Firm-specific corporate governance and analysts’ earnings forecast characteristics: Evidence from Asian stock markets

Minna Yu (Department of Accounting, Monmouth University, West Long Branch, New Jersey, USA)
Yanming Wang (School of Accounting, Shanghai University of Finance and Economics, Shanghai, China)

International Journal of Accounting & Information Management

ISSN: 1834-7649

Publication date: 6 August 2018

Abstract

Purpose

The purpose of this paper is to examine the impact of corporate governance on the capital market participants’ abilities to forecast future performance, as measured by the properties of analysts’ earnings forecasts in Asian stock markets.

Design/methodology/approach

This paper hypothesizes that higher corporate governance is associated with lower forecast errors, lower forecast dispersion and lower forecast revision volatility.

Findings

These predictions are supported with a sample of companies across eleven Asian economies over 2004-2012. The results of this paper suggest that corporate governance plays a significant role in the predictability of firm’s future performance and, therefore, improves the financial environment in Asian stock markets. Furthermore, the impact of corporate governance on analysts’ forecast properties is more pronounced in countries with strong investor protection.

Research/limitations/implications

The authors acknowledge the following limitations of this paper. First, the results of this paper may be subject to omitted-variable bias and endogeneity issue. The authors have used control variables in the regressions to reduce the omitted variable bias. The authors have run lead-lag regressions to address causality issue. Second, CLSA corporate governance scores are collected for largest companies in each jurisdiction. Therefore, the sample is biased towards the largest companies in those jurisdictions and may not be representative of the average firm in the Asia.

Originality/value

The results of this paper speak to the benefit of having strong corporate governance in terms of reducing the information asymmetry between investors and corporate management.

Keywords

Citation

Yu, M. and Wang, Y. (2018), "Firm-specific corporate governance and analysts’ earnings forecast characteristics", International Journal of Accounting & Information Management, Vol. 26 No. 3, pp. 335-361. https://doi.org/10.1108/IJAIM-03-2017-0040

Download as .RIS

Publisher

:

Emerald Publishing Limited

Copyright © 2018, Emerald Publishing Limited


1. Introduction

In this paper, we investigate whether effective corporate governance enhances the predictability of firms’ future performance, measured by financial analysts’ earnings forecast characteristics. As noted in La Porta et al. (2000), firm-specific corporate governance is a set of mechanisms through which outside investors protect themselves against expropriation by the management. Weak corporate governance has been considered a major contributor of Asian financial crisis during 1997-1998. In fact, the weak corporate governance systems explain the extent of stock market decline better than macroeconomic factors (Johnson et al., 2000). After the financial crisis, firms in the Asian capital markets have actively implemented good corporate governance structures and processes. Prior research documents that, during the post-crisis period, Asian firms with good corporate governance tend to have higher market value and lower cost of capital (Klapper and Love, 2004; Chen et al., 2009). Are these positive capital market consequences of good corporate governance because of the reduced information asymmetry between investors and management? Until now, we lack evidence on the impact of good corporate governance on such information asymmetry.

Byard et al. (2006) document the positive association between analysts’ earnings forecast accuracy and attributes of corporate governance in the USA. However, in a US setting and most regions in the globe, corporate governance systems remain unchanged for long periods of time, which has been referred to as “stickiness” (Brown et al., 2011). The stickiness of corporate governance measure has posed significant challenges for research. The post-financial-crisis period in Asia is a desirable setting for examining the impact of corporate governance. Studying the association between analysts’ forecast characteristics and corporate governance practices enhances our understanding as to whether corporate governance reduces information asymmetry between investors and management in the Asian setting.

We use analysts’ earnings forecasts as proxies[1] for investors expectation for future earnings because investor’ prediction about firm’s performance is not observable. We use three measures of analyst forecast characteristics in earnings forecasts that have been commonly studied in prior research (Lang and Lundholm, 1996; Ali et al., 2007): forecast accuracy, forecast dispersion and volatility in forecasts.

We obtain the corporate governance rating scores developed by Credit Lyonnais Securities Asia (CLSA) as a measure of the firm-level corporate governance quality. CLSA corporate governance scores have been widely used as a proxy for the quality of firms’ internal corporate governance mechanisms by prior studies (Klapper and Love, 2004; Shen and Chih, 2007; Chen et al., 2009; Yu, 2010). Extant research on the impact of corporate governance on analyst behavior has concentrated on a single corporate governance dimension, such as ownership structure (Lang et al., 2004; Ali et al., 2007). The composite score of corporate governance is superior to a single dimension because the composite score measures the overall strength of all corporate governance mechanisms (Brown et al., 2011) and using a parsimonious index is more effective than including all individual corporate governance characteristics (Brown and Caylor, 2006; Bebchuk et al., 2009).

