Chapter

Chapter 14. Corporate Governance Quality:Trends and Real Effects

Author(s):
Christopher Crowe, Simon Johnson, Jonathan Ostry, and Jeronimo Zettelmeyer
Published Date:
August 2010
Share
  • ShareShare
Show Summary Details

Gianni De Nicoló • Luc Laeven • Kenichi Ueda1

14.1. Introduction

Corporate governance reform has ranked high on policy makers’ agendas in many countries around the world since the late 1990s. New laws and regulations aimedat improving corporate governance have been introduced in many countries, and particularly in several Asian countries in the aftermath of the East Asian financial crisis of 1997–98.2

Yet, have governance practices actually improved? And, do improvements in corporate governance contribute to higher output, investment, and productivitygrowth in the corporate sector? To date, these key questions have not been addressed in the literature. This chapter addresses these questions. We first constructa composite corporate governance quality (CGQ) index and document it sevolution for major emerging markets and developed economies during theperiod 1994–2003. Then, we assess the impact of measured improvements incorporate governance quality on output growth, productivity growth, and invest mentat a country level, and on industry growth.

Our CGQ index is constructed at a country level using accounting and market data of samples of nonfinancial firms listed in the relevant domestic stock markets. Hence, it captures corporate governance quality specific to a universe of firmswhich are likely to be comparatively more exposed to market discipline. For this reason, the finding of no improvement in governance for these firms would likely signal the lack of improvements for the corporate sector as a whole. On the other hand, the finding of improvements for these firms could signal either that improve mentshave occurred in the corporate sector as a whole, or that improvements arelikely to be found especially among firms subject to market discipline. In eithercase, the evolution of the index is informative about changes in governance in thecorporate sector.

The CGQ index is a simple average of three proxy measures of outcomes ofcorporate governance in the dimensions of accounting disclosure and transparency.Disclosure and transparency are necessary, albeit not sufficient, conditions ofgood corporate governance, as the extent of information asymmetries amongmanagers and stakeholders pointed out by the corporate governance literature arelikely to be less severe with enhanced transparency and disclosure.3 By focusingon indicators capturing necessary conditions for good corporate governance, weaim at capturing in a parsimonious, yet informative way, the dynamics of dimensionsof corporate governance quality that are likely to be correlated with otherdeterminants of efficient governance arrangements. As detailed below, these indicatorsare derived from selected studies in the finance and accounting literature.

Considering outcome-based measures of corporate governance, as opposed to dejure measures, is advantageous and informative for at least two reasons. First, trackingchanges in corporate governance with de jure measures is difficult, becauseimprovements may not necessarily occur because of lags in implementation and/orenforcement, as stressed by Berglöf and Claessens (2006), who more generallypoint out that it is hard to measure enforcement of corporate governance rules. Second, firms may indeed choose to improve their corporate governance prior to or independently of the enactment of new rules whenever the benefits of good corporategovernance, especially in terms of easier and less costly access to finance, arecritical for their growth prospects. In other words, firms can choose to improvecorporate governance beyond the minimum standards set by the country.

In essence, corporate governance quality may be viewed partly as an “endogenous”firm’s choice, as pointed out by Himmelberg, Hubbard, and Palia (1999) and Coles, Lemmon, and Meschke (2006). Ultimately, shareholders’ or stakeholders’ values willbe maximized when managerial incentives are set in a right direction, and a goodcorporate governance helps it happen (e.g., Jensen, 1986 ; and Tirole, 2001). Thus, it is a broad set of underlying rules and practices that determine corporate governanceand influence managerial incentives. Our aim is not to identify and quantifyeach of these underlying factors and the specific channels through which they operateto affect corporate governance and managerial incentives. Our contribution is todevelop outcome-based corporate-governance measures based on accounting andmarket data, as those data measure the outcomes of managerial decisions.

We investigate the relationship between corporate governance quality andeconomic performance at the country-level, although most of the literature relatesmeasures of corporate governance to firm-level performance (see, for example, Gompers, Ishii, and Metrick, 2003). Our choice is supported by empiricalevidence in Doidge, Karolyi, and Stulz (2007) who show that most of the variationin firm-level governance can be explained by country-level characteristics. Furthermore, Core, Guay, and Rusticus (2006) show that investors discountvalues of weakly governed firms and that weak governance does not cause poorstock market returns at the firm level. Our work is further motivated by Bushmanand Smith (2001), who review the literature on the role of publicly reportedfinancial accounting information in the governance processes of corporations andpropose areas for future research. They argue that “the use of financial accountinginformation in corporate governance mechanisms is one channel by which financialaccounting information potentially enhances the investment decisions andproductivity of firms” and “propose cross-country research to investigate moredirectly the effects of financial accounting information on economic performancethrough its role in governance.”

Our investigation yields three main findings. First, the CGQ index exhibitsimprovements in corporate governance quality in most countries considered since1994, with the exception of a few countries, where either no significant changeshave occurred, or a worsening is recorded. Although improvements in accountingdisclosure have been more limited, corporate governance quality has improvedespecially in the dimension of transparency, that is in terms of the reliability ofaccounting and market information.

Second, the data exhibit cross-country convergence in corporate governancequality with countries that score poorly initially catching up with countries withhigh corporate governance scores.

Third, improvements in corporate governance quality affect aggregate economicactivity positively and significantly, as shown in regression analysis of percapita GDP growth, total factor productivity (TFP) growth, and the ratio ofinvestment to GDP on the CGQ index. Moreover, when we gauge the impact ofchanges in corporate governance quality on sales growth and growth opportunitiesof firms grouped by industry, we find a positive effect of improvements in corporategovernance on the growth of financially dependent industries. This result isconsistent with the idea that improvements in corporate governance quality benefitmost those industries whose growth crucially depends on external finance.

Overall, the answers to the two questions we wished to address are both positive. Actual improvements in corporate governance, as captured by our indicators, have indeed occurred in most countries, although in varying degrees and withsome notable exceptions. More importantly, improvements in corporate governancequality yield tangible benefits in terms of enhanced growth, productivity, and investment, and these benefits are large for those industries which rely moston external finance. Thus, effective implementation of corporate governancereform appears to be an important contributing factor to countries’ well-being.

The remainder of this chapter is composed of three sections. Section 14.2 details the construction of the CGQ index and its components. Section 14.3 depicts the evolution of our measures of corporate governance quality withinand across countries and regions. Section 14.4 presents country and industryregressions relating the CGQ index and its components to measures of growth, productivity growth, and investment for the economy and the corporate sector.Section 14.5 concludes.

14.2. The CGQ Index

The CGQ index is a simple average of three indicators, called Accounting Standards (AS), Earning Smoothing (ES), and Stock Price Synchronicity (SPS). These indicators are constructed from accounting and market data for samples ofnonfinancial companies listed in stock markets taken from the Worldscope and Datastream databases.

14.2.1. Accounting Standards

The first indicator is a simple measure of the amount of accounting informationfirms disclose and is constructed similarly to the index reported by the Center for International Financial Analysis and Research (CIFAR) until 1993 . This indicator captures the degree of accounting disclosure of firms in the country.

CIFAR uses information based on the top 8 to 40 companies (dependingon data availability) and on 90 items selected by professional accountants(CIFAR, 1993). Our indicator is given by the number of reported accountingitems as a fraction of 40 accounting items selected from CIFAR’s 90 itemsbased on availability in the Worldscope database. We use information for thetop 10 manufacturing companies in terms of total assets for each year and ineach country.4

14.2.2. Earning Smoothing

The second indicator is a measure of “earnings opacity” proposed by Leuz, Nanda, and Wysocki (2003) and Bhattacharya, Daouk, and Welker (2003). Ittracks the extent to which managers may conceal the true performance of firmsusing accruals to smooth fluctuations of annual profits. Specifically, it is the rankcorrelation between cash flows (before any accounting adjustments) and profits(after accounting adjustments) across a set of firms at each point in time. Thisindicator is an important complement to the first indicator, because a large numberof reported accounting items may be meaningless if accounts are seriouslymanipulated or misrepresented.

Unlike these authors, who use a pooled cross-section data for each country, ourmeasures are calculated for each year and each country. Accruals (AS) are estimated asASikt=(ΔCAiktΔCashikt)(ΔCLiktΔSTDiktΔTPikt)Depiktwhere CA denotes current assets, Cash is cash and cash equivalents, CL are currentliabilities, STD is short-term debt and the current portion of long-term debt, TP is income tax payable, and Dep denotes depreciation and amortization.

Because cash-flow statements are not widely reported in many developingcountries, cash flow from operations (ECF) are estimated by subtracting accruals(AS) from operating income (OI) :ECFikt=OIiktASikt Cross-sectional earnings smoothing is then measured by a Spearman rank order correlation betweenchanges in accruals and changes in estimated cash flow (both normalized by totalassets). It is defined for each year and each country as

The ES indicator is standardized so that its values fall in the unit interval andincreases as earning smoothness declines (i.e., firm performance is less opaque). Thus, an increase of this indicator signals an improvement in transparency.

14.2.3. Stock Price Synchronicity

The third indicator is a measure of stock price synchronicity proposed by Morck, Yeung, and Yu (2000), given by the average goodness-of-fit (R2.) of regressions of eachcompany’s stock return on country-average return in each year.5 These authors showthat after controlling for other drivers of comovements in stock prices not necessarilyrelated to corporate governance,6 more synchronous stock prices are found in countriesin which corporate governance is poor and financial systems are less developed. More recently, Jin and Meyers (2006) analyze a larger data set and find a positiverelationship between stock price synchronicity and lack of transparency. Intuitively, ifthe accounting information is opaque, investors find it difficult to distinguish goodperformers from bad performers. Ceteris paribus, in the event of a shock to the marketor the arrival of new information, the inability of investors to discriminate amongfirms would induce them to trade most stocks, prompting movements in stock pricesto become more synchronous. We use this measure of stock price synchronicity as aproxy for the degree of accounting transparency in the country.

