 0.00749 0.00457 1.64 0.1015 CDS (t-1)
AR1_2_2 -0.07083 0.02761 -2.57 0.0104 EQ (t-1)
AR2_2_1 -0.00828 0.00427 -1.94 0.0523 CDS (t-2)
AR2_2_2 -0.06177 0.02939 -2.1 0.0358 EQ (t-2)

GRANGER CAUSALITY TEST

Table 3.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 217.84 <.0001
2 2 4.48 0.1066

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

UNIVARIATE MODEL ANOVA DIAGNOSTICS

76
Table 3.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.3179 2.74318 156.26 <.0001
EQ 1diff 0.0147 0.45693 4.99 0.0005

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 2.00527 9999.99 <.0001 75.64 <.0001
EQ 1diff 1.99343 788.3 <.0001 15.03 0.0001

77
Appendix 7 – INDEX QUARTILE 4
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 66244.606 1.00000 | |********************| 0

1 66106.545 0.99792 | .|********************| 0.027227

2 65936.470 0.99535 | . |********************| 0.047092

3 65736.932 0.99234 | . |********************| 0.060717

4 65516.341 0.98901 | . |********************| 0.071739

5 65285.445 0.98552 | . |********************| 0.081220

6 65039.532 0.98181 | . |********************| 0.089647

7 64781.444 0.97791 | . |********************| 0.097292

8 64517.552 0.97393 | . |******************* | 0.104324

9 64252.633 0.96993 | . |******************* | 0.110859

10 63987.650 0.96593 | . |******************* | 0.116981

11 63719.563 0.96188 | . |******************* | 0.122751

12 63439.960 0.95766 | . |******************* | 0.128217

13 63151.039 0.95330 | . |******************* | 0.133414

14 62851.574 0.94878 | . |******************* | 0.138371

15 62545.052 0.94415 | . |******************* | 0.143112

16 62220.886 0.93926 | . |******************* | 0.147658

17 61881.771 0.93414 | . |******************* | 0.152022

18 61535.641 0.92892 | . |******************* | 0.156219

19 61178.057 0.92352 | . |****************** | 0.160262

20 60817.479 0.91807 | . |****************** | 0.164159

21 60456.240 0.91262 | . |****************** | 0.167922

78
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

22 60106.138 0.90734 | . |****************** | 0.171559

23 59765.185 0.90219 | . |****************** | 0.175081

24 59429.765 0.89713 | . |****************** | 0.178494

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 1.7801 0.9821 2.28 0.9949

1 1.3036 0.9514 1.07 0.9260

2 1.2456 0.9455 0.99 0.9150

Single Mean 0 1.0972 0.9905 0.94 0.9961 2.90 0.3252

1 0.0675 0.9608 0.04 0.9607 0.99 0.8167

2 -0.0718 0.9538 -0.04 0.9539 0.94 0.8308

Trend 0 -1.1928 0.9858 -0.72 0.9706 2.36 0.7043

1 -3.1584 0.9303 -1.23 0.9029 1.57 0.8629

2 -3.3589 0.9210 -1.27 0.8947 1.54 0.8706

Table 1.3 – 1. Difference


Graphical inspection of stationarity

79
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 66.389010 1.00000 | |********************| 0

1 27.518176 0.41450 | .|******** | 0.027237

2 12.903900 0.19437 | .|**** | 0.031571

3 10.066270 0.15163 | .|*** | 0.032447

4 8.589973 0.12939 | .|*** | 0.032968

5 10.527960 0.15858 | .|*** | 0.033343

6 4.193588 0.06317 | .|* | 0.033898

7 2.480737 0.03737 | .|* | 0.033985

8 4.156687 0.06261 | .|* | 0.034015

9 6.403292 0.09645 | .|** | 0.034101

10 4.679316 0.07048 | .|* | 0.034303

11 2.101002 0.03165 | .|* | 0.034410

12 3.369152 0.05075 | .|* | 0.034432

13 1.704481 0.02567 | .|* | 0.034487

14 -0.707715 -.01066 | .|. | 0.034501

15 3.670081 0.05528 | .|* | 0.034504

16 6.292292 0.09478 | .|** | 0.034569

17 6.849770 0.10318 | .|** | 0.034761

18 2.700210 0.04067 | .|* | 0.034988

19 -3.378693 -.05089 | *|. | 0.035023

20 -2.849529 -.04292 | *|. | 0.035078

21 -4.383748 -.06603 | *|. | 0.035117

22 -2.942426 -.04432 | *|. | 0.035209

23 -0.936881 -.01411 | .|. | 0.035250

24 0.533386 0.00803 | .|. | 0.035254

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

80
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -784.950 0.0001 -23.51 <.0001

1 -735.502 0.0001 -19.04 <.0001

2 -601.364 0.0001 -15.47 <.0001

Single Mean 0 -787.845 0.0001 -23.56 <.0001 277.47 0.0010

1 -740.726 0.0001 -19.10 <.0001 182.36 0.0010

2 -607.730 0.0001 -15.52 <.0001 120.49 0.0010

Trend 0 -790.310 0.0001 -23.60 <.0001 278.41 0.0010

1 -745.254 0.0001 -19.14 <.0001 183.15 0.0010

2 -613.312 0.0001 -15.56 <.0001 121.08 0.0010

81
EQUITY

STATIONARITY

Table 2.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 69.633586 1.00000 | |********************| 0

1 69.377450 0.99632 | .|********************| 0.027227

2 69.137749 0.99288 | . |********************| 0.047042

3 68.909153 0.98960 | . |********************| 0.060618

4 68.656393 0.98597 | . |********************| 0.071599

5 68.407848 0.98240 | . |********************| 0.081041

6 68.161865 0.97886 | . |********************| 0.089434

7 67.921730 0.97542 | . |********************| 0.097052

8 67.676962 0.97190 | . |******************* | 0.104066

9 67.418154 0.96818 | . |******************* | 0.110590

10 67.149494 0.96433 | . |******************* | 0.116704

11 66.896494 0.96069 | . |******************* | 0.122469

12 66.629758 0.95686 | . |******************* | 0.127933

13 66.350914 0.95286 | . |******************* | 0.133133

14 66.062776 0.94872 | . |******************* | 0.138096

15 65.778474 0.94464 | . |******************* | 0.142845

16 65.504037 0.94070 | . |******************* | 0.147403

17 65.207032 0.93643 | . |******************* | 0.151788

18 64.903073 0.93207 | . |******************* | 0.156012

19 64.611188 0.92787 | . |******************* | 0.160087

20 64.314360 0.92361 | . |****************** | 0.164025

21 64.006585 0.91919 | . |****************** | 0.167836

22 63.702294 0.91482 | . |****************** | 0.171527

23 63.390094 0.91034 | . |****************** | 0.175107

24 63.089742 0.90602 | . |****************** | 0.178580

82
Table 2.2 – Original Data
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.3047 0.6137 -0.72 0.4051

1 -0.3019 0.6143 -0.77 0.3851

2 -0.2974 0.6153 -0.79 0.3741

Single Mean 0 -0.4034 0.9340 -0.18 0.9386 0.26 0.9900

1 -0.0323 0.9559 -0.02 0.9561 0.30 0.9896

2 0.1801 0.9658 0.09 0.9650 0.34 0.9836

Trend 0 -2.1820 0.9655 -0.92 0.9527 2.58 0.6582

1 -1.7276 0.9765 -0.78 0.9660 2.72 0.6316

2 -1.4755 0.9814 -0.70 0.9722 2.88 0.5976

83
Table 2.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.260910 1.00000 | |********************| 0

1 -0.018514 -.07096 | *|. | 0.027237

2 -0.011020 -.04224 | *|. | 0.027374

3 0.015030 0.05761 | .|* | 0.027422

4 -0.0035280 -.01352 | .|. | 0.027511

5 0.0019666 0.00754 | .|. | 0.027516

6 -0.010959 -.04200 | *|. | 0.027518

7 0.00017762 0.00068 | .|. | 0.027565

8 0.013935 0.05341 | .|* | 0.027565

9 0.0026603 0.01020 | .|. | 0.027642

10 0.00026354 0.00101 | .|. | 0.027645

11 0.010563 0.04048 | .|* | 0.027645

12 0.0050721 0.01944 | .|. | 0.027689

13 0.0078636 0.03014 | .|* | 0.027699

14 0.0023674 0.00907 | .|. | 0.027723

15 -0.022707 -.08703 | **|. | 0.027726

16 0.018578 0.07120 | .|* | 0.027927

17 0.014063 0.05390 | .|* | 0.028062

18 -0.012234 -.04689 | *|. | 0.028139

19 0.00058535 0.00224 | .|. | 0.028196

20 0.00065749 0.00252 | .|. | 0.028197

21 -0.0091388 -.03503 | *|. | 0.028197

22 0.0054752 0.02099 | .|. | 0.028229

23 -0.0021918 -.00840 | .|. | 0.028241

24 -0.0021528 -.00825 | .|. | 0.028242

Table 2.4 – 1. Difference

84
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1442.05 0.0001 -39.38 <.0001

1 -1583.64 0.0001 -28.12 <.0001

2 -1348.18 0.0001 -21.16 <.0001

Single Mean 0 -1442.60 0.0001 -39.39 <.0001 775.60 0.0010

1 -1585.77 0.0001 -28.13 <.0001 395.70 0.0010

2 -1351.98 0.0001 -21.17 <.0001 224.11 0.0010

Trend 0 -1447.22 0.0001 -39.50 <.0001 780.22 0.0010

1 -1603.12 0.0001 -28.27 <.0001 399.60 0.0010

2 -1382.10 0.0001 -21.31 <.0001 227.08 0.0010

85
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.005 6.9081 15.34
1 1 0.0002 0.2142 3.84

VAR-MODEL

Table 3.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.22684 0.18852 1.2 0.2291 1
AR1_1_1 0.32067 0.02731 11.74 0.0001 CDS (t-1)
AR1_1_2 -5.35306 0.37236 -14.38 0.0001 EQ (t-1)
AR2_1_1 0.06593 0.02576 2.56 0.0106 CDS (t-2)
AR2_1_2 -1.82639 0.39642 -4.61 0.0001 EQ (t-2)
EQ 1 diff Constant2 -0.01107 0.01394 -0.79 0.4272 1
AR1_2_1 0.00123 0.00202 0.61 0.5428 CDS (t-1)
AR1_2_2 -0.0718 0.02754 -2.61 0.0092 EQ (t-1)
AR2_2_1 -0.00171 0.0019 -0.9 0.3683 CDS (t-2)
AR2_2_2 -0.04266 0.02931 -1.46 0.1458 EQ (t-2)

GRANGER CAUSALITY TEST

Table 3.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 215.72 <.0001
2 2 0.88 0.6456

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

86
UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.2874 6.89533 135.22 <.0001
EQ 1diff 0.008 0.5099 2.69 0.0297

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 2.01629 6964.27 <.0001 114.72 <.0001
EQ 1diff 1.99508 784.65 <.0001 22.77 <.0001

87
Appendix 8 - INDEX 02.01.2004 – 08.08.2007 (BEFORE)
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 96.651300 1.00000 | |********************| 0

1 96.093635 0.99423 | .|********************| 0.033131

2 95.184125 0.98482 | . |********************| 0.057165

3 94.066460 0.97326 | . |******************* | 0.073465

4 92.910225 0.96129 | . |******************* | 0.086467

5 91.691349 0.94868 | . |******************* | 0.097495

6 90.421234 0.93554 | . |******************* | 0.107150

7 89.099412 0.92186 | . |****************** | 0.115770

8 87.750888 0.90791 | . |****************** | 0.123565

9 86.513994 0.89511 | . |****************** | 0.130683

10 85.256492 0.88210 | . |****************** | 0.137248

11 84.119144 0.87034 | . |***************** | 0.143336

12 82.974861 0.85850 | . |***************** | 0.149024

13 81.886565 0.84724 | . |***************** | 0.154358

14 80.822554 0.83623 | . |***************** | 0.159381

15 79.802379 0.82567 | . |***************** | 0.164126

16 78.841662 0.81573 | . |**************** | 0.168624

17 77.942434 0.80643 | . |**************** | 0.172901

18 77.065083 0.79735 | . |**************** | 0.176982

19 76.140307 0.78778 | . |**************** | 0.180882

88
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

20 75.168416 0.77773 | . |**************** | 0.184610

21 74.142288 0.76711 | . |*************** | 0.188172

22 73.130918 0.75665 | . |*************** | 0.191574

23 72.159206 0.74659 | . |*************** | 0.194827

24 71.182406 0.73649 | . |*************** | 0.197943

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.3169 0.7601 0.71 0.8684

1 0.1372 0.7152 0.21 0.7467

2 0.0014 0.6833 0.00 0.6836

Single Mean 0 -0.7553 0.9076 -0.31 0.9212 0.35 0.9823

1 -5.0449 0.4307 -1.38 0.5945 1.06 0.7999

2 -9.1204 0.1661 -1.99 0.2891 2.06 0.5434

Trend 0 3.1552 0.9999 0.99 0.9999 1.87 0.8032

1 -5.0744 0.8151 -1.05 0.9353 0.95 0.9737

2 -13.7558 0.2306 -2.21 0.4855 2.57 0.6615

89
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.633255 1.00000 | |********************| 0

1 0.220137 0.34763 | .|******* | 0.033150

2 0.181942 0.28731 | .|****** | 0.036939

3 0.105308 0.16630 | . |*** | 0.039318

4 0.070679 0.11161 | . |** | 0.040084

5 0.071306 0.11260 | . |** | 0.040424

6 0.053099 0.08385 | . |** | 0.040767

7 0.0030219 0.00477 | .|. | 0.040956

8 0.023053 0.03640 | . |*. | 0.040957

9 0.020737 0.03275 | . |*. | 0.040992

10 0.034394 0.05431 | . |*. | 0.041021

11 0.065409 0.10329 | . |** | 0.041100

12 0.039727 0.06274 | . |*. | 0.041384

13 0.0027958 0.00441 | .|. | 0.041489

14 0.0077430 0.01223 | .|. | 0.041489

15 -0.037091 -.05857 | .*| . | 0.041493

16 -0.0007945 -.00125 | .|. | 0.041584

17 -0.0008221 -.00130 | .|. | 0.041584

18 0.031065 0.04906 | . |*. | 0.041584

19 0.037168 0.05869 | . |*. | 0.041647

20 0.033862 0.05347 | . |*. | 0.041738

21 0.045403 0.07170 | . |*. | 0.041813

22 0.017986 0.02840 | . |*. | 0.041948

23 -0.0048612 -.00768 | .|. | 0.041969

24 0.0081877 0.01293 | .|. | 0.041971

90
Table 1.4 – 1. Difference
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -572.633 0.0001 -19.74 <.0001

1 -379.794 0.0001 -13.21 <.0001

2 -373.728 0.0001 -11.96 <.0001

Single Mean 0 -573.105 0.0001 -19.74 <.0001 194.86 0.0010

1 -380.365 0.0001 -13.20 <.0001 87.20 0.0010

2 -374.410 0.0001 -11.95 <.0001 71.45 0.0010

Trend 0 -574.960 0.0001 -19.75 <.0001 195.19 0.0010

1 -381.972 0.0001 -13.19 <.0001 87.24 0.0010

2 -376.717 0.0001 -11.95 <.0001 71.53 0.0010

AUTOCORRELATION

Graph 2.1 – Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0065

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

91
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 904.9923 <.0001

AR(2) 905.7488 <.0001

AR(3) 905.9484 <.0001

AR(4) 905.9499 <.0001

Graph 2.2 – 1. difference

Table 2.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 1.2444

Table 2.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 116.8977 <.0001

92
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(2) 146.6671 <.0001

AR(3) 146.6821 <.0001

AR(4) 146.6958 <.0001

93
EQUITY

STATIONARITY

Table 3.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 27.870549 1.00000 | |********************| 0

1 27.750637 0.99570 | .|********************| 0.033131

2 27.637238 0.99163 | . |********************| 0.057221

3 27.530262 0.98779 | . |********************| 0.073709

4 27.430339 0.98421 | . |********************| 0.087035

5 27.318613 0.98020 | . |********************| 0.098497

6 27.209672 0.97629 | . |********************| 0.108678

7 27.109133 0.97268 | . |******************* | 0.117913

8 27.004121 0.96891 | . |******************* | 0.126414

9 26.908340 0.96548 | . |******************* | 0.134319

10 26.806943 0.96184 | . |******************* | 0.141732

11 26.690625 0.95766 | . |******************* | 0.148725

12 26.572712 0.95343 | . |******************* | 0.155346

13 26.444233 0.94882 | . |******************* | 0.161642

14 26.321292 0.94441 | . |******************* | 0.167644

15 26.191998 0.93977 | . |******************* | 0.173386

16 26.064613 0.93520 | . |******************* | 0.178890

17 25.939207 0.93070 | . |******************* | 0.184178

18 25.813160 0.92618 | . |******************* | 0.189271

19 25.677569 0.92132 | . |****************** | 0.194182

20 25.546984 0.91663 | . |****************** | 0.198922

21 25.425562 0.91227 | . |****************** | 0.203506

94
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

22 25.307743 0.90805 | . |****************** | 0.207947

23 25.180763 0.90349 | . |****************** | 0.212255

24 25.050059 0.89880 | . |****************** | 0.216435

Table 3.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.3647 0.7724 1.70 0.9786

1 0.3590 0.7709 1.67 0.9773

2 0.3592 0.7710 1.74 0.9808

Single Mean 0 -1.1151 0.8754 -0.64 0.8584 1.81 0.6078

1 -1.0379 0.8827 -0.60 0.8685 1.72 0.6298

2 -0.9126 0.8941 -0.55 0.8789 1.81 0.6057

Trend 0 -21.8472 0.0468 -3.35 0.0588 5.65 0.0829

1 -22.4760 0.0410 -3.40 0.0528 5.81 0.0742

2 -21.0890 0.0549 -3.28 0.0713 5.41 0.0959

95
Table 3.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.091781 1.00000 | |********************| 0

1 -0.0001576 -.00172 | .|. | 0.033150

2 -0.0040826 -.04448 | *|. | 0.033150

3 0.0023759 0.02589 | .|* | 0.033215

4 -0.0022178 -.02416 | .|. | 0.033237

5 -0.0020518 -.02235 | .|. | 0.033257

6 -0.0046455 -.05061 | *|. | 0.033273

7 -0.0020426 -.02226 | .|. | 0.033358

8 -0.0017135 -.01867 | .|. | 0.033374

9 -0.0027249 -.02969 | *|. | 0.033386

10 0.0028345 0.03088 | .|* | 0.033415

11 0.0012678 0.01381 | .|. | 0.033446

12 -0.0011882 -.01295 | .|. | 0.033452

13 -0.0012460 -.01358 | .|. | 0.033458

14 -0.0022198 -.02419 | .|. | 0.033464

15 -0.0012702 -.01384 | .|. | 0.033483

16 -0.0004264 -.00465 | .|. | 0.033489

17 -0.0017054 -.01858 | .|. | 0.033490

18 0.0029611 0.03226 | .|* | 0.033501

19 -0.0023682 -.02580 | *|. | 0.033535

20 -0.0000171 -.00019 | .|. | 0.033557

21 0.00042032 0.00458 | .|. | 0.033557

22 0.00042125 0.00459 | .|. | 0.033558

23 0.0020574 0.02242 | .|. | 0.033559

24 0.0020930 0.02280 | .|. | 0.033575

96
Table 3.4 – 1. Difference
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -907.502 0.0001 -30.07 <.0001

1 -984.544 0.0001 -22.13 <.0001

2 -899.190 0.0001 -17.19 <.0001

Single Mean 0 -910.565 0.0001 -30.15 <.0001 454.54 0.0010

1 -994.674 0.0001 -22.23 <.0001 247.18 0.0010

2 -917.380 0.0001 -17.30 <.0001 149.67 0.0010

Trend 0 -910.617 0.0001 -30.14 <.0001 454.13 0.0010

1 -994.841 0.0001 -22.23 <.0001 246.98 0.0010

2 -917.526 0.0001 -17.29 <.0001 149.56 0.0010

97
AUTOCORRELATION

Graph 4.1 – Original data

Table 4.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0033

Table 4.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 905.9664 <.0001

AR(2) 905.9674 <.0001

AR(3) 905.9713 <.0001

AR(4) 905.9728 <.0001

98
Graph 4.2 – 1. difference

Table 4.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 2.0004

Table 4.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 0.0027 0.9586

AR(2) 1.8083 0.4049

AR(3) 2.4132 0.4912

AR(4) 3.0831 0.5440

99
MODEL

CO-INTEGRATION

Table 5.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0049 4.5135 15.34
1 1 0.0000 0.0281 3.84

VAR-MODEL

Table 5.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 -0.02763 0.02344 -1.18 0.2398 1
AR1_1_1 0.23224 0.03338 6.96 0.0001 CDS (t-1)
AR1_1_2 -0.54424 0.07775 -7.00 0.0001 EQ (t-1)
AR2_1_1 0.18672 0.03239 5.77 0.0001 CDS (t-2)
AR2_1_2 -0.41276 0.07970 -5.18 0.0001 EQ (t-2)
EQ 1 diff Constant2 -0.02027 0.01009 -2.01 0.0448 1
AR1_2_1 -0.03330 0.01437 -2.32 0.0207 CDS (t-1)
AR1_2_2 -0.01500 0.03347 -0.45 0.6542 EQ (t-1)
AR2_2_1 -0.00927 0.01394 -0.66 0.5064 CDS (t-2)
AR2_2_2 -0.06926 0.03431 -2.02 0.0439 EQ (t-2)

GRANGER CAUSALITY TEST

Table 5.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 73.10 <0.0001
2 2 7.67 0.0216

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

100
UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 5.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.2237 0.70242 65.04 < 0.0001
EQ 1diff 0.0104 0.30240 2.37 0.0508

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 5.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 1.90525 5987.61 < 0.0001 67.30 < 0.0001
EQ 1diff 1.99396 255.38 < 0.0001 6.81 0.0092

101
Appendix 9 - INDEX 09.08.2007 – 01.05.2009 (AFTER)
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 16283.564 1.00000 | |********************| 0

1 16233.320 0.99691 | . |********************| 0.047782

2 16167.486 0.99287 | . |********************| 0.082590

3 16091.868 0.98823 | . |********************| 0.106407

4 16009.714 0.98318 | . |********************| 0.125626

5 15925.193 0.97799 | . |********************| 0.142112

6 15836.032 0.97252 | . |******************* | 0.156727

7 15742.621 0.96678 | . |******************* | 0.169947

8 15647.018 0.96091 | . |******************* | 0.182071

9 15549.658 0.95493 | . |******************* | 0.193303

10 15449.930 0.94881 | . |******************* | 0.203789

11 15346.967 0.94248 | . |******************* | 0.213636

12 15240.230 0.93593 | . |******************* | 0.222927

13 15131.410 0.92924 | . |******************* | 0.231725

14 15021.212 0.92248 | . |****************** | 0.240082

15 14909.960 0.91564 | . |****************** | 0.248042

16 14794.327 0.90854 | . |****************** | 0.255643

17 14674.078 0.90116 | . |****************** | 0.262911

18 14550.737 0.89358 | . |****************** | 0.269871

19 14425.125 0.88587 | . |****************** | 0.276544

102
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

20 14299.224 0.87814 | . |****************** | 0.282949

21 14174.374 0.87047 | . |***************** | 0.289104

22 14053.912 0.86307 | . |***************** | 0.295027

23 13935.965 0.85583 | . |***************** | 0.300737

24 13818.758 0.84863 | . |***************** | 0.306247

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.6032 0.8322 1.30 0.9512

1 0.3800 0.7760 0.48 0.8188

2 0.4015 0.7815 0.52 0.8279

Single Mean 0 -0.8374 0.9004 -0.91 0.7837 2.51 0.4291

1 -1.6141 0.8229 -1.05 0.7374 1.24 0.7543

2 -1.5736 0.8274 -1.05 0.7377 1.29 0.7405

Trend 0 -0.8293 0.9901 -0.35 0.9890 0.42 0.9900

1 -5.8926 0.7515 -1.45 0.8435 1.20 0.9373

2 -5.5195 0.7807 -1.38 0.8648 1.11 0.9524

103
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 31.228610 1.00000 | |********************| 0

1 14.845397 0.47538 | . |********** | 0.047782

2 6.365027 0.20382 | . |**** | 0.057576

3 4.164605 0.13336 | . |*** | 0.059200

4 2.977120 0.09533 | . |** | 0.059882

5 4.025380 0.12890 | . |*** | 0.060228

6 1.889053 0.06049 | . |*. | 0.060854

7 0.560985 0.01796 | .|. | 0.060992

8 0.757395 0.02425 | .|. | 0.061004

9 2.513010 0.08047 | . |** | 0.061026

10 2.910430 0.09320 | . |** | 0.061267

11 1.341970 0.04297 | . |*. | 0.061590

12 0.572590 0.01834 | .|. | 0.061659

13 -0.892116 -.02857 | .*| . | 0.061671

14 -1.080856 -.03461 | .*| . | 0.061701

15 1.832296 0.05867 | . |*. | 0.061746

16 2.495283 0.07990 | . |** | 0.061873

17 2.861507 0.09163 | . |** | 0.062108

18 0.626563 0.02006 | .|. | 0.062416

19 -1.318758 -.04223 | .*| . | 0.062431

20 -1.544742 -.04947 | .*| . | 0.062496

21 -2.624431 -.08404 | .**| . | 0.062585

22 -1.700078 -.05444 | . *| . | 0.062842

23 -0.462074 -.01480 | . | . | 0.062950

24 0.961437 0.03079 | . |* . | 0.062958

104
Table 1.4 – 1. Difference
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -226.732 0.0001 -12.36 <.0001

1 -236.584 0.0001 -10.80 <.0001

2 -199.572 0.0001 -8.90 <.0001

Single Mean 0 -228.906 0.0001 -12.42 <.0001 77.16 0.0010

1 -240.668 0.0001 -10.87 <.0001 59.13 0.0010

2 -204.843 0.0001 -8.98 <.0001 40.30 0.0010

Trend 0 -229.266 0.0001 -12.43 <.0001 77.21 0.0010

1 -241.127 0.0001 -10.88 <.0001 59.17 0.0010

2 -205.380 0.0001 -8.98 <.0001 40.38 0.0010

105
AUTOCORRELATION

Graph 2.1 – Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0019

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 435.8099 <.0001

AR(2) 435.8819 <.0001

AR(3) 435.8883 <.0001

AR(4) 435.8909 <.0001

106
Graph 2.2– 1. difference

Table 2.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 1.0460

Table 2.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 99.1505 <.0001

AR(2) 99.3824 <.0001

AR(3) 100.6910 <.0001

AR(4) 100.7847 <.0001

107
EQUITY

STATIONARITY

Table 3.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 81.133591 1.00000 | |********************| 0

1 80.615699 0.99362 | . |********************| 0.047782

2 80.161834 0.98802 | . |********************| 0.082409

3 79.746368 0.98290 | . |********************| 0.106060

4 79.293085 0.97732 | . |********************| 0.125140

5 78.871618 0.97212 | . |******************* | 0.141497

6 78.457282 0.96701 | . |******************* | 0.156002

7 77.994174 0.96131 | . |******************* | 0.169135

8 77.535316 0.95565 | . |******************* | 0.181180

9 77.027118 0.94939 | . |******************* | 0.192345

10 76.482718 0.94268 | . |******************* | 0.202761

11 75.968290 0.93634 | . |******************* | 0.212532

12 75.408974 0.92944 | . |******************* | 0.221750

13 74.879979 0.92292 | . |****************** | 0.230473

14 74.288705 0.91563 | . |****************** | 0.238762

15 73.731254 0.90876 | . |****************** | 0.246648

16 73.192745 0.90213 | . |****************** | 0.254178

17 72.594906 0.89476 | . |****************** | 0.261386

18 71.956316 0.88689 | . |****************** | 0.268287

19 71.390605 0.87991 | . |****************** | 0.274900

20 70.800064 0.87264 | . |***************** | 0.281256

21 70.195723 0.86519 | . |***************** | 0.287371

108
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

22 69.624079 0.85814 | . |***************** | 0.293258

23 69.013716 0.85062 | . |***************** | 0.298936

24 68.416420 0.84326 | . |***************** | 0.304412

Table 3.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.5589 0.5572 -1.46 0.1355

1 -0.5482 0.5595 -1.64 0.0959

2 -0.5434 0.5605 -1.80 0.0691

Single Mean 0 -1.3686 0.8495 -0.79 0.8205 1.18 0.7700

1 -1.0356 0.8826 -0.68 0.8480 1.39 0.7143

2 -0.8441 0.8999 -0.62 0.8637 1.64 0.6524

Trend 0 -11.4032 0.3443 -2.44 0.3596 2.98 0.5804

1 -8.8432 0.5163 -2.13 0.5273 2.27 0.7219

2 -7.2756 0.6395 -1.92 0.6429 1.85 0.8078

109
Table 3.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.558764 1.00000 | |********************| 0

1 -0.077341 -.13841 | ***| . | 0.047782

2 -0.046861 -.08387 | **| . | 0.048689

3 0.048535 0.08686 | . |** | 0.049017

4 -0.010686 -.01913 | .|. | 0.049368

5 -0.0019795 -.00354 | .|. | 0.049384

6 0.0019551 0.00350 | .|. | 0.049385

7 -0.016660 -.02982 | .*| . | 0.049386

8 0.046985 0.08409 | . |** | 0.049427

9 -0.0008937 -.00160 | .|. | 0.049752

10 0.00077748 0.00139 | .|. | 0.049752

11 0.020375 0.03646 | . |*. | 0.049752

12 -0.0004679 -.00084 | .|. | 0.049813

13 0.024220 0.04335 | . |*. | 0.049813

14 -0.017197 -.03078 | .*| . | 0.049899

15 -0.061003 -.10917 | **| . | 0.049943

16 0.047233 0.08453 | . |** | 0.050485

17 0.044474 0.07959 | . |** | 0.050807

18 -0.054735 -.09796 | **| . | 0.051091

19 0.010212 0.01828 | .|. | 0.051518

20 0.018388 0.03291 | . |*. | 0.051533

21 -0.039918 -.07144 | .*| . | 0.051581

22 0.014431 0.02583 | . |*. | 0.051806

23 0.0061530 0.01101 | .|. | 0.051835

24 -0.023905 -.04278 | .*| . | 0.051841

110
Table 3.4 – 1. Difference
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -495.356 0.0001 -24.02 <.0001

1 -602.590 0.0001 -17.36 <.0001

2 -481.471 0.0001 -12.41 <.0001

Single Mean 0 -497.501 0.0001 -24.10 <.0001 290.45 0.0010

1 -612.789 0.0001 -17.48 <.0001 152.79 0.0010

2 -499.018 0.0001 -12.54 <.0001 78.57 0.0010

Trend 0 -497.497 0.0001 -24.07 <.0001 289.79 0.0010

1 -612.801 0.0001 -17.46 <.0001 152.44 0.0010

2 -499.080 0.0001 -12.52 <.0001 78.39 0.0010

111
AUTOCORRELATION

Graph 4.1 – Original data

Table 4.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0068

Table 4.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 433.8382 <.0001

AR(2) 433.8706 <.0001

AR(3) 433.8850 <.0001

AR(4) 433.8866 <.0001

Graph 4.2 – 1. difference

112
Table 4.3 – 1. difference
Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 1.0460

Table 4.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 99.1505 <.0001

AR(2) 99.3824 <.0001

AR(3) 100.6910 <.0001

AR(4) 100.7847 <.0001

113
MODEL

CO-INTEGRATION

Table 5.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0494 22.4888 15.34
1 1 0.0010 0.4204 3.84

VECM-MODEL

Table 5.2 – Long-Run Parameter Beta Estimates When RANK = 1


Variable 1
CDS 0.04851
EQUITY 0.67773

Table 5.3 – Adjustment Coefficient Alpha Estimates When RANK = 1


Variable 1
CDS -0.98413
EQUITY 0.00770

Table 5.4
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS Constant1 37.04224 7.89496 4.69 0.0001 1
AR1_1_1 -0.04774 0.01022 CDS (t-1)
AR1_1_2 -0.66698 0.14277 EQ (t-1)
AR2_1_1 0.41326 0.03833 10.78 0.0001 D_CDS (t-1)
AR2_1_2 -2.12302 0.30645 -6.93 0.0001 D_EQ (t-1)
EQ Constant2 -0.34684 1.32214 -0.26 0.7932 1
AR1_2_1 0.00037 0.00171 CDS (t-1)
AR1_2_2 0.00522 0.02391 EQ (t-1)
AR2_2_1 0.00818 0.00642 1.27 0.2034 D_CDS (t-1)
AR2_2_2 -0.13617 0.05132 -2.65 0.0083 D_EQ (t-1)

114
GRANGER CAUSALITY TEST

Table 5.5
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 83.30 <0.0001
2 2 1.55 0.2125

Test 1: Group 1 Variable: CDS


Group 2 Variable: EQ
Test 2: Group 1 Variable: EQ
Group 2 Variable: CDS

UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 5.6
Variable R-Square Std. Deviation F Value Pr > F
CDS 0.3823 4.39872 66.68 < 0.0001
EQ 0.227 0.73664 2.50 0.0418

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 5.7
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 2.09803 185.35 < 0.0001 31.11 < 0.0001
EQ 2.01954 53.63 < 0.0001 0.44 0.5074

115
Appendix 10 – Before QUARTILE 1
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 57.622923 1.00000 | |********************| 0

1 57.557438 0.99886 | .|********************| 0.033131

2 57.479408 0.99751 | . |********************| 0.057342

3 57.381960 0.99582 | . |********************| 0.073977

4 57.266113 0.99381 | . |********************| 0.087462

5 57.137150 0.99157 | . |********************| 0.099085

6 56.997769 0.98915 | . |********************| 0.109437

7 56.852123 0.98662 | . |********************| 0.118846

8 56.705085 0.98407 | . |********************| 0.127521

9 56.548964 0.98136 | . |********************| 0.135601

10 56.387086 0.97855 | . |********************| 0.143185

11 56.216199 0.97559 | . |********************| 0.150346

12 56.038786 0.97251 | . |******************* | 0.157142

13 55.854393 0.96931 | . |******************* | 0.163615

14 55.664350 0.96601 | . |******************* | 0.169802

15 55.470270 0.96264 | . |******************* | 0.175731

16 55.275271 0.95926 | . |******************* | 0.181427

17 55.074719 0.95578 | . |******************* | 0.186911

18 54.870129 0.95223 | . |******************* | 0.192201

19 54.661886 0.94861 | . |******************* | 0.197312

116
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

20 54.445591 0.94486 | . |******************* | 0.202256

21 54.222330 0.94099 | . |******************* | 0.207045

22 53.995442 0.93705 | . |******************* | 0.211687

23 53.764131 0.93303 | . |******************* | 0.216192

24 53.530324 0.92898 | . |******************* | 0.220568

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.4562 0.5796 -1.18 0.2176

1 -0.4887 0.5725 -1.00 0.2832

2 -0.5708 0.5549 -0.89 0.3316

Single Mean 0 -1.0320 0.8832 -0.95 0.7733 0.86 0.8520

1 -1.3114 0.8557 -0.96 0.7705 0.71 0.8900

2 -1.9141 0.7888 -1.06 0.7339 0.71 0.8898

Trend 0 2.3518 0.9999 0.87 0.9999 1.39 0.9002

1 -0.6147 0.9923 -0.18 0.9934 0.48 0.9900

2 -6.6002 0.6959 -1.41 0.8564 1.15 0.9467

117
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.075135 1.00000 | |********************| 0

1 0.016387 0.21810 | .|**** | 0.033150

2 0.022157 0.29490 | .|****** | 0.034691

3 0.016543 0.22017 | .|**** | 0.037344

4 0.012432 0.16547 | . |*** | 0.038744

5 0.0090601 0.12058 | . |** | 0.039513

6 0.0068730 0.09148 | . |** | 0.039916

7 0.00075800 0.01009 | .|. | 0.040145

8 0.0072796 0.09689 | . |** | 0.040148

9 0.0034725 0.04622 | . |*. | 0.040404

10 0.0072617 0.09665 | . |** | 0.040462

11 0.0067836 0.09029 | . |** | 0.040715

12 0.0027813 0.03702 | . |*. | 0.040935

13 0.0026112 0.03475 | . |*. | 0.040971

14 0.0043619 0.05805 | . |*. | 0.041004

15 -0.0011480 -.01528 | .|. | 0.041094

16 0.0018474 0.02459 | .|. | 0.041100

17 0.0028234 0.03758 | . |*. | 0.041116

18 0.0012360 0.01645 | .|. | 0.041154

19 0.0021596 0.02874 | . |*. | 0.041161

20 0.0039780 0.05295 | . |*. | 0.041183

21 0.00068264 0.00909 | .|. | 0.041258

22 0.0034520 0.04594 | . |*. | 0.041260

23 -0.0003352 -.00446 | .|. | 0.041317

118
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 0.0023984 0.03192 | . |*. | 0.041317

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -702.491 0.0001 -23.46 <.0001

1 -403.927 0.0001 -13.90 <.0001

2 -291.961 0.0001 -10.96 <.0001

Single Mean 0 -703.001 0.0001 -23.47 <.0001 275.33 0.0010

1 -404.502 0.0001 -13.90 <.0001 96.66 0.0010

2 -292.583 0.0001 -10.97 <.0001 60.16 0.0010

Trend 0 -704.721 0.0001 -23.48 <.0001 275.76 0.0010

1 -405.968 0.0001 -13.90 <.0001 96.72 0.0010

2 -293.989 0.0001 -10.97 <.0001 60.21 0.0010

119
EQUITY

STATIONARITY

Table 2.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 36.707337 1.00000 | |********************| 0

1 36.542282 0.99550 | .|********************| 0.033131

2 36.387543 0.99129 | . |********************| 0.057214

3 36.241705 0.98732 | . |********************| 0.073693

4 36.105953 0.98362 | . |********************| 0.087010

5 35.952470 0.97944 | . |********************| 0.098462

6 35.802976 0.97536 | . |********************| 0.108632

7 35.667137 0.97166 | . |******************* | 0.117853

8 35.526874 0.96784 | . |******************* | 0.126341

9 35.399237 0.96436 | . |******************* | 0.134233

10 35.261981 0.96062 | . |******************* | 0.141634

11 35.109832 0.95648 | . |******************* | 0.148614

12 34.958443 0.95236 | . |******************* | 0.155224

13 34.789330 0.94775 | . |******************* | 0.161511

14 34.628966 0.94338 | . |******************* | 0.167504

15 34.458199 0.93873 | . |******************* | 0.173238

16 34.290732 0.93417 | . |******************* | 0.178735

17 34.127654 0.92972 | . |******************* | 0.184016

18 33.963641 0.92525 | . |******************* | 0.189102

19 33.790426 0.92054 | . |****************** | 0.194008

20 33.622796 0.91597 | . |****************** | 0.198745

21 33.470163 0.91181 | . |****************** | 0.203326

120
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

22 33.323339 0.90781 | . |****************** | 0.207766

23 33.163778 0.90346 | . |****************** | 0.212075

24 33.001275 0.89904 | . |****************** | 0.216259

Table 2.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.4782 0.8016 2.27 0.9948

1 0.4723 0.8001 2.30 0.9952

2 0.4725 0.8002 2.43 0.9966

Single Mean 0 0.2132 0.9671 0.14 0.9689 2.59 0.4057

1 0.3387 0.9719 0.23 0.9747 2.64 0.3924

2 0.4447 0.9756 0.32 0.9795 2.94 0.3168

Trend 0 -15.8071 0.1577 -2.80 0.1967 4.34 0.3018

1 -15.5360 0.1660 -2.80 0.1993 4.41 0.2886

2 -14.1766 0.2136 -2.67 0.2504 4.12 0.3475

121
Table 2.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.088600 1.00000 | |********************| 0

1 -0.0023829 -.02689 | *|. | 0.033150

2 -0.0048581 -.05483 | *|. | 0.033174

3 0.0026914 0.03038 | .|* | 0.033273

4 -0.0017999 -.02031 | .|. | 0.033304

5 -0.0021442 -.02420 | .|. | 0.033317

6 -0.0045221 -.05104 | *|. | 0.033336

7 -0.0022781 -.02571 | *|. | 0.033422

8 -0.0010374 -.01171 | .|. | 0.033444

9 -0.0029780 -.03361 | *|. | 0.033448

10 0.0016653 0.01880 | .|. | 0.033486

11 0.00072911 0.00823 | .|. | 0.033497

12 0.00066351 0.00749 | .|. | 0.033499

13 -0.0029670 -.03349 | *|. | 0.033501

14 -0.0013188 -.01489 | .|. | 0.033538

15 -0.0002638 -.00298 | .|. | 0.033545

16 -0.0028332 -.03198 | *|. | 0.033546

17 -0.0015079 -.01702 | .|. | 0.033579

18 0.0025829 0.02915 | .|* | 0.033588

19 -0.0023909 -.02699 | *|. | 0.033616

20 0.00017487 0.00197 | .|. | 0.033640

21 0.00004099 0.00046 | .|. | 0.033640

22 0.00099962 0.01128 | .|. | 0.033640

23 0.0018744 0.02116 | .|. | 0.033644

122
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 0.0025673 0.02898 | .|* | 0.033659

Table 2.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -928.388 0.0001 -30.78 <.0001

1 -1024.96 0.0001 -22.58 <.0001

2 -923.042 0.0001 -17.33 <.0001

Single Mean 0 -933.489 0.0001 -30.93 <.0001 478.49 0.0010

1 -1042.89 0.0001 -22.77 <.0001 259.23 0.0010

2 -955.804 0.0001 -17.53 <.0001 153.66 0.0010

Trend 0 -934.328 0.0001 -30.95 <.0001 478.98 0.0010

1 -1045.89 0.0001 -22.79 <.0001 259.81 0.0010

2 -960.946 0.0001 -17.56 <.0001 154.19 0.0010

123
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0028 3.2791 15.34
1 1 0.0008 0.7206 3.84

VAR-MODEL

Table 3.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.00164 0.00835 0.2 0.8442 1
AR1_1_1 0.13462 0.03196 4.21 0.0001 CDS (t-1)
AR1_1_2 -0.11663 0.02795 -4.17 0.0001 EQ (t-1)
AR2_1_1 0.26608 0.03166 8.4 0.0001 CDS (t-2)
AR2_1_2 -0.18489 0.02816 -6.57 0.0001 EQ (t-2)
EQ 1 diff Constant2 0.02394 0.00993 2.41 0.0162 1
AR1_2_1 -0.04781 0.03802 -1.26 0.2089 CDS (t-1)
AR1_2_2 -0.02934 0.03324 -0.88 0.3778 EQ (t-1)
AR2_2_1 0.03802 0.03766 1.01 0.3131 CDS (t-2)
AR2_2_2 -0.06026 0.0335 -1.8 0.0724 EQ (t-2)

GRANGER CAUSALITY TEST

Table 3.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 58.79 <.0001
2 2 2.14 0.3422

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

124
UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.1696 0.25023 46.12 <.0001
EQ 1diff 0.0062 0.29767 1.41 0.2291

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 2.02082 2566.59 <.0001 76.09 <.0001
EQ 1diff 1.99746 361.87 <.0001 6.41 0.0115

125
Appendix 11 – Before QUARTILE 2
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 58.131649 1.00000 | |********************| 0

1 57.992438 0.99761 | .|********************| 0.033131

2 57.787800 0.99408 | . |********************| 0.057294

3 57.514645 0.98939 | . |********************| 0.073838

4 57.215964 0.98425 | . |********************| 0.087184

5 56.900067 0.97881 | . |********************| 0.098630

6 56.576019 0.97324 | . |******************* | 0.108771

7 56.251496 0.96766 | . |******************* | 0.117944

8 55.935505 0.96222 | . |******************* | 0.126358

9 55.632577 0.95701 | . |******************* | 0.134160

10 55.320910 0.95165 | . |******************* | 0.141456

11 55.012577 0.94634 | . |******************* | 0.148317

12 54.688307 0.94077 | . |******************* | 0.154803

13 54.366800 0.93524 | . |******************* | 0.160957

14 54.044530 0.92969 | . |******************* | 0.166815

15 53.734658 0.92436 | . |****************** | 0.172409

16 53.434546 0.91920 | . |****************** | 0.177766

17 53.140912 0.91415 | . |****************** | 0.182909

18 52.860290 0.90932 | . |****************** | 0.187857

19 52.582182 0.90454 | . |****************** | 0.192628

126
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

20 52.295847 0.89961 | . |****************** | 0.197235

21 52.008974 0.89468 | . |****************** | 0.201689

22 51.721435 0.88973 | . |****************** | 0.205999

23 51.438857 0.88487 | . |****************** | 0.210175

24 51.166272 0.88018 | . |****************** | 0.214226

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.1199 0.6556 -0.25 0.5980

1 -0.2172 0.6335 -0.35 0.5578

2 -0.3277 0.6084 -0.42 0.5316

Single Mean 0 -2.0490 0.7728 -1.01 0.7498 0.51 0.9494

1 -3.3038 0.6199 -1.29 0.6368 0.83 0.8580

2 -5.2888 0.4080 -1.62 0.4707 1.32 0.7336

Trend 0 -4.1046 0.8802 -0.97 0.9459 0.67 0.9900

1 -10.2519 0.4185 -1.89 0.6619 1.88 0.8009

2 -22.2219 0.0433 -3.10 0.1078 4.82 0.2063

127
Table 1.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.261871 1.00000 | |********************| 0

1 0.058481 0.22332 | .|**** | 0.033150

2 0.069177 0.26416 | .|***** | 0.034764

3 0.029750 0.11361 | .|** | 0.036904

4 0.018803 0.07180 | .|* | 0.037286

5 0.0095151 0.03633 | .|* | 0.037438

6 -0.0031061 -.01186 | .|. | 0.037476

7 -0.0046284 -.01767 | .|. | 0.037480

8 -0.0067546 -.02579 | *|. | 0.037490

9 0.0096702 0.03693 | . |*. | 0.037509

10 0.0063721 0.02433 | .|. | 0.037549

11 0.016801 0.06416 | . |*. | 0.037566

12 -0.0020002 -.00764 | .|. | 0.037687

13 0.0021105 0.00806 | .|. | 0.037688

14 -0.011624 -.04439 | .*| . | 0.037690

15 -0.0099650 -.03805 | .*| . | 0.037748

16 -0.0072730 -.02777 | .*| . | 0.037790

17 -0.012749 -.04868 | .*| . | 0.037812

18 -0.0035827 -.01368 | .|. | 0.037881

19 0.0059590 0.02276 | .|. | 0.037886

20 -0.0015970 -.00610 | .|. | 0.037901

21 0.0023505 0.00898 | .|. | 0.037902

22 -0.0025057 -.00957 | .|. | 0.037905

23 -0.010944 -.04179 | .*| . | 0.037907

128
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 -0.0051311 -.01959 | .|. | 0.037958