Investors and regulators increasingly view effective corporate governance as crucial to mitigating the agency problem as high levels of corporate governance reduce conflicts by better aligning managers’ interest with that of investors. In contrast, in firms that lack adequate monitoring systems, managers are more likely to obscure or manipulate disclosures and make inefficient operating, investing and financing decisions, which make future performance less predictable. Therefore, good corporate governance mechanisms facilitate analysts’ judgment of the quality of financial information and the future performance. Conversely, firms with missing or ineffective corporate governance mechanisms making forecasting task more complex. This leads to our prediction that analyst earnings’ forecasts should be more accurate for firms with good corporate governance mechanisms than for firms without good corporate governance mechanisms. Applying the above reasoning, we also predict that stronger corporate governance is associated with lower analyst forecast dispersion and less volatility in analysts’ forecast revisions.

We find that, controlling for other factors, analysts’ earnings forecast error, forecast dispersion and volatility in forecast revisions are lower for firms that have implemented better corporate governance systems. This finding is consistent with that, for firms with relatively strong corporate governance mechanisms, investors tend to have more accurate beliefs about future performance; there is less asymmetry in individual investors’ beliefs about forecasted firms’ performance; and investor expectations about earnings change more smoothly over the year.

In general, Asia’s emerging markets feature a relatively weak legal system where shareholder rights are not well protected and information asymmetry is relatively high (Claessens and Fan, 2002; Fan and Wong, 2002; Haw et al., 2011; Ariff et al., 2014). However, Asian jurisdictions have institutional differences in the country-level corporate governance. We further examine how cross-country corporate governance differences explain the strength of the associations between firm-level corporate governance and analyst behavior. We predict and find that our results are primarily driven by jurisdictions with stronger country-level corporate governance.

Our paper contributes to the literature in the following ways. First, our paper adds to the research stream of how firm characteristics affect the predictability of future earnings. It specifically joins the increasing stream of research on the effect of firm-level corporate governance on analyst behavior (Lang et al., 2003; Lang et al., 2004; Byard et al., 2006; Ali et al., 2007; Gul et al., 2013).

Second, we hypothesize and provide evidence on the effect of firm-level corporate governance quality on the analysts’ forecast properties. Hope (2003b) documents the positive association between country-level corporate governance and analyst forecast accuracy. Instead, we examine the association using firm-level corporate governance and find the association is significant. We further examine the interaction of country-level corporate governance and firm-level corporate governance on analysts’ earnings forecast properties. Our findings suggest that firm-level corporate governance affects information asymmetry in common law jurisdictions, which feature strong investor protection.

Our paper has practical implications for corporate managers as well as policy-makers and capital market regulators. Specifically, the evidence provided in this paper suggests that, by adopting better corporate governance mechanisms, firms in Asian stock markets can improve the accuracy of market expectations, reduce information asymmetries and limit market surprises, especially in jurisdictions with stronger investor protection. The findings shed light on the advantage of improving corporate governance on firms’ information environment and, therefore, are relevant to the cost-benefit analysis of improving internal corporate governance.

The paper proceeds as follows. Section 2 reviews relevant literature and develops testable hypotheses. Section 3 describes the research models to test the hypotheses. Section 4 introduces our proxy for corporate governance quality as well as provides the sample distribution and univariate statistics. Section 5 presents empirical results and Section 6 runs additional robustness checks. We conclude in Section 7.

2. Literature review and hypotheses development

We establish the association between the strength of corporate governance on the predictability of earnings based on the research findings of the association between corporate governance and the quality of information provided by the management. Effective corporate governance mechanisms reduce agency problems, which in turn will enhance information quality produced by analysts and reduce information risk faced by investors. Good corporate governance gives rise to high-quality management-investor communication because the interest of managers is better aligned with that of investors. As such, managers do not have incentives to hide information or abuse accounting discretion. Good corporate governance constrains real-activities earnings management (Malik, 2011) and accrual-based earnings manipulation (Klein, 2002; Davidson et al., 2005).

In Singapore, Eng and Mak (2003) find that lower managerial ownership and significant government ownership are associated with more non-mandatory strategic, non-financial and financial disclosures that is not mandatory. Chau and Gray (2010) as well as Ho and Wong (2001) find similar evidence in Hong Kong. Therefore, we conjecture that good governance mechanisms will lead to reduction in analysts’ earnings forecast error. Prior research has also well documented that higher-quality disclosures, more disclosures on accounting policies, more readable annual reports, as well as voluntary management forecasts will enhance the forecast accuracy (Waymire, 1986; Lang and Lundholm, 1996; Hope, 2003a; Frankel et al., 2006; Lehavy et al., 2011).

We conjecture that good corporate governance mechanisms facilitate analysts’ judgment of the quality of financial information and the future performance. Conversely, firms with missing or ineffective corporate governance mechanisms making forecasting task more complex. A number of papers report that nonfinancial indicators of investments in intangible assets are important predictors of revenues (Trueman et al., 2001), operating income (Behn and Riley, 1999) and firm value (Amir and Lev, 1996). Prior research has documented the difficulty of forecasting for high-technology firms as such firms have significant research and development projects (Barron et al., 2002). Good corporate governance will ensure that managers make effective decisions, which reduce the uncertainty and complexity of operations as well as investing and financing activities. Bhat et al. (2006) document the importance of corporate governance-related disclosures for analyst forecast accuracy, even after controlling for financial transparency. It appears analysts place some importance on governance characteristics, as this enables them to assess the credibility of the firm’s disclosures. Hope (2003b) finds stronger country-level enforcement gives managers higher incentives to follow accounting rules, which reduces analysts’ uncertainty about future earnings.

