J. Japanese Int.

Economies 23 (2009) 149176

Contents lists available at ScienceDirect

Journal of The Japanese and International Economies

Corporate governance reform in emerging markets: How much, why, and with what effects?
Sudarat Ananchotikul a , Barry Eichengreen b,
a b

Bank of Thailand, Thailand University of California, Berkeley, Department of Economics, 549 Evans Hall #3880, Berkeley, CA, United States

a r t i c l e

i n f o

a b s t r a c t
Ananchotikul, Sudarat, and Eichengreen, BarryCorporate governance reform in emerging markets: How much, why, and with what effects? This paper extends previous measures of the quality of corporate governance for a cross section of emerging markets. We nd that there have been signicant improvements in a wide range of countries in the course of the last ten years. Improvements are most likely in countries with stable governments, where foreign investors lobby for reform, and where peer effects are present. We also document signicant benets in terms of the depth and liquidity of nancial markets. J. Japanese Int. Economies 23 (2) (2009) 149 176. Bank of Thailand, Thailand; University of California, Berkeley, Department of Economics, 549 Evans Hall #3880, Berkeley, CA, United States. 2009 Elsevier Inc. All rights reserved.

Article history: Received 5 February 2008 Revised 12 January 2009 Available online 21 February 2009 JEL classication: G0 L0 Keywords: Corporate governance Asia

1. Introduction In the ten years since the Asian crisis, the economic and nancial landscape in emerging markets has been transformed. Large current account decits have been eliminated. Dependence on shortterm foreign borrowing has been curtailed while central banks have accumulated foreign-exchange reserves to better bullet-proof their economies from ckle capital ows. Exchange rates have become more exible. Budget decits have come down and with them debt-to-GDP ratios and ination.

Bank of Thailand and University of California, Berkeley, respectively. We thank Kenichi Ueda for advice regarding data and Pipat Leungnaruemitchai for help in obtaining it. Corresponding author. E-mail address: [email protected] (B. Eichengreen).

0889-1583/$ see front matter doi:10.1016/j.jjie.2009.01.005

2009 Elsevier Inc.

All rights reserved.

150

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

Countries have strengthened their nancial markets, developing the infrastructure needed to encourage issuance and investment in long-term, xed-rate, domestic-currency-denominated debt securities. Equity markets have become more liquid and drawn growing attention from international investors. All this is quite different from ten years ago. One should not allow self-congratulation to breed complacency. It is easy enough to point to emerging markets where one or more items on the preceding list are missing. In addition, there are grounds for wondering how much of this progress is merely cyclical whether, as global growth slows, budget decits, ination, nancial fragility and the problems with which they are associated will reemerge. Inevitably a critical look yields something of a glass-half-empty, glass-half-full story. Still, it is hard to deny that important progress has been made in strengthening institutions and policies. Why has that progress not been faster? Why have some countries reformed more comprehensively than others? And have reforms had the expected payoff in terms of economic performance? In this paper we attempt to answer these questions for the case for corporate governance reform. This is an appropriate case study for several reasons. First, the shortcomings of corporate governance were greatly emphasized in ocial post mortems on the Asian crisis.1 These blamed principal-agent problems for the extensive leverage and dependence on short-term foreign-currency-denominated debt that rendered the corporate and nancial sectors so fragile. They pointed to inaccurate information about corporate nances as an explanation for why investors scrambled out of Asian markets. In turn this led the World Bank, IMF, OECD and BIS, among others, to stress the need for corporate governance reform.2 Second, effective corporate governance is not something that can be legislated. Rather, it emerges from the interplay of the public and private sectors. Regulators can establish guidelines, but how decisions are reached and how those taking them are held accountable depends on how rms implement those decisions and on how investors react. Corporate governance thus epitomizes the challenges of reform in a world where outcomes depend market reactions as well as ocial decisions.3 Third, high-prole management scandals in the U.S., such as Enron and Worldcom, and some of the fallout from the subprime crisis are reminders that even countries with relatively sophisticated nancial markets have corporate governance problems. Insofar as the United States cannot solve those problems, one might reasonably ask whether it is realistic to ask emerging markets to do so. Fourth, there is disagreement on how best to provide corporate governance and therefore on what reforms are desirable. Even among advanced countries with relatively sophisticated nancial markets, the United States and Europe in particular, there is disagreement on the specics of corporate governance reform. In part this reects different analytical perspectives, but in part it results from differences in economic structure for example, that bank-based nancial systems are still more prevalent in Europe.4 In the emerging-market context there similarly are questions about the suitability of one-sizets-all governance reform. Legislation and regulations tailored to the circumstances of high-income countries may yield less satisfactory results where the information environment is underdeveloped, cross shareholding is common, and family control is pervasive. Governance that relies on the accurate and timely provision of accounting information may be ill suited to economies where the supply of independent accountants and auditors is limited and the accounting professions self-regulatory body is weak.5 Attempting to prevent management from pursuing private agendas by giving large shareholders more power may not work in a setting where there the majority owner is the manager and

1 See for example World Bank (1998). This emphasis was to the exclusion of other factors, to be sure. But this emphasis on corporate governance problems was what distinguished accounts of the Asian crisis from analysis of its predecessors. 2 A representative compendium (of OECD work on reforming corporate governance) is at http://www.oecd.org/topic/ 0,2686,en_2649_37439_1_1_1_1_37439,00.html. 3 One recent study for the Asian economies (Cheung and Jang, 2005) actually reports a negative correlation between how countries rank in terms of the adequacy of corporate governance rules and regulations and the adequacy of actual practice as seen by investors. 4 On this, see Enriques and Volpin (2007). 5 See Alba et al. (1998).

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

151

the problem instead is the exploitation of outside shareholders by large block-holders.6 Observations like these prompt questions about the suitability for emerging markets of advanced-country models. In Asia they have fueled calls for a distinct approach to corporate governance.7 Our analysis of these issues makes use of the work of de Nicolo et al. (2006). These authors (NLU for short) have constructed outcome-based measures of the quality of corporate governance for a wide sample of countries for the period 19942003. Outcome based means looking at what rms and markets actually do. As a measure of rm behavior, the authors look to rms release of accounting information on items recognized as important by professional accountants. A second measure of rm behavior is earnings opacity, namely the extent to which managers conceal true performance by using accruals to limit the uctuation of prots. As their measure of market behavior, NLU compute the co-movement of the prices of the shares of different companies in the same country, on the grounds that co-movement will be higher when management is less transparent about the nancial condition and prospects of individual rms. Corporate governance quality is then the average of these indicators. In principle this index should measure how well countries are doing in strengthening corporate governance, even if the specic measures appropriate for doing so differ as a function of the structure of the economy and its nancial markets.8 In this paper we extend these measures through 2005 and use them to analyze the determinants and effects of corporate governance reform at the country level. We look rst at the economic and political determinants of corporate governance quality. This helps us to understand why some countries have stronger corporate governance than others. It speaks to the notion that this problem will solve itself in the course of economic growth and to the idea that quick xes ignore the fact that the development of corporate governance is an organic part of the larger process of economic and nancial development. In addition, these estimates point to instrumental variables with which to analyze the impact of corporate governance on nancial activity and development. Policies fostering the development of capital markets, including corporate governance reform, have attracted considerable attention in the literatures on economic development and nancial instability. Yet few analyses have acknowledged the obvious problem that the quality of institutional arrangements not only affects but also is affected by nancial development. Insofar as our analysis of the determinants of corporate governance quality points to variables that are themselves exogenous and unlikely to independently inuence nancial depth and liquidity, it suggests instrumental variables that can be used to better pin down the impact of corporate governance on larger outcomes. The results provide some evidence, mainly for Asia, that corporate governance improves with economic development. In addition they point to some specic circumstances that appear to facilitate the development of strong corporate governance practice. Corporate governance improves with the stability and development of the political system, as if governments that expect to remain in power are readier to sink the costs of reforms that only pay off down the road, and as if investors are better able to effectively communicate their interest in corporate governance reform in countries with welldeveloped political systems. There is also evidence that countries where foreign investors are more prominent push for improvements in corporate governance. Finally, corporate governance is stronger in countries with a common law tradition, where shareholders are likely to be more active and better able to represent their interests. Using these political variables as instruments for corporate governance, we nd that corporate governance quality has a positive impact on private bond market capitalization, stock market capi-