We should note that synchronicity can also occur if there is cross-subsidizationamong firms belonging to the same group. Cross-subsidization may stem fromoptimal allocation of funds in internal capital markets. Yet, in a poor governanceenvironment, cross-subsidization is likely to be associated with inefficient connectedlending: this governance-specific feature is likely captured by the SPSindicator, but it is not by the AS and ES indicators. In this sense, the SPS indicatorcomplements the two indicators previously described.

For each year, the SPS indicator is computed in five steps. First, we calculatethe weekly return rikt for each firm (t = week), dropping firms with less than 30weeks’ observations, and dropping an observation if the absolute value of rikt is greater than 0.25. Second, we calculate market capitalization-weighted weekly returns for each country k, ρkt, using weekly stock price indices from Datastream, and weekly net exchange rate appreciation rates for each country, ekt. Third, foreach firm we run the regressions:rikt=αi+βiρkt+λ(ρUSt+ekt)+εit, and retrieve the relevant goodness of fit R i2k.

Fourth, we calculate the total variation for each firm, given by

and compute the country-level common variation, given by

To avoid sample selection bias, Rk2 is computed for thesame sample size over years(but possibly for different companies) based on the rank order of market capitalization. Finally, the SPS indicator is standardized so that its range is the unitinterval, it increases as synchronicity declines (i.e., transparency improves), and iscomputed based on an equal number of (but different) firms selected by theirmarket capitalization at each date.7

Three measurement issues deserve further discussion. First, although our measuresof corporate governance are widely accepted proxies for various aspects of corporategovernance, we cannot rule out that they also capture other aspects of firm performance. For example, stock price synchronicity may be affected by abrupt declines incapital inflows, also known as sudden-stops. We try to mitigate this in our empiricalwork by investigating various subsamples of our data, such as dropping countries thatexperienced a crisis or that do not have well-developed stock markets.

Second, by construction, our CGQ index does not capture all aspects of corporategovernance but focuses on two important aspects of corporate governance:disclosure and transparency. The choice of our variables has been determined by three criteria: (1) they are based on widely accepted methodologies developed in the finance literature; (2) they are based on widely available financial data and can easilybe replicated and updated; and (3) they can be computed annually so we can track changes over time. We have considered a number of other variables considered by the corporate governance literature but these were not included because they did not meet any of the above criteria. These variables include ownership structures,8 American Depository Receipts (ADR) premiums,9 and the value of cash holdings.10

Finally, as already mentioned, we focus on de facto measures of corporate governanceand do not include de jure measures such as shareholder rights and those basedon securities laws (see La Porta and others, 1998 ; and La Porta, Lopez-de-Silanes, and Shleifer, 2006). The reason is that we want to capture changes in corporate governanceat the firm level rather than changes in laws that are generally made at thecountry level, and most importantly, we aim to capture real rather than legal changesin corporate governance quality. Changes in laws, because they are often not effectivelyenforced, may not be reflected in real changes that affect firm performance.

14.3. Trends in Corporate Governance Quality

The time series of the CGQ index and its components for 10 Asian countries(China, Hong Kong SAR, India, Indonesia, Korea, Malaysia, Pakistan, Philippines, Singapore, and Thailand), 7 Latin American countries (Argentina, Brazil, Chile, Colombia, Mexico, Peru, and Venezuela), 22 developed countries11 and 2 otheremerging markets (South Africa and Turkey) are reported in Table A14.1 to A14.4 in the Appendix of the working paper on which this chapter is based(De Nicolo, Laeven, and Ueda, 2006).

As shown in Figures 14.1, improvements in corporate governance quality havebeen recorded in most emerging market economies and developed economies, although with varying intensity. With regard to emerging market economies, it isworth noting that improvements in the CGQ index in Asian countries have beenlarger on average than those witnessed by Latin American countries, which also exhibit levels of the index generally lower than in Asia. Yet, in both emerging marketregions, as well as in some developed economies, the level of the index remainsabout 15 to 20 percent below that of the first quartile of developed economies.

Figure 14.1.CGQ index, subperiod averages. This figure shows the average of the CGQ index for each country for the two periods 1995 to 1996 and 2000 to 2003. The countries are grouped by region.

Figure 14.2.CGQ index in Asia. This figure shows the evolution of the CGQ index over the period 1995 to 2003 for a select number of countries in Asia.

Table 14.1Changes in Creditor Rights and Shareholder Protection
Change in Creditor Rights,Change in Shareholder Protection,
1995–20021997–2005
Argentina00
Australia00
Austria00
Belgium00
Brazil00
Canada00
Chile00
China0n.a.
Denmark00
Finland00
France00
Germany01
Greece00
Hong Kong SAR00
India01
Indonesia–1n.a.
Ireland00
Israel–11
Italy01
Japan–20
Korea, Rep.01
Malaysia00
Mexico01
Netherlands00
Norway00
Pakistan00
Philippines00
Portugal00
Singapore00
South Africa00
Spain01
Sweden–10
Switzerland00
Taiwan, Province of China00
Thailand–10
Turkey00
United Kingdom00
United States00
Average change–0.160.19
n.a., data not available.This table lists the changes over the period 1995 to 2002 in the index of creditor rights from Djankov and others (2007) and the changes over the period 1997 to 2005 in the index of anti-directors’ rights (shareholder protection) from Spamann (2006) for the sample of countries included in our study.
n.a., data not available.This table lists the changes over the period 1995 to 2002 in the index of creditor rights from Djankov and others (2007) and the changes over the period 1997 to 2005 in the index of anti-directors’ rights (shareholder protection) from Spamann (2006) for the sample of countries included in our study.
Table 14.2Correlation Matrix of Creditor Rights, Shareholder Protection, and CGQ Index
PANEL A. CORRELATION AMONG CHANGES IN CREDITOR RIGHTS, SHAREHOLDER PROTECTION,
AND CGQ INDEX
Change in Creditor Rights,Change in Shareholder Protection,
1995–20021997–2005
Change in Shareholder Protection,
1997–2005–0.005
(36)
Change in CGQ Index, 1995–2003–0.172–0.006
(38)(36)
PANEL B. CORRELATION BETWEEN CREDITOR RIGHTS, SHAREHOLDER PROTECTION, CGQ INDEX,
AND ITS COMPONENTS
CreditorShareholderAccountingEarnings
RightsProtectionCGQ IndexStandardsSmootding
Shareholder protection0.124
(40)
CGQ Index0.02430.057
(320)(36)
Accounting standards0.170**0.1060.538**
(354)(36)(361)
Earnings smoothing30.130**30.1070.747**0.267**
(336)(40)(361)(361)
Price synchronicity0.0220.1640.754**0.276**0.157**
(378)(40)(361)(395)(378)
CGQ, corporate governance qualityPanel A reports correlations between changes over the period 1995 to 2002 in the index of creditor rights from Djankov and others (2007), changes over the period 1997 to 2005 in the index of anti-directors’ rights (shareholder protection) from Spamann (2006), and changes over the period 1995 to 2003 in the CGQ index. Panel B reports correlations over the sample period (when data are available) between the levels of the index of creditor rights from Djankov and others (2005), the index of anti-directors’ rights (shareholder protection) from Spamann (2006), and the CGQ index and its components.Number of observations between parentheses.**Significant at 5 percent.
CGQ, corporate governance qualityPanel A reports correlations between changes over the period 1995 to 2002 in the index of creditor rights from Djankov and others (2007), changes over the period 1997 to 2005 in the index of anti-directors’ rights (shareholder protection) from Spamann (2006), and changes over the period 1995 to 2003 in the CGQ index. Panel B reports correlations over the sample period (when data are available) between the levels of the index of creditor rights from Djankov and others (2005), the index of anti-directors’ rights (shareholder protection) from Spamann (2006), and the CGQ index and its components.Number of observations between parentheses.**Significant at 5 percent.

The case of Asia is of interest with regard to the information content of our CGQ index relative to changes in de jure measures. As shown in Figures 14.2, the CGQ index exhibits an upward trend in all Asian countries except China, wherethe index exhibits a decline. However, notable improvements have been recordedin Hong Kong SAR, Malaysia, Philippines, Singapore, and Thailand, althoughimprovements in India, Indonesia, Korea, and Pakistan have been more muted.

These patterns contrast with those exhibited by measures of shareholder andcreditor rights during a similar period.12 As shown in Table 14.1, minority shareholderrights appear to have been strengthened in some countries, but not inothers. By contrast, measures of creditor rights do not appear to have improved, and they have even worsened in some countries. On average, minority shareholderrights have improved somewhat and creditor rights have deterioratedsomewhat for our sample of countries. In general, these de jure indexes of shareholderrights and creditor rights are very stable over the sample period for oursample of countries, with little or no changes in most cases. Yet, there appears tobe no relationship between the direction of change recorded by de jure type measuresand that recorded by our outcome-based measure. Panel A of Table 14.2 shows that the correlation between the CGQ index and the de jure measures islow (—0.17 in the case of creditor rights and—0.01 in the case of shareholderrights) and statistically insignificant. This difference between de jure and de factomeasures suggests the importance of taking into account the endogeneity of firms’governance choices in evaluating trends in corporate governance quality.

In which dimension has corporate governance quality changed most? As noted, each component of the CGQ index captures different, albeit complementary, aspects of corporate governance quality, as witnessed by the fact that their crosscorrelationis relatively low, ranging from 0.16 to 0.28 (Panel B of Table 14.2). Thus, it is informative to look at the evolution of each component separately.