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -701.815 0.0001 -23.64 <.0001

1 -437.399 0.0001 -14.57 <.0001

2 -421.757 0.0001 -12.91 <.0001

Single Mean 0 -701.809 0.0001 -23.63 <.0001 279.09 0.0010

1 -437.391 0.0001 -14.57 <.0001 106.08 0.0010

2 -421.738 0.0001 -12.90 <.0001 83.26 0.0010

Trend 0 -702.198 0.0001 -23.62 <.0001 278.93 0.0010

1 -437.682 0.0001 -14.55 <.0001 105.98 0.0010

2 -422.257 0.0001 -12.90 <.0001 83.21 0.0010

129
EQUITY

STATIONARITY

Table 2.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 27.489382 1.00000 | |********************| 0

1 27.367784 0.99558 | .|********************| 0.033131

2 27.249910 0.99129 | . |********************| 0.057216

3 27.137271 0.98719 | . |********************| 0.073695

4 27.028225 0.98322 | . |********************| 0.087009

5 26.910082 0.97893 | . |********************| 0.098452

6 26.796934 0.97481 | . |******************* | 0.108613

7 26.688956 0.97088 | . |******************* | 0.117826

8 26.576082 0.96678 | . |******************* | 0.126302

9 26.475172 0.96311 | . |******************* | 0.134180

10 26.367517 0.95919 | . |******************* | 0.141565

11 26.242506 0.95464 | . |******************* | 0.148528

12 26.114239 0.94998 | . |******************* | 0.155117

13 25.977882 0.94502 | . |******************* | 0.161377

14 25.847144 0.94026 | . |******************* | 0.167341

15 25.711479 0.93532 | . |******************* | 0.173043

16 25.574365 0.93034 | . |******************* | 0.178506

17 25.437217 0.92535 | . |******************* | 0.183752

18 25.300721 0.92038 | . |****************** | 0.188798

19 25.154705 0.91507 | . |****************** | 0.193660

20 25.016231 0.91003 | . |****************** | 0.198350

21 24.884236 0.90523 | . |****************** | 0.202881

130
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

22 24.754464 0.90051 | . |****************** | 0.207267

23 24.617382 0.89552 | . |****************** | 0.211518

24 24.474929 0.89034 | . |****************** | 0.215640

Table 2.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.3578 0.7706 1.59 0.9730

1 0.3550 0.7699 1.56 0.9716

2 0.3559 0.7701 1.62 0.9750

Single Mean 0 -1.3383 0.8530 -0.74 0.8357 1.70 0.6348

1 -1.3286 0.8540 -0.72 0.8386 1.65 0.6479

2 -1.2182 0.8652 -0.69 0.8477 1.72 0.6300

Trend 0 -20.7731 0.0586 -3.26 0.0738 5.33 0.1010

1 -21.5196 0.0502 -3.30 0.0666 5.48 0.0924

2 -20.4307 0.0629 -3.20 0.0851 5.15 0.1390

131
Table 2.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.099690 1.00000 | |********************| 0

1 0.00051774 0.00519 | .|. | 0.033150

2 -0.0037019 -.03713 | *|. | 0.033151

3 0.0046007 0.04615 | .|* | 0.033196

4 -0.0044189 -.04433 | *|. | 0.033267

5 -0.0029831 -.02992 | *|. | 0.033332

6 -0.0023941 -.02402 | .|. | 0.033361

7 -0.0025457 -.02554 | *|. | 0.033380

8 -0.0039311 -.03943 | *|. | 0.033401

9 -0.0028841 -.02893 | *|. | 0.033453

10 0.0052822 0.05299 | .|* | 0.033480

11 0.0014320 0.01436 | .|. | 0.033572

12 -0.0031026 -.03112 | *|. | 0.033579

13 -0.0004311 -.00432 | .|. | 0.033611

14 -0.0028909 -.02900 | *|. | 0.033611

15 -0.0000414 -.00042 | .|. | 0.033639

16 0.0029406 0.02950 | .|* | 0.033639

17 -0.0013610 -.01365 | .|. | 0.033667

18 0.0025747 0.02583 | .|* | 0.033673

19 -0.0044295 -.04443 | *|. | 0.033695

20 0.00080494 0.00807 | .|. | 0.033759

21 0.0013911 0.01395 | .|. | 0.033761

22 -0.0000714 -.00072 | .|. | 0.033768

23 0.0031187 0.03128 | .|* | 0.033768

24 0.0016737 0.01679 | .|. | 0.033800

132
Table 2.4 – 1. Difference
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -901.514 0.0001 -29.87 <.0001

1 -964.219 0.0001 -21.91 <.0001

2 -828.720 0.0001 -16.75 <.0001

Single Mean 0 -904.273 0.0001 -29.94 <.0001 448.26 0.0010

1 -973.142 0.0001 -22.00 <.0001 241.99 0.0010

2 -843.319 0.0001 -16.84 <.0001 141.86 0.0010

Trend 0 -904.301 0.0001 -29.93 <.0001 447.81 0.0010

1 -973.225 0.0001 -21.99 <.0001 241.76 0.0010

2 -843.355 0.0001 -16.84 <.0001 141.72 0.0010

133
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0064 6.2109 15.34
1 1 0.0004 0.4086 3.84

VAR-MODEL

Table 3.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.00994 0.0156 0.64 0.524 1
AR1_1_1 0.12751 0.03212 3.97 0.0001 CDS (t-1)
AR1_1_2 -0.29224 0.04941 -5.92 0.0001 EQ (t-1)
AR2_1_1 0.21824 0.03151 6.93 0.0001 CDS (t-2)
AR2_1_2 -0.30546 0.0503 -6.07 0.0001 EQ (t-2)
EQ 1 diff Constant2 0.01842 0.01052 1.75 0.0803 1
AR1_2_1 -0.04149 0.02166 -1.92 0.0557 CDS (t-1)
AR1_2_2 0.0019 0.03332 0.06 0.9546 EQ (t-1)
AR2_2_1 0.01945 0.02125 0.92 0.3603 CDS (t-2)
AR2_2_2 -0.04827 0.03392 -1.42 0.1551 EQ (t-2)

GRANGER CAUSALITY TEST

Table 3.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 71.02 <.0001
2 2 3.94 0.1392

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

134
UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.1662 0.46859 45 <.0001
EQ 1diff 0.0058 0.31602 1.31 0.2659

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 1.97433 2299.12 <.0001 56.54 <.0001
EQ 1diff 1.99629 156.57 <.0001 5.5 0.0192

135
Appendix 12 – Before QUARTILE 3
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 187.833 1.00000 | |********************| 0

1 186.489 0.99285 | .|********************| 0.033131

2 184.379 0.98161 | . |********************| 0.057112

3 181.838 0.96808 | . |******************* | 0.073329

4 179.235 0.95423 | . |******************* | 0.086224

5 176.565 0.94001 | . |******************* | 0.097127

6 173.806 0.92532 | . |******************* | 0.106647

7 170.976 0.91026 | . |****************** | 0.115123

8 168.148 0.89520 | . |****************** | 0.122769

9 165.524 0.88123 | . |****************** | 0.129737

10 162.848 0.86699 | . |***************** | 0.136149

11 160.437 0.85415 | . |***************** | 0.142080

12 158.003 0.84119 | . |***************** | 0.147609

13 155.693 0.82889 | . |***************** | 0.152781

14 153.514 0.81729 | . |**************** | 0.157640

15 151.465 0.80638 | . |**************** | 0.162224

16 149.640 0.79666 | . |**************** | 0.166566

17 147.969 0.78777 | . |**************** | 0.170697

18 146.349 0.77914 | . |**************** | 0.174643

19 144.639 0.77004 | . |*************** | 0.178417

136
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

20 142.791 0.76020 | . |*************** | 0.182029

21 140.801 0.74961 | . |*************** | 0.185481

22 138.809 0.73900 | . |*************** | 0.188777

23 136.880 0.72873 | . |*************** | 0.191927

24 134.992 0.71868 | . |************** | 0.194940

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.3170 0.7601 0.55 0.8338

1 0.0415 0.6926 0.05 0.6985

2 -0.1309 0.6531 -0.13 0.6399

Single Mean 0 -0.7244 0.9101 -0.28 0.9252 0.23 0.9900

1 -5.4322 0.3950 -1.41 0.5792 1.06 0.7986

2 -8.9930 0.1712 -1.94 0.3152 1.92 0.5777

Trend 0 4.2921 0.9999 1.32 0.9999 3.22 0.5290

1 -4.1014 0.8804 -0.83 0.9612 1.08 0.9561

2 -11.1589 0.3615 -1.83 0.6909 2.01 0.7736

137
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 1.362111 1.00000 | |********************| 0

1 0.470926 0.34573 | .|******* | 0.033150

2 0.343909 0.25248 | .|***** | 0.036900

3 0.197703 0.14514 | . |*** | 0.038752

4 0.028154 0.02067 | .|. | 0.039345

5 0.099751 0.07323 | . |*. | 0.039357

6 0.023601 0.01733 | .|. | 0.039506

7 -0.021668 -.01591 | .|. | 0.039515

8 0.053119 0.03900 | . |*. | 0.039522

9 0.024140 0.01772 | .|. | 0.039564

10 0.126122 0.09259 | . |** | 0.039573

11 0.164670 0.12089 | . |** | 0.039810

12 0.072995 0.05359 | . |*. | 0.040211

13 -0.033927 -.02491 | .|. | 0.040290

14 -0.030500 -.02239 | .|. | 0.040307

15 -0.115883 -.08508 | **| . | 0.040320

16 -0.029198 -.02144 | .|. | 0.040517

17 -0.014600 -.01072 | .|. | 0.040530

18 0.068583 0.05035 | . |*. | 0.040533

19 0.106752 0.07837 | . |** | 0.040601

20 0.077583 0.05696 | . |*. | 0.040767

21 0.110918 0.08143 | . |** | 0.040855

22 0.041552 0.03051 | . |*. | 0.041033

23 -0.053038 -.03894 | .*| . | 0.041057

138
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 -0.0083713 -.00615 | .|. | 0.041098

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -573.763 0.0001 -19.76 <.0001

1 -410.687 0.0001 -13.66 <.0001

2 -403.518 0.0001 -12.26 <.0001

Single Mean 0 -574.088 0.0001 -19.75 <.0001 195.02 0.0010

1 -411.143 0.0001 -13.65 <.0001 93.24 0.0010

2 -404.063 0.0001 -12.25 <.0001 75.09 0.0010

Trend 0 -577.236 0.0001 -19.78 <.0001 195.83 0.0010

1 -414.594 0.0001 -13.66 <.0001 93.52 0.0010

2 -409.103 0.0001 -12.27 <.0001 75.42 0.0010

139
EQUITY

STATIONARITY

Table 2.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 24.868061 1.00000 | |********************| 0

1 24.776822 0.99633 | .|********************| 0.033131

2 24.688969 0.99280 | . |********************| 0.057245

3 24.606202 0.98947 | . |********************| 0.073762

4 24.529055 0.98637 | . |********************| 0.087122

5 24.442425 0.98288 | . |********************| 0.098622

6 24.358196 0.97950 | . |********************| 0.108844

7 24.277099 0.97624 | . |********************| 0.118125

8 24.195330 0.97295 | . |******************* | 0.126672

9 24.119303 0.96989 | . |******************* | 0.134625

10 24.039717 0.96669 | . |******************* | 0.142088

11 23.948869 0.96304 | . |******************* | 0.149133

12 23.855602 0.95929 | . |******************* | 0.155810

13 23.751770 0.95511 | . |******************* | 0.162164

14 23.650381 0.95103 | . |******************* | 0.168225

15 23.544348 0.94677 | . |******************* | 0.174027

16 23.440252 0.94258 | . |******************* | 0.179592

17 23.335912 0.93839 | . |******************* | 0.184943

18 23.229661 0.93412 | . |******************* | 0.190097

19 23.115111 0.92951 | . |******************* | 0.195071

20 23.004149 0.92505 | . |******************* | 0.199874

21 22.898302 0.92079 | . |****************** | 0.204519

140
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

22 22.796213 0.91669 | . |****************** | 0.209020

23 22.683917 0.91217 | . |****************** | 0.213388

24 22.570019 0.90759 | . |****************** | 0.217626

Table 2.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.3149 0.7596 1.36 0.9565

1 0.3085 0.7580 1.31 0.9521

2 0.3104 0.7584 1.37 0.9580

Single Mean 0 -1.1488 0.8721 -0.64 0.8593 1.26 0.7495

1 -1.1466 0.8723 -0.63 0.8621 1.17 0.7705

2 -1.0164 0.8847 -0.58 0.8723 1.23 0.7549

Trend 0 -15.7683 0.1588 -2.92 0.1574 4.32 0.3074

1 -16.7532 0.1315 -3.01 0.1293 4.61 0.2469

2 -15.6324 0.1630 -2.91 0.1612 4.30 0.3107

141
Table 2.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.088051 1.00000 | |********************| 0

1 0.0013819 0.01569 | .|. | 0.033150

2 -0.0039912 -.04533 | *|. | 0.033158

3 0.0010748 0.01221 | .|. | 0.033226

4 -0.0005784 -.00657 | .|. | 0.033231

5 -0.0019226 -.02183 | .|. | 0.033232

6 -0.0025676 -.02916 | *|. | 0.033248

7 -0.0029907 -.03397 | *|. | 0.033276

8 -0.0010915 -.01240 | .|. | 0.033314

9 -0.0023782 -.02701 | *|. | 0.033319

10 0.0013283 0.01509 | .|. | 0.033343

11 0.0018369 0.02086 | .|. | 0.033351

12 0.0011156 0.01267 | .|. | 0.033365

13 0.00009915 0.00113 | .|. | 0.033370

14 -0.0022652 -.02573 | *|. | 0.033370

15 -0.0025901 -.02942 | *|. | 0.033392

16 0.00028505 0.00324 | .|. | 0.033421

17 -0.0004956 -.00563 | .|. | 0.033421

18 0.0027859 0.03164 | .|* | 0.033422

19 -0.0013389 -.01521 | .|. | 0.033455

20 0.00082030 0.00932 | .|. | 0.033463

21 -0.0005683 -.00645 | .|. | 0.033465

22 0.0018532 0.02105 | .|. | 0.033467

23 0.00011966 0.00136 | .|. | 0.033481

142
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 0.0020129 0.02286 | .|. | 0.033481

Table 2.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -892.739 0.0001 -29.58 <.0001

1 -972.858 0.0001 -22.00 <.0001

2 -925.739 0.0001 -17.39 <.0001

Single Mean 0 -894.705 0.0001 -29.63 <.0001 438.86 0.0010

1 -979.265 0.0001 -22.06 <.0001 243.38 0.0010

2 -937.757 0.0001 -17.46 <.0001 152.35 0.0010

Trend 0 -894.808 0.0001 -29.61 <.0001 438.52 0.0010

1 -979.593 0.0001 -22.06 <.0001 243.23 0.0010

2 -938.228 0.0001 -17.45 <.0001 152.29 0.0010

143
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0066 6.7961 15.34
1 1 0.0009 0.8238 3.84

VAR-MODEL

Table 3.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.032 0.03461 0.92 0.3554 1
AR1_1_1 0.24936 0.03372 7.4 0.0001 CDS (t-1)
AR1_1_2 -0.86163 0.11722 -7.35 0.0001 EQ (t-1)
AR2_1_1 0.15576 0.03264 4.77 0.0001 CDS (t-2)
AR2_1_2 -0.49192 0.12065 -4.08 0.0001 EQ (t-2)
EQ 1 diff Constant2 0.01553 0.00988 1.57 0.1165 1
AR1_2_1 -0.01922 0.00963 -2 0.0462 CDS (t-1)
AR1_2_2 0.0073 0.03347 0.22 0.8274 EQ (t-1)
AR2_2_1 -0.00057 0.00932 -0.06 0.951 CDS (t-2)
AR2_2_2 -0.06527 0.03445 -1.89 0.0585 EQ (t-2)

GRANGER CAUSALITY TEST

Table 3.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 69.48 <.0001
2 2 4.55 0.1029

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

144
UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.2092 1.03897 59.72 <.0001
EQ 1diff 0.0073 0.29668 1.67 0.1555

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 1.91223 9999.99 <.0001 77.72 <.0001
EQ 1diff 1.99806 205.59 <.0001 4.47 0.0348

145
Appendix 13 – Before QUARTILE 4
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 664.790 1.00000 | |********************| 0

1 657.743 0.98940 | .|********************| 0.033131

2 647.721 0.97432 | . |******************* | 0.056981

3 635.909 0.95656 | . |******************* | 0.073013

4 623.923 0.93853 | . |******************* | 0.085672

5 611.425 0.91973 | . |****************** | 0.096299

6 598.505 0.90029 | . |****************** | 0.105501

7 585.226 0.88032 | . |****************** | 0.113622

8 571.245 0.85929 | . |***************** | 0.120877

9 558.670 0.84037 | . |***************** | 0.127406

10 546.172 0.82157 | . |**************** | 0.133352

11 534.967 0.80472 | . |**************** | 0.138797

12 524.229 0.78856 | . |**************** | 0.143827

13 514.428 0.77382 | . |*************** | 0.148497

14 504.733 0.75924 | . |*************** | 0.152859

15 495.680 0.74562 | . |*************** | 0.156944

16 487.128 0.73276 | . |*************** | 0.160786

17 479.104 0.72069 | . |************** | 0.164410

18 471.448 0.70917 | . |************** | 0.167842

19 463.201 0.69676 | . |************** | 0.171100

146
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

20 454.589 0.68381 | . |************** | 0.174187

21 445.600 0.67029 | . |************* | 0.177109

22 437.099 0.65750 | . |************* | 0.179872

23 429.269 0.64572 | . |************* | 0.182491

24 421.268 0.63369 | . |************* | 0.184982

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.6141 0.8353 1.18 0.9397

1 0.4947 0.8058 0.75 0.8764

2 0.3758 0.7753 0.47 0.8172

Single Mean 0 -0.9251 0.8930 -0.30 0.9231 0.82 0.8605

1 -4.5370 0.4813 -1.13 0.7047 1.10 0.7904

2 -9.1416 0.1653 -1.86 0.3512 2.04 0.5469

Trend 0 -0.9157 0.9893 -0.29 0.9908 0.67 0.9900

1 -4.4972 0.8553 -1.12 0.9231 0.94 0.9753

2 -9.0824 0.5009 -1.85 0.6799 1.84 0.8098

147
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 7.079433 1.00000 | |********************| 0

1 1.548419 0.21872 | .|**** | 0.033150

2 1.505458 0.21265 | .|**** | 0.034699

3 0.837773 0.11834 | .|** | 0.036103

4 0.660591 0.09331 | .|** | 0.036527

5 0.555856 0.07852 | .|** | 0.036788

6 0.627592 0.08865 | .|** | 0.036972

7 -0.119124 -.01683 | .|. | 0.037204

8 0.0085693 0.00121 | .|. | 0.037213

9 0.0065113 0.00092 | .|. | 0.037213

10 0.145513 0.02055 | .|. | 0.037213

11 0.348424 0.04922 | .|* | 0.037225

12 0.352045 0.04973 | .|* | 0.037297

13 0.063171 0.00892 | .|. | 0.037369

14 0.101319 0.01431 | .|. | 0.037372

15 -0.325409 -.04597 | *|. | 0.037378

16 0.168053 0.02374 | .|. | 0.037440

17 -0.039499 -.00558 | .|. | 0.037456

18 0.412688 0.05829 | .|* | 0.037457

19 0.289293 0.04086 | . |*. | 0.037557

20 0.227836 0.03218 | . |*. | 0.037606

21 0.255123 0.03604 | . |*. | 0.037636

22 0.137029 0.01936 | .|. | 0.037674

23 0.071082 0.01004 | .|. | 0.037685

148
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 0.086056 0.01216 | .|. | 0.037688

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -699.779 0.0001 -23.30 <.0001

1 -483.618 0.0001 -15.14 <.0001

2 -435.221 0.0001 -12.94 <.0001

Single Mean 0 -701.192 0.0001 -23.32 <.0001 271.84 0.0010

1 -485.674 0.0001 -15.15 <.0001 114.83 0.0010

2 -437.862 0.0001 -12.96 <.0001 84.02 0.0010

Trend 0 -702.374 0.0001 -23.32 <.0001 272.01 0.0010

1 -486.998 0.0001 -15.15 <.0001 114.84 0.0010

2 -439.453 0.0001 -12.96 <.0001 84.04 0.0010

149
EQUITY

STATIONARITY

Table 2.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 22.997601 1.00000 | |********************| 0

1 22.879387 0.99486 | .|********************| 0.033131

2 22.768922 0.99006 | . |********************| 0.057189

3 22.666372 0.98560 | . |********************| 0.073638

4 22.569606 0.98139 | . |********************| 0.086920

5 22.464574 0.97682 | . |********************| 0.098334

6 22.359414 0.97225 | . |******************* | 0.108464

7 22.267532 0.96825 | . |******************* | 0.117642

8 22.168624 0.96395 | . |******************* | 0.126087

9 22.076954 0.95997 | . |******************* | 0.133932

10 21.983409 0.95590 | . |******************* | 0.141283

11 21.876005 0.95123 | . |******************* | 0.148213

12 21.768203 0.94654 | . |******************* | 0.154769

13 21.654798 0.94161 | . |******************* | 0.160998

14 21.547115 0.93693 | . |******************* | 0.166934

15 21.435233 0.93206 | . |******************* | 0.172610

16 21.327738 0.92739 | . |******************* | 0.178049

17 21.222213 0.92280 | . |****************** | 0.183274

18 21.115260 0.91815 | . |****************** | 0.188305

19 20.997159 0.91302 | . |****************** | 0.193157

20 20.883466 0.90807 | . |****************** | 0.197838

21 20.779830 0.90357 | . |****************** | 0.202361

150
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

22 20.678419 0.89916 | . |****************** | 0.206742

23 20.572181 0.89454 | . |****************** | 0.210991

24 20.461649 0.88973 | . |****************** | 0.215114

Table 2.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.2683 0.7477 1.19 0.9406

1 0.2598 0.7455 1.15 0.9363

2 0.2593 0.7454 1.20 0.9415

Single Mean 0 -3.0120 0.6552 -1.36 0.6040 1.82 0.6049

1 -2.8640 0.6733 -1.29 0.6371 1.66 0.6443

2 -2.6504 0.6996 -1.24 0.6583 1.66 0.6452

Trend 0 -30.9224 0.0064 -3.90 0.0124 7.66 0.0142

1 -32.0866 0.0049 -3.94 0.0110 7.79 0.0114

2 -30.3798 0.0072 -3.79 0.0174 7.22 0.0237

151
Table 2.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.123939 1.00000 | |********************| 0

1 0.00004643 0.00037 | .|. | 0.033150

2 -0.0053003 -.04277 | *|. | 0.033150

3 0.0027917 0.02252 | .|. | 0.033210

4 -0.0030114 -.02430 | .|. | 0.033227

5 -0.0008024 -.00647 | .|. | 0.033247

6 -0.0086840 -.07007 | *|. | 0.033248

7 0.00035273 0.00285 | .|. | 0.033410

8 -0.0014019 -.01131 | .|. | 0.033410

9 -0.0034759 -.02805 | *|. | 0.033414

10 0.0028205 0.02276 | .|. | 0.033440

11 0.0010877 0.00878 | .|. | 0.033457

12 -0.0019008 -.01534 | .|. | 0.033460

13 -0.0012112 -.00977 | .|. | 0.033467

14 -0.0019089 -.01540 | .|. | 0.033471

15 -0.0027362 -.02208 | .|. | 0.033478

16 -0.0005091 -.00411 | .|. | 0.033494

17 -0.0024396 -.01968 | .|. | 0.033495

18 0.0044506 0.03591 | .|* | 0.033508

19 -0.0038225 -.03084 | *|. | 0.033550

20 -0.0016914 -.01365 | .|. | 0.033581

21 0.0013734 0.01108 | .|. | 0.033587

22 -0.0016598 -.01339 | .|. | 0.033591

23 0.0024897 0.02009 | .|. | 0.033597

152
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 0.0024644 0.01988 | .|. | 0.033610

Table 2.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -907.002 0.0001 -30.05 <.0001

1 -984.114 0.0001 -22.13 <.0001

2 -912.975 0.0001 -17.28 <.0001

Single Mean 0 -908.661 0.0001 -30.09 <.0001 452.80 0.0010

1 -989.504 0.0001 -22.18 <.0001 245.91 0.0010

2 -922.598 0.0001 -17.33 <.0001 150.17 0.0010

Trend 0 -908.746 0.0001 -30.08 <.0001 452.35 0.0010

1 -989.782 0.0001 -22.17 <.0001 245.67 0.0010

2 -923.171 0.0001 -17.32 <.0001 150.03 0.0010

153
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0055 5.6597 15.34
1 1 0.0007 0.6602 3.84

VAR-MODEL

Table 3.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.10817 0.0834 1.3 0.195 1
AR1_1_1 0.14223 0.03349 4.25 0.0001 CDS (t-1)
AR1_1_2 -1.30794 0.23853 -5.48 0.0001 EQ (t-1)
AR2_1_1 0.16223 0.03286 4.94 0.0001 CDS (t-2)
AR2_1_2 -0.92861 0.24212 -3.84 0.0001 EQ (t-2)
EQ 1 diff Constant2 0.01916 0.01166 1.64 0.1008 1
AR1_2_1 -0.0078 0.00468 -1.66 0.0964 CDS (t-1)
AR1_2_2 -0.01506 0.03336 -0.45 0.6518 EQ (t-1)
AR2_2_1 -0.01471 0.0046 -3.2 0.0014 CDS (t-2)
AR2_2_2 -0.06957 0.03386 -2.05 0.0402 EQ (t-2)

GRANGER CAUSALITY TEST

Table 3.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 43.43 <.0001
2 2 15.67 0.0004

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

154
UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.1218 2.50058 31.3 <.0001
EQ 1diff 0.0189 0.34972 4.34 0.0018

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 1.95258 9999.99 <.0001 5.89 0.0154
EQ 1diff 1.98163 169.23 <.0001 4.79 0.0288

155
Appendix 14 – After QUARTILE 1
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 720.785 1.00000 | |********************| 0

1 718.164 0.99636 | . |********************| 0.047782

2 714.584 0.99140 | . |********************| 0.082560

3 710.372 0.98555 | . |********************| 0.106321

4 705.841 0.97927 | . |********************| 0.125457

5 701.353 0.97304 | . |******************* | 0.141839

6 696.735 0.96663 | . |******************* | 0.156338

7 692.003 0.96007 | . |******************* | 0.169435

8 687.150 0.95334 | . |******************* | 0.181430

9 682.172 0.94643 | . |******************* | 0.192528

10 677.031 0.93930 | . |******************* | 0.202872

11 671.679 0.93187 | . |******************* | 0.212569

12 666.158 0.92421 | . |****************** | 0.221700

13 660.594 0.91649 | . |****************** | 0.230329

14 655.078 0.90884 | . |****************** | 0.238509

15 649.533 0.90115 | . |****************** | 0.246289

16 643.895 0.89333 | . |****************** | 0.253705

17 638.048 0.88521 | . |****************** | 0.260788

18 632.069 0.87692 | . |****************** | 0.267560

19 626.034 0.86854 | . |***************** | 0.274043

156
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

20 619.969 0.86013 | . |***************** | 0.280258

21 614.069 0.85194 | . |***************** | 0.286221

22 608.495 0.84421 | . |***************** | 0.291953

23 603.008 0.83660 | . |***************** | 0.297474

24 597.576 0.82906 | . |***************** | 0.302799

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.4803 0.8017 0.89 0.9003

1 0.2754 0.7491 0.33 0.7805

2 0.1918 0.7282 0.21 0.7465

Single Mean 0 -1.2298 0.8638 -1.06 0.7319 1.79 0.6138

1 -2.1983 0.7544 -1.23 0.6631 1.28 0.7446

2 -2.5844 0.7071 -1.31 0.6290 1.27 0.7449

Trend 0 -1.3412 0.9834 -0.53 0.9821 0.56 0.9900

1 -5.7166 0.7654 -1.44 0.8476 1.25 0.9275

2 -7.4698 0.6238 -1.68 0.7612 1.59 0.8591

157
Table 1.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 2.212085 1.00000 | |********************| 0

1 0.898375 0.40612 | . |******** | 0.047782

2 0.544709 0.24624 | . |***** | 0.055102

3 0.291384 0.13172 | . |*** | 0.057560

4 -0.011674 -.00528 | .|. | 0.058244

5 0.102065 0.04614 | . |*. | 0.058245

6 0.025170 0.01138 | .|. | 0.058328

7 0.063267 0.02860 | . |*. | 0.058333

8 0.026385 0.01193 | .|. | 0.058365

9 0.122850 0.05554 | . |*. | 0.058371

10 0.221626 0.10019 | . |** | 0.058491

11 0.145449 0.06575 | . |*. | 0.058882

12 -0.010334 -.00467 | .|. | 0.059049

13 -0.084148 -.03804 | .*| . | 0.059050

14 -0.040344 -.01824 | .|. | 0.059106

15 0.144908 0.06551 | . |*. | 0.059119

16 0.159527 0.07212 | . |*. | 0.059284

17 0.140160 0.06336 | . |*. | 0.059484

18 0.017475 0.00790 | .|. | 0.059638

19 0.010406 0.00470 | .|. | 0.059641

20 -0.165366 -.07476 | .*| . | 0.059641

21 -0.247999 -.11211 | **| . | 0.059855

22 -0.118002 -.05334 | .*| . | 0.060333

23 -0.072118 -.03260 | .*| . | 0.060440

158
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 0.117926 0.05331 | . |*. | 0.060480

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -258.006 0.0001 -13.52 <.0001

1 -210.437 0.0001 -10.23 <.0001

2 -207.368 0.0001 -9.10 <.0001

Single Mean 0 -259.448 0.0001 -13.56 <.0001 91.91 0.0010

1 -212.581 0.0001 -10.27 <.0001 52.71 0.0010

2 -210.603 0.0001 -9.15 <.0001 41.84 0.0010

Trend 0 -259.934 0.0001 -13.56 <.0001 91.99 0.0010

1 -213.279 0.0001 -10.28 <.0001 52.82 0.0010

2 -211.575 0.0001 -9.16 <.0001 41.97 0.0010

159
EQUITY

STATIONARITY

Table 2.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 78.073690 1.00000 | |********************| 0

1 77.515844 0.99285 | . |********************| 0.047782

2 77.057624 0.98699 | . |********************| 0.082367

3 76.668905 0.98201 | . |********************| 0.105983

4 76.216812 0.97622 | . |********************| 0.125043

5 75.794785 0.97081 | . |******************* | 0.141377

6 75.383901 0.96555 | . |******************* | 0.155855

7 74.926866 0.95969 | . |******************* | 0.168961

8 74.468740 0.95383 | . |******************* | 0.180979

9 73.958577 0.94729 | . |******************* | 0.192114

10 73.439688 0.94065 | . |******************* | 0.202498

11 72.941584 0.93427 | . |******************* | 0.212239

12 72.406197 0.92741 | . |******************* | 0.221430

13 71.889793 0.92079 | . |****************** | 0.230127

14 71.303705 0.91329 | . |****************** | 0.238390

15 70.762894 0.90636 | . |****************** | 0.246249

16 70.257293 0.89988 | . |****************** | 0.253751

17 69.687850 0.89259 | . |****************** | 0.260936

18 69.077973 0.88478 | . |****************** | 0.267816

19 68.552085 0.87804 | . |****************** | 0.274408

20 68.000287 0.87098 | . |***************** | 0.280750

21 67.417485 0.86351 | . |***************** | 0.286852

160
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

22 66.886933 0.85672 | . |***************** | 0.292727

23 66.317129 0.84942 | . |***************** | 0.298397

24 65.742359 0.84206 | . |***************** | 0.303867

Table 2.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.4768 0.5749 -1.28 0.1838

1 -0.4606 0.5784 -1.46 0.1355

2 -0.4505 0.5807 -1.64 0.0961

Single Mean 0 -1.4963 0.8358 -0.78 0.8247 0.97 0.8234

1 -0.9897 0.8869 -0.60 0.8673 1.12 0.7859

2 -0.6674 0.9144 -0.47 0.8947 1.35 0.7258

Trend 0 -12.5684 0.2818 -2.62 0.2717 3.47 0.4801

1 -9.4943 0.4685 -2.29 0.4399 2.67 0.6422

2 -7.4715 0.6237 -2.05 0.5743 2.16 0.7442

161
Table 2.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.665186 1.00000 | |********************| 0

1 -0.108267 -.16276 | ***| . | 0.047782

2 -0.073244 -.11011 | **| . | 0.049031

3 0.068990 0.10372 | . |** | 0.049593

4 -0.0085352 -.01283 | .|. | 0.050085

5 -0.0055005 -.00827 | .|. | 0.050093

6 0.0089651 0.01348 | .|. | 0.050096

7 -0.019225 -.02890 | .*| . | 0.050104

8 0.057696 0.08674 | . |** | 0.050142

9 -0.024665 -.03708 | .*| . | 0.050484

10 0.015689 0.02359 | .|. | 0.050546

11 0.013558 0.02038 | .|. | 0.050571

12 -0.0021581 -.00324 | .|. | 0.050590

13 0.033582 0.05049 | . |*. | 0.050590

14 -0.026014 -.03911 | .*| . | 0.050705

15 -0.068073 -.10234 | **| . | 0.050774

16 0.049831 0.07491 | . |*. | 0.051243

17 0.049151 0.07389 | . |*. | 0.051492

18 -0.073022 -.10978 | **| . | 0.051734

19 0.012603 0.01895 | .|. | 0.052263

20 0.029424 0.04423 | . |*. | 0.052278

21 -0.059376 -.08926 | **| . | 0.052364

22 0.013100 0.01969 | .|. | 0.052710

23 0.024582 0.03696 | . |*. | 0.052727

162
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 -0.039572 -.05949 | .*| . | 0.052786

Table 2.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -506.454 0.0001 -24.59 <.0001

1 -663.648 0.0001 -18.21 <.0001

2 -525.455 0.0001 -12.76 <.0001

Single Mean 0 -508.137 0.0001 -24.66 <.0001 303.94 0.0010

1 -672.839 0.0001 -18.31 <.0001 167.54 0.0010

2 -541.743 0.0001 -12.86 <.0001 82.72 0.0010

Trend 0 -508.193 0.0001 -24.63 <.0001 303.36 0.0010

1 -673.237 0.0001 -18.29 <.0001 167.27 0.0010

2 -542.666 0.0001 -12.85 <.0001 82.61 0.0010

163
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0185 8.2601 15.34
1 1 0.0003 0.1347 3.84

VAR-MODEL

Table 3.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.0269 0.06244 0.43 0.6669 1
AR1_1_1 0.28228 0.04771 5.92 0.0001 CDS (t-1)
AR1_1_2 -0.49543 0.07867 -6.3 0.0001 EQ (t-1)
AR2_1_1 0.14359 0.04609 3.12 0.002 CDS (t-2)
AR2_1_2 -0.30462 0.07997 -3.81 0.0002 EQ (t-2)
EQ 1 diff Constant2 -0.06266 0.03858 -1.62 0.1051 1
AR1_2_1 0.0242 0.02948 0.82 0.4122 CDS (t-1)
AR1_2_2 -0.18031 0.04861 -3.71 0.0002 EQ (t-1)
AR2_2_1 0.00023 0.02848 0.01 0.9937 CDS (t-2)
AR2_2_2 -0.1283 0.04942 -2.6 0.0097 EQ (t-2)

GRANGER CAUSALITY TEST

Table 3.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 46.52 <.0001
2 2 0.82 0.6635

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

164
UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.2536 1.29441 36.61 <.0001
EQ 1diff 0.0477 0.79985 5.4 0.0003

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 2.05309 1245.59 <.0001 7.76 0.0056
EQ 1diff 1.98008 66.74 <.0001 4.32 0.0382

165
Appendix 15 – After Quartile 2
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 2667.923 1.00000 | |********************| 0

1 2657.065 0.99593 | . |********************| 0.047782

2 2641.890 0.99024 | . |********************| 0.082536

3 2624.993 0.98391 | . |********************| 0.106253

4 2607.474 0.97734 | . |********************| 0.125340

5 2590.371 0.97093 | . |******************* | 0.141675

6 2573.173 0.96449 | . |******************* | 0.156130

7 2555.968 0.95804 | . |******************* | 0.169186

8 2537.808 0.95123 | . |******************* | 0.181149

9 2519.580 0.94440 | . |******************* | 0.192215

10 2501.020 0.93744 | . |******************* | 0.202532

11 2481.215 0.93002 | . |******************* | 0.212208

12 2460.674 0.92232 | . |****************** | 0.221318

13 2440.377 0.91471 | . |****************** | 0.229926

14 2420.496 0.90726 | . |****************** | 0.238089

15 2400.351 0.89971 | . |****************** | 0.245855

16 2379.280 0.89181 | . |****************** | 0.253261

17 2357.317 0.88358 | . |****************** | 0.260332

18 2334.808 0.87514 | . |****************** | 0.267091

19 2312.443 0.86676 | . |***************** | 0.273559

166
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

20 2290.088 0.85838 | . |***************** | 0.279759

21 2268.055 0.85012 | . |***************** | 0.285709

22 2246.979 0.84222 | . |***************** | 0.291427

23 2226.165 0.83442 | . |***************** | 0.296932

24 2205.505 0.82667 | . |***************** | 0.302238

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.4968 0.8059 0.79 0.8835

1 0.2084 0.7323 0.21 0.7483

2 0.2469 0.7419 0.26 0.7621

Single Mean 0 -1.2933 0.8573 -1.01 0.7510 1.59 0.6641

1 -2.4729 0.7208 -1.24 0.6574 1.22 0.7602

2 -2.3387 0.7373 -1.22 0.6686 1.22 0.7580

Trend 0 -2.8158 0.9441 -0.93 0.9501 0.66 0.9900

1 -9.1173 0.4959 -1.93 0.6385 1.97 0.7830

2 -8.4823 0.5438 -1.83 0.6875 1.80 0.8170

167
Table 1.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 10.046357 1.00000 | |********************| 0

1 4.106206 0.40873 | . |******** | 0.047782

2 1.347605 0.13414 | . |*** | 0.055190

3 0.314194 0.03127 | . |*. | 0.055929

4 -0.290336 -.02890 | .*| . | 0.055969

5 -0.182697 -.01819 | .|. | 0.056003

6 -0.387888 -.03861 | .*| . | 0.056017

7 0.625390 0.06225 | . |*. | 0.056077

8 -0.157582 -.01569 | .|. | 0.056235

9 0.435025 0.04330 | . |*. | 0.056245

10 1.348101 0.13419 | . |*** | 0.056321

11 0.294511 0.02932 | . |*. | 0.057046

12 -0.506123 -.05038 | .*| . | 0.057081

13 -0.751542 -.07481 | .*| . | 0.057182

14 -0.222091 -.02211 | .|. | 0.057405

15 0.644425 0.06415 | . |*. | 0.057425

16 0.610202 0.06074 | . |*. | 0.057588

17 0.546611 0.05441 | . |*. | 0.057734

18 -0.377655 -.03759 | .*| . | 0.057851

19 -0.249050 -.02479 | .|. | 0.057907

20 -0.373962 -.03722 | .*| . | 0.057931

21 -0.593893 -.05912 | .*| . | 0.057985

22 -0.227774 -.02267 | .|. | 0.058123

23 -0.065339 -.00650 | .|. | 0.058143

168
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 0.387716 0.03859 | . |*. | 0.058145

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -256.964 0.0001 -13.48 <.0001

1 -275.884 0.0001 -11.71 <.0001

2 -281.750 0.0001 -10.26 <.0001

Single Mean 0 -258.236 0.0001 -13.51 <.0001 91.30 0.0010

1 -278.509 0.0001 -11.75 <.0001 68.98 0.0010

2 -286.367 0.0001 -10.30 <.0001 53.08 0.0010

Trend 0 -258.483 0.0001 -13.51 <.0001 91.26 0.0010

1 -278.929 0.0001 -11.74 <.0001 68.97 0.0010

2 -287.021 0.0001 -10.30 <.0001 53.10 0.0010

169
EQUITY

STATIONARITY

Table 2.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 71.238802 1.00000 | |********************| 0

1 70.760415 0.99328 | . |********************| 0.047782

2 70.363076 0.98771 | . |********************| 0.082390

3 70.003476 0.98266 | . |********************| 0.106032

4 69.598432 0.97697 | . |********************| 0.125108

5 69.220291 0.97167 | . |******************* | 0.141458

6 68.860132 0.96661 | . |******************* | 0.155954

7 68.462646 0.96103 | . |******************* | 0.169080

8 68.084195 0.95572 | . |******************* | 0.181122

9 67.648079 0.94960 | . |******************* | 0.192291

10 67.179778 0.94302 | . |******************* | 0.202715

11 66.743088 0.93689 | . |******************* | 0.212495

12 66.265468 0.93019 | . |******************* | 0.221725

13 65.814966 0.92386 | . |****************** | 0.230463

14 65.304258 0.91670 | . |****************** | 0.238769

15 64.840097 0.91018 | . |****************** | 0.246673

16 64.403365 0.90405 | . |****************** | 0.254225

17 63.913183 0.89717 | . |****************** | 0.261462

18 63.390413 0.88983 | . |****************** | 0.268398

19 62.951330 0.88367 | . |****************** | 0.275051

20 62.480563 0.87706 | . |****************** | 0.281458

170
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

21 61.986354 0.87012 | . |***************** | 0.287630

22 61.543600 0.86391 | . |***************** | 0.293579

23 61.056468 0.85707 | . |***************** | 0.299326

24 60.586400 0.85047 | . |***************** | 0.304878

Table 2.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.4199 0.5874 -1.09 0.2497

1 -0.4111 0.5894 -1.23 0.2027

2 -0.4076 0.5902 -1.33 0.1715

Single Mean 0 -1.5434 0.8307 -0.77 0.8248 0.76 0.8773

1 -1.0951 0.8770 -0.63 0.8610 0.83 0.8591

2 -0.8699 0.8976 -0.55 0.8791 0.92 0.8357

Trend 0 -12.2504 0.2979 -2.66 0.2558 3.61 0.4534

1 -9.6731 0.4559 -2.37 0.3968 2.89 0.5987

2 -8.3425 0.5546 -2.20 0.4873 2.51 0.6735

171
Table 2.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.649010 1.00000 | |********************| 0

1 -0.091046 -.14029 | ***| . | 0.047782

2 -0.044911 -.06920 | .*| . | 0.048713

3 0.051756 0.07975 | . |** | 0.048937

4 -0.010332 -.01592 | .|. | 0.049233

5 -0.011923 -.01837 | .|. | 0.049245

6 0.0030717 0.00473 | .|. | 0.049260

7 -0.021900 -.03374 | .*| . | 0.049261

8 0.055893 0.08612 | . |** | 0.049314

9 0.0049821 0.00768 | .|. | 0.049656

10 -0.0008081 -.00125 | .|. | 0.049659

11 0.021275 0.03278 | . |*. | 0.049659

12 -0.0073610 -.01134 | .|. | 0.049708

13 0.028697 0.04422 | . |*. | 0.049714

14 -0.033428 -.05151 | .*| . | 0.049804

15 -0.075301 -.11602 | **| . | 0.049925

16 0.040772 0.06282 | . |*. | 0.050537

17 0.042691 0.06578 | . |*. | 0.050715

18 -0.062306 -.09600 | **| . | 0.050910

19 0.018323 0.02823 | . |*. | 0.051321

20 0.015964 0.02460 | .|. | 0.051357

21 -0.043234 -.06662 | .*| . | 0.051384

22 0.024873 0.03832 | . |*. | 0.051580

23 0.0013835 0.00213 | .|. | 0.051645

172
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 -0.017504 -.02697 | .*| . | 0.051646

Table 2.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -497.142 0.0001 -24.10 <.0001

1 -590.283 0.0001 -17.17 <.0001

2 -481.288 0.0001 -12.41 <.0001

Single Mean 0 -498.306 0.0001 -24.14 <.0001 291.29 0.0010

1 -595.642 0.0001 -17.23 <.0001 148.40 0.0010

2 -490.657 0.0001 -12.47 <.0001 77.76 0.0010

Trend 0 -498.408 0.0001 -24.12 <.0001 290.83 0.0010

1 -596.215 0.0001 -17.22 <.0001 148.22 0.0010

2 -491.804 0.0001 -12.46 <.0001 77.69 0.0010

173
MODEL

CO-INTEGRATION

Table 5.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0341 15.5077 15.34
1 1 0.0008 0.3622 3.84

VECM-MODEL

Table 5.2 – Long-Run Parameter Beta Estimates When RANK = 1


Variable 1
CDS 0.06381
EQUITY 0.37507

Table 5.3 – Adjustment Coefficient Alpha Estimates When RANK = 1


Variable 1
CDS -0.51617
EQUITY 0.01512

Table 5.4
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS Constant1 11.45002 2.92259 3.92 0.0001 1
AR1_1_1 -0.03294 0.00848 CDS (t-1)
AR1_1_2 -0.1936 0.04983 EQ (t-1)
AR2_1_1 0.36738 0.04247 8.65 0.0001 D_CDS (t-1)
AR2_1_2 -0.66628 0.17202 -3.87 0.0001 D_EQ (t-1)
EQ Constant2 -0.37807 0.83665 -0.45 0.6516 1
AR1_2_1 0.00096 0.00243 CDS (t-1)
AR1_2_2 0.00567 0.01427 EQ (t-1)
AR2_2_1 0.01418 0.01216 1.17 0.244 D_CDS (t-1)
AR2_2_2 -0.13507 0.04924 -2.74 0.0063 D_EQ (t-1)

