You are on page 1of 17

Determinants of Oil Price: A Markov Regime Switching Approach

M.Thenmozhi* and N. Srinivasan**

Abstract
This study examines the impact of fundamental, financial and speculative factors on crude oil prices
and analyzes the behavior of various determinants with respect to volatile regimes. Prior research has
focused on the impact of demand, supply and other macroeconomic factors. While financialization
and speculative activity in oil derivative segment has created the need to examine the impact of 12month basis and non-commercial net long positions (futures and options) on crude oil prices. This
study differs from previous literature by examining the impact of speculative factors on oil price at
low and high volatile regimes using Markov regime-switching model. We contribute by controlling
for emerging market oil consumption and trade-weighted US dollar (Broad) Index apart from other
demand and supply factors, we also control for stock index to examine the effect on oil prices. Our
empirical findings indicate that speculation affects the oil price positively in low-volatile state and has
inverse effect in high-volatile state. At low-volatile regimes, fundamental, financial and speculative
factors have significant impact on the oil price, whereas at high volatile regimes, only the factors
pertaining to supply, S&P 500 and trade-weighted US dollar (Broad) Index have a significant effect
on the oil price. Broadly, the results imply that the effect of speculation on the oil price can only be
seen in low-volatile regimes, whereas, in high-volatile regimes, supply and financial factors play a
significant role in explaining the oil price. Our results suggest that regulators and policymakers
should consider supply dynamics while tracking or predicting the movements of the oil price,
particularly, during high-volatile periods.
Keywords: Crude oil, Speculation, Markov Regime Switching Model
*Professor, Department of Management Studies, Indian Institute of Technology Madras, Chennai 600036, India,
mtm@iitm.ac.in
**Research scholar, Department of Management Studies, Indian Institute of Technology Madras, Chennai 600036,
India, sin.iitm@gmail.com

1. Introduction
2

This index consists of the seven (euro, Canadian dollar, Japanese yen, British pound, Swiss franc, Australian dollar and
Swedish krona)

Crude oil price plays a significant role in the world economy and has always been in the forefront of
discussions, regardless of the macroeconomic situation that prevails in both developed and
developing economies. Oil is one of the highly traded commodities in the world with high
participation from financial institutions in the derivative segment and they play an active role in
establishing the price of the crude oil. The economy becomes more difficult to manage when the oil
prices remain highly volatile, as higher volatility in crude oil prices, have greater ramifications for
different players in the economy and managing current account balance for governments becomes a
challenge.
Precise estimation of the factors that affect the crude oil price helps in understanding the dynamics of
crude oil price movements and managing the risk pertaining to volatile oil prices. The determinants of
the crude oil price in the high-volatile period might be different from low-volatile period, and may
differ during different economic phases. We differ from prior research, by investigating the effect of
fundamental, financial and speculative factors on the crude oil prices at high and low volatility
regimes.
Most of the previous studies focused on oil supply and demand factors and provide evidence that
fundamental factors are the key drivers of oil prices (Chevillon and Rifflart, 2009; Peltonen et al.,
2011; Kilian, 2009). With respect to crude oil supply factors, such as OPEC production (Chevillon
and Rifflart, 2009; Lin and Tamvakis, 2010) and OECD inventory (Ye et al., 2002) are found to have
inverse relationship with oil prices. In the demand side, most of the studies have used only OECD
consumption and OECD net imports for determining the demand shock. The demand side variables
were used to extract the residual demand shock and these shocks are found to have positive impact on
oil prices (Kilian and Lee, 2014; Kilian, 2009). Since past studies considered only shocks and studied
their effect on oil price, the main effect of OPEC production, OECD inventory, OECD consumption
and OECD net imports individually has been ignored. However, past studies paid little attention to the
demand from emerging markets such as China and India, which are major oil importing countries.
Moreover, considering OECD consumption alone is not appropriate, since it does not include major
oil consuming countries like China and India.
Hence, there is a need to empirically examine the effect of individual supply related variables such as
OECD inventory and OPEC production, demand related variables such as OECD consumption,
OECD net imports and use proxy variables to account for the increasing consumption/ demand from
emerging economies such as China and India.
Contemporary Research in the field of oil price dynamics indicates that financial factors also have
significant impact on oil price apart from fundamental factors. Many studies have shown significant
impact of exchange rate on crude oil prices (Amano and Van Norden, 1998; Basher et al., 2012;
Beckmann and Czudaj, 2013). However, studies have used NEER (Breitenfellner et al., 2009;
Cifarelli and Paladino, 2010), REER (Oriavwote and Eriemo, 2012; Breitenfellner et al., 2009) and
US dollar exchange rate, which is based on the transaction between US and its trading partners.
2

This index consists of the seven (euro, Canadian dollar, Japanese yen, British pound, Swiss franc, Australian dollar and
Swedish krona)