We argue that firms with better corporate governance are subject to less uncertainties and risks, which reduce the difficulty of forecasting and therefore reduce analysts’ forecast error. Lack of adequate governance mechanisms gives rise to increased uncertainty, which in turn complicates the forecasting task of analysts. Duru and Reeb (2002) argue that earnings forecast error depends on the difficulty or complexity of the forecasting task and show that analyst earnings forecasts are less accurate for multinational enterprises than domestic firms. Chen et al. (2010) find that analysts’ earnings forecasts are less accurate for firms with high-level political connections because analysts experience more difficulty in predicting earnings.

Research in the USA has documented the relation between certain aspects of corporate governance (such as internal control, ownership structure and board composition) and analyst forecast characteristics. Xu and Tang (2012) document more accurate analyst earnings’ forecasts with firms which report internal control material weaknesses. Ali et al. (2007) document that analysts are able to make more informative forecasts for family firms and also have lower level of forecast dispersion and lower volatility in forecast revisions. Gul et al. (2013) find that the presence of women on the board increases governance and therefore lead to higher analyst forecast accuracy and lower dispersion. Given that a variety of aspects of corporate governance affects analysts’ forecasting abilities, we predict that the overall corporate governance characteristics have an impact on the ability of financial analysts to forecast earnings accurately. Therefore, our first hypothesis is:

H1.

Analysts’ one-year-ahead earnings forecast accuracy is positively related to forecasted firms’ quality of corporate governance mechanisms.

As good corporate governance leads to less agency problem from the separation of ownership and management and enhances the communication between managers and investors though public disclosures, we expect lower forecast dispersion for firms with better corporate governance. Therefore, our second hypothesis is:

H2.

Analysts’ earnings forecast dispersion is negatively associated with forecasted firms’ quality of corporate governance mechanisms.

Following the same logic, corporate governance also has an impact on the smoothness of earnings prediction during the forecasting period. Analysts are more likely to update their expectations with firms facing governance problems. Conversely, investor expectations about earnings change more smoothly over the year for firms with relatively good corporate governance mechanisms. Prior research, such as Aboody et al. (2006) and Bartov et al. (2007), provide evidence managers manipulate inputs for fair values for their own interests. However, in some situations, managers may use their private information to credibly report fair values (Barth et al., 1998). Good corporate governance reduces information asymmetry through improved timeliness of the disclosures. For example, Song et al. (2010) document that for firms with strong corporate governance, the fair values have higher value relevance. The improved timeliness of disclosure, in turn, leads to smoother adjustments in analyst forecasts. Therefore, our third hypothesis is as follows:

H3.

The volatility of analysts’ earnings forecast revisions is negatively related to the quality of corporate governance.

We next examine whether the associations between analysts’ earnings forecast characteristics and corporate governance vary across institutional environments. Hope (2003b) documents the positive effect of stronger legal enforcement on analyst forecast accuracy. Furthermore, in jurisdictions with higher level of investor protection, analysts have more incentives to improve their forecasts (Barniv et al., 2005) and analyst forecasts outperform the historical earnings forecasting model (Barniv and Myring, 2006). Therefore, we expect that, in jurisdictions with higher level of investor protection (rather than in jurisdictions with lower level of investor protection), it is easier for analysts to assess firms’ future prospects based on the strength of firm-level corporate governance.

H4.

The associations between firm-specific corporate governance strength and analysts’ earnings forecast accuracy, forecast dispersion and revision volatility are more pronounced in jurisdictions with higher level of investor protection.

3. Research models

We use the corporate governance ratings developed by CLSA as our measure of corporate governance quality (CGit). Analyst forecast error (FERRORit) is the absolute analysts’ forecast error, calculated as actual earnings minus the initial mean analyst forecasted earnings following the annual reports, deflated by beginning-of-fiscal-year price. Following prior research, we control for size (SIZEit), measured as the log of beginning-of-fiscal-year market value of equity in US $millions. We also control for the standard deviation of return on equity, computed over the preceding ten years (SD_ROEit). In addition, we add earnings surprise (EARNINGS_SURit) as a control variable, calculated as the absolute value of the difference between the current year’s earnings per share and last year’s earnings per share, deflated by beginning-of-fiscal-year price. We also use LOSSit as a control variable because analyst forecast accuracy is lower for firms that report negative earnings relative to firms that report profits (Hwang et al., 1996). LOSSit is set to 1 if firm i reports negative earnings in year t and 0 otherwise.

