Emerging Asia Equity Home Bias and Financial Integration

Abstract

Equity home bias remains a phenomenon and a puzzle. Recent studies show the importance of financial integration in explaining the observed decline of equity home bias in advanced economies. This paper takes a step in understanding this relationship in the context of Emerging Asia. Stock market ratios and the mean-variance approach are used to construct measures of equity home bias; while foreign direct investments and time-varying global betas are used to derive measures of financial integration. These measures provide evidence that equity home bias has declined in recent years and progress has been made toward greater financial integration in the region. Fixed-effects panel regression was used to determine whether the factors that contribute to the decline of the bias in advanced economies – including financial integration – are relevant for Emerging Asia. Results show that a higher initial level of equity home bias and a greater financial integration lower the bias; while a larger stock market raises it. These findings concur with those for advanced economies. As in advanced economies, better quality of institutions and larger bank assets generally lower equity home bias, although insignificantly. However, unlike in advanced economies, country-specific risks are important in explaining the decline of the bias in Emerging Asia.

Keywords:

• Portfolio equity home bias
• financial integration
• and emerging Asia

Acknowledgements

The author would like to thank Daniel Fricke of Kiel Institute for World Economy for research guidance and comments; Prof. Dr. Vasyl Golosnoy of the Institute of Statistics and Econometrics, Kiel University for valuable suggestions; and Hsio Chink Tang of the Asian Development Bank for suggesting further improvements on the paper. The author also highly appreciates the comments and suggestions of Damaris Yarcia, Benjamin Radoc, Maria Theresa Baguisi, and Maria Vida Dacumos.

Notes

¹Throughout this paper, Emerging Asia refers to a group of fast growing economies in Asia. It includes the People's Republic of China (China); Hong Kong, China; India; Indonesia; Republic of Korea (Korea); Malaysia; Philippines; Singapore; Taipei, China; and Thailand. For the purposes of comparing intra-regional trends, this paper groups Hong Kong, China; Korea; Singapore; and Taipei, China as newly-industrialized economies (NIEs); and Indonesia; Malaysia; Philippines; and Thailand as Association of Southeast Asian Nations (ASEAN-4). China and India are presented separately. The sample period begins around the time when capital account liberalization started in the region.

²Advanced economies include country members of the European Monetary Union, Japan, the United Kingdom, and the United States.

³No separate section is provided for the literature review. Discussions on theories, empirical issues, and findings related to equity home bias are included in Sections 2 and 4, while those for financial integration are in Section 3. A literature survey on the determinants of equity home bias is presented in Section 4.

⁴Stock price returns are calculated as the difference between current period stock price and the previous stock price in natural logarithm form.

⁵The first period (1990–1996) covers the period of rapid capital account liberalization in the region. The second period (1997–2001) includes the Asian financial crisis and the global dot.com bust of 2000/2001. The third period (2002–2007) refers to the years of surging capital flows in the region. The last period (2008–2010) covers the recent global economic and financial crisis of 2008/2009 and recovery. This paper uses the same period groupings in other sections for purposes of comparison. To smooth some of the annual observations for the optimal foreign portfolio weights calculated using the mean-variance approach, the averages of preceding and succeeding years were used.

⁶Both ASEAN-4 and NIEs showed a decline in equity home bias in Figure 3 for the period 1997–2001, reflecting the 1997/1998 Asia financial crisis and the 2000/2001 drop in technology-based stocks.

⁷There appears to be a weak correlation between the two computed measures of equity home bias. The correlation ranges from 0.6211 for Hong Kong and China to as low as 0.1076 for Malaysia. The average correlation for Emerging Asia is 0.3690, which is rather low. Therefore, it cannot be stated that both measures show similar patterns for Emerging Asia.

⁸The net capital outflow of about 1.8% of GDP during the Asian financial crisis is far larger than the net outflow of around 0.3% of GDP the region had during the recent global financial crisis in 2008/2009.

⁹This agreement among the 10 members of the Association for Southeast Asian Nations (ASEAN) with China, Korea, and Japan involves bilateral currency swaps and pooling of foreign exchange reserves as a safeguard against speculative attacks on local currencies during times of crises.

¹⁰Cross-market dispersion of weekly equity returns was computed using Wednesday average price of benchmark stock price index of selected economies, including those from Emerging Asia. Weekly returns were derived as the difference between current and previous week's prices in natural logarithm form. The standard deviation of weekly returns was then calculated and the values were smoothed by taking the six-month moving average values. The World Series includes the 10 countries of Emerging Asia, United States, Japan, United Kingdom, Switzerland, Germany, France, Italy, Spain, Netherlands, Australia, Canada, Brazil, Mexico, Peru, and Russia.

¹¹Another reason for focusing on global financial integration (and not regional integration) is the absence of cross-border equity asset holdings for China, and Taipei, China, which will reduce the number of countries in the sample.

¹²The conditional variance of the error terms is assumed to follow the GARCH (1,1) process.

¹³To estimate the global market betas, a weekly benchmark stock price index for each of the ten Emerging Asian economies was used. For the global stock price index, the MSCI All County (AC) World Index was used as it includes both advanced and emerging markets. The weekly returns are calculated as the difference between the present week's stock price and the previous week's stock price both in natural logarithm form.

¹⁴Regression results for equation (6) report very low R-square and the estimated coefficients are rarely significant. This suggests model misspecification due to omitted variables that could influence domestic equity returns.

¹⁵The authors proposed to use β -convergence as a measure of the speed of adjustment of deviations of countries to the long-run benchmark value. Their regression model estimated the change in interest rate spreads of European countries (over Germany) on the previous year's level interest rate. The same approach is adopted in equation (8).

¹⁶The panel dataset that will be used to estimate equation (8) contains n=10 countries, and T=21 observations for the years 1990 to 2010. The total number of observations is nT=210. The dataset is fairly large and strongly balanced as all countries are expected to have values in all time periods for each variable.

¹⁷Institutional Quality index refers to the simple average of the World Governance Indicators developed by Kaufmann, Kraay, and Mastruzzi, which includes measures for voice and accountability, political stability and absence of violence, government effectiveness, regulatory quality, rule of law, and control of corruption.

¹⁸The real foreign exchange rate was derived as the product of nominal foreign exchange rate (expressed in local currency unit per US dollar) and the price index ratio of the US consumer price index and domestic consumer price index.

¹⁹A test was conducted to check whether pooled regression is more appropriate. The test ran a pooled regression as specified in equation (8) and plotted the residuals. The scatter plot of the residuals across countries reveals heterogeneity among the error terms, suggesting the panel regression estimation is more appropriate.

²⁰Both measures of equity home bias also exhibit time variation; however, it is not as large as country variation. Hence, a one-way fixed-effects model is used.

²¹The p-values for Wooldridge's test for panel data serial correlation reject the null hypothesis that there is no first-order autocorrelation.

²²Feasible generalized least squares is equivalent to applying OLS to transformed data with the effect of standardizing the error term and de-correlating the data with regressors.

²³The Levin-Lin-Chu panel unit root test is preferred for moderate-sized panels with around 10 to 20 cross-sections and 20 to 25 observations per cross-section.

²⁴Robust standard errors reported in Tables 3 and 4 refer to the Huber-White robust variance estimator used against heteroskedasticity.

²⁵To check the robustness of the results, the crisis years of 1997 and 2008 were jointly removed, i.e. equation (8) is re-estimated dropping 1997 and 2008 observations.

²⁶The volatility of the real exchange rate used in the regression model involves two domestic macroeconomic risks, namely foreign exchange rate and inflation, to which Emerging Asia has been very vulnerable compared with advanced economies.