Examining the impact of US dollar index (TWEXMMTH US dollar index2) (Basher et al., 2012) has
limitations in accounting for the global trading of oil in dollars. However, using a comprehensive
measure such as trade weighted US dollar index (broad) would capture the complete dynamics of oil
prices. Hence, it is pertinent to examine the dollar index (broad).
Prior research have also found significant impact of stock index on crude oil prices (Basher et al.,
2012; Cifarelli and Paladino, 2010; Breitenfellner et al., 2009). If the expected future spot price is
greater than the futures price, there will be a premium to extract oil from well. But, the impact of
spread between the current spot price and a year ahead futures price of oil has not been considered by
prior studies. Considering 12-month basis, which measures the premium of holding/ storing a crude
oil rather than a derivative product may provide more insight on crude oil price discovery.
Sornette et al., (2009) postulated that the oil price shocks during crisis are due to speculative factors
and speculative trading can influence the oil prices without any change in fundamental factors
(Hamilton, 2008). Speculation variable such as feedback trading was found to have negative effect on
oil prices (Cifarelli and Paladino, 2010). While, non-commercial net long positions (futures) has been
found to have positive effect oil price (Fattouh and Mahadeva, 2012). However, with the use of call
option and put option, extensive building up of positions in put/ call can impact oil prices
significantly. Hence, capturing speculation ignoring options will not be meaningful. Non-commercial
net long positions (futures and options), which comprehensively captures the speculative activity,
would capture the impact of speculation on oil price more effectively.
Oil price series exhibit structural breaks and non-linearity (Reboredo, 2010) and may exhibit dynamic
behavior with respect to the financial events such as crisis, changes in government policy, changes in
the business cycles and economic down turns. As the behavior of the oil price determinants changes
over time, understating the oil price and its determinants behavior with respective regimes becomes
crucial. Using ordinary regression models fail to identify the dynamic linkage between oil prices and
its determinants across regimes. Markov Regime-switching model provides a flexible framework to
model structural breaks, dynamic shifts and dynamic relationships. Hence, Markov Regime-switching
model is used to examine the determinants of oil prices at high and low volatile regimes.
Most of the existing literature examines the relationship using traditional linear time series models
such as VAR/ VECM, co-integration and structural VAR. But oil price is found to have structural
breaks and may not be linear and most of the studies examine casual/ long-term relationship of crude
oil with only one or few variables. Linear models could suffer from possible misspecification or
omitted variable bias and the relationship between oil price and determinants may vary over time.
Incorporating or using the model, which accounts for structural break and capturing the effect at
different periods would help in accurate estimation of oil price. Very few studies have used nonlinear
models such as CCC - GARCH-M (Cifarelli and Paladino, 2010) and Bayesian Model averaging
(Breitenfellner et al., 2009). However, these models cannot estimate the relationship with respect to
different market phases and fail to address structural breaks in the data. Markov-regime switching
methodology is largely used for finding non-linear causality (Firouz Fallahi, 2011) and volatility
2

This index consists of the seven (euro, Canadian dollar, Japanese yen, British pound, Swiss franc, Australian dollar and
Swedish krona)

(Naifar and Dohaiman, 2013), but has not been used to estimate the effect of multiple factors on oil
prices at different regimes. Hence, the study uses Markov regime switching model to investigate the
nonlinear relationship between the crude oil and its determinants at different regimes.
Empirical results of our study indicate that speculative factors play an important role only in high
volatile state, whereas in low volatile state, fundamental, and financial factors a major role in
determining the oil price. The study provides a holistic view about price formation in the crude oil
market by incorporating determinants from fundamental, financial and speculative factors. The study
has implications for policymakers, financial institutions and investors, as we indicate the determinants
that affect the crude oil price for each regime separately.
Hence, this study differs from previous literature and extends the literature in following ways:
i.

By examining the impact of speculative activity in derivative segment on WTI crude oil
prices, controlling for fundamental factors and financial factors.

ii.

We contribute to literature by considering individual demand factors such as OECD


consumption, OECD net import and supply factors such as OPEC production, OECD
inventory, and account for increasing demand from emerging economies. Since, the
consumption data for the emerging economies is non-existing, Industrial production of China
(in $) and industrial production of India (in $) is considered as proxy for the demand from
emerging countries.

iii.

We contribute methodologically by using markov regime switching methodology to control


structural break and non-linearity. We also develop a multivariate framework to examine the
impact of fundamental, financial and speculative factors on oil prices.

The remainder of this paper is organized as follows: Section 2 presents Data and sample. Section 3
discusses the methodology and section 4 explains the empirical results. Section 5 summarizes the
findings and brings out the implications of the study.
2. Data and sample
The study focuses on WTI crude oil traded at NYMEX exchange. The fundamental variables
considered in the study are Lagged WTI oil price, OPEC production, OECD Stocks, OECD
consumption, OECD Net Imports, industrial production of China and Industrial production of India.
The speculative variable used in the study is net long positions of non-commercial traders. The
financial variables include 12-month basis, S&P 500 and trade-weighted US dollar Index (Broad).
The data is collected from databases such as Energy Information Agency, World Bank, St. Louis
Federal Reserve (USA) and the US Commodity Futures Trading Commission. Table 1 summarizes
the Variables, measurement and its source. Monthly data has been collected for April 1995 to May
2014 and the final sample comprises 229 observations. The monthly returns are calculated as the
difference in the logarithm of two consecutive prices for all the variables.
2

This index consists of the seven (euro, Canadian dollar, Japanese yen, British pound, Swiss franc, Australian dollar and
Swedish krona)

