Investigating Equilibrium Relationship between Macroeconomic Variables and Malaysian Stock Market Index through Bounds Tests Approach

The current paper conducts an empirical examination into the long-run and short-run equilibrium relationships between macroeconomic variables and the Malaysian stock market index (SMI) for the 1977-2011 period. Specifically, it employs Ng and Perron (NP) bounds statistics test to detect the boundaries of variables stationarity. Subsequently, the co-integrating relationships among variables are tested using the bounds F-statistic test. Eventually, the long-run and short-run equilibrium relationships are analyzed using Pesaran, Shin, and Smith (PSS) bounds tests Approach. The results indicate that all macroeconomic variables are co-integrated with SMI. Besides, understanding the long-run and short-run equilibrium relationships between macroeconomic variables and SMI could be highly appreciable from the perspectives of policymakers, financial economists, domestic and international investors dealing with Malaysian stock market.


Introduction
Over the past three decades, the issue relating to macroeconomic variables and stock markets generated vast volume of literature and heated debate due to the ability of these variables to enhance stock markets and economies. Specifically, the optimal macroeconomic environment promotes business profitability and boosts economic growth (Pal and Mittal, 2011). The performance of countries' economies and stock markets are measured and determined by macroeconomic variables such as exchange rate (ER), inflation rate (INF), money supply (MS), trade balance (TB) and many other variables.
However, several studies debated the influences of macroeconomic variables on matured and emerging stock markets indices. In the US stock market, Bjornald and Leitemo (2009) examined the influences of macroeconomic variables on S&P500 using vector autoregressive (VAR) Model and monthly time series data for the 1983-2002 period. They found that macroeconomic variables influenced S&P500 positively. On the other hand, Morelli (2002) scanned if the conditional volatility of macroeconomic variables explained the volatility of FTSE100 using generalized autoregressive conditional heteroscedasticity (GARCH) Model and monthly time-series data for the 1967-1995 period. He found that the conditional volatility of the macroeconomic variables did not explain the volatility of FTSE100.
Notable studies conducted to examine the equilibrium relationships between macroeconomic variables and emerging stock market indices. Gunasekarage et al. (2004) investigated the long-run and short-run equilibrium relationships between macroeconomic variables and the Sri Lankan stock market index using vector error correction model (VECM) and monthly time-series data for the 1985-2001 period. They found that macroeconomic variables are co-integrated with the Sri Lankan stock market index. In Malaysian context, few notable studies have been found in our area of interest. Ibrahim and Aziz (2003) inspected the long-run and short-run equilibrium relationships between four macroeconomic variables (industrial production index (IP), consumer price index (CPI), the broad money supply (M2), and ER) and Malaysian stock market index using monthly time series data for the 1977-1998 period. They found positive long-run and short-run equilibrium relationships between Malaysian stock market index and both of CPI and IP. Additionally, they found negative long-run and short-run equilibrium relationships between Malaysian stock market index and both of M2 and ER.
The equilibrium relationships between stock markets and macroeconomic variables holds implications for investors, as well as for policy makers to recognize and to evaluate changes in economic conditions, or to forecast the future performance of the macro-economy. Besides, the results of this study based on the Malaysian stock market are likely to hold implications for other emerging stock markets.
The reminder of this paper is organized as follows: The next section provides an overview of Malaysian economy and stock market. Section 3 reviews the previous empirical studies using various equilibrium and volatility time-series models. The data selection and methodology are discussed in section 4. Section 5 analyzes the results. Section 6 draws policy implications, while conclusions, future research and limitation are presented in last section.

Malaysian Economy Overview
Since the independence in 1957, Malaysia began its economic transition from being reliant on primary sectors (mining, quarrying, agriculture, forestry and fishing) to depend more on services, construction and manufacturing to imitate the four Asian Dragons economies; China, Singapore, South Korea, and Taiwan (Ghosh and Ariff, 2004). However, during the last three decades, many emerging countries followed financial and economic liberalization by opening their economies and domestic stock markets to international investors that leaded to enhance economic growth (Ortiz et al., 2006). In 1991, the economic and financial plan was coincided with the financial and economic liberalization. This plan improved productivity and employment in various economic and financial sectors by attracting FDI in Islamic finance, high technology industries, financial services and education system (Ghosh and Ariff, 2004)