As shown in Figures 14.3, improvements in the ES indicator have occurredin Asia, and they have been substantial in developed economies, although progresshas been either slow, or nonexistent, in the Latin American countries. Thus, in the dimension of transparency captured by the ES indicator, progress hasbeen slower on average in emerging markets. Observe that the value of the ES indicator for the median Asian and Latin American countries remains aboutone-third lower than the median of the developed country group as of 2003. Bycontrast, progress in the transparency dimension captured by the SPS indicatorhas been more pronounced in emerging markets than in developed economies (Figures 14.4). Indeed, Asian countries exhibit levels closer to those exhibited byother developed countries, whereas for Latin American countries SPS levelsremain significantly lower than those of developed countries, despite recentimprovements. Lastly, the AS indicator exhibits some improvement, albeitsmall, in most Asian countries, though the indicator exhibits virtually nochange in both Latin American and developed economies (Figures 14.5).

Despite the noted regional and country differences in the evolution of theindex, convergence toward higher values of the CGQ index has occurred, as indicatedby the negative and relatively large cross-country correlation (−0.53)between the average growth rate of the CGQ index during the 1994–2003 periodand the 1995 level. On average, countries with the fastest average rate of increaseof the index were indeed those witnessing the lowest levels of the index in 1995. For example, the gap between the CGQ index in Asian countries and thatrecorded for the United States, which is the highest among all countries in allyears, has narrowed since 1994. Notably, convergence has occurred at a relativelyfaster rate in the transparency dimension, as the correlations between initial levelsof the ES and SPS indicators and their average growth rates, equal to−0.74 and−0.67, respectively, are substantially higher in absolute value than the relevantcorrelation for the CGQ index.

In sum, corporate governance quality of nonfinancial firms listed in domesticstock markets has improved overall in almost every country considered during the1994–2003 period, and improvements have been witnessed primarily in thetransparency dimension captured by the ES and SPS indicators. Remarkably, convergence in corporate governance quality has indeed occurred within the setof countries considered.

Figure 14.3.Earning smoothing indicator, subperiod averages. This figure shows the average of the earnings smoothing indicator for each country for the two periods 1995 to 1996 and 2000 to 2003. The countries are grouped by region.

Figure 14.4.Stock price synchronicity indicator, subperiod averages. This figure shows the average of the Stock Price Synchronicity indicator for each country for the two periods 1995 to 1996 and 2000 to 2003. The countries are grouped by region.

Figure 14.5.Accounting standards indicator, subperiod averages. This figure shows the average of the accounting standards indicator for each country for the two periods 1995 to 1996 and 2000 to 2003. The countries are grouped by region.

A critical question is whether improvements in corporate governance qualityhave “real” effects. We address such questions next by measuring the impact ofour indicators on real economic outcomes.

14.4. The Real Effects of Corporate Governance Quality

Corporate governance quality may affect aggregate economic activity throughseveral channels. For example, improvements in corporate governance qualitymay impact positively on growth by lowering firms’ cost of funds and possiblyincreasing the supply of credit, thereby encouraging investment. Moreover, bettergoverned firms may align managers’ and claimholders’ interests more closely, providing stronger incentives for managers to achieve high firm productivitygrowth by the adoption of frontier technologies. As a result, capital in the corporatesector may be allocated more efficiently, and economy-wide productivitygrowth may increase.

More generally, corporate governance arrangements can be viewed as technologies that firms may adopt subject to the constraints of the institutional environmentin which they operate. Comin and Mulani (2005) formulate anendogenous growth model which embeds firms’ choices of “general innovations,”defined as innovations that are available and applicable to several firms and sectors, and whose “rents” or “benefits” are not privately appropriable, as in the caseof patentable research and development. Their model rationalizes several empiricalfacts concerning the dynamics of productivity growth at the aggregate level, as well as at an industry and firm level. Managerial and organizational innovationsare prominent examples of general innovations. If corporate governancearrangements are viewed as general innovations in the sense of Comin and Mulani, then they may have a significant impact on macroeconomic activity andproductivity growth.

To assess the link between corporate governance quality and macroeconomicactivity, we estimate two complementary statistical models that can be viewed asgeneric empirical counterparts of models of endogenous growth partly driven bygeneral innovations such as corporate governance arrangements. The first modelis a simple dynamic panel model that exploits both the time and cross-sectionaldimensions of the data. We use this setup to explore the impact of our measuresof corporate governance quality on GDP growth, on estimates of TFP growth, and on the ratio of investment to GDP. The second model exploits only thecross-sectional dimensions of the data, but expands such dimensions by includingindustry-level data. We use it to explore the impact of changes in our corporategovernance indicators on the growth of industries most dependent onexternal finance.

14.4.1. Impact on Growth, Productivity, and Investment

Our benchmark statistical setup is given by the following standard autore gressivedynamic panel model:

The dependent variable, Yit, denotes either GDP growth, TFP growth,13 andthe investment-to-GDP ratio for country i ϵ {1,..N } (N denotes the number ofcountries) in year t ∊ [1,..,T] (T denotes the terminal date of the sample). Theconstants αi capture time-invariant, unobserved country-specific effects. CGQit-1denotes the CGQ index or the vector of its components, and it is lagged becausewe assume it takes time to translate the effects of a change in corporate governancein a given year into macroeconomic outcomes. Of course, such a changewill affect values of the dependent variable beyond the subsequent year via itsautoregressive term. Xit denotes a vector that includes all other variables that affect Yit, and are log-transformed for the reasons detailed below. We include two lagsof the dependent variable to deal with potential serial correlation of the residuals.14 The errors εit are assumed to be identically, independently distributed anduncorrelated over time and across countries. Our focus is on estimates of theparameter vector β.15

We accomplish this estimation following two steps. First, using the difference operator Δxtxtxt1, equation (1) can be expressed as:

Note that if we could control exhaustively for each relevant country componentof the vector Xit, we would be able to obtain precise unbiased estimates of β using equation (2) . However, controlling for all relevant variables is likely to be a daunting task. Even if this could be done, we would rapidly exhaust our degrees of freedoms.

Alternatively, we can approximate, and control for, the effects of these variables by making assumptions on the data-generating process of Xit. Specifically, we assume that the vector Xit satisfies Δln(Xit)=Gi+vit are identically, independently distributed, and uncorrelated over time and across countries. This amounts to assuming that the continuously compounded growth rates of the variables in Xit are random. Next, define AiGiγandηitνitγ+Δit We make the further assumption that all vit are uncorrelated with CGQ it-1 and △εit. Underthis set of assumptions, we obtain the following fixed (country) effect dynamicpanel regression model in differenced variables:

Using equation (3), we effectively control for time-invariant country characteristicsby including country-fixed effects.16 Furthermore, to the extent that anytime-varying country characteristic that we have not controlled for is not correlatedwith changes in our corporate governance index, our inference remains valid.17

We estimate β by applying the difference GMM estimation procedure developedby Arellano and Bond (1991) to equation (3) . Because such estimation iscarried out on differenced variables, it is actually implemented through differencingof equation (3), which is equivalent to “double” differencing equation (1) . In this way, we are able to introduce an additional layer of country-specificeffects that are used to control for the deterministic component of all variables Xit. Only in the regressions of the ratio of investment to GDP as the dependentvariable, we have added the lagged value of GDP growth to control for businesscycle nent of all variablesnent of all variables effects.

Estimations are carried out using an unbalanced panel composed of annualdata of country-year observations for all countries listed previously for which datacould be constructed or were available (about 40 countries) during the period1993–2004. In all estimations, we treat both the lagged dependent variables andall independent variables as endogenous and instrument these variables usingtheir lags at time t-3, t-4, and so on, up to a maximum of nine lags. We reportresults for the one-step Arellano-Bond (1991) estimator, though Sargan specificationtests are based on the relevant two-step estimator.18 We first present theresults for the benchmark model, and subsequently we assess whether, and inwhat way, these results change under some modifications of the benchmarkmodel. In all the benchmark estimates, as well as in virtually all subsequent ones, the autocorrelation and specification tests indicate that coefficient estimates areunbiased and the specification of the model is satisfactory.

14.4.1.1. Benchmark Results

As shown in Table 14.3, estimates of the benchmark model yield three main results. First, GDP growth, TFP growth, and the ratio of investment to GDP vary positivelyand significantly with lagged values of the CGQ index. Second, changes incorporate governance quality have a significant economic impact on GDP, TFPgrowth, and the ratio of investment to GDP. Namely, a one-standard deviationincrease in the CGQ index in the current year results in an increase in GDP growthof about 2 percentage points (0.02 = 0.21*0.09), an increase in TFP growth ofabout 2 percentage points (0.02 = 0.24*0.09), and an increase in the ratio ofinvestment to GDP of about 1 percentage point (0.93 = 10.306*0.09), in the followingyear. These are substantial effects compared to the averages for these variables(see Table 14.9 for the summary statistics of the main regression variables). Third, the positive dynamic relationship between all measures of macroeconomicoutcomes and corporate governance quality appears to be driven by improvementsin transparency, because the coefficients associated with the SPS indicator arepositive and significant in all regressions. In this specification, the SPS indicatorappears to be the main component driving the significance of the overall CGQindex. Given the lack of significance of the AS indicator, this suggests that whatmatters for firm performance is the transparency of the accounting informationdisclosed (as captured by the SPS indicator) rather than the amount of informationdisclosed (as captured by the AS indicator).19 The Sargan test for overidentifyingrestrictions and the Arellano and Bond (1991)m2 test for second-order serial correlationof the error term both support the benchmark model specification and donot detect any problems.20