3. Methodology
We examine the stationarity of the variables using Augmented Dickey Fuller, Phillip Perron and
KPSS tests, and test for non-linearity using BDS test. The Markov regime-switching model is
developed using Matlab R2012a.
BDS Test for Non-linearity
Presence of non-linear dependence in the data is tested using the BDS non-linearity test proposed by
Brock et al. (1987, 1996).The presence of non-linearity is confirmed if the test statistic is greater than
the critical value of the standard normal distribution at the specified confidence levels. The BDS nonlinearity test based on the correlation integral of the time series is as follows:
Wm ( , T ) =
where,

T Cm ( , T ) C1 ( , T ) m

m ( , T )

(1)

represents the BDS test statistic,

stands for the standard deviation of

m is the embedding dimension, and represents the maximum difference between pairs of
observations considered in calculating the correlation integral. The BDS test statistic is
asymptotically normally distributed with zero mean and unit variance [i.e., N (0, 1)]. The null
hypothesis of the BDS procedure is that the data are independently, identically distributed (i.i.d). The
null hypothesis of linearity is rejected if the computed test statistic exceeds the critical value at the
convention level. The rejection of the null hypothesis indicates the presence of nonlinear dependence
in the data.
The Markov-Switching model
Several studies in the literature have used Markov regime-switching models to investigate nonlinearity and asymmetry (Hamilton, 1989; Gray, 1996; Krolzig, 1998; Krolzig, 2000; Artis et al.,
2004). Hamilton (1989) proposed that Regime shifts in mean and variance in a time series variable yt
can be modeled using Markov regime-switching autoregressive model (MS-AR) of pth order as
follows
(2)
where i is the slope coefficient of the autoregressive term (yt 1). and are the mean and standard
deviation based on the state at time t (st) and yt represents the WTI crude oil price. This MS-AR
methodology allows us to model the potential regime shifts in the crude oil price. Krolzig (1998)
developed MS-VAR, which is a generalization of the MS-AR model (Hamilton, 1989) and used to
find the causal relationship between endogenous variables. Several existing literature used MS-VAR
for this purpose (Krolzig, 1997; Warne, 2000; Psaradakis et al., 2005). In these models, the intercept
2

This index consists of the seven (euro, Canadian dollar, Japanese yen, British pound, Swiss franc, Australian dollar and
Swedish krona)

term, autoregressive slope coefficient and the error variance are regime dependent. The Markov
regime switching regression is modeled as follows:
(3)
Let us assume that rt as dependent variable and xt as independent variable. ut is the innovation
process, with variance v(st) based on the state st at time t, follows an irreducible two-state Markov
process defined by the transition probabilities (Pij) between states as follows:
2

Pij p[ st j s(t 1) i ], With

ij

1 for all i, j {1, 2},

j 1

Where,
P11 = p(st = 1|s(t 1) = 1)
P12 = 1 P11 = p(st = 1|s(t 1) = 2)
P21 = 1 P21 = 1 P22 = p(st = 2|s(t 1) = 1)
P22 = p(st = 2|s(t 1) = 2).

The Markov regime switching regression model captures potential regime shifts in the data. But,
ignoring structural breaks in explaining the data may lead to misleading conclusions. Incorporating
the regime-switching feature of the economic process leads to a better forecasting result (Krolzig,
2000). Our model is the generalization of the MS-VAR/ Markov regime switching regression model
proposed by Krolzig (1997). Using this model, we explain the dynamics of the WTI crude oil spot
price using potential determinants from fundamental and speculation factors.
(4)
where kj is the slope coefficient of the independent variables, which is state-dependent (st); kj is
the slope coefficient of the dummy variables; ut is the innovation process with variance v(st) based on
the state (st). The slope coefficient of the independent variables and the variance of the error term is
state dependent (st). However, the intercept and dummy variables are not state dependent.
We model Markov Regime Switching Regression (Model 3) as follows:
The fundamental, financial, and speculative factors of crude oil are analyzed to understand the
relative importance of these factors. In these models, the intercept term, independent variables' slope
coefficient and the error variance are regime dependent. The Markov regime switching regression is
modeled as follows:
2

This index consists of the seven (euro, Canadian dollar, Japanese yen, British pound, Swiss franc, Australian dollar and
Swedish krona)

Where

is two state markov process, WTI crude oil price is used as dependent variable. Net long

pos represents non-commercial traders net long position, S&P represents the S&P 500 index, Dollar
index represents the trade-weighted US dollar index (Broad), basis represents the 12-month basis
between WTI crude oil Futures and Spot, lagged WTI represents the lagged values of the crude oil
price, OPEC prod represents the OPEC Production, OECD cons represents the OECD Consumption,
OECD stk represents the OECD inventory, OECD net imp represents the total Net Imports, Ind prod
India represents the Industrial production of India, and Ind prod China represents the Industrial
production of China.

