Professional Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: While the relationship between stock market returns and macroeconomic variables have been amply
Received 11 July 2019 examined, there is a gap in the literature when it comes to the relationship between different sector
Received in revised form 3 October 2019 indices and various macroeconomic variables. This study examines how certain macroeconomic vari-
Accepted 23 October 2019
ables influence different sectors of the stock market differently in the US and Canada. Using monthly
Available online 12 November 2019
data over the 2000–2018 period, a cointegration analysis is applied to model the relationship between
industrial production, money supply, long-term interest rate, and different sector indices. Sectors that
JEL classification:
have been examined in this study include energy, financials, real estate, industrial, healthcare, consumer
C32
E22
discretionary, and consumer staples. Results suggest that there is a stable long-term relationship between
E44 the macroeconomic variables used in the study and different sector indices for the US but not for Canada.
G12 However, US money supply and interest rate can explain the Canadian stock market. The results suggest
important insights for private investors, pension funds, and governments as long-term investors often
Keywords: base their decision to invest in equities on the stated macroeconomic variables.
Cointegration
© 2019 Board of Trustees of the University of Illinois. Published by Elsevier Inc. All rights reserved.
Macroeconomic variables
Stock market returns
Sector indices
https://doi.org/10.1016/j.qref.2019.10.005
1062-9769/© 2019 Board of Trustees of the University of Illinois. Published by Elsevier Inc. All rights reserved.
E.M. Bhuiyan, M. Chowdhury / The Quarterly Review of Economics and Finance 77 (2020) 62–74 63
coefficients using the multivariate regression framework. A host influence its composite as well as sector-specific indices. The sen-
of works using the APT framework includes Chen, Roll and Ross sitivity of sectoral indices varies, however. For Canada, there is no
(1986), Poon and Taylor (1991); Hamao (1988), Martinez and Rubio association between stock indices, either composite or sectoral,
(1989), Schwert (1990), Ferson and Harvey (1991). These papers with certain macroeconomic variables considered in this study.
find a significant relationship between stock market returns and Canada is a small open economy and is heavily dependent on
money supply, interest rate, and real economic activity questioning international trade and commodity prices in contrast to the US.
the validity of the EMH. Moreover, the monetary policy is straightforward, and the only pol-
The advent of cointegration by Engle and Granger (1987) icy instrument the Bank of Canada uses is the target it sets for the
allows for an alternative approach to study long-term equilib- overnight interest rate to achieve its inflation target whereas the
rium relations between variables without having to worry about monetary policy is more involved and intricate for the US. Further-
spurious correlations. Regressions in the APT framework with non- more, the US macroeconomic variables have a strong influence in
stationary variables were susceptible to such correlations. Since its the Canadian stock indices while the opposite is not true.
development, a host of literature (e.g. Mukherjee & Naka, 1995; The remainder of this study is organized as follows. In Section
Cheung & Ng, 1998; Nasseh & Strauss, 2000; Ratanapakorn & 2, we review the existing literature that contextualizes our paper,
Sharma, 2007; Humpe & Macmillan, 2009) examine and find sig- first from the theoretical perspective and then from the empir-
nificance of macroeconomic variables in explaining stock market ical perspective. In Section 3, we describe the data, model, and
returns further questioning the validity of the EMH. The growing methodology used in this paper. Section 4 shares the results and
literature is mostly limited to the US and a handful of industrial- elaborates the relevant discussion. Further, it provides important
ized countries but not Canada. Despite a strong trading relationship insights about the differences in outcome. Section 5 concludes the
between the US and Canada, the responsiveness of Canadian stock paper.
indices due to the changes in macroeconomic variables could be
different than the US due to the independence of monetary policy
across the US and Canada. To the best of our knowledge, none of the 2. Review: theory and empirics
existing studies have examined the association between the US and
Canadian stock indices with their respective macroeconomic vari- The pioneering work by Fama (1970) suggests that based on
ables. They also ignore how macroeconomic variables in the US can the efficient market hypothesis (EMH) stock prices reflect all pub-
influence Canadian stock indices and vice versa. Additionally, the licly available information and, as such, lags of financial variables
majority of the studies, if not all, use composite stock market indices will not be able to predict future stock prices. However, Abdullah
which ignores the fact that the responsiveness of stock indices due and Hayworth (1993) argue that most previous studies that sup-
to the changes in macroeconomic variables could vary across sec- port the EMH are based on daily and weekly models that do not
tors. For example, stocks in financials could be more responsive due incorporate variables that the EMH theory predicts. Proponents of
to changes in macroeconomic variables as opposed to the stocks the EMH claim that only unanticipated changes in macroeconomic
in health care and vice-versa. In this paper, we focus on answer- variables are able to influence the stock market (e.g. Sorensen,
ing three questions using unit root, cointegration, and vector error 1982; Davidson & Froyen, 1982; Pearce & Roley, 1983). Samuelson
correction methods: (1) Is there any long-run relationship between (1998) argues that while the stock market may be “micro efficient”,
the macroeconomic variables and respective composite as well as it is “macro inefficient” which means that the EMH may be appli-
sector-specific stock market indices in the US and Canada? (2) What cable for individual stocks but not the aggregate stock market. This
is the direction of relationship between the macroeconomic vari- concept of the stock market being “micro efficient” but “macro
ables and stock indices? (3) Is there any relationship between the inefficient” became known as Samuelson’s dictum which is backed
US macroeconomic variables and Canadian stock indices and vice up by some empirical evidence such as Shiller (1981), Campbell
versa? and Shiller (1988), Vuolteenaho (2002), Cohen, Christopher, and
We contribute to the existing literature in four ways. First, while Vuolteenaho, (2003)) using vector autoregressive models.
