“Time-series evidence on corporate governance in Thailand: the effect on expected stock returns”

This paper presents an empirical evidence of a time-varying relationship between cor- porate governance and its impacts on stock returns in Thailand. The governance grades assessed by the Thai Institute of Directors are used as governance measurement for the analysis. The parameters estimated by Fama-Macbeth regression indicate that firms with higher governance ratings generate greater expected stock returns in a long run. However, on yearly basis, the positive relationship deteriorates and loses explanatory power in the most of the tested years. The coefficients of governance ratings estimated by fixed effect regression are examined for statistical difference, which confirms that effect of corporate governance on stock returns differs year by year. While there are some distinct years that governance ratings affect stock prices positively, higher governance ratings lead to lower returns in other particular years. The both positive and negative magnitudes of corporate governance’s impact on expected returns do not stay the same over time. Good governance practice at a firm does not always yield positive returns to investors. This study fills the gap by presenting empirical results that show the effects of corporate governance ratings on expected returns are inconstant and time-varying. This paper focuses on Thailand which is home to the second largest stock market in the Southeast Asia after Singapore. The East Asian Financial Crisis in 1997 started in Thailand, and Zhuang et al. (2000) blame poor corporate governance in the region as one of the causes of the crisis. Deficient firm-level governance practices led to the collapse


INTRODUCTION
This study aims to reveal an unstable relationship between corporate governance and its effect on expected stock returns. Prior researches draw contrasting conclusions on the influence of the firms' corporate governance practices on operating performance and valuation. A theory that a good governance leads to better operating profits and increased firm values is empirically proved by Gompers et al. (2003), Drobetz et al. (2004), Bebchuk et al. (2008), and Ammann et al. (2011). On the other hand, a causal relationship between weak protection shareholder rights and lower expected stock returns is denied by Core et al. (2006) and Bhagat and Bolton (2008). Core et al. (2006) further argue that a positive relationship between corporate governance and stock returns is rather period-specific. The results of Bauer et al. (2003) support such argument and the relationship can be negative depending on the tested year. This study fills the gap by presenting empirical results that show the effects of corporate governance ratings on expected returns are inconstant and time-varying. This paper focuses on Thailand which is home to the second largest stock market in the Southeast Asia after Singapore. The East Asian Financial Crisis in 1997 started in Thailand, and Zhuang et al. (2000) blame poor corporate governance in the region as one of the causes of the crisis. Deficient firm-level governance practices led to the collapse 1. LITERATURE REVIEW Gompers et al. (2003) pioneered a holistic approach to analyze a relationship between corporate governance and its effect on stock returns. They constructed the Governance Index, or alternatively called GIM Index, employing 24 provisions that limit or protect shareholder rights. According to the score of the index, firms with the weak shareholder protection are grouped to the Dictatorship portfolio, whereas those with the strong shareholder rights are included in the Democracy portfolio. Their results show that the Democracy portfolio outperformed the Dictatorship portfolio with statistical significance. However, when the sample period is divided into two, the outperformance of the Democracy portfolio loses the statistical significance, which hints that the effect of good governance on expected returns can be period-specific and non-linear. Bebchuk et al. (2008) narrowed down the constituents of the GIM Index and identified 6 governance provisions, labeled as the E-Index (Entrenchment Index), that significantly relates to the negative correlation with firm value and stock returns. Because it is costly to adapt a number of practices, fewer provisions are beneficial to both corporate managers and investors. The E-Index appears to possess explanatory force for abnormal stock returns. The long-short portfolios, which are buying good governance stocks with the E-Index equal to 0 and/or 1 and selling bad governance firms with the E-Index ranging from 3 to 6, generate positive monthly returns in the tested period. The authors do not show the results for breakdown of shorter periods of time, but the positive alpha from the E-Index is statistically significant not only for the period tested in Gompers et al. (2003), but also for the extended period in the early the 2000s.
On the contrary, Bhagat and Bolton (2008) argue that GIM index and the E-Index do not explain abnormal expected stock returns. Once the endogeneity of the relationship among corporate governance, performance, capital structure, and ownership structure is taken into account, the explanatory power of the governance index for stock return disappears. Core et al. (2006) also state that the weak shareholder protection does not cause lower expected returns. Negative influences of poor governance are already priced in the analysts' prediction. Therefore, there should not be any surprise on the day of earning announcement. Moreover, they assert that the positive relationship between corporate governance and excess stock returns of Gompers et al. (2003) is period-specific. In the period of Dotcom clash from 2000 to 2003, the Dictatorship portfolio actually posts higher returns than the Democracy portfolio even after controlling Fama and French's 3-factor (1993) and momentum factor (Carhart, 1997). Similar inconstant relationship between corporate governance and stock returns is found in other markets such as Germany (Drobetz et al., 2004) and the UK and the European market as a whole (Bauer et al., 2003). The results of Ammann et al. (2011) are based on 22 different developed countries and demonstrate some degree of consistency of corporate governance's positive influence on firm value. Nonetheless, their sample data is only 5 years, which can be argued rather short.
Because of weaker legal protections in the emerging markets, firm-level corporate governance may serve as the alternative to inadequate law enforcements (Klapper & Love, 2004). Good corporate governance practices are considered valuable intangible assets, particularly in the emerging markets, since controlling shareholders tend to expropriate from minority shareholders in an environment with less investor protection (La Porta et al., 2002;Nenova, 2003). Durnev and Kim (2005) presents that a mixed result on the relationship between firm value and corporate governance in the US market is due to the effect of legal environment because of the strong legal structure in the US.
A number of studies on the relation between corporate governance and firm value in the emerging market have been conducted, including Korea  . What these prior researches have in common is that a good corporate governance positively affects firm value and operating performance in the emerging markets. However, none of these literatures points out time-varying characteristics of the relation between corporate governance and its effect on stock returns. Perhaps due to insufficient data, most researches in the developing markets are based on single-year cross sectional regression. This study analyzes the relationship using time-series data to examine the presence of time-varying features.