Table 14.3Aggregate Economic Activity and Corporate Governance: Benchmark Model
DEPENDENT VARIABLES
GDP Growth(t)TFP Growth(t)INV to GDP(t)
(1)(2)(3)(4)(5)(6)
CGQ index(t-1)0.209*0.238**10.306**
(1.88)(2.10)(2.06)
Earnings smoothing(t-1)0.029–0.0050.352
(0.93)(–0.14)(0.34)
Price synchronicity(t-1)0.071**0.101***5.639**
(2.01)(3.09)(2.26)
Accounting standards(t-1)–0.081–0.4175.847
(–0.42)(–1.25)(0.97)
GDP growth(t-1)11.143***11.062***
(3.09)(3.26)
Dependent variable(t-1)–0.496***–0.513***–0.494***–0.525***0.195*0.168
(–6.41)(–6.63)(–7.40)(–7.25)(1.91)(1.60)
Dependent variable(t-2)–0.322***–0.317***–0.355***–0.348***–0.153***–0.149***
(–6.01)(–5.93)(–9.01)(–7.26)(–3.60)(−3.64)
Number of countries/obs.40/27140/27140/23440/23440/27140/271
M1 (p-value)0.000.000.000.000.010.01
M2 (p-value)0.330.340.330.920.930.97
Sargan test (p-value)1.001.000.851.001.001.00
CGQ, corporate governance quality; INV, investment; TFP, total factor productivity.Estimates are obtained by the (difference) GMM one-step estimator of Arellano and Bond (1991) applied to equation (3), where both the lagged dependent and independent variables are treated as endogenous and are instrumented with all their lags at t-3, t-4, and so on, up to a maximum of nine lags. Robust t-statistics are reported in brackets; * denotes significant at 10%; ** significant at 5%; *** significant at 1%; M1 and M2 is the p-value of the Arellano Bond statistics for second-order correlation of residuals; Sargan test is the p-value obtained by estimates of the two-step version of the model.
CGQ, corporate governance quality; INV, investment; TFP, total factor productivity.Estimates are obtained by the (difference) GMM one-step estimator of Arellano and Bond (1991) applied to equation (3), where both the lagged dependent and independent variables are treated as endogenous and are instrumented with all their lags at t-3, t-4, and so on, up to a maximum of nine lags. Robust t-statistics are reported in brackets; * denotes significant at 10%; ** significant at 5%; *** significant at 1%; M1 and M2 is the p-value of the Arellano Bond statistics for second-order correlation of residuals; Sargan test is the p-value obtained by estimates of the two-step version of the model.

14.4.1.2. Accounting for Financial Crises

We wish to establish whether the benchmark results are primarily driven by observationssampled during “crisis” years, defined as years characterized by eitheroutput drops, sharp currency devaluations, stock market crashes, systemic bankfailures, or combinations of all these occurrences. If this were the case, the impactof corporate governance quality on macroeconomic outcomes (parameter β)would likely be estimated imprecisely, since even shocks that are “temporary”relative to a long time span would necessarily appear as “long-lasting” in the shorttime dimension of our data. Moreover, and related to some of the components ofour CGQ index, our estimates could capture effects not necessarily related tocorporate governance. For example, the high synchronicity of individual stockreturns occurring during stock market crashes may coincide with sharp declinesin GDP per capita, generating a temporarily high comovement between GDPgrowth and synchronicity in stock returns.21 In addition, during crisis periodsfirms may try even harder not to disclose information and overstate firm performance, resulting in an unusually high reporting of accounting accruals. Again, this could generate a temporarily high comovement between ES and macroeconomicoutcomes, which would be reflected as a relatively “long-lasting” shock inour estimation.

To cope with this issue, we first defined “crises” country-years if there was either a negative value of GDP growth (an output drop), or a negative change instock market capitalization (a stock market drop), or a banking crisis, identifiedas the initial year and the year subsequent to a banking crisis date as classified by Laeven and Honohan (2005). Then, we estimated the benchmark model on asample where all crisis country-years were dropped, that is, on a sample of “noncrisis”country-years.

Table 14.4Aggregate Economic Activity and Corporate Governance: Excluding “Crisis” Country-Years Observations
DEPENDENT VARIABLES
GDP Growth (t)TFP Growth (t)INV to GDP (t)
(1)(2)(3)(4)(5)(6)
CGQ index(t-1)0.255*0.1738.035**
(1.70)(1.47)(2.03)
Earnings smoothing(t-1)0.041−0.0110.062
(1.16)(−0.36)(0.51)
Price synchronicity(t-1)0.105**0.109***4.318***
(2.01)(2.60)(2.73)
Accounting standards(t-1)−0.035−0.5962.539
(−0.23)(−1.54)(0.38)
GDP growth(t-1)10.903***12.265***
(3.03)(2.95)
Dependent variable(t-1)−0.550***−0.576***−0.407***−0.521***0.0580.016
(−4.88)(−4.95)(−6.05)(−5.93)(0.31)(0.09)
Dependent variable(t-2)−0.384***−0.409***−0.336***−0.385***−0.341***−0.308***
(−5.66)(−4.71)(−7.06)(−7.55)(−3.89)(−3.49)
Number of countries/obs.40/22540/22540/19840/19840/22540/225
M1 (p-value)0.000.020.000.020.030.00
M2 (p-value)0.060.120.630.910.130.11
Sargan test (p-value)1.001.000.781.001.001.00
CGQ, corporate governance quality; INV, investment; TFP, total factor productivity.Estimates are obtained by the (difference) GMM one-step estimator of Arellano and Bond (1991) applied to equation (3), where both the lagged dependent and independent variables are treated as endogenous and are instrumented with all their lags at t-3, t-4, and so on, up to a maximum of nine lags. Robust t-statistics are reported in brackets; * denotes significant at 10%; ** significant at 5%; *** significant at 1%; M1 and M2 is the p-value of the Arellano Bond statistics for second-order correlation of residuals; Sargan test is the p-value obtained by estimates of the two-step version of the model.
CGQ, corporate governance quality; INV, investment; TFP, total factor productivity.Estimates are obtained by the (difference) GMM one-step estimator of Arellano and Bond (1991) applied to equation (3), where both the lagged dependent and independent variables are treated as endogenous and are instrumented with all their lags at t-3, t-4, and so on, up to a maximum of nine lags. Robust t-statistics are reported in brackets; * denotes significant at 10%; ** significant at 5%; *** significant at 1%; M1 and M2 is the p-value of the Arellano Bond statistics for second-order correlation of residuals; Sargan test is the p-value obtained by estimates of the two-step version of the model.

As shown in Table 14.4, all our parameter estimates remain virtuallyunchanged, although the significance of the results of TFP growth is somewhatreduced. Thus, our results do not appear to be driven by crisis periods.

14.4.1.3. Accounting for Complex Dynamics

In the benchmark model a change in corporate governance quality affectsmacroeconomic outcomes with a one-year lag, and impacts on their futurevalues through the persistence parameters δ1 and δ2. In reality, changes incorporate governance may take a longer time to exert their effects on macroeconomicoutcomes and could be highly persistent. In addition, crisis andrecovery from crisis may well create complicated dynamic paths which mightnot be effectively captured by the simple lag structure of the benchmarkmodel. For example, if improving governance is costly to the firm, but itscost varies according to whether or not a crisis is unfolding, then a firmdynamic decision to improve governance could create a complex interactionwith the level of macroeconomic activity. Statistically, in these cases theassumption of independent distribution of the errors over time in equation (3) may be too strong.

Table 14.5Aggregate Economic Activity and Corporate Governance: Accounting for Complex Dynamics
DEPENDENT VARIABLES
GDPTFPINVGDPTFPINV
Growth (t)Growth (t)to GDP (t)Growth (t)Growth (t)to GDP (t)
(1)(2)(3)(4)(5)(6)
Whole Sample"Noncrisis" Sample
CGQ index (t-1)0.226*0.200*8.721*0.276*0.1698.589**
(1.81)(1.72)(1.90)(1.62)(1.43)(2.12)
CGQ index (t-2)0.0450.048−0.523−0.0070.0450.291
(0.59)(0.45)(−0.17)(−0.11)(0.59)(0.08)
GDP Growth (t-1)11.967***11.530***
(3.60)(3.05)
Dependent variable (t-1)−0.618***−0.722***0.093−0.643***−0.642***0.034
(−6.20)(−10.38)(0.74)(−5.54)(−9.06)(0.18)
Dependent variable (t-2)−0.469***−0.559***−0.137***−0.546***−0.516***−0.306****
(−5.38)(−8.11)(−3.59)(−7.09)(−5.75)(−3.89)
Dependent variable(t-3)−0.227***−0.281***−0.114**−0.233***−0.242***−0.002
(−3.67)(−6.26)(−2.05)(−2.92)(−4.46)(−0.05)
Number of countries/obs.40/23140/19440/23130/19630/16530/196
M1 (p-value)0.000.040.010.010.000.06
M2 (p-value)0.210.160.790.030.170.13
Sargan test (p-value)1.000.960.771.000.971.00
CGQ, corporate governance quality; INV, investment; TFP, total factor productivity.Estimates are obtained by the (difference) GMM one-step estimator of Arellano and Bond (1991) applied to equation (3), where both the lagged dependent and independent variables are treated as endogenous and are instrumented with all their lags at t-3, t-4, and so on, up to a maximum of nine lags. Robust t-statistics are reported in brackets; * denotes significant at 10%; ** significant at 5%; *** significant at 1%; M1 and M2 is the p-value of the Arellano Bond statistics for second-order correlation of residuals; Sargan test is the p-value obtained by estimates of the two-step version of the model.
CGQ, corporate governance quality; INV, investment; TFP, total factor productivity.Estimates are obtained by the (difference) GMM one-step estimator of Arellano and Bond (1991) applied to equation (3), where both the lagged dependent and independent variables are treated as endogenous and are instrumented with all their lags at t-3, t-4, and so on, up to a maximum of nine lags. Robust t-statistics are reported in brackets; * denotes significant at 10%; ** significant at 5%; *** significant at 1%; M1 and M2 is the p-value of the Arellano Bond statistics for second-order correlation of residuals; Sargan test is the p-value obtained by estimates of the two-step version of the model.

To assess whether the benchmark model is a reasonable approximation of thedata-generating process in this dimension, we augmented the lag structure of themodel subject to the constraints imposed by the time span of our data. Specifically, we estimated equation (1) with two additional lagged values of the corporategovernance indicator, and one additional autoregressive term (at time t-3), bothfor the whole sample and the “noncrisis”sample defined previously.

