Cointegration and causality between stock index and macroeconomic variables in an emerging market.

Brahmasrene, Tantatape ; Jiranyakul, Komain

ABSTRACT

This study examined the relationship between stock market index and selected macroeconomic variables during the post-financial liberalization (pre-financial crisis) and postfinancial crisis in Thailand. In the empirical analysis, unit root, cointegration and Granger causality tests were performed. The post-financial liberalization results showed that the stock market index, the industrial production index, money supply, exchange rate, and world oil prices contained a unit root and were integrated of order one. Johansen cointegration test was then employed. The results showed at least one cointegrating or long-run relation between the stock market index and a set of macroeconomic variables. Money supply had a positive impact on the stock market index while the industrial production index, the exchange rate and oil prices had a negative impact. During the post-financial crisis, all variables were integrated at different orders. Cointegration existed between the stock market index and macroeconomic variables. In addition, the Granger causality test indicated money supply was the only variable positively affecting the stock market returns.

INTRODUCTION

The Stock Exchange of Thailand has been considered an emerging stock market since its inauguration in April 1975. The market capitalization of Thailand Stock Exchange is small while bond trading and other financial innovations have emerged in just the last few years. Like other emerging stock markets in Asia, liberalization in the Thai financial markets, both money and capital markets, reduced the regulation for foreign investors who were interested in investing in Thailand. The financial liberalization in 1992 included lifting capital control measures and allowing banks to lend and borrow more freely in both in- and off-shore transactions. In addition, the Thai government urged capital inflows in both portfolio and foreign direct investment. As a result, the volume of stock trading increased substantially in recent years. Equity instruments are a crucial source of funds for business firms. A continuous increase in private investment via issuing new stocks can be a conduit of GDP expansion and, thus, a high employment rate.

Under the fixed exchange rate regime prior to the financial crisis in 1997, Thailand saw large capital inflows, especially in terms of portfolio investment. This nearly offset the huge current account deficits. Additionally, large capital inflows caused domestic financial institutions to lend a large number of loans to both firms and individual borrowers. The ratio between loans and deposits in the banking system was as high as 1.35 in mid-1990 compared to 0.75 in early 1990. Many analysts believed this was due to the overheating of the Thai economy. In late 1996, private investment accounted for more than 40 percent of the national income. Such phenomena showed that domestic borrowers relied more on foreign capital inflows than domestic savings. During this period, the domestic interest rate rose and caused a wide gap between domestic and foreign interest rates. This interest rate differential induced large capital inflows mostly in portfolio investment. The financial crisis in 1997 had a devastating impact on the Thai economy. A significant effect related to exchange rate risk under the floating exchange rate regime began in July 1997. Other than real economic activity (e.g., real GDP or the industrial production index) that could affect an investment decision in common stocks, the risk generated from exchange rate fluctuations may also distort the portfolio investment decision. The main objective of this study was to investigate the effects of macroeconomic variables on stock market index/returns in Thailand during the post-financial liberalization prior to the financial crisis (January 1992-June 1997) and after the financial crisis (July 1997-December 2003). The stock market return represents the change in stock market index.

REVIEW OF THE LITERATURE

The literature review consists of two sections. The first reviews the literature on factors affecting stock market returns with emphasis on money supply and inflation. The second focuses on the long-run relationship or cointegration between stock market returns and macroeconomic variables.

Factors Affecting the Stock Market

Chen, Roll and Ross (1986) employed a multivariate arbitrage pricing theory (APT) to analyze the relationship between the market returns and macroeconomic factors, including measures of industrial production, the money supply, inflation, and interest rate and exchange rate variables. They confirmed a strong relationship between the market returns and these variables. Hamao (1988) found that inflationary expectations cause a change in the risk premium and in the term structure of interest rate. In turn, these variables have a significant impact upon stock returns in the Japanese market. Fung and Lie (1990) concluded that the response of the stock market index to changes in domestic production and money supply was weak in Korea. In other words, investors did not perceive a change in economic conditions could affect stock prices. Dhakal, Kandil, and Sharma (1993) adopted a vector autoregression (VAR) model to test the impact of a change in the money supply on a change in the stock market index under a money market equilibrium condition. They discovered a significant relationship between these two variables in the United States. A study by Abdullah and Haywarth (1993) also found that a change in the market index was influenced by the rate of inflation and by the change in the money supply. On the relationship between inflation and stock returns, Fama (1981) indicated that most economic factors, except inflation, exhibited a positive correlation with the stock market index. The negative correlation between inflation and real equity returns was partially explained by the proxy hypothesis. In brief, inflation and real equity returns react in an opposite manner to news about future real output growth. Aarstol (2000) confirmed that this negative relationship persisted even when output growth was controlled. Rapach (2001) examined the effects of money supply, aggregate spending, and aggregate supply shocks on real U.S. stock prices in a structural VAR model. One of the main findings was that real stock returns were negatively correlated with inflation.