earlier studies limited to the US and some other industrialized As mentioned above, the existing literature largely uses two key
countries but not Canada, we extend it to comparable Canadian approaches to examine the link between macroeconomic variables
indices to investigate whether the domestic macroeconomic vari- and stock market indices. While the earlier studies (Chen et al.,
ables can influence stock indices differently in the US and Canada. 1986; Fama & French, 1989; Fama, 1981; Hamao, 1988; Poon &
Second, although there have been many studies examining the Taylor, 1991; Schwert, 1990; Ferson & Harvey, 1991) use the Arbi-
association between composite stock market indices with macroe- trage Pricing Theory (APT) framework developed by Ross (1976),
conomic variables, there is a dearth in those extending the analysis more recent studies use cointegration technique. The APT assumes
to include sector specific stock market indices. Third, while most that an asset’s return is a linear function of economic and financial
of the previous studies limited their discussion to how domestic variables and is basically a linear multifactor model that aims to
macroeconomic variables can influence its stock market indices, measure the risk attached to the factors that influence the price of
we will investigate how the US and Canadian macroeconomic vari- the asset. When asset’s returns are regressed on different factors
ables can influence each other. Fourth, we cover the longest time in a multivariate regression framework, the coefficients capture
since the introduction of sector specific stock market indices. the risk/sensitivity of each factor. Chen et al. (1986) using monthly
This study uses the key composite indices, i.e. S&P500 for the data from the US for the period of 1958–1984 and employing the
US and TSX for Canada. Our sectoral indices include, energy (Enr), APT framework observe that while the relationship between stock
financials (Fin), consumer discretionary (CD), consumer staples returns, industrial production, changes in risk premium and the
(CS), real estate (RE), health care (HC), industrials (Ind), materi- changes in the term structure are positive, both expected and unex-
als (Mat), utilities (Uti), and technology (Tec). The macroeconomic pected inflation are negatively associated with stock returns due to
variables that have been used in this study are money supply, their impact on future dividends and discount rate.
long-term interest rate, and real economic activity. Applying the Poon and Taylor (1991) investigate whether the results found by
cointegration and vector error correction approach, our key results Chen et al. (1986) above for the US can be extended to the UK market
suggest that there are substantial variations between how macroe- using monthly data for the period of 1965–1984. The authors find
conomic variables are associated with stock market indices in the that comparable variables for the UK do not affect UK stock returns.
US and Canada. We find that macroeconomic variables in the US The authors suggest that there could be other macroeconomic fac-
64 E.M. Bhuiyan, M. Chowdhury / The Quarterly Review of Economics and Finance 77 (2020) 62–74
tors at work or that the methodology is inadequate to capture such short-term call money rate for the discount rate in the valuation
pricing relationships. The authors question whether the price of model of stock returns, since long-term bond rate has a negative
assets changes in a linear manner and consequently if the APT relationship with the TSE while call money rate has a positive rela-
is an appropriate framework and also suggest adopting Granger- tionship. Lee and Gan (2006) examine the relationship between
causality tests to investigate such relationships. The authors further several macroeconomic variables including money supply, long-
point out that if size of a firm is an important factor in being influ- and short-term interest rates, exchange rate, and inflation rate with
enced by macroeconomic variables in the US then it might not be the officially published index of the New Zealand Stock Exchange
possible to replicate the test for other countries as many uncon- (NZSE40) using cointegration framework on monthly observations
trollable variables such as company legislation, tax laws, other from January 1990 to January 2003. The results show that these
economic variables vary from country to country. Hamao (1988) macroeconomic variables have a long-term equilibrium relation-
replicates Chen et al. (1986) study for Japanese markets as a test ship with the NZSE40. However, based on the Granger causality
of robustness using monthly data for the period of 1975–1984. tests the authors suggest that money supply and the long-term
The factors investigated in this study include industrial produc- interest rate are not important in determining stock returns in the
tion, interest rates, inflation, investor confidence, and exchange short-run.
rate. Results show that other than industrial production, all other Cheung and Ng (1998) study on Germany, Italy, Japan, and the US
variables have an impact on equity returns in Japan. The sample suggest that the effect of money supply and gross national product
period for Hamao (1988) is much shorter than the sample period for on stock market indices are ambiguous given the different direc-
Chen et al., (1986) and Poon and Taylor (1991). Martinez and Rubio tion of influence these variables have over different countries. For
(1989) replicate the above studies using monthly observations for instance, money supply has a positive association with the compos-
Spain and find no significant relationship between macroeconomic ite indices for Germany and Italy while it has an inverse relationship
variables and stock returns. for Japan and the US. The authors conclude that the inconclusive
The development of cointegration analysis by Engle and Granger result is perhaps due to perceived differences in the monetary pol-
(1987) allows an alternative approach to study the long-term equi- icy implemented across countries. The authors also compare the
librium relations between the stock market and different economic VECM with VAR and finds that inclusion of error correction term
variables. Prior to the development of cointegration analysis, lin- in the VECM is an improvement over VAR based on the value of
ear regressions were used on non-stationary time series data and adjusted R2 . In a similar study, Nasseh and Strauss (2000) inves-
Granger and Newbold (1974) shows that such approach could lead tigate if the stock indices and macroeconomic variables exhibit a
to spurious correlations. If a set of time series variables are inte- cointegrating relationship for Germany, France, Italy, Netherlands,
grated of the same order and their linear combination produces Switzerland, and the UK. Results show that there is a long-term
a stationary series, then the set of those variables are said to be equilibrium relationship between the stock indices of each country
cointegrated. Cointegration points to the existence of a long-term with their domestic industrial production index, long- and short-
equilibrium relationship and through error correction modeling, term interest rates, and inflation. Generally, industrial production
the co-movement among the variables and the adjustment process has a positive association while the long-term interest rate has a
toward long-term equilibrium can be examined. Employing the negative association with the stock index. The direction of influ-
cointegration approach, a host of literature challenges the conclu- ence is consistent across all countries. Nesseh and Strauss also test
sions drawn by the EMH and provide evidence that macroeconomic whether German macroeconomic variables can influence the stock
variables help to predict the time series of stock returns (Cheung indices for France, Italy, Netherlands, Switzerland, and the UK given
& Ng, 1998; Humpe & Macmillan, 2009; Maysami & Koh, 2000; the strong economic linkage of these economies with Germany.
