Munich Personal RePEc Archive

Macroeconomic Forces and Stock

Prices:Evidence from the Bangladesh
Stock Market

Khan, Mashrur Mustaque and Yousuf, Ahmed Sadek

24 April 2013

Online at https://mpra.ub.uni-muenchen.de/46528/
MPRA Paper No. 46528, posted 25 Apr 2013 05:12 UTC
 Macroeconomic Forces and Stock Prices:
Evidence from the Bangladesh Stock Market

Mashrur Mustaque Khan

Research Analyst
International Food Policy Research Institute
mashrur.m.khan@gmail.com
House: 10A, Road: 35, Gulshan – 2,
Dhaka 1212, Bangladesh

Ahmed Sadek Yousuf

Senior Research Associate
Policy Research Institute of Bangladesh
ahmed.sadek87@gmail.com
House: 16, Road: 10/A, Banani,
Dhaka 1213, Bangladesh

April 21, 2013

1
Abstract
The study examines the influence of a selective set of macroeconomic forces on stock market
prices in Bangladesh. The Dhaka Stock Exchange All-Share Price Index (DSI) is used to
represent the prices in the stock market while deposit interest rates, exchange rates, consumer
price index (CPI), crude oil prices and broad money supply (M2) are selected to represent the
macroeconomic variables affecting the stock prices. Using monthly data from 1992m1-2011m6,
several time-series techniques were used which include Cointegration, Vector Error Correction
Model (VECM), Impulse Response Functions (IRF) and Variance Decompositions (VDC).
Cointegration analysis, along with the VECM, suggests that interest rates, crude oil prices and
money supply are positively related to stock prices, exchange rates are negatively related to
stock prices, and CPI is insignificant in influencing the stock prices, in the long-run. Both the
IRF and VDC suggest that shocks to macroeconomic variables explain a small proportion of the
forecast variance error of the DSI, but these effects persist for a long period.

Keywords: Asset Prices, Macroeconomic Factors, Dhaka Stock Exchange, Cointegration

JEL Classification: C22, E44, E58, G12

2
Introduction

Stock markets, where shares and bonds are traded and issued through exchanges

and over-the-counter markets, form an integral part of the financial markets and are

important for the development of an economy. Stock markets contribute to the economy

by providing businesses with access to capital and investors with opportunities for

capital gains. Research has shown that stock market development contributes

significantly to the economic growth of a country (Levine and Zervos, 1996)1. Since

stock market prices are subject to fluctuations, it is essential to determine the forces

influencing the stock prices for efficient functioning and development of the stock

market and country.

There are many reasons for there to be an interest to determine the forces

influencing the stock prices. Firstly, this may interest investors, so they can forecast

stock prices accurately to make apt decisions regarding their stock portfolio for

maximum gains. Secondly, businesses may find this useful; as stock price is an

indication of the financial health of the companies, businesses may be interested to

determine the future stock prices as it will allow them to assess their ability to issue

bonds or obtain financing in the future. Thirdly, policymakers and economists may find

this useful, so they can predict stock prices as prices of stocks reflect changes in

economic activity in the long run (Cheung and Lai, 1999).

Stock prices are expected discounted dividends, i.e. discounted value of future

cash flows derived from a stock. Theoretically, stock prices are modelled as:

n
E (CF )
P
t 1 (1  R) t

1
For a debate on the relationship between financial/stock market development and economic
growth, see Gurley and Shaw (1955, 1960, 1967), McKinnon (1973), Shaw (1973), Goldsmith
(1969) and Levine (2004)

3
where P refers to the stock price, CF refers to the expected cash flows derived from a

stock and R refers to the discount rate. Hence, any forces influencing the R or CF will

affect the stock prices. However, theoretical models do not provide an ‘identity’ of these

exogenous economic forces (Bodhurta et al., 1989), i.e. do not identify the economic

forces influencing the stock prices. Macroeconomic variables are potential candidates

for these forces because macroeconomic changes simultaneously affect many firms’

cash flows and may influence the risk-adjusted discount rate (Shiller, 1981; Leroy and

Porter, 1981; Flannery and Protopapadakis, 2002), or simply, discount rate.

The purpose of the research will be to try to find a long-term relationship

between macroeconomic forces and stock prices for an emerging stock market in a less

developed country. The study will focus on an emerging market because the behaviour

of stock prices in these countries is not tied to economic fundamentals (Gunasekarage et

al., 2004) and, therefore, makes it difficult to predict the forces affecting the stock

prices. Moreover, studies on emerging markets have shown that returns and risks in

these markets are higher relative to those in stock markets in developed countries

(Harvey, 1995a). It will be interesting to determine what factors cause these higher risks

and returns and study the relationship between macroeconomic forces and stock prices

in emerging stock markets.

This research will study the relationship between a selective set of

macroeconomic variables and stock prices in the Bangladesh stock market. The

Bangladesh stock market is an established capital market and deemed as the next Asian

success story by JPMorgan Chase, Citigroup, and Merrill Lynch (Bloomberg, 1997); its

stocks have performed well in recent years and prices gained nearly 50% over one year

in 2010 (2nd highest in the world after Sri Lanka)2. However, the stock market is still

2
Bespoke Investment Group (2010)

4
developing and the analysis made in this study can, therefore, be used to shed light on

other emerging stock markets.

For the purpose of the study, the stocks from the Dhaka Stock Exchange (DSE),

the primary stock market of Bangladesh, will be considered as it covers majority of the

stocks in the country and will allow a more comprehensive analysis. The DSE uses

three share price indices - DSE All-Share Price Index (DSI), DSE General Price

(DGEN) Index, and DSE-203. The DSI is a statistical compilation of all the stocks in the

DSE including Z-category shares and will be used as a proxy for stock prices. The study

will cover the last two decades since the DSE became very active during this period due

to developments in the capital markets of Bangladesh, such as exemption of tax

dividends on stock market investments to increase stock trades, off-loading of shares of

government-owned companies, allowing investment of black money4 into the capital

market, etc.

The next section reviews the existing literature on the topic and discusses the

methodologies used in the papers. Section 3 describes the macroeconomic variables

used in the study along with their hypothesized relationships with the stock prices.

Section 4 discusses the sources from which the data were collected and provides

descriptive statistics of the data. Section 5 explains the econometric model and

methodologies used in the study. Section 6 provides the empirical results with

interpretations. Section 7 concludes the paper and offers further remarks.

3
DGEN Index excludes Z-category securities, and no mutual fund, bond and debenture are
considered in this index. DSI is formed with all enlisted securities excluding mutual fund, bond
and debenture. DSE-20 is structured with the 20 best enlisted companies depending on
performance and specific criteria
4
Money undisclosed to the tax authorities and on which due taxes have not been paid

5
Literature Review

Chen, Roll and Ross (1986) (CRR) was one of the pioneer papers that tried to

identify the macroeconomic variables that influenced stock returns and determined this

relationship for the New York Stock Exchange (NYSE). They used a regression

framework to test whether macroeconomic innovations such as monthly growth in

industrial production, expected inflation and unexpected inflation5, and an interest rate

spread variable have systematic influences on stock market returns. To this end, they

estimated a Vector Autoregression (VAR) model of lagged stock market returns and

used the residuals of the macroeconomic variables as unanticipated innovations arising

from them. They found that industrial production, changes in the risk premium, term

structure, unexpected inflation and changes in anticipated inflation were significant in

explaining expected stock returns in the NYSE. They also used value-weighted NYSE

index as a macroeconomic variable, and found that it has an insignificant influence on

expected returns.

Poon and Taylor (1991) used the dataset for the London stock market and the

same macroeconomic variables as CRR and found that no significant relationship exists

between stock returns and the macroeconomic variables. Diacogiannis (1986) and

Cheng (1995), similarly, determined that there is no conclusive result regarding the

relationship of relevant macroeconomic variables with the capital market of the U.K.

Günsel and Çukur (2007), using the same variables as CRR, looked into different

industries in the U.K. and found that macroeconomic factors have a significant effect in

the U.K. stock exchange market; however, each factor affect different industry in

different manner.

5
Expected inflation was constructed following Fama and Gibbons (1984)

6
 Further work on the topic has extended the analysis by incorporating different

framework/setting and conducted the study using different econometric methods.

Bodhurta et al. (1989) undertook an analysis for seven major industrial countries –

United States, Japan, United Kingdom, Germany, France, Canada and Australia. They

chose the same explanatory variables as CRR, and to incorporate an international

setting, introduced deviations from Purchasing Power Parity, typified as real exchange

rate changes, and interest rate parity. They were able to demonstrate that several of the

international analogs of the CRR domestic variables - stock index returns, industrial

production, bond returns, unanticipated inflation and oil prices are significant in

explaining the average stock returns in the cross-section sample.