These problems of self-dealing or tunneling are the focus of Johnson et al. (2000). See the discussion in Khan (2003). The authors argument is that in Asian countries where corporations are heavily controlled by insiders often members of the founding family banking systems need to be strengthened to provide a counterweight and deal effectively with agency and monitoring problems. The implication is that corporate governance in Asia should focus more on strengthening banking systems and less on the release of accounting information and activism on the part of individual shareholders. Similarly, Hofstetter (2005) argues that rules developed to address the agency risks of companies with dispersed ownership do not necessarily have the same merits for the case of family-controlled companies. The author criticizes U.S.-like mandatory bid rules that undercut the potential eciency of control premiums and suggests default rules allowing for opt-out solutions that offer necessary exibility in the context of insider-controlled companies. 8 We will have more to say about the strengths and weaknesses of this measure of corporate governance quality.
7

152

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

Table 1 Evolution of corporate governance quality (CGQ), 19952005. 1995 All Asia Latin America Europe Others Emerging Advanced Memorandum: United States Japan 58.7 57.0 52.4 60.9 62.8 55.4 61.0 72.2 57.2 1996 60.2 58.8 54.7 62.2 63.3 57.5 62.1 72.6 59.3 1997 58.0 57.7 50.0 59.9 62.9 54.5 60.5 72.3 59.8 1998 58.9 58.2 53.2 60.0 63.6 56.2 60.8 71.9 62.0 1999 60.7 59.7 55.4 61.9 65.3 57.5 63.0 74.8 64.2 2000 61.1 60.7 56.7 61.7 65.7 58.3 63.2 77.7 63.8 2001 61.0 59.9 54.3 61.6 69.0 57.4 63.5 76.5 62.9 2002 63.2 60.7 58.9 64.6 68.6 59.3 65.9 76.7 65.7 2003 63.9 62.2 59.1 65.2 69.0 60.4 66.4 74.6 64.0 2004 64.7 61.3 60.6 65.9 72.0 60.9 67.3 79.8 64.4 2005 65.2 61.7 62.4 65.5 74.0 61.6 67.8 79.9 66.4

Note: See Appendix Table A2 for grouping of sample countries by region. Source: Corporate governance indices from 19952003 from de Nicolo et al. (2006), extended through 2005 by authors.

talization, the number of listed companies, and the turnover ratio on the stock market but not, plausibly, for public bond market capitalization. Our results thus support the notion that corporate governance reform can make a difference for nancial development. In other words, even if nancial development also affects the quality of corporate governance, as is plausible, this is not all that is going on. 2. The evolution of corporate governance The NLU indicator of corporate governance quality, as mentioned, has three components. First, the share of the 40 most important accounting items, as identied by the Center for International Financial Analysis and Research of Princeton University, on which the largest companies (top ten manufacturing companies as measured by total assets in each country) disclose information. Second, a measure of earnings smoothing, constructed as one minus the Spearman rank correlation between changes in estimated operating cash ow and changes in accounting accruals (both normalized by total assets) for each country and year.9 Third, a measure of stock price synchronicity constructed from the covariation of each rms weekly return with the market capitalization-weighted weekly return. The overall index and its components are available for 41 countries, including 19 emerging markets (ten in Asia), annually for the period 19942003. Using these authors sources and following their methods, we updated the three subindices through 2005. Like any summary of something as multi-faceted as corporate governance, this measure is not without its limitations. The number of items on which rms disclose accounting information tells us nothing about the accuracy of their information. Estimating earnings smoothing is more dicult for emerging markets than advanced countries because of limited information on cash ow and the need to use an accruals-based proxy instead. Finally, individual stock prices can move together to a greater or lesser extent for reasons other than the limited availability of information on individual rms outcomes and prospects, for example because of changes in the prevalence of common shocks. Still, this outcome-based measure tells us more about what rms do than the statute- and regulationbased alternatives. And it has strengths relative to its predecessors, for example that it does not focus exclusively on share price co-movements (as in Morck et al., 2000) or accounting practice (as in Cheung and Jang, 2005). The evolution of corporate governance for the full sample, individual regions, and emerging and advanced countries is shown in Table 1 and Fig. 1. There appear to have been improvements in corporate governance both Asia and Latin America, although progress may have been a bit slower in

9 This captures the degree to which insiders reduce the uctuation of reported earnings by altering the accounting component of earnings, namely accruals. A high correlation between year-to-year changes in accounting accruals and year-to-year changes in operating cash ows indicates use of accounting discretion to conceal poor current performance.

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

153

Fig. 1. CGQ index by region.

the Asian case.10 There is relatively little improvement (in the Latin case, even retrogression) prior to 1998 but faster progress subsequently; this suggests that subsequent reforms were prompted not merely by the Asian crisis which was of course felt most intensely in Asia but by the general push by the multilaterals and more generally by the pressures of nancial globalization.11 There have been improvements in the quality of corporate governance as measured by this index in the majority of Asian countries (Fig. 2). These are most dramatic in Hong Kong, Singapore and, interestingly, Malaysia, Thailand and the Philippines. But this improvement has been only marginal in the case of Indonesia, where there has been some improvement in accounting standards, while earnings smoothing and stock price synchronicity, reecting continued low levels of transparency, show little if any improvement.12 There appears to have been deterioration in the case of China, where there has been little evident improvement in accounting practice and stock price co-movements have risen further.13 To see what is driving these trends, we look at the components of the CGQ index separately in Fig. 3. In Asia, the improvement appears to be driven mainly by increases in accounting transparency and a rising correlation between the accruals and cash-ow proxies. Stock price synchronicity is relatively volatile over time, and it shows no discernible movement one way or the other. In Latin America, in contrast, the rise in the CGQ index is driven mainly by a decline over time in stock price synchronicity. Given that co-movements in stock prices on national markets may also reect other things (such as the nature of country-specic shocks), one may want to be more skeptical about this case.14 In any case, it will be important to examine the robustness of results by examining the im-

10 We suggest some reasons below to regard the Latin American results with special caution. Fortunately, our focus in this paper is on the Asian case. 11 We report some further evidence on this below, where we show that corporate governance depends more on pressure from foreign investors and appears to have been affected less by the incidence of nancial crises. 12 The Indonesian Accounting Standards Board and the Indonesian Institute of Accountants have issued some 50-plus standards compatible with international practice, but this does not mean that rms always disclose the requisite information on this basis. For example, none of the Indonesian rms in our sample reported accounting information on Research and Development (Expenses), the Unconsolidated Parent Companys Net Prot, Funds from/for Other Operating Activities, or Foreign Currency Translation Gain/Loss in the most recent year. 13 While recent uctuations in the Shanghai stock market are outside the sample period, observers will not be surprised by this last result. 14 In addition, as we point out below, we have a more limited number of observations for Latin America than for Asia which makes it hard to use regression analysis to test for the signicance of such patterns.

154

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

(a)

(b) Fig. 2. CGQ index in Asia.

pact not just of the overall CGQ index on variables of interest but also the effects of the individual components (in Section 5 below). 3. Determinants The next question is why progress has been more rapid in some countries than others. To answer it, we regress corporate governance quality on a vector of country-specic economic and political characteristics. Information on the denition of the variables can be found in Appendix Table A1. The equations are estimated with random effects; the Hausman and BreuschPagan tests show that random effects are preferred to xed effects and simple pooling.15

15

The key results carry over when we estimate these relationships instead using xed effects.

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

155

(a)

(b) Fig. 3. Evolution of the three components of the CGQ index by region, 19952005.

The benchmark specication is in column 1 of Table 4. Countries with more stable governments have better corporate governance, reecting the greater willingness of politicians to sink the up-front costs of reform that only yield returns down the road. It is less clear whether more or less democratic government (as measured by Polity) makes a difference; while this variable is signicantly positive in the parsimonious specication of column 1, it goes to zero when additional country characteristics are added in column 2.16 Past capital inows scaled by GDP translate into stronger present corporate governance, as if foreign investors press for improvements. Countries with a common law tradition have better corporate governance, reecting the stronger rights and voice of outside shareholders and, presumably, greater activism. Countries with low lend-

16 A zero coecient is consistent with the literature on whether democracies are more or less able and likely to undertake economic and nancial reforms which is perhaps best characterized as inconclusive. For discussion of these and other links running from the political system to corporate governance, see Roe (2003).