To test H1 regarding the association between analysts’ forecast errors and the quality of corporate governance, we use the following model:

(1) FERRORit=a0+a1CGit+a2SIZEit+a3SD_ROEit+a4LOSSit+a5EARNINGS_SURit+εit

Following prior research, we measure analysts’ earnings forecast dispersion (DISPERSIONit) as the standard deviation of the initial individual analysts’ earnings forecasts following the annual reports, deflated by beginning-of-fiscal-year price. We use the same set of control variables as in Model (1). To test for the association between analyst forecast dispersion and the quality of corporate governance (i.e. H2), we use the model as follows:

(2) DISPERSIONit=b0+b1CGit+b2SIZEit+b3SD_ROEit+b4LOSSit+b5EARNINGS_SURit+εit

Our measure of volatility in forecast revisions is the standard deviation of monthly forecast revisions over the fiscal year, deflated by beginning-of-fiscal-year price, where forecast revision is defined as current-month median forecast minus previous-month median forecast. Our model to test H3 is:

(3) VOL_REVit=c0+c1CGit+c2SIZEit+c3SD_ROEit+c4LOSSit+c5EARNINGS_SURit+εit

To test H4, we use three alternative proxies for the institutional environment: legal system (LEGALc), anti-director rights index (ANTIDIRECTOR_RIGHTSc) developed in La Porta et al. (1998) and the revised anti-director rights index (R_ANTIDIRECTOR_RIGHTSc) developed in Djankov et al. (2008). Investor protection is the fundamental factor in describing differences in corporate governance regimes across jurisdictions than some of the more customary classifications such as bank- or market-centeredness (La Porta et al., 2000). According to La Porta et al. (1998), the common versus code legal system is highly correlated with anti-director rights index, which measures how strongly the legal system favors minority shareholders against managers or dominant shareholders in the corporate decision-making process, including the voting process[2].

Djankov et al. (2008) provide a revised anti-director rights index. Spamann (2010) makes corrections for 33 of the 46 countries analyzed in La Porta et al. (1998). The correlation between corrected and original values is only 0.53. Consequently, many empirical results established using the original index may not be replicable with corrected values. Therefore, in this paper, we also use the revised anti-director rights index developed in Djankov et al. (2008) as a measure of institutional environment.

We interact CGit with ENVIRONMENTc in the above three models.

(4) FERRORit= d0+d1CGit+d2CGGit*ENVIRONMENTc+d3ENVIRONMENTc+d4SIZEit+d5SD_ROEit+d6LOSSSit+d7EARNINGS_SURit+εit
(5) DISPERSIONit=e0+e1CGit+e2CGit*ENVIRONMENTc+e3ENVIRONMENTc+e4SIZEit+e5SD_ROEit+e6LOSSit+e7EARNINGS_SURit+εit
(6) VOL_REVit=f0+f1CGit+f2CGit*ENVIRONMENTc+f3ENVIRONMENTc+f4SIZEit+f5SD_ROEit+f5LOSSit+f6EARNINGS_SURit+εit

We expect CGit*ENVIRONMENTc to be negative in Models (4), (5) and (6) to support H4.

4. Data and sample

We construct our sample over 2004-2012. Our sample starts from 2004 because CLSA has provided governance ratings for Asian-Pacific area starting from 2004[3]. As reported in Table I, our initial sample is composed of 6,658 firm-year observations from 13 Asian-Pacific economies. As we examine our research issue in Asia, we delete 144 non-Asian firm-year observations, which consist of 10 firm-year observations (two distinct firms) from USA and 134 firm-year observations (40 distinct firms) from Australia. To obtain data for our models, we merge our CLSA data set with I/B/E/S International. We remove firms without analyst coverage and firm-year observations that have missing data in I/B/E/S International database. Our final sample includes 3,393 observations representing 786 distinct firms (Tables I to IV).

Table II provides sample distribution across jurisdictions. Our sample includes eleven Asian economies: China, Hong Kong, Indonesia, India, Japan, Korea, Malaysia, the Philippines, Singapore, Thailand and Taiwan. The jurisdictions most highly represented in the sample are India (16.1 per cent) and China (15.6 per cent), followed by Japan (10.9 per cent), Taiwan (10.4 per cent) and Hong Kong (9.9 per cent). We also provide the numbers of distinct firms from each jurisdiction and the percentages. Our sample covers 88 firms from China (13.56 per cent), 86 firms from Japan (13.25 per cent), 82 firms from India (12.63 per cent), 76 firms from Taiwan (11.71 per cent) and 67 firms from Hong Kong (10.32 per cent). Firms from the other six jurisdictions represent less than 10 per cent of the sample. The firm-specific distribution by jurisdiction is mostly consistent with the firm-year distribution by jurisdiction except for Japan. Although the number of Japanese firms represents 13.25 per cent of the total number of distinct firms in the sample, total firm-year observations only account for 10.9 per cent of the sample.

In Table III, we report our sample distribution by both jurisdiction and year to further describe our sample. In general, our sample is not concentrated in specific years. All jurisdictions have firm representations in each sample year except for Japan, which does not have firm representations until 2008. This evidence explains the findings of Japan in Table II. In addition, the number of sample firms increases by year. This is mainly because CLSA has provided corporate governance ratings for an increasing number of firms year by year.

In Table IV, we present sample distribution by jurisdiction and industry. In CLSA data, companies are classified into 18 industries: automotive, capital goods, conglomerates, consumer, financial service, healthcare, hotels and leisure, infrastructure, insurance, internet, materials, media, petroleum/chemical, power, property, technology, telecommunications and transportation. Consumer, financial service, materials and technology industries each represent more than 10 per cent of the sample. Petroleum/chemical, power, property and telecommunications each represent more than 5 per cent. All other industries represent less than 5 per cent of the sample. It seems that our sample is not highly concentrated in a certain group of industries.