Cointegration

Long-run relationships between the stock market index and various macroeconomic variables are commonly observed. Mookerjee and Naka (1995) showed that short-run relationships among these variables existed in the Japanese stock market. However, this might not be the case for a small open economy. Mookerjee and Yu (1997) further found that not all macroeconomic variables were cointegrated with stock prices in Singapore. Cheung and Ng (1998) obtained evidence of cointegration between stock market indices and various macroeconomic variables, including oil prices. Cointegration between stock market returns and several macroeconomic variables also existed in South Korea (Kwon & Shin, 1999). However, the stock market indices were found not to be leading indicators of macroeconomic variables, such as the production index, money supply, exchange rate, and the trade balance. In the case of Malaysia, Ibrahim (1999) indicated that stock prices had a long-run relationship with consumer prices, credit aggregates, and official reserves. In 2003, Ibrahim found cointegration between returns and the money supply in the Malaysian equity market to be a major influence on equity prices. Groenewold (2004) analyzed the relationship between share prices and real output using structural VAR models without considering other macroeconomic variables. One of the major results showed that a macroeconomic boom caused an overvaluation in stock prices.

CONCEPTUAL FRAMEWORK

An early theory of arbitrage pricing uses a functional form to test the relationship between stock index and macroeconomic variables. All individual stocks are affected by common factors. The multifactor model, as in the arbitrage pricing theory (APT), stipulates various factors that can influence the returns of all assets in the stock market. Market index can be affected by macroeconomic variables, such as changes in interest rate, money supply, economic growth, and inflation. By and large, the APT model has a drawback as it assumes the constant term to be a risk-free rate of return. The functional form of multiple regression that is widely used in empirical studies is:

(1) SE[T.sub.t]=[[beta].sub.0]+[[beta].sub.1]I[P.sub.t]+[[beta].sub.2] [M.sub.2]+[[beta].sub.3][P.sub.t]+[[beta].sub.4] E[X.sub.t]+[[beta].sub.5]I[N.sub.t]+[[beta].sub.6]O[P.sub.t][e.sub.t]

where

* SE[T.sub.t] denotes the market index of overall market value of listed stocks in the Stock Exchange of Thailand. This is the sum of market value (share outstanding multiplied by market price) of all stocks being traded. A change in the index represents capital gains/losses. Rate of return ()SET) is measured as the sum of capital gains/losses for each period. Dividends are not available for inclusion in this study.

* I[P.sub.t] is the logarithm of the total industrial production index, a proxy for real economic activity.

* M[2.sub.t] is the logarithm of changes in the broad definition of money supply.

* [P.sub.t] is the logarithm of the inflation rate.

* E[X.sub.t] is the logarithm of the nominal exchange rate measured in terms of Thai baht per U.S. dollar.

* I[N.sub.t] is the logarithm of the long-term interest rate.

* O[P.sub.t] is the logarithm of oil price measured in U.S. dollar per barrel.

* [e.sub.t] is a disturbance term.

The ordinary least squares (OLS) estimate can be applied to Equation (1) if all variables are stationary. If variables are not stationary, the typical OLS regression will yield spurious results or will not be meaningful (Gujarati 2003).