Mukherjee and Naka (1995) used a different methodology to determine the

relationship between macroeconomic forces and stock returns. They employed

Johansen’s (1991) Vector Error Correction Model to examine whether relevant

macroeconomic variables and the Tokyo Stock Exchange (TSE) index were

cointegrated for the period 1971-1990. They found that a cointegrating relationship

exists and that stock prices contributed to the relation. The relationships between stock

index and exchange rates, inflation, money supply, and industrial production were as

hypothesized and the same as existing literature. Nasseh and Strauss (2000) used

Johansen’s cointegration procedure and variance decompositions method for 1962-1995

for six European countries: France, Germany, Italy, The Netherlands, Switzerland and

the U.K., and found support for the existence of a long-run cointegrating relationship

between stock prices and domestic interest rates, consumer prices, real industrial

production, business surveys of manufacturing orders, and International (German)

movements in stock prices.

There have only been a few studies focused on the emerging markets in less

developed countries. Wongbangpo and Sharma (2004) studied the stock markets of the

7
five ASEAN countries, namely Indonesia, Malaysia, Singapore, Philippines and

Thailand, and their relationship with select macroeconomic variables. They found that

in the long-run, the stock prices were positively related to growth in output, and

negatively to the aggregate price level. A negative long-run relationship between stock

prices and interest rates was observed in Philippines, Singapore and Thailand. High

inflation in Indonesia and Philippines was found to influence the long-run negative

relation between stock prices and the money supply, while the money growth in

Malaysia, Singapore and Thailand was found to be responsible for the positive effect on

their stock markets. Lastly, the exchange rate variable was positively related to stock

prices in Indonesia, Malaysia, and Philippines, which can be explained by the high

competition in the world exporting market.

Mookerjee and Yu (1997) and Maysami and Koh (2000) studied the Singapore

stock market and found that changes in Singapore’s stock market levels form a

cointegrating relationship with changes in price levels, money supply, short- and long-

term interest rates. Gunasekarage et al. (2004) examined the long-run relationship

between macroeconomic factors and all-share price index from 1985-2001 for the

Colombo Stock Exchange and found that the consumer price index and treasury-bill rate

coefficients are significant and negative, money supply coefficient is significant and

positive, but exchange rate had no influence on share price index. Frimpong (2009)

conducted a study on Ghana for the period 1990-2006 and found that exchange rates are

positively related to the Ghana Stock Exchange All-Share index (GSE), and inflation

and money supply are negatively related to the GSE.

8
Macroeconomic Forces and the DSI: Hypothesized Relations

This section covers the macroeconomic variables chosen for the study and their

hypothesized relationships with the DSI. The variables chosen were based on financial

theory and established literature - deposit interest rates (IR), exchange rate (ER),

consumer price index (CPI), crude oil prices (OP) and broad money supply (M2)6.

The intuition behind the relationship between deposit interest rates and stock

prices forms the basis for the hypothesized negative relationship between the variables.

Interest rates represent the opportunity cost for investors in the equity markets (Asprem,

1989). An increase in the interest rates results in high opportunity cost of holding cash

and leads investors to substitute between stocks and other interest-bearing securities.

Moreover, the interest rates, through their effect on the risk-free rate, will cause an

increase in the discount rate (Mukherjee and Naka, 1995). Thus, stock prices are

expected to fall and vice versa.

Solnik (1987), Soenen and Hennigar (1988) and Ma and Kao (1990), among

others, indicate that exchange rates play a significant role in affecting the performance

of a stock market. For this study, a positive relation is hypothesized between exchange

rate (against the U.S. dollar)7 and stock prices based on the classic theory of Hume

(1752). As goods in the Bangladesh economy become relatively more expensive in the

international market due to an appreciation of the Bangladeshi Taka (falling exchange

rate) against the U.S. dollar, demand for exports reduce and, at the same time, demand

for imports increase, thus leading to lower Taka-denominated cash flows into the

Bangladeshi companies and hence, lower stock prices. This is evident from the

6
Amount of money in circulation in the economy
7
A weighted average exchange rate against the U.S. Dollar was compiled (and not against a
basket of currencies) since most international transactions are conducted in U.S. Dollar

9
theoretical model of stock valuation. The opposite should hold when the Bangladeshi

Taka depreciates against the U.S. Dollar.

The relation between consumer price index and stock returns has been generally

theorized to be negative (Fama and Schwert, 1977). A fall in consumer price index

lowers the nominal risk-free rate and decreases the discount rate in the stock valuation

model, leading to higher stock prices. Mukherjee and Naka (1995) suggest that the

effect of a lower discount rate would be neutralized if cash flows decrease with the CPI.

DeFina (1991), however, documents that cash flows do not decrease at the same rate as

inflation or CPI, and, hence, it is expected that the fall in discount rate will lead to

higher stock prices. It must be noted though, that prices, in general, may be subject to

greater fluctuations in the developing countries, which may render the relationship

insignificant. Thus, a negative or insignificant relationship is expected between CPI and

stock prices.

Crude oil prices are used in this study following Hassan and Hisham (2010). Oil

prices serve as an input for production in sectors, such as agriculture and manufacturing,

and thereby account for real economic activities taking place in the country. A high oil

price will result in reduced economic activities, and lower expected cash flows and

stock prices. Hence, a negative relationship is hypothesized between oil prices and stock

prices. The opposite relationship is expected to hold as well. It should be noted that

crude oil price is an external factor - the objective is to see whether international factors

play a role in the stock markets of Bangladesh.

The relationship between money supply and stock prices is not straightforward

because changes in money supply have important direct effects on stock prices via

portfolio changes, and indirect effects via its effects on real activity variables

(Mookerjee and Yu, 1987). As money growth rate is likely to be positively related to

inflation, it will increase the discount rate and, hence, lower stock prices. However,

10
since prices are considered constant in this study, the effect of money supply on stock

prices may be through other mechanisms. Sellin (2001) argues that a positive money

supply shock will alter expectations about future monetary policy and lead people to

anticipate tightening monetary policy in the future. The subsequent increase in bidding

for bonds will drive up the current rate of interest. As the interest rate goes up, the

discount rate increases, and the present value of future earnings decline, leading to a fall

in stock prices. The increase in money supply may also lead to a boost in companies’

cash flows resulting from the increased money supply (Mukherjee and Naka, 1995;

Chaudhuri and Smiles, 2004), known as corporate earnings effect, which is likely to

increase stock prices. For this study, a negative relationship is expected between money

supply and stock prices since prices and interest rates are subject to greater fluctuations

in developing countries.

11
Data

4.1 Data Sources

For the purpose of this paper, monthly data has been collected for the period

January 1992 to June 2011. The period was chosen purposefully since the Bangladesh

economy has undergone major changes during this period, such as trade liberalisation in

the 1990s, capital market developments in the 2000s, etc., and it will be interesting to

analyse the relationship between the macroeconomic variables and stock index during

this period. Firstly, the monthly Dhaka Stock Exchange All-Share Price index data is

obtained from the Dhaka Stock Exchange and its website8. Next, five macroeconomic

variables have been chosen. These include the deposit interest rate, the exchange rate

(domestic currency for US dollar), consumer price index (with a base year of 2005), per

barrel price of crude oil in U.S. Dollar, and broad money supply in local currency. Data

on consumer price index, exchange rate, deposit interest rates and crude oil prices were

collected from the International Financial Statistics of the International Monetary Fund.

Data on broad money was collected from the Monthly Trends publications of the

Bangladesh Bank. Other variables were also considered for the study initially, such as

call money rate and industrial production index, but due to unreliability and

unavailability of data, they had to be excluded from the study.

8
www.dsebd.org

12
4.2 Data Statistics

This sub-section provides descriptive statistics and time-plots (attached in Figure

A-1a:f in the Appendix) for the data under study. The purpose is to observe the trends

that the variables have displayed over the period and analyse the changes that have

taken place. For the sake of the study, all the variables (except interest rates9) have been

converted into natural logarithms10. The following table gives a summary of the

descriptive statistics of the variables:

Table 0:1: Descriptive Statistics of Variables - January 1992 to June 2011

Summary Statistics – Logged Variables (except IR)

Variable DSI IR ER CPI OP M2

Mean 6.99 8.22 3.98 4.42 3.44 13.75

St. Dev. 0.76 1.21 0.22 0.32 0.65 0.80

Maximum 8.87 11.39 4.30 5.05 4.89 15.30

Minimum 5.66 5.77 3.66 3.93 2.34 12.47

Note: DSI is Dhaka Stock Exchange All-Share Price Index, IR is deposit interest rate,
ER is exchange rate, CPI is consumer price index, OP is oil prices and M2 is broad
money. All the variables (except interest rates) are in natural logarithms
Source: Dhaka Stock Exchange, Bangladesh Bank and International Financial Statistics

The time plot in Figure A-1a shows that in the span of 1992M1-2011M6, the

DSI has registered an upward trend. The DSI series shows spikes in 1996 and 2010,

both were due to bubbles11 in the stocks.