DATA AND METHODOLOGY
The sample universe of this study is all the stocks listed on the Stock Exchange of Thailand (SET). REITs (Real Estate Investment Trust), property funds, and infrastructure funds are excluded. The corporate governance ratings are taken from the "Corporate Governance of Thai Listed Companies", an annual report published by the Thai Institute of Directors (IOD). The IOD examines each listed company on the SET according to a predetermined set of governance criteria and scores them 0-100. The numerical scores are not disclosed, but each company is given the corporate governance stars, as shown in Table 1 required to submit the year-end financial statements within three months from the end of the accounting period. Therefore, in January of year 1, t + all the accounting related information of year t is not available to the public. For this reason, in the regression analysis, the log of total assets, leverage ratio, capital expenditure-to-sales ratio, sales growth ratio, and return on assets are lagged one year. Book-to-market is calculated by dividing market capitalization of a stock in January of year 1 t + by book value of the stock of year Prior researches indicate that some firm-level characteristics are related to equity return and corporate governance. In order to avoid the likely extent of omitted variable bias, several control variables are included in the regression. The log of firms' total assets is included as a control variable because larger firms tend to have a higher rating on the governance level (Gompers et al., 2003;Drobetz et al., 2004). Bigger corporations get more exposure to the public. Hence, they have more incentives to retain a higher standard of corporate governance to prevent scandals, as well as keep a good public image.  2008), firms' leverage ratio, measured by debt-to-assets ratio in this study, is controlled. Leverage possibly affects firms' governance practice , and governance has an influence on firms' access to credit (Bhojraj & Sengupta, 2003). To control endogenous factor of growth opportunities, the past 3-years' sales growth ratio is included in the equation. Drobetz et al. (2004) write that growth firms desire access to capital markets to raise additional funds to keep expanding their business. The level of capital expenditure, estimated by capital expenditure by total assets, is controlled, as Gompers et al. (2003) report that firms with higher governance practices have higher capital expenditure. Return on assets, labeled as ROA, is proxy for profitability, which is likely to have direct effects on stock returns. Among other variables, ROA is more powerful measure of operating performance (Barber & Lyon, 1996). The log of firms' number of years listed on the market, labeled as LnFIRMAGE, is added to the control variables (Bauer et al., 2003, and others). Governance practic-es of older firms are different from those of younger firms . Most studies find that the coefficients of this variable to be negative, because younger firms are more likely to grow faster (Ararat et al., 2016). Book-to-market ratio (Fama & French, 1993) is included in the regression. This ratio remains as one of the most powerful explanatory factors for stock returns. Finally, industry group is controlled, because there is always a possibility that firm's corporate governance rating merely reflects an industry that the firm belongs to (Bebchuk et al., 2008). That is, it happens that better governance ratings are more common in certain industries. Table 2 reports descriptive statistics for all the variables for the full sample period between 2008 and 2015. The mean of the monthly geometric mean stock return is positive at 1.60, while the median is 1.22. The maximum return is 12.84, while the minimum is -7.44. These suggest that stock returns in the sample period are skewed to the right, which is normal for stock return distribution. Although not reported in the table, the skewness of the stock return is 0.48. Among the all control variables, sales growth ratio shows the greatest cross-sectional variation with the standard deviation of 23.77. This might be due to the fact that the sample firms include small cap firms, of which some of those small firms' sales growth can be unpredictably high or low in some years.