Univariate statistics are presented in Tables V to VIIII. Table V provides simple statistics for the whole sample. The means (medians) of analyst forecast error (FERRORit), forecast dispersion (DISPERSIONit) and volatility in forecast revisions (VOL_REVit) are 0.015 (0.005), 0.009 (0.005) and 0.008 (0.004), respectively. The standard deviations of these three variables are 0.057, 0.012 and 0.017, which suggest that analysts’ earnings forecast characteristics do not have wide variations in our sample. The corporate governance score ranges from 0.009 to 0.971, with a mean of 0.543 and a median of 0.551. The standard deviation of corporate governance scores is 0.146. Among the control variables, SIZEit exhibits significant variation as indicated by the standard deviation of 2.601 and range of 5.422 to 19.56. SD_ROEit has a standard deviation of is 0.245 and ranges from 0 to 675.49. More than two-thirds of the firm-year observations report loss during the sample period.

In Table VI, we provide means, medians and standard deviations of our three dependent variables (i.e. FERRORit, DISPERSIONit, VOL_REVit) and the main variable of interest (CGit) for each jurisdiction. As seen from the means and medians, Korea has the highest analysts’ earnings forecast errors, forecast dispersion and revision volatility among the sample Asian jurisdictions. Analysts’ earnings forecast errors are relatively low in Singapore and Taiwan. The Philippines has the largest standard deviation of FERRORit. Companies from Thailand have the highest corporate governance score, followed by Hong Kong, Singapore and Malaysia. Companies in China and Indonesia have relatively lower corporate governance scores, as indicated by both the means and medians.

In Table VII, we compare the means and medians of all variables for common law jurisdictions and code law jurisdictions. We also test for differences and provide statistics from t-tests and Wilcoxon rank-sum tests. First, the means and medians of the three analyst forecast characteristic variables are lower significantly in common law jurisdictions. This is consistent with prior literature that, in countries with stronger investor protection, analyst forecast accuracy is higher (Hope, 2003a and 2003b) and analysts have more incentives to improve their forecasts (Barniv et al., 2005). Second, firms in common law jurisdictions have better corporate governance than those in code law jurisdictions as both the mean and median of CGit are significantly higher in common law jurisdictions than in code law jurisdictions.

We provide the means of corporate governance score by jurisdiction and year in Table VIII. For every jurisdiction as well as for the whole sample, the corporate governance score changes year by year during 2004-2012. This contrasts with the USA where corporate governance systems are sticky (Brown et al., 2011). For example, in the whole sample, the average corporate governance score ranges from 0.528 (the lowest) to 0.607 (the highest) during the nine-year period.

Table IX presents pairwise correlation matrix with Pearson correlation above the diagonal and Spearman correlation below. The three measures of analysts’ earnings forecasts characteristics are significantly correlated in both Pearson correlation and Spearman correlations. Specifically, the pairwise Pearson (Spearman) correlation between FERRORit and DISPERSIONit is 0.378 (0.515); the Pearson (Spearman) correlation between FERRORit and VOL_REVit is 0.382 (0.455); the Pearson (Spearman) correlation between DISPERSIONit and VOL_REVit is 0.538 (0.462). Although significantly correlated, the moderate correlation coefficients indicate that the three measures used in this paper represent different characteristics of analysts’ earnings forecasts.

As suggested by both Pearson and Spearman correlations, the corporate governance score has a negative and significant correlation with all three measures of analysts’ earnings forecast characteristics. The Pearson (Spearman) correlation coefficient between CGit and FERRORit is −0.031 (−0.114). The Pearson (Spearman) correlation coefficient between CGit and DISPERSIONit is −0.095 (−0.142). The Pearson (Spearman) correlation coefficient between CGit and VOL_REVit is −0.057 (−0.084). These pair-wise correlation results provide preliminary support for our predictions regarding analyst forecast characteristics and the forecast firm’s corporate governance quality.

Other noteworthy findings are as follows. First, LEGALc is significantly and positively correlated with CGit in both Pearson and Spearman correlations (Pearson correlation coefficient = 0.132 with p-value of 0.001 and Spearman correlation coefficient = 0.106 with p-value of 0.001). ANTIDIRECTOR_RIGHTSc has a positive correlation with CGit in Pearson correlation (Pearson correlation coefficient = 0.056, p-value = 0.001) yet an insignificant correlation with CGit in Spearman correlation (Spearman correlation coefficient = 0.025, p-value = 0.155). The correlation between R_ANTIDIRECTOR_RIGHTSc and CGit is insignificant in both the Pearson and Spearman correlations. Second, our finding in Table IX is consistent with Tables XIII and XIV of Hung (2001), which presents the Pearson (Spearman) correlation coefficient for legal system and anti-director rights index to be 0.78 (0.83), both significant at 0.01 level. Third, in Pearson and Spearman correlations, LEGALc and ANTIDIRECTOR_RIGHTSc are both negatively associated with FERRORit, DISPERSIONit and VOL_REVit, which is consistent with prior literature that documents analyst forecast quality is higher in countries with better investor protection (Hope, 2003a, 2003b; Barniv et al., 2005). However, the correlations between R_ANTIDIRECTOR_RIGHTSc and analyst forecast characteristics are mixed. In Pearson correlation, R_ANTIDIRECTOR_RIGHTSc and analyst forecast characteristics are not significantly correlated. In Spearman correlation, R_ANTIDIRECTOR_RIGHTSc is negatively correlated with VOL_REVit, but not significantly correlated with FERRORit or DISPERSIONit. Fourth, the correlation coefficients among explanatory variables suggest that multicollinearity is not a concern in the regression analyses.