are the coefficients of the independent

variable, respectively, that have to be estimated. ut is the innovation process with variance v(st) based
on the state (st).
4. Empirical Results
It is evident from the Figure 5.1 (in the Appendix) that OECD Consumption, OECD Net Imports and
Inventory are reducing post the financial crisis. It is also observed that supply determinants such as
OPEC production, which measures the level of current production, are seen increasing post the 2008
financial crisis. The demand factors, i.e., Industrial production in India and China, that measures
demand from emerging economies, are increasing with the crude oil prices.
The financial and speculative determinants, i.e., S&P 500 and Non-commercial Net long positions are
moving along with the WTI crude oil price. Basis has inverse relationship with the oil prices before
financial crisis, but post-crisis it has a positive relationship. Trade-weighted dollar index has an
inverse relationship, i.e., weak dollar tends to increase the oil price.
The summary statistics and unit root tests for the logarithmic return series are presented in the Table
5.1. Excess kurtosis persisted in almost all the variables except OECD consumption and OECD Net
Imports. Jarque-bera test statistic confirms that all the determinants are normally distributed except
OECD consumption and OECD Net Imports.
Unit root analysis is accomplished using Augmented Dickey Fuller, Phillip Perron and KPSS tests to
confirm stationarity of the variables. The results indicated that all the variables are stationary in
atleast two of the tests and most of the variables are stationary in all the three tests. Thus, the results
of these three tests confirm that the variables are free from unit root (i.e.) variables are stationary.
Table 5.1: Descriptive Statistics
Determinants
WTI SPOT OIL PRICE
2

Mean

Std. Dev.

Skewness

Kurtosis

Jarque-Bera

ADF

PP

KPSS

0.007144

0.081903

-0.77216

4.88251

56.57035*

-11.85362*

-11.83093*

0.036135*

This index consists of the seven (euro, Canadian dollar, Japanese yen, British pound, Swiss franc, Australian dollar and
Swedish krona)

OPEC PRODUCTION
OECD INVENTORY

0.001263

0.015505

-0.465211

5.998215

94.03293*

-15.39638*

-15.49786*

0.086145*

0.000406

0.008967

-0.501476

4.574866

33.26335*

-3.552168*

-14.84646*

0.046155*

OECD CONSUMPTION

-0.000182

0.028483

-0.214342

2.989326

1.754559

-2.916049**

-32.53229*

0.285203*

OECD NET IMPORTS


INDUSTRIAL PRODUCTION
INDIA
INDUSTRIAL PRODUCTION
CHINA

-0.000651

0.039688

-0.029506

2.543648

2.020347

-17.74676*

-30.41094*

0.5*

0.009582

0.056493

0.702814

4.493807

40.1442*

-2.301233

-28.4986*

0.235417*

0.004629

0.054431

-0.408255

3.75903

11.85852*

-3.013915**

-44.14877*

0.143771*

S&P 500

0.005786

0.04503

-0.877515

4.530197

51.73146*

-13.6528*

-13.73022*

0.184718*

DOLLAR INDEX

0.000435

0.012268

0.386009

4.800334

36.61341*

-10.36025*

-10.45809*

0.340712*

12 M BASIS

0.010592

0.698915

0.04897

7.448997

188.9553*

-16.40647*

-20.58826*

0.116019*

NC NET LONG POSITION

0.009224

0.560974

0.869

10.46097

559.969*

-18.22633*

-18.4153*

0.113867*

Notes: *** - 10% significance, ** - 5% significance, * - 1% significance

The non-linearity in the returns series is tested using BDS test (see Table 5.2) and the test statistics are
significantly greater than the critical values, rejecting the null hypothesis of I.D.D. The residuals of
linear OLS model are also tested for non-linearity and found that the residual series is non-linearly
dependent on the oil prices.
Table 5.2: BDS Test Results
Dimension

BDS Statistic

Prob.

0.014999***

0.0030

0.021711***

0.0069

0.025868***

0.0069

0.025151***

0.0118

6
0.022178***
0.0214
Note: This table reports the results of BDS test.
*Denotes the rejections of the null hypothesis at 10% significance level;
**Denotes the rejections of the null hypothesis at 5% significance level;
***Denotes the rejections of the null hypothesis at 1% significance level.

The study considers two states in the Markov Regime-Switching Model. These two states represent
the high- and low-volatility regimes in the model. Chen and Chen (2007) used two-state Markov
Regime-Switching Model to differentiate bull and bear markets. Li (2007) used two-state Markov
Regime-Switching Model to distinguish low and high uncertainty in stock markets. In our model, the
regimes are classified as high and low-volatile regimes. The regimes are classified based on their
model variance. Higher the model variances, the state is termed as high-volatility state and lower the
model variance, the state is termed to be low-volatility state.
Table 5.3: Effect of speculative, fundamental and financial variables on oil price
MODEL 1
Determinants

State 1

State 2

MODEL 2
State 1

State 2

MODEL 3
State 1

State 2

Lagged WTI spot oil price

0.2050*

-0.1286*

OPEC Production

0.6273**

0.6723*

This index consists of the seven (euro, Canadian dollar, Japanese yen, British pound, Swiss franc, Australian dollar and
Swedish krona)

OECD Stocks

- 2.8084*

8.6105*

OECD Consumption

-0.1244

1.9411*

OECD Net Imports

-0.1984

-1.0853*

Industrial Production INDIA

0.2164*

-1.3195*

Industrial Production CHINA

-0.1084

-0.2137*

S&P 500
Trade weighted US Dollar Index
(Broad)

1.353912*

-0.168533

0.4047*

-0.5342*

-2.36167*

0.609047

-0.9107**

2.3189*

12 Month Basis

0.012282

-0.002821

-0.0076

-0.0193*

Non Commercial Net Long Positions

0.007961

-0.279252***

-0.041312*

0.030444*

-0.0017

0.0118*

Intercept

0.01242*

-0.246028*

-0.017704

0.014666*

-0.0026

-0.0186*

Model variance

0.005303*

0.005438

0.002333*

0.004916*

0.004089**

0.000005*

175.47

5.2

1.28

5.02

16.68

1.61

0.9954

0.1235

0.94

0.62

0.0046

0.8765

0.06

0.38

Expected duration of Regime


Transition Probabilities Matrix
(std. error, p-value)

Notes: *** - 10% significance; ** - 5% significance; * - 1% significance.