The deposit interest rates were fairly stable in the period under consideration,

with low standard deviation, as seen in Table 4.1 and Figure A-1b. The deposit interest

9
Interest rates were kept in their original forms due to the nature of the data
10
Natural logarithms are taken as they stabilize variance, handle covariates which are only
positive, and make a symmetric distribution so coefficients are not influenced by extreme values
11
A bubble means that the prices of the stocks were above their fundamental values

13
rates were lower during the periods of the stock market crashes, as banks were forced

then to lower the interest rates that they pay out on deposits to customers.

The exchange rate of Bangladeshi Taka against the U.S. Dollar has been on an

upward trend for the entire period, as seen in Figure A-1c. The Bangladesh economy is

highly reliant on imports for luxury products and raw materials. Since these transactions

are conducted in U.S. Dollars, the exchange rate of the Bangladeshi Taka against the

U.S. Dollar has been rising. However, the appreciation of the U.S. Dollar against the

Bangladeshi Taka has ceased since the global financial crisis in 2006-2007, as

transactions in U.S. Dollar have reduced.

The consumer price index, which is taken to account for inflation12, has risen

steadily over the entire period, as seen in Figure A-1d.

The crude oil prices data in Figure A-1e show that the prices have remained

mostly stable until 2006. Since then, crude oil prices have seen major fluctuations with

record-high prices during the recent global financial crisis. The prices were lower in

2008; a stronger US dollar in the international market and a decline in European

demand are among the likely causes of the decline.

The broad money supply data in Figure A-1f shows that M2 has steadily increased

for the entire period except for two shocks in 1996 and 2006. However, the overall trend

in the broad money data has been upward.

12
Precisely, difference of log CPI is inflation

14
Econometric Methodology

The purpose of the study is to determine if a long-run relationship exists

between the DSI and macroeconomic factors for Bangladesh. The econometric model to

be used for the paper is as follows:

ln DSIt  0  1IRt   2 ln ERt  3 ln CPI t   4 ln OPt  5 ln M 2t   t

where the variables are as they have already been defined, and εt is the error term in the

model. β0 represents the constant term in the model and β1, β2, β3, β4 and β5 represent

long-run parameters.

Time-series econometrics requires an analysis of the time-series properties and

paths of the economic variables in a regression equation before estimation in order to

assess if a long-run relationship can be estimated for the model. A long-run relationship

exists if the variables are non-stationary in levels and stationary in first differences.

More specifically, it should be ensured that the variables in the study are integrated of

order d, where d≥1, i.e., they should be stationary in differenced forms, denoted as I(d).

Unit root, or non-stationarity of the variables, in the time series involved is

important before proceeding further with the analysis. When testing for stationarity of

the variables involved, conventional unit root tests like Augmented Dickey Fuller

(1979) (ADF) and the Phillip-Perron (1988) tests, in the presence of structural breaks in

the time series of the variables, may present test statistics that are strongly misleading.

This may result in the inferences drawn from them to be highly inaccurate. For example

in the event of a standard ADF regression:

p
yt    t    j yt  j  yt 1   t
j 1

A deterministic level shift will cause the primary coefficient of concern in the ADF

regression, ρ, to be biased towards 1 while a change in the trend slope makes the

15
estimator tend to 1 in probability as the sample size increases. Hence, the ADF test may

indicate presence of unit root even when the time-series is stationary around the

deterministic break component. In fact, such flaws may be extended to other classes of

traditional unit root tests as well (Perron, 1989, 1997 and Zivot and Andrews, 1992,

Perron and Vogelsang, 1992).

Perron (1989) first proposed a solution to this problem, where he figured for

inclusion of an exogenous structural break at time Tb, the time of break is known a

priori. Perron used a modified Dickey-Fuller (DF) unit root tests which included

dummy variables to account for one known, or exogenous structural break. The break

point of the trend function is fixed (exogenous) and chosen independently of the data.

Perron’s (1989) unit root tests allowed for a structural break under both the null and

alternative hypothesis. However the usage of an exogenous break was criticized and

thus Zivot and Andrews (1992), and Perron and Vogelsang (1992) both formulated unit

root tests where the break was endogenously derived. Of these two, however, Zivot and

Andrews (1992) allowed the structural break not under the null hypothesis of a unit root

but only under the alternative. This is thus a very undesirable feature, and subsequently

Vogelsang and Perron (1998) showed that if a unit root exists and a break occurs in the

trend function the Zivot and Andrews test will either diverge or will not be invariant to

the break parameters (Haldrup. et al., 2012).

Essentially these classes of tests follow in the methodology first established by

Perron (1989), and the following equations are estimated:

16
 Where the intercept dummy DUt represents a change in the level; DUt =1 if (t >

TB) and zero otherwise; the slope dummy DTt (also DTt*) represents a change in the

slope of the trend function; DT* = t-TB (or DTt*= t if t > TB) and zero otherwise; the

crash dummy (DTB) = 1 if t = TB +1, and zero otherwise; and TB is the break date.

Each of the three models has a unit root with a break under the null hypothesis, as the

dummy variables are incorporated in the regression under the null. The alternative

hypothesis is a broken trend stationary process.

However, Perron had suggested that most economic time series may be

adequately modeled using either model A or model C. Hence, subsequent literature has

primarily applied model A and / or model C (Waheed et al., 2006).

Although Perron and Vogelsang (1992) adopted a similar methodology, they had

tailored the tests for usage on raw (non-trending) data. This was improved subsequently

by Perron (1997), where the methodology was updated to test for unit root in trending

data series. Hence, we apply the unit root tests as derived by Perron (1997)

Structural Breaks in time series data by a shock occur either instantaneously or

gradually. Instantaneous change to the new trend function is modeled in Additive

Outlier (AO) model and the change that takes place gradually is modeled in

Innovational Outlier (IO) model. From the viewpoint of this paper, it is reasonable to

follow the IO model, as policy reforms at macro level do not cause the target variable to

respond instantaneously to the policy actions.

17
 Thus following Perron (1997), we ran an IO model to test for stationarity under

presence of structural break:

It may be seen that the preceding model conforms to Model C from the three

equations originally outlined by Perron (1989). As stated, in our tests we accounted for

structural breaks in both the slope and intercept. This unit root test was ran under the

Eviews add-in program ‘ppuroot’ which after having been provided with a maximal lag

length (usually either 4 or 12, based on the frequency being either based on quarterly or

yearly intervals), internally derives the optimal lag length for the test.

After testing the variables for unit root, the next step entails determining if

cointegration exists among the variables. Engle and Granger (1987) suggest that a long-

term equilibrium relation between stock prices and macroeconomic factors can be

determined using cointegration analysis. If two or more series individually have unit

root, but some linear combination of them has a stationary process, then the series is

said to be cointegrated. The Johansen (1991) method is an extension of Engle and

Granger procedure, allowing for more than one cointegrating equation and it this

procedure which will be undertaken. Suppose for a multivariate case,

Yt    1Yt 1   2Yt 2  ut

where Yt is a vector of k variables, i = 1,2 Θi is k x k. This can be manipulated and

written as:

Yt     2 Yt 1   k Yt 1  ut

where ∆Yt = Yt – Yt-1, and ∏k = Θ1 + Θ2 - Ik. If the rank of ∏k is zero, then there are no

cointegrating vectors. In the presence of cointegration, ∏k has rank r ≤ k – 1, and then,

∏k = αβ’, where α is k x r and β’ is r x k. Then, this can be written as:

Yt    2Yt 1   ' Yt u t

18
where β’ is the cointegrating matrix, β’Yt represents the r linear combination and α

represents the speed of adjustment towards the long-run equilibrium relationship. In

order to perform Johansen tests, we need to compute the k eigenvalues of ̂ k , which is

the estimate of ∏k. It is assumed that ˆ1  ˆ2  ...  ˆk is the squared canonical

relationship ordered from the largest to the smallest. If there are r cointegrating

relationships, then log(1 + λj) = 0 for j=r+1,…,k. Test for H0 : r ≤ r0 versus HA : r > r0,

i.e. under the null, the number of cointegrating vector is at most r0, under the alternative,

it is larger than r0.

k
trace(r0 )  T  log(1  ˆ j )
j r0 1

This is called a Trace test. The maximum eigenvalue test is also conducted to test for

the number of cointegrating relationships. Under the maximum eigenvalue test,

H0 : r ≤ r0 versus HA : r = r0 + 1, i.e. under the null, the number of cointegrating vector

is at most r0, under the alternative, it is equal to r0 + 1.

max (r0 )  T log(1  ˆr0  1)

If at least one cointegrating relationship exists among the variables, a causal relationship

among them can be determined by estimating the Vector Error Correction Model

(VECM). In this study, the relevant short-run VECM equation with a lag length p is

modeled as:

p p p p
 ln DSI t   0  1t   ln DSI t i   2t  IRt i   3t   ln ERt i   4t   ln CPI t i
i 1 i 1 i 1 i 1
p p
  5t   ln OPt i   6t   ln M 2 t i   t 1  1t
i 1 i 1

where the variables are I(1) and as previously defined, α1, α2, α3, α4, α5 and α6 represent

short-run elasticities, εt-1 is the error correction term, with its coefficient φ, which

conveys the long-run information contained in the data and denotes the speed of

19
adjustment to long-run equilibrium after a shock to the system. The VECM builds on

cointegration by incorporating error correction terms that account for short-run

dynamics, and, if a long-run equilibrium condition is valid and cointegration exists, it

explains short-run fluctuations (as represented by the α1, …. α6) in the dependent variable

(Frimpong, 2009). The optimal number of lags is determined by lag length in VAR

using AIC.