156

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

(c) Fig. 3. (continued) Table 2 Summary statistics. Variable Corporate governance quality (0 to 100) Accounting standards (0 to 100) Earnings smoothing (0 to 100) Stock price synchronicity (0 to 100) Private bond market capitalization (% GDP) Public bond market capitalization (% GDP) Stock market capitalization (% GDP) Number of listed companies Stock turnover ratio Government stability index (0 to 12) Polity index (10 to 10) Cumulative capital inows (% GDP) English legal origin dummy Real GDP per capita (in log) Domestic credit by banking sector (% GDP) Lending interest rates Number of parent enterprises Financial openness Trade openness Exchange rate stability index (0 to 10) Cumulative years under IMF programs Currency crisis indicator Banking crisis indicator Obs 451 435 451 451 439 439 451 451 451 451 451 451 451 451 451 451 451 451 451 451 451 451 451 Mean 61.42 84.21 19.22 82.25 24.86 36.47 77.88 858 72.41 8.73 7.74 33.04 0.34 8.84 97.16 13.02 1483 7.03 77.56 8.96 1.34 0.04 0.12 Std. dev. 7.17 4.06 12.04 10.35 26.42 24.86 72.51 1428 68.74 1.82 4.07 40.42 0.47 1.92 45.97 14.76 2097 10.23 56.73 1.82 3.79 0.20 0.32 Min 28.60 66.30 0.00 38.50 0.00 0.21 3.61 50 1.31 4 7 0.63 0 0.12 8.58 1.68 0 0.07 16.30 0 0 0 0 Max 89.20 93.35 101.01 96.00 145.62 147.89 566.18 8851 623.59 12 10 299.88 1 10.90 258.50 103.30 9356 96.38 383.06 10 23 1 1

Note: Data on private and public bond market capitalization are unavailable for Israel throughout the sample period (1995 2005).

ing rates also appear to have relatively strong corporate governance. Low lending rates may reect stable economic conditions, enabling the government to devote more time and resources to corporate governance improvement rather than to other more pressing economic problems. Doidge et al. (2004) observe that where external nance is more readily available, the incentives are stronger for rms to improve corporate governance. Finally, the number of parent enterprises (multinational enterprises with subsidiaries abroad) enters positively in the benchmark specication. The number of parent enterprises is a proxy for the number of large corporations, which in turn reects the level of corporate sector development. Countries with more large corporations could feel a greater need to

Table 3 Correlation matrix. Corporate governance quality Government stability Polity index Cumulative capital inows (% GDP) Log real GDP per capita English legal origin dummy Domestic credit by banks (% GDP) Lending rate (%) Number of parent enterprises Financial openness Trade openness Exchange rate stability Banking crisis dummy Years under IMF programs
** ***

Government stability

Polity index

Cumulative Log real capital GDP per inows capita (% GDP)

English legal origin dummy

Domestic credit by banks (% GDP)

Lending rate (%)

Number of parent enterprises

Financial openness

Trade openness

Exchange rate stability

Banking crisis dummy

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

0.112** 0.262*** 0.126*** 0.328

1.000

0.154*** 0.194***
0.022 0.155*** 0.198***

1.000 0.147*** 0.388

1.000 0.249*** 0.308*** 0.212*** 1.000 0.073 0.495*** 1.000 0.155*** 1.000

0.242*** 0.310***

0.094**
0.185***

0.274 0.213***

***

0.152 0.002

***

0.098 0.182***
**

0.187 0.145***
***

0.558 0.285***
***

0.187 0.267***
***

0.414*** 0.338***
0.226*** 0.285*** 0.185***

1.000 0.163***

1.000 0.009 0.119** 0.056 1.000 0.546*** 0.104**

0.123*** 0.078 0.126***

0.199*** 0.152*** 0.059

0.011 0.196*** 0.005 0.195*** 0.140

0.607*** 0.805*** 0.118**

0.289*** 0.190*** 0.388***

0.082 0.324*** 0.150***

0.192*** 0.254*** 0.538***

1.000 0.131*** 0.019

1.000

0.203*** 0.162
***

0.016 0.057

0.109** 0.124
***

0.311*** 0.335
***

0.074 0.150
***

0.043 0.366
***

0.013 0.198
***

0.138*** 0.144
***

0.196***
***

1.000 0.173***

0.183

0.035

indicates signicance level at 5%. indicates signicance level at 1%.

157

158

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

place them under rigorous governance standards and may also feel that rms are better able to comply with stringent regulation. Conversely, it is often said that some countries hesitate to enforce rigid governance standards because their many small- and medium-sized companies would nd it costly to comply with the rules. A number of other variables do not appear to signicantly inuence the quality of corporate governance.17 These include GDP per capita, as a measure of aspects of general economic and nancial development not adequately captured by other variables, although this specic result is sensitive to sample, as we explain below. Other insignicant variables include domestic credit provided by the banking sector, included on the grounds that an active banking sector is sometimes identied as pushing for improved corporate governance; the incidence of recent banking crises; the stability of the exchange rate, and the number of recent years in which the country was under an IMF program.18 Another interesting result is the negative coecient on the dummy variable for Asia. That is, after controlling for more than a dozen economic and political characteristics, the quality of corporate governance in Asia continues to lag behind that in other countries. Since the majority of non-Asian countries in the sample are advanced economies, this would appear to reect the differential between Asia and that grouping. To shed further light on this, we estimated the same equation separately for Asian and non-Asian countries. The results are in columns 3 and 4 of Table 4. English legal origin continues to be associated with stronger corporate governance, and past capital inows are positively associated with corporate governance quality. Government stability is positively associated with the quality of corporate governance in both subsamples, although not signicantly, perhaps reecting fewer degrees of freedom. There are also a number of interesting differences between Asia and other regions. The zero coecient on credit provided by the banking sector in the full-sample estimates appears to conate the strong negative effect of this variable in Asia with a weak positive effect in other regions. It does not appear that banks are active agents for improvements in corporate governance in Asia; this may reect the tendency for banks to be connected to the enterprises to which they lend and to be agents of the governments development strategy more than independent investors, something that has been emphasized in accounts of the Asian crisis.19 In addition, whereas the number of large enterprises entered positively in the full sample, it enters negatively in the Asian subsample. In Asia large enterprises are often owned by business groups controlled by wealthy families. These enterprises are reluctant to disclose information to outsiders and use their political connections to lobby against positive changes in governance practice. Where nancial openness has no effect on corporate governance quality in the full sample, it has a signicant negative effect in the Asian subsample. This is consistent with rm-level evidence for Thailand in Ananchotikul (2007), who shows that when foreign investors become insiders they may not press for greater transparency and management accountability. In practice, this negative association of nancial openness with corporate governance quality is driven by the observations for China and Indonesia (which have opened nancially but not seen notable improvements in corporate governance) and Hong Kong (which may be mirroring Chinas inuence).

17 That is, they enter with coecients insignicantly different from zero in the multivariate regressions. However, as shown in the correlation matrix (Table 3), each of them has statistically signicant correlation with the measure of corporate governance quality. 18 Any tendency for IMF tutelage to lead to improved corporate governance appears to be neutralized in the aggregate by cases of countries that were continuously under Fund programs and in which transparency problems were rife. This variable is cumulative years under an IMF program where we start counting in 1960. Its maximum value is 23, this observation belonging to Argentina, which was continuously under IMF programs from 1983 through 2005. Similarly, while it is sometimes suggested that institutional strengthening (including better corporate governance) is a prerequisite for moving to greater exchange rate exibility (since rms then must limit currency mismatches and other exchange-rate-related balance-sheet risks), this effect seems to be neutralized in the aggregate by the tendency for some countries with weak governance to exhibit relatively high levels of currency instability. 19 See e.g. Goldstein (1998).

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

159

Table 4 Determinants of corporate governance quality (rst stage, GLS random effects). Dependent variable: corporate governance quality (CGQ) Full sample (1) Excluded instruments: Government stability Polity index Cumulative capital inows (% GDP) Included instruments: Log GDP per capita English legal origin dummy Domestic credit provided by banking sector (% GDP) Lending interest rate (%) Number of parent enterprises Financial openness Trade openness Exchange rate stability Dummy for banking crisis in previous year Years under IMF programs Dummy for Asia Constant 55.991 (36.18) 47.10 0.000 451 41 0.073 0.234* (1.87) 0.247** (2.20) 0.046*** (5.13) (2) 0.130 (1.01) 0.063 (0.53) 0.039*** (3.40) Asia (3) 0.075 (0.39) 0.337*** (5.73) 0.033*** (2.92) 1.260*** (4.40) 2.465*** (3.13) 0.049*** (5.32) 0.180 (1.61) 0.001*** (4.35) 0.096** (2.32) 0.002 (0.21) 0.300 (0.88) 0.556 (0.69) 0.069 (0.79) Non-Asia (4) 0.059 (0.34) 0.018 (0.03) 0.074*** (2.81)

0.608 (1.43) 4.300** (2.30) 0.001 (0.06) 0.140*** (2.73) 0.001*** (3.53) 0.026 (1.37) 0.013 (0.96) 0.136 (0.66) 0.105 (0.15) 0.004 (0.03) 5.564*** (3.21) 64.424 (12.29)
15.67 0.001 451 41 0.203

0.455
(0.93) 5.827*** (2.66) 0.022* (1.67) 0.135** (2.43) 0.001*** (3.21) 0.010 (0.46) 0.012 (0.37) 0.184 (0.73) 1.461 (1.32) 0.101 (0.37)

59.642 (11.37) 45.48 0.000 121 11 0.498

62.102 (7.97) 9.87 0.020 330 30 0.263

F-statistics for excluded instruments p-value Observations Number of country ID R2

Notes: Heteroskedasticity-consistent z-statistics in parentheses. * indicates signicance level at 10%. ** indicates signicance level at 5%. *** indicates signicance level at 1%.