WTI oil price is valued / priced in US $. OPEC production, OECD net imports and OECD consumption are measured as thousand
barrels per day, whereas OECD inventory is measured as million barrels per month. Industrial production of China and India are
measured as value in US $. Trade weighted US Dollar Index (Broad): A weighted average of the foreign exchange value of the U.S.
dollar against the currencies of a broad group of major U.S. trading partners. Broad currency index includes the Euro Area, Canada,
Japan, Mexico, China, United Kingdom, Taiwan, Korea, Singapore, Hong Kong, Malaysia, Brazil, Switzerland, Thailand, Philippines,
Australia, Indonesia, India, Israel, Saudi Arabia, Russia, Sweden, Argentina, Venezuela, Chile and Colombia.
12 Month Basis: Difference between spot price and 12 month ahead futures price (i.e.) S t Ft+12. S&P 500: A free float market
capitalization-weighted index, also called a market-value-weighted index is a index whose components are weighted according to the
total market value of their outstanding shares. Net Long Positions (Non-Commercial) - Oil (Futures & Options): Difference between
total long positions minus short positions in both futures and options. Non-Commercial Traders: Traders who do not involve directly
in the production, distribution or management of the underlying commodities.

a.) Effect of speculation on the oil price: Measuring the marginal effect of speculation
The impact of speculative factors on the WTI oil price is estimated using Markov regime-switching
regression model, where state 1 is low-volatile regime and state 2 is high-volatile regime and the
results are presented in Table 5.3. It is observed from the results that in low-volatile regime the impact
of non-commercial traders net long position on the WTI oil price is insignificant, whereas in highvolatile regime it is significant at 10%. This indicates that speculative factor has a limited effect on
the oil price. The difference in the model variance between the two states is minimal which indicates
that the variations in crude oil price explained by speculative factors in both the states is nearly the
same and have minimal significance.
The duration of the regime 1 is 175.47 and regime 2 is 5.2. This shows that the volatility of crude oil
conditioned on speculative factor remained in the low-volatile regime for most of the times. The level
of speculation captured by non-commercial traders net long position is consistent irrespective of the
regime.

This index consists of the seven (euro, Canadian dollar, Japanese yen, British pound, Swiss franc, Australian dollar and
Swedish krona)

It is observed from Figure 5.2 that the transition between the states have occurred only once, i.e., in
the middle of 2007 and returned to its initial state by 2008. The results indicate that speculation is not
the only factor that contributes to the changes in the WTI oil price.
Fig 5.2: Conditional Means, Volatilities, Smoothed Transition Probalities for Model 1
0.4
Explained Variable #1

0.2
0
-0.2
-0.4

50

100

150

200

250

0.0745
Conditional Std of Equation #1

0.074
0.0735

S m o o t h e d S t a t e s P ro b a b ilit ie s

0.073

0.0725

50

100

150

200

250

1.5
State 1
State 2

1
0.5
0

50

100

Time

150

200

250

b.) Effect of speculation on the oil price controlling for financial factors
The impact of speculative factor on WTI oil prices controlling for financial factors Model 2 of Table
5.3 indicate that in both low-volatile regime and high-volatile regime the impact of non-commercial
traders net long position on the WTI oil price is significant at 1%. It implies that interaction between
the financial factors and the speculative factor explains the WTI oil price formation.
The impact of dollar index on the WTI oil price in high-volatile regime is insignificant, whereas in
low-volatile regime it is significant at 1%. This indicates that dollar index plays a significant role in
the times of normal market phases. However, during bullish and bearish market periods dollar index
fails to explain the change in the oil price. The negative sign in the low-volatile regime indicates that
the US dollar appreciation have negative effect () on oil price, whereas depreciation of the US dollar
have positive effect (+) on the oil price.
Basis has insignificant impact on the WTI oil price in both high-volatile regime and low-volatile
regime. This indicates that basis have no significant role in the determination of oil price during all
market phases (Normal, Bearish, and Bullish).
The impact of S&P 500 on WTI oil price is insignificant in low-volatile regime, whereas in highvolatile regime it is significant at 1%. This indicates that stock market index explains the crude oil
price only in times of normal market phases. However, during bullish and bearish market periods,
2

This index consists of the seven (euro, Canadian dollar, Japanese yen, British pound, Swiss franc, Australian dollar and
Swedish krona)

stock market index fails to explain the dynamics in the oil price. It is expected to have an inverse
relationship with the oil price, but a positive relationship is observed between the stock market index
and oil price. Rise in the price of oil might be due to the increased consumption of the crude oil
during economic expansion.
The duration of the regime 1 is 1.28 and regime 2 is 5.02. This shows that volatility of the crude oil
conditioned on combined speculative and financial factor is high, that is, the oil price is more volatile.
The level of speculation is significant irrespective of the regime when controlled for financial factors.
The difference in the model variance between the two states indicates that the variation in crude oil
price is explained more by speculative and financial factors in high-volatile states than in the lowvolatile state.
Fig 5.3: Conditional Means, Volatilities, Smoothed Transition Probalities for Model 2
0.4
Explained Variable #1