Some systematic factors in the economy may play a major role in affecting the stock market index. In particular, a different period of time can capture different responses of stock prices to varying levels of macroeconomic activity. When Thailand experienced financial crisis, the policy makers shifted from a fixed foreign exchange rate regime prior to the crisis to a flexible rate regime after the crisis. Policymakers became more prudent in exercising monetary policy tools. The cointegration test indicates the presence or absence of long-run equilibrium relationships among variables. Cointegration among variables may or may not exist due to changes in their orders of integration when the regime shifts. Therefore, this research distinguishes the effects of macroeconomic variables on market returns in two periods: the post-financial liberalization before financial crisis (January 1992-June 1997) and the post-financial crisis (July 1997-December 2003). Results are expected to be different due to these different circumstances.

DATA AND METHODOLOGY

The Bank of Thailand Economic Bulletin provides monthly data on the industrial production index, the consumer price index (price level), money supply, interest rates, and nominal exchange rates from January 1992 to December 2003. The price level series are adjusted to the base period of 1998. Data used for the stock market index are obtained from Stock Exchange of Thailand index. The Energy Information Administration is the source for oil prices.

The relationship between the stock market index and crucial macroeconomic variables in equation (1) can be applied if all variables are stationary in level or trend. If they are not stationary in level, but stationary in first differences, they may or may not be cointegrated. If they are cointegrated, the error correction mechanism (ECM) can be used to determine the short- run deviation from the long-run equilibrium. If they are not cointegrated, the Granger causality can be employed to navigate direction of causation.

In practice, the most widely used method of estimation is based on the condition that many economic variables are known to be integrated of order one or I(1), with or without cointegration. The Phillips & Perron (PP) unit root test (Phillips & Perron, 1988) for time series is performed to determine the order of integration of each variable. Furthermore, Johansen cointegration tests (Johansen, 1991 & 1995) are conducted to determine whether the stock market index and a set of macroeconomic factors are cointegrated. If cointegration exists, there is a long-run relationship among the variables in question. If cointegration does not exist, Granger bivariate causality tests are employed to determine the direction of causation between stock market returns (stationary first differences of stock market index, DSET) and each of the relevant macroeconomic variables.

The Johansen's cointegration test employs the maximum likelihood procedure to determine the existence of cointegrating vectors. In nonstationary time series, a vector autoregressive (VAR) form is indicated in equation (2).

(2) [DELTA][Z.sub.t][product][Z.sub.t] + [k.summation over (i=1)] [[GAMMA].sub.i][[DELTA][Z.sub.t-i] + [e.sub.t]

Where

* [Z.sub.t] is a vector of nonstationary variables.

* [[GAMMA].sub.i] is the matrix of short-run parameters.

* [product] = [alpha][[beta].sup.1], is the information on the coefficient matrix between the levels of the series.

The relevant elements of the " matrix are adjusted coefficients and the $ matrix contains the cointegrating vectors. According to Johansen and Juselius (1990), there are two likelihood ratio test statistics to test for the number of cointegrating vectors, i.e. the maximum eigenvalue statistic and the trace statistic. The two test statistics are compared with the critical values. If the maximum eigenvalue statistic and the trace statistic are greater than the critical values, cointigrating relation(s) will be present. The Johansen procedure bases on the error correction mechanism (ECM) representation of the vector autoregressive model.

The equation below is used to test the causation from each of the macroeconomic factors ([X.sub.t]) to stock market returns.

(3) [DELTA]SE[T.sub.t] = [[alpha.sub.t][DELTA]SE[T.sub.t-i] + [k.summation over (j=1)][[beta.sub.j][X.sub.t-j] + [e.sub.t]

The equation used to test the causation from stock market return to a change in each macroeconomic variable is

(4) [X.sub.t] = [[alpha.sub.0] + [k.summation over (i=1)][[alpha.sub.i][X.sub.t-I] + [k.summation over (j=1)][[beta.sub.j]SE[T.sub.t-j] + [u.sub.t]

Equation (3) postulates that stock market returns ([DELTA]SET) are related to the previous [DELTA]SET and to an independent macroeconomic variable ([x.sub.t]), and equation (4) postulates a similar behavior for [x.sub.t]. According to the Granger causality test, if the set of estimated coefficients on the lagged x in (3) is statistically significant and the set of estimated coefficients on the lagged [DELTA]SET in (4) is statistically insignificant, then the unidirectional causality from x to [DELTA]SET exists. In contrast, if the set of estimated coefficients on the lagged x in (3) is not statistically different from zero and the set of estimated coefficients on the lagged [DELTA]SET in (4) is statistically different from zero, then unidirectional causality from [DELTA]SET to x exists. If the set of [DELTA]SET and x coefficients are insignificant in both regressions, independence occurs. Bi-directional causality is present when both sets of [DELTA]SET and x coefficients are significant in both regressions. The power of the test is valid if the [b.sub.j] coefficients are significantly different from zero.