Mukherjee & Naka, 1995; Nasseh & Strauss, 2000; Ratanapakorn & Results show that the German short-term interest rate, stock prices,
Sharma, 2007). and industrial production significantly affect stock prices in the
Employing the cointegration and the Vector Error Correction other five economies while the reverse is not true.
Model (VECM) for monthly data from US, Ratanapakorn and Sharma The literature examining the relationship between macroe-
(2007) find that the S&P 500 has a positive relationship with money conomic variables and sector-specific stock indices is scanty.
supply, industrial production, inflation, exchange rate, and the Maysami, Howe, and Hamzah, (2004) study whether Singapore’s
short-term interest rate but a negative relationship with the long- financials and real estate along with a composite index form a
term interest rate. Further, the authors claim the stock market in cointegrating relationship with inflation, money supply, short-
the US does not seem to be efficient and can be forecasted by the and long-term interest rates, and exchange rate using monthly
information provided by these variables. Humpe and Macmillan observations from January 1989 to December 2001. For the finan-
(2009) conduct a similar exercise using the S&P 500 and the Nikkei cials sector, money supply, inflation, and short-term interest rate
225 for the US and Japan respectively. The authors find that for the have a positive relationship, while industrial production, long-term
US, consumer price index and the long-term interest rate have an interest rate, and exchange rate have a negative relationship with
inverse relationship, but industrial production and money supply the sector index. For the real estate sector, industrial production,
are positively associated with stock price although effect of money money supply, inflation, and short-term interest rate have a pos-
supply is not statistically significant. For Japan, stock prices are neg- itive relationship and long-term interest rate and exchange rate
atively related to money supply and positively related to industrial have a negative relationship with the real estate index. For the com-
production and the authors claim that the consumer price index posite index, money supply, industrial production, inflation, and
indirectly has a negative impact on stock prices through its effect short-term interest rate have a positive relationship while long-
on industrial production first. The authors attribute the contrast- term interest rate and exchange rate have a negative relationship
ing result for money supply to the liquidity trap that the Japanese with the index.
economy faced during the 1990 s where increasing money supply A survey of various studies has shown that there is a gap in
and falling interest rates were unable to pull the Japanese economy the literature when it comes to exploring the relationship between
out of its slump. Mukherjee and Naka (1995) also find the positive macroeconomic variables and sector specific stock indices. To the
relationship between the Tokyo Stock Exchange (TSE) with indus- best of our knowledge, none of the existing studies have explored
trial production, money supply, short-term interest rate, exchange the sensitivity of sectoral indices with macroeconomic indicators.
rate and negative with long-term bond rate, and inflation. The Despite having apparent similarities between the US and Canada in
authors suggest that long-term bond rate is a better proxy than macroeconomic structure, no research has investigated a compara-
E.M. Bhuiyan, M. Chowdhury / The Quarterly Review of Economics and Finance 77 (2020) 62–74 65
tive picture of how domestic and international factors can influence consequently the unit price of its shares (Ratanapakorn & Sharma,
stock market indices for both countries. Moreover, to the best of our 2007).
knowledge, we did not encounter any recent studies on this topic Our study proceeds in the following steps. First, we test the sta-
incorporating recent data since the great recession in the US. This tionarity of all the series using the Augmented Dickey Fuller (ADF)
could be important because the financial crisis of 2007–2008 could method. Second, we determine the order of integration of the series
have altered the economy fundamentally in such a way that the if it contains a unit root. Having all the series integrated of order
long-term relationship between the S&P 500 and different macroe- 1 indicates the appropriateness of using the cointegration tests.
conomic variables after the crisis may behave differently compared Third, the study proceeds with an unrestricted Vector Auto Regres-
to the period before the crisis. Hence, we address those gaps in the sion (VAR) model to determine the lag length to be employed in the
following sections. cointegration tests. Fourth, the Johansen cointegration tests iden-
tify whether the variables are cointegrated or not. Finally, if the
variables are cointegrated, we estimate a stable long-run relation-
3. Data and methods ship. Subsequently, we run a Vector Error Correction Model (VECM)
to determine the speed of adjustment for any variables deviating
3.1. The model from its long-run path. If variables are not cointegrated, VAR models
only capture the short-term relationships. Our choice of empirical
A pertinent issue is the mechanism behind how variables were method is based on the characteristics of the data and existing lit-
chosen to examine their relationship with the stock market. Most erature (Greene, 2012; Mukherjee & Naka, 1995; Nasseh & Strauss,
authors base their choice of variables through “simple and intu- 2000). As we are interested in examining the long-run relationship
itive financial theory” as noted by Chen et al. (1986). The financial between stock market indices and macroeconomic variables, there-
theory involves how the stock prices are determined in terms fore, we use the cointegration technique proposed by Johansen
of expected cash flow and discount rate. If there exists a pos- (1995). The cointegration method has several characteristics that
sibility that a change in macroeconomic variable results in a make it a preferred technique over several other techniques includ-
change in expected cash flow or discount rate then there is a ing the linear regression. It takes care of the issues related to the
corresponding change in stock prices and as such, the relation- spurious relationships due to the non-stationary nature of data.
ship between that variable and stock prices may be examined. If the variables are integrated of I(1), using any technique which
Among others, Mukherjee and Naka (1995), Cheung and Ng (1998), does not accommodate for the non-stationarity of data would be
Maysami and Koh (2000), Nasseh and Strauss (2000), Ratanapakorn misleading. Differencing of the non-stationary variables does not
and Sharma (2007), Humpe and Macmillan (2009) use such an always resolve the problems of spurious relationships (Greene,
approach to decide which variables to incorporate in their stud- 2012). The cointegration also resolves the issues related to reverse
ies. causality between variables. The flexible functional form of the
Theory suggests that an increased money supply may affect the Johansen (1995) cointegration treats all variables as endogenous
discount rate positively or negatively and thereby the consequent and avoids the arbitrary choice of the dependent variable in the
impact on stock prices is inconclusive (Mukherjee & Naka, 1995). On cointegrating equation. Hence, the use of cointegration is more
one hand, increased money supply may lower the interest rate due appropriate in our case.