As shown in Table 14.5, the qualitative results obtained previously remainunchanged. In addition, these estimates provide some useful insights. First, higherlaggedvalues of the CGQ index do not enter significantly, indicating that the benchmarkspecification of the lagged structure for this variable is not off the mark. Second, all three lags of the dependent variable are statistically significant, indicating a highpersistence of the impact of improvements in corporate governance. Third, the coefficientsfor the “noncrisis” sample are comparable in size to those of the whole sample, suggesting that during crisis periods the effects of improvements in corporate governancemay not be all that different from during “noncrisis” periods.22

14.4.1.4. Accounting for Financial Development

Identifying the impact of corporate governance quality per se on aggregate economicactivity is complicated because other interrelated factors may be at play. Among these, financial development is of particular importance, because suchdevelopment may be both a function, and a potentially important determinant, of corporate governance quality. For example, if firms cannot achieve the potentialreduction in borrowing costs arising from improvements in corporate governancebecause the capacity of the financial sector to price risk is underdeveloped, then their incentives to improve corporate governance may be limited. There isalso a growing literature on how transparency improves the operation of banksand other financial intermediaries (e.g., Barth, Caprio, and Levine, 2004, 2006 ;Demirgüç-Kunt, Laeven, and Levine, 2004 ; and Beck, Demirgüç-Kunt, and Levine, 2007). Given that banks as creditors and equity holders play an importantrole in governing firms, an improvement in the transparency and governance ofbanks may also increase the likelihood that banks exert sound governance over thefirms they fund. In addition, financial development per se may be an importantdeterminant of macroeconomic outcomes. In terms of the benchmark model, variables related to financial development may be dynamically correlated with ourcorporate governance indicators, making it necessary to take them explicitly intoaccount to mitigate the potential biases in the estimates.

To account for financial development and its potential interaction with corporategovernance, we consider the following extension of the benchmark model:

where FDit-1 is the lagged value of a proxy measure of financial development, given by the sum of private credit and stock market capitalization to GDP. Asbefore, we estimate parameters β1, β2, and β3 by applying the difference GMMestimator to the regression:

As shown in Table 14.6, although the qualitative results are essentially thesame as those obtained with the benchmark model, they provide evidence of thecomplementarities between corporate governance and financial development wehave emphasized. In fact, the interaction terms between these variables are positiveand significant in all regressions except the ones with the ratio of investmentto GDP as the dependent variable. That is, the economic impact of improvementsin corporate governance on macroeconomic outcomes is greater the moredeveloped is the financial sector. Furthermore, although the “autonomous”impact of the SPS indicator continues to be positive and significant in all regressions, now the ES measure too exhibits a positive and significant effect on GDPgrowth and TFP growth when interacted with the financial development proxy. This result is consistent with the role of a developed financial sector in enhancingthe impact of improvements in corporate governance. The SPS measure exhibitsa positive and significant effect only on TFP growth when interacted with thefinancial development proxy. The AS measure enters positively and significantlyin the investment to GDP regression at low levels of financial development andturns negative or insignificant at high levels of financial development. However, the AS variables turn insignificant when not controlling for the ES and SPS variablesand interaction terms, suggesting that this result is because of collinearitybetween the three governance measures.

Table 14.6Aggregate Economic Activity and Corporate Governance: Accounting for Financial Development
DEPENDENT VARIABLES
GDP Growth (t)TFP Growth (t)INV to GDP (t)
(1)(2)(3)(4)(5)(6)
CGQ index(t-1)0.1470.134*8.169**
(1.57)(1.79)(2.13)
Financial development(t-1)0.027**0.029**0.026**0.024**1.0191.424**
(2.19)(2.54)(2.58)(2.39)(1.58)(2.29)
CGQ index(t-1)*0.736***0.527***−2.925
Financial development(t-1)(3.56)(5.91)(−0.13)
Earnings smoothing(t-1)−0.0050.002−0.572
(−0.22)(0.08)(−0.69)
Price synchronicity(t-1)0.080**0.054**5.671**
(2.00)(2.07)(2.29)
Accounting standards(t-1)−0.125−0.4317.535*
(−0.98)(−1.67)(1.75)
Earnings smoothing(t-1)*0.295***0.190***4.294
Financial development (t-1)(3.26)(3.42)(0.73)
Price synchronicity(t-1)*0.1370.223**−3.484
Financial development (t-1)(0.70)(2.19)(−0.29)
Accounting standards(t-1)*−0.0140.538−49.194***
Financial development (t-1)(−0.03)(1.14)(−3.02)
GDP growth(t-1)11.705***11.585***
(3.21)(3.45)
Dependent variable(t-1)−0.578***−0.589***−0.549***−0.567***0.1420.108
(−6.95)(−7.32)(−9.41)(−9.68)(1.31)(0.90)
Dependent variable(t-2)−0.379***−0.364***−0.377***−0.365***−0.162***−0.147***
(−6.44)(−5.84)(−8.53)(−7.07)(−3.72)(−2.93)
Number of countries/obs.39/25139/25139/21639/25139/25139/251
M1 (p-value)0.000.000.010.000.010.01
M2 (p-value)0.380.370.490.880.630.93
Sargan test (p-value)1.001.001.001.001.001.00
CGQ, corporate governance quality; INV, investment; TFP, total factor productivity.Estimates are obtained by the (difference) GMM one-step estimator of Arellano and Bond (1991) applied to equation (3), where both the lagged dependent and independent variables are treated as endogenous and are instrumented with all their lags at t-3, t-4, and so on, up to a maximum of nine lags.. Robust t-statistics are reported in brackets; * denotes significant at 10%; ** significant at 5%; *** significant at 1%; M1 and M2 is the p-value of the Arellano Bond statistics for second-order correlation of residuals; Sargan test is the p-value obtained by estimates of the two-step version of the model.
CGQ, corporate governance quality; INV, investment; TFP, total factor productivity.Estimates are obtained by the (difference) GMM one-step estimator of Arellano and Bond (1991) applied to equation (3), where both the lagged dependent and independent variables are treated as endogenous and are instrumented with all their lags at t-3, t-4, and so on, up to a maximum of nine lags.. Robust t-statistics are reported in brackets; * denotes significant at 10%; ** significant at 5%; *** significant at 1%; M1 and M2 is the p-value of the Arellano Bond statistics for second-order correlation of residuals; Sargan test is the p-value obtained by estimates of the two-step version of the model.

14.4.1.5. Long-Run Effects

The “long-run” effects of corporate governance quality on macroeconomic outcomesare given by estimates of the parameter β/(1δ1δ2). Although we have established that this parameter is positive and significant, the time span of ourdata is too short to allow us to measure these long-run effects with precision. Moreover, pinning down such effects with cross-country regressions would bedifficult, because one would need to identify through theory, and control explicitlyfor, a host of country-specific, possibly endogenous, variables. Besides, withonly about 40 observations in the sample and the potential need to considermany control variables, our degrees of freedom would be rapidly exhausted andthe precision of our estimates would likely be unsatisfactory. Rather than pursuingthis avenue, we complement the foregoing analysis by expanding the crosssectionaldimension of the data in order to focus on the differential effect ofimprovements of corporate governance on long-run industry growth.

14.4.2. Impact on Growth of Financial Dependent Industries

As noted, the returns of good corporate governance are likely to be the largestwhen firms are able to attain easier and less costly access to finance, and suchaccess is critical for their growth prospects. Therefore, we would expect that thebenefits of improvements in corporate governance quality would be the largest forfinancially dependent industries.

To assess this conjecture, consider the following industry-level counterpart ofmodel (1):

with t ϵ [1,..,T ], i ϵ {1,..,N }, j ϵ{1,..,M }, where M is the number of industries, Yijt is the continuously compounded real growth rate of industry j in country i, α^i and α^j are fixed country and industry effects respectively, Xit are firm-specificvariables, Zjt are country-specific variables, Wijt are firm-country-specific variables, and ηijt is the error term.

Under the assumption that all right-hand side variables grow at a constant deterministic rate during the period [1,..,T]. in a steady state we obtain the following regression model:

whereαi=(1δ)1(α^i+Xiβ1),αj=(1δ)1(α^j+β1),β^=(1δ)1β3andεij=T1t=1Tηijt and all variables without time subscript denote their relevant constant growth rates.

As noted previously, precise and unbiased estimates of β^ would be obtained ifwe could control exhaustively for each relevant component of the vector Wij. Yet, this is a task even more difficult than that we faced before, because it wouldrequire identification of a host of country- and industry-specific variables. Forthese reasons, we employ an approach similar to the one developed by Rajan and Zingales (1998), and estimate the following benchmark industry-level regression:

where Growthij is real sales growth over the period 1995 to 2003 of industry j incountry i, calculated at the ISIC industry level and weighted by the lagged valueof market capitalization of individual firms, Shareij is the share of the industry intotal real sales of the country in 1995, CGQ i is the level of corporate governanceof country i in 1995, FDi is the level of financial development of country i in1995, and EDj is the Rajan and Zingales (1998) measure of external financialdependence, calculated at the two- or three-digit ISIC industry level over theperiod 1980 to 1989. The sample consists of the 36 manufacturing industriescovered by Rajan and Zingales (1998) in a total of 34 countries. We include theindustry share in total sales to capture a potential convergence effect, becauseindustries that are large relative to other industries in the country are expected togrow at lower rates. Rajan and Zingales (1998) show that the growth of financiallydependent industries is disproportionally higher in countries with morefinancial development.