Impulse Response Functions (IRF) and Variance Decompositions (VDC) will be

constructed after estimating the VECM. IRF is a useful tool for characterizing the

dynamic responses implied by estimated VECM. Consider a first-order VAR for the n-

vector yt:

yt    Ay t 1   t

where μ is the vector of intercepts and εt ~ IN (0,  ). The IRF of a shock to a variable,

for instance, IR, on variable DSI after k periods is:

IRF ( DSI , IR, k )  [ E ( y DSI,t |  IR,t k  1,  \ IR,t k  0,  t k 1  0,....,  t  0]

 E[ y DSI,t |  IR,t k  0,  \ IR,t k  0,  t k 1.  0,...,  t  0]

where ε\IR,t is the vector εt excluding the IR element. The IRF measures the effect of a

shock of 1 unit occurring at period t-k in say, IR, or Deposit Interest Rate, on DSI, k

periods later, assuming there are no other shocks at period t-k, or in the other

intervening periods (t–k+1,…t). The IRF shows impulse responses of the select variable

in the VECM system in regards to the time paths of the variable’s own error shock

against the error shocks to other variables in the system. Since the innovations of error

terms are likely to be contemporaneously correlated, a mechanism of Impulse Response

via the Generalized Impulse Response method is implemented, to ensure

orthogonalization of the innovations, so as to ensure that interpretation of an impulse

response is straightforward. This helps to ensure to rule out contemporaneous

correlations amongst the error terms, the presence of which would imply that a shock to

20
one variable is likely to be accompanied by shocks to some of the other variables.

Unlike other mechanisms of orthogonalization of innovations where the interpretation

of specific impulses rests on the ordering of variables within the VAR system, the

Generalized Impulse Response has no such concern on the VAR ordering. The VDC is

implemented to show the percentage of the movement of the t-step ahead forecast error

variance of the select variable in the VECM system that is attributed to its own error

shock in contrast to error shocks to other variables in the system (Gunasekarage, 2004).

21
Empirical Results

6.1 Results of Unit Root Tests

It is essential to confirm the order of integration of the variables before the

model is estimated and tested for cointegration. As mentioned earlier, Perron’s (1997)

Innovative Outlier unit root test was utilized for this and was conducted for all the

variables with structural breaks allowed for both in intercept and time trend, with the

optimal lag length chosen by the test program on Eviews itself. The null hypothesis for

this test implies presence of nonstationarity in the presence of structural break in both

intercept and trend. Where the test statistic for the variable was greater than the critical

value for the test, the null of unit root was rejected and vice versa.

Table 0:1: Results of Unit Root Tests

Variable Perron's (1997) Unit Root Test

Levels
Test-Statistic Break Point
Log(DSI) -5.427 1996M10
IR -3.686 2004M04
Log(ER) -3.404 2004M11
Log(CPI) -4.751 2001M01
Log(OP) -4.139 1999M02
Log(M2) -2.602 2003M06
First Differences
Log(∆DSI) -15.223*** 1996M10
∆IR -6.963*** 2008M06
Log(∆ER) -15.096*** 2006M03
Log(∆CPI) -10.189*** 1995M03
Log(∆OP) -12.377 *** 2008M06
Log(∆M2) -15.283*** 2006M06
The critical values for t-statistics at 1% and 5% significance levels this test (inclusive of
structural break in both intercept and slope under null hypothesis) is -6.32 and -5.59.
 *** and ** denote the rejection of unit root/non-stationarity for the Perron
(1997) unit root test at 1% and 5% significance levels respectively

22
 The variables show non-stationarity in levels and stationarity in first differences,

after allowing for structural breaks in both the intercept and the slope. Furthermore,

since the primary objective is to see whether the null hypothesis of unit root in presence

of structural breaks is violated, the column which presents the derived structural break

events do not merit elaboration. As can be observed from Table 6.1 above, all the

variables are non-stationary in levels, and stationary in first differences, a necessary pre-

condition for cointegration analysis.

6.2 Results of Optimum Lag Length Tests

In choosing the specification of the cointegration model, it is necessary to

specify the number of lags in the autoregressive specification (Chaudhuri and Smiles,

2004). For this purpose, the Likelihood Ratio, Final Prediction Error, Akaike, Schwarz

and Hannan-Quinn Information Criterion (HQIC) were used to determine the

appropriate lag length. The AIC, SIC and HQIC are chosen based on lowest values over

the lags considered (allowed for a maximum of ten lags in this case). The Akaike

criterion suggests that a lag of five is optimal, whereas the Schwarz criterion indicates a

lag of one. Since the number of observations considered in the study is 234, i.e. below

250, the AIC is a better fit for the model. With a small sample such as the one

considered for this study, the SIC may choose too small a model, and it is a bigger

mistake to select too few lags (dynamic misspecification) than to select too many, which

is why the optimal lag length is chosen by the AIC.

23
 Table 0:2: VECM Lag Order Selection Criteria

Lag LR FPE AIC SIC HQIC

0 4.6-9 -2.16 -2.07 -2.12
1 4374.6 2.1-17 -21.37 -20.73* -21.11
2 123.86 1.7-17 -21.60 -20.41 -21.12*
3 66.87 1.7-17 -21.58 -19.84 -20.88
4 74.43 1.7-17 -21.59 -19.30 -20.67
5 86.98 1.6-17* -21.66* -18.82 -20.51
6 57.24 1.7-17 -21.59 -18.21 -20.22
7 75.44 1.7-17 -21.60 -17.68 -20.02
8 61.34* 1.8-17 -21.56 -17.08 -19.75
9 38.71 2.2-17 -21.41 -16.38 -19.38
-17
10 44.92 2.5 -21.29 -15.71 -19.04
*indicates lag order selected by the criterion

6.3 Results of Johansen Cointegration Tests

Table 6.3 shows the results for the Johansen Cointegration test performed to

investigate the long-run relationships of the variables in the model. However, the

number of cointegrating vectors generated by the Johansen test may be sensitive to the

lag length. Hence, the optimum lag length estimated in the previous section via AIC

(five) is used to determine the number of cointegrating relations.

However, a question also arises regarding the proper assumption of trends in the

Johansen Cointegration framework. Here, since the most of the variables in question

come to exhibit some form of trending, we select the trend assumption as constant,

wherein we exclude the possibility that the levels of the data have quadratic trends, and

we restrict the cointegrating equations to be stationary around constant means. However

this specification still puts a linear time trend in the levels of the data. (Stata Time

Series Reference Manual, Release 11, 2009).

24
 Table 0:3: Results for Johansen Cointegration Test

No. of λmax 95% Critical 99% Critical λtrace 95% Critical 99% Critical
CE(s) [H0] Statistic Value [Max.] Value [Max.] Statistic Value [Trace] Value [Trace]
r=0 41.58 39.37 45.10 104.51 94.15 103.18
r≤1 32.59 33.46 38.77 62.93* 68.52 76.07
r≤2 14.57 27.07 32.24 30.35 47.21 54.46
r≤3 10.35 20.97 25.52 15.78 29.68 35.65
r≤4 5.36 14.07 18.63 5.42 15.41 20.04
r≤5 0.06 3.76 6.65 0.06 3.76 6.65
 r denotes the number of cointegrating relationships
 CE refers to cointegrating equations
 * indicates that the number of CE selected by the estimator correspond to this row

The first column in Table 6.3 shows the null hypothesis assumed for the

Maximum Eigenvalue and Trace Tests. The value of λtrace under the null of r = 0 (no

cointegration) is 104.51, which is greater than 94.15, the 5% critical value reported from

Osterwald-Lenum (1992), so the null of no cointegration can be rejected in support of

one cointegrating equation. For r ≤ 1, the λtrace statistic is less than the critical value at

1% and 5% significance levels, which forms the basis for accepting the null hypothesis

of at most one cointegrating vector. An alternative measure that is used to determine the

number of cointegrating vectors is λmax. The λmax shows that at the 5% significance level,

the null hypothesis of no cointegrating vector is rejected since the value of 41.58 is

greater than 39.37. Similar to the λtrace statistic, for r ≤ 1 and other values of r, the λmax

measure is less than the critical value at 5% significance level. Therefore, it can be

assumed that there is at least one cointegrating vector. According to both λtrace and λmax

statistics, it can be confirmed that there is at least one long-run equilibrium relationship

between the Dhaka Stock Exchange All-Share Price Index and macroeconomic

variables. Lags of six and seven were considered to check for robustness; they also

indicated one cointegrating relationship.