5. Empirical results

The pooled regression results are provided in Table X. We first estimate the three research models with ordinary least squares approach and report results on the left panel of the table. We control for country and year fixed effects by adding country and year dummy variables. The t-statistics are adjusted using the White procedure to alleviate the concern of heteroscedasticity. Industry features may play a role in the impact of corporate governance (Januszewski et al., 2002; Byard et al., 2006; and Giroud and Mueller, 2011). Therefore, the association between corporate governance and analyst forecast characteristics may vary by industry. We run our results by controlling for industry fixed effects.

The regression results in Table X are obtained from the whole sample: 3,393 firm-year observations. As shown in Table X, the estimated coefficient on CGit in model (1) is −0.008 with White-adjusted t-statistic of −3.01, significant at the 0.01 level. This evidence supports H1 that analysts’ one-year-ahead earnings forecast errors are lower for firms with higher-quality of corporate governance. In support of H2 that analysts’ earnings forecast dispersion is smaller for firms with higher-quality of corporate governance, the estimated coefficient on CGit in Model (2) is negative and significant (coefficient estimate = −0.007 and White-adjusted t-statistic = −5.75). This evidence supports H2. The H3 predicts that the volatility of analysts’ earnings forecast revisions is negatively related to the quality of corporate governance. As shown in the table, the estimated coefficient on CGit in Model (3) is −0.006 and White-adjusted t-statistic is −3.09. This evidence supports H3.

We next examine whether cross-country differences explain the strength of the association between corporate governance quality and analyst behavior. To test H4, we classify the 11 jurisdictions based on the strength of investor protection. In La Porta et al. (1998), China is not classified into common or code law. Anti-director rights index and revised anti-director rights index is not provided for China, either. As such, China is not in our sample for testing H4, thereby reducing our sample to 2,865 firm-year observations. In Table XI, we present the comparison of original anti-director rights scores by La Porta et al. (1998) and the revised scores by Djankov et al. (2008) for the jurisdictions we study except China. As it shows, the index for seven jurisdictions have changed (Tables XI and XII).

We report ordinary least squares estimation results using the three alterative measures of institutional environment. When legal origin (LEGALc) is the measure of institutional environment, the coefficient estimates on CGit in Model (4) is −0.014 and is marginally significant with a t-statistic of −1.38. The coefficient estimate on CGit*ENVIRONMENTc is −0.011 and is significant at the 0.01 level with a t-statistic of −2.82. In Model (5), the coefficient estimate on CGit*ENVIRONMENTc is −0.008 with a t-statistic of −3.17. In Model (6), the coefficient estimate on CGit*ENVIRONMENTc is −0.004 with a t-statistic of −2.27. As such, in support of H4, the association between analyst behavior and the strength of corporate governance is more pronounced where the institutional environment features strong investor protection. ENVIRONMENTc is positive in Model (6) when the volatility in analyst forecast revisions (VOL_REVit) is the dependent variable, but not significant when FERRORit or DISPERSIONit is the dependent variable.

When we replace legal origin (LEGALc) with anti-director rights index (ANTIDIRECTOR_RIGHTSc) and revised anti-director rights index (R_ANTIDIRECTOR_RIGHTSc), our results are similar. That is, we find CGit*ENVIRONMENTc to be negatively associated with FERRORit, DISPERSIONit and VOL_REVit in all three models. Specifically, when anti-director rights index (ANTIDIRECTOR_RIGHTSc) is used as the measure of institutional environment (ENVIRONMENTc), the coefficient on CGit*ENVIRONMENTc in Model (4) is −0.006 (with t-statistic of −2.84), the coefficient on CGit*ENVIRONMENTc in Model (5) is −0.006 (with t-statistic of −4.44), and the coefficient on CGit*ENVIRONMENTc in Model (6) is −0.005 (with t-statistic of −2.36), all significant at the 0.01 level. When revised anti-director rights index (R_ANTIDIRECTOR_RIGHTSc) is used as the measure of institutional environment (ENVIRONMENTc), the coefficient on CGit*ENVIRONMENTc in Model (4) is −0.003 (with t-statistic of −5.25), significant at the 0.01 level. The coefficient on CGit*ENVIRONMENTc in Model (5) is −0.005 (with t-statistic of −2.26), and the coefficient on CGit*ENVIRONMENTc in Model (6) is −0.007 (with t-statistic of −1.94), both significant at the 0.05 level. Taken together, these results strengthen our inference that corporate governance has a stronger impact on analyst behavior in jurisdictions with stronger investor protection.