EMPIRICAL RESULTS

The Post-Financial Liberalization (Pre-Financial Crisis)

The results of the unit root test during the period of the post-financial liberalization (January 1992-June 1997) are reported in Table 1. The PP tests show the industrial production index is trend stationary, that is, I(0). Without trend, the industrial production index is nonstationary at level, but its first difference is stationary, I(1). The logarithm of each of the remaining variables contains a unit root at level. However, tests of first differences indicate stationarity or the absence of unit roots. Therefore, all variables are integrated of order one, I(1) during the pre-financial crisis. Money supply is chosen as a representative financial variable because the correlations among financial variables are high: 0.99 between money supply and consumer price index, 0.63 between money supply and interest rate, and 0.65 between consumer price index and interest rate.

The Johansen cointregation test is employed as shown in Table 2. Cointegration among the stock market index, the industrial production index, money supply, nominal exchange rate and oil price is performed using up to four lags length. This optimal lag length is determined by generally accepted techniques. The maximum Eigenvalue and Trace statistics show an acceptance of one and two cointegrating relation, respectively, at the 5 percent level among all five series. There exists at least one cointegrating relation among these series. The long-run relationship between the stock market index and four macroeconomic variables is:

(5) SE[T.sub.t] = -1.078 I[P.sub.t] + 0.975M[2.sub.t] - 8.447E[X.sub.t] - 1.496O[P.sub.t] (0.655) (0.358) (2.212) (0.169)

The number in parenthesis is standard error. The error correction mechanism (ECM) is employed when cointegration exists. Equation (6) below shows the short-run deviation from the long-run equilibrium:

(6)[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

ECT in equation (6) is the error correction term. Lagged first differences coefficients at 10 percent significant level are customarily omitted. The number in parentheses is t-statistic.

The Post-Financial Crisis

The results of the unit root tests for the post-financial crisis (July 1997-December 2003) are reported in Table 3. The PP tests show that the variables are integrated at different orders. The stock market index, industrial production, nominal interest rate, and oil prices are integrated of order one, I(1). This means their first differences are stationary. Money supply, consumer price index, and nominal exchange rate are integrated of order zero, I(0), or are stationary series.

DISCUSSION

The Post-Financial Liberalization (Pre-Financial Crisis) During the pre-financial crisis, there was a long-run relationship between the stock market index (SET) and the following macroeconomic factors: industrial production index (IP), money supply (M2), nominal exchange rate (EX), and oil prices (OP) from Table 2 & Equation 5. Money supply positively influenced stock market index while the industrial production index, nominal exchange rate and international oil prices negatively influenced the stock market index. According to the data from Bank of Thailand, the ratio of M2 to GDP increased from approximately 0.80 before the crisis to 1.20 by the end of 1999. The ratio has changed slightly since that date. The fiscal position revealed a budgetary surplus until the crisis prompted expansionary monetary and fiscal policies to stimulate the economy. The nominal interest rate was somewhat manipulated to a low level in order to stimulate private investment spending.

Contrary to the existing theory, the industrial production index negatively affected the stock market index. Throughout the crisis, capacity utilization and the industrial production index declined as firms were reluctant to expand their levels of production. In spite of a decline in the industrial production index, the stock market index continued to rise because speculative motive in financial and real estate sectors dominated the buy and hold strategy. It should be noted here that financial and real estate sectors are the major component of the stock market index. Furthermore, the nominal exchange rate adversely affected the stock market index as the inflow in portfolio investment plunged when Thai baht depreciated against the US dollar. Higher international oil prices due to foreseeable higher costs of production sent a negative signal to the stock market.