25
6.4 Results of Long-Run Cointegration Model

Table 6.4 shows the long-run cointegrating model. The long-run relationships
were as hypothesized in the study and these are reported below:
Table 0:4: Long Run Cointegrating Model

Regressor Coefficient Std. Error z-statistics

Constant 1.39 - -
Interest Rate 0.12 0.07 -1.79*
Exchange Rate -5.26 1.33 3.94***
Consumer Price Index -3.01 2.26 1.33
Crude Oil Prices 0.70 0.21 -3.28***
Broad Money Supply 2.66 1.19 -2.23**
***, ** and * denote significance of variables at the 1%, 5% and 10% levels
respectively. Note: Dependent Variable - DSI

According to the table, the actual long-run relationship can be represented by:

DSI - 0.12IR + 5.26ER + 3.01CPI - 0.69OP - 2.66M2 - 1.39 = εi,t (1)

Which, after rearranging becomes:

DSI = 0.12IR - 5.26ER - 3.01CPI + 0.69OP + 2.66M2 + 1.39 + εi,t (2)

As may be inferred from the estimates, the model posits that interest rates and

DSI index is positively related and impact of interest rates is statistically significant at

10%. This was unexpected as, theoretically, high deposit interest rates lead investors to

invest less in risky assets and, consequently, lower stock prices are expected. However,

this converse result is not uncommon according to the literature. Mukherjee and Naka

(1995), Maysami and Koh (2000), and Bulmash and Trivoli (1991) found a positive

relation between the short-term interest rates and stock market prices, and a negative

relationship between long-term interest rates and stock prices. The relationship between

deposit interest rates and stock prices in this study are, therefore, consistent with the

results of the short-term interest rates. One possible explanation for this is that if the

interest rates is increased now, it means that they will fall in the future. When investors

26
find that interest rates have increased, they buy more stocks now since they know that

lower interest rates in the future will result in even higher stock prices then, hence,

higher capital gains for investors. Higher demand for stocks now leads to higher current

stock prices.

The exchange rate and the DSI are significantly and negatively related in the

long-run. This was not hypothesized since, theoretically, it is expected that increasing

exchange rates (Taka depreciation against the U.S. Dollar) will result in money inflows,

and, consequently higher investment in the DSE stocks, and, higher DSI. Maysami and

Koh (2000) report the same relation as in this study. Maysami and Koh (2000) found

that the Singapore Dollar exchange rate (against the U.S. Dollar) and the Singapore

stock market are negatively related. They state that an appreciation of the Singaporean

Dollar lowers imported inputs and allows the exporters in the country to be more

competitive internationally. This is received as favourable news for the Singapore stock

markets and, hence, positive stock returns are generated as a result. Perhaps, this also

holds true for the Bangladesh exchange rate. Depreciation in Taka, instead of resulting

in money inflows, results in increased imported materials leading to the exports being

uncompetitive in the world economy. Consequently, this will lower stock demand and,

hence, lower stock prices. Mukherjee and Naka (1995) and Brown and Otsuki (1990)

report a positive relation for the Japanese stock market. Gunasekarage (2004) found that

exchange rates have no significant relationship with the Colombo stock prices. This was

due to limited participation by the foreign investors.

It was hypothesized that the relationship between CPI and the DSI in the long-

run will be either negative or insignificant due to large price changes. Long-run

cointegrating model shows that the relationship between CPI and stock prices is

negative. The relationship, though, was statistically insignificant, which confirms that

large price changes in Bangladesh affect the theorized relationship. Price fluctuations in

27
developing countries are more prevalent due to lower regulations, competition, etc. and

explain why the relationship was found insignificant. The relationship found here is

consistent with evidence in the literature - Lintner (1973), Oudet (1973), Bodie (1976),

Nelson (1976), Mukherjee and Naka (1995) and Gunasekerage (2004) found a negative

relationship between CPI and stock prices.

The relationship between crude oil prices and the DSI was found to be positive

in the long-run, contrary to the hypothesis. Chen et al. (1986) found an insignificant

relationship between oil prices and the NYSE. Hassan and Hisham (2010) found a

negative relationship between crude oil prices and the Jordan Stock Exchange.

However, this is consistent with evidence found recently. It has been recently seen that

crude oil prices and stock prices are positively related. For e.g., the Standard & Poor’s

(S&P) 500 Index and the oil prices from 1998-2008 have demonstrated a positive

relationship with each other with a correlation of 0.55, and 0.86 since 2008 (Smirnov,

2012). For the DSI, the relationship was positive with oil prices and consistent with

recent results, though, inconsistent with theory. This is expected as the DSI is likely to

be strongly influenced by other stock price indices, such as S&P 500 index, etc.

The relationship between money supply and stock prices was found to be

positive and significant at 5% level in the long-run. This is consistent with most studies

in the literature, such as Bulmash and Trivoli (1991), Mukherjee and Naka (1995) and

Gunesekarage (2004). This means that the corporate earnings effect plays a strong role

for the Bangladesh economy. An increase in money supply leads to a boost in

companies’ cash flows, which results in increased stock prices. This result is

inconsistent with Frimpong (2009), who found a negative relationship between the

Ghana stock prices and money supply.

28
6.5 Results of Error Correction Mechanism:

Table 6.5 below reports the short-run results of the Vector Error Correction

Model regarding the short run dynamics of DSI, the stock price index. The sign and

magnitude of the error correction coefficient (adjustment coefficient) indicate the

direction and speed of adjustment towards the long-run equilibrium path. A negative

error correction coefficient implies that in the event of a positive deviation of the model

from the long-run equilibrium, in the absence of variation in the independent variables,

which implies that the value of DSI is above its equilibrium, is corrected by changes in

the dependent variable. This confirms the existence of a long–run relationship. The size

of the coefficient of the error correction term in this study implies that about 5.3% of the

disequilibrium in the long-run relationship is corrected every month as DSI settles back

into its equilibrium value. The error term coefficient was significant at the 5% level.

The short-run results indicate that DSI positively affects the DSI at the first lag;

the results from the latter lags are insignificant. CPI positively affects the stock prices at

the third lag, but it was insignificant for other lags. Interest rates, exchange rates, oil

prices and money supply mostly affect the DSI negatively, but these variables are also

statistically insignificant at most lags. Other lags were also considered for robustness

and better results. Lags of four and seven revealed a negative sign for the error

correction term but it was not significant. The other variables, similarly, did not result in

more significant or robust estimates. Impulse Response Functions are employed later in

an attempt to explain the short-run results better.

29
 Table 0:5: Vector Error Correction Model

Regressors Coefficient Standard Error z-statistics

Speed of Adjustment -0.053 0.022 -2.42**
∆DSIt-1 0.172 0.071 2.44**
∆DSIt-2 -0.012 0.071 -0.17
∆DSIt-3 0.110 0.071 1.55
∆DSIt-4 -0.002 0.072 -0.02
∆IRt-1 -0.025 0.024 -1.03
∆IRt-2 0.028 0.024 1.15
∆IRt-3 -0.023 0.025 -0.92
∆IRt-4 0.017 0.025 0.65
∆ERt-1 -0.968 0.924 -1.05
∆ERt-2 0.093 0.935 0.10
∆ERt-3 -0.576 0.937 -0.61
∆ERt-4 -0.708 0.958 -0.74
∆CPIt-1 -0.172 0.883 -0.20
∆CPIt-2 0.247 0.895 0.28
∆CPIt-3 1.803 0.903 2.00**
∆CPIt-4 -0.020 0.889 -0.02
∆OPt-1 0.090 0.087 1.04
∆OPt-2 0.074 0.088 0.84
∆OPt-3 -0.055 0.088 -0.62
∆OPt-4 -0.023 0.087 -0.26
∆M2t-1 -0.186 0.282 -0.66
∆M2t-2 -0.026 0.319 -0.08
∆M2t-3 0.026 0.308 0.08
∆M2t-4 0.045 0.259 0.17
Constant 0.001 0.015 0.05

30
The following table below details the short run adjustment coefficients for the other

variables in question:

Table 6.6 : Adjustment Coefficients of the other Variables

Variables Coefficients Standard Error Z-Stat

∆IR 0.222 0.062 3.57***

∆ER 0.001 0.002 0.58

∆CPI 0.006 0.002 3.73 ***

∆OP 0.044 0.017 2.53**

∆M2 0.003 0.005 0.67

With respect to the original signs of the coefficients of the variables in the long run

relationship (as shown in Equation 1), it may be inferred here that the adjustment

coefficients mostly point to the validity of the cointegrating relationship between DSI

and the other variables. For interest rate, given its negative sign in Equation 1, a positive

unit disequilibrium from the long run relationship (an unit increase in DSI) from last

month results in interest rate adjusting back to equilibrium by 22% correction of the one

unit disequilibrium in the current month (at the same time as DSI is adjusting).