By augmenting the specification in Rajan and Zingales (1998) with the interactionbetween the level of corporate governance quality and external financialdependence, we can disentangle the effect of corporate governance from the effectof financial development on growth and assess the differential effect of corporategovernance quality on growth net of the impact of financial development. Following Rajan and Zingales (1998) we use the sum of private credit and stockmarket capitalization to GDP as the measure of financial development.23

This specification has the advantage over a pure cross-country regression inthat it controls for country- and industry-fixed effects. However, it rests on theassumption that the vector Wij is only composed of three elements, the variable Shareij and the interaction terms CGQi *EDj and FDi*EDj, that is, Wijβ^=γ*Shareij+β*CGQi*EDj+δ*FDi*EDj.. In what follows, we also consider aricher specification of the vector Wij which includes a triple interaction betweenfinancial dependence, corporate governance, and financial development.

It is worth stressing that this specification only allows us to measure the differentialeffect of corporate governance (controlling for financial development) onoutcome measures of economic performance, but not level effects. That is, we canmeasure whether improvements in corporate governance disproportionatelyaffect the growth of industries that are most likely to benefit from such improvements, but we cannot measure whether improvements in corporate governancedirectly affect the growth of all industries.

The industry characteristic of interest is the degree of external financial dependence, measured as the share of investment not financed by operating cash-flow(see Rajan and Zingales, 1998), because we expect industries that rely more onoutside finance to benefit most from improvements in corporate governancebecause it should help them to attract external financing for investment.

Table 14.7 reports the regression results of our basic specification in model (8). In addition to using our overall index of corporate governance, we also runregressions for each component of the governance index. We include the level ofthe AS, ES, and SPS indicators in 1995 in the regressions. To reduce outliers, werestrict growth rates in real sales to -1 and +2.

Table 14.7Industry Growth, Financial Dependence, and Corporate Governance
(1)(2)(3)(4)(5)
Share in industry sales0.0530.0530.0500.0490.054
(0.034)(0.032)(0.033)(0.032)(0.033)
CGQ index * Financial dependence0.409***
(0.121)
Earnings smoothing * Financial dependence0.088*0.119**
(0.044)(0.052)
Price synchronicity * Financial dependence0.148**0.171***
(0.060)(0.056)
Accounting standards * Financial dependence0.1240.068
(0.197)(0.210)
Financial development * Financial dependence0.0070.0090.0060.0060.007
(0.007)(0.007)(0.007)(0.007)(0.007)
Number of countries3333333333
Number of industries3636363636
Number of country-industry observations586586586586586
R-squared0.580.570.570.570.58
CGQ, corporate governance quality.The dependent variable is real sales growth over the period 1995 to 2003, calculated at the ISIC industry level. Share in industry sales is the share of the industry in total real sales of the country in 1995. Financial dependence is the Rajan and Zingales, (1998) measure of external financial dependence, calculated at the 2- or 3-digit ISIC industry level. CGQ index is our country-level index of corporate governance, and is the average of the Earnings smoothing, Price synchronicity, and Accounting standards indicators. We include the value of these corporate governance scores in 1995 in the regressions. Financial development is private credit plus stock market capitalization to GDP in 1995. We lose two countries because of missing data on financial development: we do not have data on private credit for China and we do not have data on stock market capitalization for Ireland. All regressions include country- and industry-fixed effects. We report White’s heteroskedasticity-consistent standard errors between brackets. Standard errors are corrected for clustering at the industry level. * denotes significant at 10%; ** significant at 5%; and *** significant at 1%.
CGQ, corporate governance quality.The dependent variable is real sales growth over the period 1995 to 2003, calculated at the ISIC industry level. Share in industry sales is the share of the industry in total real sales of the country in 1995. Financial dependence is the Rajan and Zingales, (1998) measure of external financial dependence, calculated at the 2- or 3-digit ISIC industry level. CGQ index is our country-level index of corporate governance, and is the average of the Earnings smoothing, Price synchronicity, and Accounting standards indicators. We include the value of these corporate governance scores in 1995 in the regressions. Financial development is private credit plus stock market capitalization to GDP in 1995. We lose two countries because of missing data on financial development: we do not have data on private credit for China and we do not have data on stock market capitalization for Ireland. All regressions include country- and industry-fixed effects. We report White’s heteroskedasticity-consistent standard errors between brackets. Standard errors are corrected for clustering at the industry level. * denotes significant at 10%; ** significant at 5%; and *** significant at 1%.

We find a strong and positive effect of corporate governance on the growth offinancially dependent industries. The effect is statistically significant at the5 percent level (column 1 of Table 14.7). We also find a disproportionate positiveand significant effect on the growth of financial dependent industries for the indicesof earnings smoothing and stock price synchronicity. For the index of accountingstandards, the effect, though positive, is not measured precisely and is notstatistically significant from zero. The interaction between financial developmentand financial dependence also does not enter significantly. In column 5, we includeall three subcomponents of the corporate governance index and find similar results.

The economic effect of the result on stock price synchronicity is also significant. Take the regression in column 1 of Table 14.7 . The coefficient of this regressionsuggests that an industry at the 75th percentile of financial dependence in acountry at the 75th percentile of the corporate governance index has a growth ratethat is 0.10 (= 0.409*(0.625*0.452 - 0.571*0.070)) higher than an industry atthe 25th percentile of financial dependence in a country at the 25th percentile ofcorporate governance (see Table 14.10 for the summary statistics of the mainregression variables). This is a substantial effect compared to the average growthrate of 0.10 (i.e., about one time average growth).

In Table 14.8, we wish to assess whether the effect depends on whether countrieswere affected by a banking crisis or not, as (see Table 14.11)we did in the panel regressions inTable 14.4 . As noted, we wish to check that the results are not driven by thecrisis countries. In fact, Kroszner, Laeven, and Klingebiel (2007) show that thepositive effect of financial development on the growth of financially dependentindustries identified by Rajan and Zingales (1998) disappears during crises, andthe same may be true for the effect of corporate governance on the growth offinancially dependent industries. We use data from Laeven and Honohan (2005) to identify banking crises. In columns 1 to 5, we rerun the regressions for thesubset of countries without banking crises during the sample period 1995–2003. We confirm our previous result on the overall corporate governance index(although the size of the effect decreases somewhat), suggesting that the effect wefind is not driven by crisis countries. The interaction between earnings smoothingand financial dependence now enters positively and significantly at the 5 percentlevel (column 2), though the interaction between stock price synchronicity andfinancial dependence no longer enters significantly at the 10 percent level.

Table 14.8Industry Growth, Financial Dependence, and Corporate Governance: Excluding Crisis Countries
(1)(2)(3)(4)(5)
Share in industry sales0.0260.0280.0250.0220.026
(0.047)(0.047)(0.045)(0.038)(0.046)
CGQ index * Financial dependence0.281**
(0.110)
Earnings smoothing * Financial dependence0.144**0.116
(0.064)(0.090)
Price synchronicity * Financial dependence0.0860.047
(0.092)(0.109)
Accounting standards * Financial dependence0.2250.133
(0.239)(0.260)
Financial development * Financial dependence0.0030.0050.0050.0020.003
(0.008)(0.008)(0.008)(0.011)(0.008)
Number of countries2525252525
Number of industries3636363636
Number of country-industry observations453453453453453
R-squared0.470.470.470.470.47
CGQ, corporate governance quality.The dependent variable is real sales growth over the period 1995 to 2003, calculated at the ISIC industry level. Share in industry sales is the share of the industry in total real sales of the country in 1995. Financial dependence is the Rajan and Zingales (1998) measure of external financial dependence, calculated at the 2- or 3-digit ISIC industry level. CGQ index is our country-level index of corporate governance, and is the average of the Earnings smoothing, Price synchronicity, and Accounting standards indicators. We include the value of these corporate governance scores in 1995 in the regressions. Financial development is private credit plus stock market capitalization to GDP in 1995. We lose two countries because of missing data on financial development: we do not have data on private credit for China and we do not have data on stock market capitalization for Ireland. All regressions include country- and industry-fixed effects. We report White’s heteroskedasticity-consistent standard errors between brackets. Standard errors are corrected for clustering at the industry level. * denotes significant at 10%; ** significant at 5%; and *** significant at 1%.
CGQ, corporate governance quality.The dependent variable is real sales growth over the period 1995 to 2003, calculated at the ISIC industry level. Share in industry sales is the share of the industry in total real sales of the country in 1995. Financial dependence is the Rajan and Zingales (1998) measure of external financial dependence, calculated at the 2- or 3-digit ISIC industry level. CGQ index is our country-level index of corporate governance, and is the average of the Earnings smoothing, Price synchronicity, and Accounting standards indicators. We include the value of these corporate governance scores in 1995 in the regressions. Financial development is private credit plus stock market capitalization to GDP in 1995. We lose two countries because of missing data on financial development: we do not have data on private credit for China and we do not have data on stock market capitalization for Ireland. All regressions include country- and industry-fixed effects. We report White’s heteroskedasticity-consistent standard errors between brackets. Standard errors are corrected for clustering at the industry level. * denotes significant at 10%; ** significant at 5%; and *** significant at 1%.

Thus far, we have tested whether the governance effect we identify is independentof the financial development effect identified by Rajan and Zingales (1998). In Table 14.9, we further corroborate the results in Rajan and Zingales (1998) byincluding a triple interaction term between financial dependence, corporate governancequality, and financial development to test whether the effect we find forcorporate governance is conditional on the level of financial development of thecountry. This is indeed what we find. The sign of the coefficient on the tripleinteraction term is positive, suggesting that the effect we found previously on theinteraction between financial dependence and the CGQ index is more pronouncedin more financially developed countries, although the effect is notstatistically significant in all cases.24 The estimates indicate that this effect operatesonly at high levels of financial development.