to increased liquidity and consequently resulting in an increase in
stock prices due to lower discount rate (Ratanapakorn & Sharma, 3.2. Unit root tests
2007). On the other hand, increased money supply may result in
inflationary expectations and the interest rate may increase as a Economic and financial variables usually exhibit trending
consequence, which increases the discount rate and consequently behavior i.e. the series are non-stationary in their mean. Existence
results in lower stock prices (Dhakal et al., 1993). There are other of a unit root in a series means that the series is not station-
theoretical channels through which money supply may influence ary. Consequently, using non-stationary time series in a regression
stock prices as well. For instance, portfolio-balance model suggests framework may lead to spurious regression and futile economic and
that an increased money supply may cause a portfolio shift from statistical inference unless cointegration tests are performed where
non-interest-bearing money to financial assets such as equities all the series need to be integrated of the same order. For the pur-
(Friedman & Schwartz, 1963; Friedman, 1961). Also, an increase pose of this study all the time series variables must be integrated
in money supply may act as a stimulus for economic growth which of order 1. The Augmented Dickey Fuller (ADF) test has been used
may increase expected cashflows (Mukherjee & Naka, 1995). in this paper to test for a unit root. Given the nature of the variables
Several studies attempt to theoretically justify that industrial in this paper, the null hypothesis under the ADF test contains a unit
production, as a proxy for real economic activity, has a pos- root with a constant and a time trend. The following regression,
itive association with stock prices (Dhakal et al., 1993; Fama, which corresponds to a random walk with drift and time trend, is
1990; Schwert, 1990). In theory, however, the lead-lag relationship estimated to implement the ADF test:
between the industrial production and stock prices is unclear. First,
increased output may reflect accumulation of real assets which
P
Yt = ∝ + ıt + Yt−1 + i Y t−i + εt (1)
increases the productive capacity of an economy and consequently
i =1
increases the ability of firms to generate higher expected cash flows
in the future, and this may enhance stock price (Maysami et al., Where is the difference operator, Y corresponds to the variable,
2004). Alternatively, higher stock prices indicate increased wealth ıt represents the time trend, P is the number of lags and repre-
which may increase the demand for consumption and investment sents the error term. The number of lags is chosen by the Akaike
goods (Fama, 1990). Information Criteria (AIC). The t-statistic of is compared to the
The theory on the impact of interest rate on stock prices is ADF critical values to determine if the series contains unit root.
straightforward and suggests an inverse relationship. A higher
interest rate will increase the discount rate via its effect on the 3.3. VAR
nominal risk-free rate which will result in lower stock prices (Chen
et al., 1986). Additionally, an increased interest rate may raise the Vector Auto Regressive (VAR) models attempt to capture linear
financing cost which will reduce the profitability of the firm and interdependencies among multiple time series. These models are
66 E.M. Bhuiyan, M. Chowdhury / The Quarterly Review of Economics and Finance 77 (2020) 62–74
extensions of univariate Auto Regressive (AR) models by allowing ables. This test is based on a general VAR model with n variables
for more than one dynamic variable. While in a standard regression (Johansen, 1995):
framework, dependent and independent variables interact simul-
taneously, in VAR models, evolution of each variable is based on the
K
lagged values of itself, and lagged values of other variables in the Yt = v + i Yt−i + εt (6)
model. The VAR model in this study is as follows: i=1
p
p The following equation is derived after subtracting Yt-1 from
Indext = ∝0 + ˇ1i Indext−i + ˇ2i IP t−i both sides:
i=1 i =1
K−1
p
p Yt = v + Y t−1 + џi Yt−i + εt (7)
+ ˇ3i M1t−i + ˇ4i LIRt−i + εt (2) i=1
i =1 i=1
K
K
Here, = i − I and џi = − j
p
p
I=1 j=i+1
IP t = 0 + 1i Indext−i + 2i IP t−i In the above equations, Y is the n by 1 vector of endogenous
i=1 i =1 variables, v is a vector of parameters, is the vector of residuals.
is the n by n matrix of parameters for endogenous variables, and I
p
p
+ 3i M1t−i + 4i LIRt−i + t (3) is the identity matrix of dimension n. The matrix has rank 0 ≤
r < n, where
r is the number of linearly independent cointegrating
i =1 i=1
vectors. can be written as the product of ∝ and ’ matrices where
∝ represents the parameters for the speed of long-run adjustment
p
p and  contains r cointegrating vectors.
M1t = 0 + 1i Indext−i + 2i IP t−i The trace test and maximum eigenvalue test determines the
i=1 i =1
number of cointegrating vectors. The null hypothesis for the trace
test is that there are at most r cointegrating relations against the
p
p
alternative j cointegrating equations where r = 0, 1,2, . . ., j – 1. The
+ 3i M1t−i + 4i LIRt−i + D t (4) null hypothesis for maximum eigenvalue test is that there are r
i =1 i=1 cointegrating relations against the alternative of r + 1 cointegrating
relations.
p
p
test the dynamics among the variables. + 3i M1t−i + 4i LIRt−i + 2 zt−1 + t (8)
The Johansen cointegration test is applied in this study to i =1 i=1
determine the existence of a long-run relationship among the vari-
E.M. Bhuiyan, M. Chowdhury / The Quarterly Review of Economics and Finance 77 (2020) 62–74 67
p
p Table 1
M1t = ∝0 + 1i Indext−i + 2i IP t−i ADF Unit Root Test for US Variables for 2000–2018.