However, adjustment coefficients for the other variables are minuscule in magnitude,

with only CPI and Oil prices possessing statistically significant adjustment coefficients.

The positive adjustment coefficient for CPI while being small, is statistically significant,

and given its negative sign in the long run model (Equation 1), the estimate for CPI

instead points to steady divergence from equilibrium in the event of disequilibrium from

the long run relationship. However, since the long run estimate of CPI was statistically

insignificant, formal and meaningful interpretations may have to be withheld regarding

its dynamics.

31
 For a more comprehensive picture regarding the short run dynamics of the

variables, interpretations of Impulse Response Functions would be more meaningful.

6.6 Diagnostic Tests for the VECM

For statistical accuracy and efficiency of the residuals in the VECM, diagnostic

tests are performed. The results for the diagnostic tests are attached in Table A-1. With

regard to the test for autocorrelated residuals, the p-value corresponding to Lagrange

Multiplier (LM) statistics is significantly higher than the 10% level of significance.

Hence, we can accept the null hypothesis of no autocorrelation in the residuals. The

Eigenvalue Stability Circle (ESC) reports stability of the residuals in the VECM. In the

case of a two-variable model with one lag, stability conditions are such that one of the

eigenvalues is equal to one, and the other is smaller than one in absolute terms. The

result reported in the ESC in Table A-1 is exact generalization of this condition to the

six-variable model with five lags. With regard to the Jarque-Bera test, we reject the null

hypothesis of normality since the p-value is 0.00. However, this will not significantly

distort results as, more importantly, the residuals are stable and non-autocorrelated.

6.7 Robustness of Results in the VECM

To check for robustness in the VECM estimation results, lags of six and seven

were considered for the long-run model. All the variables reported the same directional

relationship with the stock prices at six lags; however, oil prices and money supply were

insignificant in explaining stock price changes. Lags of seven reported the same

relationships between stock prices and the macroeconomic variables as the fifth lag, but

the relationship between money supply and stock prices was insignificant. This shows

32
that the relationships between stock prices and interest rates, exchange rates and

consumer price index were robust. However, money supply and oil prices were sensitive

to lag lengths and the estimates for these variables are not robust.

6.8 Results of Impulse Response Functions and Variance Decompositions

The results for the Impulse Response Functions are reported in Figure A-2.

Impulse Response Analysis entails analysing the incremental impact of a shock to the

whole system. Hence, we are primarily concerned about the behaviour of the short-run

dynamics of the model in the presence of any external shock emanating from any of the

variables particularly on DSI. Figure A-2 shows the effect of macroeconomic shocks on

the DSI. Four years of data are forecasted and their effects plotted via Impulse Response

Functions.

An interest rate shock of one Generalized Standard Deviation causes the DSI to

fluctuate initially, however, from the 15th month onwards, the effect from the shock

recedes and in the 30th month, the effect settles to an apparent permanent level of about

0.06 units over the baseline. An exchange rate shock results in a big dip in the DSI in

the initial periods, however with passage of time, this impact gradually withers away

and around the 30th period, the impact settles just above the baseline, implying that a

shock emanating from interest rates has no tangible or apparent long run effect on DSI,

which is in contrast to the relationship derived in the long run model earlier. With

regards to shock from exchange rate, the first 6 months witness a continuous slide in

DSE before settling at a permanent value of 0.04 units below the baseline. With regards

to a unit Generalized S.D. shock from CPI, the impact lessens after the fifth month,

before becoming negative after the 15th momth, and at the end of the time-span (4 years;

48 months) the impact settles permanently on a constant value below the baseline. Oil

33
prices shock lead to a large increase in the DSI initially and the effect increases with

time before again settling at a permanent level of 0.06 units over the baseline. Similarly

Money supply shock leads to a slight fall in the DSI in the first month, but then

increases the DSI over time, which supports the cash flow effect of money supply,

before settling in at a constant level after the 8th month. In general, most of the variables

seem to have a permanent effect on DSI in the long run as can be inferred from the

impulse response diagrams, which also reflects the dynamics of the long run relations

model derived earlier. However, only interest rates seemingly do not have a permanent

level of impact on DSI.

Our views are further reinforced from the inferences which we can draw from

the Cumulative Impulse response functions for DSI (Figure A-3, Appendix), which

cumulatively sums up impulse responses of a particular variable, in this case, DSI over

successive time periods, so as to glean the aggregate impact of a shock at the end of n-

periods. In keeping with the inferences drawn from the individual impulse response

functions as well as the long run relation model derived earlier, it can be seen that at the

end of the 4 year projected span, most of the variables’ cumulative impact over the 4

year span on DSI is either positive (for shocks from DSI, oil prices and Money Supply)

or negative (for exchange rate). The cumulative responses of DSI to interest rates and

CPI are more of a mixed bag, with the deposit interest rates, at the end of the 4 year old,

having a cumulative shock of just below the baseline of zero. This may appear at odds

with the coefficient for interest rates in the long run model, which suggested a positive

relationship between interest rates and DSI. However from the theoretical standpoint, it

merits justification, since here, we extract a discernible, negative, albeit small final

impact of interest rates on DSI, which is also predicted by theory. Similarly, for CPI, its

cumulative impact too seems to be indistinguishable from zero at the end of 4th year,

and furthermore, since the coefficient for CPI is statistically insignificant at 10 percent

34
in the long run model, this particular outcome maybe is line with earlier findings and

also has theoretical justifications as well (which were elaborated earlier).

Table 6.7 reports the Variance Decomposition test results. Twenty months of the

model are forecasted and the results indicate that most of the variations in the DSI are

explained by the DSI itself. In the 5th period, 94.58% of the variation in the DSI was

explained by shocks to itself, 2.09% by ER and 2.54% by OP. At the end of the 10th

period, 4.98% and 5.65% of the variations in the DSI were explained by shocks to ER

and OP respectively. The results obtained from VDC combined with IRF indicate that

the macroeconomic shocks from other variables explain a minority of the forecast error

variance in the DSI, though, exchange rates and oil prices have a significant influence

on the stock prices. It must be noted though that most of the shocks from the

macroeconomic variables are permanent and persist for a long period, as is to be

expected in conventional VEC system.

35
 Table 0.7: Variance Decomposition of DSI

Period Std. Error DSI IR ER CPI OP M2

1 0.099 100.000 0.000 0.000 0.000 0.000 0.000

2 0.151 99.032 0.199 0.336 0.015 0.414 0.005
3 0.188 97.691 0.129 0.595 0.011 1.553 0.020
4 0.223 96.372 0.108 1.059 0.249 2.127 0.086
5 0.253 94.578 0.085 2.098 0.443 2.541 0.255
6 0.281 92.929 0.069 2.985 0.513 3.063 0.441
7 0.305 91.460 0.064 3.680 0.511 3.617 0.667
8 0.326 89.925 0.066 4.206 0.476 4.292 1.034
9 0.346 88.560 0.068 4.613 0.442 4.988 1.328
10 0.363 87.356 0.073 4.976 0.411 5.650 1.535
11 0.380 86.291 0.078 5.239 0.382 6.299 1.711
12 0.395 85.270 0.083 5.479 0.355 6.934 1.879
13 0.410 84.255 0.085 5.732 0.330 7.546 2.053
14 0.423 83.282 0.084 5.977 0.310 8.136 2.210
15 0.436 82.348 0.082 6.211 0.294 8.714 2.351
16 0.448 81.449 0.079 6.417 0.282 9.279 2.493
17 0.459 80.579 0.076 6.605 0.273 9.833 2.633
18 0.470 79.731 0.073 6.784 0.268 10.375 2.769
19 0.481 78.910 0.069 6.949 0.265 10.908 2.898
20 0.491 78.109 0.067 7.102 0.265 11.434 3.022

36
6.9 Robustness of Results in the IRF and VDC

To test for robustness, a VAR model was constructed in first differences, and

IRF and VDC drawn from its estimates. The IRF demonstrated short-run results very

similar to the IRF from the VECM model. The results for the VDC from the VECM and

VAR models are also similar. A very low percentage of the changes in stock prices is

explained by the variables, though exchange rate explains a high proportion of the

changes in stock prices.