0.2
0
-0.2
-0.4

50

100

150

200

250

0.08
Conditional Std of Equation #1

0.07

S m o o t h e d S t a t e s P ro b a b ilit ie s

0.06
0.05
0.04

50

100

150

200

250

1
State 1
State 2
0.5

50

100

Time

150

200

250

Figure 5.3 portrays that there is a frequent switchover between the two states. The results indicate that
speculation combined with financial factors explains the changes in the WTI oil price.
c.) Relative importance of determinants of the oil price
The impact of speculative factors on the WTI oil price controlling for both fundamental and financial
factors analyzed using Markov Regime-Switching Model (Model 3 of Table 5.3) shows that the
impact of non-commercial traders net long position on the WTI oil price is insignificant in lowvolatile regime, whereas it is significant at 1% in high-volatile regime. It implies that speculative
factor explains the movements in WTI oil price only in high-volatile regime when controlled for both
financial and fundamental factors.
The dollar index is significant at 5% in low-volatile regime, whereas in high-volatile regime it is
significant at 1%. This indicates that the dollar index has significant impact on the WTI oil price
2

This index consists of the seven (euro, Canadian dollar, Japanese yen, British pound, Swiss franc, Australian dollar and
Swedish krona)

regardless of state. This implies that dollar index plays a significant role in all market phases. In lowvolatile regime, it has significant negative impact, whereas in high-volatile regime it has significant
positive impact.
Basis has insignificant impact on the WTI oil price in low-volatile regime, whereas it is significant at
1% in high-volatile regime. This indicates that basis have significant role in determining the oil price
in bearish and bullish market phases and its effect is insignificant during normal market phase.
S&P 500 has significant impact on the WTI oil price irrespective of its state and is significant at 1%
in both regimes. This indicates that stock market index explains changes in crude oil price in all
market phases. A significant positive relationship is observed in low-volatile period, whereas a
significant negative relationship observed in high-volatile regime.
The empirical results indicate that the Lagged WTI oil price plays a vital role in estimating the price
of crude oil. The lagged WTI oil price is significant at 1% in both regimes. There is a significant
positive relation between the lagged oil price and oil price in low-volatile regime whereas in highvolatile regime it is negative relation.
The OPEC production has significant positive relationship in both high- and low-volatile regimes. It
implies that the OPEC production is lower than the equilibrium level. Our result suggests that OPEC
production is one of the significant factor in determining the oil price in all market phases. The
impact OECD stock on the oil price in both high- and low-volatile regimes is significant at 1%. The
relationship in low-volatile period is negative, whereas in high-volatile period it is positive.
The OECD consumption has insignificant impact on the WTI oil price in low-volatile regime,
whereas in high-volatile regime it is significant at 1%. This indicates that OECD consumption has
significant role in the oil price determination only during bearish and bullish market phases and its
impact is insignificant during normal market phase.
The impact of the OECD net imports on WTI oil price in low-volatile regime is insignificant, whereas
in high-volatile regime it is significant at 1%. The relationship between the crude oil and OECD
import is negative. OECD consumption plays a significant role in determination of oil price only
during bearish and bullish market phases.
The relationship between the crude oil and Industrial production of India is negative and is significant
at 1% in high-volatile regime. In low-volatile regime, it is positive and is significant at 1%. This
implies that Industrial production of India plays a key role in estimating the WTI crude oil price.
The impact of Industrial production of China is insignificant in low-volatile regime, whereas it is
negative and is significant at 1% in high-volatile regime. This implies that Industrial production of
China plays a significant role in determination of the oil price only during bearish and bullish market
phases.
2

This index consists of the seven (euro, Canadian dollar, Japanese yen, British pound, Swiss franc, Australian dollar and
Swedish krona)

The duration of the state 1 is 16.68 and state 2 is 1.61. This indicates that the volatility of the crude oil
explained by combination speculative, fundamental, and financial factors. The level of speculation is
significant in high-volatile regime indicating that speculative factor has significance in determining
the oil price only in high-volatile phase when combined with other factors. The difference in the
model variance between the two states indicate that the variation in crude oil price is better explained
by speculative, fundamental, and financial factors in low-volatile state compared to the high-volatile
state.
From Model 3 of Table 5.3, The predominant factors influencing oil price are OPEC production,
followed by high OECD inventory. Moreover, high dollar index reduces oil prices in low volatile
state. But, high OPEC production, high S&P 500 index, high Industrial production of China and India
leads to increase in oil price in low volatility state.
But, during high volatile regime, high OECD inventories and dollar index decreases oil price. High
OECD consumption and greater non-commercial net long positions reduces oil price. Whereas high
OPEC production and high S&P 500 leads to increase in oil prices. But, increase in OECD net
imports, Industrial production in India and China, S&P 500 and 12M basis reduces oil prices in high
volatility state.
Thus, the nature of impact of fundamental, financial and speculation factors on WTI crude oil price
varies with volatility states.
Fig 5.4: Conditional Means, Volatilities, Smoothed Transition Probalities for Model 3
0.5
Explained Variable #1