The dynamics of equation (6) indicates short-run impact of changes in industrial production and oil prices on stock market returns (also known as capital gain or loss). Stock market returns are not affected by changes in money supply and nominal exchange rate. The variables that play an important role are industrial production (or real activities) and oil prices. The error correction term (ECT) is significant at the 1 percent level. Its value indicates that about 0.56 of the discrepancy between the actual and the long-run or equilibrium, value of SET is corrected or eliminated each month. Case in point, the stock market index is related to some macroeconomic variables: industrial production index, money supply, nominal exchange rate, and international oil prices in the long run. There exists cointegration among these variables. The economic bubbles or pre-crisis relationships will not reappear if banks prevent it by restricting loans only to those firms that have fundamental strength. However, it must be said that what routinely occurs in a well-developed stock market may not appear in an emerging stock market.

The Post-Financial Crisis

The Johansen cointegration test in Table 4 shows no cointegration among four nonstationary variables. Thus, the Granger causality test is implemented. This test requires a stationary pair-wise series. Table 5 exhibits results from bivariate causality tests.

The post-financial crisis shows no cointegration between stock market index (SET) and crucial macroeconomic variables (Table 4). At the height of the crisis, the Bank of Thailand decided to no longer "peg" the nation's exchange rate. The exchange rate fluctuated erratically until 1998, when the fluctuations subsided. This structural break changed the behavior of investors and business firms. Due to the instability induced by the financial crisis, the variables become integrated at different orders post-crisis. As a result, it caused a change in the relationship between stock market returns and macroeconomic variables.

Unidirectional causality exists between stock market returns (DSET) and the following macroeconomic factors: money supply (M2), change in nominal interest rate and nominal exchange rate (Table 5). As indicated in Table 5, the money supply causes stock market returns to change in the same direction. In other words, money supply is a precursor of stock market returns. The money supply became the only variable to affect stock market return in the post-crisis period. This is due in part to its primary role in economic stimulus, while followed by expansionary fiscal policy. Additionally, information about economic conditions is not effectively transmitted among investors in the stock market. Furthermore, stock market return is a leading indicator of movements in nominal interest and nominal exchange rates under the managed float regime (post-financial crisis) with the highly significant causation from stock market return to nominal exchange rate and nominal interest rate.

Also noteworthy is the evidence showing that there are no relationships between stock market returns and the following macroeconomic factors: industrial production index, and oil prices. In order to prevent adverse supply shocks, the government controls the price of gasoline used in real activity. Thus, world oil prices did not have a significant impact. The erratic behavior of the stock market was considered to be a temporary or transitory phenomenon as the economy gradually heals itself and adjusts toward long-run stability.

CONTRIBUTIONS

This paper makes two main important contributions to the literature concerning the long-run relationship between the stock market and macroeconomic variables. First, no existing research has studied this relationship in Thailand using a unit root test and cointegration test in the period that contains a structural break. The post-financial liberalization (and pre-financial crisis) and post-financial crisis periods are examined to control for the structural break that may result from changes in policy regime. Second, in the absence of cointegration after the financial crisis, the results from causality testing yield different notions from the existing literature. In summary, relationships exist among stock market return, money supply, nominal interest and exchange rate in the post- financial crisis. The industrial production index is not an indicator of stock market expansion at all after the financial crisis. Oil price shocks do not have an impact on the stock market, as generally believed. The estimated results should be stable and statistically acceptable since well-known and acceptable econometric methods were employed in the analysis.

IMPLICATIONS

This study finds that the stock index is cointegrated with some macroeconomic variables in the pre-financial crisis, but not in the post-financial crisis. During the post-financial liberalization prior to the financial crisis in Thailand, the industrial production index adversely affected stock market index (equation 5). This is contradictory to the belief that there is a positive linkage between real activities and stock market. The structural break has caused a change in the relationship between the stock market index and crucial macroeconomic variables. At the height of the financial crisis, the money supply played an important role. This suggests an expansionary monetary policy may be able to stimulate the stock market. An increase in money supply will increase stock market returns. However, the evidence obtained here suggests that this policy will be effective only in the short run. Additionally, understanding the stock market reaction to various macroeconomic variables over time, especially during an economic crisis, should provide valuable insight to both practitioners and researchers. For example, stock market returns may be employed as a leading indicator of change in nominal interest and exchange rates. The practical implication of this research is that in the recovery from an economic crisis, especially if it has significant financial implications, investors should spend more time and effort acquiring the knowledge associated with monetary policy and its effects on the economy.