Table 14.9Controlling for Financial Development
(1)(2)(3)(4)(5)(6)
No CrisisNo CrisisNo Crisis
Share in industry sales0.0550.0550.0540.0290.0250.023
(0.033)(0.033)(0.032)(0.046)(0.045)(0.044)
CGQ index * Financial dependence−0.337−0.846*
(0.442)(0.470)
Earnings smoothing * Financial dependence−0.112−0.211−0.401−0.584
(0.165)(0.256)(0.243)(0.704)
Price synchronicity * Financial dependence−0.110−0.106−0.448**−0.331
(0.180)(0.224)(0.175)(0.238)
Accounting standards * Financial dependence−0.116−0.295−0.042−0.051
(0.198)(0.357)(0.186)(0.566)
Financial development * Financial dependence−0.268*−0.267−0.373**−0.468**−0.594**−0.568*
(0.157)(0.173)(0.166)(0.185)(0.227)(0.294)
CGQ index * Financial development0.468*0.4660.789**0.992**
* Financial dependence(0.263)(0.289)(0.312)(0.378)
Earnings smoothing * Financial development0.2610.449
* Financial dependence(0.175)(0.413)
Price synchronicity * Financial development0.1180.257*
* Financial dependence(0.111)(0.132)
Accounting standards * Financial development0.302*0.357
* Financial dependence(0.172)(0.272)
Number of countries333333252525
Number of industries363636363636
Number of country-industry observations586586586453453453
R-squared0.580.580.580.480.480.48
CGQ, corporate governance quality.The dependent variable is real sales growth over the period 1995 to 2003, calculated at the ISIC industry level. Share in industry sales is the share of the industry in total real sales of the country in 1995. Financial dependence is the Rajan and Zingales (1998) measure of external financial dependence, calculated at the 2- or 3-digit ISIC industry level. CGQ index is our country-level index of corporate governance, and is the average of the Earnings smoothing, Price synchronicity, and Accounting standards indicators. We include the value of these corporate governance scores in 1995 in the regressions. Financial development is private credit plus stock market capitalization to GDP in 1995. We lose two countries because of missing data on financial development: we do not have data on private credit for China and we do not have data on stock market capitalization for Ireland. In regressions (4) to (6), we report results for the subset of countries that did not experience a banking crisis during the period 1995–2003. We use data from Laeven and Honohan (2005) to identify banking crises. All regressions include country- and industry-fixed effects. We report White’s heteroskedasticity-consistent standard errors between brackets. Standard errors are corrected for clustering at the industry level. * denotes significant at 10%; ** significant at 5%; and *** significant at 1%.
CGQ, corporate governance quality.The dependent variable is real sales growth over the period 1995 to 2003, calculated at the ISIC industry level. Share in industry sales is the share of the industry in total real sales of the country in 1995. Financial dependence is the Rajan and Zingales (1998) measure of external financial dependence, calculated at the 2- or 3-digit ISIC industry level. CGQ index is our country-level index of corporate governance, and is the average of the Earnings smoothing, Price synchronicity, and Accounting standards indicators. We include the value of these corporate governance scores in 1995 in the regressions. Financial development is private credit plus stock market capitalization to GDP in 1995. We lose two countries because of missing data on financial development: we do not have data on private credit for China and we do not have data on stock market capitalization for Ireland. In regressions (4) to (6), we report results for the subset of countries that did not experience a banking crisis during the period 1995–2003. We use data from Laeven and Honohan (2005) to identify banking crises. All regressions include country- and industry-fixed effects. We report White’s heteroskedasticity-consistent standard errors between brackets. Standard errors are corrected for clustering at the industry level. * denotes significant at 10%; ** significant at 5%; and *** significant at 1%.

14.5. Conclusion

The chapter has constructed new measures of corporate governance quality for abroad set of developed and emerging market countries based on recent advancesin the finance literature. Contrary to existing indicators of corporate governancebased on tracking changes in de jure governance laws and regulations, our indexreflects the actual outcome of governance in the marketplace. This is important, because legal changes may not necessarily reflect actual outcomes owing to implementationlags, and because corporate governance quality may be an importantfirm decision, which can change relatively independently of the institutional environmentin which firms operate. For a large set of countries during the period1994 to 2003, our measures indicate that corporate governance quality hasimproved in almost all countries, and there is evidence of convergence.

We have gauged the “real” effects of corporate governance quality throughestimation of a set of dynamic panel regression models for GDP growth, TFPgrowth, and the ratio of investment to GDP, and cross-sectional regressions ofgrowth at the industry level.

Overall, our evidence suggests that improvements in corporate governancequality have a positive and significant effect on all measures of macroeconomicoutcomes considered. This is true especially in the transparency dimension, asshown by the positive and significant impact of the SPS indicator, although alldimensions of corporate governance captured by our measures show a similarimpact in countries with more developed financial sectors. In addition, at theindustry level, we find that improvements in corporate governance appear topositively affect the performance of industries that depend on external finance.

These results are consistent with the notion that well-governed firms incorporatebetter managerial incentives that are likely to spur corporate sector growthand improve its productivity independently of the level of financial development. However, we also find that a higher level of financial development boosts thepositive effects of improvements of corporate governance on macroeconomicoutcomes, consistent with the notion that well-governed firms are better able toattract outside financing.

In sum, these findings suggest that it is actual, not necessarily legal, changes incorporate governance that really matter. Thus, our findings call for additionalwork to collect new data and compare a broad set of de jure and outcome-basedmeasures of corporate governance rules and practices. We believe that such comparisonswould enhance our understanding of the drivers of improvements in thequality of corporate governance and their real effects.

Appendix I
Table 14.10Summary Statistics of Main Variables in Country Panel Regressions
VariableObs.MeanMedianSth. Dev.25th Percentile75th Percentile
GDP growth3280.0210.0210.0350.0070.038
TFP growth2870.0090.0120.032−0.0000.023
INV to GDP32821.49320.7134.69418.76923.504
CGQ index3280.5530.5640.0950.4930.618
CGQ, corporate governance quality; INV, investment; TFP, total factor productivity.
CGQ, corporate governance quality; INV, investment; TFP, total factor productivity.
Table 14.11Summary Statistics of Main Variables in Industry Panel Regressions
VariableObs.MeanMedianSth. Dev.25th Percentile75th Percentile
Real sales growth6010.0930.0860.0930.0330.141
CGQ index360.5960.5950.0490.5710.625
Earnings smoothing360.1460.1210.1080.0800.187
Price synchronicity360.8100.8240.1150.7820.902
Accounting standards360.8340.8390.0410.8100.860
Financial development341.4181.1360.9050.7442.074
Financial dependence360.3190.2310.4060.0700.452
CGQ, corporate governance quality.
CGQ, corporate governance quality.
References

    ArellanoM.BondS.1991“Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations,”Review of Economic StudiesVol. 58pp. 277–97.

    BarthJ. R.G.Caprio R.Levine2004“Bank Supervision and Regulation: What Works Best?”Journal of Financial IntermediationVol. 13pp. 205–48.

    BarthJ. R.G.Caprio R.Levine2006Rethinking Bank Regulation: Till Angels Govern(Cambridge, MA:Cambridge University Press).

    BechtM.P.BoltonA. Roëll2003“Corporate Governance and Control,”in Handbook of the Economics of Finance(pp. 1–109)ed. by G. M. Constantinides, M. Harris, and R. M. Stulz(Amsterdam: North-Holland).

    BeckT. A.Demirgüç-Kunt R. Levine2007“Bank Supervision and Corruption in Lending.”Journal of Monetary EconomicsVol. 53No. 8pp. 2131–63.

    BerglöfE.S.Claessens2006“Enforcement and Good Corporate Governance in Developing Countries and Transition Economies,”The World Bank Research ObserverVol. 21pp. 121–50.

    BhattacharyaU. H.Daouk and M. Welker2003“The World Price of Earnings Opacity,”Accounting ReviewVol. 78pp. 641–78.

    BondS.2002“Dynamic Panel Data Models: A Guide to Micro Data Methods and Practice,”Working Paper CWP09/02(London, UK:Institute of Fiscal Studies, University College of London).

    BushmanR. A.Smith2001“Financial Accounting Information and Corporate Governance,”Journal of Accounting and EconomicsVol. 32pp. 237–333.

    CIFAR1993.International Accounting and Auditing Trends3rd ed.(Princeton, NJ:Center for International Financial Analysis & Research, Inc.).

    ClaessensS.J.Fan2003“Corporate Governance in Asia: A Survey”(unpublished;Amsterdam: University of Amsterdam).

    ClaessensS.S. Djankov andL.Lang2000“The Separation of Ownership and Control in East Asian Corporations,”Journal of Financial EconomicsVol. 58pp. 81–112.

    ColesJ. M.LemmonF.Meschke2006“Structural Models and Endogeneity in Corporate Finance: The Link Between Managerial Ownership and Performance”(unpublished;Tucson: Arizona State University).

    CominD.S.Mulani2005“A Theory of Growth and Volatility at the Aggregate and Firm Level.”NBER Working Paper No. 11503(Cambridge, MA:National Bureau of Economic Research).

    CoreJ.W.Guay T.Rusticus2006“Does Weak Governance Cause Weak Stock Returns? An Examination of Firm Operating Performance and Investors’ Expectations,”Journal of FinanceVol. 61pp. 655–87.

    De NicoloG. L.LaevenK.Ueda2006“Corporate Governance Quality: Trends and Real Effects,”IMF Working Paper 06/293(Washington:International Monetary Fund).

    Demirgüç-KuntA. L.LaevenR.Levine2004“Regulations, Market Structure, Institutions, and the Cost of Financial Intermediation,”Journal of Money Credit and BankingVol. 36pp. 593–622.

    DoidgeC.A. Karolyi and R. Stulz2004“Why Are Foreign Firms Listed in the US Worth More?”Journal of Financial EconomicsVol. 72pp. 519.

    DoidgeC.A. Karolyi and R. Stulz2007“Why Do Countries Matter So Much for Corporate Governance?”Journal of Financial EconomicsVol. 86pp 1–39.

    DjankovS.C. McLiesh and A. Shleifer2007“Private Credit in 129 Countries,”Journal of Financial EconomicsVol. 84pp. 299–329.

    GompersP.J. Ishii and A. Metrick2003“Corporate Governance and Equity Prices,”Quarterly Journal of EconomicsVol. 118pp. 107–55.