174
GRANGER CAUSALITY TEST

Table 5.5
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 37.43 <.0001
2 2 1.43 0.4895

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 5.6
Variable R-Square Std. Deviation F Value Pr > F
CDS 0.2361 2.77433 33.3 <.0001
EQ 0.0227 0.79421 2.5 0.0421

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 5.7
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 2.05964 855.32 <.0001 13.49 0.0003
EQ 2.02698 79.4 <.0001 1.15 0.2847

175
Appendix 16 – After QUARTILE 3
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 9193.773 1.00000 | |********************| 0

1 9167.349 0.99713 | . |********************| 0.047782

2 9126.475 0.99268 | . |********************| 0.082602

3 9080.588 0.98769 | . |********************| 0.106408

4 9031.673 0.98237 | . |********************| 0.125607

5 8981.707 0.97693 | . |********************| 0.142070

6 8929.362 0.97124 | . |******************* | 0.156658

7 8874.669 0.96529 | . |******************* | 0.169850

8 8820.258 0.95937 | . |******************* | 0.181945

9 8766.415 0.95352 | . |******************* | 0.193149

10 8710.694 0.94746 | . |******************* | 0.203613

11 8650.613 0.94092 | . |******************* | 0.213441

12 8587.522 0.93406 | . |******************* | 0.222710

13 8523.396 0.92708 | . |******************* | 0.231481

14 8460.250 0.92022 | . |****************** | 0.239809

15 8397.657 0.91341 | . |****************** | 0.247739

16 8332.293 0.90630 | . |****************** | 0.255312

17 8263.938 0.89886 | . |****************** | 0.262555

18 8193.826 0.89124 | . |****************** | 0.269489

19 8124.197 0.88366 | . |****************** | 0.276136

176
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

20 8054.882 0.87612 | . |****************** | 0.282519

21 7986.462 0.86868 | . |***************** | 0.288655

22 7919.980 0.86145 | . |***************** | 0.294563

23 7854.361 0.85431 | . |***************** | 0.300260

24 7788.701 0.84717 | . |***************** | 0.305759

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.3119 0.7585 0.54 0.8321

1 0.0144 0.6860 0.02 0.6877

2 0.1067 0.7076 0.12 0.7216

Single Mean 0 -1.0958 0.8769 -0.90 0.7870 1.02 0.8110

1 -2.3500 0.7359 -1.18 0.6846 0.91 0.8375

2 -2.0046 0.7777 -1.11 0.7142 0.89 0.8440

Trend 0 -1.2249 0.9852 -0.47 0.9846 0.41 0.9900

1 -6.8493 0.6742 -1.58 0.7989 1.38 0.9022

2 -5.3685 0.7923 -1.35 0.8734 1.06 0.9590

177
Table 1.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 30.919984 1.00000 | |********************| 0

1 14.082069 0.45544 | . |********* | 0.047782

2 4.025087 0.13018 | . |*** | 0.056835

3 2.588448 0.08371 | . |** | 0.057512

4 1.153636 0.03731 | . |*. | 0.057789

5 2.040216 0.06598 | . |*. | 0.057844

6 1.061069 0.03432 | . |*. | 0.058016

7 -1.128977 -.03651 | .*| . | 0.058062

8 -1.470985 -.04757 | .*| . | 0.058115

9 1.628640 0.05267 | . |*. | 0.058204

10 4.349257 0.14066 | . |*** | 0.058312

11 2.151438 0.06958 | . |*. | 0.059082

12 0.475260 0.01537 | .|. | 0.059269

13 -2.008876 -.06497 | .*| . | 0.059278

14 -1.587709 -.05135 | .*| . | 0.059440

15 1.399204 0.04525 | . |*. | 0.059541

16 2.222676 0.07188 | . |*. | 0.059620

17 1.508954 0.04880 | . |*. | 0.059817

18 -0.910646 -.02945 | .*| . | 0.059908

19 -0.779712 -.02522 | .*| . | 0.059941

20 -0.979613 -.03168 | .*| . | 0.059965

21 -0.948482 -.03068 | .*| . | 0.060004

22 -0.613229 -.01983 | .|. | 0.060039

23 -0.028188 -.00091 | .|. | 0.060054

178
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 1.582776 0.05119 | . |*. | 0.060054

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -237.130 0.0001 -12.75 <.0001

1 -286.626 0.0001 -11.92 <.0001

2 -230.956 0.0001 -9.49 <.0001

Single Mean 0 -237.790 0.0001 -12.76 <.0001 81.43 0.0010

1 -288.157 0.0001 -11.94 <.0001 71.26 0.0010

2 -232.927 0.0001 -9.51 <.0001 45.21 0.0010

Trend 0 -238.098 0.0001 -12.76 <.0001 81.44 0.0010

1 -288.667 0.0001 -11.94 <.0001 71.28 0.0010

2 -233.587 0.0001 -9.51 <.0001 45.27 0.0010

179
EQUITY

STATIONARITY

Table 2.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 80.500786 1.00000 | |********************| 0

1 80.038245 0.99425 | . |********************| 0.047782

2 79.612465 0.98897 | . |********************| 0.082444

3 79.211675 0.98399 | . |********************| 0.106127

4 78.789013 0.97874 | . |********************| 0.125236

5 78.398010 0.97388 | . |******************* | 0.141627

6 78.011717 0.96908 | . |******************* | 0.156170

7 77.573768 0.96364 | . |******************* | 0.169343

8 77.131567 0.95815 | . |******************* | 0.181431

9 76.642697 0.95207 | . |******************* | 0.192638

10 76.115488 0.94552 | . |******************* | 0.203097

11 75.613381 0.93929 | . |******************* | 0.212910

12 75.065185 0.93248 | . |******************* | 0.222169

13 74.544006 0.92600 | . |******************* | 0.230932

14 73.976168 0.91895 | . |****************** | 0.239259

15 73.434504 0.91222 | . |****************** | 0.247186

16 72.917233 0.90580 | . |****************** | 0.254756

17 72.347945 0.89872 | . |****************** | 0.262006

18 71.736314 0.89113 | . |****************** | 0.268952

19 71.175922 0.88416 | . |****************** | 0.275611

20 70.599699 0.87701 | . |****************** | 0.282013

21 70.007597 0.86965 | . |***************** | 0.288172

180
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

22 69.440384 0.86261 | . |***************** | 0.294103

23 68.844195 0.85520 | . |***************** | 0.299824

24 68.266272 0.84802 | . |***************** | 0.305342

Table 2.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.6199 0.5444 -1.56 0.1128

1 -0.6128 0.5459 -1.71 0.0822

2 -0.6140 0.5456 -1.89 0.0566

Single Mean 0 -1.1037 0.8762 -0.70 0.8446 1.26 0.7491

1 -0.8827 0.8965 -0.62 0.8625 1.49 0.6909

2 -0.7461 0.9081 -0.58 0.8721 1.78 0.6149

Trend 0 -7.9768 0.5833 -2.05 0.5750 2.10 0.7567

1 -6.4670 0.7053 -1.83 0.6892 1.68 0.8411

2 -5.3139 0.7964 -1.64 0.7746 1.35 0.9070

181
Table 2.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.463158 1.00000 | |********************| 0

1 -0.052117 -.11253 | **| . | 0.047782

2 -0.039721 -.08576 | **| . | 0.048383

3 0.030843 0.06659 | . |*. | 0.048729

4 -0.0088870 -.01919 | .|. | 0.048936

5 0.0028092 0.00607 | .|. | 0.048953

6 0.0085925 0.01855 | .|. | 0.048955

7 -0.010600 -.02289 | .|. | 0.048971

8 0.039246 0.08474 | . |** | 0.048996

9 -0.0020832 -.00450 | .|. | 0.049329

10 -0.0026002 -.00561 | .|. | 0.049330

11 0.022955 0.04956 | . |*. | 0.049331

12 0.0019856 0.00429 | .|. | 0.049445

13 0.012072 0.02606 | . |*. | 0.049446

14 -0.0098142 -.02119 | .|. | 0.049477

15 -0.039096 -.08441 | **| . | 0.049498

16 0.044015 0.09503 | . |** | 0.049825

17 0.040881 0.08827 | . |** | 0.050238

18 -0.034906 -.07537 | **| . | 0.050590

19 0.0024852 0.00537 | .|. | 0.050846

20 0.013148 0.02839 | . |*. | 0.050847

21 -0.029651 -.06402 | .*| . | 0.050884

22 0.011746 0.02536 | . |*. | 0.051067

23 0.0034855 0.00753 | .|. | 0.051096

182
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 -0.015811 -.03414 | .*| . | 0.051098

Table 2.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -483.729 0.0001 -23.42 <.0001

1 -581.080 0.0001 -17.06 <.0001

2 -485.503 0.0001 -12.44 <.0001

Single Mean 0 -486.198 0.0001 -23.51 <.0001 276.46 0.0010

1 -592.229 0.0001 -17.19 <.0001 147.79 0.0010

2 -505.889 0.0001 -12.59 <.0001 79.22 0.0010

Trend 0 -486.194 0.0001 -23.49 <.0001 275.83 0.0010

1 -592.256 0.0001 -17.17 <.0001 147.46 0.0010

2 -505.971 0.0001 -12.57 <.0001 79.03 0.0010

183
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0214 9.7581 15.34
1 1 0.0008 0.3312 3.84

VAR-MODEL

Table 3.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.02283 0.21866 0.1 0.9169 1
AR1_1_1 0.42645 0.03928 10.86 0.0001 CDS (t-1)
AR1_1_2 -2.90776 0.32088 -9.06 0.0001 EQ (t-1)
EQ 1 diff Constant2 -0.05391 0.03232 -1.67 0.0961 1
AR1_2_1 0.00787 0.00581 1.36 0.1758 CDS (t-1)
AR1_2_2 -0.10727 0.04743 -2.26 0.0242 EQ (t-1)

GRANGER CAUSALITY TEST

Table 3.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 82.12 <.0001
2 2 1.84 0.1751

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

184
UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.3341 4.55348 108.86 <.0001
EQ 1diff 0.017 0.6731 3.76 0.0242

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 2.0782 662.74 <.0001 12.21 0.0005
EQ 1diff 2.00482 24.15 <.0001 0.72 0.3978

185
Appendix 17 – After QUARTILE 4
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 92329.999 1.00000 | |********************| 0

1 92040.342 0.99686 | . |********************| 0.047782

2 91670.245 0.99285 | . |********************| 0.082588

3 91237.326 0.98817 | . |********************| 0.106404

4 90756.945 0.98296 | . |********************| 0.125621

5 90261.314 0.97759 | . |********************| 0.142101

6 89737.036 0.97192 | . |******************* | 0.156705

7 89187.934 0.96597 | . |******************* | 0.169911

8 88622.581 0.95985 | . |******************* | 0.182018

9 88039.893 0.95354 | . |******************* | 0.193229

10 87439.867 0.94704 | . |******************* | 0.203689

11 86820.993 0.94033 | . |******************* | 0.213505

12 86180.011 0.93339 | . |******************* | 0.222760

13 85522.556 0.92627 | . |******************* | 0.231517

14 84849.210 0.91898 | . |****************** | 0.239829

15 84169.794 0.91162 | . |****************** | 0.247738

16 83466.296 0.90400 | . |****************** | 0.255282

17 82730.595 0.89603 | . |****************** | 0.262489

18 81974.804 0.88785 | . |****************** | 0.269382

19 81194.672 0.87940 | . |****************** | 0.275982

186
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

20 80415.240 0.87095 | . |***************** | 0.282307

21 79641.538 0.86257 | . |***************** | 0.288376

22 78895.803 0.85450 | . |***************** | 0.294208

23 78167.311 0.84661 | . |***************** | 0.299821

24 77446.464 0.83880 | . |***************** | 0.305230

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.6186 0.8359 1.36 0.9568

1 0.4359 0.7904 0.60 0.8462

2 0.4313 0.7892 0.59 0.8424

Single Mean 0 -0.7693 0.9062 -0.81 0.8141 2.33 0.4753

1 -1.4062 0.8455 -0.94 0.7751 1.17 0.7726

2 -1.4953 0.8359 -0.98 0.7607 1.21 0.7608

Trend 0 -1.4106 0.9823 -0.60 0.9784 0.37 0.9900

1 -5.6105 0.7737 -1.49 0.8320 1.18 0.9404

2 -5.7181 0.7653 -1.48 0.8354 1.19 0.9393

187
Table 1.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 188.678 1.00000 | |********************| 0

1 80.057769 0.42431 | . |******** | 0.047782

2 35.853886 0.19003 | . |**** | 0.055724

3 28.275728 0.14986 | . |*** | 0.057185

4 23.708719 0.12566 | . |*** | 0.058074

5 30.783547 0.16315 | . |*** | 0.058692

6 11.243504 0.05959 | . |*. | 0.059718

7 6.184162 0.03278 | . |*. | 0.059854

8 11.804051 0.06256 | . |*. | 0.059895

9 17.590736 0.09323 | . |** | 0.060044

10 12.952858 0.06865 | . |*. | 0.060374

11 4.637757 0.02458 | .|. | 0.060552

12 8.492896 0.04501 | . |*. | 0.060574

13 3.541938 0.01877 | .|. | 0.060651

14 -3.724335 -.01974 | .|. | 0.060664

15 9.875348 0.05234 | . |*. | 0.060679

16 17.733699 0.09399 | . |** | 0.060782

17 19.579603 0.10377 | . |** | 0.061112

18 6.291425 0.03334 | . |*. | 0.061513

19 -11.918689 -.06317 | .*| . | 0.061555

20 -10.305621 -.05462 | .*| . | 0.061703

21 -14.690940 -.07786 | **| . | 0.061813

22 -9.754400 -.05170 | .*| . | 0.062036

23 -4.551015 -.02412 | .|. | 0.062135

188
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 0.543776 0.00288 | .|. | 0.062156

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -248.958 0.0001 -13.18 <.0001

1 -238.555 0.0001 -10.83 <.0001

2 -191.340 0.0001 -8.71 <.0001

Single Mean 0 -251.239 0.0001 -13.25 <.0001 87.75 0.0010

1 -242.598 0.0001 -10.90 <.0001 59.37 0.0010

2 -196.313 0.0001 -8.78 <.0001 38.57 0.0010

Trend 0 -251.441 0.0001 -13.24 <.0001 87.70 0.0010

1 -242.777 0.0001 -10.89 <.0001 59.32 0.0010

2 -196.473 0.0001 -8.78 <.0001 38.59 0.0010

189
EQUITY

STATIONARITY

Table 2.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 95.922428 1.00000 | |********************| 0

1 95.273907 0.99324 | . |********************| 0.047782

2 94.664170 0.98688 | . |********************| 0.082388

3 94.074389 0.98073 | . |********************| 0.105995

4 93.463665 0.97437 | . |******************* | 0.125008

5 92.888544 0.96837 | . |******************* | 0.141287

6 92.304478 0.96228 | . |******************* | 0.155705

7 91.657818 0.95554 | . |******************* | 0.168737

8 91.011762 0.94881 | . |******************* | 0.180669

9 90.322081 0.94162 | . |******************* | 0.191708

10 89.578214 0.93386 | . |******************* | 0.201992

11 88.878457 0.92657 | . |******************* | 0.211620

12 88.124174 0.91870 | . |****************** | 0.220688

13 87.419104 0.91135 | . |****************** | 0.229253

14 86.638317 0.90321 | . |****************** | 0.237381

15 85.877721 0.89528 | . |****************** | 0.245101

16 85.121708 0.88740 | . |****************** | 0.252457

17 84.303323 0.87887 | . |****************** | 0.259481

18 83.436491 0.86983 | . |***************** | 0.266191

19 82.642984 0.86156 | . |***************** | 0.272603

20 81.828865 0.85307 | . |***************** | 0.278750

21 81.014255 0.84458 | . |***************** | 0.284648

190
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

22 80.227315 0.83638 | . |***************** | 0.290313

23 79.401521 0.82777 | . |***************** | 0.295763

24 78.596815 0.81938 | . |**************** | 0.301006

Table 2.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.7626 0.5159 -1.89 0.0557

1 -0.7567 0.5170 -2.08 0.0358

2 -0.7528 0.5178 -2.19 0.0276

Single Mean 0 -1.7161 0.8113 -1.10 0.7183 1.99 0.5613

1 -1.5247 0.8327 -1.08 0.7242 2.33 0.4751

2 -1.4325 0.8427 -1.08 0.7264 2.53 0.4219

Trend 0 -11.0438 0.3656 -2.23 0.4730 2.57 0.6624

1 -8.8382 0.5167 -1.94 0.6298 2.02 0.7736

2 -7.8242 0.5954 -1.80 0.7042 1.77 0.8238

191
Table 2.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.541377 1.00000 | |********************| 0

1 -0.058587 -.10822 | **| . | 0.047782

2 -0.026014 -.04805 | .*| . | 0.048338

3 0.041173 0.07605 | . |** | 0.048447

4 -0.014682 -.02712 | .*| . | 0.048719

5 0.0070659 0.01305 | .|. | 0.048753

6 -0.015748 -.02909 | .*| . | 0.048761

7 -0.013014 -.02404 | .|. | 0.048801

8 0.040379 0.07459 | . |*. | 0.048828

9 0.0051241 0.00946 | .|. | 0.049087

10 -0.012731 -.02352 | .|. | 0.049092

11 0.026606 0.04915 | . |*. | 0.049117

12 0.0059796 0.01105 | .|. | 0.049229

13 0.027678 0.05112 | . |*. | 0.049235

14 -0.0022410 -.00414 | .|. | 0.049356

15 -0.058947 -.10888 | **| . | 0.049357

16 0.048890 0.09031 | . |** | 0.049902

17 0.045641 0.08431 | . |** | 0.050274

18 -0.048975 -.09046 | **| . | 0.050596

19 0.0077841 0.01438 | .|. | 0.050964

20 0.010863 0.02006 | .|. | 0.050973

21 -0.033097 -.06113 | .*| . | 0.050991

22 0.0048681 0.00899 | .|. | 0.051158

23 -0.0038721 -.00715 | .|. | 0.051162

192
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 -0.021301 -.03935 | .*| . | 0.051164

Table 2.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -481.015 0.0001 -23.26 <.0001

1 -531.838 0.0001 -16.32 <.0001

2 -421.818 0.0001 -11.89 <.0001

Single Mean 0 -484.311 0.0001 -23.40 <.0001 273.77 0.0010

1 -544.993 0.0001 -16.49 <.0001 135.94 0.0010

2 -442.895 0.0001 -12.06 <.0001 72.71 0.0010

Trend 0 -484.590 0.0001 -23.38 <.0001 273.40 0.0010

1 -546.043 0.0001 -16.48 <.0001 135.86 0.0010

2 -444.399 0.0001 -12.06 <.0001 72.71 0.0010

193
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0294 14.2603 15.34
1 1 0.0028 1.2339 3.84

VAR-MODEL

Table 3.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.35171 0.54879 0.64 0.5219 1
AR1_1_1 0.39976 0.03973 10.06 0.0001 CDS (t-1)
AR1_1_2 -6.93038 0.74121 -9.35 0.0001 EQ (t-1)
EQ 1 diff Constant2 -0.06795 0.03516 -1.93 0.0539 1
AR1_2_1 0.0022 0.00255 0.86 0.388 CDS (t-1)
AR1_2_2 -0.10547 0.04749 -2.22 0.0269 EQ (t-1)

GRANGER CAUSALITY TEST

Table 3.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 87.42 <.0001
2 2 0.75 0.3875

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

194
UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.3181 11.38451 101.21 <0.0001
EQ 1diff 0.0136 0.72938 2.98 0.0516

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 2.1646 123.17 <.0001 17.32 <.0001
EQ 1diff 2.00557 32.28 <.0001 0.31 0.5798

195
Appendix 18 – INVESTMENT GRADE
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 3084.198 1.00000 | |********************| 0

1 3076.764 0.99759 | .|********************| 0.027227

2 3067.104 0.99446 | . |********************| 0.047082

3 3056.428 0.99100 | . |********************| 0.060687

4 3045.271 0.98738 | . |********************| 0.071686

5 3034.000 0.98372 | . |********************| 0.081144

6 3022.343 0.97994 | . |********************| 0.089549

7 3010.419 0.97608 | . |********************| 0.097174

8 2998.626 0.97225 | . |******************* | 0.104188

9 2987.268 0.96857 | . |******************* | 0.110710

10 2975.942 0.96490 | . |******************* | 0.116823

11 2964.192 0.96109 | . |******************* | 0.122588

12 2951.695 0.95704 | . |******************* | 0.128052

13 2938.791 0.95285 | . |******************* | 0.133249

14 2925.662 0.94860 | . |******************* | 0.138208

15 2912.058 0.94419 | . |******************* | 0.142952

16 2897.457 0.93945 | . |******************* | 0.147503

17 2881.838 0.93439 | . |******************* | 0.151874

18 2866.037 0.92926 | . |******************* | 0.156077

19 2850.196 0.92413 | . |****************** | 0.160126

20 2834.131 0.91892 | . |****************** | 0.164032

21 2818.055 0.91371 | . |****************** | 0.167804

22 2802.441 0.90864 | . |****************** | 0.171453

196
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

23 2787.451 0.90378 | . |****************** | 0.174986

24 2772.853 0.89905 | . |****************** | 0.178413

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 1.8073 0.9832 1.89 0.9863

1 0.9913 0.9115 0.63 0.8525

2 1.2858 0.9497 0.92 0.9045

Single Mean 0 1.2268 0.9923 0.88 0.9953 1.95 0.5713

1 -0.4693 0.9295 -0.21 0.9352 0.59 0.9235

2 0.1095 0.9627 0.05 0.9621 0.74 0.8815

Trend 0 -1.0740 0.9874 -0.61 0.9778 2.69 0.6377

1 -3.6732 0.9050 -1.28 0.8912 1.70 0.8379

2 -2.7808 0.9458 -1.09 0.9295 1.71 0.8346

197
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 4.388184 1.00000 | |********************| 0

1 1.993354 0.45425 | .|********* | 0.027237

2 0.511146 0.11648 | .|** | 0.032373

3 0.201439 0.04590 | .|* | 0.032682

4 0.141152 0.03217 | .|* | 0.032730

5 0.110958 0.02529 | .|* | 0.032753

6 -0.079133 -.01803 | .|. | 0.032768

7 -0.259706 -.05918 | *|. | 0.032775

8 -0.280468 -.06391 | *|. | 0.032854

9 0.302948 0.06904 | .|* | 0.032947

10 0.614446 0.14002 | .|*** | 0.033054

11 0.302948 0.06904 | .|* | 0.033491

12 -0.076140 -.01735 | .|. | 0.033596

13 -0.238066 -.05425 | *|. | 0.033603

14 -0.129233 -.02945 | *|. | 0.033668

15 0.302628 0.06896 | .|* | 0.033687

16 0.405072 0.09231 | .|** | 0.033791

17 0.217240 0.04951 | .|* | 0.033978

18 -0.053443 -.01218 | .|. | 0.034031

19 0.036822 0.00839 | .|. | 0.034035

20 -0.071216 -.01623 | .|. | 0.034036

21 -0.238140 -.05427 | *|. | 0.034042

22 -0.228327 -.05203 | *|. | 0.034106

23 -0.077410 -.01764 | .|. | 0.034165

24 0.095700 0.02181 | .|. | 0.034172

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

198
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -732.709 0.0001 -22.41 <.0001

1 -916.668 0.0001 -21.35 <.0001

2 -801.754 0.0001 -17.50 <.0001

Single Mean 0 -734.426 0.0001 -22.44 <.0001 251.72 0.0010

1 -920.852 0.0001 -21.39 <.0001 228.69 0.0010

2 -807.644 0.0001 -17.53 <.0001 153.73 0.0010

Trend 0 -737.135 0.0001 -22.48 <.0001 252.73 0.0010

1 -927.542 0.0001 -21.45 <.0001 230.02 0.0010

2 -817.294 0.0001 -17.60 <.0001 154.88 0.0010

AUTOCORRELATION

Graph 2.1 – Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0014

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

199
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1346.9729 <.0001

AR(2) 1347.3508 <.0001

AR(3) 1347.3641 <.0001

AR(4) 1347.3662 <.0001

Graph 2.2 – 1. difference

Table 2.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 1.0903

Table 2.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

200
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 278.4716 <.0001

AR(2) 291.9942 <.0001

AR(3) 294.5321 <.0001

AR(4) 294.5422 <.0001

201
EQUITY

STATIONARITY

Table 3.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 48.992883 1.00000 | |********************| 0

1 48.783824 0.99573 | .|********************| 0.027227

2 48.607518 0.99213 | . |********************| 0.047024

3 48.454051 0.98900 | . |********************| 0.060585

4 48.272662 0.98530 | . |********************| 0.071559

5 48.096257 0.98170 | . |********************| 0.080994

6 47.925004 0.97820 | . |********************| 0.089380

7 47.753871 0.97471 | . |******************* | 0.096992

8 47.586496 0.97129 | . |******************* | 0.104000

9 47.403435 0.96756 | . |******************* | 0.110520

10 47.216761 0.96375 | . |******************* | 0.116631

11 47.035896 0.96006 | . |******************* | 0.122392

12 46.843762 0.95613 | . |******************* | 0.127852

13 46.643688 0.95205 | . |******************* | 0.133047

14 46.433104 0.94775 | . |******************* | 0.138005

15 46.232559 0.94366 | . |******************* | 0.142748

16 46.046788 0.93987 | . |******************* | 0.147300

17 45.838990 0.93563 | . |******************* | 0.151680

18 45.619511 0.93115 | . |******************* | 0.155900

19 45.418883 0.92705 | . |******************* | 0.159970

20 45.216568 0.92292 | . |****************** | 0.163904

21 45.003360 0.91857 | . |****************** | 0.167712

22 44.801865 0.91446 | . |****************** | 0.171401

23 44.588919 0.91011 | . |****************** | 0.174980

24 44.380207 0.90585 | . |****************** | 0.178454

202
Table 3.2 – Original Data
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.1411 0.6509 -0.33 0.5682

1 -0.1268 0.6541 -0.33 0.5671

2 -0.1148 0.6569 -0.33 0.5676

Single Mean 0 -3.3425 0.6155 -1.20 0.6769 0.73 0.8844

1 -2.4995 0.7183 -1.01 0.7522 0.52 0.9467

2 -1.9575 0.7838 -0.86 0.7997 0.39 0.9761

Trend 0 -2.2556 0.9634 -0.78 0.9659 1.74 0.8289

1 -1.3523 0.9834 -0.53 0.9823 1.91 0.7954

2 -0.7578 0.9910 -0.32 0.9899 2.20 0.7369

203
Table 3.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.281315 1.00000 | |********************| 0

1 -0.032450 -.11535 | **|. | 0.027237

2 -0.022693 -.08067 | **|. | 0.027597

3 0.022773 0.08095 | .|** | 0.027771

4 -0.0033311 -.01184 | .|. | 0.027946

5 -0.0043126 -.01533 | .|. | 0.027949

6 -0.0037160 -.01321 | .|. | 0.027956

7 -0.0067897 -.02414 | .|. | 0.027960

8 0.018273 0.06496 | .|* | 0.027976

9 -0.0009268 -.00329 | .|. | 0.028087

10 0.0049757 0.01769 | .|. | 0.028088

11 0.0094626 0.03364 | .|* | 0.028096

12 0.0027615 0.00982 | .|. | 0.028126

13 0.0093474 0.03323 | .|* | 0.028128

14 -0.0056024 -.01991 | .|. | 0.028157

15 -0.024467 -.08697 | **|. | 0.028168

16 0.020153 0.07164 | .|* | 0.028366

17 0.016384 0.05824 | .|* | 0.028500

18 -0.019827 -.07048 | *|. | 0.028588

19 0.00065560 0.00233 | .|. | 0.028717

20 0.0054657 0.01943 | .|. | 0.028717

21 -0.014639 -.05204 | *|. | 0.028727

22 0.0090231 0.03207 | .|* | 0.028797

23 0.0024442 0.00869 | .|. | 0.028823

24 -0.0056864 -.02021 | .|. | 0.028825

Table 3.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

204
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1502.37 0.0001 -41.20 <.0001

1 -1817.38 0.0001 -30.12 <.0001

2 -1490.62 0.0001 -21.87 <.0001

Single Mean 0 -1502.39 0.0001 -41.18 <.0001 848.03 0.0010

1 -1817.51 0.0001 -30.11 <.0001 453.40 0.0010

2 -1490.86 0.0001 -21.86 <.0001 238.92 0.0010

Trend 0 -1505.59 0.0001 -41.26 <.0001 851.40 0.0010

1 -1832.20 0.0001 -30.22 <.0001 456.66 0.0010

2 -1516.77 0.0001 -21.97 <.0001 241.30 0.0010

AUTOCORRELATION

Graph 4.1 – Original data

Table 4.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0057

Table 4.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

205
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1339.9142 <.0001

AR(2) 1340.0097 <.0001

AR(3) 1340.0674 <.0001

AR(4) 1340.0940 <.0001

Graph 4.2 – 1. difference

Table 4.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 2.2304

Table 4.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

206
Alternative LM Pr > LM

AR(1) 17.9382 <.0001

AR(2) 30.0042 <.0001

AR(3) 34.9291 <.0001

AR(4) 34.9353 <.0001

MODEL

CO-INTEGRATION

Table 5.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0044 6.3969 15.34
1 1 0.0003 0.4611 3.84

VAR-MODEL

Table 5.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.06147 0.04741 1.30 0.1951 1
AR1_1_1 0.40062 0.02760 14.51 0.0001 CDS (t-1)
AR1_1_2 -1.26317 0.09215 -13.71 0.0001 EQ (t-1)
AR2_1_1 -0.04711 0.02586 -1.82 0.0687 CDS (t-2)
AR2_1_2 -0.49717 0.09690 -5.13 0.0001 EQ (t-2)
EQ 1 diff Constant2 -0.00356 0.01433 -0.25 0.8039 1
AR1_2_1 0.01952 0.00834 2.34 0.0194 CDS (t-1)
AR1_2_2 -0.11200 0.02784 -4.02 0.0001 EQ (t-1)
AR2_2_1 -0.01088 0.00781 -1.39 0.1641 CDS (t-2)
AR2_2_2 -0.07091 0.02928 -2.42 0.0156 EQ (t-2)

GRANGER CAUSALITY TEST

Table 5.3
Granger Causality Wald Test

207
Test DF Chi-Square Pr > ChiSq
1 2 195.67 < 0.0001
2 2 5.59 0.0612

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 5.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.3163 1.73640 155.12 < 0.0001
EQ 1diff 0.0263 0.52466 9.07 < 0.0001

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 5.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 2.00886 9999.99 < 0.0001 144.79 < 0.0001
EQ 1diff 1.98920 1679.99 < 0.0001 44.75 < 0.0001

208
Appendix 19 – HIGH YIELD
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 85423.524 1.00000 | |********************| 0

1 85287.062 0.99840 | .|********************| 0.027227

2 85112.836 0.99636 | . |********************| 0.047108

3 84902.928 0.99391 | . |********************| 0.060753

4 84667.323 0.99115 | . |********************| 0.071802

5 84418.758 0.98824 | . |********************| 0.081314

6 84148.142 0.98507 | . |********************| 0.089777

7 83858.590 0.98168 | . |********************| 0.097460

8 83557.352 0.97815 | . |********************| 0.104534

9 83249.760 0.97455 | . |******************* | 0.111112

10 82940.000 0.97093 | . |******************* | 0.117277

11 82625.747 0.96725 | . |******************* | 0.123091

12 82298.611 0.96342 | . |******************* | 0.128602

13 81962.872 0.95949 | . |******************* | 0.133846

14 81617.249 0.95544 | . |******************* | 0.138851

15 81267.567 0.95135 | . |******************* | 0.143642

16 80904.913 0.94710 | . |******************* | 0.148239

17 80531.886 0.94274 | . |******************* | 0.152659

18 80149.121 0.93826 | . |******************* | 0.156915

19 79758.068 0.93368 | . |******************* | 0.161020

20 79363.274 0.92906 | . |******************* | 0.164984

21 78968.125 0.92443 | . |****************** | 0.168818

22 78585.901 0.91996 | . |****************** | 0.172530

209
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

23 78209.090 0.91555 | . |****************** | 0.176129

24 77834.134 0.91116 | . |****************** | 0.179622

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 1.4402 0.9632 1.84 0.9847

1 0.9145 0.8985 0.72 0.8711

2 0.7629 0.8690 0.56 0.8373

Single Mean 0 0.6222 0.9809 0.57 0.9889 2.26 0.4899

1 -0.3871 0.9351 -0.22 0.9336 0.82 0.8619

2 -0.6816 0.9136 -0.36 0.9135 0.75 0.8793

Trend 0 -1.5910 0.9793 -1.00 0.9428 1.98 0.7809

1 -3.6213 0.9077 -1.41 0.8574 1.53 0.8724

2 -4.1140 0.8798 -1.49 0.8322 1.53 0.8715

210
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 75.843021 1.00000 | |********************| 0

1 33.521921 0.44199 | .|********* | 0.027237

2 18.839581 0.24840 | .|***** | 0.032120

3 13.040797 0.17194 | .|*** | 0.033515

4 10.014724 0.13205 | .|*** | 0.034163

5 15.239986 0.20094 | .|**** | 0.034539

6 9.739829 0.12842 | .|*** | 0.035396

7 7.037779 0.09279 | .|** | 0.035740

8 8.080389 0.10654 | .|** | 0.035918

9 7.131033 0.09402 | .|** | 0.036152

10 6.477075 0.08540 | .|** | 0.036333

11 3.676871 0.04848 | .|* | 0.036481

12 3.161184 0.04168 | .|* | 0.036529

13 -0.782964 -.01032 | .|. | 0.036565

14 -3.324278 -.04383 | *|. | 0.036567

15 2.821680 0.03720 | .|* | 0.036606

16 5.366284 0.07076 | .|* | 0.036634

17 8.651856 0.11408 | .|** | 0.036735

18 3.275851 0.04319 | .|* | 0.036997

19 -3.494169 -.04607 | *|. | 0.037034

20 -3.454293 -.04555 | *|. | 0.037077

21 -7.090035 -.09348 | **|. | 0.037118

22 -2.512721 -.03313 | *|. | 0.037292

23 0.335228 0.00442 | .|. | 0.037314

24 4.556487 0.06008 | .|* | 0.037315

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

211
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -748.551 0.0001 -22.74 <.0001

1 -648.256 0.0001 -17.88 <.0001

2 -565.516 0.0001 -15.08 <.0001

Single Mean 0 -750.910 0.0001 -22.78 <.0001 259.50 0.0010

1 -652.004 0.0001 -17.92 <.0001 160.62 0.0010

2 -570.346 0.0001 -15.12 <.0001 114.37 0.0010

Trend 0 -752.561 0.0001 -22.81 <.0001 260.05 0.0010

1 -654.645 0.0001 -17.94 <.0001 161.01 0.0010

2 -573.837 0.0001 -15.14 <.0001 114.69 0.0010

AUTOCORRELATION

Graph 2.1 – Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0009

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

212
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1347.4659 <.0001

AR(2) 1347.6512 <.0001

AR(3) 1347.6627 <.0001

AR(4) 1347.6662 <.0001

Graph 2.2 – 1. difference

Table 2.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 1.1148

Table 2.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

213
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 263.6621 <.0001

AR(2) 268.8385 <.0001

AR(3) 271.8410 <.0001

AR(4) 273.2075 <.0001

214
EQUITY

STATIONARITY

Table 3.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 70.777886 1.00000 | |********************| 0

1 70.597011 0.99744 | .|********************| 0.027227

2 70.412903 0.99484 | . |********************| 0.047078

3 70.229255 0.99225 | . |********************| 0.060693

4 70.036249 0.98952 | . |********************| 0.071717

5 69.843923 0.98680 | . |********************| 0.081210

6 69.651976 0.98409 | . |********************| 0.089659

7 69.453624 0.98129 | . |********************| 0.097337

8 69.251590 0.97844 | . |********************| 0.104413

9 69.043442 0.97549 | . |********************| 0.111002

10 68.824463 0.97240 | . |******************* | 0.117185

11 68.613047 0.96941 | . |******************* | 0.123021

12 68.392735 0.96630 | . |******************* | 0.128559

13 68.165863 0.96310 | . |******************* | 0.133835

14 67.933990 0.95982 | . |******************* | 0.138877

15 67.700951 0.95653 | . |******************* | 0.143711

16 67.465957 0.95321 | . |******************* | 0.148355

17 67.217635 0.94970 | . |******************* | 0.152828

18 66.970561 0.94621 | . |******************* | 0.157142

19 66.723602 0.94272 | . |******************* | 0.161310

20 66.470294 0.93914 | . |******************* | 0.165343

21 66.212346 0.93549 | . |******************* | 0.169251

22 65.956019 0.93187 | . |******************* | 0.173042

23 65.693431 0.92816 | . |******************* | 0.176723

24 65.439377 0.92457 | . |****************** | 0.180300

215
Table 3.2 – Original Data
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.4453 0.5821 -0.99 0.2895

1 -0.4548 0.5800 -1.00 0.2844

2 -0.4558 0.5798 -1.02 0.2771

Single Mean 0 0.6423 0.9814 0.38 0.9823 0.71 0.8886

1 0.6313 0.9811 0.37 0.9819 0.72 0.8859

2 0.6836 0.9825 0.41 0.9835 0.77 0.8733

Trend 0 -2.3566 0.9604 -1.14 0.9201 3.15 0.5436

1 -2.3189 0.9615 -1.12 0.9241 3.03 0.5684

2 -2.2115 0.9647 -1.09 0.9293 3.05 0.5646

216
Table 3.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.146492 1.00000 | |********************| 0

1 0.00085645 0.00585 | .|. | 0.027237

2 -0.0024640 -.01682 | .|. | 0.027238

3 0.0030013 0.02049 | .|. | 0.027245

4 0.00087008 0.00594 | .|. | 0.027257

5 0.0030234 0.02064 | .|. | 0.027258

6 0.00098242 0.00671 | .|. | 0.027269

7 0.0022099 0.01509 | .|. | 0.027271

8 0.0043070 0.02940 | .|* | 0.027277

9 0.0055878 0.03814 | .|* | 0.027300

10 0.0032663 0.02230 | .|. | 0.027340

11 0.0048234 0.03293 | .|* | 0.027353

12 0.0020255 0.01383 | .|. | 0.027383

13 0.0024162 0.01649 | .|. | 0.027388

14 0.0029703 0.02028 | .|. | 0.027395

15 -0.0077973 -.05323 | *|. | 0.027406

16 0.010799 0.07372 | .|* | 0.027483

17 0.0034274 0.02340 | .|. | 0.027629

18 0.0012897 0.00880 | .|. | 0.027644

19 0.0029948 0.02044 | .|. | 0.027646

20 -0.0008556 -.00584 | .|. | 0.027657

21 -0.0053996 -.03686 | *|. | 0.027658

22 0.0051839 0.03539 | .|* | 0.027695

23 -0.0018675 -.01275 | .|. | 0.027728

24 -0.0013158 -.00898 | .|. | 0.027732

Table 3.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

217
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1337.81 0.0001 -36.44 <.0001

1 -1380.18 0.0001 -26.25 <.0001

2 -1292.03 0.0001 -20.85 <.0001

Single Mean 0 -1339.12 0.0001 -36.46 <.0001 664.82 0.0010

1 -1384.35 0.0001 -26.28 <.0001 345.36 0.0010

2 -1299.86 0.0001 -20.89 <.0001 218.16 0.0010

Trend 0 -1344.08 0.0001 -36.58 <.0001 669.10 0.0010

1 -1399.82 0.0001 -26.41 <.0001 348.82 0.0010

2 -1329.17 0.0001 -21.03 <.0001 221.13 0.0010

AUTOCORRELATION

Graph 4.1 – Original data

Table 4.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0021

Table 4.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

218
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1346.1939 <.0001

AR(2) 1346.1939 <.0001

AR(3) 1346.1949 <.0001

AR(4) 1346.1960 <.0001

Graph 4.2 – 1. difference

Table 4.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 1.9877

Table 4.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

219
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 0.0461 0.8300

AR(2) 0.4291 0.8069

AR(3) 1.0067 0.7996

AR(4) 1.0461 0.9027

MODEL

CO-INTEGRATION

Table 5.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0068 9.4004 15.34
1 1 0.0002 0.2226 3.84

VAR-MODEL

Table 5.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.20606 0.20294 1.02 0.3101 1
AR1_1_1 0.41559 0.02337 17.78 0.0001 CDS (t-1)
AR1_1_2 -6.38043 0.53152 -12.00 0.0001 EQ (t-1)
EQ 1 diff Constant2 -0.01221 0.01046 -1.17 0.2433 1
AR1_2_1 0.00055 0.00120 0.46 0.6470 CDS (t-1)
AR1_2_2 0.00705 0.02740 0.26 0.7969 EQ (t-1)

GRANGER CAUSALITY TEST

220
Table 5.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 144.10 < 0.0001
2 2 0.21 0.6469

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 5.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.2735 7.43388 252.97 < 0.0001
EQ 1diff 0.0002 0.38319 0.13 0.8799

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 5.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 2.12445 9999.99 < 0.0001 73.76 < 0.0001
EQ 1diff 1.99966 297.67 < 0.0001 2.13 0.1450

221
Appendix 20 – Investment grade QUARTILE 1
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 502.463 1.00000 | |********************| 0

1 501.548 0.99818 | .|********************| 0.027227

2 500.319 0.99573 | . |********************| 0.047101

3 498.876 0.99286 | . |********************| 0.060732

4 497.304 0.98973 | . |********************| 0.071763

5 495.714 0.98657 | . |********************| 0.081254

6 494.036 0.98323 | . |********************| 0.089695

7 492.302 0.97978 | . |********************| 0.097358

8 490.514 0.97622 | . |********************| 0.104411

9 488.692 0.97259 | . |******************* | 0.110971

10 486.834 0.96889 | . |******************* | 0.117120

11 484.922 0.96509 | . |******************* | 0.122918

12 482.922 0.96111 | . |******************* | 0.128412

13 480.860 0.95701 | . |******************* | 0.133638

14 478.784 0.95287 | . |******************* | 0.138626

15 476.637 0.94860 | . |******************* | 0.143399

16 474.425 0.94420 | . |******************* | 0.147977

17 472.149 0.93967 | . |******************* | 0.152378

18 469.823 0.93504 | . |******************* | 0.156614

19 467.475 0.93037 | . |******************* | 0.160699

222
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

20 465.111 0.92566 | . |******************* | 0.164644

21 462.782 0.92103 | . |****************** | 0.168458

22 460.547 0.91658 | . |****************** | 0.172150

23 458.357 0.91222 | . |****************** | 0.175730

24 456.190 0.90791 | . |****************** | 0.179206

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 1.1246 0.9310 1.34 0.9549

1 0.5999 0.8320 0.45 0.8125

2 0.4367 0.7911 0.30 0.7740

Single Mean 0 0.5891 0.9800 0.43 0.9843 1.02 0.8110

1 -0.8027 0.9037 -0.38 0.9109 0.45 0.9633

2 -1.2333 0.8638 -0.53 0.8827 0.46 0.9602

Trend 0 -1.2043 0.9857 -0.76 0.9680 2.45 0.6845

1 -3.0231 0.9361 -1.21 0.9062 1.56 0.8650

2 -3.5685 0.9105 -1.32 0.8818 1.55 0.8679

223
Table 1.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.692405 1.00000 | |********************| 0

1 0.289976 0.41879 | .|******** | 0.027237

2 0.166654 0.24069 | .|***** | 0.031655

3 0.089532 0.12931 | .|*** | 0.032985

4 0.0072197 0.01043 | .|. | 0.033359

5 0.038950 0.05625 | .|* | 0.033361

6 0.022676 0.03275 | .|* | 0.033432

7 0.029558 0.04269 | .|* | 0.033455

8 0.017419 0.02516 | .|* | 0.033496

9 0.048652 0.07026 | .|* | 0.033510

10 0.077225 0.11153 | .|** | 0.033619

11 0.054657 0.07894 | .|** | 0.033892

12 -0.0001472 -.00021 | .|. | 0.034028

13 -0.022086 -.03190 | *|. | 0.034028

14 0.0060886 0.00879 | .|. | 0.034051

15 0.046451 0.06709 | .|* | 0.034052

16 0.054059 0.07807 | .|** | 0.034150

17 0.047894 0.06917 | .|* | 0.034282

18 0.011929 0.01723 | .|. | 0.034386

19 0.0073732 0.01065 | .|. | 0.034392

20 -0.049491 -.07148 | *|. | 0.034395

21 -0.071233 -.10288 | **|. | 0.034505

22 -0.034785 -.05024 | *|. | 0.034731

23 -0.0086443 -.01248 | .|. | 0.034785

224
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 0.051381 0.07421 | .|* | 0.034789

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -781.421 0.0001 -23.45 <.0001

1 -664.336 0.0001 -18.19 <.0001

2 -655.246 0.0001 -16.17 <.0001

Single Mean 0 -782.519 0.0001 -23.46 <.0001 275.17 0.0010

1 -666.033 0.0001 -18.20 <.0001 165.71 0.0010

2 -657.768 0.0001 -16.19 <.0001 131.07 0.0010

Trend 0 -785.005 0.0001 -23.50 <.0001 276.13 0.0010

1 -669.810 0.0001 -18.24 <.0001 166.43 0.0010

2 -663.611 0.0001 -16.24 <.0001 131.80 0.0010

225
EQUITY

STATIONARITY

Table 2.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 49.608374 1.00000 | |********************| 0