Finally, per capita GDP matters for Asia, as if there are additional differences in the region, presumably associated with the general level of economic development, not adequately captured by the other explanatory variables something that does not appear to be true of the rest of the sample (as noted above). In sum, corporate governance quality varies across countries for both systematic and idiosyncratic reasons. Systematic reasons include the legal system and how effectively it empowers outsiders to lobby for information disclosure and representation; the presence of foreign investors, who are likely to lobby for improvements in corporate governance; and political stability and development, which inuences the readiness of government to invest in governance reform. In Asia, in addition, the dominance of bank nance and the number of parent enterprises are negatively associated with corporate governance quality.

160

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

(a)

(b) Fig. 4. CGQ index and nancial development pooled sample, 19952005.

4. Effects We now examine the impact of corporate governance quality on nancial development. Although there is no question about the existence of a relationship (Fig. 4), one can question the direction of causality. We treat corporate governance as endogenous, recognizing that its quality can be affected by as well as affecting nancial development. The net effect of this inuence running in the opposite direction is uncertain a priori. On the one hand, as nancial markets grow their participants will press, out of self-interest, for institutional reform, and being more numerous they may be correspondingly more likely to succeed; this suggests

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

161

(c)

(d) Fig. 4. (continued)

a positive inuence running from nancial development to corporate governance quality. In addition, both rms and regulators in more nancially-advanced economies may be able to afford more demanding corporate governance practices, which are costly to implement and comply with. On the other hand, countries where nancial development is lagging may have particular incentive to raise corporate governance quality in order to jump-start their markets; this would point to a negative relationship between the two variables. We construct the tted value of corporate governance using all the exogenous variables in the second stage as included instruments and the political variables and our measures of the presence

162

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

(e) Fig. 4. (continued)

of foreign investors as excluded instruments.20 We are not aware of previous arguments or evidence that these particular political variables are important for nancial development which is why we are comfortable about omitting them from the second stage and using them as instruments for corporate governance.21 Omitting cumulative capital ows from the second stage is likely to be more controversial. Fortunately, dropping this variable from our list of instruments does not alter our key results, as we show below.22 In the second stage regressions we consider the impact of corporate governance, along with a vector of controls, on the capitalization of private bond markets (as a percent of GDP), government bond markets (as a percent of GDP), stock market capitalization (as a percent of GDP), the number of listed companies, and the turnover ratio on the stock market (in per cent). The correlation between these additional explanatory variables and the index of corporate governance quality is shown in Fig. 5. The key nding (in Table 5) is that the exogenous component of corporate governance as we measure it has a positive effect on all of these variables, but this effect is weakest it is insignicant at conventional condence levels in the case of public-sector bond market capitalization.23 This makes sense: stronger corporate governance will work directly to make investment by outsiders in private corporations more attractive, but it will stimulate investment in public debt securities only indirectly insofar as private and public bond markets are complements (they utilize the same market infrastructure, have a similar customer base, etc.).24

20 All exogenous variables in the system should be used as instruments for any endogenous variable in the rst stage in order to obtain unbiased estimates in the second stage regression. 21 Admittedly, other political variables such as the structure of interest group politics have been linked to nancial development (see for example Rajan and Zingales, 2003; Carney, 2007), but not the specic political variables here. And those other political variables have typically been linked to the choice between a bank- and market-based nancial system, which is not our focus. 22 See the section on robustness. Dropping this variable does however create some other sources of econometric discomfort, as we explain there, requiring us to modify the specication slightly. 23 As noted in Table 2 above, we lack some bond market data for Israel. This accounts for the small reduction in number of usable observations in some columns of Table 5 (and Tables 8 and 9 below). 24 When we include private bond market capitalization as a determinant of public bond market capitalization, there is only weak evidence of this last effect. When we substitute public bond market size lagged one year in the rst column, the key results do not change. In addition, lagged public bond market size is not signicant (coecient = 0.021, t-statistic = 0.66).

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

163

(a)

(b) Fig. 5. CGQ index and explanatory variables cross section, average 19952005.

When we estimate the same equations by pooled OLS, ignoring the possibility of simultaneity, the signicant positive effects on private bond market capitalization, stock market capitalization, the number of listed companies and the measure of stock liquidity remain (column 1 of Table 6). But only the

Similarly, when we add private bond market capitalization lagged one year in the second column, the key results remain the same and the lagged private bond market size is not signicant (coecient = 0.044, t-statistic = 0.58).

164

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

(c)

(d) Fig. 5. (continued)

positive effect on private bond market capitalization remains when we use xed- or random-effects panel estimators without instrumental variables (columns 2 and 3). This underscores the importance of acknowledging the endogeneity of corporate governance.25

25 In the case of private bond market, although the effect is positive and signicant across all regression models, the fact that the coecient on corporate governance quality in the OLS regressions is smaller than in the IV regressions suggests that the reverse relationship from bond market size to corporate governance quality may be negative.

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

165

(e)

(f) Fig. 5. (continued)

5. Robustness We performed a number of sensitivity analyses to check robustness. To start, we compared GLS regression results based on country-clustering with those obtained using HuberWhite-sandwich variance estimator. Reassuringly, estimates obtained using the two methods do not differ signicantly. Second, we estimated our equations using xed rather than random effects. This requires dropping legal origin, the Asia region dummy, and the number of parent enterprises because these variables do

166

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

Table 5 Effects of corporate governance quality (second stage, IV regressions). Dependent variable Private bond market cap (% GDP) Corporate governance quality (tted) Log GDP per capita English legal origin dummy Domestic credit provided by banking sector (% GDP) Lending interest rate (%) Number of parent enterprises Financial openness Trade openness Exchange rate stability Dummy for banking crisis in previous year Years under IMF programs Dummy for Asia Public bond market size (% GDP) Constant Observations Number of country ID R2 1.509*** (4.46) 2.938*** (4.89) 5.454 (0.95) 0.172*** (6.49) 0.173*** (3.04) 0.005** (2.48) 0.026 (0.75) 0.001 (0.02) 0.381** (2.33) 0.978 (1.07) 0.164*** (2.69) 0.421 (0.07) 0.038 (1.17) 122.943 (5.40) 439 40 0.656 Public bond market cap (% GDP) 0.366 (0.71) 6.057*** (3.57) 1.219 (0.13) 0.094*** (2.75) 0.203* (1.74) 0.004** (2.25) 0.134** (2.25) 0.117*** (2.90) 0.906*** (2.80) 0.418 (0.28) 0.088 (0.54) 23.452 (1.63) Stock market cap (% of GDP) 12 215*** (4.33) 14.117*** (4.25) 27.908 (1.36) 0.082 (1.37) 2.132*** (4.48) 0.008* (1.72) 1.006*** (3.11) 0.483** (2.25) 3.800*** (4.46) 13.119*** (2.97) 0.922* (1.86) 61.071** (2.32) Number of listed companies (in log) 0.053*** (3.21) 0.092** (2.33) 0.788*** (2.61) 0.001 (0.77) 0.010*** (3.21) 0.001** (2.10) 0.003** (2.22) 0.000 (0.17) 0.015** (2.01) 0.017 (0.47) 0.011*** (3.58) 1.001*** (3.62) Stock turnover ratio (%) 3.826** (2.27) 3.123 (0.75) 5.905 (0.31) 0.128 (1.28) 0.577 (1.46) 0.008 (1.28) 0.308** (2.34) 0.349*** (2.53) 0.732 (0.56) 4.220 (0.46) 1.673 (1.19) 57.944** (1.96)

64.088 (1.71) 439 40 0.011

890.114 (4.55)
451 41 0.472

0.991 (0.86) 451 41 0.313

167.657 (1.41)
451 41 0.227

Notes: Corporate governance quality is the tted value from regressing the CGQ index on a set of instrumental variables as in column 2 of Table 4. Instruments used are: government stability, polity index, and cumulative capital inows as a percent of GDP. Heteroskedasticity-consistent z-statistics in parentheses.
* ** ***

indicates signicance level at 10%. indicates signicance level at 5%. indicates signicance level at 1%.