CONCLUSION

This study examines the relationship between the stock market and several macroeconomic variables in Thailand. The Phillips & Perron (PP) test is used to test for unit roots in the variables in question. Cointegration tests between the stock market index and a set of the macroeconomic variables are performed for two periods, the post-financial liberalization and post-financial crisis periods. The existing literature indicates that real economic activity has a strong and positive effect on the stock market index. Money supply has a positive influence on stock market returns while inflation has a negative impact. Furthermore, oil price shocks and nominal exchange rate movements have been found to adversely affect stock market returns. Contrary to these findings, this study has found cointegration between stock market index and crucial macroeconomic variables during the pre-financial crisis only. During the post-fiancial crisis, causality between stock market return and macroeconomic variables is observed (to some extent) only for the money supply, change in nominal interest rate and exchange rate variables. In order to generalize the results obtained above, several suggestions for future research may be offered. The empirical model may be estimated with additional and/or alternative economic and financial factors. Studies encompassing various regions should be conducted when more data are available. Such research will contribute toward improving our understanding of the emerging financial markets responses to the frequently occurring phenomena of economic crisis induced by globalization.

REFERENCES

Abdullah, D. A. & S. C. Hayworth (1993). Macroeconometrics of stock price fluctuations. Quarterly Journal of Business and Economics, 32, 49-63.

Akaike, H. (1969). Fitting autoregression for predictions. Annals of the Institute Statistical Mathematics, 21, 243-247.

Aarstol, M. (2000). Inflation, agency costs, and equity return. Journal of Economics and Business, 52, 387-403.

Bank of Thailand 1992-2004. Economic Bulletin, Bangkok: Thailand.

Chen, N. F., R. Roll & S. Ross (1986). Economic forces and the stock market. Journal of Business, 59, 383-403.

Cheung, Y. & L. K. Ng (1998). International evidence on the stock market and aggregate economic activity. Journal of Empirical Finance, 5, 281-296.

Chung, S. K. & S. S. Tai (1999). Cointegration and causality between macroeconomic variables and stock market returns. Global Finance Journal, 10, 71-81.

Dhakal, D., M. Kandiland & S. Shama (1993). Causality between money supply and share prices: A VAR investigation. Quarterly Journal of Business and Economics, (32), 52-74.

Energy Information Administration, www.eia.doe.org.

Fama, E. F. (1981). Stock return, real activity, inflation, and money. American Economic Review, 71, 545-565.

Fung, H. G. & C. J. Lie (1990). Stock market and economic activity: A causal analysis. In S. L. Rhee & R. P. Chang (Eds.), Pacific-Basin Capital Markets Research, Amsterdam: North-Holland.

Groenewold, N. (2004). Fundamental share prices and aggregate real output. Applied Financial Economics, 14, 651-661.

Gujarati, Damodar N. (2003). Basic Econometrics, New York: McGraw-Hill, Inc.

Hamao, Y. (1988). An empirical examination of the arbitrage pricing theory using Japanese data. Japan and the World Economy, 1, 45-61.

Ibrahim, M. H. (1999). Macroeconomic variables and stock prices in Malaysia: An empirical analysis. Asian Economic Journal, 13, 219-231.

Ibrahim, M. H. (2003). Macroeconomic Forces and capital market integration: A VAR analysis for Malaysia. Journal of the Asia Pacific Economy, 8, 19-40.

Johansen, S. (1991). Estimation and hypothesis testing of cointegration vectors in gaussian vector autoregressive models. Econometrica, 59, 1551-1580.

Johansen, S. (1995). Likelihood-based inference in cointegrated vector autoregressive models, Oxford: Oxford University.

Kwon, C. S. & T. S. Shin (1999). Cointegration and causality between macroeconomic variables and stock market return. Global Finance Journal, 10, 78-81.

MacKinnon, J. G. (1996). Numerical distribution functions for unit root and cointegration tests. Journal of Applied Econometrics, 11, 601-618.

Mookerjee, T. K. & A. Naka (1995). Dynamic relations between macroeconomic variables and the Japanese stock market: An application of a vector error-correction model. Journal of Financial Research, 18, 223-237.