1 49.393857 0.99568 | .|********************| 0.027227

2 49.230859 0.99239 | . |********************| 0.047022

3 49.103609 0.98983 | . |********************| 0.060590

4 48.938957 0.98651 | . |********************| 0.071580

5 48.778504 0.98327 | . |********************| 0.081034

6 48.626029 0.98020 | . |********************| 0.089443

7 48.478680 0.97723 | . |********************| 0.097079

8 48.337157 0.97437 | . |******************* | 0.104116

9 48.170760 0.97102 | . |******************* | 0.110670

10 48.007206 0.96772 | . |******************* | 0.116815

11 47.845624 0.96447 | . |******************* | 0.122614

12 47.671883 0.96096 | . |******************* | 0.128114

13 47.491594 0.95733 | . |******************* | 0.133350

14 47.302330 0.95352 | . |******************* | 0.138351

15 47.127856 0.95000 | . |******************* | 0.143140

16 46.976576 0.94695 | . |******************* | 0.147740

17 46.807484 0.94354 | . |******************* | 0.152173

18 46.624472 0.93985 | . |******************* | 0.156449

19 46.467639 0.93669 | . |******************* | 0.160580

20 46.311614 0.93354 | . |******************* | 0.164581

21 46.137369 0.93003 | . |******************* | 0.168460

226
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

22 45.982151 0.92690 | . |******************* | 0.172224

23 45.824115 0.92372 | . |****************** | 0.175883

24 45.662296 0.92046 | . |****************** | 0.179444

Table 2.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.0224 0.6779 -0.05 0.6657

1 -0.0032 0.6823 -0.01 0.6803

2 0.0115 0.6857 0.04 0.6947

Single Mean 0 -5.1424 0.4219 -1.71 0.4235 1.49 0.6895

1 -3.9003 0.5506 -1.51 0.5266 1.17 0.7718

2 -3.1278 0.6413 -1.38 0.5920 0.99 0.8188

Trend 0 -4.4360 0.8595 -1.17 0.9157 1.51 0.8747

1 -2.4714 0.9568 -0.76 0.9679 1.40 0.8984

2 -1.2115 0.9856 -0.42 0.9867 1.54 0.8693

227
Table 2.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.331318 1.00000 | |********************| 0

1 -0.050977 -.15386 | ***|. | 0.027237

2 -0.034881 -.10528 | **|. | 0.027874

3 0.033239 0.10032 | .|** | 0.028167

4 -0.0034228 -.01033 | .|. | 0.028431

5 -0.0092297 -.02786 | *|. | 0.028434

6 -0.0051636 -.01558 | .|. | 0.028454

7 -0.0094634 -.02856 | *|. | 0.028461

8 0.028587 0.08628 | .|** | 0.028482

9 -0.0049431 -.01492 | .|. | 0.028675

10 0.0053475 0.01614 | .|. | 0.028681

11 0.011173 0.03372 | .|* | 0.028688

12 0.0024461 0.00738 | .|. | 0.028717

13 0.0080542 0.02431 | .|. | 0.028718

14 -0.013055 -.03940 | *|. | 0.028734

15 -0.027977 -.08444 | **|. | 0.028774

16 0.019328 0.05834 | .|* | 0.028957

17 0.016059 0.04847 | .|* | 0.029044

18 -0.029514 -.08908 | **|. | 0.029104

19 0.00093307 0.00282 | .|. | 0.029306

20 0.012967 0.03914 | .|* | 0.029306

21 -0.019554 -.05902 | *|. | 0.029345

22 0.0017536 0.00529 | .|. | 0.029433

23 0.0079362 0.02395 | .|. | 0.029433

228
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 -0.011606 -.03503 | *|. | 0.029448

Table 2.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1554.27 0.0001 -42.84 <.0001

1 -2025.84 0.0001 -31.80 <.0001

2 -1630.86 0.0001 -22.50 <.0001

Single Mean 0 -1554.30 0.0001 -42.83 <.0001 917.07 0.0010

1 -2026.03 0.0001 -31.79 <.0001 505.35 0.0010

2 -1631.17 0.0001 -22.49 <.0001 252.86 0.0010

Trend 0 -1556.25 0.0001 -42.87 <.0001 919.01 0.0010

1 -2036.47 0.0001 -31.86 <.0001 507.52 0.0010

2 -1650.26 0.0001 -22.56 <.0001 254.43 0.0010

229
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0042 7.884 15.34
1 1 0.0016 2.174 3.84

VAR-MODEL

Table 3.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.01892 0.01978 0.96 0.339 1
AR1_1_1 0.31869 0.02709 11.76 0.0001 CDS (t-1)
AR1_1_2 -0.34147 0.03527 -9.68 0.0001 EQ (t-1)
AR2_1_1 0.119 0.02644 4.5 0.0001 CDS (t-2)
AR2_1_2 -0.22013 0.0358 -6.15 0.0001 EQ (t-2)
EQ 1 diff Constant2 0.00232 0.0154 0.15 0.8801 1
AR1_2_1 0.03407 0.0211 1.61 0.1066 CDS (t-1)
AR1_2_2 -0.16955 0.02746 -6.17 0.0001 EQ (t-1)
AR2_2_1 0.00446 0.02059 0.22 0.8284 CDS (t-2)
AR2_2_2 -0.11988 0.02787 -4.3 0.0001 EQ (t-2)

GRANGER CAUSALITY TEST

Table 3.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 113.84 <.0001
2 2 3.61 0.1643

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

230
UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.2448 0.72485 108.69 <.0001
EQ 1diff 0.0433 0.56442 15.18 <.0001

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 2.04192 9999.99 <.0001 62.95 <.0001
EQ 1diff 1.98076 3972.27 <.0001 88.29 <.0001

Appendix 21 – Investment grade QUARTILE 2


CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 939.825 1.00000 | |********************| 0

1 936.517 0.99648 | .|********************| 0.027227

2 932.407 0.99211 | . |********************| 0.047047

3 928.440 0.98789 | . |********************| 0.060603

4 924.588 0.98379 | . |********************| 0.071551

5 920.928 0.97989 | . |********************| 0.080960

6 917.199 0.97593 | . |********************| 0.089320

7 913.548 0.97204 | . |******************* | 0.096903

231
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

8 909.289 0.96751 | . |******************* | 0.103880

9 904.879 0.96282 | . |******************* | 0.110357

10 900.478 0.95813 | . |******************* | 0.116418

11 895.724 0.95308 | . |******************* | 0.122124

12 890.958 0.94800 | . |******************* | 0.127518

13 886.412 0.94317 | . |******************* | 0.132640

14 882.078 0.93856 | . |******************* | 0.137521

15 877.696 0.93389 | . |******************* | 0.142190

16 873.038 0.92894 | . |******************* | 0.146667

17 868.222 0.92381 | . |****************** | 0.150965

18 863.275 0.91855 | . |****************** | 0.155099

19 858.463 0.91343 | . |****************** | 0.159081

20 853.639 0.90830 | . |****************** | 0.162922

21 848.732 0.90308 | . |****************** | 0.166634

22 844.037 0.89808 | . |****************** | 0.170223

23 839.372 0.89312 | . |****************** | 0.173700

24 834.910 0.88837 | . |****************** | 0.177071

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.8347 0.8836 0.63 0.8521

1 0.3738 0.7749 0.23 0.7523

2 0.6588 0.8459 0.45 0.8106

232
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Single Mean 0 -0.9048 0.8949 -0.39 0.9090 0.61 0.9200

1 -2.4059 0.7297 -0.83 0.8110 0.70 0.8929

2 -1.5386 0.8317 -0.59 0.8702 0.63 0.9140

Trend 0 -3.8975 0.8925 -1.33 0.8792 1.52 0.8743

1 -6.1951 0.7289 -1.70 0.7513 1.84 0.8098

2 -4.8444 0.8318 -1.49 0.8339 1.60 0.8584

233
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 3.798696 1.00000 | |********************| 0

1 0.810236 0.21329 | .|**** | 0.027237

2 -0.238015 -.06266 | *|. | 0.028449

3 -0.157488 -.04146 | *|. | 0.028551

4 -0.204482 -.05383 | *|. | 0.028596

5 -0.022999 -.00605 | .|. | 0.028671

6 -0.182041 -.04792 | *|. | 0.028672

7 0.551859 0.14528 | .|*** | 0.028731

8 0.151876 0.03998 | .|* | 0.029271

9 0.083286 0.02192 | .|. | 0.029311

10 0.450323 0.11855 | .|** | 0.029324

11 -0.058759 -.01547 | .|. | 0.029677

12 -0.329272 -.08668 | **|. | 0.029683

13 -0.296283 -.07800 | **|. | 0.029870

14 -0.115810 -.03049 | *|. | 0.030021

15 0.191880 0.05051 | .|* | 0.030044

16 0.118444 0.03118 | .|* | 0.030107

17 0.132146 0.03479 | .|* | 0.030131

18 -0.169679 -.04467 | *|. | 0.030160

19 -0.052144 -.01373 | .|. | 0.030209

20 -0.027704 -.00729 | .|. | 0.030214

21 -0.101196 -.02664 | *|. | 0.030215

22 0.0090787 0.00239 | .|. | 0.030233

23 -0.089485 -.02356 | .|. | 0.030233

234
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 -0.0015458 -.00041 | .|. | 0.030247

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1058.84 0.0001 -29.52 <.0001

1 -1327.23 0.0001 -25.74 <.0001

2 -1333.56 0.0001 -21.09 <.0001

Single Mean 0 -1059.69 0.0001 -29.53 <.0001 436.12 0.0010

1 -1329.91 0.0001 -25.76 <.0001 331.72 0.0010

2 -1338.97 0.0001 -21.11 <.0001 222.74 0.0010

Trend 0 -1060.65 0.0001 -29.54 <.0001 436.44 0.0010

1 -1332.98 0.0001 -25.78 <.0001 332.20 0.0010

2 -1345.32 0.0001 -21.13 <.0001 223.26 0.0010

235
EQUITY

STATIONARITY

Table 2.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 44.844663 1.00000 | |********************| 0

1 44.647222 0.99560 | .|********************| 0.027227

2 44.488345 0.99205 | . |********************| 0.047020

3 44.353527 0.98905 | . |********************| 0.060580

4 44.186297 0.98532 | . |********************| 0.071556

5 44.027244 0.98177 | . |********************| 0.080991

6 43.872389 0.97832 | . |********************| 0.089379

7 43.723788 0.97501 | . |********************| 0.096993

8 43.579815 0.97179 | . |******************* | 0.104005

9 43.420151 0.96823 | . |******************* | 0.110531

10 43.256329 0.96458 | . |******************* | 0.116649

11 43.099049 0.96107 | . |******************* | 0.122419

12 42.930035 0.95731 | . |******************* | 0.127890

13 42.762244 0.95356 | . |******************* | 0.133096

14 42.584697 0.94960 | . |******************* | 0.138067

15 42.418048 0.94589 | . |******************* | 0.142827

16 42.269451 0.94257 | . |******************* | 0.147397

17 42.100513 0.93881 | . |******************* | 0.151800

18 41.920663 0.93480 | . |******************* | 0.156044

19 41.761996 0.93126 | . |******************* | 0.160142

20 41.603973 0.92774 | . |******************* | 0.164107

21 41.436278 0.92400 | . |****************** | 0.167950

236
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

22 41.280537 0.92052 | . |****************** | 0.171677

23 41.114605 0.91682 | . |****************** | 0.175298

24 40.948170 0.91311 | . |****************** | 0.178817

Table 2.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.0683 0.6674 -0.15 0.6311

1 -0.0492 0.6718 -0.13 0.6407

2 -0.0348 0.6751 -0.10 0.6500

Single Mean 0 -4.6507 0.4699 -1.58 0.4929 1.25 0.7503

1 -3.6052 0.5845 -1.40 0.5867 0.98 0.8211

2 -2.9869 0.6585 -1.28 0.6414 0.82 0.8603

Trend 0 -3.2510 0.9260 -0.94 0.9500 1.54 0.8686

1 -1.8202 0.9745 -0.60 0.9785 1.59 0.8587

2 -0.9375 0.9891 -0.34 0.9894 1.82 0.8138

237
Table 2.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.288279 1.00000 | |********************| 0

1 -0.038134 -.13228 | ***|. | 0.027237

2 -0.024430 -.08474 | **|. | 0.027709

3 0.028743 0.09971 | .|** | 0.027901

4 -0.0066580 -.02310 | .|. | 0.028164

5 -0.0052689 -.01828 | .|. | 0.028178

6 -0.0077020 -.02672 | *|. | 0.028187

7 -0.0058400 -.02026 | .|. | 0.028206

8 0.019035 0.06603 | .|* | 0.028216

9 0.00059994 0.00208 | .|. | 0.028331

10 0.00073779 0.00256 | .|. | 0.028331

11 0.010828 0.03756 | .|* | 0.028331

12 -0.0053395 -.01852 | .|. | 0.028368

13 0.010252 0.03556 | .|* | 0.028377

14 -0.0087579 -.03038 | *|. | 0.028410

15 -0.025992 -.09016 | **|. | 0.028434

16 0.019273 0.06685 | .|* | 0.028645

17 0.013756 0.04772 | .|* | 0.028761

18 -0.022627 -.07849 | **|. | 0.028820

19 -0.0016696 -.00579 | .|. | 0.028978

20 0.0090012 0.03122 | .|* | 0.028979

21 -0.013082 -.04538 | *|. | 0.029003

22 0.0082303 0.02855 | .|* | 0.029056

23 0.0046544 0.01615 | .|. | 0.029077

24 -0.0046967 -.01629 | .|. | 0.029084

238
Table 2.4 – 1. Difference
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1525.19 0.0001 -41.91 <.0001

1 -1878.12 0.0001 -30.62 <.0001

2 -1468.18 0.0001 -21.76 <.0001

Single Mean 0 -1525.20 0.0001 -41.89 <.0001 877.60 0.0010

1 -1878.15 0.0001 -30.61 <.0001 468.47 0.0010

2 -1468.22 0.0001 -21.75 <.0001 236.54 0.0010

Trend 0 -1527.71 0.0001 -41.96 <.0001 880.20 0.0010

1 -1890.31 0.0001 -30.70 <.0001 471.10 0.0010

2 -1488.51 0.0001 -21.84 <.0001 238.39 0.0010

239
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0044 6.8841 15.34
1 1 0.0007 0.9235 3.84

VAR-MODEL

Table 3.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.05215 0.05079 1.03 0.3047 1
AR1_1_1 0.21157 0.02696 7.85 0.0001 CDS (t-1)
AR1_1_2 -0.35312 0.09614 -3.67 0.0002 EQ (t-1)
AR2_1_1 -0.1243 0.02678 -4.64 0.0001 CDS (t-2)
AR2_1_2 -0.62769 0.09631 -6.52 0.0001 EQ (t-2)
EQ 1 diff Constant2 0.00031 0.01442 0.02 0.9828 1
AR1_2_1 -0.00471 0.00765 -0.62 0.5381 CDS (t-1)
AR1_2_2 -0.14933 0.0273 -5.47 0.0001 EQ (t-1)
AR2_2_1 0.02337 0.0076 3.07 0.0022 CDS (t-2)
AR2_2_2 -0.09689 0.02735 -3.54 0.0004 EQ (t-2)

GRANGER CAUSALITY TEST

Table 3.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 50.21 <.0001
2 2 9.45 0.0089

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

240
UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.0918 1.86208 33.87 <.0001
EQ 1diff 0.0349 0.52879 12.14 <.0001

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 2.03305 9999.99 <.0001 4.21 0.0405
EQ 1diff 1.98825 2358.18 <.0001 39.46 <.0001

241
Appendix 22 – Investment grade QUARTILE 3
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 3409.107 1.00000 | |********************| 0

1 3402.557 0.99808 | .|********************| 0.027227

2 3393.563 0.99544 | . |********************| 0.047098

3 3383.231 0.99241 | . |********************| 0.060723

4 3372.171 0.98917 | . |********************| 0.071746

5 3361.035 0.98590 | . |********************| 0.081228

6 3349.329 0.98247 | . |********************| 0.089661

7 3337.303 0.97894 | . |********************| 0.097315

8 3324.990 0.97533 | . |********************| 0.104360

9 3312.776 0.97174 | . |******************* | 0.110911

10 3300.526 0.96815 | . |******************* | 0.117052

11 3287.869 0.96444 | . |******************* | 0.122845

12 3274.532 0.96052 | . |******************* | 0.128335

13 3260.739 0.95648 | . |******************* | 0.133558

14 3246.757 0.95238 | . |******************* | 0.138543

15 3232.481 0.94819 | . |******************* | 0.143314

16 3217.700 0.94385 | . |******************* | 0.147891

17 3202.649 0.93944 | . |******************* | 0.152291

18 3187.350 0.93495 | . |******************* | 0.156528

19 3171.853 0.93041 | . |******************* | 0.160614

242
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

20 3156.258 0.92583 | . |******************* | 0.164561

21 3140.903 0.92133 | . |****************** | 0.168378

22 3125.992 0.91695 | . |****************** | 0.172074

23 3111.173 0.91261 | . |****************** | 0.175659

24 3096.307 0.90825 | . |****************** | 0.179139

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 1.2285 0.9436 1.29 0.9505

1 0.4587 0.7967 0.30 0.7721

2 0.5406 0.8174 0.36 0.7888

Single Mean 0 0.3575 0.9727 0.25 0.9756 1.16 0.7736

1 -1.3185 0.8551 -0.57 0.8745 0.58 0.9289

2 -1.1778 0.8693 -0.52 0.8842 0.59 0.9248

Trend 0 -1.8646 0.9735 -0.99 0.9433 1.70 0.8374

1 -4.7262 0.8400 -1.56 0.8071 1.68 0.8412

2 -4.4411 0.8592 -1.51 0.8268 1.60 0.8567

243
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 5.171733 1.00000 | |********************| 0

1 2.288622 0.44253 | .|********* | 0.027237

2 0.892617 0.17260 | .|*** | 0.032131

3 0.385088 0.07446 | .|* | 0.032811

4 0.048013 0.00928 | .|. | 0.032936

5 0.126094 0.02438 | .|. | 0.032938

6 -0.089364 -.01728 | .|. | 0.032952

7 0.034853 0.00674 | .|. | 0.032959

8 0.014309 0.00277 | .|. | 0.032960

9 0.331227 0.06405 | .|* | 0.032960

10 0.541578 0.10472 | .|** | 0.033052

11 0.296426 0.05732 | .|* | 0.033297

12 0.046212 0.00894 | .|. | 0.033370

13 -0.300704 -.05814 | *|. | 0.033372

14 -0.262967 -.05085 | *|. | 0.033447

15 0.166807 0.03225 | .|* | 0.033504

16 0.157594 0.03047 | .|* | 0.033527

17 0.261354 0.05054 | .|* | 0.033548

18 -0.0047610 -.00092 | .|. | 0.033604

19 -0.173335 -.03352 | *|. | 0.033604

20 -0.239515 -.04631 | *|. | 0.033629

21 -0.208829 -.04038 | *|. | 0.033676

22 0.068947 0.01333 | .|. | 0.033712

23 0.197183 0.03813 | .|* | 0.033716

244
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 0.331658 0.06413 | .|* | 0.033748

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -749.227 0.0001 -22.76 <.0001

1 -790.919 0.0001 -19.84 <.0001

2 -766.433 0.0001 -17.19 <.0001

Single Mean 0 -750.478 0.0001 -22.78 <.0001 259.46 0.0010

1 -793.425 0.0001 -19.86 <.0001 197.17 0.0010

2 -770.322 0.0001 -17.22 <.0001 148.22 0.0010

Trend 0 -751.856 0.0001 -22.80 <.0001 259.90 0.0010

1 -796.168 0.0001 -19.88 <.0001 197.63 0.0010

2 -774.769 0.0001 -17.24 <.0001 148.68 0.0010

245
EQUITY

STATIONARITY

Table 2.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 60.148607 1.00000 | |********************| 0

1 59.907426 0.99599 | .|********************| 0.027227

2 59.694891 0.99246 | . |********************| 0.047032

3 59.502837 0.98926 | . |********************| 0.060600

4 59.292962 0.98577 | . |********************| 0.071577

5 59.089419 0.98239 | . |********************| 0.081018

6 58.889341 0.97906 | . |********************| 0.089413

7 58.682918 0.97563 | . |********************| 0.097036

8 58.480458 0.97227 | . |******************* | 0.104053

9 58.264682 0.96868 | . |******************* | 0.110583

10 58.051485 0.96513 | . |******************* | 0.116704

11 57.839884 0.96162 | . |******************* | 0.122478

12 57.620700 0.95797 | . |******************* | 0.127952

13 57.387854 0.95410 | . |******************* | 0.133163

14 57.144967 0.95006 | . |******************* | 0.138137

15 56.905296 0.94608 | . |******************* | 0.142899

16 56.680814 0.94235 | . |******************* | 0.147469

17 56.432884 0.93822 | . |******************* | 0.151867

18 56.173068 0.93390 | . |******************* | 0.156105

19 55.932752 0.92991 | . |******************* | 0.160193

20 55.689769 0.92587 | . |******************* | 0.164146

21 55.439945 0.92172 | . |****************** | 0.167972

246
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

22 55.201929 0.91776 | . |****************** | 0.171681

23 54.948562 0.91355 | . |****************** | 0.175280

24 54.699534 0.90941 | . |****************** | 0.178775

Table 2.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.1361 0.6520 -0.30 0.5783

1 -0.1251 0.6545 -0.30 0.5786

2 -0.1169 0.6564 -0.30 0.5783

Single Mean 0 -3.2369 0.6282 -1.21 0.6729 0.74 0.8829

1 -2.6587 0.6987 -1.08 0.7269 0.59 0.9262

2 -2.2313 0.7510 -0.97 0.7666 0.48 0.9574

Trend 0 -2.4252 0.9582 -0.89 0.9556 1.93 0.7916

1 -1.8251 0.9744 -0.73 0.9702 2.02 0.7715

2 -1.3785 0.9830 -0.59 0.9791 2.17 0.7416

247
Table 2.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.319006 1.00000 | |********************| 0

1 -0.026180 -.08207 | **|. | 0.027237

2 -0.020171 -.06323 | *|. | 0.027420

3 0.015082 0.04728 | .|* | 0.027528

4 -0.0049749 -.01560 | .|. | 0.027588

5 -0.0022632 -.00709 | .|. | 0.027594

6 0.00021297 0.00067 | .|. | 0.027596

7 -0.0054037 -.01694 | .|. | 0.027596

8 0.013583 0.04258 | .|* | 0.027603

9 -0.0062445 -.01957 | .|. | 0.027652

10 0.0089357 0.02801 | .|* | 0.027662

11 0.0056597 0.01774 | .|. | 0.027683

12 0.0085449 0.02679 | .|* | 0.027692

13 0.0093567 0.02933 | .|* | 0.027711

14 0.0022251 0.00698 | .|. | 0.027734

15 -0.025174 -.07891 | **|. | 0.027735

16 0.020368 0.06385 | .|* | 0.027901

17 0.017208 0.05394 | .|* | 0.028010

18 -0.020363 -.06383 | *|. | 0.028086

19 0.0014855 0.00466 | .|. | 0.028194

20 0.00006288 0.00020 | .|. | 0.028194

21 -0.016675 -.05227 | *|. | 0.028194

22 0.013098 0.04106 | .|* | 0.028266

23 0.0015955 0.00500 | .|. | 0.028310

24 -0.0056807 -.01781 | .|. | 0.028311

248
Table 2.4 – 1. Difference
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1457.53 0.0001 -39.84 <.0001

1 -1677.45 0.0001 -28.94 <.0001

2 -1493.55 0.0001 -21.89 <.0001

Single Mean 0 -1457.54 0.0001 -39.82 <.0001 792.91 0.0010

1 -1677.50 0.0001 -28.93 <.0001 418.60 0.0010

2 -1493.67 0.0001 -21.88 <.0001 239.42 0.0010

Trend 0 -1460.81 0.0001 -39.90 <.0001 796.16 0.0010

1 -1690.79 0.0001 -29.04 <.0001 421.53 0.0010

2 -1518.89 0.0001 -21.99 <.0001 241.75 0.0010

249
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0039 5.6938 15.34
1 1 0.0003 0.4602 3.84

VAR-MODEL

Table 3.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.05466 0.05236 1.04 0.2967 1
AR1_1_1 0.36708 0.02727 13.46 0.0001 CDS (t-1)
AR1_1_2 -1.19504 0.09395 -12.72 0.0001 EQ (t-1)
AR2_1_1 0.01132 0.02586 0.44 0.6616 CDS (t-2)
AR2_1_2 -0.57607 0.09845 -5.85 0.0001 EQ (t-2)
EQ 1 diff Constant2 -0.00275 0.01536 -0.18 0.8578 1
AR1_2_1 0.00084 0.008 0.11 0.9161 CDS (t-1)
AR1_2_2 -0.08849 0.02756 -3.21 0.0014 EQ (t-1)
AR2_2_1 0.00586 0.00759 0.77 0.4401 CDS (t-2)
AR2_2_2 -0.0674 0.02888 -2.33 0.0197 EQ (t-2)

GRANGER CAUSALITY TEST

Table 3.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 180.54 <.0001
2 2 0.85 0.6539

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

250
UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.292 1.91835 138.28 <.0001
EQ 1diff 0.0123 0.56269 4.18 0.0023

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 2.00981 9999.99 <.0001 83.55 <.0001
EQ 1diff 1.99412 1357.09 <.0001 19.13 <.0001

251
Appendix 23 – Investment grade QUARTILE 4
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 11327.099 1.00000 | |********************| 0

1 11292.506 0.99695 | .|********************| 0.027227

2 11248.312 0.99304 | . |********************| 0.047062

3 11201.366 0.98890 | . |********************| 0.060637

4 11152.874 0.98462 | . |********************| 0.071601

5 11103.849 0.98029 | . |********************| 0.081019

6 11053.195 0.97582 | . |********************| 0.089380

7 11001.750 0.97128 | . |******************* | 0.096956

8 10952.574 0.96694 | . |******************* | 0.103919

9 10906.185 0.96284 | . |******************* | 0.110387

10 10860.145 0.95878 | . |******************* | 0.116446

11 10811.359 0.95447 | . |******************* | 0.122158

12 10759.727 0.94991 | . |******************* | 0.127567

13 10706.245 0.94519 | . |******************* | 0.132706

14 10651.147 0.94032 | . |******************* | 0.137606

15 10594.302 0.93531 | . |******************* | 0.142290

16 10531.457 0.92976 | . |******************* | 0.146777

17 10461.207 0.92356 | . |****************** | 0.151079

18 10390.410 0.91731 | . |****************** | 0.155208

19 10318.778 0.91098 | . |****************** | 0.159176

252
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

20 10245.692 0.90453 | . |****************** | 0.162995

21 10171.625 0.89799 | . |****************** | 0.166675

22 10099.162 0.89159 | . |****************** | 0.170223

23 10031.251 0.88560 | . |****************** | 0.173651

24 9966.067 0.87984 | . |****************** | 0.176967

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 1.9642 0.9880 1.72 0.9799

1 1.1626 0.9358 0.69 0.8645

2 1.5662 0.9718 1.07 0.9260

Single Mean 0 1.4090 0.9942 0.86 0.9950 1.60 0.6620

1 -0.2591 0.9432 -0.11 0.9469 0.57 0.9300

2 0.5335 0.9784 0.25 0.9758 0.80 0.8656

Trend 0 -1.0972 0.9871 -0.55 0.9813 2.73 0.6289

1 -3.4530 0.9163 -1.17 0.9148 1.78 0.8208

2 -2.3167 0.9616 -0.90 0.9542 1.90 0.7962

253
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 22.444984 1.00000 | |********************| 0

1 8.266438 0.36830 | .|******* | 0.027237

2 0.369527 0.01646 | .|. | 0.030710

3 0.630405 0.02809 | .|* | 0.030716

4 0.921542 0.04106 | .|* | 0.030735

5 1.101543 0.04908 | .|* | 0.030776

6 -0.485999 -.02165 | .|. | 0.030834

7 -2.403179 -.10707 | **|. | 0.030845

8 -1.677017 -.07472 | *|. | 0.031120

9 1.284469 0.05723 | .|* | 0.031253

10 3.461288 0.15421 | .|*** | 0.031330

11 0.910641 0.04057 | .|* | 0.031888

12 -0.261202 -.01164 | .|. | 0.031927

13 -0.406072 -.01809 | .|. | 0.031930

14 -0.767317 -.03419 | *|. | 0.031937

15 1.624650 0.07238 | .|* | 0.031965

16 2.868850 0.12782 | .|*** | 0.032086

17 0.849502 0.03785 | .|* | 0.032461

18 -0.090472 -.00403 | .|. | 0.032494

19 0.857432 0.03820 | .|* | 0.032495

20 0.393329 0.01752 | .|. | 0.032528

21 -0.811579 -.03616 | *|. | 0.032535

22 -1.806075 -.08047 | **|. | 0.032565

23 -0.931477 -.04150 | *|. | 0.032712

254
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 -0.051896 -.00231 | .|. | 0.032751

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -848.525 0.0001 -24.86 <.0001

1 -1116.85 0.0001 -23.56 <.0001

2 -880.501 0.0001 -18.13 <.0001

Single Mean 0 -850.118 0.0001 -24.88 <.0001 309.63 0.0010

1 -1121.45 0.0001 -23.59 <.0001 278.36 0.0010

2 -886.547 0.0001 -18.17 <.0001 164.99 0.0010

Trend 0 -853.348 0.0001 -24.94 <.0001 311.00 0.0010

1 -1130.88 0.0001 -23.68 <.0001 280.26 0.0010

2 -899.012 0.0001 -18.24 <.0001 166.41 0.0010

255
EQUITY

STATIONARITY

Table 2.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 63.889123 1.00000 | |********************| 0

1 63.660551 0.99642 | .|********************| 0.027227

2 63.446510 0.99307 | . |********************| 0.047046

3 63.247145 0.98995 | . |********************| 0.060625

4 63.019905 0.98639 | . |********************| 0.071612

5 62.796971 0.98291 | . |********************| 0.081061

6 62.582782 0.97955 | . |********************| 0.089460

7 62.364244 0.97613 | . |********************| 0.097086

8 62.145482 0.97271 | . |******************* | 0.104108

9 61.912686 0.96906 | . |******************* | 0.110640

10 61.667046 0.96522 | . |******************* | 0.116762

11 61.431077 0.96153 | . |******************* | 0.122534

12 61.180835 0.95761 | . |******************* | 0.128005

13 60.920578 0.95354 | . |******************* | 0.133210

14 60.647591 0.94926 | . |******************* | 0.138177

15 60.384459 0.94514 | . |******************* | 0.142930

16 60.127727 0.94113 | . |******************* | 0.147490

17 59.842956 0.93667 | . |******************* | 0.151876

18 59.545589 0.93201 | . |******************* | 0.156100

19 59.252232 0.92742 | . |******************* | 0.160172

20 58.957161 0.92280 | . |****************** | 0.164104

21 58.653168 0.91805 | . |****************** | 0.167907

256
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

22 58.352582 0.91334 | . |****************** | 0.171587

23 58.036853 0.90840 | . |****************** | 0.175154

24 57.733683 0.90365 | . |****************** | 0.178612

Table 2.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.3621 0.6007 -0.73 0.4011

1 -0.3584 0.6015 -0.77 0.3840

2 -0.3447 0.6046 -0.79 0.3760

Single Mean 0 -0.8789 0.8972 -0.37 0.9113 0.29 0.9900

1 -0.5275 0.9253 -0.24 0.9311 0.30 0.9900

2 -0.2340 0.9447 -0.11 0.9463 0.31 0.9883

Trend 0 -1.7256 0.9766 -0.72 0.9707 1.83 0.8107

1 -1.3360 0.9837 -0.59 0.9789 1.88 0.8003

2 -1.0278 0.9880 -0.49 0.9841 2.07 0.7615

257
Table 2.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.262058 1.00000 | |********************| 0

1 -0.016326 -.06230 | *|. | 0.027237

2 -0.015489 -.05910 | *|. | 0.027342

3 0.019298 0.07364 | .|* | 0.027437

4 -0.0015903 -.00607 | .|. | 0.027583

5 -0.0039281 -.01499 | .|. | 0.027584

6 -0.0032804 -.01252 | .|. | 0.027590

7 -0.0037095 -.01416 | .|. | 0.027594

8 0.014564 0.05558 | .|* | 0.027600

9 0.0051635 0.01970 | .|. | 0.027683

10 0.0048379 0.01846 | .|. | 0.027693

11 0.010887 0.04154 | .|* | 0.027702

12 0.0046501 0.01774 | .|. | 0.027748

13 0.010457 0.03990 | .|* | 0.027757

14 -0.0024690 -.00942 | .|. | 0.027799

15 -0.020638 -.07875 | **|. | 0.027802

16 0.023426 0.08939 | .|** | 0.027967

17 0.017917 0.06837 | .|* | 0.028178

18 -0.0049409 -.01885 | .|. | 0.028301

19 -0.0028220 -.01077 | .|. | 0.028310

20 -0.0005105 -.00195 | .|. | 0.028313

21 -0.0082415 -.03145 | *|. | 0.028313

22 0.011737 0.04479 | .|* | 0.028339

23 -0.0019900 -.00759 | .|. | 0.028391

258
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 -0.0020338 -.00776 | .|. | 0.028393

Table 2.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1430.43 0.0001 -39.04 <.0001

1 -1620.96 0.0001 -28.45 <.0001

2 -1317.88 0.0001 -21.00 <.0001

Single Mean 0 -1430.93 0.0001 -39.04 <.0001 762.25 0.0010

1 -1622.92 0.0001 -28.46 <.0001 404.86 0.0010

2 -1321.19 0.0001 -21.01 <.0001 220.62 0.0010

Trend 0 -1434.25 0.0001 -39.12 <.0001 765.32 0.0010

1 -1635.85 0.0001 -28.55 <.0001 407.69 0.0010

2 -1342.31 0.0001 -21.10 <.0001 222.68 0.0010

259
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0043 5.8276 15.34
1 1 0.0001 0.0943 3.84

VAR-MODEL

Table 3.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.11758 0.11221 1.05 0.2949 1
AR1_1_1 0.341 0.02736 12.46 0.0001 CDS (t-1)
AR1_1_2 -2.91174 0.22259 -13.08 0.0001 EQ (t-1)
AR2_1_1 -0.09187 0.02572 -3.57 0.0004 CDS (t-2)
AR2_1_2 -0.88628 0.23471 -3.78 0.0002 EQ (t-2)
EQ 1 diff Constant2 -0.01102 0.01391 -0.79 0.4285 1
AR1_2_1 0.00906 0.00339 2.67 0.0076 CDS (t-1)
AR1_2_2 -0.05236 0.02759 -1.9 0.0579 EQ (t-1)
AR2_2_1 -0.00645 0.00319 -2.02 0.0432 CDS (t-2)
AR2_2_2 -0.03867 0.02909 -1.33 0.1841 EQ (t-2)

GRANGER CAUSALITY TEST

Table 3.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 176.88 <.0001
2 2 8.37 0.0152

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

260
UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.2509 4.10995 112.29 <.0001
EQ 1diff 0.014 0.50947 4.77 0.0008

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 1.99664 9999.99 <.0001 62.06 <.0001
EQ 1diff 1.9918 654.28 <.0001 16.84 <.0001

261
Appendix 24 – High yield QUARTILE 1
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 7511.242 1.00000 | |********************| 0

1 7500.041 0.99851 | .|********************| 0.027227

2 7481.705 0.99607 | . |********************| 0.047111

3 7458.275 0.99295 | . |********************| 0.060749

4 7432.324 0.98949 | . |********************| 0.071778

5 7405.379 0.98591 | . |********************| 0.081263

6 7376.989 0.98213 | . |********************| 0.089693

7 7347.556 0.97821 | . |********************| 0.097339

8 7317.153 0.97416 | . |******************* | 0.104372

9 7286.253 0.97005 | . |******************* | 0.110908

10 7254.185 0.96578 | . |******************* | 0.117028

11 7219.722 0.96119 | . |******************* | 0.122794

12 7183.941 0.95643 | . |******************* | 0.128250

13 7147.503 0.95157 | . |******************* | 0.133433

14 7111.885 0.94683 | . |******************* | 0.138372

15 7077.137 0.94221 | . |******************* | 0.143094

16 7042.444 0.93759 | . |******************* | 0.147621

17 7007.723 0.93296 | . |******************* | 0.151972

18 6972.435 0.92827 | . |******************* | 0.156160

19 6937.516 0.92362 | . |****************** | 0.160198

262
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

20 6902.524 0.91896 | . |****************** | 0.164098

21 6868.094 0.91438 | . |****************** | 0.167869

22 6835.555 0.91004 | . |****************** | 0.171522

23 6804.287 0.90588 | . |****************** | 0.175064

24 6772.682 0.90167 | . |****************** | 0.178505

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.0352 0.6750 -0.03 0.6724

1 -0.7192 0.5247 -0.45 0.5177

2 -1.0858 0.4585 -0.60 0.4560

Single Mean 0 -1.6961 0.8141 -0.92 0.7810 0.62 0.9152

1 -3.7183 0.5714 -1.37 0.5990 1.03 0.8077

2 -4.8598 0.4490 -1.57 0.4965 1.31 0.7349

Trend 0 -3.6050 0.9086 -1.34 0.8783 0.89 0.9792

1 -7.8572 0.5953 -1.96 0.6203 1.93 0.7904

2 -10.0751 0.4311 -2.21 0.4853 2.44 0.6874

263
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 18.760912 1.00000 | |********************| 0

1 6.960842 0.37103 | .|******* | 0.027237

2 4.611080 0.24578 | .|***** | 0.030759

3 2.371473 0.12640 | .|*** | 0.032183

4 0.947983 0.05053 | .|* | 0.032549

5 0.972947 0.05186 | .|* | 0.032607

6 0.671858 0.03581 | .|* | 0.032668

7 0.636583 0.03393 | .|* | 0.032697

8 0.434368 0.02315 | .|. | 0.032723

9 1.206442 0.06431 | .|* | 0.032735

10 2.528536 0.13478 | .|*** | 0.032829

11 0.980040 0.05224 | .|* | 0.033237

12 0.197468 0.01053 | .|. | 0.033298

13 -1.214275 -.06472 | *|. | 0.033300

14 -1.316276 -.07016 | *|. | 0.033393

15 -0.523750 -.02792 | *|. | 0.033503

16 -0.128012 -.00682 | .|. | 0.033520

17 0.538912 0.02873 | .|* | 0.033521

18 -0.494296 -.02635 | *|. | 0.033539

19 -0.081449 -.00434 | .|. | 0.033554

20 -0.437197 -.02330 | .|. | 0.033555

21 -1.622783 -.08650 | **|. | 0.033567

22 -1.151821 -.06139 | *|. | 0.033732

23 0.438625 0.02338 | .|. | 0.033815

264
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 0.881303 0.04698 | .|* | 0.033827

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -846.762 0.0001 -24.84 <.0001

1 -656.117 0.0001 -18.08 <.0001

2 -654.571 0.0001 -16.17 <.0001

Single Mean 0 -847.016 0.0001 -24.83 <.0001 308.29 0.0010

1 -656.514 0.0001 -18.08 <.0001 163.42 0.0010

2 -655.170 0.0001 -16.17 <.0001 130.78 0.0010

Trend 0 -847.015 0.0001 -24.82 <.0001 308.07 0.0010

1 -656.501 0.0001 -18.07 <.0001 163.31 0.0010

2 -655.147 0.0001 -16.17 <.0001 130.68 0.0010

265
EQUITY

STATIONARITY

Table 2.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 35.457634 1.00000 | |********************| 0

1 35.356951 0.99716 | .|********************| 0.027227

2 35.257364 0.99435 | . |********************| 0.047069

3 35.163974 0.99172 | . |********************| 0.060674

4 35.058019 0.98873 | . |********************| 0.071690

5 34.951731 0.98573 | . |********************| 0.081172

6 34.846329 0.98276 | . |********************| 0.089607

7 34.737026 0.97968 | . |********************| 0.097269

8 34.624533 0.97650 | . |********************| 0.104327

9 34.508339 0.97323 | . |******************* | 0.110896

10 34.387545 0.96982 | . |******************* | 0.117056

11 34.270569 0.96652 | . |******************* | 0.122868

12 34.146322 0.96302 | . |******************* | 0.128381

13 34.017567 0.95939 | . |******************* | 0.133628

14 33.886009 0.95568 | . |******************* | 0.138640

15 33.757680 0.95206 | . |******************* | 0.143441

16 33.630633 0.94847 | . |******************* | 0.148051

17 33.495637 0.94467 | . |******************* | 0.152489

18 33.362426 0.94091 | . |******************* | 0.156767

19 33.225082 0.93704 | . |******************* | 0.160899

20 33.084517 0.93307 | . |******************* | 0.164894

21 32.944227 0.92912 | . |******************* | 0.168763

266
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

22 32.808166 0.92528 | . |******************* | 0.172513

23 32.666047 0.92127 | . |****************** | 0.176153

24 32.524969 0.91729 | . |****************** | 0.179690

Table 2.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.3815 0.5963 -0.81 0.3665

1 -0.3894 0.5946 -0.83 0.3555

2 -0.3783 0.5970 -0.87 0.3411

Single Mean 0 -0.9024 0.8951 -0.41 0.9043 0.36 0.9814

1 -0.8495 0.8997 -0.39 0.9078 0.37 0.9792

2 -0.5493 0.9237 -0.27 0.9265 0.38 0.9780

Trend 0 -3.6489 0.9063 -1.42 0.8558 2.06 0.7646

1 -3.5113 0.9134 -1.38 0.8672 1.97 0.7826

2 -3.0804 0.9337 -1.29 0.8890 2.00 0.7771

267
Table 2.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.124150 1.00000 | |********************| 0

1 -0.0012972 -.01045 | .|. | 0.027237

2 -0.0083755 -.06746 | *|. | 0.027240

3 0.0083042 0.06689 | .|* | 0.027363

4 0.0017225 0.01387 | .|. | 0.027484

5 0.0013413 0.01080 | .|. | 0.027490

6 -0.0002762 -.00222 | .|. | 0.027493

7 0.00046627 0.00376 | .|. | 0.027493

8 0.0021695 0.01748 | .|. | 0.027493

9 0.0013204 0.01064 | .|. | 0.027502

10 0.00020920 0.00169 | .|. | 0.027505

11 0.0055332 0.04457 | .|* | 0.027505

12 0.00083788 0.00675 | .|. | 0.027558

13 0.0020005 0.01611 | .|. | 0.027559

14 -0.0002371 -.00191 | .|. | 0.027566

15 -0.0074222 -.05978 | *|. | 0.027566

16 0.0062582 0.05041 | .|* | 0.027662

17 0.0011039 0.00889 | .|. | 0.027731

18 0.0066736 0.05375 | .|* | 0.027733

19 0.0015350 0.01236 | .|. | 0.027810

20 -0.0041110 -.03311 | *|. | 0.027814

21 -0.0063512 -.05116 | *|. | 0.027843

22 0.0040707 0.03279 | .|* | 0.027913

23 0.0028116 0.02265 | .|. | 0.027941

268
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 -0.0012224 -.00985 | .|. | 0.027955

Table 2.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1360.51 0.0001 -37.07 <.0001

1 -1555.44 0.0001 -27.87 <.0001

2 -1269.89 0.0001 -20.73 <.0001

Single Mean 0 -1361.08 0.0001 -37.07 <.0001 687.07 0.0010

1 -1557.60 0.0001 -27.88 <.0001 388.59 0.0010

2 -1273.47 0.0001 -20.74 <.0001 215.10 0.0010

Trend 0 -1363.16 0.0001 -37.11 <.0001 688.63 0.0010

1 -1565.30 0.0001 -27.93 <.0001 390.13 0.0010

2 -1285.92 0.0001 -20.80 <.0001 216.25 0.0010

269
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.007 9.6759 15.34
1 1 0.0002 0.2709 3.84

VAR-MODEL

Table 3.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.01617 0.10558 0.15 0.8783 1
AR1_1_1 0.27135 0.02701 10.05 0.0001 CDS (t-1)
AR1_1_2 -2.52039 0.30161 -8.36 0.0001 EQ (t-1)
AR2_1_1 0.13491 0.0263 5.13 0.0001 CDS (t-2)
AR2_1_2 -1.4798 0.30795 -4.81 0.0001 EQ (t-2)
EQ 1 diff Constant2 -0.00775 0.0096 -0.81 0.42 1
AR1_2_1 0.00027 0.00246 0.11 0.9113 CDS (t-1)
AR1_2_2 -0.01073 0.02744 -0.39 0.6957 EQ (t-1)
AR2_2_1 -0.00289 0.00239 -1.21 0.2265 CDS (t-2)
AR2_2_2 -0.07058 0.02801 -2.52 0.0119 EQ (t-2)

GRANGER CAUSALITY TEST

Table 3.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 90.2 <.0001
2 2 1.59 0.4526

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

270
UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.2049 3.87107 86.37 <.0001
EQ 1diff 0.0059 0.35212 1.98 0.0958

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 2.00789 9999.99 <.0001 26.68 <.0001
EQ 1diff 1.99178 173.71 <.0001 6.76 0.0094

271
Appendix 25 – High yield QUARTILE 2
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 40654.840 1.00000 | |********************| 0