6. Additional robustness tests

6.1 Addressing endogeneity issue with Granger (1969) causality test

Yu (2010) documents that analyst following is positively associated with the quality of corporate governance. As analyst forecast accuracy may be correlated with analyst following, the higher forecast accuracy for better-governed firms may indicate that firms with higher analyst following and higher quality forecasts tend to adopt effective corporate governance mechanisms. To rule out this possibility, we use Granger (1969) causality test. We rerun all the models with a lagged value of corporate governance (i.e. CGit-1) (Tables XIII and XIV).

We first check the correlations between CGit and CGit−1. Pearson and Spearman correlations are presented in Table XIII. For the whole sample, the Pearson correlation between CGit and CGit−1 is 0.320 and the Spearman correlation between CGit and CGit−1 is 0.265, both of which are significant at the 1 per cent level. We also present the correlations by country. For each country, the correlation coefficients are lower than 0.5. That is, the lead and lag corporate governance scores are not highly correlated, which suggests that we are able to use Granger (1969) causality test (i.e. lead-lag regressions) to address the causality concern.

In Table XIV, the coefficient on CGit−1 in Model (1) is −0.006 (with t-statistic of −2.87), and the coefficient on CGit−1 in Model (2) is −0.005 (with t-statistic of −3.64), significant at 0.01 level. The coefficient on CGit−1 in Model (3) is −0.003 (with t-statistic of −1.56), significant at 0.10 level. Overall, the results using lagged value of corporate governance are qualitatively the same as our main results. The negative correlations between FERRORit, DISPERSIONit and VOL_REVit and lagged corporate governance scores help rule out the possibility that firms with higher quality analyst forecasts choose to implement good corporate governance mechanisms.

6.2 Deleting jurisdictions with significant number of observations

Among the 11 jurisdictions, China, Hong Kong, India, Japan and Taiwan each accounts for close to or above 10 per cent of the sample. To preclude the possibility that our results are driven by a single or a subgroup of the Asian sample, we delete China, Hong Kong, India, Japan and Taiwan one by one from the sample and rerun the results. Our results still hold.

6.3 Deleting cross-listed firm-year observations

Our results may be driven by firms publicly traded on US stock exchanges. Because of more stringent listing requirements, firms cross-listed on the US stock exchanges tend to have better corporate governance mechanisms and therefore tend to have higher analyst forecast accuracy (Lang et al., 2003). To rule out the possibility that our results are driven by cross-listed firms, we delete the cross-listed firm-year observations and rerun our regressions. Our results are robust to using the non-cross-listed subsample.

6.4 Alternative measures of analysts’ earnings forecast accuracy and forecast dispersion

Our empirical results regarding analyst forecast accuracy and forecast dispersion are generated using the earliest earnings forecasts generated by financial analysts before earnings releases. To run a robustness check, we use alternative measures of analysts’ earnings forecast error and dispersion. Following Ali et al. (2007), we use a 12-month average of earnings forecasts when calculating forecast error and dispersion. We also use the most recent forecasts before earnings announcement as used in Behn et al. (2008). Our results remain qualitatively unchanged.

6.5 Alternative treatment of correlation among country and year

In our main analyses, we have added country and year dummy variables to control for country and year fixed effects. Alternatively, we account for the possibility that the residuals could be correlated across country and time. Following prior research (Petersen, 2005; Gow et al., 2010), we estimate the coefficients and standard errors by adjusting for correlation by jurisdiction and year clusters. The Rogers standard errors results generated from clustering remain qualitatively the same as those controlling for fixed effects. We also use Fama and MacBeth (1973) regression to correct for cross-sectional correlation. The results remain qualitatively the same as our main results.

6.6 Using robust regressions instead of ordinary least squares

To control the influence of outliers, we estimate all our research models using robust regression method, particularly MM regression method as it performs the best among all methods (Leone et al., 2013). The robust regression results are qualitatively similar to the ordinary least squared results. The consistency between the two alternative estimation approaches speaks to the robustness of our empirical results.

7. Conclusions

In this paper, we examine the effect of forecasted company’s corporate governance quality on analyst forecasting behavior. We find that the accuracy of analysts’ earnings forecast improves and analysts’ earnings forecast revisions are less volatile for firms with higher corporate governance ratings. In addition, using forecast dispersion as a proxy for investors’ diverse beliefs in predicted earnings, we find that analysts’ earnings forecast dispersion is narrower for firms with higher corporate governance ratings. We also find that the relation between corporate governance and analyst forecasting behavior is more pronounced in jurisdictions with strong investor protection. This paper contributes to both the corporate governance literature and the analyst forecast literature. By documenting the benefit of having good corporate governance mechanisms in the Asia setting, this paper has important implications for capital market regulators and firm management.

We acknowledge the following limitations of our paper. First, our results may be subject to omitted-variable bias and endogeneity issue. We have used control variables in our regressions to reduce the omitted variable bias. We have run lead-lag regressions to address causality issue. Second, CLSA corporate governance scores are collected for largest companies in each jurisdiction. Therefore, our sample is biased towards the largest companies in those jurisdictions and may not be representative of all firms in Asia.