37
 7. Conclusion

The study investigates the long-term relations between macroeconomic variables

and the Dhaka stock market prices using Johansen’s methodology of multivariate

cointegration analysis and Vector Error Correction Model. The macroeconomic forces

were represented by interest rates, exchange rates, consumer price index, crude oil

prices and money supply while the Dhaka Stock Exchange All-Share Price Index was

used to represent changes in the Dhaka stock market prices.

The main findings revealed that there is a long-term relationship between the

stock prices and macroeconomic variables. According to the cointegration analysis and

the VECM estimated in the study, the stock prices and macroeconomic variables are

related significantly, though not in accordance with the hypothesized relationships. The

interest rate was positively related with the stock prices; this was unexpected as higher

interest rates, theoretically, shift investors away from stocks and vice versa. The

exchange rates are negatively related with the stock prices – meaning that Taka

depreciation leads to higher imported inputs and, hence, lower exports and lower stock

prices and vice versa. The consumer price index is found to be negative, but

insignificant in explaining the stock prices. This was hypothesized as large price

changes in Bangladesh may render an insignificant relationship between CPI and stock

prices. The relationship between crude oil prices and stock prices was found to be

positive and significant; this is consistent with recent results from other stock

exchanges. The broad money supply is positively related with the stock prices which

confirms that the corporate earnings effect leads to a boost in companies’ cash flows

and, hence, higher stock prices. The macroeconomic variables, except oil prices and

money supply, were robust to different lag lengths.

38
 The short-term results of the VECM revealed that around 5.3% of the

disequilibrium in the long-run model is corrected every month as DSI reverts back to its

equilibrium. However, the DSI and macroeconomic variables were insignificant at most

lags in the ECM specification for DSI. Since the VECM results were inconclusive, the

Impulse Response Functions and Variance Decompositions were undertaken. An

interest rate shock causes the DSI to fluctuate initially, but a tendency to converge to

equilibrium just above the baseline is seen. An exchange rate shock results in a large fall

in the DSI initially, but the gap remains steady afterwards, implying a persistent effect

on DSI. A consumer price index shock leads to an increase in the DSI initially, but then

the relationship becomes negative and also in the long run assumes a persistent or

constant negative effect on DSI. Oil prices shock lead to a large increase in the DSI

initially and the effect increases slowly over time before too settling on a permanent

impact level. Money supply shock increases the DSI over time, which supports the

corporate earnings effect, but settles to a constant level in the long run. Overall, the

results for the impulse response function seems to imply that most of the

macroeconomic variables have a permanent and long run impact on DSI, and thus

largely validates the results in the long run model. The cumulative impulse response

functions further affirm our findings, with important differences being the cumulative

impacts of deposit interest rates and CPI, the first of which was found to be small and

negative, the latter in accordance with economic theory, while CPI shows no discernible

cumulative impact on DSI after 4 years, a reflection of the statistical insignificance of

CPI in the long run model as well as economic theory. The Variance Decomposition

results indicate that the macroeconomic shocks emanating from variables apart from

DSI explain a small proportion of the forecast error variance in the DSI, though,

exchange rates and oil prices have a significant influence on the stock prices. Most of

the shocks from the macroeconomic variables to the DSI are permanent and are

39
persistent. Diagnostic tests in the VECM were performed to ensure that there was no

autocorrelation in the residuals and that they were stable.

In light of the analysis made in the study, policymakers and economists in

Bangladesh need to be careful when they try to influence the economy through changes

in key macroeconomic variables comprising the interest rates, exchange rates, consumer

price index and money supply. The important takeaway from this study is that

developing country macroeconomics or dynamics is different from developed country

macroeconomics. The theorized macroeconomic relationships with the stock prices do

not hold for Bangladesh mostly and this should be considered when policies are made.

Stock markets are inefficient and underdeveloped in developing countries and, hence,

central banks of these countries should be careful in designing its monetary policy.

Country-specific traits should be taken into account and research should be conducted

before any policy is implemented. They should also be aware of international factors

such as crude oil prices as these also have a significant impact on the economy. Further

research on the topic may also want to consider other macroeconomic variables such as

call money rate, long-term government bond rate and industrial production index, which

are important economic drivers and may significantly influence the stock prices. The

Bangladesh stock market is an established capital market and its development is crucial

for the growth of the country. Thus, the government and policymakers should aim to

influence the key macroeconomic variables in a way that ensures that stock prices are

stable and stock markets are performing in an efficient and effective manner.

40
References

Asprem, M., 1989, Stock Prices, Asset Portfolios, and Macroeconomic Variables in ten

European Countries, Journal of Banking and Finance 13, pp. 589-612

Bespoke Investment Group, 2010, 2010 Country Stock Market Performance,

http://www.bespokeinvest.com/thinkbig/2010/9/3/2010-country-stock-market-

performance.html, last accessed June 20, 2012

Bloomberg, 1997, Bangladesh Stock Market Lures Investors Hungry for Next Vietnam,

retrieved from

http://www.bloomberg.com/apps/news?pid=newsarchive&sid=ajYCq0jTs6Fo&re

fer=india, last accessed June 10, 2012

Bodhurta, J. N., Cho, D. C. and Senbet, L. W., 1989, Economic Forces and the Stock

Market: An International Perspective, The Global Finance Journal 1(1) pp. 21-46

Bodie, Z., 1976, Common stocks as a hedge against inflation, Journal of Finance 31,

pp. 459-470

Brown, S. J. and Otsuki, T., 1990, Macroeconomic factors and the Japanese equity

markets: The CAPMD Project, Japanese Capital Markets, Harper and Row: New

York

Bulmash, S. B. and Trivoli, G. W., 1991, Time-lagged interactions between stock prices

and selected economic variables, Journal of Portfolio Management pp. 61-67

41
Chaudhuri, K. and Smiles, S., 2004, Stock market and aggregate economic activity:

evidence from Australia, Applied Financial Economics 14, pp. 121-129

Cheng, A. C. S., 1995, The UK Stock Market and Economic Factors: A New Approach,

Journal of Business Finance and Accounting 22(1), pp. 129-142

Chen, N-F., Roll, R. and Ross, S. A., 1986, Economic forces and the stock market,

Journal of Business 59, 383-403

Cheung, Y-W, and Lai, K. S., 1999, Macroeconomic determinants of long-term stock

market comovements among major EMS countries, Applied Financial Economics

9(1), pp. 73-85

DeFina, R. H., 1991, Does inflation depress the stock market?, Business Review,

Federal Reserve Bank of Philadelphia, 3-12

Diacogiannis, G. P., 1986, Arbitrage Pricing Model: A Critical Examination of its

Empirical Applicability for the London Stock Exchange, Journal of Business

Finance and Accounting 13 (4), pp. 489-504

Dickey, D. A. and Fuller, W. A., 1979, Distribution of the Estimators for

Autoregressive Time Series with a Unit Root, Journal of the American Statistical

Association 74, pp. 427-431

42
Engle, R. F. and Granger, C. W. J., 1987, Co-integration and Error Correction:

Representation, Estimation, and Testing, Econometrica 55(2), pp. 251-76

Fama, E. F. and Gibbons, M., 1984, A comparison of inflation forecasts, Journal of

Monetary Economics 13, pp. 327-248

Fama, E. F. and Schwert, G. W., 1977, Asset Returns and Inflation, Journal of

Financial Economics 5, pp. 115-146

Flannery, M. J. and Protopapadakis, A. A., 2002, Macroeconomic Factor do influence

aggregate stock returns, The Review of Financial Studies 15(3), pp. 751-782

Frimpong, J. M., 2009, Economic Forces and the Stock Market in a Developing

Economy: Cointegration Evidence from Ghana, European Journal of Economics,

Finance and Administrative Sciences 16, pp. 123-135

Goldsmith, R. W., 1969, Financial Structure and Development, New Haven: Yale

University Press

Gunasekarage, A., Pisedtasalasai, A. and Power, D. M., 2004, Macro-economic

influences on the stock market: Evidence from an Emerging Market in South

Asia, Journal of Emerging Market Finance 3(3) 285-304

Günsel, N. and Çukur, S., 2007, The Effects of Macroeconomic Factors on the London

Stock Returns: A Sectoral Approach, International Research Journal of Finance

and Economics 10, pp. 140-152

43
Gurley, J. G. and Shaw, E. S., 1955, Financial Aspects of Economic Development,

American Economic Review 45, pp. 515-538

Gurley, J. G. and Shaw, E. S., 1960, Money in a Theory of Finance, Brookings

Institution, Washington D.C.