1 40587.907 0.99835 | .|********************| 0.027227

2 40501.111 0.99622 | . |********************| 0.047106

3 40393.185 0.99356 | . |********************| 0.060749

4 40275.889 0.99068 | . |********************| 0.071791

5 40155.150 0.98771 | . |********************| 0.081296

6 40031.031 0.98466 | . |********************| 0.089752

7 39901.884 0.98148 | . |********************| 0.097431

8 39767.784 0.97818 | . |********************| 0.104503

9 39631.328 0.97482 | . |******************* | 0.111084

10 39495.327 0.97148 | . |******************* | 0.117254

11 39354.948 0.96803 | . |******************* | 0.123076

12 39206.516 0.96438 | . |******************* | 0.128596

13 39053.007 0.96060 | . |******************* | 0.133850

14 38898.231 0.95679 | . |******************* | 0.138866

15 38741.388 0.95293 | . |******************* | 0.143670

16 38581.419 0.94900 | . |******************* | 0.148281

17 38416.564 0.94494 | . |******************* | 0.152717

18 38250.884 0.94087 | . |******************* | 0.156992

19 38085.678 0.93681 | . |******************* | 0.161117

272
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

20 37921.030 0.93276 | . |******************* | 0.165106

21 37759.368 0.92878 | . |******************* | 0.168967

22 37605.002 0.92498 | . |****************** | 0.172710

23 37454.510 0.92128 | . |****************** | 0.176344

24 37305.164 0.91761 | . |****************** | 0.179877

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 1.2490 0.9458 1.40 0.9605

1 0.6976 0.8547 0.51 0.8258

2 0.5296 0.8147 0.36 0.7892

Single Mean 0 0.2879 0.9701 0.23 0.9746 1.59 0.6626

1 -0.7823 0.9054 -0.41 0.9052 0.74 0.8808

2 -1.0905 0.8778 -0.53 0.8822 0.71 0.8898

Trend 0 -2.4481 0.9575 -1.26 0.8975 1.69 0.8396

1 -5.0043 0.8204 -1.68 0.7622 1.77 0.8228

2 -5.6430 0.7725 -1.77 0.7217 1.86 0.8056

273
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 46.672310 1.00000 | |********************| 0

1 18.751123 0.40176 | .|******** | 0.027237

2 9.903302 0.21219 | .|**** | 0.031326

3 5.529930 0.11848 | .|** | 0.032375

4 2.583513 0.05535 | .|* | 0.032695

5 2.167791 0.04645 | .|* | 0.032764

6 1.792537 0.03841 | .|* | 0.032813

7 0.287380 0.00616 | .|. | 0.032846

8 1.964802 0.04210 | .|* | 0.032847

9 3.060730 0.06558 | .|* | 0.032887

10 5.606956 0.12013 | .|** | 0.032984

11 3.779712 0.08098 | .|** | 0.033307

12 1.045868 0.02241 | .|. | 0.033453

13 -3.352059 -.07182 | *|. | 0.033464

14 -2.338052 -.05010 | *|. | 0.033578

15 -0.492114 -.01054 | .|. | 0.033634

16 1.896041 0.04062 | .|* | 0.033636

17 1.660646 0.03558 | .|* | 0.033672

18 -1.200615 -.02572 | *|. | 0.033700

19 -2.772783 -.05941 | *|. | 0.033715

20 -4.146000 -.08883 | **|. | 0.033792

21 -5.682302 -.12175 | **|. | 0.033965

22 -3.828857 -.08204 | **|. | 0.034287

23 -1.204143 -.02580 | *|. | 0.034433

274
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 0.982835 0.02106 | .|. | 0.034447

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -803.778 0.0001 -23.92 <.0001

1 -703.868 0.0001 -18.60 <.0001

2 -671.046 0.0001 -16.18 <.0001

Single Mean 0 -805.660 0.0001 -23.95 <.0001 286.82 0.0010

1 -707.110 0.0001 -18.63 <.0001 173.55 0.0010

2 -675.854 0.0001 -16.21 <.0001 131.40 0.0010

Trend 0 -806.739 0.0001 -23.96 <.0001 287.13 0.0010

1 -709.089 0.0001 -18.64 <.0001 173.79 0.0010

2 -678.905 0.0001 -16.22 <.0001 131.63 0.0010

275
EQUITY

STATIONARITY

Table 2.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 124.771 1.00000 | |********************| 0

1 124.330 0.99647 | .|********************| 0.027227

2 123.897 0.99300 | . |********************| 0.047047

3 123.466 0.98954 | . |********************| 0.060624

4 123.039 0.98612 | . |********************| 0.071603

5 122.621 0.98277 | . |********************| 0.081048

6 122.208 0.97946 | . |********************| 0.089446

7 121.769 0.97594 | . |********************| 0.097072

8 121.328 0.97241 | . |******************* | 0.104092

9 120.866 0.96870 | . |******************* | 0.110621

10 120.378 0.96479 | . |******************* | 0.116740

11 119.909 0.96103 | . |******************* | 0.122508

12 119.426 0.95716 | . |******************* | 0.127975

13 118.923 0.95313 | . |******************* | 0.133176

14 118.406 0.94899 | . |******************* | 0.138140

15 117.886 0.94482 | . |******************* | 0.142891

16 117.359 0.94060 | . |******************* | 0.147449

17 116.791 0.93604 | . |******************* | 0.151832

18 116.223 0.93149 | . |******************* | 0.156051

19 115.668 0.92704 | . |******************* | 0.160120

20 115.109 0.92257 | . |****************** | 0.164050

21 114.532 0.91794 | . |****************** | 0.167852

276
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

22 113.956 0.91332 | . |****************** | 0.171533

23 113.352 0.90848 | . |****************** | 0.175101

24 112.764 0.90377 | . |****************** | 0.178561

Table 2.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.4375 0.5839 -0.79 0.3730

1 -0.4333 0.5848 -0.81 0.3674

2 -0.4356 0.5843 -0.82 0.3619

Single Mean 0 0.1368 0.9639 0.06 0.9630 0.36 0.9817

1 0.2589 0.9690 0.13 0.9677 0.39 0.9771

2 0.3163 0.9712 0.16 0.9697 0.41 0.9730

Trend 0 -2.9444 0.9394 -1.21 0.9079 3.12 0.5491

1 -2.7609 0.9466 -1.16 0.9165 3.18 0.5377

2 -2.6538 0.9505 -1.13 0.9222 3.16 0.5412

277
Table 2.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.410194 1.00000 | |********************| 0

1 -0.010679 -.02603 | *|. | 0.027237

2 -0.0040858 -.00996 | .|. | 0.027255

3 -0.014974 -.03650 | *|. | 0.027258

4 -0.0060378 -.01472 | .|. | 0.027294

5 0.0064897 0.01582 | .|. | 0.027300

6 0.015714 0.03831 | .|* | 0.027307

7 0.0013094 0.00319 | .|. | 0.027347

8 0.015492 0.03777 | .|* | 0.027347

9 0.013021 0.03174 | .|* | 0.027386

10 0.0084513 0.02060 | .|. | 0.027413

11 0.0028317 0.00690 | .|. | 0.027424

12 0.0075123 0.01831 | .|. | 0.027426

13 0.0064374 0.01569 | .|. | 0.027435

14 0.0054731 0.01334 | .|. | 0.027441

15 -0.012288 -.02996 | *|. | 0.027446

16 0.037654 0.09180 | .|** | 0.027470

17 0.012509 0.03049 | .|* | 0.027697

18 -0.010617 -.02588 | *|. | 0.027722

19 -0.0030667 -.00748 | .|. | 0.027740

20 0.0032489 0.00792 | .|. | 0.027741

21 -0.010260 -.02501 | *|. | 0.027743

22 0.028973 0.07063 | .|* | 0.027760

23 -0.0038685 -.00943 | .|. | 0.027893

278
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 0.0025134 0.00613 | .|. | 0.027895

Table 2.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1381.33 0.0001 -37.63 <.0001

1 -1408.51 0.0001 -26.52 <.0001

2 -1576.15 0.0001 -22.26 <.0001

Single Mean 0 -1382.07 0.0001 -37.64 <.0001 708.44 0.0010

1 -1410.92 0.0001 -26.53 <.0001 351.98 0.0010

2 -1582.09 0.0001 -22.28 <.0001 248.16 0.0010

Trend 0 -1387.01 0.0001 -37.76 <.0001 713.05 0.0010

1 -1426.73 0.0001 -26.67 <.0001 355.56 0.0010

2 -1621.92 0.0001 -22.44 <.0001 251.78 0.0010

279
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0071 9.6885 15.34
1 1 0 0.0577 3.84

VAR-MODEL

Table 3.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.16664 0.16485 1.01 0.3123 1
AR1_1_1 0.3756 0.02423 15.5 0.0001 CDS (t-1)
AR1_1_2 -2.59833 0.25847 -10.05 0.0001 EQ (t-1)
EQ 1 diff Constant2 -0.01628 0.01748 -0.93 0.3519 1
AR1_2_1 0.00341 0.00257 1.33 0.1847 CDS (t-1)
AR1_2_2 -0.02211 0.02741 -0.81 0.42 EQ (t-1)

GRANGER CAUSALITY TEST

Table 3.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 101.06 <.0001
2 2 1.76 0.1845

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

280
UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.2201 6.04214 189.65 <.0001
EQ 1diff 0.002 0.64076 1.34 0.263

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.5
Variable Durbin Normality Pr > Chi Sq F Value Pr > F
Watson Chi Square
CDS 1diff 2.12578 9999.99 <.0001 33.22 <.0001
EQ 1diff 1.99848 1198.79 <.0001 3.77 0.0523

281
Appendix 26 – High yield QUARTILE 3
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 61724.615 1.00000 | |********************| 0

1 61557.140 0.99729 | .|********************| 0.027227

2 61355.813 0.99403 | . |********************| 0.047073

3 61126.710 0.99031 | . |********************| 0.060669

4 60875.636 0.98625 | . |********************| 0.071657

5 60617.874 0.98207 | . |********************| 0.081098

6 60335.315 0.97749 | . |********************| 0.089480

7 60050.098 0.97287 | . |******************* | 0.097074

8 59766.513 0.96828 | . |******************* | 0.104051

9 59486.621 0.96374 | . |******************* | 0.110528

10 59212.347 0.95930 | . |******************* | 0.116591

11 58934.959 0.95480 | . |******************* | 0.122303

12 58648.056 0.95016 | . |******************* | 0.127709

13 58359.581 0.94548 | . |******************* | 0.132846

14 58065.359 0.94072 | . |******************* | 0.137744

15 57771.391 0.93595 | . |******************* | 0.142427

16 57475.978 0.93117 | . |******************* | 0.146915

17 57169.727 0.92621 | . |******************* | 0.151227

18 56863.886 0.92125 | . |****************** | 0.155375

19 56554.179 0.91623 | . |****************** | 0.159373

282
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

20 56247.402 0.91126 | . |****************** | 0.163231

21 55937.269 0.90624 | . |****************** | 0.166959

22 55631.982 0.90129 | . |****************** | 0.170567

23 55331.775 0.89643 | . |****************** | 0.174061

24 55032.991 0.89159 | . |****************** | 0.177451

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 1.5577 0.9713 1.69 0.9782

1 1.2597 0.9470 1.06 0.9255

2 1.0909 0.9264 0.83 0.8908

Single Mean 0 0.9077 0.9873 0.57 0.9890 1.55 0.6744

1 0.0686 0.9608 0.03 0.9604 0.83 0.8599

2 -0.4507 0.9308 -0.20 0.9360 0.70 0.8913

Trend 0 -1.6432 0.9782 -0.79 0.9655 1.92 0.7932

1 -3.0705 0.9341 -1.15 0.9190 1.63 0.8518

2 -3.8497 0.8953 -1.30 0.8863 1.59 0.8603

283
Table 1.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 114.719 1.00000 | |********************| 0

1 27.603739 0.24062 | .|***** | 0.027237

2 17.055785 0.14867 | .|*** | 0.028771

3 13.711945 0.11953 | .|** | 0.029335

4 13.554669 0.11816 | .|** | 0.029694

5 19.797134 0.17257 | .|*** | 0.030041

6 3.084640 0.02689 | .|* | 0.030767

7 0.099773 0.00087 | .|. | 0.030785

8 2.469368 0.02153 | .|. | 0.030785

9 0.017917 0.00016 | .|. | 0.030796

10 4.970823 0.04333 | .|* | 0.030796

11 2.313235 0.02016 | .|. | 0.030841

12 0.607009 0.00529 | .|. | 0.030851

13 0.607911 0.00530 | .|. | 0.030852

14 -6.495992 -.05663 | *|. | 0.030852

15 -2.830516 -.02467 | .|. | 0.030929

16 5.547689 0.04836 | .|* | 0.030944

17 0.762149 0.00664 | .|. | 0.031000

18 1.194696 0.01041 | .|. | 0.031001

19 -3.111747 -.02713 | *|. | 0.031004

20 -2.233390 -.01947 | .|. | 0.031021

21 -2.991302 -.02608 | *|. | 0.031030

22 -3.210828 -.02799 | *|. | 0.031047

23 1.294386 0.01128 | .|. | 0.031065

284
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 2.951091 0.02572 | .|* | 0.031068

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1020.21 0.0001 -28.62 <.0001

1 -836.001 0.0001 -20.37 <.0001

2 -695.437 0.0001 -16.52 <.0001

Single Mean 0 -1022.33 0.0001 -28.65 <.0001 410.54 0.0010

1 -839.822 0.0001 -20.41 <.0001 208.26 0.0010

2 -700.351 0.0001 -16.55 <.0001 136.98 0.0010

Trend 0 -1024.86 0.0001 -28.70 <.0001 411.80 0.0010

1 -844.266 0.0001 -20.45 <.0001 209.07 0.0010

2 -706.138 0.0001 -16.59 <.0001 137.62 0.0010

285
EQUITY

STATIONARITY

Table 2.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 41.552011 1.00000 | |********************| 0

1 41.398080 0.99630 | .|********************| 0.027227

2 41.239009 0.99247 | . |********************| 0.047042

3 41.083645 0.98873 | . |********************| 0.060607

4 40.914119 0.98465 | . |********************| 0.071572

5 40.750021 0.98070 | . |********************| 0.080994

6 40.581055 0.97663 | . |********************| 0.089364

7 40.409389 0.97250 | . |******************* | 0.096954

8 40.237494 0.96836 | . |******************* | 0.103933

9 40.067248 0.96427 | . |******************* | 0.110419

10 39.885365 0.95989 | . |******************* | 0.116494

11 39.709277 0.95565 | . |******************* | 0.122217

12 39.525711 0.95123 | . |******************* | 0.127636

13 39.337288 0.94670 | . |******************* | 0.132787

14 39.147224 0.94213 | . |******************* | 0.137700

15 38.956279 0.93753 | . |******************* | 0.142398

16 38.761295 0.93284 | . |******************* | 0.146902

17 38.556757 0.92792 | . |******************* | 0.151230

18 38.349716 0.92293 | . |****************** | 0.155393

19 38.142904 0.91796 | . |****************** | 0.159405

20 37.927667 0.91278 | . |****************** | 0.163276

21 37.707491 0.90748 | . |****************** | 0.167016

286
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

22 37.490888 0.90226 | . |****************** | 0.170632

23 37.272274 0.89700 | . |****************** | 0.174133

24 37.066462 0.89205 | . |****************** | 0.177525

Table 2.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.5311 0.5634 -0.83 0.3584

1 -0.5440 0.5607 -0.84 0.3528

2 -0.5400 0.5615 -0.86 0.3451

Single Mean 0 -1.1863 0.8685 -0.48 0.8931 0.38 0.9780

1 -1.2368 0.8634 -0.49 0.8903 0.39 0.9759

2 -1.0565 0.8810 -0.43 0.9014 0.39 0.9762

Trend 0 -4.4263 0.8601 -1.50 0.8288 2.20 0.7367

1 -4.4472 0.8588 -1.49 0.8313 2.13 0.7502

2 -4.1590 0.8770 -1.44 0.8499 2.08 0.7609

287
Table 2.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.189931 1.00000 | |********************| 0

1 0.0015656 0.00824 | .|. | 0.027237

2 -0.0053765 -.02831 | *|. | 0.027239

3 0.0084861 0.04468 | .|* | 0.027260

4 -0.0031889 -.01679 | .|. | 0.027315

5 0.0057777 0.03042 | .|* | 0.027322

6 0.00017547 0.00092 | .|. | 0.027347

7 0.0014205 0.00748 | .|. | 0.027347

8 -0.0013416 -.00706 | .|. | 0.027349

9 0.0090502 0.04765 | .|* | 0.027350

10 0.0040392 0.02127 | .|. | 0.027412

11 0.0049598 0.02611 | .|* | 0.027424

12 0.0036766 0.01936 | .|. | 0.027443

13 0.00006975 0.00037 | .|. | 0.027453

14 0.0017492 0.00921 | .|. | 0.027453

15 -0.0054486 -.02869 | *|. | 0.027455

16 0.0084391 0.04443 | .|* | 0.027477

17 0.0047749 0.02514 | .|* | 0.027530

18 -0.0005983 -.00315 | .|. | 0.027547

19 0.0055815 0.02939 | .|* | 0.027548

20 0.0042298 0.02227 | .|. | 0.027571

21 -0.0070671 -.03721 | *|. | 0.027584

22 0.0016596 0.00874 | .|. | 0.027622

23 -0.0066275 -.03489 | *|. | 0.027624

288
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 -0.0083826 -.04413 | *|. | 0.027656

Table 2.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1335.33 0.0001 -36.36 <.0001

1 -1411.39 0.0001 -26.54 <.0001

2 -1230.39 0.0001 -20.50 <.0001

Single Mean 0 -1335.89 0.0001 -36.37 <.0001 661.25 0.0010

1 -1413.24 0.0001 -26.55 <.0001 352.36 0.0010

2 -1233.57 0.0001 -20.51 <.0001 210.31 0.0010

Trend 0 -1338.04 0.0001 -36.41 <.0001 662.77 0.0010

1 -1420.08 0.0001 -26.60 <.0001 353.69 0.0010

2 -1245.26 0.0001 -20.56 <.0001 211.40 0.0010

289
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0118 16.0802 15.34
1 1 0 0.0256 3.84

VECM-MODEL

Table 3.2 – Long-Run Parameter Beta Estimates When RANK = 1


Variable 1
CDS 0.00997
EQUITY 0.41243

Table 3.3 – Adjustment Coefficient Alpha Estimates When RANK = 1


Variable 1
CDS -0.99311
EQUITY -0.01783

Table 3.4
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS Constant1 13.60703 3.60652 3.77 0.0002 1
AR1_1_1 -0.00991 0.00268 CDS (t-1)
AR1_1_2 -0.40959 0.1107 EQ (t-1)
AR2_1_1 0.23625 0.02509 9.42 0.0001 D_CDS (t-1)
AR2_1_2 -6.98465 0.62163 -11.24 0.0001 D_EQ (t-1)
EQ Constant2 0.22946 0.15935 1.44 0.1501 1
AR1_2_1 -0.00018 0.00012 CDS (t-1)
AR1_2_2 -0.00736 0.00489 EQ (t-1)
AR2_2_1 0.00142 0.00111 1.28 0.2 D_CDS (t-1)
AR2_2_2 0.01404 0.02747 0.51 0.6093 D_EQ (t-1)

290
GRANGER CAUSALITY TEST

Table 3.5
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 154.09 <.0001
2 2 3.83 0.1476

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

UNIVARIATE MODEL ANOVA DIAGNOSTICS


Table 3.6
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.1547 9.85109 61.4 <.0001
EQ 1diff 0.0031 0.43525 1.03 0.3898

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.7
Variable Durbin Watson Normality Chi Pr > Chi Sq F Value Pr > F
Square
CDS 1diff 2.08987 8697.37 <.0001 85.01 <.0001
EQ 1diff 1.99759 9999.99 <.0001 0 0.9901

291
Appendix 27 – High yield QUARTILE 4
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 344555 1.00000 | |********************| 0

1 343966 0.99829 | .|********************| 0.027227

2 343216 0.99612 | . |********************| 0.047104

3 342326 0.99353 | . |********************| 0.060745

4 341308 0.99058 | . |********************| 0.071787

5 340206 0.98738 | . |********************| 0.081290

6 339005 0.98389 | . |********************| 0.089741

7 337706 0.98012 | . |********************| 0.097410

8 336354 0.97620 | . |********************| 0.104465

9 334967 0.97218 | . |******************* | 0.111022

10 333560 0.96809 | . |******************* | 0.117162

11 332139 0.96397 | . |******************* | 0.122949

12 330660 0.95967 | . |******************* | 0.128430

13 329141 0.95526 | . |******************* | 0.133640

14 327550 0.95065 | . |******************* | 0.138609

15 325937 0.94597 | . |******************* | 0.143361

16 324258 0.94109 | . |******************* | 0.147916

17 322538 0.93610 | . |******************* | 0.152289

18 320754 0.93092 | . |******************* | 0.156497

19 318930 0.92563 | . |******************* | 0.160549

292
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

20 317070 0.92023 | . |****************** | 0.164458

21 315211 0.91484 | . |****************** | 0.168231

22 313411 0.90961 | . |****************** | 0.171880

23 311629 0.90444 | . |****************** | 0.175412

24 309852 0.89928 | . |****************** | 0.178835

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 1.3060 0.9517 1.45 0.9639

1 0.8157 0.8798 0.62 0.8491

2 0.7015 0.8556 0.50 0.8238

Single Mean 0 0.3243 0.9715 0.26 0.9762 1.70 0.6354

1 -0.6197 0.9185 -0.34 0.9167 0.84 0.8573

2 -0.8321 0.9012 -0.43 0.9014 0.79 0.8685

Trend 0 -1.9662 0.9711 -1.07 0.9321 1.48 0.8821

1 -3.8753 0.8938 -1.44 0.8487 1.42 0.8946

2 -4.2603 0.8708 -1.50 0.8283 1.45 0.8873

293
Table 1.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 396.325 1.00000 | |********************| 0

1 144.393 0.36433 | .|******* | 0.027237

2 68.832811 0.17368 | .|*** | 0.030639

3 57.224735 0.14439 | .|*** | 0.031361

4 51.652492 0.13033 | .|*** | 0.031851

5 69.867417 0.17629 | .|**** | 0.032244

6 43.018773 0.10854 | .|** | 0.032951

7 34.404776 0.08681 | .|** | 0.033215

8 39.234585 0.09900 | .|** | 0.033383

9 38.980332 0.09835 | .|** | 0.033600

10 18.962957 0.04785 | .|* | 0.033813

11 18.940390 0.04779 | .|* | 0.033863

12 20.423191 0.05153 | .|* | 0.033913

13 18.888973 0.04766 | .|* | 0.033971

14 -2.943059 -.00743 | .|. | 0.034021

15 14.709741 0.03712 | .|* | 0.034022

16 24.564671 0.06198 | .|* | 0.034052

17 52.625897 0.13278 | .|*** | 0.034136

18 14.989989 0.03782 | .|* | 0.034517

19 -9.236894 -.02331 | .|. | 0.034548

20 -18.135824 -.04576 | *|. | 0.034559

21 -25.759664 -.06500 | *|. | 0.034604

22 -2.464054 -.00622 | .|. | 0.034695

23 6.185165 0.01561 | .|. | 0.034695

294
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 23.113500 0.05832 | .|* | 0.034701

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -853.717 0.0001 -24.98 <.0001

1 -770.305 0.0001 -19.51 <.0001

2 -623.327 0.0001 -15.68 <.0001

Single Mean 0 -855.870 0.0001 -25.02 <.0001 312.94 0.0010

1 -774.217 0.0001 -19.55 <.0001 191.11 0.0010

2 -628.235 0.0001 -15.72 <.0001 123.60 0.0010

Trend 0 -857.003 0.0001 -25.03 <.0001 313.26 0.0010

1 -776.351 0.0001 -19.56 <.0001 191.38 0.0010

2 -630.844 0.0001 -15.73 <.0001 123.77 0.0010

295
EQUITY

STATIONARITY

Table 2.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 89.397348 1.00000 | |********************| 0

1 89.197337 0.99776 | .|********************| 0.027227

2 88.984572 0.99538 | . |********************| 0.047088

3 88.770199 0.99298 | . |********************| 0.060714

4 88.553808 0.99056 | . |********************| 0.071750

5 88.335740 0.98812 | . |********************| 0.081257

6 88.120787 0.98572 | . |********************| 0.089724

7 87.901114 0.98326 | . |********************| 0.097421

8 87.667966 0.98066 | . |********************| 0.104519

9 87.431250 0.97801 | . |********************| 0.111131

10 87.188836 0.97530 | . |********************| 0.117338

11 86.949015 0.97261 | . |******************* | 0.123200

12 86.701332 0.96984 | . |******************* | 0.128766

13 86.443349 0.96696 | . |******************* | 0.134072

14 86.180348 0.96401 | . |******************* | 0.139146

15 85.919210 0.96109 | . |******************* | 0.144012

16 85.659966 0.95819 | . |******************* | 0.148690

17 85.395544 0.95524 | . |******************* | 0.153199

18 85.136602 0.95234 | . |******************* | 0.157553

19 84.880944 0.94948 | . |******************* | 0.161764

20 84.619785 0.94656 | . |******************* | 0.165843

21 84.355297 0.94360 | . |******************* | 0.169801

296
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

22 84.094523 0.94068 | . |******************* | 0.173645

23 83.834912 0.93778 | . |******************* | 0.177382

24 83.577420 0.93490 | . |******************* | 0.181020

Table 2.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.8726 0.4956 -1.85 0.0619

1 -0.9156 0.4878 -1.83 0.0647

2 -0.9141 0.4881 -1.80 0.0691

Single Mean 0 0.3719 0.9732 0.28 0.9772 2.20 0.5059

1 0.2002 0.9666 0.14 0.9688 2.02 0.5519

2 0.1842 0.9659 0.13 0.9679 1.95 0.5715

Trend 0 -2.3231 0.9614 -1.07 0.9327 1.27 0.9239

1 -2.5374 0.9546 -1.10 0.9273 1.14 0.9480

2 -2.6173 0.9518 -1.12 0.9240 1.16 0.9440

297
Table 2.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.117697 1.00000 | |********************| 0

1 0.0067586 0.05742 | .|* | 0.027237

2 0.0018342 0.01558 | .|. | 0.027326

3 -0.0025664 -.02180 | .|. | 0.027333

4 0.00091810 0.00780 | .|. | 0.027346

5 -0.0019025 -.01616 | .|. | 0.027348

6 -0.0023301 -.01980 | .|. | 0.027355

7 0.0086935 0.07386 | .|* | 0.027365

8 0.00046719 0.00397 | .|. | 0.027513

9 0.0018450 0.01568 | .|. | 0.027513

10 0.00031439 0.00267 | .|. | 0.027520

11 0.0055094 0.04681 | .|* | 0.027520

12 0.0035612 0.03026 | .|* | 0.027579

13 0.0034600 0.02940 | .|* | 0.027604

14 -0.0002438 -.00207 | .|. | 0.027627

15 -0.0028602 -.02430 | .|. | 0.027627

16 0.0020053 0.01704 | .|. | 0.027643

17 0.0014771 0.01255 | .|. | 0.027651

18 0.0054540 0.04634 | .|* | 0.027655

19 0.0035528 0.03019 | .|* | 0.027712

20 -0.0033426 -.02840 | *|. | 0.027737

21 -0.0074319 -.06314 | *|. | 0.027758

22 -0.0025698 -.02183 | .|. | 0.027865

23 0.0011668 0.00991 | .|. | 0.027877

298
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 0.00097040 0.00824 | .|. | 0.027880

Table 2.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1265.45 0.0001 -34.56 <.0001

1 -1226.13 0.0001 -24.75 <.0001

2 -1302.15 0.0001 -20.92 <.0001

Single Mean 0 -1269.64 0.0001 -34.66 <.0001 600.67 0.0010

1 -1237.38 0.0001 -24.85 <.0001 308.86 0.0010

2 -1326.17 0.0001 -21.05 <.0001 221.47 0.0010

Trend 0 -1270.88 0.0001 -34.68 <.0001 601.23 0.0010

1 -1240.70 0.0001 -24.88 <.0001 309.41 0.0010

2 -1333.28 0.0001 -21.07 <.0001 222.03 0.0010

299
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0048 6.435 15.34
1 1 0 0.0012 3.84

VAR-MODEL

Table 3.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.49202 0.50237 0.98 0.3276 1
AR1_1_1 0.35887 0.02519 14.25 0.0001 CDS (t-1)
AR1_1_2 -7.4463 1.46134 -5.1 0.0001 EQ (t-1)
EQ 1 diff Constant2 -0.01866 0.00936 -1.99 0.0465 1
AR1_2_1 -0.00011 0.00047 -0.24 0.8107 CDS (t-1)
AR1_2_2 0.05714 0.02723 2.1 0.0361 EQ (t-1)

GRANGER CAUSALITY TEST

Table 3.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 25.96 <.0001
2 2 0.06 0.8107

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

300
UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.1493 18.38679 117.94 <.0001
EQ 1diff 0.0033 0.34263 2.26 0.105

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.5
Variable Durbin Watson Normality Chi Pr > Chi Sq F Value Pr > F
Square
CDS 1diff 2.04406 9999.99 <.0001 20.07 <.0001
EQ 1diff 2.00051 9999.99 <.0001 5.25 0.0221

301
Appendix 28 - COMMUNICATION
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 14682.404 1.00000 | |********************| 0

1 14652.068 0.99793 | .|********************| 0.027227

2 14616.901 0.99554 | . |********************| 0.047093

3 14574.863 0.99268 | . |********************| 0.060722

4 14527.620 0.98946 | . |********************| 0.071750

5 14481.698 0.98633 | . |********************| 0.081238

6 14430.502 0.98284 | . |********************| 0.089677

7 14382.557 0.97958 | . |********************| 0.097335

8 14334.905 0.97633 | . |********************| 0.104387

9 14287.629 0.97311 | . |******************* | 0.110950

10 14242.613 0.97005 | . |******************* | 0.117106

11 14193.427 0.96670 | . |******************* | 0.122918

12 14143.129 0.96327 | . |******************* | 0.128430

13 14091.049 0.95972 | . |******************* | 0.133679

14 14035.766 0.95596 | . |******************* | 0.138693

15 13980.491 0.95219 | . |******************* | 0.143494

16 13916.921 0.94786 | . |******************* | 0.148104

17 13851.838 0.94343 | . |******************* | 0.152534

18 13786.363 0.93897 | . |******************* | 0.156800

19 13719.664 0.93443 | . |******************* | 0.160914

20 13656.314 0.93011 | . |******************* | 0.164888

21 13592.526 0.92577 | . |******************* | 0.168732

22 13533.697 0.92176 | . |****************** | 0.172456

302
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

23 13476.333 0.91786 | . |****************** | 0.176071

24 13419.076 0.91396 | . |****************** | 0.179583

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.2186 0.7353 0.16 0.7316

1 -0.0333 0.6754 -0.02 0.6762

2 -0.4392 0.5835 -0.24 0.5989

Single Mean 0 -1.3205 0.8549 -0.69 0.8483 0.70 0.8932

1 -1.7847 0.8040 -0.83 0.8103 0.71 0.8902

2 -2.5363 0.7138 -1.03 0.7450 0.80 0.8663

Trend 0 -4.7374 0.8392 -1.63 0.7826 1.46 0.8867

1 -5.8155 0.7590 -1.78 0.7138 1.70 0.8381

2 -7.4829 0.6251 -2.00 0.6033 2.07 0.7630

303
Table 1.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 40.297991 1.00000 | |********************| 0

1 4.450522 0.11044 | .|** | 0.027237

2 5.820296 0.14443 | .|*** | 0.027567

3 4.737915 0.11757 | .|** | 0.028123

4 -1.472683 -.03654 | *|. | 0.028485

5 5.123193 0.12713 | .|*** | 0.028520

6 -3.513280 -.08718 | **|. | 0.028937

7 -0.424492 -.01053 | .|. | 0.029131

8 0.043319 0.00107 | .|. | 0.029134

9 -1.528580 -.03793 | *|. | 0.029134

10 4.745930 0.11777 | .|** | 0.029171

11 0.048729 0.00121 | .|. | 0.029521

12 1.502929 0.03730 | .|* | 0.029521

13 2.174063 0.05395 | .|* | 0.029556

14 -1.008726 -.02503 | *|. | 0.029629

15 7.736375 0.19198 | .|**** | 0.029645

16 1.121142 0.02782 | .|* | 0.030553

17 0.285186 0.00708 | .|. | 0.030572

18 0.766373 0.01902 | .|. | 0.030573

19 -4.486843 -.11134 | **|. | 0.030582

20 -1.503531 -.03731 | *|. | 0.030881

21 -4.825896 -.11976 | **|. | 0.030915

22 -2.174566 -.05396 | *|. | 0.031257

23 -0.289037 -.00717 | .|. | 0.031326

24 0.561459 0.01393 | .|. | 0.031327

Table 1.4 – 1. Difference

304
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1197.37 0.0001 -32.81 <.0001

1 -912.362 0.0001 -21.33 <.0001

2 -716.033 0.0001 -16.77 <.0001

Single Mean 0 -1198.20 0.0001 -32.82 <.0001 538.56 0.0010

1 -913.919 0.0001 -21.33 <.0001 227.58 0.0010

2 -717.914 0.0001 -16.78 <.0001 140.82 0.0010

Trend 0 -1198.45 0.0001 -32.81 <.0001 538.35 0.0010

1 -914.399 0.0001 -21.33 <.0001 227.51 0.0010

2 -718.507 0.0001 -16.78 <.0001 140.78 0.0010

AUTOCORRELATION

Graph 2.1 – Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0020

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

305
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1346.0697 <.0001

AR(2) 1346.4940 <.0001

AR(3) 1346.5316 <.0001

AR(4) 1346.5419 <.0001

Graph 2.2 – 1. difference

Table 2.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 1.1620

Table 2.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

306
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 235.7924 <.0001

AR(2) 251.0236 <.0001

AR(3) 255.3014 <.0001

AR(4) 255.3499 <.0001

307
EQUITY

STATIONARITY

Table 3.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 45.968950 1.00000 | |********************| 0

1 45.803658 0.99640 | .|********************| 0.027227

2 45.639695 0.99284 | . |********************| 0.047045

3 45.484399 0.98946 | . |********************| 0.060619

4 45.327341 0.98604 | . |********************| 0.071597

5 45.172940 0.98268 | . |********************| 0.081041

6 45.015651 0.97926 | . |********************| 0.089439

7 44.864888 0.97598 | . |********************| 0.097062

8 44.717761 0.97278 | . |******************* | 0.104083

9 44.557716 0.96930 | . |******************* | 0.110617

10 44.403421 0.96594 | . |******************* | 0.116744

11 44.254279 0.96270 | . |******************* | 0.122525

12 44.105826 0.95947 | . |******************* | 0.128010

13 43.955432 0.95620 | . |******************* | 0.133234

14 43.803779 0.95290 | . |******************* | 0.138228

15 43.660217 0.94978 | . |******************* | 0.143014

16 43.515359 0.94663 | . |******************* | 0.147616

17 43.371623 0.94350 | . |******************* | 0.152049

18 43.229747 0.94041 | . |******************* | 0.156329

19 43.091419 0.93740 | . |******************* | 0.160468

20 42.949395 0.93431 | . |******************* | 0.164477

21 42.796412 0.93099 | . |******************* | 0.168366

22 42.640821 0.92760 | . |******************* | 0.172139

23 42.476744 0.92403 | . |****************** | 0.175806

308
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 42.320503 0.92063 | . |****************** | 0.179370

Table 3.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.4486 0.5814 -0.87 0.3398

1 -0.4682 0.5771 -0.93 0.3123

2 -0.4620 0.5784 -0.97 0.2978

Single Mean 0 -1.2007 0.8670 -0.49 0.8912 0.43 0.9688

1 -1.0472 0.8819 -0.44 0.9006 0.47 0.9601

2 -0.7754 0.9060 -0.34 0.9167 0.48 0.9575

Trend 0 -3.7042 0.9033 -1.31 0.8845 1.74 0.8282

1 -3.3601 0.9209 -1.22 0.9049 1.56 0.8653

2 -2.9773 0.9380 -1.14 0.9214 1.53 0.8726

309
Table 3.3 – 1. Difference
Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.206263 1.00000 | |********************| 0

1 -0.0054558 -.02645 | *|. | 0.027237

2 -0.0098772 -.04789 | *|. | 0.027256

3 0.0020345 0.00986 | .|. | 0.027318

4 -0.0049274 -.02389 | .|. | 0.027321

5 0.0036782 0.01783 | .|. | 0.027336

6 -0.0072596 -.03520 | *|. | 0.027345

7 -0.0035381 -.01715 | .|. | 0.027378

8 0.012363 0.05994 | .|* | 0.027386

9 -0.0072009 -.03491 | *|. | 0.027484

10 0.00028553 0.00138 | .|. | 0.027516

11 -0.0001571 -.00076 | .|. | 0.027517

12 0.00010310 0.00050 | .|. | 0.027517

13 -0.0003427 -.00166 | .|. | 0.027517

14 -0.0018955 -.00919 | .|. | 0.027517

15 -0.0026837 -.01301 | .|. | 0.027519

16 0.0014514 0.00704 | .|. | 0.027523

17 0.0036217 0.01756 | .|. | 0.027525

18 -0.0049976 -.02423 | .|. | 0.027533

19 0.0020756 0.01006 | .|. | 0.027549

20 0.0081115 0.03933 | .|* | 0.027552

21 -0.0028189 -.01367 | .|. | 0.027593

22 0.0070027 0.03395 | .|* | 0.027598

23 -0.0022440 -.01088 | .|. | 0.027629

24 -0.0086645 -.04201 | *|. | 0.027632

Table 3.4 – 1. Difference

310
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1381.96 0.0001 -37.69 <.0001

1 -1521.52 0.0001 -27.58 <.0001

2 -1483.13 0.0001 -21.84 <.0001

Single Mean 0 -1382.65 0.0001 -37.70 <.0001 710.67 0.0010

1 -1524.09 0.0001 -27.59 <.0001 380.73 0.0010

2 -1488.49 0.0001 -21.86 <.0001 238.96 0.0010

Trend 0 -1384.41 0.0001 -37.73 <.0001 711.82 0.0010

1 -1530.18 0.0001 -27.63 <.0001 381.82 0.0010

2 -1500.96 0.0001 -21.91 <.0001 239.97 0.0010

AUTOCORRELATION

Graph 4.1 – Original data

Table 4.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0042

Table 4.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

311
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1343.2477 <.0001

AR(2) 1343.2602 <.0001

AR(3) 1343.3052 <.0001

AR(4) 1343.3128 <.0001

Graph 4.2 – 1. difference

Table 4.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 2.0820

Table 4.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

312
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 2.4661 0.1163

AR(2) 12.7721 0.0017

AR(3) 14.6266 0.0022

AR(4) 15.4389 0.0039

MODEL

CO-INTEGRATION

Table 5.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0046 6.2066 15.34
1 1 0 0.0218 3.84

VAR-MODEL

Table 5.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.12303 0.17046 0.72 0.4706 1
AR1_1_1 0.13201 0.02712 4.87 0.0001 CDS (t-1)
AR1_1_2 -2.03795 0.37921 -5.37 0.0001 EQ (t-1)
AR2_1_1 -0.01039 0.01238 -0.84 0.4015 CDS (t-2)
AR2_1_2 -0.0014 0.00197 -0.71 0.4773 EQ (t-2)
EQ 1 diff Constant2 -0.02358 0.02753 -0.86 0.392 1
AR1_2_1 0.12303 0.17046 0.72 0.4706 CDS (t-1)
AR1_2_2 0.13201 0.02712 4.87 0.0001 EQ (t-1)
AR2_2_1 -2.03795 0.37921 -5.37 0.0001 CDS (t-2)
AR2_2_2 -0.01039 0.01238 -0.84 0.4015 EQ (t-2)

GRANGER CAUSALITY TEST

Table 5.3

313
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 28.88 <.0001
2 2 0.51 0.4772

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 5.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.033 6.25178 22.92 <.0001
EQ 1diff 0.0011 0.45394 0.72 0.4845

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 5.5
Variable Durbin Watson Normality Chi Pr > Chi Sq F Value Pr > F
Square
CDS 1diff 2.04458 9999.99 <.0001 91.42 <.0001
EQ 1diff 2.00205 9999.99 <.0001 0.01 0.9145

314
Appendix 29 – CONSUMER, CYCLICAL
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 29017.914 1.00000 | |********************| 0

1 28980.490 0.99871 | .|********************| 0.027227

2 28924.407 0.99678 | . |********************| 0.047117

3 28852.379 0.99430 | . |********************| 0.060771

4 28769.759 0.99145 | . |********************| 0.071825

5 28681.634 0.98841 | . |********************| 0.081340

6 28583.582 0.98503 | . |********************| 0.089803

7 28476.389 0.98134 | . |********************| 0.097484

8 28362.269 0.97741 | . |********************| 0.104551

9 28245.854 0.97339 | . |******************* | 0.111118

10 28128.320 0.96934 | . |******************* | 0.117269

11 28007.296 0.96517 | . |******************* | 0.123065

12 27881.637 0.96084 | . |******************* | 0.128554

13 27749.759 0.95630 | . |******************* | 0.133772

14 27612.810 0.95158 | . |******************* | 0.138747

15 27471.716 0.94672 | . |******************* | 0.143503

16 27324.163 0.94163 | . |******************* | 0.148061

17 27172.452 0.93640 | . |******************* | 0.152435

18 27015.414 0.93099 | . |******************* | 0.156641

19 26852.460 0.92538 | . |******************* | 0.160691

20 26686.471 0.91966 | . |****************** | 0.164594

21 26521.905 0.91398 | . |****************** | 0.168360

22 26364.400 0.90856 | . |****************** | 0.171999

315
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

23 26211.364 0.90328 | . |****************** | 0.175520

24 26059.317 0.89804 | . |****************** | 0.178933

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.8309 0.8829 0.94 0.9075

1 -0.0289 0.6764 -0.02 0.6771

2 -0.3836 0.5959 -0.22 0.6070

Single Mean 0 -0.1558 0.9493 -0.12 0.9451 1.05 0.8027

1 -1.8372 0.7979 -0.84 0.8077 0.68 0.8983

2 -2.5129 0.7166 -1.02 0.7462 0.77 0.8735

Trend 0 -1.8748 0.9733 -1.08 0.9304 1.05 0.9606

1 -4.9083 0.8272 -1.63 0.7822 1.45 0.8879

2 -6.0278 0.7422 -1.78 0.7128 1.67 0.8431

316
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 34.074309 1.00000 | |********************| 0

1 17.096778 0.50175 | .|********** | 0.027237

2 11.196734 0.32860 | .|******* | 0.033397

3 7.521272 0.22073 | .|**** | 0.035715

4 3.543226 0.10399 | .|** | 0.036713

5 7.102234 0.20843 | .|**** | 0.036931

6 5.414740 0.15891 | . |*** | 0.037794

7 4.587221 0.13462 | . |*** | 0.038286

8 2.912201 0.08547 | . |** | 0.038636

9 2.961379 0.08691 | . |** | 0.038776

10 4.837847 0.14198 | . |*** | 0.038920

11 2.841561 0.08339 | . |** | 0.039302

12 3.466192 0.10172 | . |** | 0.039433

13 1.038760 0.03049 | . |*. | 0.039627

14 0.914435 0.02684 | . |*. | 0.039645

15 2.961615 0.08692 | . |** | 0.039658

16 2.744660 0.08055 | . |** | 0.039799

17 4.269887 0.12531 | . |*** | 0.039920

18 3.335564 0.09789 | . |** | 0.040211

19 -0.130749 -.00384 | .|. | 0.040387

20 -2.011055 -.05902 | .*| . | 0.040388

21 -3.404472 -.09991 | **| . | 0.040451

22 -2.378244 -.06980 | .*| . | 0.040634

23 0.183770 0.00539 | .|. | 0.040723

24 2.032354 0.05964 | . |*. | 0.040723

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

317
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -668.774 0.0001 -21.06 <.0001

1 -536.357 0.0001 -16.25 <.0001

2 -496.350 0.0001 -14.29 <.0001

Single Mean 0 -669.815 0.0001 -21.07 <.0001 221.96 0.0010

1 -537.770 0.0001 -16.26 <.0001 132.20 0.0010

2 -498.217 0.0001 -14.30 <.0001 102.25 0.0010

Trend 0 -670.267 0.0001 -21.07 <.0001 221.96 0.0010

1 -538.381 0.0001 -16.26 <.0001 132.19 0.0010

2 -499.062 0.0001 -14.30 <.0001 102.24 0.0010

AUTOCORRELATION

Graph 2.1 – Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0012

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

318
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1347.0559 <.0001

AR(2) 1347.3864 <.0001

AR(3) 1347.4135 <.0001

AR(4) 1347.4173 <.0001

Graph 2.2 – 1. difference

Table 2.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 1.0259

Table 2.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

319
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 319.2120 <.0001

AR(2) 332.8922 <.0001

AR(3) 334.5532 <.0001

AR(4) 337.7913 <.0001

320
EQUITY

STATIONARITY

Table 3.1 – Original data


Graphical inspection of stationarity

Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 31.951825 1.00000 | |********************| 0

1 31.811025 0.99559 | .|********************| 0.027227

2 31.672993 0.99127 | . |********************| 0.047019

3 31.545778 0.98729 | . |********************| 0.060561

4 31.407407 0.98296 | . |********************| 0.071504

5 31.269161 0.97863 | . |********************| 0.080903

6 31.138105 0.97453 | . |******************* | 0.089248

7 31.009531 0.97051 | . |******************* | 0.096815

8 30.879039 0.96642 | . |******************* | 0.103777

9 30.736977 0.96198 | . |******************* | 0.110247

10 30.590273 0.95739 | . |******************* | 0.116303

11 30.447221 0.95291 | . |******************* | 0.122005

12 30.298851 0.94827 | . |******************* | 0.127403

13 30.140526 0.94331 | . |******************* | 0.132532

14 29.974050 0.93810 | . |******************* | 0.137419

15 29.815442 0.93314 | . |******************* | 0.142087

16 29.664001 0.92840 | . |******************* | 0.146559

17 29.489763 0.92294 | . |****************** | 0.150856

18 29.314683 0.91747 | . |****************** | 0.154985

19 29.141046 0.91203 | . |****************** | 0.158960

20 28.965767 0.90655 | . |****************** | 0.162793

21 28.782052 0.90080 | . |****************** | 0.166493

22 28.615183 0.89557 | . |****************** | 0.170067

23 28.446656 0.89030 | . |****************** | 0.173528

321
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 28.277447 0.88500 | . |****************** | 0.176882

Table 3.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.2285 0.6310 -0.46 0.5144