p
p Indices
S&P 500 −2.376 −5.663 ***
+ 3i M1t−i + 4i LIRt−i + 3 zt−1 + ët (9)
Enr −1.701 −10.094 ***
i =1 i=1 Fin −1.837 −3.834 **
CD −1.944 −6.855 ***
CS −2.577 −10.492 ***
p
p
RE −2.159 −4.813 ***
LIRt = ∝0 + 1i Indext−i + 2i IP t−i HC −1.323 −10.395 ***
i=1 i =1 Ind −2.803 −5.740 ***
Uti −2.133 −7.722***
p
p
Mat −3.274 −10.546 ***
+ 3i M1t−i + 4i LIRt−i + 4 zt−1 + t (10) Tec −3.625** NA
Macroeconomic Variables
i =1 i=1
IP −3.071 −4.242 ***
Where: Indext represents changes in composite or sector indices M1 −1.901 −5.050***
from one time period to the next; IPt represents changes in indus- LIR −3.064 −7.729 ***
trial production from one time period to the next; M1t represents Notes: The numerical values represent ADF test statistic. ***≤0.001, **≤0.005,
changes in money supply from one time period to the next; LIRt *≤.01. All the level series are in natural logarithmic form.
represents changes in the long-term interest rate from one time
period to the next. All the variables are in natural logarithmic form. in the S&P 500. For the US sector indices, the weight of any single
‘p’ denotes the number of lagged differences. ‘z’ is the error correc- index constituent is capped at 20 %. For Canada, constituents of all
tion term. , , ë, and t represent error terms. The coefficient , of the sector indices are selected from a stock pool of the benchmark
the error correction term measures the speed of adjustment when S&P/TSX composite index. For Canadian indices, the weight of any
there is a deviation from the equilibrium. The coefficient vectors , single index constituent is capped at 25 %. All the sector indices
, , capture the short-run dynamics between the variables. are based upon the Global Industry Classification Standards (GICS).
The GICS is an industry classification system developed by MSCI
4. Data, results, and discussion (formerly Morgan Stanley Capital International) and Standard and
Poor’s (S&P) for use by the global financial community.
4.1. Data There are various forms of indices such as uncapped, equal
weight, 35 % capped, and 20 % capped index for almost all the sec-
This study uses monthly observations for all the stock indices tors for the US. However, forms of sector indices data are limited
and macroeconomic variables over the period ranging from January for Canada. For comparison purposes across all sectors with the US,
2000 to June 2018 for the United States and from January 2000 to only 25 % capped index is available. For the US, 20 % capped indices
April 2018 for Canada. The sample for the US contains 222 observa- have been used. The rationale for using capped sector indices is
tions, the sample for Canada includes 220 observations. Data for that, the number of stocks listed and consequently the market cap-
the US sector indices are available from January 2000 and data italization of sector indices is much lower compared to the original
for most Canadian sectors are available from 1997. For comparison benchmark index from which sector indices were created. Move-
purposes with the US, this study uses Canadian data from January ment in any stock, which forms much of a sector in terms of market
2000. The data for all the macroeconomic variables, industrial pro- capitalization, could be due to company specific reasons and not
duction index, long-term interest rate, and narrow money supply due to changes in macroeconomic variables. As a result, movement
have been collected from the Federal Reserve Bank of St. Louis for in uncapped sector indices is subject to a lot of noise due to com-
both countries. Prior studies use M1 money supply and the indus- pany specific factors, specifically those companies which have a
trial production index to capture the effects of money supply and large weight in the sector. Such a problem does not exist for the
real economic activity respectively, and to be consistent, this paper benchmark indices for the US or Canada due to the sheer number of
also uses M1 money supply and the industrial production index. stocks listed in each of them. Market capitalization of an individual
Following Humpe and Macmillan (2009), seasonally adjusted data stock listed in any of the benchmark indices is very low compared
for M1 and Industrial Production are used in this study as these to the entire index and as a result, movements in the price of a
variables exhibit strong seasonality. To capture the effect of the stock due to company specific factors will hardly cause a dent in
long-term interest rate, most studies use 10-year or 5-year govern- the direction of the entire index.
ment bond rate. This paper uses benchmark 10-year government
bond rate, since data for this series is available for both Canada and 4.2. Unit root tests
the US at the Federal Reserve Bank of St. Louis. While most studies
use additional variables such as the short-term interest rate, infla- This study uses the Augmented Dickey Fuller (ADF) test to detect
tion, and exchange rate, they generally test the relationship against the presence of unit root in the variables. The ADF tests are under
a single benchmark index of a country. Since this paper investigates the null of a unit root with a constant and a time trend, which
ten different indices for the US and ten different indices for Canada, was selected based on the trending behavior of all the series.
attention was given to a parsimonious model with sufficient num- Tables 1 and 2 display the results of the ADF test for the different
ber of macroeconomic variables. stock indices and macroeconomic variables for the US and Canada
All the stock market composite and sector indices data for both respectively. For the purpose of robustness check, results of the
the countries have been obtained from Bloomberg and are based Phillips-Perron test to detect the presence of unit root are presented
on the closing price of the indices on the last business day of each in Appendix B. Table 1 shows that for the US, all macroeconomic
month. All the level series of indices and macroeconomic variables series, S&P500 and all sector indices other than the technology
are expressed in natural logarithmic form. Variable description and sector index contain unit root and are therefore non-stationary
summary statistics are presented in Appendix A. Companies listed processes, but the first difference of most of the series turn out
in all the sector indices for the US are selected from the stocks listed to be stationary. This suggests that all the variables other than the
68 E.M. Bhuiyan, M. Chowdhury / The Quarterly Review of Economics and Finance 77 (2020) 62–74
Table 2 Table 3
ADF Unit Root Test for Canadian Variables for 2000–2018. Cointegration Test for US Indices and Macroeconomic Variables for 2000–2018.