Mookerjee, R. & Q.Yu (1997). Macroeconomic variables and stock prices in a small open economy: The case of Singapore. Pacific-Basin Finance Journal, 5, 377-388.

Phillips, P. C. B. & P. Perron (1988). Testing for a unit root in time series regression, Biometrika, 75, 335- 346.

Rapach, D.E. (2001). Macro shocks and real stock prices. Journal of Economics and Business, 53, 5-26.

Tantatape Brahmasrene, Purdue University North Central Komain Jiranyakul, National Institute of Development Administration

Table 1: Unit Root Tests
(January 1992-June 1997)

 PP Test

Variables Without Trend With Trend

Stock Market Index:
SET -0.525(3) [0.88] -0.403(2) [0.99]
[DELTA] SET -7.394(3) *** [0.00] -8.106(1) *** [0.00]

Industrial Production:
IP -1.631(3) [0.46] -4.765(3) *** [0.00]
[DELTA] IP -11.076 *** [0.00]

Money Supply(M2):
M2 -0.160(1) [0.94] -2.577(2) [0.29]
[DELTA] M2 -7.398(2) *** [0.00] -7.337(2) *** [0.00]
Consumer Price Index:

P -0.301(2) [0.98] -2.338(1) [0.41]
[DELTA] P -6.665(3) *** [0.00] -6.627(3) *** [0.00]
Interest Rate:

IN -1.436(4) [0.56] -2.510(4) [0.32]
[DELTA] IN -4.835(0) *** [0.00] -4.809(0) *** [0.00]
Exchange Rate:

EX -1.420(2) [0.57] -1.078(0) [0.92]
[DELTA] EX -5.711(3) *** [0.00] -5.696(4) *** [0.00]
Oil Price:

OP -1.976(2) [0.30] -2.081(3) [0.55]
[DELTA] OP -4.877(3) *** [0.00] -4.896(2) *** [0.00]

 PP Test

Variables Without Trend Order of Integration

Stock Market Index: I(1)
SET -0.525(3) [0.88]
[DELTA] SET -7.394(3) *** [0.00]

Industrial Production: I(1) or I(0) with trend
IP -1.631(3) [0.46] stationarity
[DELTA] IP -11.076 *** [0.00]

Money Supply(M2): I(1)
M2 -0.160(1) [0.94]
[DELTA] M2 -7.398(2) *** [0.00]
Consumer Price Index: I(1)

P -0.301(2) [0.98]
[DELTA] P -6.665(3) *** [0.00]
Interest Rate: I(1)

IN -1.436(4) [0.56]
[DELTA] IN -4.835(0) *** [0.00]
Exchange Rate: I(1)

EX -1.420(2) [0.57]
[DELTA] EX -5.711(3) *** [0.00]
Oil Price: I(1)

OP -1.976(2) [0.30]
[DELTA] OP -4.877(3) *** [0.00]

Note The number in parentheses is the optimal bandwidth determined by
 Newey-West using Bartlett Kernel. The number in brackets is
 one-sided p-values of accepting the null hypothesis of a unit root
 (MacKinon, 1996).

*** significant at 1 percent level

Table 2: Johansen Cointegration Test results
(January 1992-June 1997)

Cointegration Maximum Eigenvalue
rank (r) Statistics Trace Statistics

r=0 34.27 (33.46) ** 85.73 (68.52) **
r[less than or equal to]1 23.25 (27.07) 51.46 (47.21) **
r[less than or equal to]2 18.95 (20.97) 28.22 (29.08)
r[less than or equal to]3 9.17 (14.07) 9.27 (15.41)
r[less than or equal to]4 0.10 (3.76) 0.10 (3.76)

Note The number in parenthesis is the critical value at 5 percent level.