Malaysian Stock Market Overview
Malaysian stock market is known as Bursa Malaysia, and considered one of the largest stock markets in South East Asia with 829 listed companies offering a wide range of investment opportunities to domestic and international investors (Bursa Malaysia, 2012). On July 6, 2009 Malaysian stock market followed the latest FTSE Bursa Malaysia index methodology that calculated and disseminated on a real time basis every 15 seconds instead of 60 second to insure that Malaysian stock market reflects the changes in national and global economy (Bursa Malaysia, 2012). However, Malaysian stock market is considered the second among the largest South East Asian stock markets according to its domestic market capitalization. Figure 4  Furthermore, Figure 5 reveals that SMI recorded an annual growth rate of 5.7% for the 1977-2011 period. Before the onslaught of AFC in 1997-98, the performance of SMI rose sharply to reach the first peak in 1993 and the second peak in 1996 with 1275 points and 1238 points respectively. The weakness of SMI in the year 2008 was due to the GFC, whilst SMI declined from 1445 points in 2007 to 877 points in 2008. However, SMI achieved more than 1400 points at the end of 2011. Besides, Figure 3 shows that the trading volume of shares in Malaysian stock market was vivid that achieved a growth rate of 6.7% for the 1993-2011 period.  1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999  The trading volume started at RM20.6 billion and fell gradually to reach the first sharp decline in 1995 at a value of RM8.24 billion, then, increased slowly to reach the first peak in 2007 with a value of RM55.8 billion. The trading volume remained stable over 2007-2009 period. However, the trading volume declined harshly from RM28.6 billion in 2010 to RM27 billion in 2011.
One of the key objectives of this paper is to analyze the long-run and short-run equilibrium relationships between macroeconomic variables (GDP, PPI, CPI, ER and M3) and SMI by applying the equilibrium time-series bounds tests Approach developed by PSS (2001). Therefore, the formulation of paper hypotheses is essentially based on the previous empirical studies that conducted using VAR and VEC equilibrium time-series models as follows: : There are significant long-run equilibrium relationships between macroeconomic variables (GDP, PPI, CPI, ER, and M3) and SMI.

Data Selection
Annual time-series data for the macroeconomic variables and SMI for the 1977-2011 period were collected. SMI represents the annual figures of Malaysian stock market index was obtained from Bursa Malaysia (www.bursamalaysia.com). M3 (RM billion) and ER (RM/US$) were collected from Bank Negara Malaysia (www.bnm.gov.my). GDP denotes the proxy of Malaysian real economic activity (RM billion) was obtained from the World Bank economic indicators data base (http://databank.worldbank.org/ddp/home.do). PPI and CPI were collected from Malaysian department of statistics (www.statistics.gov.my). According to the theory variables (M3, ER, GDP, PPI and CPI) could influence stock market indices either positively or negatively (Chaudhuri and Smiles, 2004;Hanousek and Kocenda, 2011;Ibrahim and Aziz, 2003). To stabilize the variables variances and to remove the seasonality, the variables transformations into natural logarithmic forms were used (Montgomery et al., 2008), except ER to make this variable simultaneous with other variables (Chen et al., 1986).