zero in first differences for both the US and Canada, the Johansen S&P 500 r=0 37.0308** 18.3683**
cointegration test can be employed to see whether there exists r=1 18.6626 10.6666
r=2 7.9959 5.2682
a long-run relationship between the stock indices and industrial
Enr r=0 31.6042** 13.7036**
production, long-term interest rate and money supply. The trace r=1 17.9006 9.1548
statistic and the maximum eigenvalue statistic are compared with r=2 8.7458 5.9831
the critical value at 5 % level of significance. Tables 3 and 4 display Fin r=0 29.0896** 12.7180**
results using the Johansen cointegration test for the US and Canada, r=1 16.3716 9.1436
r=2 7.2280 5.2000
respectively. CD r=0 37.2123** 20.1998**
Choosing the number of lags is subjective and is dependent on r=1 17.0125 11.1437
a multitude of factors such as context of study, empirical evidence, r=2 5.8688 3.1538
and theory. Too few lags could result in residual autocorrelation CS r=0 36.8213** 23.8302**
r=1 12.9911 7.0412
and too many lags would lead to losing observations and error
r=2 5.9499 5.3737
in forecasts (Stock & Watson, 2001). This paper has tested for lag RE r=0 31.8286** 14.9835**
lengths between 1 and 12 and the Akaike Information Criteria (AIC) r=1 16.8450 9.0688
was chosen to determine optimal lag length. The AIC tends to be r=2 7.7762 5.5110
more accurate with monthly data in the context of Vector Auto HC r=0 40.1864** 25.7687**
r=1 14.4176 8.6298
Regression (VAR) models (Ivanov & Kilian, 2001). r=2 5.7878 3.3508
The trace and maximum eigenvalue statistics for the S&P 500 Ind r=0 37.0716** 16.2832*
and all the sector indices for the United States reject the null of r=1 20.7884 13.9328
0 rank, that is “Ho: no cointegrating vector”. For all the indices, r=2 6.8556 4.6315
Mat r=0 34.4853** 17.3312**
both the trace test and maximum eigenvalue test fail to reject
r=1 17.1541 9.2622
the null hypothesis that a rank of 1 exists. The results imply that r=2 7.8919 4.9849
there exists a singular cointegrating vector between the indices and Uti r=0 36.1843** 14.4494**
the macroeconomic variables. This means that there is evidence of r=1 21.7349 12.3348
long-term relationship among the indices and industrial produc- r=2 9.4001 5.9017
tion, money supply, and the long-term interest rates for the US. For Note: *** ***≤0.001, **≤0.005, *≤.01.
robustness check, different lag lengths suggested by other informa-
E.M. Bhuiyan, M. Chowdhury / The Quarterly Review of Economics and Finance 77 (2020) 62–74 69
1 2
The sign of the normalized cointegrating coefficients are reversed to enable The results of the full VECM with equations for industrial production, money
proper interpretation. For details, see Johansen (1995). supply, and the long-term interest rate are available upon request.
70 E.M. Bhuiyan, M. Chowdhury / The Quarterly Review of Economics and Finance 77 (2020) 62–74
there is a presence of cointegration implying predictive causality these foreign markets display significantly higher returns when the
and hence market inefficiency. However, at the same time Canadian Federal Reserve is following an expansive monetary policy. Using
variables fail to explain the US indices in a cointegrating framework. the ARCH framework, Hamao, Masulis, and Ng, (1990)) report the
The discrepancy between the countries could be explained in the presence of price volatility spillovers from New York to Tokyo and
following manner. The monetary policy is much more involved for London markets but not the opposite. Harvey (1991) finds that vari-
the US, and the Federal Reserve, beyond utilizing traditional mone- ables that represent the US business cycle can explain foreign stock
tary policy instruments, also engages in non-traditional monetary returns while Campbell and Hamao (1992) find that the US eco-
policy. So, for the US, according to Dhakal, Kandil, and Subhash nomic variables improve the forecasts of Japanese stock returns.
(1993), even if information regarding changes in stock of money The above studies highlight the dominant role of the US in the world
is readily available to traders, the implications of these changes economy.
on the direction of monetary policy may not be readily apparent. Considering the financial crisis of 2007/2008, this study repeats
Consequently, information regarding policy changes may not be the cointegration test using data from January 2010 for the US and
reflected in the stock prices. In contrast, the monetary policy for finds that energy, consumer staples, and health care sectors no
Canada is straightforward. The Bank of Canada’s only monetary pol- longer have a long-term relationship with money supply and the
icy instrument is the target it sets for the overnight interest rate, long-term interest rate. Results are not as robust as before for real
which is a market determined rate at which commercial banks lend estate, materials, and utilities sectors because the trace test and
funds to each other for a very short period of time. Given that the the max eigenvalue test show conflicting outcomes. Only the S&P
Bank of Canada’s strategies are much easier to interpret compared 500, financials, and consumer discretionary sectors have long-term
to the Fed’s policies, it is much easier for the traders to assess the relationship with money supply and the interest rate according to
direction of Canadian monetary policy. As a result, stock prices in both trace and max statistics. The industrial production index was
Canada readily incorporate information regarding policy changes dropped as the ADF test showed that the series does not contain
compared to the US. Stocks in the S&P/TSX are inter-listed in the a unit root at level form. Results may indicate the possibility that
US exchanges, moreover the Canadian economy is highly integrated the 2007/2008 crisis might have fundamentally changed the econ-
to the US economy with the US being Canada’s largest trading part- omy in such a way that the relationship between the stock indices
ner; Canada exports around 30 % of its GDP, equivalent to 80 % and money supply and the interest rate has weakened. As a part
of its exports, to the US. International trade is very important to of robustness check, since the US money supply and the long-term
Canada relative to the US whose exports account for around 10–12% interest rate have a long-term equilibrium relationship with Cana-
of its GDP compared to Canada’s 30–35%. Such reasons provided the dian indices using 2000–2018 data, test is now repeated using data
motivation to test whether the US variables can explain Canadian from January 2010 onward. Results show that US money supply and
stock returns. It turns out that the US money supply, M1, and long- the long-term interest rate are no longer able to explain any of the
term interest rate are cointegrated with all the Canadian sectors Canadian stock indices. At the same time, just as before, Canadian
including the composite S&P/TSX. At the same time the Canadian money supply and the long-term interest rate cannot explain any
variables are not cointegrated with the US indices. of the US indices using data from January 2010.