not vary over time. The rst-stage results still hold for the other variables except that the coecient after trade openness now becomes positive and signicant. The key results from the second stage are unchanged; the remaining coecients all have the same sign and continue to differ signicantly from zero at standard condence levels. Third, we added to the rst stage dummy variables for Latin America and for the South Africa Turkey pair (column 1 of Table 7). The signicant negative coecient on the dummy for the Asian countries remains when the two additional regional dummies are present; and the Asia dummy in fact becomes larger in absolute value. The additional coecients on the Latin America and South AfricaTurkey dummies do not differ signicantly from zero at standard condence levels. Fourth, we estimated alternative specications for the rst stage, as reported in columns 2 and 3 of Table 7. We dropped the insignicant exchange rate, banking crisis, and IMF dummy variables; next we dropped the measures of nancial openness, trade openness and domestic credit provided by the banking sector. The key results carried over. Fifth, we explored further the robustness of the insignicant nancial crisis and IMF program variables. We substituted whether a country was under an IMF program in the current year for cumulative

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

167

Table 6 Alternative specications for the effects of corporate governance quality on nancial development. Dependent variable Coecient on corporate governance quality index Pooled OLS Private bond market cap (% GDP) Public bond market cap (% GDP) Stock market cap (% of GDP) Log number of listed companies Stock turnover ratio (%) 0.443*** (3.37) 0.143 (0.87) 1.232*** (3.52) 0.025*** (4.03) 0.438 (1.38) OLS/FE 0.188*** (3.12) 0.100 (1.21) 0.065 (0.26) 0.005** (2.17) 0.232 (0.76) GLS/RE 0.196*** (3.07) 0.126 (1.34) 0.259 (1.03) 0.003 (1.32) 0.281 (0.92) IV/FE 1.393*** (3.71) 0.055 (0.10) 11.940*** (3.60) 0.025* (1.66) 9.028*** (2.89) IV/RE 1.509*** (4.46) 0.366 (0.71) 12.215*** (4.33) 0.053*** (3.21) 3.826** (2.27)

Notes: For the OLS and GLS models, each nancial development variable is regressed on the CGQ index and the set of independent variables analogous to columns 1 through 5 in Table 5. For the IV models, rst stage regressions are analogous to column 2 of Table 4 (with English dummy origin, number of parent enterprises, and dummy for Asia dropped for the xed effects IV regression). CGQ Index in the second stage IV regressions is the tted value from the rst stage. Instruments used are: government stability, polity index, and cumulative capital inows as a percent of GDP. Heteroskedasticity-consistent z-statistics in parentheses.
* ** ***

indicates signicance level at 10%. indicates signicance level at 5%. indicates signicance level at 1%.

years under a program. We interacted the IMF program indicator with a dummy variable for 1998 and after to see whether Fund programs negotiated in conjunction with the Asian crisis, which tended to place more emphasis than their predecessors on corporate governance reform, had differential effects. None of these alternatives was particularly supportive of the notion that IMF programs mattered for corporate governance quality. The one additional variable that did matter was the dummy variable for 1998 and after when it was not interacted with Fund programs. Evidently, there was a change in attitudes following the Asian crisis that did have an impact on corporate governance quality. Sixth, there is the worry that exceptional nancial volatility may have a positive impact on measured stock price synchronicity and therefore on the index of corporate governance quality. Similarly, the proxy for earnings opacity and therefore the measure of corporate governance quality may show a spurious increase in periods of nancial disruption. In turn these measured changes in the index may result in a spurious association with potential determinants of corporate governance quality such as interest rates, nancial openness and exchange rate variability. The crisis indicator is designed to control for these periods of exceptional volatility but may do so imperfectly. As an alternative we therefore excluded the crisis years entirely. The results are robust to this change.26 Seventh, we dropped cumulative capital inows from the rst stage. Some will worry that cumulative capital ows are not really exogenous with respect to the quality of corporate governance, although we derive some reassurance from the fact the instrument is heavily driven by capital ows in the past while the corporate governance quality indicator is contemporaneous. A more telling objection is that cumulative capital ows are strongly correlated with time and may be picking up the effects of other factors that also display a trend. Indeed, when we add a time trend to the rst stage, the cumulative capital ows variable loses its signicance. Reassuringly, however, our key results in particular, the sign and signicance level of the corporate governance variable in the second stage remain the same when cumulative capital ows are excluded and whether or not a time trend is added.

26 The main difference is that the magnitude and signicance of the coecient on tted CGQ in the stock market capitalization/GDP regression decline sharply, possibly reecting a strong relationship between low CG quality and a reduction in the stock market size of the crisis countries during crises.

168

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

Table 7 Determinants of corporate governance quality robustness checks. Dependent variable: corporate governance quality (CGQ) (1) Excluded instruments: Government stability Polity index Cumulative capital inows (% GDP) Average CGQ in other countries in the same region Included instruments: Log GDP per capita English legal origin dummy Lending interest rate (%) Number of parent enterprises Domestic credit provided by banking sector (% GDP) Financial openness Trade openness Exchange rate stability Dummy for banking crisis in previous year Years under IMF programs Dummy for Asia Dummy for Latin America Dummy for South Africa and Turkey Dummy for IMF program participation Dummy for post-1997 period (Dummy for IMF program)* (Dummy for post-1997 period) Time trend Constant 66.633 (11.34) 15.75 0.001 451 41 0.225 63.188 (14.32) 15.79 0.001 451 41 0.206 63.307 (14.83) 30.43 0.000 451 41 0.215 0.127 (0.99) 0.051 (0.44) 0.042*** (3.50) (2) 0.132 (1.02) 0.067 (0.55) 0.039*** (3.38) (3) 0.127 (0.98) 0.081 (0.68) 0.042*** (4.71) (4) 0.476*** (2.58) 0.007 (0.06) 0.024*** (2.57) (5) 0.268** (2.13) 0.072 (0.60) (6) 0.337*** (2.58) 0.045 (0.37) 0.006 (0.69) (7) 0.124 (1.02) 0.017 (0.15)

0.373*** (4.59)

0.687 (1.34) 3.549* (1.87) 0.131** (2.46) 0.001*** (3.12) 0.002 (0.23) 0.027 (1.38) 0.010 (0.74) 0.123 (0.60) 0.118 (0.17) 0.034 (0.26) 6.254*** (2.95) 3.577 (1.21) 0.263 (0.06)

0.637
(1.56) 4.281** (2.39) 0.128*** (3.01) 0.001*** (3.62) 0.002 (0.24) 0.026 (1.35) 0.012 (0.91)

0.580
(1.51) 4.450*** (2.58) 0.130*** (3.09) 0.001*** (3.78)

0.076
(0.19) 5.016*** (2.99) 0.095** (2.20) 0.001*** (3.64)

0.344
(0.94) 5.084*** (3.05) 0.136*** (3.19) 0.001*** (3.34)

0.254 (0.68) 5.564*** (3.27) 0.017 (0.42) 0.001*** (3.36)

0.087 (0.21) 3.877** (2.34) 0.085 (1.89) 0.001*** (2.71)

5.541***
(3.32)

4.940***
(3.09)

4.287*** (2.70)

4.482*** (2.85)

3.699** (2.38)

3.018* (1.83)

0.992 (0.54) 4.133*** (4.80) 1.148 (0.65) 0.774*** (7.74) 51.497 (11.87) 7.02 0.071 451 41 0.305

61.541 (14.57) 13.70 0.003 451 41 0.253

61.437 (14.76) 4.92 0.085 451 41 0.234

37.583 (5.25) 22.73 0.000 451 41 0.271

F-stat for excluded instruments p-value Observations Number of country ID R2 Notes: All regressions are * indicates signicance ** indicates signicance *** indicates signicance

estimated using GLS random effects. Heteroskedasticity-consistent-z-statistics in parentheses. level at 10%. level at 5%. level at 1%.