Gurley, J. G. and Shaw, E. S., 1967, Financial Development and Economic

Development, Economic Development and Cultural Change 15(3), pp. 257-268

Haldrup, N., Kruse, R. and Terasvirta, T., 2012, Unit Roots, Nonlinearities, and

Structural Breaks, Department of Economics, Aarhus University, and CREATES

Research Paper 2012-14

Harvey, C. R., 1995a, Predictable risk and returns in emerging markets, Review of

Financial Studies 8, 773-816

Hassan, G. and Hisham, A., 2010, Can Macroeconomic Factors Explain Equity Returns

in the Long Run? The Case of Jordan, MPRA Paper 22713, University Library of

Munich, Germany

Hume, D., 1752, Political Discourses, Edinburgh: A. Kincaid & A. Donaldson

Johansen, S., 1991, Estimation and hypothesis testing of cointegrating vectors in

Gaussian vector autoregressive models, Econometrica 59, 1551-80

44
LeRoy, S. and Porter, R. D., 1981, The Present-Value Relation: Tests based on Implied

Variance Bounds, Econometrica 49, pp. 555-574

Levine, R. and Zervos, S., 1996, Stock Market Development and Long-Run Growth,

The World Bank Economic Review 10(2), pp. 323-339

Levine, R., 2004, Finance and Growth: Theory and Evidence, National Bureau of

Economic Research, Working Paper 10766

Lintner, J., 1973, Inflation and common stock prices in a cyclical context, National

Bureau of Economic Research 53rd Annual Report, New York

Ma, C. K. and Kao, G. W., 1990, On Exchange Rate Changes and Stock Price

Reactions, Journal of Business Finance and Accounting 17, pp. 441-449

Maysami, R. C. and Koh, T. S., 2000, A Vector Error Correction Model of the

Singapore stock market. International Review of Economics and Finance 9, pp.

79-96

Mookerjee, R. and Yu, Q., 1997, Macroeconomic variables and stock prices in a small

open economy: The case of Singapore, Pacific-Basin Finance Journal 5, pp. 377-

388

Mukherjee, T. and Naka, A., 1995, Dynamic Relations between Macroeconomic

Variables and the Japanese Stock Market: An Application of a Vector Error

Correction Model, The Journal of Financial Research 18 (2) 223-237

45
Nasseh, A. and Strauss, J., 2000, Stock prices and domestic and international

macroeconomic activity: A cointegration approach, The Quarterly Review of

Economics and Finance 40, pp. 229-245

Nelson, C. R., 1976, Inflation and rates of return on common stocks, Journal of Finance

31, pp. 471-83

Osterwald-Lenum, M., 1992, A Note with Quantiles of the Asymptotic Distribution of

the Maximum Likelihood Cointegration Rank Test Statistic, Oxford Bulletin of

Economics and Statistics 54, pp. 461-472

Oudet, B. A., 1973, The variation of the returns on stock in periods of inflation, Journal

of Financial and Quantitative Analysis 8, pp. 247-58

Perron, P., 1989, Testing for a Unit Root in a Time Series with a Changing Mean,

Journal of Business & Economic Statistics, American Statistical Association 8(2),

pp. 153-162

Perron, P. and Vogelsang, T. J., 1992, Non-stationarity and Level Shifts with an

Application to Purchasing Power Parity, Journal of Business and Economic

Statistics 10, pp. 301-320

Perron, P., 1997, Further Evidence on Breaking Trend Functions in Macroeconomic

Variables, Journal of Econometrics 80(2), pp. 355-385

46
Phillips, P. C. B. and Perron, P., 1988, Testing for Unit Roots in Time Series

Regression, Biometrika 75, pp. 335-346

Poon, S. and Taylor, S. J., 1991, Macroeconomic Factors and the UK stock market, The

Journal of Business Finance and Accounting 18(5) pp. 619-636

Sellin, P., 2001, Monetary Policy and the Stock Market: Theory and Empirical

Evidence, Journal of Economic Surveys 15 (4), pp. 491-541

Shaw, E. S., 1973, Financial Deepening in Economic Development, New York, Oxford

University Press

Shiller, R. J., 1981, Do stock prices move too much to be justified by subsequent

changes in dividends?, The American Economic Review 71, pp. 421-436

Soenen, L. A. and Hennigar, E. S., 1988, An Analysis of Exchange Rates and Stock

Prices: the U.S. Experience between 1980 and 1986, Akron Business and

Economic Review 19(4), pp. 7-16

Solnik, B., 1987, Using Financial Prices to Test Exchange Rate Models: A Note,

Journal of Finance 42, 141-149

Smirnov, D., 2012, Price of Oil Versus the Stock Market, Seeking Alpha retrieved from

http://seekingalpha.com/article/542981-price-of-oil-versus-the-stock-market,

last accessed on July 5, 2012

47
StataCorp, Stata Time Series Reference Manual Release 11, College Station,

Stata Press, 2009.

Waheed, M., Alam, T. and Pervaiz, S., 2006, Structural Breaks and Unit Root: Evidence

from the Pakistani Macroeconomic Time Series, MRPA Paper No. 1797,

University of Munich Library Germany

Wongbangpo, P. and Sharma, S. C., 2002, Stock Market and Macroeconomic

Fundamental Dynamic Interactions: ASEAN-5 Countries, Journal of Asian

Economics 13, pp. 27-51

Zivot, E. and Andrews, K., 1992, Further Evidence on The Great Crash, The Oil Price

Shock, and The Unit Root Hypothesis, Journal of Business and Economic

Statistics 10(10), pp. 251-270

48
Appendix

LOG(DSI)
9.0

8.5

8.0

7.5

7.0

6.5

6.0

5.5
1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

Figure A-1a: Time-Series Plot of Log(DSI) in Levels

IR
12

11

10

5
1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

Figure A-1b: Time-Series Plot of Interest Rates in Levels

49
 LOG(ER)
4.4

4.3

4.2

4.1

4.0

3.9

3.8

3.7

3.6
1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

Figure A-1c: Time-Series Plot of Log(ER) in Levels

LOG(CPI)
5.2

5.0

4.8

4.6

4.4

4.2

4.0

3.8
1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

Figure A-1d: Time-Series Plot of Log(CPI) in Levels

50
 LOG(OP)
5.0

4.5

4.0

3.5

3.0

2.5

2.0
1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

Figure A-1e: Time-Series Plot of Log(OP) in Levels

LOG(M2)
15.5

15.0

14.5

14.0

13.5

13.0

12.5

12.0
1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

Figure A1-f: Time-Series Plot of Log(M2) in Levels

51
 Tests p-value

Autocorrelation Test

LM(5) 0.77

Normality Test

Jarque-Bera 0.00

Stability Test

Eigenvalue Stability Circle

Table A-1: Diagnostic Tests for the VECM

52
 Response of LOG(DSI) to LOG(DSI) Response of LOG(DSI) to IR
.12 .12

.10 .10

.08 .08

.06 .06

.04 .04

.02 .02

.00 .00

-.02 -.02

-.04 -.04
5 10 15 20 25 30 35 40 45 5 10 15 20 25 30 35 40 45

Response of LOG(DSI) to LOG(ER) Response of LOG(DSI) to LOG(CPI)

.12 .12

.10 .10

.08 .08

.06 .06

.04 .04

.02 .02

.00 .00

-.02 -.02

-.04 -.04
5 10 15 20 25 30 35 40 45 5 10 15 20 25 30 35 40 45

Response of LOG(DSI) to LOG(OP) Response of LOG(DSI) to LOG(M2)

.12 .12

.10 .10

.08 .08

.06 .06

.04 .04

.02 .02

.00 .00

-.02 -.02

-.04 -.04
5 10 15 20 25 30 35 40 45 5 10 15 20 25 30 35 40 45

Figure A-2: Impulse Response Functions

53
 Accumulated Response of DSI to DSI Accumulated Response of DSI to IR
4 4

3 3

2 2

1 1

0 0

-1 -1

-2 -2
5 10 15 20 25 30 35 40 45 5 10 15 20 25 30 35 40 45

Accumulated Response of DSI to ER Accumulated Response of DSI to CPI

4 4

3 3

2 2

1 1

0 0

-1 -1

-2 -2
5 10 15 20 25 30 35 40 45 5 10 15 20 25 30 35 40 45

Accumulated Response of DSI to OP Accumulated Response of DSI to M2

4 4

3 3

2 2

1 1

0 0

-1 -1

-2 -2
5 10 15 20 25 30 35 40 45 5 10 15 20 25 30 35 40 45

Figure A-3: Cumulative Impulse Response Functions

54