** significant at 5 percent level.

Table 3: PP Test for Unit Root
(July 1997-January 2003)

 PP Test Statistics

Variables Without Trend With Trend

Stock Market Index:
SET -1.730(2) [0.41] -1.585(0) [0.79]
[DELTA] SET -9.359(0) *** [0.00] -9.640(2) *** [0.00]

Industrial Production:
IP -0.474(3) [0.89] -4.506(4) *** [0.00]
[DELTA] IP -15.126 *** [0.00]

Money Supply(M2):
M2 -3.079(4) ** [0.03] -4.265(4) *** [0.01]

Consumer Price Index:
P -4.390(5) *** [0.00] -5.490(4) *** [0.00]

Interest Rate:
IN -0.422(5) [0.90] -1.951(5) [0.62]
[DELTA]IN -5.882(4) *** [0.00] -5.866(4) *** [0.00]

Exchange Rate:
EX -5.116(1) *** [0.00] -4.671(2) *** [0.00]

Oil Price:
OP -1.146 (1) [0.69] -2.013(2) *** [0.58]
[DELTA]OP -6.142 (5) *** [0.00]

Variables Order of Integration

Stock Market Index: I(1)
SET
[DELTA] SET

Industrial Production: I(1) or I(0) with trend
IP stationarity
[DELTA] IP

Money Supply(M2): I(0)
M2

Consumer Price Index: I(0)
P

Interest Rate: I(1)
IN
[DELTA]IN

Exchange Rate: I(0)
EX

Oil Price: I(1) or I(0) with trend
OP stationarity
[DELTA]OP

Note The number in parentheses is the optimal bandwidth determined by
Newey-West using Bartlett kernel. The number in brackets is one-sided
p-values of accepting the null hypothesis of a unit root (MacKinnon,
1996).

*** significant at 1 percent level and

Table 4: Johansen Cointegration Test
(July 1997-December 2003)

 Maximum Eigenvalue
Cointegration rank (r) Statistics Trace Statistics

r=0 18.90 (27.07) 41.44 (47.21)
r[less than or equal to]1 11.59 (20.97) 22.54 (29.68)
r[less than or equal to]2 7.64 (14.07) 10.95 (15.41)
r[less than or equal to]3 3.30 (3.76) 3.30 (3.76)

Note The number in parenthesis is critical values at the 5 percent
level.

Table 5: Granger Causality F-Statistics
(July 1997-December 2003)

 F-statistic Optimal Lag

[DELTA]IP [right arrow] [DELTA]SET 0.72 (0.40)
[DELTA]SET [right arrow] 1.46 (0.23) 1
 OL68\f"Symbol"\s12IP

 M2 [right arrow] [DELTA]SET 4.18 (0.04) **
[DELTA]SET [right arrow] M2 0.21 (0.65) 1

P [right arrow] [DELTA]SET 2.38 (0.13)
[DELTA]SET [right arrow] P 0.51 (0.48) 1

[DELTA]IN [right arrow] [DELTA]SET 2.56 (0.11)
[DELTA]SET [right arrow] 6.88 (0.01) *** 1
 OL68\f"Symbol"\s12IN

EX [right arrow] [DELTA]SET 0.97 (0.38)
[DELTA]SET [right arrow] EX 7.89 (0.00) *** 2

[DELTA]OP [right arrow] [DELTA]SET 0.55 (0.46)
[DELTA]SET [right arrow] 0.01 (0.93) 1
 OL68\f"Symbol"\s12OP

 F-statistic AIC

[DELTA]IP [right arrow] [DELTA]SET 0.72 (0.40)
[DELTA]SET [right arrow] 1.46 (0.23) -4.54
 OL68\f"Symbol"\s12IP

 M2 [right arrow] [DELTA]SET 4.18 (0.04) **
[DELTA]SET [right arrow] M2 0.21 (0.65) -7.96

P [right arrow] [DELTA]SET 2.38 (0.13)
[DELTA]SET [right arrow] P 0.51 (0.48) -9.49

[DELTA]IN [right arrow] [DELTA]SET 2.56 (0.11)
[DELTA]SET [right arrow] 6.88 (0.01) *** -5.72
 OL68\f"Symbol"\s12IN

EX [right arrow] [DELTA]SET 0.97 (0.38)
[DELTA]SET [right arrow] EX 7.89 (0.00) *** -5.09

[DELTA]OP [right arrow] [DELTA]SET 0.55 (0.46)
[DELTA]SET [right arrow] 0.01 (0.93) -3.37
 OL68\f"Symbol"\s12OP

Note Numbers in the parentheses are probabilities of accepting the
null hypotheses of no causality.

*** significant at 1 percent level,

** significant at 5 percent level, and

* significant at 10 percent level.