    HimmelbergC.G. Hubbard and D. Palia1999“Understanding the Determinants of Managerial Ownership and the Link Between Ownership and Performance,”Journal of Financial EconomicsVol. 53pp. 353–84.

    JensenM.1986“Agency Costs of Free Cash Flow, Corporate Finance and Takeovers,”American Economic ReviewVol. 76pp. 323–29.

    JinL.and S. Myers2006“R 2 around the world: new theory and new tests,”Journal of Financial EconomicsVol. 79pp. 257–59.

    KlenowP.and A. Rodriguez-Clare2005“Externalities and Growth,” in Handbook of Economic Growth (pp. 817–61)ed. by P. Aghion and S. N. Durlauf (Amsterdam:Elsevier).

    KrosznerR.L. Laeven and D. Klingebiel2007“Banking Crises, Financial Dependence, and Growth,”Journal of Financial EconomicsVol. 84pp. 187–228.

    LaevenL. and P. Honohan2005Systemic Financial Crises: Containment and Resolution(Cambridge, MA:Cambridge University Press).

    La PortaR. F. Lopez-de-Silanes A. Shleifer and R. Vishny1998“Law and Finance,”Journal of Political EconomyVol. 106pp. 1113–55.

    La PortaR. F.Lopez-de-Silanes A.Shleifer2006“What Works in Securities Laws?”Journal of FinanceVol. 61pp. 1–32.

    LeuzC.D.NandaP.Wysocki2003“Earnings Management and Investor Protection: An International Comparison,”Journal of Financial EconomicsVol. 69pp. 505–27.

    MorckR.B.Yeung W.Yu2000“The Information Content of Stock Markets: Why Do Emerging Markets Have Synchronous Price Movements?”Journal of Financial EconomicsVol. 58pp. 215–60.

    Organisation for Economic Co-operation and Development2003White Paper on Corporate Governance in Asia(Paris:Organisation for Economic Co-operation and Development).

    PinkowitzL.R.Stulz and R.Williamson2006“Does the Contribution of Corporate Cash Holdings and Dividends to Firm Value Depend on Governance? A Cross-Country Analysis,”Journal of FinanceVol. 61pp. 2725–51.

    RajanR.L.Zingales1998“Financial Dependence and Growth,”American Economic ReviewVol. 88pp. 559–86.

    ShleiferA.R.Vishny1997“A Survey of Ccorporate Governance,”Journal of FinanceVol. 52pp. 737–83.

    SpamannH.2006“On the Insignificance and/or Endogeneity of La Porta et al.’s ‘Antidirector Rights Index’ Under Consistent Coding,”John M. Olin Center Discussion PaperNo. 7Harvard Law School.

    TiroleJ.2001“Corporate Governance,”EconometricaVol. 61pp. 1–35.

    ZingalesL.1998“Corporate Governance,”in The New Palgrave Dictionary of Economics and the Law(pp. 497–502)ed. by P. Newman (New York:Macmillan).

This chapter is a slightly revised version of an article that appeared in The Journal of Financial Intermediation 17 (2008): 198–228. Reprinted with permission from Elsevier.

We would like to thank Ross Levine, two anonymous referees, Franklin Allen, Thorsten Beck, Patrick Bolton, Stijn Claessens, Simon Johnson, Laura Kodres, Jonathan Ostry, Eswar Prasad, Raghuram Rajan, René Stulz, and seminar participants at the International Monetary Fund, the2006 IMF/World Bank annual meeting in Singapore, and the Sixth Annual Conference on Emerging Markets at the Darden School of Business at the University of Virgina for comments on earlierdrafts of this chapter. Excellent research assistance by Nese Erbil, Juanita Riano, Junko Sekine, and Wellian Wiranto is gratefully acknowledged.

We checked the robustness of the AS indicator by constructing variants in several ways. For example, eliminating the accounting items that are reported by 95 percent of all firms in 1995, we constructthe index using only those 16 items that are reported by less than 95 percent of all firms. This indexhas more variation, compared to the original index that is based on 40 accounting items, but thecorrelation with the original index is very high, more than 0.95. We also constructed an alternativeindex using the percentage of the 10 largest firms in each country (in terms of market capitalization)that reports all of the 24 items that are reported by 95 percent or more of all firms, but there is verylittle variation. We calculate these two variants using a threshold of 85 percent instead of 95 percent, but this does not alter our findings. Finally, we constructed an index based on the 100 largest firms(or fewer when 100 are not available) instead of the 10 largest firms, but sample selection biasappears severe, as the number of firms covered by Worldscope typically grows over time in emergingmarket economies. However, using this alternative measure of accounting standards based on thelargest 100 firms does not alter our main findings.

Morck, Yeung, and Yu (2000) report a second measure, given by the share of stocks whose pricesmove in the same direction (either up or down). Our results are invariant to the use of this measure.

Synchronicity may be observed if a country specializes in specific industries. In this case, industryspecificshocks would drive overall movements of stock prices, in contrast with the case of a highlydiversified country. In addition, if aggregate shocks are large (e.g., overall boom and bust, oil shocks, or currency crisis), then stock prices may move more in those countries which are most sensitive toaggregate shocks.

This selection criterion takes into account changes in stock price synchronicity because of changesin the number of firms that are listed in the stock exchange at each point in time. This is importantespecially in the case of countries that experienced a crisis. By construction, a balanced samplewould not reflect exits of bankrupt firms (possibly characterized by poor corporate governance) andentry of new firms (possibly characterized by good corporate governance).

Although there exists a large literature on the relationship between ownership and firm performance(see, for example, Shleifer and Vishny, 1997 ; Claessens, Djankov, and Lang, 2000 ; and Gompers, Ishii, and Metrick, 2003), we do not consider ownership structures of firms because data on ownershipstructures is not widely available and because ownership structures do not change much overtime, except when dramatic events such as mergers and acquisitions occur.

We do not consider ADR premiums because this measure does not exhibit a consistent patternacross countries. In theory, ADR premiums (i.e., the stock price premium of ADRs over domesticallylisted shares) for foreign firms cross-listed in the United States should be higher for firms withworse corporate governance (see, for example, Doidge, Karolyi, and Stulz, 2004). However, of allthe countries included in our study we find that U.K. and Canadian firms display the highest premiums, and that premiums of firms in these countries are significantly higher than for firms incountries that one would expect to have poor corporate governance.

The value of cash holdings measure proposed by Pinkowitz, Stulz, and Williamson (2005) is basedon the premise that one dollar in the corporate balance sheet is valued less than one dollar by thestock market in countries where corporate governance is weak. This measure appears highly volatilewhen estimated over time and therefore does not seem to capture well changes in corporate governanceover time.

The developed countries are: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Israel, Italy, Japan, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, United Kingdom, and United States.

Time series data on creditor rights are compiled by Djankov, McLeish, and Shleifer (2005). Timeseries data on minority shareholders’ rights are from Spamann (2006) following the work by LaPorta and others (1998).

We estimate TFP growth based on the standard method used by Klenow and Rodriguez-Clare(2005) without correcting for changes in educational attainments, as they vary little in the sampleperiod for our sample countries. The underlying data are from Penn World Table 6.1 and the IMF’s World Economic Outlook database.

When including only one lag of the dependent variable, we obtain similar results but the Arellanoand Bond (1991) test for second-order serial correlation rejects the null hypothesis of zero autocorrelationin several specifications of our basic model.

As stressed by Bond (2002), one advantage of this type of autoregressive-distributed lag model isthat it does not require modeling the series on the right-hand side of the equation to estimate therelevant coefficients.

When including time-fixed effects, all specification tests remain valid. Moreover, the sign of thecoefficients is unchanged, but there is a loss of precision in the estimates, as significance levels drop. This is not surprising, because these variables in part capture the synchronicity of changes in corporategovernance arrangements across several countries.

Standard control variables used in the growth literature, such as schooling and population growth, tend not to vary much over short periods of time, and are unlikely to be highly correlated with ourcorporate governance index. However, a large aggregate shock, for example one resulting in a currencycrisis, may be correlated with the CGQ index, and below we conduct robustness checks toaccount for this possibility.

One assumption underlying the Arellano and Bond (1991) model is the stationarity of the dependentand independent variables. Standard Dickey-Fuller unit root tests indeed confirm that thevariables of interest are stationary.

Note, however, that greater synchronicity may also be related to factors other than accounting transparency(as mentioned earlier), such as the degree of industry specialization or the degree of uncertaintyabout monetary policy in the country, and should thus be interpreted with this caveat in mind.

We should note that the test of second-order serial correlation indicates the absence of second-orderserial correlation. If the errors in the model in levels [equation (1) ] are serially uncorrelated, thenthe double-differenced model [equation (3) ] may exhibit second-order serial correlation. This notbeing the case, our results suggest that the errors in the model in levels are close to a random walk, which may generate only first-order serial correlation in the double-differenced model.

The data on stock price synchronicity and GDP growth for the East Asian crisis countries duringthe crisis years 1998–99 are consistent with this: the stock market crash in late 1998 coincided withhigh synchronicity in stock returns, and a sharp decline in GDP was recorded with one-year lag in1999, generating an exceptionally high comovement between stock price synchronicity and GDPper capita growth during the crisis period.

Note that in Table 14.4 and 14.5, there is evidence of second-order serial correlation for somespecifications, consistent with the assumptions of lack of serial correlation of the disturbances in themodel in levels (see footnote 20).

We obtain similar results when using private credit to GDP as a measure of financial development.

We have also analyzed whether the effect is different for the subset of East Asian countries in oursample. The reason for focusing on the East Asian countries is that growth and changes in corporategovernance quality may have followed different paths in these countries in response to governanceproblems arising from the East Asian financial crisis in 1997 to 1998. However, we do not find asignificant difference in the results for East Asian countries compared to the rest of the world.

    Other Resources Citing This Publication