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

169

The cost of dropping the capital ows variable is a problem of weak instruments, according to the F-test of joint signicance of the two political variables in the rst stage regression.27 The textbook treatment for this problem is to nd other, more powerful instruments. One possibility is to build on work on peer effects and policy diffusion (Simmons and Elkins, 2003), where it is argued that the probability of a country adopting a particular policy reform is increasing with the number of its neighbors who have already done so. One can argue that there is a logic for including this variable, constructed as the quality of corporate governance in other countries in the same region, insofar as countries compete for foreign capital partly on the basis of how well they represent the interests of investors. This variable is plausibly exogenous except perhaps for countries large enough to inuence the quality of corporate governance throughout the region. It is plausible to exclude it from the second stage, there being no reason to expect the quality of corporate governance elsewhere to have a rstorder impact on the subject countrys nancial development. A limitation of this variable is that it is not clear that a countrys economic neighbors are also its geographic neighbors in other words, that the relevant peer group is made up of the countries in the same geographical neighborhood. This is why we relegate estimates using this instrument to this section on sensitivity analysis. In any case, constructed as it is here this peer-effects measure is a strong instrument. Adding it leaves the sign and signicance of the key corporate governance variable in the second stage unchanged and eliminates the weak-instrument problem. Eighth, we ran a series of cross section regressions on data for select years and on the 1995 2005 country averages. This is another way for controlling for the possibility of a spurious correlation between secular improvements in corporate governance quality and increases in cumulative capital ows. As might be expected, cumulative capital ows are no longer important in the cross sections; lending rates also lose their prior signicance. In contrast, other variables, notably English legal origin, remain highly signicant.28 An interpretation is that legal origin explains cross-country differences in corporate governance quality, while cumulative capital inows and lending rates are correlated with within country changes over time. Ninth, we looked separately at the impact on market development of the individual components of our corporate governance index, having rst estimated their determinants using the same specication as above. As shown in Table 8, all three elements (adoption of accounting standards, tendency not to smooth earnings and share-price non-synchronicity) generally have the expected positive effect on private bond market capitalization, stock market capitalization, the number of listed companies and the stock turnover ratio.29 This reassures us that the results do not hinge on the behavior of any one component of our corporate governance measure. Finally, we considered the number of listed companies scaled by real GDP and by total number of business registrations rather than simply the number of listed companies to take into account the effect of country size and corporate sector size on this variable. For both specications the sign and signicance of most variables, including corporate governance quality, remained unchanged, except for the log of real GDP per capita, the coecient on which now became negative and signicant.30

27 The F-statistic for the excluded instruments is 15.67 with the measure of cumulative capital inows included in the rst stage and 1.33 without. The cutoff for weak instruments is a threshold of 10.00 as suggested by Staiger and Stock (1997) for the case of a single endogenous regressor. 28 The same is true of nancial openness and exchange rate stability depending on the year considered. 29 With an exception of stock price non-synchronicity, which has negative effect on the turnover ratio. Analogous regressions for regional subsamples reveal that this strong, negative effect prevails in the Asian economies, offsetting the positive effect for non-Asian subsample. See below for further discussion of this result for Asian economies. The number of observations varies across columns because we possess some components but not others of the CGQ index in 1994 for some of our countries. 30 We include these specications in the section on robustness rather than the results section above because it not clear that real GDP is an appropriate variable to use as a scaling factor for number of listed rms. Using the total number of business registrations as a scaling factor is more appropriate, but due to incomplete time-series data on this variable, doing so reduces the number of observations greatly (from 451 to 193 country-years). In addition, since we already include the number of parent companies in the list of controls, we feel that it is justied to enter number of listed companies as a simple number. Data on total number of registrations are obtained from the World Bank Group Database on Entrepreneurship available online at http://econ.worldbank.org/WBSITE/EXTERNAL/EXTDEC/EXTRESEARCH/0contentMDK:21164814~pagePK:64214825~piPK: 64214943~theSitePK:469382,00.html.

170 S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

Table 8 Effects of different components of CGQ index (second stage, IV regressions). Dependent variable Private bond market cap (% GDP) CGQ Components: Accounting standards (tted) Earnings smoothing (tted) Price non-synchronicity (tted) Control variables Observations Number of country ID R2 Y 453 40 0.655 Y 439 40 0.657 3.161*** (3.82) 0.912*** (4.60) 0.819*** (4.02) Y 475 40 0.658 Public bond market cap (% GDP) Stock market cap (% GDP) Number of listed companies (in log) 0.142*** (3.51) 6.966*** (4.38) 0.439 (1.44) Y 475 40 0.013 8.663*** (4.40) Y 489 41 0.476 0.033*** (3.39) 0.027** (2.51) Y 490 41 0.296 Stock turnover ratio (%)

1.328
(0.89) 0.104 (0.33)

19.071*** (3.48)

49.230*** (5.21) 4.564*** (3.97)

Y 453 40 0.005

Y 439 40 0.010

Y 465 41 0.452

Y 451 41 0.471

Y 466 41 0.281

Y 451 41 0.312

Y 465 41 0.221

Y 451 41 0.226

2.912 (1.45) Y
489 41 0.230

Notes: All regressions are analogous to the full specication in Table 5 with the CGQ index replaced by each of its three components. Results on other independent variables omitted. Instruments used are: government stability, polity index, and cumulative capital inows as a percent of GDP. Heteroskedasticity-consistent z-statistics in parentheses.
** ***

indicates signicance level at 5%. indicates signicance level at 1%.

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

171

6. Corporate governance and capital market development in Asia Having looked separately in Section 4 at the determinants of corporate governance practice in Asia, here we look separately at the impact in Asia of corporate governance reform on nancial development.31 We estimate the same specication as in the Table 5 full-sample regressions separately for the Asian and non-Asian subsamples, again using instrumental variables.32 Results are in Table 9. The relatively high R 2 for the Asian subsample indicates that the independent variables used in our analysis account for most of the differences in capital market development. The positive effects of corporate governance on stock market capitalization and the number of listed companies are strong and signicant only for the Asian economies; it would appear that this result is driven mainly by the Asian subsample. On the other hand, the impact of corporate governance quality on private bond market capitalization and stock market turnover is negative for the Asian subsample, in contrast to the non-Asian subsample where the full-sample results carry over. This negative association with private bond market capitalization may indicate that efforts to improve corporate governance in the region have not had a payoff in terms of the growth of this market. But the negative effect of corporate governance quality on private bond market size in the Asian subsample turns out to be driven by two outliers: Malaysia and the Philippines.33 When these two countries are dropped, we get a positive effect of corporate governance quality on private bond market development that is statistically signicant at standard condence levels for the Asian subsample as well. The negative association of corporate governance quality with stock market turnover may reect the tendency for provident funds and other buy-and-hold investors to dominate the market in countries like Singapore and Hong Kong where both corporate governance and equity markets are relatively sophisticated. In addition, it is argued the prevalence of momentum traders in Asian stock exchanges may account for the association of high turnover with poor corporate governance, since such investors are unlikely to pay much attention to the governance of companies.34 In both the Asian and non-Asian subsamples, more parent enterprises contribute to larger private bond markets, larger stock markets, more listed companies and higher stock turnover, as expected. Recall from the analysis of determinants of corporate governance that the number of parent enterprises has a negative effect on governance quality; we attributed this to the inuence of big business groups in Asia, which may have opposed improvements in corporate governance. Thus, the number of parent companies has a mixed impact on nancial market development: a positive direct impact and a negative indirect impact operating via the quality of corporate governance. In the Asian subsample, trade openness is positively associated with private bond issuance and equity market capitalization. Interestingly, nancial openness appears to be good for stock market capitalization but bad for bond market capitalization. Insofar as this reects a differential tendency for countries to open their stock and bond markets to foreign investors, a more nuanced measure of nancial openness may shed more light on the pattern. Corporates rely more on bond issuance

31 Latin America is characterized by many of the same capital-market and corporate-governance problems as Asia, including concentrated ownership and weak shareholder rights (Capaul, 2003). It would be interesting to analyze that region separately. However, the small sample size together with the strong similarity of economic and political attributes of the seven Latin American countries in the sample limit the variation in the nancial variables which in turn undermines the precision in the subsample regression analyses. We have analyzed the Latin American case using a different methodology in Ananchotikul and Eichengreen (2008). 32 We of course estimate the rst stage separately for the Asian and non-Asian subsamples. 33 Malaysia had relatively low corporate governance score over the sample period (56.9 compared with the regions average of 59.8), but very high private bond market capitalization as a percent of GDP (49.74 percent compared with the regions average of 18.20 percent). The very aggressive measures taken by the Malaysian government and central bank to promote the development of the corporate bond market, by establishing an ecient trading platform and mandating extensive price transparency in transactions, are the subject of Bin Ibrahim and Wong (2005). On the other hand, the Philippines scored relatively well in the corporate governance index (61.0) but had extremely small private bond market (0.09 percent of GDP). Espenila (2005) points to a variety of problems that account for the relatively small size of the market: the absence of an ecient trading platform like Malaysias (all trading is over the counter), outmoded bankruptcy laws, and bank dominance of the nancial sector. 34 Momentum traders refers to investors who make their investment decisions based not on company fundamentals but on the dynamics of the stock market index or individual stock prices. For what it is worth, this negative relationship is especially strong in China, Pakistan, and Malaysia.