S m o o t h e d S t a t e s P ro b a b ilit ie s

0
-0.5

50

100

150

200

250

0.1
Conditional Std of Equation #1

0.05
0

50

100

150

200

250

2
State 1
State 2

1
0

50

100

Time

150

200

250

5. Conclusion
This paper has analyzed the impact of fundamental, financial and speculative factors on the WTI
crude oil prices using regime-switching methodology. The empirical results suggest that the
2

This index consists of the seven (euro, Canadian dollar, Japanese yen, British pound, Swiss franc, Australian dollar and
Swedish krona)

speculative variable when controlled for fundamental and financial factors have a significant role in
explaining the WTI oil price in high volatility state. At high-volatility regimes, we observe that
factors pertaining to speculation combined with financial and fundamental factors have a significant
effect on the oil price, while at low-volatility regimes, it is the fundamental and financial variables
that have a significant role in WTI crude oil price discovery.
OPEC Inventory and trade-weighted US dollar Index (Broad) are dominant at influencing WTI oil
price, but the impact varies depending on the volatility regime. The nature of impact of other
determinants tends to change with volatility regimes. During high volatile state, high value of noncommercial net long positions increases the crude oil price significantly, while it has no significant
impact during low volatile regime.
The results of the study indicate the speculation has significant impact on the oil price only in highvolatility regime, whereas, during low-volatile regimes, supply and financial factors plays a dominant
role in explaining oil price. Thus, the price discovery of WTI crude oil price varies with high and low
volatile regimes and the results are of greater significance for the policymakers, regulators and
investors. It is important for the regulators and policymakers to look into the varying supply
dynamics, financial factors as well as speculative activity based on the volatility sate, to track or
predict the movements of the oil price.
6. References
Amano, R. ., & van Norden, S. (1998). Oil prices and the rise and fall of the US real exchange
rate. Journal of International Money and Finance, 17(2), 299316. http://doi.org/10.1016/S02615606(98)00004-7.
Artis, M., Krolzig, H. M., & Toro, J. (2004). The European business cycle. Oxford Economic
Papers, 56(1), 1-44.
Basher, S. A., Haug, A. A., & Sadorsky, P. (2012). Oil prices, exchange rates and emerging stock
markets. Energy Economics, 34(1), 227240. http://doi.org/10.1016/j.eneco.2011.10.005.
Beckmann, J., & Czudaj, R. (2013). Oil prices and effective dollar exchange rates. International
Review of Economics and Finance, 27, 621636. http://doi.org/10.1016/j.iref.2012.12.002.
Breitenfellner, A., Cuaresma, J., & Keppel, C. (2009). Determinants of Crude Oil Prices: Supply,
Demand, Cartel or Speculation? Monet Policy Econ Q, 4, 111136. Retrieved from
http://www.nationalbank.at/en/img/mop_2009_q4_analyses_06_tcm16-181766.pdf
Chevillon, G., & Rifflart, C. (2009). Physical market determinants of the price of crude oil and the
market premium. Energy Economics, 31(4), 537549. http://doi.org/10.1016/j.eneco.2009.01.002.
Cifarelli, G., & Paladino, G. (2010). Oil price dynamics and speculation. A multivariate financial
approach. Energy Economics, 32(2), 363372. http://doi.org/10.1016/j.eneco.2009.08.014.
2

This index consists of the seven (euro, Canadian dollar, Japanese yen, British pound, Swiss franc, Australian dollar and
Swedish krona)

Fattouh, B., Kilian, L., & Mahadeva, L. (2012). The Role of Speculation in Oil Markets: What
Have We Learned So Far?, hrsg. vom Oxford Institute for Energy Studies. Working Paper 45. Am
1. August im Internet unter: http://www. oxfordenergy.
org/wpcms/wpcontent/uploads/2012/03/WPM-45. pdf.
Fallahi, F. (2011). Causal relationship between energy consumption (EC) and GDP: A Markovswitching (MS) causality. Energy, 36(7), 41654170. http://doi.org/10.1016/j.energy.2011.04.027.
Gray, S. F. (1996). Modeling the conditional distribution of interest rates as a regime-switching
process. Journal of Financial Economics, 42(1), 27-62.
Hamilton, J. D. (2008). Understanding Crude Oil Prices. Energy Journal, 30(2), 179206.
http://doi.org/10.1073/pnas.0703993104.
Hamilton, J. D. (1989). A new approach to the economic analysis of nonstationary time series and
the business cycle. Econometrica: Journal of the Econometric Society, 357-384.
Kilian, L., & Lee, T. K. (2014). Quantifying the speculative component in the real price of oil:
The role of global oil inventories. Journal of International Money and Finance, 42, 71-87.
Kilian, L. (2008). Exogenous oil supply shocks: how big are they and how much do they matter
for the US economy?. The Review of Economics and Statistics, 90(2), 216-240.
Krolzig, H. M. (2000). Predicting Markov-switching vector autoregressive processes. Nuffield
College.
Krolzig, H. M. (1998). Econometric modelling of Markov-switching vector autoregressions using
MSVAR for Ox.
Lin, S. X., & Tamvakis, M. (2010). OPEC announcements and their effects on crude oil prices.
Energy Policy, 38(2), 10101016. http://doi.org/10.1016/j.enpol.2009.10.053.
Naifar, N., & Al Dohaiman, M. S. (2013). Nonlinear analysis among crude oil prices, stock
markets' return and macroeconomic variables. International Review of Economics and Finance,
27, 416-431.
Oriavwote, V. E., & Eriemo, N. O. (2012). Oil prices and the real exchange rate in Nigeria.
International Journal of Economics and Finance, 4(6), 198.
Peltonen, T. a., Sousa, R. M., & Vansteenkiste, I. S. (2011). Fundamentals, Financial Factors, and
the Dynamics of Investment in Emerging Markets. Emerging Markets Finance and Trade, 47(6),
88105. http://doi.org/10.2753/REE1540-496X4703S205.
Psaradakis, Z., Ravn, M. O., & Sola, M. (2005). Markov switching causality and the money
output relationship. Journal of Applied Econometrics, 20(5), 665-683.