RESULTS
The results of the equation (1) are reported in Table  3. The results for industry dummy variables are omitted from the table. The time-series mean of the dummy variables for Excellent and Very Good take positive significant coefficients, whereas that for Good does not, suggesting that only CG ratings with Very Good or higher can lead to higher expected return. However, the dummy variables for Excellent do not get statistical significance in the cross-sectional regressions. There are even two years that the dummy for Excellent gets negative coefficients without significance. The Very Good dummy gets the positive significance only in 2008. The outcome confirms inconsistent influence of the CG ratings on the stock returns. Receiving a high score on the CG ratings does not always result in higher returns.
Besides the CG dummy variables, ROA and bookto-market show significant positive coefficients in the time-series mean. The coefficient of BM is 0.946 with strong significance, which indicates powerful effect on stock return.
The results of cross-sectional regressions and time-series mean in Table 3 show that the effect of the CG ratings is not constant but rather time-varying. However, that assumption was made based on only 8 years of data, which is not sufficient enough to conclude the time-varying characteristics. Alternatively, parameters of the CG ratings are estimated using fixed effect regression and then the parameters are compared by z-test for difference. Table 4 reports the results of the fixed effects regression. Z-test is conducted for a null hypothesis that there is no difference in the coefficients between any pair of parameters.      LnASSETS is log of assets, / CAPEX ASSETS is capital expenditure to assets, SALESG is an average past 3-year sales growth, ROA is return on assets, LnFIRMAGE is log of firm age, and BM is book-to-market, ID is industry dummy variable. Industry grouping is according to the Industry Classification Benchmark (ICB). There are actually 10 industry groups, but the results for each industry dummy is omitted from this table. The data are from 2008 to 2015. Standard error for the coefficients are in the in the parentheses. *, **, and *** indicate significance at 0.05, 0.01, and 0.0001 level, respectively. Z-test is for a null hypothesis that there is no difference in coefficients. P-values are in the parentheses. *, **, and *** indicate significance at 0.05, 0.01, and 0.001 level, respectively.

CONCLUSION
Since Gompers et al. (2003) first explored the field of the corporate governance index and its effect on increased firm value, a number of researches have been conducted. The literature generally supports the theory that better governance practices positively affect the firms' value creation. Nevertheless, some scholars, including Core et al. (2006), propose opposing views. In a broad sense, the results of this study support both pros and cons of the positive relationship. This research fills the gap and claims that the relationship between the corporate governance and its effect on stock returns is non-linear and time-varying.