The results are not surprising. For instance, Cheung, He, and
Ng, (1997)) find that the US variables are able to explain Euro-
pean and Pacific Rim’s stock market movement, but the reverse 5. Conclusion
is not true. Similarly, Eun and Shim (1989) report that shocks in
the US stock market are transmitted to other international mar- This paper finds that there is a long-term equilibrium rela-
kets while no single foreign market can significantly explain the tionship between macroeconomic variables that represent money
US stock movements. Conover, Jensen, and Johnson, (1999)) find supply, real economic activity, and long-term interest rate and S&P
that some foreign stock markets are more strongly related to the 500 and all sector indices for the US for the 2000–2018 time period.
US monetary environment than to local monetary conditions and Money supply generally has a positive relationship, while the long-
term interest rate has a negative relationship with the indices
72 E.M. Bhuiyan, M. Chowdhury / The Quarterly Review of Economics and Finance 77 (2020) 62–74
Note: In US, market cap in USD billions. (Source: S&P Global). In Canada, market cap.
Appendix A in CAD billions. (Source: TMX Money).
Table A3 Table B2
Summary Statistics of US Stock Indices & Macroeconomic Variables (Natural Log Phillips-Perron Unit Root test for Canadian Variables for 2000–2018.
form):2000–2018
Index Level First Difference
Index Mean Std. Dev. Min Max
S&P/TSX −2.587 −12.089***
SPX 7.247466 .3014701 6.599993 7.945842 Enr −1.896 −13.650***
Enr 5.322002 .4419402 4.432137 5.977051 Fin −2.521 −11.821***
Fin 4.603877 .3200504 3.48765 5.099531 CD −1.400 −12.218***
CD 4.732602 .4396156 3.872325 5.710311 CS −1.517 −14.760***
CS 5.026783 .3525152 4.437568 5.687236 RE −1.507 −13.913***
RE 5.000024 .2997437 3.939054 5.415433 HC −2.265 −12.413***
HC 4.947539 .3645726 4.418684 5.761136 Ind −2.331 −11.561***
Ind 4.952174 .3433907 4.219934 5.733775 Uti −2.668 −15.567***
Mat 4.884961 .3376732 4.238329 5.535829 Mat −1.490 −15.105***
Uti 4.838592 .2602242 4.167254 5.327551 Tec −4.199*** NA
Tec 4.028656 .4099992 3.161458 4.958394 Macroeconomic Variables Level First Difference
Macroeconomic Variables Mean Std. Dev. Min Max IP −1.259 −13.943***
IP 4.592483 .051341 4.466705 4.682121 M1 −1.733 −13.550***
M1 7.507704 .3929106 6.992648 8.20732 LIR −2.552 −11.200***
LIR 1.186284 .3709966 .4054651 1.874874
Notes: The numerical values represent the PP Z(t) test statistic. ***≤0.001,
**≤0.005, *≤.01. All the level series are in natural logarithmic form.
Table A4
Summary Statistics Canadian Stock Indices & Macroeconomic Variables (Natural Log
Appendix C
Form): 2000–2018.
Notes: The numerical values represent the PP Z(t) test statistic. ***≤0.001,
**≤0.005, *≤.01. All the level series are in natural logarithmic form.
74 E.M. Bhuiyan, M. Chowdhury / The Quarterly Review of Economics and Finance 77 (2020) 62–74
Table C2 Ferson, W. E., & Harvey, C. R. (1991). The variation of economic risk premiums. The
Long-Term Estimates for Canadian Indices with US Money Supply and Interest Rate. Journal of Political Economy, 99(2), 385–415.
Friedman, M. (1961). The lag in effect of monetary policy. The Journal of Political
Indices M1 LIR Constant Economy, 69(5), 447–466.
S&P 500 5.76** 11.08** −66.07036 Friedman, M., & Schwartz, A. J. (1963). Money and business cycles. The State of
(1.96) (2.05) Monetary Economics NBER, 32–78.
Granger, C. J. (1986). Developments in the study of cointegrated economic
Enr −16.18** −29.93** 151.9373
variables. Oxford Bulletin of Economics and Statistics, 48(3), 213–228.
(5.23) (5.49)
Granger, C. W. J., & Newbold, P. (1974). Spurious regressions in econometrics.
Fin 1.173* 3.26** −18.00484 Journal of Econometrics, 2(2), 111–120.
(0.683) (0.739) Greene, W. H. (2012). Econometric analysis. Pearson Education Inc.
CD −0.259 1.143** −4.249018 Hamao, Y. (1988). An empirical examination of the arbitrage pricing theory. Japan
(0.345) (0.361) and the World Economy, 1(1), 45–61.
CS 2.46** 6.57** −31.8472 Hamao, Y., Masulis, R., & Ng, W. (1990). Correlations in price changes and volatility
(1.19) (1.252) across international stock markets. The Review of Financial Studies, 3(2),
RE 2.95** 6.641** −35.42252 281–307.
(1.173) (1.23) Harvey, C. (1991). The world price of covariance risk. The Journal of Finance, 46(1),
HC 3.16** 6.61** −36.19436 111–157.
(1.212) (1.271) Humpe, A., & Macmillan, P. (2009). Can macroeconomic variables explain
Ind −9.391** −14.54** 83.50051 long-term stock market movements? A comparison of the US and Japan.
Applied Financial Economics, 19(2), 111–119.
(2.39) (2.508)
Ivanov, V., & Kilian, L. (2001). A practitioner’s guide to lag-order selection for vector
Mat −8.55** −14.77** 76.83764
autoregressions. CEPR Discussion Papers, 2685.
(2.634) (2.76)
Johansen, S. (1995). Likelihood−based inference in cointegrated vector autoregressive
Note: ***≤0.001, **≤0.005, *≤.01. models. Oxford (United Kingdom): Oxford University Press.
Kwon, C. S., & Shin, T. S. (1999). Co-integration and causality between
macroeconomic variables and stock market returns. Global Finance Journal,
References 10(1), 71–81.
Lee, M., & Gan, C. (2006). Macroeconomic variables and stock market interactions:
New Zealand evidence. Investment Management and Financial Innovations, 3(4),
Abdullah, D. A., & Hayworth, S. C. (1993). Macro-econometrics of stock price
89 –101.
fluctuations. Quarterly Journal of Business and Economics, 32(1), 50–67.