1 -0.2290 0.6309 -0.47 0.5130

2 -0.2233 0.6322 -0.48 0.5071

Single Mean 0 -3.6656 0.5775 -1.21 0.6714 0.77 0.8738

1 -3.6174 0.5831 -1.20 0.6768 0.76 0.8773

2 -3.1108 0.6434 -1.09 0.7220 0.64 0.9092

Trend 0 -4.4812 0.8565 -1.46 0.8422 1.88 0.8017

1 -4.4238 0.8603 -1.45 0.8463 1.86 0.8055

2 -3.8832 0.8934 -1.34 0.8765 1.77 0.8235

322
Table 3.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.215997 1.00000 | |********************| 0

1 -0.0012283 -.00569 | .|. | 0.027237

2 -0.012075 -.05590 | *|. | 0.027238

3 0.0088087 0.04078 | .|* | 0.027323

4 -0.0005853 -.00271 | .|. | 0.027368

5 -0.0052600 -.02435 | .|. | 0.027368

6 -0.0056982 -.02638 | *|. | 0.027384

7 0.0014765 0.00684 | .|. | 0.027403

8 0.010603 0.04909 | .|* | 0.027404

9 0.0018952 0.00877 | .|. | 0.027469

10 0.0034230 0.01585 | .|. | 0.027471

11 0.0026288 0.01217 | .|. | 0.027478

12 0.0047228 0.02187 | .|. | 0.027482

13 0.0084746 0.03923 | .|* | 0.027495

14 -0.0040477 -.01874 | .|. | 0.027537

15 -0.014459 -.06694 | *|. | 0.027546

16 0.019093 0.08840 | .|** | 0.027666

17 0.0055751 0.02581 | .|* | 0.027875

18 -0.0022356 -.01035 | .|. | 0.027893

19 -0.0009068 -.00420 | .|. | 0.027896

20 0.0032985 0.01527 | .|. | 0.027896

21 -0.019917 -.09221 | **|. | 0.027902

22 0.0012734 0.00590 | .|. | 0.028128

23 0.0046886 0.02171 | .|. | 0.028128

24 -0.0007033 -.00326 | .|. | 0.028141

Table 3.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

323
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1354.59 0.0001 -36.89 <.0001

1 -1513.98 0.0001 -27.50 <.0001

2 -1337.26 0.0001 -21.11 <.0001

Single Mean 0 -1354.66 0.0001 -36.88 <.0001 680.11 0.0010

1 -1514.26 0.0001 -27.49 <.0001 377.83 0.0010

2 -1337.78 0.0001 -21.10 <.0001 222.69 0.0010

Trend 0 -1356.25 0.0001 -36.91 <.0001 681.23 0.0010

1 -1519.85 0.0001 -27.53 <.0001 378.90 0.0010

2 -1347.76 0.0001 -21.15 <.0001 223.58 0.0010

AUTOCORRELATION

Graph 4.1 – Original data

Table 4.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0075

Table 4.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

324
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1338.3298 <.0001

AR(2) 1338.3298 <.0001

AR(3) 1338.3540 <.0001

AR(4) 1338.3656 <.0001

Graph 4.2 – 1. difference

Table 4.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 2.0058

Table 4.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

325
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 0.0126 0.9106

AR(2) 3.1390 0.2081

AR(3) 4.6771 0.1970

AR(4) 4.8506 0.3030

MODEL

CO-INTEGRATION

Table 5.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0041 6.1175 15.34
1 1 0.0005 0.6406 3.84

VAR-MODEL

Table 5.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.09082 0.12985 0.7 0.4844 1
AR1_1_1 0.37935 0.02709 14 0.0001 CDS (t-1)
AR1_1_2 -3.36513 0.28089 -11.98 0.0001 EQ (t-1)
AR2_1_1 0.13587 0.02605 5.22 0.0001 CDS (t-2)
AR2_1_2 -1.27632 0.29353 -4.35 0.0001 EQ (t-2)
EQ 1 diff Constant2 -0.00378 0.01269 -0.3 0.7659 1
AR1_2_1 -0.00106 0.00265 -0.4 0.6895 CDS (t-1)
AR1_2_2 -0.00738 0.02746 -0.27 0.788 EQ (t-1)
AR2_2_1 0.0005 0.00255 0.2 0.8431 CDS (t-2)
AR2_2_2 -0.0594 0.02869 -2.07 0.0386 EQ (t-2)

GRANGER CAUSALITY TEST

326
Table 5.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 158.49 <.0001
2 2 0.16 0.9233

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 5.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.339 4.75817 171.91 <.0001
EQ 1diff 0.0033 0.46512 1.1 0.3529

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 5.5
Variable Durbin Watson Normality Chi Pr > Chi Sq F Value Pr > F
Square
CDS 1diff 2.01632 9999.99 <.0001 29.57 <.0001
EQ 1diff 1.99565 307.23 <.0001 17.72 <.0001

327
Appendix 30 – CONSUMER, NON-CYCLICAL
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 1633.901 1.00000 | |********************| 0

1 1630.750 0.99807 | .|********************| 0.027227

2 1626.870 0.99570 | . |********************| 0.047097

3 1622.281 0.99289 | . |********************| 0.060729

4 1617.234 0.98980 | . |********************| 0.071761

5 1611.992 0.98659 | . |********************| 0.081253

6 1606.416 0.98318 | . |********************| 0.089695

7 1600.590 0.97961 | . |********************| 0.097356

8 1594.508 0.97589 | . |********************| 0.104408

9 1588.271 0.97207 | . |******************* | 0.110964

10 1581.871 0.96816 | . |******************* | 0.117106

11 1575.248 0.96410 | . |******************* | 0.122897

12 1568.373 0.95989 | . |******************* | 0.128381

13 1561.262 0.95554 | . |******************* | 0.133595

14 1554.038 0.95112 | . |******************* | 0.138569

15 1546.684 0.94662 | . |******************* | 0.143327

16 1539.198 0.94204 | . |******************* | 0.147889

17 1531.601 0.93739 | . |******************* | 0.152272

18 1523.908 0.93268 | . |******************* | 0.156491

19 1516.164 0.92794 | . |******************* | 0.160559

20 1508.286 0.92312 | . |****************** | 0.164486

21 1500.350 0.91826 | . |****************** | 0.168283

22 1492.488 0.91345 | . |****************** | 0.171957

328
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

23 1484.650 0.90865 | . |****************** | 0.175517

24 1476.788 0.90384 | . |****************** | 0.178971

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 1.3886 0.9591 2.15 0.9929

1 1.1375 0.9327 1.23 0.9449

2 0.9554 0.9056 0.89 0.9007

Single Mean 0 0.9569 0.9882 0.81 0.9943 2.41 0.4532

1 0.1706 0.9654 0.10 0.9659 0.99 0.8166

2 -0.4201 0.9329 -0.22 0.9341 0.75 0.8782

Trend 0 -0.9370 0.9891 -0.61 0.9777 2.24 0.7283

1 -2.1763 0.9657 -1.00 0.9422 1.45 0.8876

2 -2.9950 0.9373 -1.19 0.9110 1.33 0.9120

329
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 1.675718 1.00000 | |********************| 0

1 0.569205 0.33968 | .|******* | 0.027237

2 0.398543 0.23783 | .|***** | 0.030216

3 0.282866 0.16880 | .|*** | 0.031575

4 0.154114 0.09197 | .|** | 0.032237

5 0.125739 0.07504 | .|** | 0.032431

6 0.099923 0.05963 | .|* | 0.032560

7 0.151693 0.09052 | .|** | 0.032641

8 0.080654 0.04813 | .|* | 0.032826

9 0.169968 0.10143 | .|** | 0.032879

10 0.204094 0.12179 | .|** | 0.033110

11 0.107614 0.06422 | .|* | 0.033441

12 0.074660 0.04455 | .|* | 0.033532

13 -0.010363 -.00618 | .|. | 0.033576

14 -0.038500 -.02298 | .|. | 0.033577

15 0.053917 0.03218 | .|* | 0.033588

16 0.064971 0.03877 | .|* | 0.033611

17 0.101591 0.06063 | .|* | 0.033644

18 0.062455 0.03727 | .|* | 0.033725

19 0.063807 0.03808 | .|* | 0.033756

20 0.062353 0.03721 | .|* | 0.033788

21 0.013308 0.00794 | .|. | 0.033818

22 0.029328 0.01750 | .|. | 0.033820

23 0.070542 0.04210 | .|* | 0.033826

24 0.123666 0.07380 | .|* | 0.033865

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

330
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -885.089 0.0001 -25.62 <.0001

1 -661.008 0.0001 -18.04 <.0001

2 -563.470 0.0001 -15.07 <.0001

Single Mean 0 -887.870 0.0001 -25.66 <.0001 329.31 0.0010

1 -665.042 0.0001 -18.08 <.0001 163.46 0.0010

2 -568.502 0.0001 -15.12 <.0001 114.27 0.0010

Trend 0 -890.641 0.0001 -25.71 <.0001 330.48 0.0010

1 -668.928 0.0001 -18.11 <.0001 164.10 0.0010

2 -573.279 0.0001 -15.15 <.0001 114.77 0.0010

AUTOCORRELATION

Graph 2.1 – Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0010

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

331
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1347.3538 <.0001

AR(2) 1347.4890 <.0001

AR(3) 1347.5181 <.0001

AR(4) 1347.5244 <.0001

Graph 2.2 – 1. difference

Table 2.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 1.3172

Table 2.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

332
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 156.0723 <.0001

AR(2) 179.9187 <.0001

AR(3) 184.4572 <.0001

AR(4) 184.4999 <.0001

333
EQUITY

STATIONARITY

Table 3.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 18.092169 1.00000 | |********************| 0

1 17.983046 0.99397 | .|********************| 0.027227

2 17.886927 0.98866 | . |********************| 0.046968

3 17.808047 0.98430 | . |********************| 0.060458

4 17.710328 0.97889 | . |********************| 0.071355

5 17.610147 0.97336 | . |******************* | 0.080698

6 17.510073 0.96783 | . |******************* | 0.088977

7 17.411115 0.96236 | . |******************* | 0.096465

8 17.315443 0.95707 | . |******************* | 0.103337

9 17.214158 0.95147 | . |******************* | 0.109712

10 17.113252 0.94589 | . |******************* | 0.115667

11 17.015820 0.94051 | . |******************* | 0.121265

12 16.917251 0.93506 | . |******************* | 0.126557

13 16.812377 0.92926 | . |******************* | 0.131579

14 16.702740 0.92320 | . |****************** | 0.136357

15 16.597426 0.91738 | . |****************** | 0.140914

16 16.499703 0.91198 | . |****************** | 0.145274

17 16.395486 0.90622 | . |****************** | 0.149458

18 16.291171 0.90045 | . |****************** | 0.153477

19 16.195093 0.89514 | . |****************** | 0.157344

20 16.096691 0.88970 | . |****************** | 0.161075

21 15.995259 0.88410 | . |****************** | 0.164678

22 15.903662 0.87904 | . |****************** | 0.168160

23 15.810072 0.87386 | . |***************** | 0.171532

24 15.714682 0.86859 | . |***************** | 0.174801

334
Table 3.2 – Original Data
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.1533 0.6481 -0.42 0.5314

1 -0.1441 0.6502 -0.43 0.5264

2 -0.1323 0.6529 -0.45 0.5211

Single Mean 0 -4.2363 0.5134 -1.26 0.6522 0.83 0.8587

1 -3.2765 0.6234 -1.06 0.7325 0.61 0.9174

2 -2.3044 0.7421 -0.84 0.8082 0.41 0.9720

Trend 0 -3.1773 0.9294 -0.92 0.9528 1.61 0.8563

1 -2.1453 0.9665 -0.68 0.9738 1.64 0.8486

2 -1.0952 0.9872 -0.39 0.9879 1.87 0.8033

335
Table 3.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.152119 1.00000 | |********************| 0

1 -0.013887 -.09129 | **|. | 0.027237

2 -0.016131 -.10604 | **|. | 0.027463

3 0.016695 0.10975 | .|** | 0.027765

4 0.0022070 0.01451 | .|. | 0.028085

5 -0.0016776 -.01103 | .|. | 0.028090

6 -0.0014991 -.00985 | .|. | 0.028094

7 -0.0042769 -.02812 | *|. | 0.028096

8 0.0086475 0.05685 | .|* | 0.028117

9 -0.0003261 -.00214 | .|. | 0.028202

10 0.00007010 0.00046 | .|. | 0.028202

11 -0.0000356 -.00023 | .|. | 0.028202

12 0.0044216 0.02907 | .|* | 0.028202

13 0.0033278 0.02188 | .|. | 0.028225

14 -0.0024364 -.01602 | .|. | 0.028237

15 -0.0098904 -.06502 | *|. | 0.028244

16 0.0051587 0.03391 | .|* | 0.028355

17 0.0011622 0.00764 | .|. | 0.028385

18 -0.0089422 -.05878 | *|. | 0.028386

19 0.0055776 0.03667 | .|* | 0.028476

20 -0.0005873 -.00386 | .|. | 0.028511

21 -0.0097703 -.06423 | *|. | 0.028512

22 0.0014967 0.00984 | .|. | 0.028619

23 0.0047765 0.03140 | .|* | 0.028621

24 -0.0042776 -.02812 | *|. | 0.028647

Table 3.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

336
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1469.89 0.0001 -40.20 <.0001

1 -1851.52 0.0001 -30.40 <.0001

2 -1384.74 0.0001 -21.35 <.0001

Single Mean 0 -1469.98 0.0001 -40.19 <.0001 807.54 0.0010

1 -1851.95 0.0001 -30.39 <.0001 461.92 0.0010

2 -1385.42 0.0001 -21.35 <.0001 227.87 0.0010

Trend 0 -1472.61 0.0001 -40.25 <.0001 810.05 0.0010

1 -1864.18 0.0001 -30.48 <.0001 464.60 0.0010

2 -1404.29 0.0001 -21.43 <.0001 229.69 0.0010

AUTOCORRELATION

Graph 4.1 – Original data

Table 4.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0084

Table 4.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

337
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1336.6511 <.0001

AR(2) 1336.7374 <.0001

AR(3) 1336.8694 <.0001

AR(4) 1336.9580 <.0001

Graph 4.2 – 1. difference

Table 4.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 2.1824

Table 4.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

338
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 11.2352 0.0008

AR(2) 29.0186 <.0001

AR(3) 39.8060 <.0001

AR(4) 40.4463 <.0001

MODEL

CO-INTEGRATION

Table 5.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0052 8.5067 15.34
1 1 0.0012 1.5514 3.84

VAR-MODEL

Table 5.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.03975 0.03223 1.23 0.2177 1
AR1_1_1 0.24989 0.02709 9.22 0.0001 CDS (t-1)
AR1_1_2 -0.62065 0.08353 -7.43 0.0001 EQ (t-1)
AR2_1_1 0.14650 0.02661 5.50 0.0001 CDS (t-2)
AR2_1_2 -0.34168 0.08484 -4.03 0.0001 EQ (t-2)
EQ 1 diff Constant2 -0.00307 0.01057 -0.29 0.7715 1
AR1_2_1 -0.00960 0.00888 -1.08 0.2800 CDS (t-1)
AR1_2_2 -0.10597 0.02738 -3.87 0.0001 EQ (t-1)
AR2_2_1 -0.00031 0.00872 -0.04 0.9720 CDS (t-2)
AR2_2_2 -0.12239 0.02781 -4.40 0.0001 EQ (t-2)

GRANGER CAUSALITY TEST

339
Table 5.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 65.41 <0.0001
2 2 1.33 0.5132

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 5.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.1738 1.17947 70.52 < 0.0001
EQ 1diff 0.0225 0.38661 7.72 < 0.0001

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 5.5
Variable Durbin Watson Normality Chi Pr > Chi Sq F Value Pr > F
Square
CDS 1diff 2.03497 9999.99 < 0.0001 24.12 < 0.0001
EQ 1diff 1.97926 4026.34 < 0.0001 85.44 < 0.0001

340
Appendix 31 - ENERGY
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 4218.331 1.00000 | |********************| 0

1 4210.151 0.99806 | .|********************| 0.027227

2 4198.563 0.99531 | . |********************| 0.047097

3 4183.585 0.99176 | . |********************| 0.060719

4 4166.137 0.98763 | . |********************| 0.071729

5 4147.489 0.98321 | . |********************| 0.081186

6 4127.398 0.97844 | . |********************| 0.089579

7 4106.642 0.97352 | . |******************* | 0.097179

8 4085.129 0.96842 | . |******************* | 0.104158

9 4062.965 0.96317 | . |******************* | 0.110631

10 4041.496 0.95808 | . |******************* | 0.116682

11 4020.042 0.95299 | . |******************* | 0.122375

12 3998.276 0.94783 | . |******************* | 0.127758

13 3976.780 0.94274 | . |******************* | 0.132868

14 3955.488 0.93769 | . |******************* | 0.137738

15 3934.776 0.93278 | . |******************* | 0.142391

16 3914.119 0.92788 | . |******************* | 0.146851

17 3893.536 0.92300 | . |****************** | 0.151134

18 3873.733 0.91831 | . |****************** | 0.155257

19 3854.524 0.91376 | . |****************** | 0.159232

20 3836.123 0.90939 | . |****************** | 0.163073

21 3818.315 0.90517 | . |****************** | 0.166790

22 3801.472 0.90118 | . |****************** | 0.170393

341
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

23 3784.634 0.89719 | . |****************** | 0.173890

24 3767.416 0.89311 | . |****************** | 0.177288

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.5789 0.8269 0.58 0.8421

1 0.0065 0.6846 0.00 0.6846

2 -0.5529 0.5588 -0.30 0.5772

Single Mean 0 -0.3434 0.9380 -0.20 0.9361 0.38 0.9773

1 -2.1219 0.7642 -0.83 0.8106 0.52 0.9479

2 -3.9259 0.5477 -1.23 0.6626 0.89 0.8443

Trend 0 -1.6660 0.9778 -0.89 0.9557 1.61 0.8563

1 -3.7522 0.9007 -1.35 0.8753 1.44 0.8897

2 -5.7701 0.7626 -1.67 0.7646 1.68 0.8409

342
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 9.254701 1.00000 | |********************| 0

1 3.485519 0.37662 | .|******** | 0.027237

2 2.977391 0.32172 | .|****** | 0.030859

3 2.013368 0.21755 | .|**** | 0.033254

4 1.090005 0.11778 | .|** | 0.034294

5 1.333421 0.14408 | .|*** | 0.034593

6 0.487660 0.05269 | .|* | 0.035035

7 0.575578 0.06219 | .|* | 0.035094

8 0.623688 0.06739 | .|* | 0.035175

9 -0.560165 -.06053 | *|. | 0.035271

10 0.164246 0.01775 | .|. | 0.035348

11 0.065369 0.00706 | .|. | 0.035355

12 -0.573618 -.06198 | *|. | 0.035356

13 -0.757417 -.08184 | **|. | 0.035436

14 -1.059353 -.11447 | **|. | 0.035576

15 -0.366081 -.03956 | *|. | 0.035848

16 -0.113254 -.01224 | .|. | 0.035881

17 -0.748300 -.08086 | **|. | 0.035884

18 -0.668465 -.07223 | *|. | 0.036019

19 -0.917097 -.09910 | **|. | 0.036126

20 -0.693784 -.07497 | *|. | 0.036327

21 -0.890681 -.09624 | **|. | 0.036442

22 -0.040501 -.00438 | .|. | 0.036630

23 0.467134 0.05048 | .|* | 0.036630

24 0.348329 0.03764 | .|* | 0.036682

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

343
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -839.189 0.0001 -24.68 <.0001

1 -547.349 0.0001 -16.54 <.0001

2 -481.258 0.0001 -14.26 <.0001

Single Mean 0 -839.651 0.0001 -24.68 <.0001 304.53 0.0010

1 -547.930 0.0001 -16.54 <.0001 136.82 0.0010

2 -481.931 0.0001 -14.26 <.0001 101.64 0.0010

Trend 0 -841.129 0.0001 -24.70 <.0001 305.08 0.0010

1 -549.521 0.0001 -16.56 <.0001 137.07 0.0010

2 -484.053 0.0001 -14.28 <.0001 101.90 0.0010

AUTOCORRELATION

Graph 2.1 – Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0022

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

344
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1346.0414 <.0001

AR(2) 1346.4614 <.0001

AR(3) 1346.5739 <.0001

AR(4) 1346.5815 <.0001

Graph 2.2 – 1. difference

Table 2.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 1.2465

Table 2.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

345
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 191.2203 <.0001

AR(2) 242.1242 <.0001

AR(3) 245.1931 <.0001

AR(4) 246.3835 <.0001

346
EQUITY

STATIONARITY

Table 3.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 249.378 1.00000 | |********************| 0

1 248.417 0.99614 | .|********************| 0.027227

2 247.609 0.99290 | . |********************| 0.047037

3 246.940 0.99022 | . |********************| 0.060614

4 246.184 0.98719 | . |********************| 0.071609

5 245.425 0.98415 | . |********************| 0.081072

6 244.686 0.98119 | . |********************| 0.089491

7 243.937 0.97818 | . |********************| 0.097139

8 243.229 0.97534 | . |********************| 0.104185

9 242.490 0.97238 | . |******************* | 0.110747

10 241.705 0.96923 | . |******************* | 0.116905

11 240.929 0.96612 | . |******************* | 0.122717

12 240.148 0.96299 | . |******************* | 0.128231

13 239.418 0.96006 | . |******************* | 0.133484

14 238.648 0.95697 | . |******************* | 0.138508

15 237.855 0.95379 | . |******************* | 0.143326

16 237.143 0.95094 | . |******************* | 0.147956

17 236.390 0.94792 | . |******************* | 0.152420

18 235.585 0.94469 | . |******************* | 0.156729

19 234.870 0.94182 | . |******************* | 0.160894

20 234.126 0.93884 | . |******************* | 0.164931

21 233.357 0.93575 | . |******************* | 0.168846

22 232.604 0.93274 | . |******************* | 0.172647

23 231.820 0.92959 | . |******************* | 0.176343

24 231.056 0.92653 | . |******************* | 0.179939

347
Table 3.2 – Original Data
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.0747 0.6660 -0.09 0.6525

1 -0.0144 0.6797 -0.02 0.6766

2 0.0375 0.6917 0.06 0.7017

Single Mean 0 -5.0474 0.4308 -1.87 0.3476 1.86 0.5930

1 -4.2772 0.5090 -1.77 0.3972 1.70 0.6344

2 -3.6810 0.5757 -1.70 0.4303 1.62 0.6547

Trend 0 -4.9522 0.8241 -1.29 0.8889 1.74 0.8285

1 -3.4287 0.9175 -1.00 0.9426 1.62 0.8535

2 -2.2030 0.9649 -0.72 0.9709 1.67 0.8424

348
Table 3.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 1.353705 1.00000 | |********************| 0

1 -0.151396 -.11184 | **|. | 0.027237

2 -0.135305 -.09995 | **|. | 0.027575

3 0.078080 0.05768 | .|* | 0.027843

4 0.0075908 0.00561 | .|. | 0.027931

5 -0.0091749 -.00678 | .|. | 0.027932

6 0.0043845 0.00324 | .|. | 0.027933

7 -0.047377 -.03500 | *|. | 0.027934

8 0.031397 0.02319 | .|. | 0.027966

9 0.033343 0.02463 | .|. | 0.027980

10 0.0094551 0.00698 | .|. | 0.027996

11 0.0021680 0.00160 | .|. | 0.027998

12 -0.043065 -.03181 | *|. | 0.027998

13 0.041987 0.03102 | .|* | 0.028025

14 0.020932 0.01546 | .|. | 0.028050

15 -0.084157 -.06217 | *|. | 0.028056

16 0.040848 0.03017 | .|* | 0.028158

17 0.055943 0.04133 | .|* | 0.028182

18 -0.092526 -.06835 | *|. | 0.028227

19 0.019329 0.01428 | .|. | 0.028350

20 0.014307 0.01057 | .|. | 0.028355

21 -0.018658 -.01378 | .|. | 0.028358

22 0.031202 0.02305 | .|. | 0.028363

23 -0.019480 -.01439 | .|. | 0.028377

24 -0.045500 -.03361 | *|. | 0.028382

Table 3.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

349
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1497.57 0.0001 -41.03 <.0001

1 -1880.57 0.0001 -30.63 <.0001

2 -1680.60 0.0001 -22.70 <.0001

Single Mean 0 -1497.81 0.0001 -41.02 <.0001 841.25 0.0010

1 -1881.74 0.0001 -30.63 <.0001 469.08 0.0010

2 -1682.90 0.0001 -22.70 <.0001 257.68 0.0010

Trend 0 -1499.90 0.0001 -41.06 <.0001 843.14 0.0010

1 -1891.72 0.0001 -30.70 <.0001 471.13 0.0010

2 -1703.43 0.0001 -22.77 <.0001 259.33 0.0010

AUTOCORRELATION

Graph 4.1 – Original data

Table 4.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0054

Table 4.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

350
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1338.7672 <.0001

AR(2) 1338.8374 <.0001

AR(3) 1338.8993 <.0001

AR(4) 1338.9091 <.0001

Graph 4.2 – 1. difference

Table 4.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 2.2225

Table 4.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

351
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 16.8785 <.0001

AR(2) 34.1514 <.0001

AR(3) 35.6111 <.0001

AR(4) 35.6486 <.0001

MODEL

CO-INTEGRATION

Table 5.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0046 9.1014 15.34
1 1 0.0022 2.9383 3.84

VAR-MODEL

Table 5.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.05478 0.07212 0.76 0.4477 1
AR1_1_1 0.21235 0.02732 7.77 0.0001 CDS (t-1)
AR1_1_2 -0.59805 0.06335 -9.44 0.0001 EQ (t-1)
AR2_1_1 0.19033 0.02713 7.02 0.0001 CDS (t-2)
AR2_1_2 -0.40986 0.06532 -6.27 0.0001 EQ (t-2)
AR3_1_1 0.08264 0.02648 3.12 0.0018 CDS (t-3)
AR3_1_2 -0.29584 0.06493 -4.56 0.0001 EQ (t-3)
EQ 1 diff Constant2 0.01656 0.03143 0.53 0.5985 1
AR1_2_1 0.00793 0.01191 0.67 0.5054 CDS (t-1)
AR1_2_2 -0.11887 0.02761 -4.31 0.0001 EQ (t-1)
AR2_2_1 0.00214 0.01182 0.18 0.8563 CDS (t-2)
AR2_2_2 -0.10533 0.02847 -3.70 0.0002 EQ (t-2)
AR3_2_1 0.00766 0.01154 0.66 0.5072 CDS (t-3)
AR3_2_2 0.04008 0.02830 1.42 0.1568 EQ (t-3)

GRANGER CAUSALITY TEST

352
Table 5.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 120.13 <0.0001
2 2 1.71 0.6353

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 5.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.2499 2.64234 74.31 < 0.0001
EQ 1diff 0.0277 1.15150 6.34 < 0.0001

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 5.5
Variable Durbin Watson Normality Chi Pr > Chi Sq F Value Pr > F
Square
CDS 1diff 2.00236 9999.99 < 0.0001 42.89 < 0.0001
EQ 1diff 2.00283 2873.70 < 0.0001 37.40 < 0.0001

353
Appendix 32 - FINANCE
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 38766.778 1.00000 | |********************| 0

1 38567.323 0.99486 | .|********************| 0.027227

2 38334.250 0.98884 | . |********************| 0.046996

3 38094.764 0.98267 | . |********************| 0.060484

4 37853.425 0.97644 | . |********************| 0.071344

5 37618.029 0.97037 | . |******************* | 0.080644

6 37377.817 0.96417 | . |******************* | 0.088879

7 37137.588 0.95797 | . |******************* | 0.096321

8 36910.486 0.95212 | . |******************* | 0.103142

9 36697.294 0.94662 | . |******************* | 0.109464

10 36486.180 0.94117 | . |******************* | 0.115373

11 36269.514 0.93558 | . |******************* | 0.120930

12 36043.235 0.92975 | . |******************* | 0.126182

13 35821.442 0.92402 | . |****************** | 0.131162

14 35601.362 0.91835 | . |****************** | 0.135902

15 35382.205 0.91269 | . |****************** | 0.140427

16 35140.060 0.90645 | . |****************** | 0.144757

17 34857.523 0.89916 | . |****************** | 0.148905

18 34570.961 0.89177 | . |****************** | 0.152877

19 34282.021 0.88431 | . |****************** | 0.156686

20 33987.025 0.87670 | . |****************** | 0.160343

21 33684.687 0.86891 | . |***************** | 0.163858

22 33384.971 0.86117 | . |***************** | 0.167239

354
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

23 33106.608 0.85399 | . |***************** | 0.170494

24 32849.223 0.84735 | . |***************** | 0.173636

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 3.9158 0.9998 2.55 0.9976

1 2.8577 0.9983 1.35 0.9565

2 3.3172 0.9994 1.74 0.9807

Single Mean 0 3.2145 0.9998 1.72 0.9997 3.46 0.1818

1 1.6568 0.9962 0.65 0.9910 1.26 0.7482

2 2.3257 0.9988 1.01 0.9967 1.80 0.6085

Trend 0 -0.2084 0.9952 -0.08 0.9952 3.06 0.5609

1 -3.3659 0.9206 -0.92 0.9526 2.04 0.7691

2 -2.0390 0.9693 -0.62 0.9776 2.20 0.7354

355
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 99.791817 1.00000 | |********************| 0

1 30.140767 0.30204 | .|****** | 0.027237

2 -0.205277 -.00206 | .|. | 0.029617

3 0.897811 0.00900 | .|. | 0.029617

4 4.415779 0.04425 | .|* | 0.029620

5 9.943450 0.09964 | .|** | 0.029669

6 0.302452 0.00303 | .|. | 0.029916

7 -7.720181 -.07736 | **|. | 0.029916

8 -4.685696 -.04695 | *|. | 0.030064

9 7.485209 0.07501 | .|** | 0.030118

10 12.417338 0.12443 | .|** | 0.030257

11 1.996119 0.02000 | .|. | 0.030634

12 -6.509141 -.06523 | *|. | 0.030644

13 -2.743187 -.02749 | *|. | 0.030746

14 -6.058306 -.06071 | *|. | 0.030765

15 3.697904 0.03706 | .|* | 0.030853

16 14.689218 0.14720 | .|*** | 0.030886

17 5.990550 0.06003 | .|* | 0.031403

18 2.402203 0.02407 | .|. | 0.031488

19 -0.350144 -.00351 | .|. | 0.031501

20 1.957254 0.01961 | .|. | 0.031501

21 -1.217113 -.01220 | .|. | 0.031511

22 -9.668025 -.09688 | **|. | 0.031514

23 -9.900200 -.09921 | **|. | 0.031734

24 -3.119820 -.03126 | *|. | 0.031963

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

356
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -937.219 0.0001 -26.79 <.0001

1 -1146.58 0.0001 -23.91 <.0001

2 -1004.08 0.0001 -19.08 <.0001

Single Mean 0 -939.995 0.0001 -26.84 <.0001 360.25 0.0010

1 -1154.19 0.0001 -23.98 <.0001 287.47 0.0010

2 -1015.78 0.0001 -19.16 <.0001 183.46 0.0010

Trend 0 -944.304 0.0001 -26.92 <.0001 362.46 0.0010

1 -1166.21 0.0001 -24.09 <.0001 290.13 0.0010

2 -1034.53 0.0001 -19.27 <.0001 185.66 0.0010

AUTOCORRELATION

Graph 2.1 – Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0026

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

357
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1345.2585 <.0001

AR(2) 1345.5477 <.0001

AR(3) 1345.5749 <.0001

AR(4) 1345.5789 <.0001

Graph 2.3 – 1. difference

Table 2.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 1.3955

Table 2.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

358
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 123.0217 <.0001

AR(2) 135.8956 <.0001

AR(3) 138.3713 <.0001

AR(4) 139.4880 <.0001

359
EQUITY

STATIONARITY

Table 3.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 139.068 1.00000 | |********************| 0

1 138.551 0.99628 | .|********************| 0.027227

2 138.076 0.99287 | . |********************| 0.047041

3 137.620 0.98959 | . |********************| 0.060617

4 137.137 0.98612 | . |********************| 0.071598

5 136.678 0.98282 | . |********************| 0.081043

6 136.229 0.97959 | . |********************| 0.089443

7 135.799 0.97650 | . |********************| 0.097071

8 135.358 0.97333 | . |******************* | 0.104098

9 134.911 0.97011 | . |******************* | 0.110639

10 134.444 0.96675 | . |******************* | 0.116774

11 133.987 0.96346 | . |******************* | 0.122564

12 133.524 0.96014 | . |******************* | 0.128055

13 133.063 0.95682 | . |******************* | 0.133285

14 132.578 0.95334 | . |******************* | 0.138283

15 132.128 0.95010 | . |******************* | 0.143072

16 131.674 0.94683 | . |******************* | 0.147675

17 131.178 0.94327 | . |******************* | 0.152109

18 130.649 0.93947 | . |******************* | 0.156385

19 130.125 0.93569 | . |******************* | 0.160514

20 129.587 0.93183 | . |******************* | 0.164507

21 129.042 0.92791 | . |******************* | 0.168375

22 128.523 0.92418 | . |****************** | 0.172124

23 127.980 0.92027 | . |****************** | 0.175764

24 127.458 0.91652 | . |****************** | 0.179300

360
Table 3.2 – Original Data
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.5161 0.5667 -0.96 0.3012

1 -0.5023 0.5696 -1.01 0.2822

2 -0.4938 0.5715 -1.03 0.2727

Single Mean 0 0.4490 0.9758 0.21 0.9733 0.57 0.9318

1 0.7722 0.9845 0.39 0.9825 0.73 0.8856

2 0.9302 0.9878 0.49 0.9864 0.83 0.8593

Trend 0 -2.2636 0.9632 -0.91 0.9529 2.38 0.6987

1 -1.7913 0.9752 -0.78 0.9659 2.54 0.6664

2 -1.5668 0.9797 -0.71 0.9712 2.68 0.6388

361
Table 3.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.469981 1.00000 | |********************| 0

1 -0.035676 -.07591 | **|. | 0.027237

2 -0.016783 -.03571 | *|. | 0.027393

3 0.0081230 0.01728 | .|. | 0.027428

4 -0.021869 -.04653 | *|. | 0.027436

5 -0.0031651 -.00673 | .|. | 0.027494

6 -0.026933 -.05731 | *|. | 0.027496

7 0.0033529 0.00713 | .|. | 0.027584

8 0.023700 0.05043 | .|* | 0.027585

9 -0.0069680 -.01483 | .|. | 0.027654

10 0.025006 0.05321 | .|* | 0.027660

11 0.0022158 0.00471 | .|. | 0.027735

12 -0.0035226 -.00750 | .|. | 0.027736

13 0.0068179 0.01451 | .|. | 0.027737

14 -0.023805 -.05065 | *|. | 0.027743

15 -0.029988 -.06381 | *|. | 0.027812

16 0.033078 0.07038 | .|* | 0.027920

17 0.038891 0.08275 | .|** | 0.028051

18 0.0019163 0.00408 | .|. | 0.028232

19 0.0029425 0.00626 | .|. | 0.028232

20 0.0011593 0.00247 | .|. | 0.028233

21 -0.033727 -.07176 | *|. | 0.028233

22 0.0092364 0.01965 | .|. | 0.028368

23 0.0069787 0.01485 | .|. | 0.028379

24 -0.0036346 -.00773 | .|. | 0.028384

Table 3.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

362
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1448.09 0.0001 -39.55 <.0001

1 -1570.02 0.0001 -28.00 <.0001

2 -1508.73 0.0001 -21.96 <.0001

Single Mean 0 -1449.27 0.0001 -39.57 <.0001 783.05 0.0010

1 -1574.41 0.0001 -28.03 <.0001 392.73 0.0010

2 -1517.93 0.0001 -21.99 <.0001 241.86 0.0010

Trend 0 -1453.41 0.0001 -39.68 <.0001 787.32 0.0010

1 -1589.84 0.0001 -28.15 <.0001 396.28 0.0010

2 -1550.28 0.0001 -22.13 <.0001 244.89 0.0010

AUTOCORRELATION

Graph 4.1 – Original data

Table 4.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0034

Table 4.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

363
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1344.4342 <.0001

AR(2) 1344.4613 <.0001

AR(3) 1344.4695 <.0001

AR(4) 1344.4700 <.0001

Graph 4.2 – 1. difference

Table 4.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 2.1514

364
Table 4.4 – 1. difference
Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 7.7687 0.0053

AR(2) 10.1012 0.0064

AR(3) 10.2759 0.0164

AR(4) 13.1233 0.0107

MODEL

CO-INTEGRATION

Table 5.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0046 9.1014 15.34
1 1 0.0022 2.9383 3.84

VAR-MODEL

Table 5.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.32171 0.24455 1.32 0.1886 1
AR1_1_1 0.26240 0.02720 9.65 0.0001 CDS (t-1)
AR1_1_2 -4.54196 0.35951 -12.63 0.0001 EQ (t-1)
AR2_1_1 -0.08049 0.02567 -3.14 0.0017 CDS (t-2)
AR2_1_2 -1.74065 0.37845 -4.60 0.0001 EQ (t-2)
EQ 1 diff Constant2 -0.02440 0.01868 -1.31 0.1917 1
AR1_2_1 -0.00024 0.00208 -0.12 0.9075 CDS (t-1)
AR1_2_2 -0.08120 0.02746 -2.96 0.0032 EQ (t-1)
AR2_2_1 0.00418 0.00196 2.13 0.0333 CDS (t-2)
AR2_2_2 -0.03814 0.02891 -1.32 0.1873 EQ (t-2)

365
GRANGER CAUSALITY TEST

Table 5.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 170.05 < 0.0001
2 2 4.83 0.0892

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 5.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.2020 8.94695 84.86 < 0.0001
EQ 1diff 0.0111 0.68344 3.75 0.0048

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 5.5
Variable Durbin Watson Normality Chi Pr > Chi Sq F Value Pr > F
Square
CDS 1diff 2.00133 9999.99 < 0.0001 48.50 < 0.0001
EQ 1diff 1.99951 1083.81 < 0.0001 84.00 < 0.0001

366
Appendix 33 - INDUSTRIALS
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 3908.722 1.00000 | |********************| 0

1 3899.947 0.99775 | .|********************| 0.027227

2 3889.413 0.99506 | . |********************| 0.047087

3 3877.232 0.99194 | . |********************| 0.060706

4 3863.875 0.98853 | . |********************| 0.071722

5 3850.199 0.98503 | . |********************| 0.081196

6 3835.705 0.98132 | . |********************| 0.089617

7 3820.774 0.97750 | . |********************| 0.097257

8 3805.195 0.97351 | . |******************* | 0.104286

9 3789.467 0.96949 | . |******************* | 0.110818

10 3773.721 0.96546 | . |******************* | 0.116936

11 3757.590 0.96133 | . |******************* | 0.122703

12 3740.542 0.95697 | . |******************* | 0.128165

13 3723.018 0.95249 | . |******************* | 0.133357

14 3705.439 0.94799 | . |******************* | 0.138308

15 3687.561 0.94342 | . |******************* | 0.143043

16 3668.954 0.93866 | . |******************* | 0.147584

17 3650.357 0.93390 | . |******************* | 0.151945

18 3631.714 0.92913 | . |******************* | 0.156142

19 3613.011 0.92435 | . |****************** | 0.160188

20 3594.960 0.91973 | . |****************** | 0.164094

21 3576.881 0.91510 | . |****************** | 0.167872

22 3559.110 0.91056 | . |****************** | 0.171530

367
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

23 3541.344 0.90601 | . |****************** | 0.175076

24 3523.541 0.90146 | . |****************** | 0.178518

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 1.9360 0.9873 2.34 0.9958

1 1.4465 0.9637 1.11 0.9309

2 1.3146 0.9525 0.94 0.9084

Single Mean 0 1.4981 0.9950 1.25 0.9985 2.87 0.3340

1 0.4293 0.9751 0.23 0.9743 0.89 0.8437

2 0.1596 0.9649 0.08 0.9641 0.76 0.8778

Trend 0 -0.7764 0.9908 -0.51 0.9833 3.56 0.4604

1 -2.4732 0.9567 -1.02 0.9393 1.86 0.8051

2 -2.8844 0.9418 -1.12 0.9243 1.78 0.8206

368
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 4.150487 1.00000 | |********************| 0

1 1.768612 0.42612 | .|********* | 0.027237

2 0.967455 0.23309 | .|***** | 0.031800

3 0.525220 0.12654 | .|*** | 0.033043

4 0.198726 0.04788 | .|* | 0.033401

5 0.241685 0.05823 | .|* | 0.033452

6 -0.023423 -.00564 | .|. | 0.033527

7 0.169823 0.04092 | .|* | 0.033528

8 0.201247 0.04849 | .|* | 0.033565

9 0.353473 0.08516 | .|** | 0.033616

10 0.456020 0.10987 | .|** | 0.033776

11 0.376410 0.09069 | .|** | 0.034040

12 0.055861 0.01346 | .|. | 0.034219

13 -0.258181 -.06220 | *|. | 0.034223

14 -0.320003 -.07710 | **|. | 0.034307

15 0.049741 0.01198 | .|. | 0.034435

16 -0.061631 -.01485 | .|. | 0.034438

17 0.112526 0.02711 | .|* | 0.034443

18 -0.080162 -.01931 | .|. | 0.034459

19 -0.039146 -.00943 | .|. | 0.034467

20 0.0033157 0.00080 | .|. | 0.034469

21 -0.048974 -.01180 | .|. | 0.034469

22 0.080161 0.01931 | .|. | 0.034472

23 0.051850 0.01249 | .|. | 0.034480

24 0.203616 0.04906 | .|* | 0.034483

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

369
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -770.584 0.0001 -23.23 <.0001

1 -676.465 0.0001 -18.36 <.0001

2 -659.416 0.0001 -16.20 <.0001

Single Mean 0 -773.006 0.0001 -23.27 <.0001 270.71 0.0010

1 -680.336 0.0001 -18.40 <.0001 169.30 0.0010

2 -665.210 0.0001 -16.24 <.0001 131.95 0.0010

Trend 0 -776.877 0.0001 -23.34 <.0001 272.38 0.0010

1 -686.582 0.0001 -18.47 <.0001 170.62 0.0010

2 -674.977 0.0001 -16.32 <.0001 133.25 0.0010

AUTOCORRELATION

Graph 2.1 – Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0011

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

370
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1347.4668 <.0001

AR(2) 1347.7135 <.0001

AR(3) 1347.7196 <.0001

AR(4) 1347.7197 <.0001

Graph 2.2 – 1. difference

Table 2.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 1.1474

Table 2.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

371
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 244.7730 <.0001

AR(2) 249.1645 <.0001

AR(3) 249.2433 <.0001

AR(4) 249.9192 <.0001

372
EQUITY

STATIONARITY

Table 3.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 102.702 1.00000 | |********************| 0

1 102.341 0.99649 | .|********************| 0.027227

2 102.017 0.99333 | . |********************| 0.047048

3 101.713 0.99037 | . |********************| 0.060633

4 101.376 0.98709 | . |********************| 0.071627

5 101.043 0.98384 | . |********************| 0.081087

6 100.710 0.98061 | . |********************| 0.089499

7 100.369 0.97728 | . |********************| 0.097138

8 100.024 0.97393 | . |******************* | 0.104172

9 99.646275 0.97025 | . |******************* | 0.110716

10 99.254531 0.96643 | . |******************* | 0.116849

11 98.871142 0.96270 | . |******************* | 0.122631

12 98.466101 0.95876 | . |******************* | 0.128111

13 98.038953 0.95460 | . |******************* | 0.133324

14 97.589951 0.95023 | . |******************* | 0.138298

15 97.163812 0.94608 | . |******************* | 0.143056

16 96.748125 0.94203 | . |******************* | 0.147621

17 96.308266 0.93775 | . |******************* | 0.152012

18 95.846270 0.93325 | . |******************* | 0.156241

19 95.400022 0.92890 | . |******************* | 0.160320

20 94.960549 0.92462 | . |****************** | 0.164261

21 94.494908 0.92009 | . |****************** | 0.168075

22 94.036340 0.91562 | . |****************** | 0.171768

23 93.557545 0.91096 | . |****************** | 0.175349

24 93.093834 0.90645 | . |****************** | 0.178823

373
Table 3.2 – Original Data
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.1453 0.6499 -0.29 0.5804

1 -0.1296 0.6535 -0.28 0.5835

2 -0.1199 0.6557 -0.28 0.5862

Single Mean 0 -2.7657 0.6856 -1.13 0.7047 0.64 0.9079

1 -2.2703 0.7463 -1.01 0.7501 0.52 0.9484

2 -2.0158 0.7769 -0.95 0.7741 0.45 0.9635

Trend 0 -1.4899 0.9811 -0.58 0.9795 1.95 0.7867

1 -0.9381 0.9891 -0.40 0.9875 2.17 0.7414

2 -0.6425 0.9921 -0.29 0.9909 2.38 0.7001

374
Table 3.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.452876 1.00000 | |********************| 0

1 -0.038776 -.08562 | **|. | 0.027237

2 -0.020440 -.04513 | *|. | 0.027436

3 0.022012 0.04860 | .|* | 0.027491

4 0.0023457 0.00518 | .|. | 0.027554

5 0.0033104 0.00731 | .|. | 0.027555

6 0.0026680 0.00589 | .|. | 0.027557

7 -0.0048373 -.01068 | .|. | 0.027557

8 0.032224 0.07115 | .|* | 0.027561

9 0.0071505 0.01579 | .|. | 0.027696

10 0.011255 0.02485 | .|. | 0.027703

11 0.019335 0.04269 | .|* | 0.027720

12 0.0079126 0.01747 | .|. | 0.027768

13 0.022104 0.04881 | .|* | 0.027777

14 -0.015250 -.03367 | *|. | 0.027840

15 -0.030969 -.06838 | *|. | 0.027870

16 0.025607 0.05654 | .|* | 0.027995

17 0.028974 0.06398 | .|* | 0.028079

18 -0.023880 -.05273 | *|. | 0.028187

19 -0.0094179 -.02080 | .|. | 0.028260

20 0.0081998 0.01811 | .|. | 0.028271

21 -0.011560 -.02553 | *|. | 0.028280

22 0.016788 0.03707 | .|* | 0.028297

23 -0.0021127 -.00467 | .|. | 0.028333

24 -0.0056673 -.01251 | .|. | 0.028334

Table 3.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

375
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1462.36 0.0001 -39.97 <.0001