Methodology Specification
Over the past decades, economists have been aware that estimating time-series econometric models with non-stationary variables leads definitely to spurious results . If a variable contains a www.ccsenet.org/ijef International Journal of Economics and Finance Vol. 4, No. 10;2012 unit root, then, it is non-stationary and it is combination with other non-stationary variable leads to non-meaningful or spurious regression (Brooks, 2008). Evidence from the past studies suggested the presence of unit root in most of financial and economic variables (Montgomery et al., 2008). Therefore, it need to differentiate variables either I(1) or I(d) to achieve their stationarity. For this reason, it is necessary to test the presence of unit roots and to differentiate the variables. The current study employs NP bounds statistics test to detect individually the variables stationarity, since this test is more appropriate with small sample sizes. However, the concept of co-integration reveals the existence of long-run equilibrium relationship among variables, where there is no tendency change, since economic and financial variables are in balance . Besides, the concept of co-integration that deals with the long-run relationship among a group of variables, where these variables either I(0), I(1) or I(d) ) is used to determine the appropriate model. The present study uses the bounds F-statistics test to identify the co-integration among variables.
In time series methodologies, numerous studies examined the long-run and short-run equilibrium relationships between macroeconomic variables and stock market indices using VAR and VEC Models. Under VAR Model, however, all variables are stationary and not co-integrated at the same level either I(1) or I(2) (Engle and Granger, 1987). On another vein, under VECM all variables are stationary and co-integrated at the same level either I(1) or I(2) (Johansen and Juselius, 1990). To examine the long-run and short-run equilibrium relationships between macroeconomic variables (LGDP, LPPI, LCPI, ER and LM3) and LSMI, PSS bounds tests approach is applied. This approach has many advantages compared with VAR and VEC Models. In other words, under PSS approach all variables are stationary and co-integrated at the same level either I(1), I(0) or mutually co-integrated. Furthermore, this approach is more appropriate with small sample sizes (PSS, 2001).In matrix elements, this approach could be formulated as the following system: Δ represents the backshift operator; 1, … . ., 6) and 1, … . . , 6 denote intercepts and error terms respectively; ECM t-1 represents the one period lagged error term which used to link the long-run equilibrium of the variables with their short-run and to insure the co-integration; 1, … . . , 6 denote the coefficients of ECM t-1 ; , 1, … . . ,6 and Γ , 1, … . . , 6 represent the long-run and short-run coefficients of variables respectively.
The H 0 of no co-integration among the variables is tested by setting the of the one lagged variables equal to zero i.e., : = 0, against the H 1 of co-integration among variables where of one lagged variables are not equal to zero i.e., : 0.
The calculated F-statistics are compared with the critical values tabulated at statistical tables in PSS (2001). If the calculated F-statistics are greater than the upper bounds, then the H 0 are definitely rejected, which means that the variables included in the models are shared long-run relationships among themselves (PSS, 2001).If the calculated F-statistics are smaller than the lower bounds, then the H 0 are accepted, which means that the variables included in the models are not shared long-run relationships among themselves (PSS, 2001). However, if the calculated F-statistics fall between the upper and the lower bounds values, then, the decisions are inconclusive to either accept or reject the H 0 (PSS, 2001).
However, after specifying stationarity and co-integration tests we proceed on analyzing the long-run and short-run coefficients as well as the error correction terms of the above system. (1) *, *** denote significance at 1 and 10% level respectively.

Stationarity Bounds Statistics Test
(2) MZ , MZ and MSB are the enhancements of Phillips and Perron (1988) statistics tests. They used to correct the size distortions when residuals are negatively correlated.
Source: Output of Eviews 7.2 Econometric Software. Table 1 shows that all variables are non-stationary at the lower bound except LM3 which is stationary at the 1% significant level. At the upper bound, however, all the variables are stationary, but with different significance levels. Specifically, LGDP, LPPI, LCPI and ER are stationary at 10% significance level, while LSMI is stationary at 1% significance level. The results indicate that all variables employed in regression are stationary and would not cause spurious regression results.

Bounds F-statistics Test
Given that all the variables are stationary, we can apply the bounds F-statistics as suggested by PSS (2001) to test the H 0 of no co-integration among variables. Table 2 provides the results of calculated and critical values of bounds F-statistics test.  Table; (2). *, **, ***, **** denote significance at 1, 2.5, 5 and 10% levels, respectively.
Source: Output of Micro-fit 4.1 Econometric Software. Table 2 reveals that the H 0 of no co-integration in all models are rejected, but with different significance levels.
The H 0 is rejected in LSMI t and LCPI t models at 2.5% and 5% significance level respectively. www.ccsenet.org/ijef International Journal of Economics and Finance Vol. 4, No. 10;2012 For the LGDP t , LPPI t and ER t models the decisions are inconclusive to either accept or reject H 0 and in this study we reject the H 0 . Also, it concludes that all variables are co-integrated with LSMI t in the long-run and any deviations may occur in the short-run. The results are confirmed with others results studies (Chaudhuri and Smiles, 2004;Hanousek and Kocenda, 2011;Ibrahim and Aziz, 2003).

Long-run and Error Correction Analyses
The presence of long-run equilibrium relationships among variables does not imply the perfect co-integration (Engle and Granger, 1987). That is, if the error correction terms are significant with negative signs, then, the error correction terms are stable and co-integration, the causality in the long-run is achieved. Table 3 provides the long-run coefficients and error correction terms of the models in the present paper (See, Section 4.2).  Table 3 shows that at 1% significance level, LGDP t-1 is positively associated with LM3 t model, and the reverse does hold. LPPI t-1 and LCPI t-1 are negatively associated with LSMI t model, while LPPI t-1 and ER t-1 are negatively associated with LM3 t model. However, LPPI t-1 is negatively associated with ER t model. At 5% significance level, LCPI t-1, LGDP t-1 and ER t-1 are positively associated with LM3 t, LSMIt and LCPI t models respectively, while ER t-1 is negatively associated with LPPI t model. At 10% significance level, LGDP t-1, ER t-1 and LM3 t-1 are negatively associated with LPPI t, LSMI t and LCPI t models, while ER t-1 is positively associated with LGDP t model.
Furthermore, Table 3 illustrates that the coefficients of (ECM t-1 )'s are significant with appropriate signs. Thus, the perfect co-integration and causality among variables are achieved. Further, the highly significant of ECM t-1 implies a quite speed of achieving the long-run equilibrium. Specifically, the ∆LSMI t model records the highest ECM t-1 in absolute value among other models suggesting that 137% of any previous disequilibrium in the long-run would be shortly corrected back in the current year, while, LPPI t model records the lowest ECM t-1 in absolute value suggesting a very low speed toward its long-run equilibrium.