172

Table 9 Effects of corporate governance quality, IV regressions Asian and non-Asian subsamples. Second stage Private bond market cap (% GDP) Asia Corporate governance quality (tted) Log GDP per capita English legal origin dummy Domestic credit provided by banking sector (% GDP) Lending interest rate (%) Number of parent enterprises Financial openness Trade openness Exchange rate stability Dummy for banking crisis in previous year Years under IMF programs Public bond market size (% GDP) Constant Non-Asia 1.268*** (2.95) 2.018** (2.39) 10.333 (1.29) 0.224*** (6.99) 0.110* (1.73) 0.006** (2.46) 0.036 (0.83) 0.103* (1.73) 0.382** (2.36) 0.579 (0.63) 0.093 (0.52) 0.097** (2.45) 92.173 (3.28) 319 29 0.662 Public bond market cap (% GDP) Asia 3.244*** (4.10) 4.592** (2.17) 2.133 (0.60) 0.141** (2.08) 1.877*** (3.39) 0.004*** (3.93) 0.403* (1.59) 0.198*** (4.54) 0.458 (0.35) 5.351 (1.46) 0.934*** (3.31) Non-Asia Stock market cap (% of GDP) Asia 7.650*** (3.27) 11.278* (1.93) 20.119*** (2.63) 0.782*** (4.64) 3.086** (2.55) 0.011*** (3.34) 2.741*** (3.03) 0.680*** (6.31) 11.777*** (4.22) 21.905* (1.92) 0.910 (1.12) Non-Asia 3.696 (1.46) 3.860 (1.34) 18.342 (0.78) 0.093 (0.77) 0.802* (1.83) 0.003 (0.56) 0.714** (2.15) 0.584*** (2.90) 1.632* (1.67) 22.123*** (4.26) 1.128 (1.17) Number of listed companies (in log) Asia 0.096*** (4.04) 0.330*** (5.17) 1.299*** (8.19) 0.011*** (6.66) 0.000 (0.04) 0.001*** (9.51) 0.016*** (2.63) 0.006*** (5.89) 0.061 (1.51) 0.305** (2.14) 0.035*** (3.47) Non-Asia 0.001 (0.04) 0.021 (0.52) 1.298*** (4.12) 0.000 (0.24) 0.002 (0.76) 0.001** (2.47) 0.004*** (3.15) 0.004** (1.98) 0.008 (0.90) 0.047 (0.94) 0.024*** (2.97) Stock turnover ratio (%) Asia Non-Asia 4.268*** (2.65) 1.415 (0.29) 31.556* (1.70) 0.126 (1.25) 0.601* (1.90) 0.000 (0.14) 0.176 (1.24) 0.075 (0.40) 0.141 (0.16) 6.434 (0.84) 3.198*** (3.56)

1.624***
(3.16) 3.154*** (3.05) 2.815* (1.82) 0.026 (0.91) 0.079 (0.35) 0.005*** (6.86) 0.715*** (2.92) 0.152*** (5.33) 0.906 (1.63) 3.691* (1.75) 0.327*** (2.74) 0.285*** (4.43) 67.961 (2.34) 120 11 0.807

1.193**
(2.42) 9.206*** (5.82) 15.041 (1.38) 0.009 (0.29) 0.360*** (3.50) 0.005** (2.16) 0.145*** (2.70) 0.075 (1.01) 0.419 (1.37) 2.623 (1.44) 0.667** (2.05)

14.848***
(4.36) 7.078 (0.89) 80.305*** (4.70) 1.010*** (4.35) 3.766* (1.91) 0.027*** (3.60) 0.028 (0.05) 0.070 (0.57) 3.836 (0.73) 12.798 (0.67) 6.111*** (2.81)

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

161.584
(3.09) 120 11 0.566

181.873 (5.59) 319 29 0.030

611.987
(4.16) 121 11 0.848

259.846
(1.52) 330 30 0.147

1.325 (0.88) 121 11 0.705

5.118 (4.05) 330 30 0.384

1023.329 (4.88) 121 11 0.537

195.822 (1.74)
330 30 0.199

Observations Number of country ID R2

Notes: Corporate governance quality for Asia (non-Asia) is the tted value from regressing the CGQ index on a set of instrumental variables exclusively for Asian (non-Asian) subsample as in column 2 (3) of Table 4. Instruments used are: government stability, polity index, and cumulative capital inows as a percent of GDP. Heteroskedasticity-consistent z-statistics in parentheses.
* ** ***

indicates signicance level at 10%. indicates signicance level at 5%. indicates signicance level at 1%.

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

173

in the aftermath of banking crises but issue less equity, presumably reecting depressed valuations. This behavior was prominent in, inter alia, Korea in 19981999, when the major conglomerates issued large numbers of bonds to nance restructuring in the wake of the countrys crisis. Public bond market size seems to be another important determinant of private bond market capitalization: larger public bond markets relative to GDP are associated with high corporate governance issuance as a percentage of GDP. This conrms a nding of previous studies (e.g. Eichengreen and Leungnaruemitchai, 2006) and attests to the benchmark function and liquidity-enhancing effects of public bond markets. It also highlights an impediment to private bond market development in Asia, namely that government bond markets are often not well developed, because countries have relied on off-budget borrowing from abroad for nancing and because economies such as Singapore and Hong Kong have had scal surpluses for a long period of time, creating no need to issue bonds.35 7. Conclusion The Asian crisis pointed up the need to strengthen corporate governance in emerging market economies. Corporate governance is only one item on the reform agenda, which includes improved supervision and regulation of nancial institutions and capital markets; adapting macroeconomic policies, including exchange rate policy, to the now more open nancial environment and strengthening competition policy. But it is arguably one of the more challenging items. Even high-income countries with relatively sophisticated nancial markets continue to grapple with shortcomings in their corporate governance arrangements. There is little agreement among scholars and practitioners on the ecient design of such arrangements. And there remains the perception, in Asia in particular, that corporate governance arrangements suitable for the advanced countries are not appropriate for the very different circumstances of emerging markets. All this said, there have been improvements in the quality of corporate governance in the last ten years. While the task is dicult, in other words, progress is possible. We show here that the relevant reforms have a signicant payoff in terms of the development of equity markets and corporate bond markets. That is to say, corporate governance reform pays. But progress has been far from uniform. We nd that improvements are most likely in countries with stable governments prepared to sink the up-front costs of institutional reforms with deferred payoffs, where foreign investors are prepared to lobby for reform, and where other countries in the region are undertaking analogous reforms, which is suggestive of peer effects. That said, the case of China, which has nothing if not governmental stability, which is increasingly open to foreign investment, and which is surrounded by other Asian countries undertaking the relevant reforms but which has made little progress on the corporate-governance front suggests that these conditions may not be sucient or that they may suce only when they reach a critical threshold. The question what emerging markets can do to eliminate this shortfall. One view would be that effective corporate governance is an organic part of the larger process of economic and nancial development and that emerging markets can close their corporate governance quality gap only as their per capital incomes and levels of nancial development converge with those of the advanced countries. Our results support a more optimistic conclusion. The quality of corporate governance depends on more than just per capita income as a measure of the general level of economic development. It in fact depends also on other factors, which suggests that there are some very specic things that emerging markets can do to promote it. Macroeconomic stability is good for the development of corporate governance. So too is political stability, which gives investors voice and governments an incentive to invest in the future. And corporate governance quality does not simply reect the level of nancial development; in addition it can affect it. The results here suggest that it can affect it in decidedly positive ways.

35 This problem has been the subject of other analyses: see inter alia McCauley (2003). Finally, there remain a variety of interesting effects of years spent under an IMF program as in the full-sample estimates in Table 5 above. We leave verifying their robustness and their interpretation to future work.

174

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

Acknowledgments We thank Kenichi Ueda for advice regarding data and Pipat Leungnaruemitchai for help in obtaining it.
Appendix Table A1 Description of variables and data sources. Variable CGQ index Description Unweighted average of the indicators of accounting standards, earning smoothing, and stock price synchronicity, ranging from 0 (worst) to 100 (best) Number of reported accounting items as a percentage of 40 accounting items Rank correlation between operating cash ow and accounting accruals across a set of rms at each point in time, standardized, ranging from 0 (most opaque performance) to 100 (least opaque performance) Average R-squared of regressions of each companys stock return on country-average return in each year, standardized, ranging from 0 (maximum synchronicity) to 100 (minimum synchronicity) Private domestic debt securities issued by nancial institutions and corporations (as a percentage of GDP) Source de Nicolo et al. (2006); updated by authors. de Nicolo et al. (2006); updated by authors. de Nicolo et al. (2006); updated by authors.