Martinez, M. A., & Rubio, G. (1989). Arbitrage pricing with macroeconomic variables:
Asprem, M. (1989). Stock prices, asset portfolios and macroeconomic variables in
An empirical investigation using Spanish data, Working Paper. Universidad del
ten European countries. Journal of Banking and Finance, 13(4-5), 589–612.
Pais Vasco.
Bureau of Labor Statistics, Monthly Labor Rlabor review, Accessed online https://
Maysami, R. C., & Koh, T. S. (2000). A vector error correction model of the Singapore
www.bls.gov/opub/mlr/2016/article/current-employment-statistics-survey-
stock market. International Review of Economics & Finance, 9(1), 79–96.
100-years-of-employment-hours-and-earnings.htm, Accessed on June 4, 2019.
Maysami, R. C., Howe, L. C., & Hamzah, M. A. (2004). Relationship between
Campbell, J. Y., & Hamao, Y. (1992). Predictable stock returns in the United States
macroeconomic variables and stock market indices: Cointegration evidence
and Japan: A study of long-term capital market integration. The Journal of
from stock exchange of Singapore’s all-s indices. Journal of Pengurusan, 24,
Finance, 47(1), 43–69.
47–77.
Campbell, J. Y., & Shiller, R. J. (1988). The Dividend-Price ratio and expectations of
Mukherjee, T. K., & Naka, A. (1995). Dynamic relations between macroeconomic
future dividends and discount factors. The Review of Financial Studies, 1(3),
variables and the Japanese stock market: An application of a vector error
195–228.
correction model. The Journal of Financial Research, 18(2), 223–237.
Chen, N. F., Roll, R., & Ross, S. (1986). Economic forces and the stock market. Journal
Nasseh, A., & Strauss, J. (2000). Stock prices and domestic and international
of Business, 59(3), 383–403.
macroeconomic activity: A cointegration approach. The Quarterly Review of
Cheung, Y. W., He, J., & Ng, L. K. (1997). Common predictable components in
Economics and Finance, 40(2), 229–245.
regional stock markets. Journal of Business & Economic Statistics, 15(1), 35–42.
Pearce, D., & Roley, V. (1983). The reaction of stock prices to unanticipated changes
Cheung, Y. W., & Ng, L. K. (1998). International evidence on the stock market and
in money: A note. The Journal of Finance, 38(4), 1323–1333.
aggregate economic activity. Journal of Empirical Finance, 5(3), 281–296.
Poon, S., & Taylor, S. (1991). Macroeconomic factors and the UK stock market.
Cohen, R. B., Christopher, P., & Vuolteenaho, T. (2003). The value spread. The Journal
Journal of Business Finance & Accounting, 18(5), 619–636.
of Finance, 58(2), 609–641.
Ratanapakorn, O., & Sharma, S. C. (2007). Dynamic analysis between the US stock
Conover, C. M., Jensen, G. R., & Johnson, R. R. (1999). Monetary environments and
returns and the macroeconomic variables. Applied Financial Economics, 17(5),
international stock returns. Journal of Banking & Finance, 23(9), 1357–1381.
369–377.
Davidson, L. S., & Froyen, R. T. (1982). Monetary policy and stock returns: Are stock
Ross, S. A. (1976). The arbitrage theory of capital asset pricing. Journal of Economic
markets efficient? Federal Reserve Bank of St. Louis Review, 64(3), 3–12.
Theory, 13, 341–360.
Dhakal, D., Kandil, M., & Subhash, C. S. (1993). Causality between the money supply
Samuelson, P. A. (1998). Summing upon business cycles: Opening address.
and share prices: A VAR investigation. Quarterly Journal of Business and
COnference Series-Federal Reserve Bank of Boston, 42, 33–36.
Economics, 32(3), 52–74.
Schwert, G. W. (1990). Stock returns and real activity: A century of evidence. The
Dibooglu, S. (1995). Multiple cointegration and structural models: Applications to
Journal of Finance, 45(4), 1237–1257.
exchange rate determination. Dissertations. Iowa State University. https://lib.dr.
Shiller, R. J. (1981). Do stock prices move too much to be justified by subsequent
iastate.edu/rtd/10420
changes in dividends? The American Economic Review, 71(3), 421–436.
Engle, R. F., & Granger, C. W. J. (1987). Cointegration and error correction:
Sorensen, E. (1982). Rational expectations and the impact of money upon stock
Representation, estimation, and testing. Econometrica, 55(2), 251–276.
prices. The Journal of Financial and Quantitative Analysis, 17(5), 649–662.
Eun, C. S., & Shim, S. (1989). International transmission of stock market
Stock, J. H., & Watson, M. W. (2001). Vector autoregressions. The Journal of
movements. The Journal of Financial and Quantitative Analysis, 24(2), 241–256.
Economic Perspectives, 15(4), 101–116.
Fama, E. F. (1970). Efficient capital markets: A review of theory and empirical work.
Vuolteenaho, T. (2002). What drives firm-level stock returns? The Journal of
The Journal of Finance, 25(2), 383–417. http://dx.doi.org/10.2307/2325486
Finance, 57(1), 233–264.
Fama, E. F. (1981). Stock returns, real activity, inflation and money. The American
Washington Post (2014). Accessed online from https://www.washingtonpost.com/
Economic Review, 71(4), 545–565.
blogs/govbeat/wp/2014/09/03/watch-the-u-s-transition-from-a-
Fama, E. F. (1990). Stock returns, expected returns, and real activity. The Journal of
manufacturing-economy-to-a-service-economy-in-one-gif/?utm term=.
Finance, 45(4), 1089–1108.
b712fbbca43d, Accessed on December 10, 2018.
Fama, E. F., & French, K. R. (1989). Business conditions and expected returns on
Young, P. (2006). Industrial production and stock returns Doctoral dissertation.
stocks and bonds. Journal of Financial Economics, 25(1), 23–49.
Faculty of Business Administration-Simon Fraser University.