This index consists of the seven (euro, Canadian dollar, Japanese yen, British pound, Swiss franc, Australian dollar and
Swedish krona)

Reboredo, J. C. (2010). Nonlinear effects of oil shocks on stock returns: a Markov-switching


approach. Applied Economics. http://doi.org/10.1080/00036840802314606.
Sornette, D., Woodard, R., & Zhou, W. X. (2009). The 20062008 oil bubble: Evidence of
speculation, and prediction. Physica A: Statistical Mechanics and its Applications, 388(8), 15711576.
Warne, A. (2000). Causality and regime inference in a Markov switching VAR (No. 118). Sveriges
Riksbank Working Paper Series.
Ye, M., Zyren, J., & Shore, J. (2002). Forecasting crude oil spot price using OECD petroleum
inventory levels. International Advances in Economic Research, 8(March 2001), 324333.
http://doi.org/10.1007/BF02295507.
APPENDIX
Table 1: Data Description
DETERMINANTS
Oil Benchmark
Lagged WTI - SPOT (NYMEX)
Fundamental - Demand Variables
Consumption (OECD)
Net Imports (OECD)
CHINA Industrial Production (US$)
INDIA Industrial Production (US$)
Fundamental - Supply Variables
OPEC Production
OECD Inventory/ Stock
Financialization - Variables
Trade weighted US Dollar Index (Broad)

S&P 500
12 Month Basis - Oil (Futures)
(NYMEX)
Speculation - Variables
Net Long Positions (Non-Commercial)
Oil (Futures & Options) (NYMEX)

MEASURE

SOURCE

In US $

Energy Information Administration

000 barrels per day


000 barrels per day
In US $
In US $

Energy Information Administration


Energy Information Administration
World bank
World bank

000 barrels per day


Million barrels/ month (at the end
of counting period)

Energy Information Administration


Energy Information Administration

Weighted average of foreign


exchange value of major U.S.
trading partners.
Capitalization-weighted index of
500 stocks Index value
St Ft+12

St. Louis Federal Reserve (U.S.A.)

Open long positions minus short


positions ( In futures & options)

Commodity Futures Trading


Commission

St. Louis Federal Reserve (U.S.A.)


Energy Information Administration

Fig 5.1: Plot of Determinants of WTI oil price

This index consists of the seven (euro, Canadian dollar, Japanese yen, British pound, Swiss franc, Australian dollar and
Swedish krona)

WTI CRUDE OIL SPOT PRICE


140
120
100

2.0

1.5

1.0

0.5

80

60

-1

40

-2

-1

20

-3

-2

-4

-3

0
96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14

96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14

0.0

-1

-0.5

-2

-1

-1.0

-1

-3

-2

-1.5

-2

-4

-3

-2.0

WTI CRUDE OIL SPOT PRICE


TOTAL OIL RIG COUNT
4

WTI CRUDE OIL SPOT PRICE


OPEC PRODUCTION

4 2.8

2.4

2.0

1.6

1 1.2

0.8

-1

-1

0.4

-2

-1

-2

-1 0.0

-1

-3

-2

-3

-2

-0.4

-2

-4

-3

-4

-3 -1.2

-0.8
96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14

WTI CRUDE OIL SPOT PRICE


OECD IMPORTS

-1

-2

-2
96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14

WTI CRUDE OIL SPOT PRICE


TRADE WEIGHTED DOLLAR INDEX

-1

-2

-1

-3

-2
96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14
WTI CRUDE OIL SPOT PRICE
S & P 500

2.0

1.5

1.0

0.5

0.0

-0.5

-1

-1.0

-2

-1.5

-3
96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14

WTI CRUDE OIL SPOT PRICE


INDUSTRIAL PRODUCTION IN INDIA

-1

-4
96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14

-3

WTI CRUDE OIL SPOT PRICES


OECD INVENTORY

96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14

WTI CRUDE OIL SPOT PRICE


OECD CONSUMPTION

96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14

-3

96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14

WTI CRUDE OIL SPOT PRICE


INDUSTRIAL PRODUCTION IN CHINA

-1

-2

-3

-1

-1

-1

-4

-2

-2

-2

-3

-3

-5
96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14
WTI

BASIS

-3
96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14
WTI CRUDE OIL SPOT PRICE
NON COMMERCIAL NET LONG POSITION

This index consists of the seven (euro, Canadian dollar, Japanese yen, British pound, Swiss franc, Australian dollar and
Swedish krona)