1 -1624.45 0.0001 -28.47 <.0001

2 -1430.01 0.0001 -21.56 <.0001

Single Mean 0 -1462.37 0.0001 -39.96 <.0001 798.31 0.0010

1 -1624.47 0.0001 -28.46 <.0001 405.10 0.0010

2 -1430.08 0.0001 -21.56 <.0001 232.34 0.0010

Trend 0 -1466.29 0.0001 -40.06 <.0001 802.29 0.0010

1 -1639.83 0.0001 -28.58 <.0001 408.55 0.0010

2 -1458.71 0.0001 -21.68 <.0001 235.05 0.0010

AUTOCORRELATION

Graph 4.1 – Original data

Table 4.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0044

Table 4.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

376
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1341.5048 <.0001

AR(2) 1341.5424 <.0001

AR(3) 1341.5547 <.0001

AR(4) 1341.5627 <.0001

Graph 4.2 – 1. difference

Table 4.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 2.1708

Table 4.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

377
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 9.8855 0.0017

AR(2) 13.6253 0.0011

AR(3) 15.8020 0.0012

AR(4) 15.9573 0.0031

MODEL

CO-INTEGRATION

Table 5.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0035 4.8609 15.34
1 1 0.0001 0.1502 3.84

VAR-MODEL

Table 5.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.06584 0.04824 1.36 0.1725 1
AR1_1_1 0.31913 0.02729 11.69 0.0001 CDS (t-1)
AR1_1_2 -0.71310 0.07302 -9.77 0.0001 EQ (t-1)
AR2_1_1 0.09035 0.02640 3.42 0.0006 CDS (t-2)
AR2_1_2 -0.46635 0.07451 -6.26 0.0001 EQ (t-2)
EQ 1 diff Constant2 -0.00135 0.01832 -0.07 0.9411 1
AR1_2_1 0.01561 0.01036 1.51 0.1321 CDS (t-1)
AR1_2_2 -0.08133 0.02772 -2.93 0.0034 EQ (t-1)
AR2_2_1 -0.01704 0.01002 -1.70 0.0893 CDS (t-2)
AR2_2_2 -0.04629 0.02829 -1.64 0.1021 EQ (t-2)

378
GRANGER CAUSALITY TEST

Table 5.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 121.51 <0.0001
2 2 3.69 0.1619

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 5.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.2526 1.76556 113.29 < 0.0001
EQ 1diff 0.0128 0.67035 4.34 0.0017

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 5.5
Variable Durbin Watson Normality Chi Pr > Chi Sq F Value Pr > F
Square
CDS 1diff 2.00200 9999.99 < 0.0001 72.70 < 0.0001
EQ 1diff 1.99387 3173.27 < 0.0001 2.77 0.0963

379
Appendix 34 – MATERIALS
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 8778.257 1.00000 | |********************| 0

1 8759.437 0.99786 | .|********************| 0.027227

2 8735.818 0.99517 | . |********************| 0.047091

3 8706.796 0.99186 | . |********************| 0.060711

4 8675.155 0.98825 | . |********************| 0.071724

5 8642.960 0.98459 | . |********************| 0.081193

6 8608.155 0.98062 | . |********************| 0.089608

7 8572.005 0.97650 | . |********************| 0.097238

8 8534.490 0.97223 | . |******************* | 0.104254

9 8496.768 0.96793 | . |******************* | 0.110771

10 8459.596 0.96370 | . |******************* | 0.116873

11 8421.415 0.95935 | . |******************* | 0.122622

12 8381.491 0.95480 | . |******************* | 0.128065

13 8339.299 0.94999 | . |******************* | 0.133238

14 8296.205 0.94509 | . |******************* | 0.138168

15 8251.391 0.93998 | . |******************* | 0.142879

16 8205.933 0.93480 | . |******************* | 0.147392

17 8160.215 0.92959 | . |******************* | 0.151724

18 8113.992 0.92433 | . |****************** | 0.155889

19 8068.053 0.91910 | . |****************** | 0.159900

20 8022.561 0.91391 | . |****************** | 0.163769

21 7977.688 0.90880 | . |****************** | 0.167507

22 7936.141 0.90407 | . |****************** | 0.171123

380
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

23 7894.749 0.89935 | . |****************** | 0.174628

24 7852.938 0.89459 | . |****************** | 0.178028

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 1.8943 0.9861 1.98 0.9891

1 1.1834 0.9384 0.79 0.8833

2 0.9671 0.9075 0.60 0.8455

Single Mean 0 1.2365 0.9924 0.99 0.9966 2.31 0.4788

1 0.0535 0.9601 0.03 0.9599 0.72 0.8867

2 -0.3373 0.9383 -0.16 0.9409 0.65 0.9061

Trend 0 -0.9722 0.9887 -0.61 0.9778 2.98 0.5776

1 -2.8328 0.9438 -1.14 0.9203 1.75 0.8271

2 -3.2916 0.9242 -1.23 0.9024 1.60 0.8570

Table 1.3 – 1. Difference


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 10.055716 1.00000 | |********************| 0

1 4.208079 0.41848 | .|******** | 0.027237

2 2.305715 0.22929 | .|***** | 0.031649

3 1.096662 0.10906 | .|** | 0.032858

4 0.150518 0.01497 | .|. | 0.033126

5 0.327792 0.03260 | .|* | 0.033131

6 0.135289 0.01345 | .|. | 0.033155

7 0.435272 0.04329 | .|* | 0.033159

8 0.332518 0.03307 | .|* | 0.033201

381
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

9 0.601926 0.05986 | .|* | 0.033225

10 1.185240 0.11787 | .|** | 0.033305

11 0.660461 0.06568 | .|* | 0.033613

12 0.775218 0.07709 | .|** | 0.033708

13 -0.372267 -.03702 | *|. | 0.033838

14 -0.512555 -.05097 | *|. | 0.033869

15 0.428474 0.04261 | .|* | 0.033925

16 0.346977 0.03451 | .|* | 0.033965

17 0.968031 0.09627 | .|** | 0.033991

18 -0.301828 -.03002 | *|. | 0.034193

19 -0.896999 -.08920 | **|. | 0.034212

20 -0.817334 -.08128 | **|. | 0.034384

21 -1.094416 -.10884 | **|. | 0.034527

22 0.042595 0.00424 | .|. | 0.034780

23 0.294423 0.02928 | .|* | 0.034781

24 0.776731 0.07724 | .|** | 0.034799

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -780.473 0.0001 -23.41 <.0001

1 -673.768 0.0001 -18.18 <.0001

2 -686.641 0.0001 -16.27 <.0001

Single Mean 0 -782.591 0.0001 -23.45 <.0001 274.92 0.0010

1 -677.346 0.0001 -18.21 <.0001 165.85 0.0010

2 -692.435 0.0001 -16.31 <.0001 133.01 0.0010

Trend 0 -785.866 0.0001 -23.51 <.0001 276.26 0.0010

1 -682.667 0.0001 -18.26 <.0001 166.77 0.0010

382
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

2 -701.459 0.0001 -16.37 <.0001 133.97 0.0010

AUTOCORRELATION

Graph 2.1 – Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0011

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

383
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1347.3421 <.0001

AR(2) 1347.5942 <.0001

AR(3) 1347.6044 <.0001

AR(4) 1347.6045 <.0001

Graph 2.2 – 1. difference

Table 2.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 1.1618

Table 2.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

384
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 236.3672 <.0001

AR(2) 241.8185 <.0001

AR(3) 241.9057 <.0001

AR(4) 244.6713 <.0001

385
EQUITY

STATIONARITY

Table 3.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 82.265507 1.00000 | |********************| 0

1 81.880211 0.99532 | .|********************| 0.027227

2 81.542050 0.99121 | . |********************| 0.047011

3 81.230599 0.98742 | . |********************| 0.060553

4 80.894581 0.98334 | . |********************| 0.071499

5 80.571173 0.97940 | . |********************| 0.080906

6 80.259535 0.97562 | . |********************| 0.089263

7 79.916965 0.97145 | . |******************* | 0.096845

8 79.591327 0.96749 | . |******************* | 0.103818

9 79.248333 0.96332 | . |******************* | 0.110299

10 78.913534 0.95925 | . |******************* | 0.116369

11 78.594098 0.95537 | . |******************* | 0.122090

12 78.242578 0.95110 | . |******************* | 0.127511

13 77.884626 0.94675 | . |******************* | 0.132666

14 77.511682 0.94221 | . |******************* | 0.137583

15 77.126933 0.93754 | . |******************* | 0.142286

16 76.768805 0.93318 | . |******************* | 0.146794

17 76.354520 0.92815 | . |******************* | 0.151128

18 75.945160 0.92317 | . |****************** | 0.155296

19 75.569350 0.91860 | . |****************** | 0.159312

20 75.192813 0.91403 | . |****************** | 0.163191

21 74.790800 0.90914 | . |****************** | 0.166943

22 74.404646 0.90445 | . |****************** | 0.170574

23 73.978291 0.89926 | . |****************** | 0.174092

24 73.578274 0.89440 | . |****************** | 0.177502

386
Table 3.2 – Original Data
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.2774 0.6199 -0.47 0.5122

1 -0.2669 0.6223 -0.48 0.5058

2 -0.2443 0.6274 -0.47 0.5107

Single Mean 0 -4.3772 0.4984 -1.37 0.5991 0.96 0.8249

1 -3.6457 0.5798 -1.22 0.6673 0.78 0.8712

2 -3.1415 0.6397 -1.12 0.7103 0.66 0.9023

Trend 0 -3.6256 0.9075 -1.08 0.9299 1.23 0.9301

1 -2.8603 0.9428 -0.92 0.9525 1.11 0.9525

2 -2.2763 0.9628 -0.78 0.9663 1.12 0.9518

387
Table 3.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.622678 1.00000 | |********************| 0

1 -0.044814 -.07197 | *|. | 0.027237

2 -0.035961 -.05775 | *|. | 0.027377

3 0.016279 0.02614 | .|* | 0.027468

4 -0.0050114 -.00805 | .|. | 0.027486

5 -0.0078409 -.01259 | .|. | 0.027488

6 0.014364 0.02307 | .|. | 0.027492

7 -0.016833 -.02703 | *|. | 0.027507

8 0.012666 0.02034 | .|. | 0.027526

9 -0.018773 -.03015 | *|. | 0.027537

10 0.00072302 0.00116 | .|. | 0.027562

11 0.030825 0.04950 | .|* | 0.027562

12 0.00042150 0.00068 | .|. | 0.027628

13 0.0089060 0.01430 | .|. | 0.027628

14 0.013307 0.02137 | .|. | 0.027633

15 -0.044552 -.07155 | *|. | 0.027646

16 0.054861 0.08810 | .|** | 0.027783

17 0.0049485 0.00795 | .|. | 0.027989

18 -0.030125 -.04838 | *|. | 0.027991

19 -0.0059309 -.00952 | .|. | 0.028053

20 0.018304 0.02940 | .|* | 0.028055

21 -0.0031168 -.00501 | .|. | 0.028078

22 0.037270 0.05986 | .|* | 0.028079

23 -0.017136 -.02752 | *|. | 0.028173

24 0.0022805 0.00366 | .|. | 0.028193

Table 3.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

388
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1443.89 0.0001 -39.44 <.0001

1 -1637.28 0.0001 -28.59 <.0001

2 -1547.24 0.0001 -22.12 <.0001

Single Mean 0 -1443.95 0.0001 -39.43 <.0001 777.38 0.0010

1 -1637.53 0.0001 -28.58 <.0001 408.31 0.0010

2 -1547.76 0.0001 -22.11 <.0001 244.52 0.0010

Trend 0 -1445.29 0.0001 -39.45 <.0001 778.29 0.0010

1 -1643.02 0.0001 -28.61 <.0001 409.32 0.0010

2 -1559.37 0.0001 -22.16 <.0001 245.44 0.0010

AUTOCORRELATION

Graph 4.1 – Original data

Table 4.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0076

Table 4.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

389
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1338.3227 <.0001

AR(2) 1338.3792 <.0001

AR(3) 1338.4137 <.0001

AR(4) 1338.4181 <.0001

Graph 4.2 – 1. difference

Table 4.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 2.1432

Table 4.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

390
Alternative LM Pr > LM

AR(1) 6.9823 0.0082

AR(2) 12.3551 0.0021

AR(3) 12.7537 0.0052

AR(4) 12.8492 0.0120

MODEL

CO-INTEGRATION

Table 5.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0024 3.2938 15.34
1 1 0.0000 0.0388 3.84

VAR-MODEL

Table 5.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.08409 0.07623 1.10 0.2702 1
AR1_1_1 0.33918 0.02729 12.43 0.0001 CDS (t-1)
AR1_1_2 -0.85551 0.09787 -8.74 0.0001 EQ (t-1)
AR2_1_1 0.10626 0.02714 3.91 0.0001 CDS (t-2)
AR2_1_2 -0.39305 0.09925 -3.96 0.0001 EQ (t-2)
EQ 1 diff Constant2 -0.00908 0.02145 -0.42 0.6722 1
AR1_2_1 0.01706 0.00768 2.22 0.0265 CDS (t-1)
AR1_2_2 -0.07155 0.02754 -2.60 0.0095 EQ (t-1)
AR2_2_1 0.00308 0.00764 0.40 0.6870 CDS (t-2)
AR2_2_2 -0.04716 0.02792 -1.69 0.0915 EQ (t-2)

GRANGER CAUSALITY TEST

Table 5.3
Granger Causality Wald Test

391
Test DF Chi-Square Pr > ChiSq
1 2 86.32 < 0.0001
2 2 7.17 0.0277

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 5.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.2291 2.79033 99.64 < 0.0001
EQ 1diff 0.0144 0.78508 4.90 < 0.0006

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 5.5
Variable Durbin Watson Normality Chi Pr > Chi Sq F Value Pr > F
Square
CDS 1diff 2.01153 9999.99 < 0.0001 88.99 < 0.0001
EQ 1diff 1.99761 1160.02 < 0.0001 17.97 < 0.0001

392
Appendix 35 - TECHNOLOGY
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 24993.343 1.00000 | |********************| 0

1 24943.373 0.99800 | .|********************| 0.027227

2 24877.597 0.99537 | . |********************| 0.047095

3 24803.204 0.99239 | . |********************| 0.060719

4 24718.809 0.98902 | . |********************| 0.071742

5 24629.097 0.98543 | . |********************| 0.081223

6 24539.159 0.98183 | . |********************| 0.089648

7 24442.865 0.97798 | . |********************| 0.097293

8 24342.943 0.97398 | . |******************* | 0.104326

9 24237.792 0.96977 | . |******************* | 0.110862

10 24126.997 0.96534 | . |******************* | 0.116982

11 24012.732 0.96077 | . |******************* | 0.122745

12 23889.545 0.95584 | . |******************* | 0.128198

13 23762.091 0.95074 | . |******************* | 0.133377

14 23628.127 0.94538 | . |******************* | 0.138309

15 23491.411 0.93991 | . |******************* | 0.143019

16 23349.498 0.93423 | . |******************* | 0.147527

17 23202.561 0.92835 | . |******************* | 0.151849

18 23048.094 0.92217 | . |****************** | 0.156000

19 22892.944 0.91596 | . |****************** | 0.159990

20 22738.844 0.90980 | . |****************** | 0.163831

21 22583.976 0.90360 | . |****************** | 0.167534

22 22430.654 0.89747 | . |****************** | 0.171109

393
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

23 22276.580 0.89130 | . |****************** | 0.174563

24 22120.750 0.88507 | . |****************** | 0.177905

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.5964 0.8312 0.44 0.8082

1 -0.0460 0.6725 -0.03 0.6746

2 -0.0470 0.6723 -0.03 0.6745

Single Mean 0 -0.5247 0.9255 -0.30 0.9226 0.62 0.9170

1 -1.5824 0.8268 -0.70 0.8452 0.59 0.9264

2 -1.5859 0.8264 -0.70 0.8451 0.59 0.9263

Trend 0 -3.0016 0.9370 -1.30 0.8862 1.42 0.8945

1 -4.7601 0.8377 -1.60 0.7929 1.60 0.8578

2 -4.7670 0.8372 -1.60 0.7932 1.60 0.8581

394
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 56.783705 1.00000 | |********************| 0

1 14.154276 0.24927 | .|***** | 0.027237

2 3.467087 0.06106 | .|* | 0.028880

3 4.565089 0.08039 | .|** | 0.028975

4 4.336537 0.07637 | .|** | 0.029140

5 -1.252798 -.02206 | .|. | 0.029288

6 2.923055 0.05148 | .|* | 0.029301

7 2.650856 0.04668 | .|* | 0.029368

8 4.394004 0.07738 | .|** | 0.029423

9 5.169427 0.09104 | .|** | 0.029573

10 1.650272 0.02906 | .|* | 0.029780

11 1.221450 0.02151 | .|. | 0.029801

12 1.606996 0.02830 | .|* | 0.029813

13 3.235877 0.05699 | .|* | 0.029833

14 -0.273949 -.00482 | .|. | 0.029913

15 -0.182206 -.00321 | .|. | 0.029914

16 4.474932 0.07881 | .|** | 0.029914

17 8.294841 0.14608 | .|*** | 0.030068

18 0.899745 0.01585 | .|. | 0.030590

19 -2.398409 -.04224 | *|. | 0.030596

20 0.262446 0.00462 | .|. | 0.030639

21 -1.166632 -.02055 | .|. | 0.030640

22 0.877324 0.01545 | .|. | 0.030650

23 1.940857 0.03418 | .|* | 0.030656

24 0.168153 0.00296 | .|. | 0.030684

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

395
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1010.00 0.0001 -28.40 <.0001

1 -1008.63 0.0001 -22.35 <.0001

2 -827.718 0.0001 -17.58 <.0001

Single Mean 0 -1010.88 0.0001 -28.41 <.0001 403.47 0.0010

1 -1010.73 0.0001 -22.36 <.0001 249.98 0.0010

2 -830.675 0.0001 -17.59 <.0001 154.68 0.0010

Trend 0 -1011.75 0.0001 -28.41 <.0001 403.68 0.0010

1 -1012.89 0.0001 -22.37 <.0001 250.23 0.0010

2 -833.741 0.0001 -17.60 <.0001 154.86 0.0010

AUTOCORRELATION

Graph 2.1 – Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0025

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

396
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1345.4414 <.0001

AR(2) 1345.6153 <.0001

AR(3) 1345.6154 <.0001

AR(4) 1345.6320 <.0001

Graph 2.2 – 1. difference

Table 2.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 1.5333

Table 2.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

397
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 73.2529 <.0001

AR(2) 73.4016 <.0001

AR(3) 80.2641 <.0001

AR(4) 82.7528 <.0001

398
EQUITY

STATIONARITY

Table 3.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 36.997872 1.00000 | |********************| 0

1 36.867020 0.99646 | .|********************| 0.027227

2 36.742200 0.99309 | . |********************| 0.047047

3 36.626338 0.98996 | . |********************| 0.060626

4 36.495172 0.98641 | . |********************| 0.071614

5 36.363491 0.98285 | . |********************| 0.081062

6 36.236296 0.97942 | . |********************| 0.089461

7 36.113911 0.97611 | . |********************| 0.097085

8 35.992719 0.97283 | . |******************* | 0.104106

9 35.866040 0.96941 | . |******************* | 0.110640

10 35.738431 0.96596 | . |******************* | 0.116766

11 35.613937 0.96259 | . |******************* | 0.122547

12 35.480939 0.95900 | . |******************* | 0.128029

13 35.328127 0.95487 | . |******************* | 0.133248

14 35.180268 0.95087 | . |******************* | 0.138227

15 35.037164 0.94700 | . |******************* | 0.142994

16 34.889043 0.94300 | . |******************* | 0.147570

17 34.730842 0.93873 | . |******************* | 0.151971

18 34.579925 0.93465 | . |******************* | 0.156210

19 34.440694 0.93088 | . |******************* | 0.160302

20 34.292489 0.92688 | . |******************* | 0.164261

21 34.137799 0.92270 | . |****************** | 0.168093

22 33.989442 0.91869 | . |****************** | 0.171806

23 33.833186 0.91446 | . |****************** | 0.175410

24 33.677837 0.91026 | . |****************** | 0.178909

399
Table 3.2 – Original Data
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.3978 0.5927 -0.90 0.3255

1 -0.4114 0.5896 -0.98 0.2907

2 -0.4036 0.5914 -1.02 0.2760

Single Mean 0 -1.0679 0.8799 -0.44 0.8996 0.45 0.9642

1 -0.7971 0.9042 -0.35 0.9155 0.50 0.9527

2 -0.5068 0.9268 -0.23 0.9318 0.52 0.9458

Trend 0 -3.4967 0.9142 -1.26 0.8971 1.65 0.8470

1 -2.9965 0.9372 -1.13 0.9217 1.47 0.8828

2 -2.6019 0.9524 -1.04 0.9364 1.47 0.8835

400
Table 3.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.159942 1.00000 | |********************| 0

1 -0.0081138 -.05073 | *|. | 0.027237

2 -0.0088257 -.05518 | *|. | 0.027307

3 0.011377 0.07113 | .|* | 0.027389

4 -0.0002268 -.00142 | .|. | 0.027526

5 -0.0028186 -.01762 | .|. | 0.027526

6 -0.0065333 -.04085 | *|. | 0.027534

7 -0.0004433 -.00277 | .|. | 0.027579

8 0.0054786 0.03425 | .|* | 0.027580

9 -0.0021378 -.01337 | .|. | 0.027611

10 0.0013237 0.00828 | .|. | 0.027616

11 0.0077277 0.04832 | .|* | 0.027618

12 0.013927 0.08708 | .|** | 0.027680

13 -0.0044285 -.02769 | *|. | 0.027883

14 -0.0004518 -.00282 | .|. | 0.027903

15 0.00065713 0.00411 | .|. | 0.027903

16 0.0072044 0.04504 | .|* | 0.027904

17 -0.0016054 -.01004 | .|. | 0.027958

18 -0.0088390 -.05526 | *|. | 0.027960

19 0.0053766 0.03362 | .|* | 0.028041

20 0.00081319 0.00508 | .|. | 0.028071

21 -0.0095241 -.05955 | *|. | 0.028072

22 0.0055897 0.03495 | .|* | 0.028165

23 0.0026859 0.01679 | .|. | 0.028198

24 0.0023691 0.01481 | .|. | 0.028205

Table 3.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

401
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1414.55 0.0001 -38.63 <.0001

1 -1585.81 0.0001 -28.16 <.0001

2 -1291.42 0.0001 -20.86 <.0001

Single Mean 0 -1415.35 0.0001 -38.64 <.0001 746.52 0.0010

1 -1588.87 0.0001 -28.18 <.0001 397.02 0.0010

2 -1296.43 0.0001 -20.88 <.0001 218.06 0.0010

Trend 0 -1417.07 0.0001 -38.67 <.0001 747.72 0.0010

1 -1595.19 0.0001 -28.22 <.0001 398.20 0.0010

2 -1306.69 0.0001 -20.93 <.0001 218.98 0.0010

AUTOCORRELATION

Graph 4.1 – Original data

Table 4.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0039

Table 4.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

402
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1343.6134 <.0001

AR(2) 1343.6221 <.0001

AR(3) 1343.6382 <.0001

AR(4) 1343.6642 <.0001

Graph 4.2 – 1. difference

Table 4.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 2.0982

Table 4.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

403
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 3.4402 0.0636

AR(2) 7.9636 0.0187

AR(3) 14.8920 0.0019

AR(4) 14.9742 0.0048

MODEL

CO-INTEGRATION

Table 5.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0145 19.8253 15.34
1 1 0.0001 0.1111 3.84

VECM-MODEL

Table 5.2 – Long-Run Parameter Beta Estimates When RANK = 1


Variable 1
CDS 0.37201
EQUITY 0.01378

Table 5.3 – Adjustment Coefficient Alpha Estimates When RANK = 1


Variable 1
CDS -0.00032
EQUITY -0.87574

404
Table 5.4
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS Constant1 -0.00577 0.15211 -0.04 0.9697 1
AR1_1_1 -0.00012 0.00404 CDS (t-1)
AR1_1_2 0 0.00015 EQ (t-1)
AR2_1_1 -0.0502 0.02744 -1.83 0.0676 D_CDS (t-1)
AR2_1_2 0.00054 0.00144 0.37 0.7091 D_EQ (t-1)
EQ Constant2 12.37582 2.75475 4.49 0.0001 1
AR1_2_1 -0.32578 0.07323 CDS (t-1)
AR1_2_2 -0.01206 0.00271 EQ (t-1)
AR2_2_1 -1.12926 0.49695 -2.27 0.0232 D_CDS (t-1)
AR2_2_2 0.24236 0.02617 9.26 0.0001 D_EQ (t-1)

GRANGER CAUSALITY TEST

Table 5.5
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 27.64 <.0001
2 2 0.13 0.9363

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 5.6
Variable R-Square Std. Deviation F Value Pr > F
CDS 0.0027 0.39893 0.9 0.461
EQ 0.0814 7.22477 29.73 <.0001

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 5.7
Variable Durbin Watson Normality Chi Pr > Chi Sq F Value Pr > F
Square
CDS 2.00696 2885.56 <.0001 0.36 0.5471
EQ 2.00653 9999.99 <.0001 17.42 <.0001

405
Appendix 36 - UTILITIES
CDS

STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 5269.100 1.00000 | |********************| 0

1 5254.448 0.99722 | .|********************| 0.027227

2 5236.872 0.99388 | . |********************| 0.047071

3 5216.790 0.99007 | . |********************| 0.060664

4 5195.090 0.98595 | . |********************| 0.071648

5 5172.598 0.98169 | . |********************| 0.081084

6 5149.365 0.97728 | . |********************| 0.089462

7 5125.314 0.97271 | . |******************* | 0.097054

8 5100.845 0.96807 | . |******************* | 0.104030

9 5076.781 0.96350 | . |******************* | 0.110506

10 5053.481 0.95908 | . |******************* | 0.116567

11 5029.550 0.95454 | . |******************* | 0.122277

12 5005.013 0.94988 | . |******************* | 0.127681

13 4979.988 0.94513 | . |******************* | 0.132816

14 4954.142 0.94023 | . |******************* | 0.137712

15 4928.560 0.93537 | . |******************* | 0.142391

16 4903.013 0.93052 | . |******************* | 0.146875

17 4877.518 0.92568 | . |******************* | 0.151182

18 4851.810 0.92080 | . |****************** | 0.155327

19 4825.678 0.91584 | . |****************** | 0.159322

20 4799.633 0.91090 | . |****************** | 0.163178

21 4773.904 0.90602 | . |****************** | 0.166905

22 4749.265 0.90134 | . |****************** | 0.170511

406
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

23 4724.370 0.89662 | . |****************** | 0.174008

24 4699.036 0.89181 | . |****************** | 0.177399

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 1.5397 0.9701 1.56 0.9711

1 1.1954 0.9398 0.93 0.9060

2 0.9860 0.9107 0.68 0.8632

Single Mean 0 0.7816 0.9848 0.47 0.9858 1.37 0.7189

1 -0.1898 0.9473 -0.09 0.9489 0.75 0.8794

2 -0.8141 0.9028 -0.34 0.9169 0.67 0.9004

Trend 0 -1.0042 0.9883 -0.51 0.9829 1.60 0.8575

1 -2.3241 0.9614 -0.92 0.9528 1.26 0.9243

2 -3.1122 0.9323 -1.10 0.9279 1.23 0.9301

407
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 10.647735 1.00000 | |********************| 0

1 2.745465 0.25784 | .|***** | 0.027237

2 1.779954 0.16717 | .|*** | 0.028991

3 1.081041 0.10153 | .|** | 0.029698

4 0.783882 0.07362 | .|* | 0.029954

5 0.626287 0.05882 | .|* | 0.030088

6 0.646695 0.06074 | .|* | 0.030173

7 0.202470 0.01902 | .|. | 0.030264

8 -0.474517 -.04457 | *|. | 0.030272

9 -0.059165 -.00556 | .|. | 0.030321

10 0.380493 0.03573 | .|* | 0.030322

11 0.072110 0.00677 | .|. | 0.030353

12 0.022059 0.00207 | .|. | 0.030354

13 -0.372897 -.03502 | *|. | 0.030354

14 -0.477776 -.04487 | *|. | 0.030384

15 -0.262204 -.02463 | .|. | 0.030433

16 -0.108146 -.01016 | .|. | 0.030448

17 0.307008 0.02883 | .|* | 0.030451

18 0.404897 0.03803 | .|* | 0.030471

19 -0.169795 -.01595 | .|. | 0.030506

20 -0.302990 -.02846 | *|. | 0.030512

21 -0.518988 -.04874 | *|. | 0.030532

22 0.264432 0.02483 | .|. | 0.030590

23 0.494075 0.04640 | .|* | 0.030605

24 0.024570 0.00231 | .|. | 0.030657

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

408
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -997.747 0.0001 -28.14 <.0001

1 -800.524 0.0001 -19.98 <.0001

2 -722.357 0.0001 -16.83 <.0001

Single Mean 0 -999.636 0.0001 -28.17 <.0001 396.72 0.0010

1 -803.762 0.0001 -20.01 <.0001 200.26 0.0010

2 -726.901 0.0001 -16.86 <.0001 142.14 0.0010

Trend 0 -1001.82 0.0001 -28.21 <.0001 397.77 0.0010

1 -807.382 0.0001 -20.05 <.0001 200.93 0.0010

2 -732.159 0.0001 -16.90 <.0001 142.76 0.0010

AUTOCORRELATION

Graph 2.1 – Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0020

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

409
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1346.0281 <.0001

AR(2) 1346.1967 <.0001

AR(3) 1346.2296 <.0001

AR(4) 1346.2332 <.0001

Graph 2.2 – 1. difference

Table 2.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 1.4841

Table 2.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

410
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 89.6325 <.0001

AR(2) 104.3452 <.0001

AR(3) 106.1970 <.0001

AR(4) 107.0599 <.0001

411
EQUITY

STATIONARITY

Table 3.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 35.478533 1.00000 | |********************| 0

1 35.320726 0.99555 | .|********************| 0.027227

2 35.177059 0.99150 | . |********************| 0.047018

3 35.049269 0.98790 | . |********************| 0.060566

4 34.911052 0.98400 | . |********************| 0.071520

5 34.770659 0.98005 | . |********************| 0.080936

6 34.632784 0.97616 | . |********************| 0.089301

7 34.500936 0.97245 | . |******************* | 0.096889

8 34.381309 0.96907 | . |******************* | 0.103872

9 34.244012 0.96520 | . |******************* | 0.110371

10 34.121553 0.96175 | . |******************* | 0.116460

11 33.999180 0.95830 | . |******************* | 0.122206

12 33.864873 0.95452 | . |******************* | 0.127655

13 33.721592 0.95048 | . |******************* | 0.132840

14 33.578072 0.94643 | . |******************* | 0.137789

15 33.431651 0.94231 | . |******************* | 0.142527

16 33.291457 0.93835 | . |******************* | 0.147073

17 33.142647 0.93416 | . |******************* | 0.151446

18 32.997347 0.93007 | . |******************* | 0.155658

19 32.863716 0.92630 | . |******************* | 0.159725

20 32.727063 0.92245 | . |****************** | 0.163659

21 32.586493 0.91848 | . |****************** | 0.167468

22 32.449157 0.91461 | . |****************** | 0.171162

23 32.301806 0.91046 | . |****************** | 0.174747

24 32.153029 0.90627 | . |****************** | 0.178229

412
Table 3.2 – Original Data
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -0.2145 0.6342 -0.48 0.5087

1 -0.1980 0.6379 -0.48 0.5074

2 -0.1840 0.6411 -0.49 0.5049

Single Mean 0 -2.8041 0.6809 -0.99 0.7597 0.54 0.9406

1 -2.1598 0.7597 -0.83 0.8106 0.40 0.9739

2 -1.6244 0.8222 -0.68 0.8505 0.30 0.9896

Trend 0 -1.8390 0.9741 -0.63 0.9767 1.52 0.8739

1 -1.1361 0.9866 -0.42 0.9866 1.68 0.8406

2 -0.5480 0.9928 -0.22 0.9925 1.95 0.7875

413
Table 3.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.211708 1.00000 | |********************| 0

1 -0.018419 -.08700 | **|. | 0.027237

2 -0.016377 -.07735 | **|. | 0.027442

3 0.0092569 0.04372 | .|* | 0.027603

4 0.0023712 0.01120 | .|. | 0.027655

5 -0.0045681 -.02158 | .|. | 0.027658

6 -0.0062197 -.02938 | *|. | 0.027671

7 -0.012775 -.06034 | *|. | 0.027694

8 0.020908 0.09876 | .|** | 0.027791

9 -0.016635 -.07857 | **|. | 0.028050

10 0.0033217 0.01569 | .|. | 0.028213

11 0.011674 0.05514 | .|* | 0.028219

12 0.0087755 0.04145 | .|* | 0.028299

13 -0.0008556 -.00404 | .|. | 0.028344

14 0.0080236 0.03790 | .|* | 0.028345

15 -0.0075815 -.03581 | *|. | 0.028382

16 0.0049712 0.02348 | .|. | 0.028416

17 -0.0023877 -.01128 | .|. | 0.028430

18 -0.0089748 -.04239 | *|. | 0.028434

19 0.0027290 0.01289 | .|. | 0.028480

20 0.0016056 0.00758 | .|. | 0.028485

21 -0.0036523 -.01725 | .|. | 0.028486

22 0.0077155 0.03644 | .|* | 0.028494

23 0.0029206 0.01380 | .|. | 0.028529

24 -0.0054324 -.02566 | *|. | 0.028533

Table 3.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

414
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -1464.17 0.0001 -40.01 <.0001

1 -1736.89 0.0001 -29.43 <.0001

2 -1577.57 0.0001 -22.26 <.0001

Single Mean 0 -1464.30 0.0001 -40.00 <.0001 800.04 0.0010

1 -1737.42 0.0001 -29.43 <.0001 432.96 0.0010

2 -1578.66 0.0001 -22.25 <.0001 247.61 0.0010

Trend 0 -1467.29 0.0001 -40.07 <.0001 802.94 0.0010

1 -1750.42 0.0001 -29.52 <.0001 435.80 0.0010

2 -1604.60 0.0001 -22.35 <.0001 249.87 0.0010

AUTOCORRELATION

Graph 4.1 – Original data

Table 4.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0060

Table 4.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

415
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1340.2072 <.0001

AR(2) 1340.2602 <.0001

AR(3) 1340.3101 <.0001

AR(4) 1340.3161 <.0001

Graph 4.2 – 1. difference

Table 4.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 2.1731

Table 4.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

416
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 10.2127 0.0014

AR(2) 20.0209 <.0001

AR(3) 21.1660 <.0001

AR(4) 21.3425 0.0003

MODEL

CO-INTEGRATION

Table 5.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0054 7.7109 15.34
1 1 0.0003 0.3954 3.84

VAR-MODEL

Table 5.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 0.10443 0.08551 1.22 0.2222 1
AR1_1_1 0.23780 0.02658 8.95 0.0001 CDS (t-1)
AR1_1_2 -0.81805 0.18857 -4.34 0.0001 EQ (t-1)
EQ 1 diff Constant2 -0.00399 0.01252 -0.32 0.7499 1
AR1_2_1 -0.00310 0.00389 -0.80 0.4260 CDS (t-1)
AR1_2_2 -0.09091 0.02760 -3.29 0.0010 EQ (t-1)

GRANGER CAUSALITY TEST

417
Table 5.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 18.82 < 0.0001
2 2 0.63 0.4258

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff
Test 2: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff

UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 5.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.0794 3.13547 57.95 < 0.0001
EQ 1diff 0.0080 0.45894 5.45 0.0044

UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 5.5
Variable Durbin Watson Normality Chi Pr > Chi Sq F Value Pr > F
Square
CDS 1diff 2.07182 9999.99 < 0.0001 12.13 0.0005
EQ 1diff 2.01630 1700.56 < 0.0001 32.63 < 0.0001

418
Appendix 37 – 1 WEEK
STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.073985 1.00000 | |********************| 0

1 0.065260 0.88208 | .|****************** | 0.027267

2 0.055616 0.75172 | . |*************** | 0.043594

3 0.049283 0.66612 | . |************* | 0.052352

4 0.047160 0.63742 | . |************* | 0.058314

5 0.048133 0.65058 | . |************* | 0.063283

6 0.048105 0.65020 | . |************* | 0.068074

7 0.047308 0.63942 | . |************* | 0.072545

8 0.047246 0.63860 | . |************* | 0.076620

9 0.046595 0.62978 | . |************* | 0.080480

10 0.045301 0.61230 | . |************ | 0.084065

11 0.044116 0.59629 | . |************ | 0.087318

12 0.043785 0.59181 | . |************ | 0.090294

13 0.043181 0.58364 | . |************ | 0.093134

14 0.042467 0.57399 | . |*********** | 0.095814

15 0.041226 0.55722 | . |*********** | 0.098338

16 0.040508 0.54752 | . |*********** | 0.100658

17 0.040174 0.54300 | . |*********** | 0.102848

18 0.040247 0.54399 | . |*********** | 0.104958

19 0.039792 0.53784 | . |*********** | 0.107034

20 0.038928 0.52616 | . |*********** | 0.109025

419
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

21 0.038041 0.51418 | . |********** | 0.110897

22 0.037560 0.50767 | . |********** | 0.112655

23 0.036616 0.49491 | . |********** | 0.114343

24 0.035887 0.48506 | . |********** | 0.115925

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -54.2277 <.0001 -5.25 <.0001

1 -62.7774 <.0001 -5.58 <.0001

2 -43.7181 <.0001 -4.62 <.0001

Single Mean 0 -158.486 0.0001 -9.18 <.0001 42.12 0.0010

1 -201.011 0.0001 -10.03 <.0001 50.33 0.0010

2 -151.235 0.0001 -8.47 <.0001 35.86 0.0010

Trend 0 -235.331 0.0001 -11.32 <.0001 64.12 0.0010

1 -320.749 0.0001 -12.62 <.0001 79.65 0.0010

2 -258.276 0.0001 -10.80 <.0001 58.34 0.0010

420
AUTOCORRELATION

Graph 2.1 – Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.2354

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation
Godfrey's Serial Correlation
Test

Alternative LM Pr > LM

AR(1) 1045.7741 <.0001

AR(2) 1049.9741 <.0001

AR(3) 1055.1575 <.0001

AR(4) 1065.2651 <.0001

421
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.1664 249.8462 29.38
1 1 0.0039 5.3764 15.34
2 2 0.0001 0.1229 3.84

VECM-MODEL

Table 3.2 – Long-Run Parameter Beta Estimates When RANK = 1

Variable 1
CDS -0.00842
EQUITY -0.05680
oneweek 4.62755

Table 3.3 – Adjustment Coefficient Alpha Estimates When RANK = 1


Variable 1
CDS 0.79743
EQUITY -0.00897
oneweek -0.03940

Table 3.4
Equation Parameter Estimate Std. Error t-value Pr > Variable
|t|
CDS Constant1 1.39677 0.15630 8.94 0.0001 1
AR1_1_1 -0.00671 0.00073 CDS (t-1)
AR1_1_2 -0.04529 0.00491 EQ (t-1)
AR1_1_3 3.69015 0.39969 Oneweek (t-1)
EQUITY Constant2 -0.01831 0.02445 -0.75 0.4541 1
AR1_2_1 0.00008 0.00011 CDS (t-1)
AR1_2_2 0.00051 0.00077 EQ (t-1)
AR1_2_3 -0.04150 0.06254 Oneweek (t-1)
Oneweek Constant3 -0.05933 0.00622 -9.53 0.0001 1
AR1_3_1 0.00033 0.00003 CDS (t-1)
AR1_3_2 0.00224 0.00020 EQ (t-1)
AR1_3_3 -0.18233 0.01592 Oneweek (t-1)

422
GRANGER CAUSALITY TEST

Table 3.5
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 4 241.02 <0.0001
2 2 234.83 <0.0001
3 2 9.72 0.0079
4 4 96.32 <0.0001
5 2 26.46 <0.0001
6 2 77.60 <0.0001
7 4 16.62 0.0023
8 2 0.89 0.6398
9 2 15.80 0.0004

Test 1: Group 1 Variable: CDS


Group 2 Variable: EQ oneweek
Test 2: Group 1 Variable: CDS
Group 2 Variable: EQ
Test 3: Group 1 Variable: CDS
Group 2 Variable: oneweek
Test 4: Group 1 Variable: oneweek
Group 2 Variable: CDS EQ
Test 5: Group 1 Variable: oneweek
Group 2 Variable: CDS
Test 6: Group 1 Variable: oneweek
Group 2 Variable: EQ
Test 7: Group 1 Variable: EQ
Group 2 Variable: CDS oneweek
Test 8: Group 1 Variable: EQ
Group 2 Variable: CDS
Test 9: Group 1 Variable: EQ
Group 2 Variable: oneweek

UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.6
Variable R-Square Std. Deviation F Value Pr > F
CDS 0.0597 3.16527 28.33 < 0.0001
EQ 0.0003 0.49525 0.15 0.9320
oneweek 0.0890 0.12604 43.62 < 0.0001

423
UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.7
Variable Durbin Watson Normality Chi Pr > Chi Sq F Value Pr > F
Square
CDS 1.10073 9999.99 < 0.0001 247.66 < 0.0001
EQ 2.21045 1770.06 < 0.0001 27.27 < 0.0001
oneweek 1.73654 1663.80 < 0.0001 67.67 < 0.0001

424
APPENDIX 38 - 1 MONTH
STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.248190 1.00000 | |********************| 0

1 0.246165 0.99184 | .|********************| 0.027431

2 0.241324 0.97234 | . |******************* | 0.047253

3 0.234456 0.94466 | . |******************* | 0.060462

4 0.226020 0.91067 | . |****************** | 0.070701

5 0.216462 0.87216 | . |***************** | 0.079036

6 0.206170 0.83069 | . |***************** | 0.085973

7 0.195553 0.78792 | . |**************** | 0.091814

8 0.184913 0.74504 | . |*************** | 0.096768

9 0.174311 0.70233 | . |************** | 0.100992

10 0.163994 0.66076 | . |************* | 0.104603

11 0.154174 0.62119 | . |************ | 0.107698

12 0.144935 0.58397 | . |************ | 0.110361

13 0.136360 0.54942 | . |*********** | 0.112662

14 0.128561 0.51799 | . |********** | 0.114660

15 0.121655 0.49017 | . |********** | 0.116408

16 0.115705 0.46620 | . |********* | 0.117950

17 0.110614 0.44568 | . |********* | 0.119329

18 0.106324 0.42840 | . |********* | 0.120575

19 0.102802 0.41421 | . |******** | 0.121715

20 0.099928 0.40263 | . |******** | 0.122771

425
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

21 0.097486 0.39279 | . |******** | 0.123760

22 0.095326 0.38409 | . |******** | 0.124695

23 0.093391 0.37629 | . |******** | 0.125582

24 0.091692 0.36944 | . |******* | 0.126427

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -3.2199 0.2178 -1.21 0.2067

1 -21.1882 0.0011 -3.23 0.0013

2 -21.9638 0.0009 -3.28 0.0011

Single Mean 0 -10.8354 0.1103 -2.37 0.1510 2.82 0.3460

1 -65.5748 0.0019 -5.75 <.0001 16.51 0.0010

2 -70.1872 0.0019 -5.86 <.0001 17.18 0.0010

Trend 0 -14.5938 0.1987 -2.65 0.2572 3.56 0.4604

1 -94.6111 0.0008 -6.83 <.0001 23.34 0.0010

2 -103.912 0.0001 -7.02 <.0001 24.64 0.0010

426
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.0039264 1.00000 | |********************| 0

1 0.0028085 0.71530 | .|************** | 0.027451

2 0.0020284 0.51662 | . |********** | 0.039048

3 0.0015710 0.40012 | . |******** | 0.043898

4 0.0011231 0.28604 | . |****** | 0.046565

5 0.00073806 0.18798 | . |**** | 0.047871

6 0.00032246 0.08213 | . |** | 0.048424

7 0.00002264 0.00577 | .|. | 0.048529

8 -0.0000297 -.00756 | .|. | 0.048529

9 -0.0002787 -.07099 | .*| . | 0.048530

10 -0.0004949 -.12605 | ***| . | 0.048608

11 -0.0005798 -.14767 | ***| . | 0.048854

12 -0.0006571 -.16735 | ***| . | 0.049189

13 -0.0007769 -.19787 | ****| . | 0.049616

14 -0.0008958 -.22815 | *****| . | 0.050207

15 -0.0009608 -.24472 | *****| . | 0.050983

16 -0.0008681 -.22108 | ****| . | 0.051860

17 -0.0008105 -.20644 | ****| . | 0.052566

18 -0.0007724 -.19673 | ****| . | 0.053173

19 -0.0006569 -.16731 | ***| . | 0.053719

20 -0.0004399 -.11204 | **| . | 0.054110

21 -0.0002834 -.07219 | .*| . | 0.054285

22 -0.0002283 -.05816 | .*| . | 0.054357

23 -0.0002373 -.06044 | .*| . | 0.054404

427
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 -0.0001474 -.03755 | .*| . | 0.054454

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.

Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -377.020 0.0001 -14.81 <.0001

1 -369.872 0.0001 -13.58 <.0001

2 -325.816 0.0001 -12.03 <.0001

Single Mean 0 -377.033 0.0001 -14.81 <.0001 109.64 0.0010

1 -369.887 0.0001 -13.58 <.0001 92.19 0.0010

2 -325.831 0.0001 -12.02 <.0001 72.26 0.0010

Trend 0 -377.062 0.0001 -14.80 <.0001 109.57 0.0010

1 -369.926 0.0001 -13.57 <.0001 92.14 0.0010

2 -325.875 0.0001 -12.02 <.0001 72.22 0.0010

428
AUTOCORRELATION

Graph 2.1 – Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0158

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 1306.4204 <.0001

AR(2) 1317.0495 <.0001

AR(3) 1317.0868 <.0001

AR(4) 1317.1560 <.0001

Graph 2.2 – 1. difference

429
Table 2.3 – 1. difference
Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.5687

Table 2.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 679.3119 <.0001

AR(2) 679.3663 <.0001

AR(3) 681.3697 <.0001

AR(4) 682.6528 <.0001

430
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0532 78.6827 29.38
1 1 0.0045 6.2634 15.34
2 2 0.0003 0.3458 3.84

VECM-MODEL

Table 3.2 – Long-Run Parameter Beta Estimates When RANK = 1


Variable 1
CDS -0.00688
EQUITY -0.03744
1month 2.55641

Table 3.3 – Adjustment Coefficient Alpha Estimates When RANK = 1


Variable 1
CDS 0.14858
EQUITY -0.01490
1month -0.00851

431
Table 3.4
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS Constant1 0.17045 0.08254 2.07 0.0391 1
AR1_1_1 -0.00102 0.00051 CDS (t-1)
AR1_1_2 -0.00556 0.00275 EQ (t-1)
AR1_1_3 0.37983 0.18800 1month (t-1)
AR2_1_1 0.41672 0.02376 17.54 0.0001 D_CDS (t-1)
AR2_1_2 -1.97129 0.15326 -12.86 0.0001 D_EQ (t-1)
AR2_1_3 4.29509 1.26134 3.41 0.0007 D_1month (t-1)
EQ Constant2 -0.01475 0.01521 -0.97 0.3324 1
AR1_2_1 0.00010 0.00009 CDS (t-1)
AR1_2_2 0.00056 0.00051 EQ (t-1)
AR1_2_3 -0.03810 0.03465 1month (t-1)
AR2_2_1 0.01041 0.00438 2.38 0.0176 D_CDS (t-1)
AR2_2_2 -0.11503 0.02824 -4.07 0.0001 D_EQ (t-1)
AR2_2_3 -0.55690 0.23245 -2.40 0.0167 D_1month (t-1)
1MONTH Constant3 -0.00434 0.00132 -3.29 0.0010 1
AR1_3_1 0.00006 0.00001 CDS (t-1)
AR1_3_2 0.00032 0.00004 EQ (t-1)
AR1_3_3 -0.02176 0.00301 1month (t-1)
AR2_3_1 0.00110 0.00038 2.88 0.0040 D_CDS (t-1)
AR2_3_2 -0.00021 0.00245 -0.08 0.9334 D_EQ (t-1)
AR2_3_3 0.71244 0.02020 35.26 0.0001 D_1month (t-1)

432
GRANGER CAUSALITY TEST

Table 3.5
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 201.69 < 0.0001
2 1 200.39 < 0.0001
3 1 0.97 0.3255
4 2 196.85 < 0.0001
5 1 19.08 < 0.0001
6 1 184.42 < 0.0001
7 2 5.64 0.0595
8 1 2.52 0.1127
9 1 1.57 0.2107

Test 1: Group 1 Variable: CDS


Group 2 Variable: EQ 1month
Test 2: Group 1 Variable: CDS
Group 2 Variable: EQ
Test 3: Group 1 Variable: CDS
Group 2 Variable: 1month
Test 4: Group 1 Variable: 1month
Group 2 Variable: CDS EQ
Test 5: Group 1 Variable: 1month
Group 2 Variable: CDS
Test 6: Group 1 Variable: 1month
Group 2 Variable: EQ
Test 7: Group 1 Variable: EQ
Group 2 Variable: CDS 1month
Test 8: Group 1 Variable: EQ
Group 2 Variable: CDS
Test 9: Group 1 Variable: EQ
Group 2 Variable: 1month

UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.6
Variable R-Square Std. Deviation F Value Pr > F
CDS 0.3353 2.67798 110.89 < 0.0001
EQ 0.0176 0.49352 3.94 0.0007
1MONTH 0.5317 0.04290 249.55 < 0.0001

433
UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.7
Variable Durbin Watson Normality Chi Pr > Chi Sq F Value Pr > F
Square
CDS 2.07066 9999.99 < 0.0001 152.66 < 0.0001
EQ 2.01491 1565.71 < 0.0001 25.41 < 0.0001
1MONTH 2.06359 3403.63 < 0.0001 99.25 < 0.0001

434
APPENDIX 39 - 2 MONTHS
STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.531914 1.00000 | |********************| 0

1 0.530206 0.99679 | .|********************| 0.027639

2 0.526385 0.98961 | . |********************| 0.047771

3 0.520676 0.97887 | . |********************| 0.061468

4 0.513245 0.96490 | . |******************* | 0.072404

5 0.504260 0.94801 | . |******************* | 0.081638

6 0.493880 0.92850 | . |******************* | 0.089655

7 0.482282 0.90669 | . |****************** | 0.096722

8 0.469598 0.88285 | . |****************** | 0.103011

9 0.455922 0.85714 | . |***************** | 0.108637

10 0.441429 0.82989 | . |***************** | 0.113686

11 0.426273 0.80139 | . |**************** | 0.118224

12 0.410609 0.77195 | . |*************** | 0.122304

13 0.394536 0.74173 | . |*************** | 0.125971

14 0.378188 0.71099 | . |************** | 0.129264

15 0.361692 0.67998 | . |************** | 0.132218

16 0.345210 0.64900 | . |************* | 0.134863

17 0.328814 0.61817 | . |************ | 0.137228

18 0.312583 0.58766 | . |************ | 0.139339

19 0.296683 0.55777 | . |*********** | 0.141220

20 0.281183 0.52863 | . |*********** | 0.142893

435
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

21 0.266167 0.50040 | . |********** | 0.144379

22 0.251703 0.47320 | . |********* | 0.145698

23 0.237847 0.44715 | . |********* | 0.146867

24 0.224671 0.42238 | . |******** | 0.147904

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.1569 0.7201 0.11 0.7171

1 -13.2107 0.0114 -2.47 0.0131

2 -17.6842 0.0033 -2.88 0.0041

Single Mean 0 -2.4425 0.7252 -1.00 0.7560 0.88 0.8457

1 -39.8871 0.0019 -4.44 0.0003 9.89 0.0010

2 -53.8009 0.0019 -5.11 <.0001 13.06 0.0010

Trend 0 -4.4264 0.8601 -1.48 0.8367 1.16 0.9446

1 -60.3445 0.0008 -5.48 <.0001 15.04 0.0010

2 -82.7019 0.0008 -6.32 <.0001 19.96 0.0010

436
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.0024325 1.00000 | |********************| 0

1 0.0020871 0.85800 | .|***************** | 0.027661

2 0.0018729 0.76996 | . |*************** | 0.043493

3 0.0017436 0.71682 | . |************** | 0.052904

4 0.0015745 0.64729 | . |************* | 0.059875

5 0.0014286 0.58729 | . |************ | 0.065009

6 0.0012555 0.51614 | . |********** | 0.068949

7 0.0011081 0.45556 | . |********* | 0.071845

8 0.0010119 0.41600 | . |******** | 0.074022

9 0.00085322 0.35076 | . |******* | 0.075789

10 0.00067936 0.27929 | . |****** | 0.077021

11 0.00056458 0.23210 | . |***** | 0.077792

12 0.00043621 0.17933 | . |**** | 0.078321

13 0.00028333 0.11648 | . |**. | 0.078634

14 0.00014139 0.05813 | . |* . | 0.078766

15 -5.014E-6 -.00206 | . | . | 0.078799

16 -0.0000882 -.03626 | . *| . | 0.078799

17 -0.0002111 -.08680 | .**| . | 0.078812

18 -0.0003610 -.14839 | ***| . | 0.078885

19 -0.0004167 -.17130 | ***| . | 0.079098

20 -0.0004953 -.20363 | ****| . | 0.079381

21 -0.0005766 -.23705 | *****| . | 0.079780

22 -0.0006252 -.25704 | *****| . | 0.080317

23 -0.0006986 -.28719 | ******| . | 0.080944

437
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 -0.0007177 -.29504 | ******| . | 0.081720

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -185.098 0.0001 -9.97 <.0001

1 -143.204 0.0001 -8.45 <.0001

2 -111.338 0.0001 -7.30 <.0001

Single Mean 0 -185.201 0.0001 -9.97 <.0001 49.72 0.0010

1 -143.293 0.0001 -8.45 <.0001 35.72 0.0010

2 -111.410 0.0001 -7.30 <.0001 26.62 0.0010

Trend 0 -185.216 0.0001 -9.97 <.0001 49.68 0.0010

1 -143.306 0.0001 -8.45 <.0001 35.69 0.0010

2 -111.422 0.0001 -7.29 <.0001 26.60 0.0010

438
AUTOCORRELATION

Graph 2.1 – Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0046

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 1301.3596 <.0001

AR(2) 1305.3418 <.0001

AR(3) 1305.5098 <.0001

AR(4) 1305.5713 <.0001

439
Graph 2.2 – 1. difference

Table 2.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.2838

Table 2.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 962.3832 <.0001

AR(2) 967.9526 <.0001

AR(3) 972.7554 <.0001

AR(4) 973.3088 <.0001

440
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0488 72.2666 29.38
1 1 0.0046 6.9768 15.34
2 2 0.0007 0.9060 3.84

VECM-MODEL

Table 3.2 – Long-Run Parameter Beta Estimates When RANK = 1


Variable 1
CDS -0.00880
EQUITY -0.074243
2month 1.82569

Table 3.3 – Adjustment Coefficient Alpha Estimates When RANK = 1


Variable 1
CDS 0.26071
EQUITY -0.00043
2month -0.00434

441
Table 3.4
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS Constant1 0.33034 0.10015 3.30 0.0010 1
AR1_1_1 -0.00229 0.00064 CDS (t-1)
AR1_1_2 -0.01106 0.00311 EQ (t-1)
AR1_1_3 0.47598 0.13371 2month (t-1)
AR2_1_1 0.37331 0.02770 13.48 0.0001 D_CDS (t-1)
AR2_1_2 -1.90363 0.16753 -11.36 0.0001 D_EQ (t-1)
AR2_1_3 15.32882 3.27589 4.68 0.0001 D_2month (t-1)
AR3_1_1 0.04779 0.02590 1.85 0.0652 D_CDS (t-2)
AR3_1_2 -0.93876 0.16272 -5.77 0.0001 D_EQ (t-2)
AR3_1_3 -17.59361 3.25707 -5.40 0.0001 D_2month (t-2)
EQ Constant2 -0.00709 0.01879 -0.38 0.7061 1
AR1_2_1 0.00000 0.0012 CDS (t-1)
AR1_2_2 0.00002 0.00058 EQ (t-1)
AR1_2_3 -0.00079 0.02509 2month (t-1)
AR2_2_1 0.00462 0.00520 0.89 0.3740 D_CDS (t-1)
AR2_2_2 -0.10399 0.03144 -3.31 0.0010 D_EQ (t-1)
AR2_2_3 0.05315 0.61479 0.09 0.9311 D_2month (t-1)
AR3_2_1 -0.00310 0.00486 -0.64 0.5241 D_CDS (t-2)
AR3_2_2 -0.07950 0.03054 -2.60 0.0093 D_EQ (t-2)
AR3_2_3 -0.37619 0.61126 -0.62 0.5384 D_2month (t-2)
2MONTH Constant3 -0.00386 0.00093 -4.13 0.0001 1
AR1_3_1 0.00004 0.00001 CDS (t-1)
AR1_3_2 0.00018 0.00003 EQ (t-1)
AR1_3_3 -0.00792 0.00125 2month (t-1)
AR2_3_1 -0.00041 0.00026 -1.57 0.1164 D_CDS (t-1)
AR2_3_2 0.00016 0.00156 0.10 0.9188 D_EQ (t-1)
AR2_3_3 0.73492 0.03055 24.05 0.0001 D_2month (t-1)
AR3_3_1 0.00041 0.00024 1.71 0.0874 D_CDS (t-2)
AR3_3_2 0.00213 0.00152 1.40 0.1609 D_EQ (t-2)
AR3_3_3 0.15759 0.03038 5.19 0.0001 D_2month (t-2)

442
GRANGER CAUSALITY TEST

Table 3.5
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 199.02 < 0.0001
2 1 197.64 < 0.0001
3 1 2.07 0.1499
4 2 336.15 < 0.0001
5 1 1.80 0.1795
6 1 336.19 < 0.0001
7 2 3.01 0.2225
8 1 2.47 0.1164
9 1 0.23 0.6347

Test 1: Group 1 Variable: CDS


Group 2 Variable: EQ 2month
Test 2: Group 1 Variable: CDS
Group 2 Variable: EQ
Test 3: Group 1 Variable: CDS
Group 2 Variable: 2month
Test 4: Group 1 Variable: 2month
Group 2 Variable: CDS EQ
Test 5: Group 1 Variable: 2month
Group 2 Variable: CDS
Test 6: Group 1 Variable: 2month
Group 2 Variable: EQ
Test 7: Group 1 Variable: EQ
Group 2 Variable: CDS 2month
Test 8: Group 1 Variable: EQ
Group 2 Variable: CDS
Test 9: Group 1 Variable: EQ
Group 2 Variable: 2month

UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.6
Variable R-Square Std. Deviation F Value Pr > F
CDS 0.3602 2.64581 81.01 < 0.0001
EQ 0.0186 0.49654 2.73 0.0037
2MONTH 0.7501 0.02468 431.80 < 0.0001

443
UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.7
Variable Durbin Watson Normality Chi Pr > Chi Sq F Value Pr > F
Square
CDS 1.99729 9999.99 < 0.0001 127.90 < 0.0001
EQ 1.99330 1354.08 < 0.0001 25.94 < 0.0001
2MONTH 2.03388 7517.35 < 0.0001 53.45 < 0.0001

444
APPENDIX 40 - 3 MONTHS
STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.866057 1.00000 | |********************| 0

1 0.864648 0.99837 | .|********************| 0.027886

2 0.861369 0.99459 | . |********************| 0.048247

3 0.856332 0.98877 | . |********************| 0.062179

4 0.849632 0.98103 | . |********************| 0.073394

5 0.841376 0.97150 | . |******************* | 0.082967

6 0.831665 0.96029 | . |******************* | 0.091385

7 0.820606 0.94752 | . |******************* | 0.098921

8 0.808302 0.93331 | . |******************* | 0.105743

9 0.794822 0.91775 | . |****************** | 0.111966

10 0.780271 0.90095 | . |****************** | 0.117670

11 0.764758 0.88303 | . |****************** | 0.122917

12 0.748368 0.86411 | . |***************** | 0.127755

13 0.731195 0.84428 | . |***************** | 0.132222

14 0.713334 0.82366 | . |**************** | 0.136349

15 0.694902 0.80238 | . |**************** | 0.140165

16 0.676015 0.78057 | . |**************** | 0.143692

17 0.656728 0.75830 | . |*************** | 0.146952

18 0.637165 0.73571 | . |*************** | 0.149964

19 0.617416 0.71290 | . |************** | 0.152745

20 0.597548 0.68996 | . |************** | 0.155311

445
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

21 0.577637 0.66697 | . |************* | 0.157676

22 0.557780 0.64405 | . |************* | 0.159855

23 0.538048 0.62126 | . |************ | 0.161860

24 0.518525 0.59872 | . |************ | 0.163704

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.1349 0.7148 0.13 0.7234

1 -12.1163 0.0155 -2.37 0.0174

2 -15.4430 0.0062 -2.68 0.0073

Single Mean 0 -1.6238 0.8222 -0.93 0.7784 0.81 0.8642

1 -35.8057 0.0019 -4.23 0.0007 8.98 0.0010

2 -45.7997 0.0019 -4.74 0.0001 11.23 0.0010

Trend 0 -2.3087 0.9618 -1.06 0.9339 0.57 0.9900

1 -56.4852 0.0008 -5.30 <.0001 14.06 0.0010

2 -74.0417 0.0008 -5.98 <.0001 17.91 0.0010

446
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.0020455 1.00000 | |********************| 0

1 0.0018770 0.91763 | .|****************** | 0.027907

2 0.0017567 0.85881 | . |***************** | 0.045721

3 0.0016762 0.81947 | . |**************** | 0.056914

4 0.0015634 0.76431 | . |*************** | 0.065462

5 0.0014608 0.71416 | . |************** | 0.072077

6 0.0013556 0.66271 | . |************* | 0.077392

7 0.0012430 0.60770 | . |************ | 0.081692

8 0.0011779 0.57584 | . |************ | 0.085140

9 0.0010670 0.52162 | . |********** | 0.088121

10 0.00095714 0.46793 | . |********* | 0.090494

11 0.00088224 0.43131 | . |********* | 0.092359

12 0.00078754 0.38501 | . |******** | 0.093915

13 0.00068004 0.33246 | . |******* | 0.095136

14 0.00056573 0.27658 | . |****** | 0.096037

15 0.00044485 0.21748 | . |**** | 0.096655

16 0.00037237 0.18204 | . |**** | 0.097036

17 0.00025546 0.12489 | . |** . | 0.097301

18 0.00014942 0.07305 | . |* . | 0.097426

19 0.00008416 0.04114 | . |* . | 0.097469

20 0.00001169 0.00572 | . | . | 0.097482

21 -0.0000750 -.03665 | . *| . | 0.097482

22 -0.0001373 -.06712 | . *| . | 0.097493

23 -0.0002121 -.10368 | . **| . | 0.097529

447
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 -0.0002657 -.12988 | .***| . | 0.097615

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -105.604 0.0001 -7.42 <.0001

1 -85.2933 <.0001 -6.53 <.0001

2 -67.4533 <.0001 -5.73 <.0001

Single Mean 0 -105.666 0.0001 -7.42 <.0001 27.51 0.0010

1 -85.3431 0.0019 -6.52 <.0001 21.28 0.0010

2 -67.4916 0.0019 -5.73 <.0001 16.39 0.0010

Trend 0 -105.670 0.0001 -7.41 <.0001 27.49 0.0010

1 -85.3459 0.0008 -6.52 <.0001 21.27 0.0010

2 -67.4930 0.0008 -5.72 <.0001 16.38 0.0010

448
AUTOCORRELATION

Graph 2.1 – Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0024

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 1281.2887 <.0001

AR(2) 1283.3909 <.0001

AR(3) 1283.5555 <.0001

AR(4) 1283.6081 <.0001

449
Graph 2.2 – 1. difference

Table 2.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.1647

Table 2.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 1081.1943 <.0001

AR(2) 1083.4717 <.0001

AR(3) 1086.0351 <.0001

AR(4) 1087.5049 <.0001

450
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0430 64.9481 29.38
1 1 0.0060 8.5714 15.34
2 2 0.0007 0.8921 3.84

VECM-MODEL

Table 3.2 – Long-Run Parameter Beta Estimates When RANK = 1


Variable 1
CDS -0.00971
EQUITY -0.04623
3month 1.53781

Table 3.3 – Adjustment Coefficient Alpha Estimates When RANK = 1


Variable 1
CDS 0.26629
EQUITY 0.01035
3month -0.00289

451
Table 3.4
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS Constant1 0.37985 0.11094 3.42 0.0006 1
AR1_1_1 -0.00258 0.00073 CDS (t-1)
AR1_1_2 -0.01231 0.00345 EQ (t-1)
AR1_1_3 0.40951 0.11491 3month (t-1)
AR2_1_1 0.35995 0.02823 12.75 0.0001 D_CDS (t-1)
AR2_1_2 -1.99046 0.16876 -11.79 0.0001 D_EQ (t-1)
AR2_1_3 18.04658 4.69941 3.84 0.0001 D_3month (t-1)
AR3_1_1 0.03230 0.02647 1.22 0.2225 D_CDS (t-2)
AR3_1_2 -0.85932 0.16370 -5.25 0.0001 D_EQ (t-2)
AR3_1_3 -13.89495 4.68056 -2.97 0.0030 D_3month (t-2)
EQ Constant2 0.00441 0.02069 0.21 0.8311 1
AR1_2_1 -0.00010 0.00014 CDS (t-1)
AR1_2_2 -0.00048 0.00064 EQ (t-1)
AR1_2_3 0.01591 0.02143 3month (t-1)
AR2_2_1 0.00491 0.00526 0.93 0.3510 D_CDS (t-1)
AR2_2_2 -0.11257 0.03147 -3.58 0.0004 D_EQ (t-1)
AR2_2_3 -0.36582 0.87632 -0.42 0.6764 D_3month (t-1)
AR3_2_1 -0.00282 0.00494 -0.57 0.5683 D_CDS (t-2)
AR3_2_2 -0.07956 0.03052 -2.61 0.0093 D_EQ (t-2)
AR3_2_3 -0.10603 0.87280 -0.12 0.9033 D_3month (t-2)
3MONTH Constant3 -0.00307 0.00073 -4.21 0.0001 1
AR1_3_1 0.00003 0.00000 CDS (t-1)
AR1_3_2 0.00013 0.00002 EQ (t-1)
AR1_3_3 -0.00444 0.00076 3month (t-1)
AR2_3_1 -0.00012 0.00019 -0.66 0.5078 D_CDS (t-1)
AR2_3_2 0.00084 0.00111 0.75 0.4516 D_EQ (t-1)
AR2_3_3 0.81322 0.03093 26.29 0.0001 D_3month (t-1)
AR3_3_1 0.00022 0.00017 1.28 0.1998 D_CDS (t-2)
AR3_3_2 0.00151 0.00108 1.40 0.1613 D_EQ (t-2)
AR3_3_3 0.12297 0.03081 3.99 0.0001 D_3month (t-2)

452
GRANGER CAUSALITY TEST

Table 3.5
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 4 247.44 < 0.0001
2 2 225.74 < 0.0001
3 2 11.47 0.0032
4 4 325.92 < 0.0001
5 2 22.23 < 0.0001
6 2 319.83 < 0.0001
7 4 2.98 0.5620
8 2 0.83 0.6610
9 2 2.23 0.3284

Test 1: Group 1 Variable: CDS


Group 2 Variable: EQ 3month
Test 2: Group 1 Variable: CDS
Group 2 Variable: EQ
Test 3: Group 1 Variable: CDS
Group 2 Variable: 3month
Test 4: Group 1 Variable: 3month
Group 2 Variable: CDS EQ
Test 5: Group 1 Variable: 3month
Group 2 Variable: CDS
Test 6: Group 1 Variable: 3month
Group 2 Variable: EQ
Test 7: Group 1 Variable: EQ
Group 2 Variable: CDS 3month
Test 8: Group 1 Variable: EQ
Group 2 Variable: CDS
Test 9: Group 1 Variable: EQ
Group 2 Variable: 3month

UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.6
Variable R-Square Std. Deviation F Value Pr > F
CDS 0.3570 2.67545 78.48 < 0.0001
EQ 0.0200 0.49890 2.88 0.0022
3MONTH 0.8486 0.01761 792.17 < 0.0001

453
UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.7
Variable Durbin Watson Normality Chi Pr > Chi Sq F Value Pr > F
Square
CDS 1.98988 9999.99 < 0.0001 133.68 < 0.0001
EQ 1.99172 1290.59 < 0.0001 24.42 < 0.0001
3MONTH 2.02718 9999.99 < 0.0001 32.08 < 0.0001

454
APPENDIX 41 - 6 MONTHS
STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 1.809238 1.00000 | |********************| 0

1 1.808140 0.99939 | .|********************| 0.028583

2 1.806124 0.99828 | . |********************| 0.049487

3 1.803229 0.99668 | . |********************| 0.063854

4 1.799478 0.99461 | . |********************| 0.075502

5 1.794901 0.99208 | . |********************| 0.085539

6 1.789527 0.98911 | . |********************| 0.094473

7 1.783385 0.98571 | . |********************| 0.102585

8 1.776506 0.98191 | . |********************| 0.110051

9 1.768903 0.97771 | . |********************| 0.116990

10 1.760606 0.97312 | . |******************* | 0.123485

11 1.751643 0.96817 | . |******************* | 0.129599

12 1.742040 0.96286 | . |******************* | 0.135379

13 1.731828 0.95721 | . |******************* | 0.140863

14 1.721031 0.95125 | . |******************* | 0.146081

15 1.709678 0.94497 | . |******************* | 0.151057

16 1.697804 0.93841 | . |******************* | 0.155812

17 1.685424 0.93157 | . |******************* | 0.160363

455
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

18 1.672558 0.92445 | . |****************** | 0.164724

19 1.659233 0.91709 | . |****************** | 0.168910

20 1.645455 0.90947 | . |****************** | 0.172930

21 1.631236 0.90161 | . |****************** | 0.176795

22 1.616594 0.89352 | . |****************** | 0.180512

23 1.601539 0.88520 | . |****************** | 0.184090

24 1.586089 0.87666 | . |****************** | 0.187536

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 0.2270 0.7374 0.47 0.8162

1 -4.4604 0.1469 -1.45 0.1385

2 -5.4299 0.1092 -1.60 0.1029

Single Mean 0 -0.7300 0.9097 -0.91 0.7866 1.21 0.7601

1 -13.4288 0.0586 -2.64 0.0850 3.51 0.1697

2 -16.1257 0.0300 -2.88 0.0489 4.15 0.0770

Trend 0 0.4114 0.9978 0.37 0.9989 1.54 0.8700

1 -23.5391 0.0330 -3.29 0.0681 5.49 0.0917

2 -29.1902 0.0095 -3.65 0.0262 6.73 0.0369

456
Table 1.3 – 1. Difference
Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.00095885 1.00000 | |********************| 0

1 0.00091191 0.95105 | .|******************* | 0.028595

2 0.00087550 0.91307 | . |****************** | 0.047925

3 0.00085488 0.89157 | . |****************** | 0.060499

4 0.00082369 0.85904 | . |***************** | 0.070428

5 0.00079483 0.82894 | . |***************** | 0.078529

6 0.00076420 0.79700 | . |**************** | 0.085385

7 0.00073467 0.76620 | . |*************** | 0.091265

8 0.00071753 0.74833 | . |*************** | 0.096381

9 0.00068889 0.71845 | . |************** | 0.101021

10 0.00066239 0.69082 | . |************** | 0.105115

11 0.00063779 0.66516 | . |************* | 0.108764

12 0.00060749 0.63357 | . |************* | 0.112041

13 0.00057828 0.60310 | . |************ | 0.114933

14 0.00054929 0.57287 | . |*********** | 0.117492

15 0.00052040 0.54274 | . |*********** | 0.119754

16 0.00050676 0.52851 | . |*********** | 0.121749

17 0.00048477 0.50557 | . |********** | 0.123611

18 0.00045809 0.47775 | . |********** | 0.125290

19 0.00044196 0.46093 | . |********* | 0.126771

20 0.00042725 0.44559 | . |********* | 0.128134

21 0.00040718 0.42465 | . |******** | 0.129395

22 0.00039216 0.40899 | . |******** | 0.130529

23 0.00037329 0.38931 | . |******** | 0.131573

457
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

24 0.00035523 0.37047 | . |******* | 0.132512

Table 1.4 – 1. Difference


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -59.5204 <.0001 -5.52 <.0001

1 -49.7353 <.0001 -4.98 <.0001

2 -35.2386 <.0001 -4.16 <.0001

Single Mean 0 -59.5872 0.0019 -5.52 <.0001 15.21 0.0010

1 -49.7905 0.0019 -4.98 <.0001 12.39 0.0010

2 -35.2753 0.0019 -4.16 0.0009 8.64 0.0010

Trend 0 -59.7806 0.0008 -5.53 <.0001 15.27 0.0010

1 -49.9615 0.0008 -4.99 0.0002 12.44 0.0010

2 -35.4115 0.0023 -4.17 0.0052 8.68 0.0010

458
AUTOCORRELATION

Graph 2.1 – Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0005

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 1222.5264 <.0001

AR(2) 1222.7872 <.0001

AR(3) 1222.8792 <.0001

AR(4) 1222.9264 <.0001

459
Graph 2.2 – 1. difference

Table 2.3 – 1. difference


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0977

Table 2.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 1106.4120 <.0001

AR(2) 1107.3362 <.0001

AR(3) 1110.5645 <.0001

AR(4) 1111.4819 <.0001

460
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0285 44.3796 29.38
1 1 0.0063 9.0300 15.34
2 2 0.0011 1.3209 3.84

VECM-MODEL

Table 3.2 – Long-Run Parameter Beta Estimates When RANK = 1


Variable 1
CDS -0.01552
EQUITY -0.05548
6month 1.50251

Table 3.3 – Adjustment Coefficient Alpha Estimates When RANK = 1


Variable 1
CDS 0.30572
EQUITY -0.01209
6month -0.00093

461
Table 3.4
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS Constant1 0.50212 0.13037 3.85 0.0001 1
AR1_1_1 -0.00475 0.00120 CDS (t-1)
AR1_1_2 -0.01696 0.00429 EQ (t-1)
AR1_1_3 0.45935 0.11620 6month (t-1)
AR2_1_1 0.36206 0.02859 12.67 0.0001 D_CDS (t-1)
AR2_1_2 -1.78069 0.17471 -10.19 0.0001 D_EQ (t-1)
AR2_1_3 57.51969 9.19858 6.25 0.0001 D_6month (t-1)
AR3_1_1 0.03667 0.02670 1.37 0.1698 D_CDS (t-2)
AR3_1_2 -0.82033 0.16546 -4.96 0.0001 D_EQ (t-2)
AR3_1_3 -50.23325 9.21508 -5.45 0.0001 D_6month (t-2)
EQ Constant2 -0.02338 0.02450 -0.95 0.3402 1
AR1_2_1 0.00019 0.00023 CDS (t-1)
AR1_2_2 0.00067 0.00081 EQ (t-1)
AR1_2_3 -0.01817 0.02184 6month (t-1)
AR2_2_1 0.00544 0.00537 1.01 0.3118 D_CDS (t-1)
AR2_2_2 -0.10231 0.03283 -3.12 0.0019 D_EQ (t-1)
AR2_2_3 0.93143 1.72867 0.54 0.5901 D_6month (t-1)
AR3_2_1 -0.00252 0.00502 -0.50 0.6156 D_CDS (t-2)
AR3_2_2 -0.07439 0.03110 -2.39 0.0169 D_EQ (t-2)
AR3_2_3 -1.34563 1.73178 -0.78 0.4373 D_6month (t-2)
6MONTH Constant3 -0.00121 0.00046 -2.65 0.0081 1
AR1_3_1 0.00001 0.00000 CDS (t-1)
AR1_3_2 0.00005 0.00001 EQ (t-1)
AR1_3_3 -0.00140 0.00041 6month (t-1)
AR2_3_1 -0.00014 0.00010 -1.38 0.1687 D_CDS (t-1)
AR2_3_2 -0.00043 0.00061 -0.70 0.4812 D_EQ (t-1)
AR2_3_3 0.85175 0.03212 26.52 0.0001 D_6month (t-1)
AR3_3_1 0.00010 0.00009 1.08 0.2824 D_CDS (t-2)
AR3_3_2 0.00138 0.00058 2.38 0.0175 D_EQ (t-2)
AR3_3_3 0.11381 0.03218 3.54 0.0004 D_6month (t-2)

462
GRANGER CAUSALITY TEST

Table 3.5
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 198.14 < 0.0001
2 1 188.65 < 0.0001
3 1 5.21 0.0225
4 2 339.69 < 0.0001
5 1 3.00 0.0830
6 1 339.97 < 0.0001
7 2 3.56 0.1684
8 1 2.44 0.1181
9 1 0.37 0.5415

Test 1: Group 1 Variable: CDS


Group 2 Variable: EQ 6month
Test 2: Group 1 Variable: CDS
Group 2 Variable: EQ
Test 3: Group 1 Variable: CDS
Group 2 Variable: 6month
Test 4: Group 1 Variable: 6month
Group 2 Variable: CDS EQ
Test 5: Group 1 Variable: 6month
Group 2 Variable: CDS
Test 6: Group 1 Variable: 6month
Group 2 Variable: EQ
Test 7: Group 1 Variable: EQ
Group 2 Variable: CDS 6month
Test 8: Group 1 Variable: EQ
Group 2 Variable: CDS
Test 9: Group 1 Variable: EQ
Group 2 Variable: 6month

UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.6
Variable R-Square Std. Deviation F Value Pr > F
CDS 0.3742 2.70228 80.47 < 0.0001
EQ 0.0204 0.50784 2.80 0.0029
6MONTH 0.9073 0.00944 1316.82 < 0.0001

463
UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.7
Variable Durbin Watson Normality Chi Pr > Chi Sq F Value Pr > F
Square
CDS 2.00244 9435.96 < 0.0001 135.85 < 0.0001
EQ 1.99488 1116.15 < 0.0001 25.01 < 0.0001
6MONTH 2.02677 9999.99 < 0.0001 37.76 < 0.0001

464
APPENDIX 41 - 1 YEAR
STATIONARITY

Table 1.1 – Original data


Graphical inspection of stationarity
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 3.003629 1.00000 | |********************| 0

1 2.990176 0.99552 | .|********************| 0.030220

2 2.976239 0.99088 | . |********************| 0.052186

3 2.961828 0.98608 | . |********************| 0.067207

4 2.946960 0.98113 | . |********************| 0.079327

5 2.931660 0.97604 | . |********************| 0.089727

6 2.915940 0.97081 | . |******************* | 0.098949

7 2.899814 0.96544 | . |******************* | 0.107296

8 2.883286 0.95993 | . |******************* | 0.114955

9 2.866357 0.95430 | . |******************* | 0.122057

10 2.849069 0.94854 | . |******************* | 0.128690

11 2.831413 0.94266 | . |******************* | 0.134924

12 2.813426 0.93668 | . |******************* | 0.140810

13 2.795129 0.93058 | . |******************* | 0.146390

14 2.776537 0.92439 | . |****************** | 0.151696

15 2.757634 0.91810 | . |****************** | 0.156756

16 2.738450 0.91171 | . |****************** | 0.161592

17 2.718991 0.90524 | . |****************** | 0.166224

18 2.699250 0.89866 | . |****************** | 0.170666

19 2.679233 0.89200 | . |****************** | 0.174934

465
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

20 2.658953 0.88525 | . |****************** | 0.179040

21 2.638433 0.87842 | . |****************** | 0.182994

22 2.617697 0.87151 | . |***************** | 0.186805

23 2.596785 0.86455 | . |***************** | 0.190482

24 2.575691 0.85753 | . |***************** | 0.194032

Table 1.2 – Original Data


Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 2.9531 0.9986 16.22 0.9999

1 1.2223 0.9428 0.74 0.8738

2 0.7859 0.8737 0.42 0.8027

Single Mean 0 4.2514 0.9999 12.71 0.9999 144.59 0.0010

1 -1.0131 0.8850 -0.33 0.9179 0.68 0.8986

2 -2.4907 0.7193 -0.70 0.8456 0.71 0.8888

Trend 0 2.9332 0.9999 5.84 0.9999 87.76 0.0010

1 -7.1350 0.6528 -1.72 0.7445 2.22 0.7328

2 -9.8734 0.4450 -2.05 0.5718 2.65 0.6452

Table 1.3 – 1. Difference


Graphical inspection of stationarity

466
Autocorrelations

Lag Covariance Correlation -1 9 8 7 6 5 4 3 2 1 0 1 2 3 4 5 6 7 8 9 1 Std Error

0 0.00034929 1.00000 | |********************| 0

1 0.00033634 0.96294 | .|******************* | 0.030234

2 0.00032609 0.93358 | . |******************* | 0.051081

3 0.00031938 0.91439 | . |****************** | 0.064828

4 0.00031036 0.88854 | . |****************** | 0.075704

5 0.00030262 0.86640 | . |***************** | 0.084702

6 0.00029485 0.84414 | . |***************** | 0.092449

7 0.00028803 0.82461 | . |**************** | 0.099244

8 0.00028326 0.81095 | . |**************** | 0.105321

9 0.00027522 0.78793 | . |**************** | 0.110882

10 0.00026814 0.76767 | . |*************** | 0.115887

11 0.00026211 0.75041 | . |*************** | 0.120446

12 0.00025384 0.72674 | . |*************** | 0.124646

13 0.00024573 0.70352 | . |************** | 0.128461

14 0.00023803 0.68148 | . |************** | 0.131936

15 0.00023164 0.66317 | . |************* | 0.135115

16 0.00022731 0.65079 | . |************* | 0.138058

17 0.00022216 0.63604 | . |************* | 0.140834

18 0.00021631 0.61927 | . |************ | 0.143436

19 0.00021234 0.60793 | . |************ | 0.145859

20 0.00020815 0.59594 | . |************ | 0.148157

21 0.00020154 0.57701 | . |************ | 0.150333

22 0.00019739 0.56513 | . |*********** | 0.152344

23 0.00019195 0.54954 | . |*********** | 0.154248

24 0.00018550 0.53108 | . |*********** | 0.156027

467
Table 1.4 – 1. Difference
Results from Dickey Fuller test. If “Pr < Tau” < 0.05 H0: Non-stationary is rejected.
Augmented Dickey-Fuller Unit Root Tests

Type Lags Rho Pr < Rho Tau Pr < Tau F Pr > F

Zero Mean 0 -34.7698 <.0001 -4.16 <.0001

1 -28.6835 <.0001 -3.73 0.0002

2 -21.3073 0.0011 -3.19 0.0015

Single Mean 0 -37.7618 0.0019 -4.30 0.0006 9.26 0.0010

1 -31.1459 0.0019 -3.85 0.0027 7.43 0.0010

2 -23.0134 0.0057 -3.26 0.0174 5.36 0.0256

Trend 0 -41.6029 0.0008 -4.46 0.0018 9.98 0.0010

1 -34.2762 0.0030 -3.97 0.0099 7.95 0.0080

2 -25.1600 0.0232 -3.34 0.0612 5.64 0.0835

468
AUTOCORRELATION

Graph 2.1 – Original data

Table 2.1 – Original data


Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0001

Table 2.2 – Original data


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 1094.3297 <.0001

AR(2) 1094.3461 <.0001

AR(3) 1094.3571 <.0001

AR(4) 1094.3649 <.0001

Graph 2.2 – 1. difference

469
Table 2.3 – 1. difference
Durbin Watson test statistic. Close to zero corresponds to positive correlation, close to 4
corresponds to negative autocorrelation
Durbin-Watson 0.0715

Table 2.4 – 1. difference


Breusch-Godfrey test. If “Pr > LM” < 0.05 we reject H0: No autocorrelation

Godfrey's Serial Correlation


Test

Alternative LM Pr > LM

AR(1) 1017.0594 <.0001

AR(2) 1017.7053 <.0001

AR(3) 1019.2109 <.0001

AR(4) 1019.6531 <.0001

470
MODEL

CO-INTEGRATION

Table 3.1 – Original data


Johansen’s co-integration test. If Trace value > Critical value (5%) we reject H0: No co-
integration.
H0: Rank = r H1: Rank > r Eigenvalue Trace 5% Critical Value
0 0 0.0137 28.1125 29.38
1 1 0.0096 13.0483 15.34
2 2 0.0023 2.5450 3.84

VAR-MODEL

Table 3.2
Equation Parameter Estimate Std. Error t-value Pr > |t| Variable
CDS 1 diff Constant1 -0.0363 0.09350 -0.35 0.7272 1
AR1_1_1 0.42536 0.02568 16.56 0.0001 CDS (t-1)
AR1_1_2 -2.22591 0.16738 -13.30 0.0001 EQ (t-1)
AR1_1_3 -13.84346 4.93399 -2.81 0.0051 1year (t-1)
EQ 1 diff Constant2 0.01259 0.01693 0.74 0.4570 1
AR1_2_1 0.00667 0.00465 1.43 0.1517 CDS (t-1)
AR1_2_2 -0.10801 0.03030 -3.56 0.0004 EQ (t-1)
AR1_2_3 -0.13485 0.89319 -0.15 0.8800 1year (t-1)
1year 1 diff Constant3 0.00009 0.00013 0.71 0.4759 1
AR1_3_1 -0.00001 0.00004 -0.34 0.7302 CDS (t-1)
AR1_3_2 0.00518 0.00023 22.12 0.0001 EQ (t-1)
AR1_3_3 0.97052 0.00690 140.63 0.0001 1year (t-1)

471
GRANGER CAUSALITY TEST

Table 3.3
Granger Causality Wald Test

Test DF Chi-Square Pr > ChiSq


1 2 180.57 < 0.0001
2 1 171.62 < 0.0001
3 1 3.21 0.0731
4 2 497.87 < 0.0001
5 1 5.81 0.0160
6 1 498.15 < 0.0001
7 2 2.35 0.3094
8 1 2.33 0.1273
9 1 0.29 0.5912

Test 1: Group 1 Variable: CDS 1 diff


Group 2 Variable: EQ 1 diff oneyear
Test 2: Group 1 Variable: CDS 1 diff
Group 2 Variable: EQ 1 diff
Test 3: Group 1 Variable: CDS 1 diff
Group 2 Variable: oneyear
Test 4: Group 1 Variable: oneyear
Group 2 Variable: CDS 1 diff EQ 1 diff
Test 5: Group 1 Variable: oneyear
Group 2 Variable: CDS 1 diff
Test 6: Group 1 Variable: oneyear
Group 2 Variable: EQ 1 diff
Test 7: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff oneyear
Test 8: Group 1 Variable: EQ 1 diff
Group 2 Variable: CDS 1 diff
Test 9: Group 1 Variable: EQ 1 diff
Group 2 Variable: oneyear

UNIVARIATE MODEL ANOVA DIAGNOSTICS

Table 3.4
Variable R-Square Std. Deviation F Value Pr > F
CDS 1diff 0.3370 2.94173 184.54 < 0.0001
EQ 1diff 0.0147 0.533253 5.43 0.0010
1YEAR 1diff 0.9517 0.00411 7160.01 < 0.0001

472
UNIVARIATE MODEL WHITE NOISE DIAGNOSTICS

Table 3.5
Variable Durbin Watson Normality Chi Pr > Chi Sq F Value Pr > F
Square
CDS 1diff 2.11745 6545.11 < 0.0001 199.03 < 0.0001
EQ 1diff 2.01135 955.87 < 0.0001 17.20 < 0.0001
1YEAR 1diff 2.09761 3482.87 < 0.0001 6.88 0.0089

473
Appendix 43 – OVERVIEW OF RESULTS
Rank 1st lag 2nd lag 3rd lag R-square
4.1 Index, After - Q4 -6.93038 31.81

3.5 High yield - Q4 -7.446 14.93


3.7 High yield -6.38 27.35
3.4 High yield - Q3 -6.985 15.47
3.4 Index - Q4 -5.35306 -1.82639 28.74
2.7 Financial -4.542 -1.741 20.2
2.8 consumer, cyk -3.36513 -1.27632 33.9
2.6 Index, after - Q3 -2.90776 33.41
2.6 Index - exogenous -2.39199 -0.9392 36.74
2.6 Index, After -2.12302 38.23
2.3 Investment grade - Q4 -2.912 -0.886 25.09
2.4 Index -2.25953 -0.83542 34.17
2.4 Index -Q3 -2.39542 -0.84414 31.79
2.0 High yield - Q2 -2.598 22.01
1.9 High yield - Q1 -2.52 -1.48 20.49
1.9 Investment grade -1.263 -0.497 31.63
1.8 Investment grade - Q3 -1.195 -0.576 29.2
1.4 Industrial -0.713 -0.466 25.26
1.4 Materials -0.856 -0.393 22.91
1.3 Index -Q2 -0.77904 -0.5991 22.68
0.9 communication -2.03795 3.3
1.4 Energy -0.598 -0.41 -0.296 24.99
1.3 Index, Before - Q3 -0.862 -0.492 20.92
1.3 Index, After - Q2 -0.66628 23.61
1.3 Index, After -Q1 -0.49543 -0.30462 25.36
1.1 Index, Before - Q4 -1.308 -0.929 12.18
1.2 Index, Before -0.544 -0.413 22.37
1.2 Index -Q1 -0.40939 -0.26813 24.04
1.2 Investment grade -Q1 -0.341 -0.22 24.48
1.0 consumer, non-cyk -0.62065 -0.34168 17.38
0.8 Technology -1.12926 8.14
0.9 Index, Before - Q2 -0.292 -0.305 16.62
0.7 Utility -0.818 7.94
0.8 Index, Before -Q1 -0.117 -0.185 16.96
0.6 Investment grade - Q2 -0.353 -0.628 9.18

474
Appendix 44 – QUARTILE RANKING
Index

3.2 Index - Q4 -5.35306 -1.82639 28.74


2.0 Index -Q3 -2.39542 -0.84414 31.79
2.0 Index -2.25953 -0.83542 34.17
1.0 Index -Q2 -0.77904 -0.5991 22.68
0.9 Index -Q1 -0.40939 -0.26813 24.04

Before

2.9 Index, Before - Q4 -1.308 -0.929 12.18


2.5 Index, Before - Q3 -0.862 -0.492 20.92
2.0 Index, Before -0.544 -0.413 22.37
1.3 Index, Before - Q2 -0.292 -0.305 16.62
1.0 Index, Before -Q1 -0.117 -0.185 16.96

After

4.1 Index, After - Q4 -6.93038 31.81


2.2 Index, after - Q3 -2.90776 33.41
2.0 Index, After -2.12302 38.23
0.9 Index, After - Q2 -0.66628 23.61
0.9 Index, After -Q1 -0.49543 -0.30462 25.36

High yield

2.0 High yield -6.38 27.35


1.7 High yield - Q4 -7.446 14.93
1.7 High yield - Q3 -6.985 15.47
1.2 High yield - Q2 -2.598 22.01
1.1 High yield - Q1 -2.52 -1.48 20.49

Investment grade

3.1 Investment grade - Q4 -2.912 -0.886 25.09


2.0 Investment grade -1.263 -0.497 31.63
1.9 Investment grade - Q3 -1.195 -0.576 29.2
1.0 Investment grade -Q1 -0.341 -0.22 24.48
0.6 Investment grade - Q2 -0.353 -0.628 9.18

475

You might also like