Accounting standards index Earning smoothing index

Stock price synchronicity index

de Nicolo et al. (2006); updated by authors.

Private bond market capitalization

Public bond market capitalization

Public domestic debt securities issued by government (as a percentage of GDP)

Stock market capitalization Number of listed companies Stock turnover ratio Government stability

Value of listed shares as a percentage of GDP Number of companies listed on the national stock market Ratio of total value of shares traded to stock market capitalization Assessment of the governments ability to carry out its declared program(s) and its ability to stay in oce, ranging from 0 (least stable) to 12 (most stable) Combined scores of polity regime characteristics, ranging from 10 (strongly autocratic) to 10 (strongly democratic) Stock of inward foreign direct investment (as a percentage of GDP) Stock of outward direct investment (as a percentage of GDP) Dummy variables indicating country law originated from English law, German law, French law, and Scandinavian law Log of deated GDP over total population Private domestic credit provided by deposit money banks and other nancial institutions (as a percentage of GDP) Average lending rates paid by commercial banks (in percent) Parent corporations are those enterprises that control assets of other entities outside of their respective home countries. Typically, control of assets requires ownership of at least 10% of a corporations shares or voting power (equity capital stake), or its equivalent for an unincorporated enterprise

BIS Domestic and International Securities Statistics; supplementary data are from Beck et al. (2000) and national statistical databases BIS Domestic and International Securities Statistics; supplementary data are from Beck et al. (2000) and national statistical databases World Development Indicator (WDI) World Development Indicator (WDI) Beck et al. (2000) International Country Risk Guide (ICRG) Polity IV project http://www.cidcm.umd.edu/polity/ United Nations Conference on Trade and Development (UNCTAD) database United Nations Conference on Trade and Development (UNCTAD) database La Porta et al. (1998) World Development Indicator (WDI) World Development Indicator (WDI) World Development Indicator (WDI) and Global Financial Data (GFD) United Nations Conference on Trade and Development (UNCTAD). World Investment Report 2005: Transnational Corporations and the Internationalization of R&D Annex Table A.I.8. Available online at http://www.unctad.org/en/docs/ wir2005_en.pdf

Polity index Cumulative capital inows Cumulative capital outows Legal origin Log of real GDP per capita Domestic credit by banking sector Lending rate

Number of parent enterprises

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

175

Appendix Table A1 (continued) Variable Financial openness Trade openness Exchange stability Description The sum of foreign direct investment and portfolio investment inows and outows (as a percentage of GDP) Total value of exports plus imports (as a percentage of GDP) Assessment of the appreciation or depreciation of a currency against the U.S. dollar over year, ranging from 0 (least stable) to 10 (most stable) Cumulative number of years a country has been under IMF agreements Dummy variable indicating an incidence of a currency crisis Source Raw data from International Financial Statistics (IFS) Raw data from World Development Indicator (WDI) International Country Risk Guide (ICRG) Vreeland (2003); updated data provided by James Vreeland Glick et al. (2006) and Ranciere et al. (2006) Caprio et al. (2003) Banking Crises Database, World Bank http://www1.worldbank.org/nance/ html/database_sfd.html. Updated banking crises data provided by Enrica Detragiache.

Years under IMF program(s) Currency crisis indicator Banking crisis indicator

Dummy variable indicating an incidence of a banking crisis

Appendix Table A2 Sample countries. Asia China Hong Kong India Indonesia Japan Korea Malaysia Pakistan Philippines Singapore Thailand Latin America Argentina Brazil Chile Colombia Mexico Peru Venezuela Europe Austria Belgium Denmark Finland France Germany Greece Ireland Israel Italy Netherlands Norway Portugal Spain Sweden Switzerland United Kingdom Other Australia Canada New Zealand South Africa Turkey United States

References
Alba, P., Claessens, Stijn, Djankov, S., 1998. Thailands corporate nancing and governance structures: Impact on rms competitiveness. Unpublished manuscript. World Bank. Ananchotikul, Sudarat, 2007. Does foreign investment really improve corporate governance? Evidence from Thailand. Unpublished manuscript. University of California, Berkeley. Ananchotikul, Sudarat, Eichengreen, Barry, 2008. Plumbing for Latin American capital markets. In: New Financing Trends in Latin America: A Bumpy Road Towards Stability. In: BIS Papers, No. 36. Bank for International Settlements, Basel, pp. 110139. Beck, Thorsten, Demirg-Kunt, Asli, Levine, Ross, 2000. A new database on nancial development and structure. World Bank Econ. Rev. 14, 597605. Bin Ibrahim, Muhammad, Wong, Adrian, 2005. The corporate bond market in Malaysia. In: BIS Papers, No. 26. Bank for International Settlements, Basel, pp. 114128. Capaul, Mierta, 2003. Corporate governance in Latin America. Unpublished manuscript. Latin America and the Caribbean Region, World Bank. Caprio, Gerard, Klingebiel, Daniela, Laeven, Luc, Noguera, Guillermo, 2003. Banking Crisis Database. World Bank. Available online at http://www1.worldbank.org/nance/html/database_sfd.html. Carney, Richard, 2007. Contested capitalism: Financial policies and implications for China. Unpublished manuscript. Nanyang Technological University, Singapore.

176

S. Ananchotikul, B. Eichengreen / J. Japanese Int. Economies 23 (2009) 149176

Cheung, Stephen, Jang, Hasung, 2005. Evaluation of Corporate Governance in East Asian Economies. PECC Macro Corporate Governance Scorecard Protect, Seoul, PECC. de Nicolo, Gianni, Laeven, Luc, Ueda, Kenichi, 2006. Corporate governance quality: Trends and real effects. IMF Working Paper No. WP/06/293 (December). Doidge, Craig, Karolyi, G. Andrew, Stulz, Rene M., 2004. Why do countries matter so much for corporate governance? Finance Working Paper No. 50/2004. European Corporate Governance Institute (September). Eichengreen, Barry, Leungnaruemitchai, Pipat, 2006. What Doesnt Asia Have Bigger Bond Markets? BIS Papers, No. 30. BIS, Basel. Enriques, Luca, Volpin, Paolo, 2007. Corporate governance reforms in continental Europe. J. Econ. Perspect. 21, 117140. Espenila, Nestor, 2005. The corporate bond market in the Philippines. In: BIS Papers, No. 26, pp. 135145. Glick, Reuven, Guo, Xueyan, Hutchison, Michael, 2006. Currency crises, capital-account liberalization, and selection bias. Rev. Econ. Statist. 88, 698714. Goldstein, Morris, 1998. The Asian Financial Crisis. Institute for International Economics, Washington, DC. Hofstetter, Karl, 2005. One size does not t all: Corporate governance for Controlled Companies. Unpublished manuscript. Harvard Law School (June). Johnson, Simon, La Porta, Rafael, Lopez-de-Silanes, Florencio, Shleifer, Andri, 2000. Tunneling. Amer. Econ. Assoc. Papers Proc. 90, 2227. Khan, Haider, 2003. Corporate governance of family-based businesses in Asia: Which road to take? Unpublished manuscript. University of Denver. La Porta, Rafael, Lopez-de-Silanes, Florencio, Shleifer, Andrei, Vishny, Robert, 1998. Law and nance. J. Polit. Economy 106, 1113 1155. McCauley, Robert, 2003. Unifying public debt markets in East Asia. BIS Quart. Rev. December, 8998. Morck, Randall, Yeung, Bernard, Yu, Wayne, 2000. The information content of stock markets: Why do emerging markets have synchronous price movements? J. Finan. Econ. 58, 215260. Rajan, Raghuram, Zingales, Luigi, 2003. The great reversals: The politics of nancial development in the 20th century. J. Finan. Econ. 69, 550. Ranciere, Romain, Tornell, Aaron, Westermann, Frank, 2006. Decomposing the effects of nancial liberalization: Crises vs. Growth. NBER Working Papers No. 12806 (December). Roe, Mark J., 2003. Political determinants of corporate governance: Political context, corporate impact. Discussion Paper No. 451. John M. Olin Center for Law, Economics, and Business, Harvard Law School (December). Simmons, Beth, Elkins, Zachary, 2003. Globalization and policy diffusion: Explaining three decades of liberalization. In: Kahler, Miles, Lake, David (Eds.), Governance in a Global Economy. Princeton University Press, Princeton, pp. 275304. Staiger, Douglas, Stock, James, 1997. Instrumental variables regression with weak instruments. Econometrica 65, 557586. Vreeland, James, 2003. The IMF and Economic Development. Cambridge University Press, Cambridge. World Bank, 1998. East Asia: The Road to Recovery. World Bank, Washington, DC.