Short-run Analysis
After conducting long-run and error correction analyses we proceed for analyzing the short-run status of the models. So, table 4 presents the results of the short-run analysis. These results reveal that the lagged variables in ∆LSMI t model are significant and the variables are Granger cause ∆LSMI t in the short-run, while the lagged variables of ∆LCPI t are not Granger cause ∆LGDP t , ∆LPPI t and ∆LM3 t models. On the other hand, the lagged variables of ∆LSMI t , and ∆LM3 t are Granger cause ∆LCPI t model, while the lagged variables of ∆LGDP t and ∆LSMI t are not Granger cause ∆ER t and ∆LM3 t models respectively.

Policy Implications
The results of this study suggest that Malaysian policy makers should pay more attention to the effects of monetary policies and economic activity measurement on stock market. Specifically, the results are confirmed with earlier results of Ibrahim (1999) who found that the money supplies are positively and negatively associated with SMI in both long-run and short-run. The contraction of money supply leads to lower interest rate, lower firm investment, and then, decreases the attractiveness of investors to invest in stock market. In sharp contrast, the expansion of money supply leads to raise stock market index due to more liquidity, more output and portfolio adjustment (Hsing and Hsieh, 2012). However, a further increase in money supply reduces stock market index because of inflationary influences and its negative effects on stock market index. Besides, the results are confirmed with the results of Hanousek and Kocenda (2011) who revealed that the GDP is positively and negatively associated with the emerging European stock market indices. A high GDP in a particular industry, such as the Malaysian manufacturing industry is a sign that the firms in that industry are performing well, thereby leads to increase their share prices in stock market. In sharp contrast, a low GDP is a sign that these firms are not performing well, which lead to decrease their share prices in stock market. Indeed, the results are consistence with the results of Ibrahim and Aziz (2003) study who found that the ER is negatively associated with Malaysian stock market index in the long-run and short-run.
In addition, Malaysian policy makers, also, should be aware to the importance of inward FDI in enhancing economic growth. The attraction of FDI into Malaysia creates jobs and employment in economic and financial sectors, improves production and management techniques and moreover, increases the competitiveness in domestic and international markets.

Conclusions, Future Research and Limitation
The present paper examines the long-run and short-run equilibrium relationships between macroeconomic variables (GDP, PPI, CPI, ER, and M3) and SMI using annual time-series data for the 1977-2011 period. However, it employs NP bounds statistics test to detect the variables stationarity and bounds F-statistic for testing the co-integrating relationships among variables. The long-run and short-run equilibrium relationships among variables are analyzed using PSS bounds tests Approach. Results of NP test show that the H0 of non-stationary is rejected at all cases except one variable. More specifically, the variables SMI, GDP, PPI, CPI, ER, and M3 are stationary at the upper bound, while the variable GDP is stationary at the lower bound. The results of bounds F-statistics test reveal that all variables are co-integrated with SMI. In addition, the results www.ccsenet.org/ijef International Journal of Economics and Finance Vol. 4, No. 10;2012 of PSS bounds tests Approach show the presence of long-run and short-run equilibrium relationships between all macroeconomic variables and SMI. In particular, PPI, CPI, ER, and M3 are negatively associated with SMI in the long-run, while GDP is positively associated. Additionally, GDP, PPI, and ER are negatively associated with SMI in the short-run, while CPI and M3are positively associated.
The study adds to the existing literature and focuses on the long-run and short-run equilibrium relationships between macroeconomic variables and stock market index in Malaysia as an emerging stock market rather than matured stock markets (case of US or the UK) which have been frequently studied in the past. Finally, the results of this paper are of particular interest and importance to policy makers, financial economists, domestic investors and international investors dealing with Malaysian economy and stock market. In fact, future research could broaden this study by adding more variables that have significant influences on stock prices such as oil prices and inward FDI or including more than one country to draw robust results.
From statistical perspective, the main limitation of this study is the small sample size of 35 observations which has a limiting factor, since the number of lags that use, consumes the number of observations and leads to specification errors in the analysis