Sample description: initial sample

No. of observations
CLSA 2004-2012 6,658
less: deleting US and Australian observations 144
Less: lost observations when merged with I/B/E/S International or missing data for variable calculations 3,121
3,393

Sample description: sample distribution by jurisdiction

Jurisdiction No. of observations (%) No. of firms (%)
China 528 15.6 88 13.56
Hong Kong 335 9.9 67 10.32
Indonesia 208 6.1 38 5.86
India 543 16.1 82 12.63
Japan 369 10.9 86 13.25
Korea 255 7.5 57 8.78
Malaysia 191 5.6 33 5.08
The Philippines 152 4.5 40 6.16
Singapore 272 8.0 54 8.32
Thailand 188 5.5 28 4.31
Taiwan 352 10.4 76 11.71
Total 3,393 100 786 100

Sample description: sample distribution by jurisdiction and year

Total 2004 2005 2006 2007 2008 2009 2010 2011 2012
China 528 24 34 40 48 56 62 77 87 100
Hong Kong 335 28 30 30 31 34 34 41 46 61
Indonesia 208 16 17 20 21 21 21 23 26 43
India 543 34 39 42 45 53 71 79 80 100
Japan 369 0 0 0 0 24 25 94 99 127
Korea 255 15 16 19 24 30 29 22 36 64
Malaysia 191 17 15 16 18 21 21 27 27 29
The Philippines 152 11 11 11 11 16 15 19 19 30
Singapore 272 19 21 24 26 32 34 37 39 40
Thailand 188 12 16 17 18 19 25 26 26 29
Taiwan 352 18 27 35 38 42 40 43 49 60
3,393 194 226 254 280 348 377 488 534 683

Sample description: sample distribution by jurisdiction and industry (percentage)

Whole sample (%) China (%) Hong Kong (%) Indonesia (%) India (%) Japan (%) Korea (%) Malaysia (%) The Philippines (%) Singapore (%) Thailand (%) Taiwan (%)
Auto 4 4.6 0 0 8.66 11.38 9.41 0 0 0 0 0.26
Capital goods 3.83 1.15 3.09 3.86 6.81 3.52 5.10 9.95 1.97 8.09 0 0.52
Conglomerates 3.91 0.66 7.87 3.85 3.13 5.15 1.18 0 10.53 15.81 0 0
Consumer 12.58 18.23 18.26 16.74 9.94 15.18 12.16 4.71 11.84 6.62 16.49 3.88
Financial service 14.01 7.55 17.98 18.45 18.60 4.88 16.47 23.04 23.03 6.62 23.94 10.85
Healthcare 1.52 0 0 0 8.29 0 0.39 4.19 0 0 0 0
Hotels/leisure 1.32 0 5.62 0 0 0 0 9.42 0 2.21 0 0.78
Infrastructure 0.62 0 2.53 2.15 0.55 0 0 0 3.29 0 0 0
Insurance 1.91 5.09 0.56 0 0 0 6.67 0 0 0 0 4.65
Internet 1.01 3.94 0 0 0.92 1.08 0.78 0 0 0 0 0.26
Materials 10.52 16.91 2.25 27.04 9.76 11.65 9.02 12.04 3.95 4.78 4.79 7.75
Media 1.66 0 0 5.15 3.13 0.81 0 0 3.29 3.31 6.91 0
Petro/Chems 5.49 7.06 0 1.29 9.02 2.17 9.02 3.66 0 0 15.96 8.27
Power 6.89 14.45 4.78 3.00 5.71 5.15 3.53 9.42 21.71 0 12.23 0
Property 9.76 8.70 28.93 9.87 4.05 4.34 0 10.47 8.55 20.96 19.68 1.02
Technology 12.52 3.28 4.21 0 7.37 29.81 17.65 0 0 0 0 55.56
Telecoms 5.01 4.76 0.55 8.17 3.50 0.54 7.84 11.52 11.84 8.46 0 6.20
Transport 3.43 3.62 3.37 0.43 0.56 4.34 0.78 1.58 0 23.14 0 0

Descriptive statistics: descriptive statistics for the whole sample

Mean Standard Minimum 5th percentile 25th percentile Median 75th percentile 95th percentile Maximum
FERRORit 0.015 0.057 0 0 0.002 0.005 0.013 0.056 2.549
DISPERSIONit 0.009 0.012 0 0.001 0.003 0.005 0.010 0.026 0.237
VOL_REVit 0.008 0.017 0 0 0.002 0.004 0.008 0.026 0.073
CGit 0.543 0.146 0.009 0.096 0.449 0.551 0.646 0.776 0.971
SIZEit 11.66 2.601 5.422 6.891 9.732 11.42 13.11 16.66 19.56
SD_ROEit 10.77 0.245 0 1.913 3.917 6.370 10.96 26.22 67.55
LOSSit 0.064 0.245 0 0 0 0 0 1 1
EARNINGS_SURit 0.039 0.103 0 0 0.007 0.017 0.039 0.133 3.625

Descriptive statistics: means, medians and standard deviations of key variables of interest by jurisdiction

FERRORit DISPERSIONit VOL_REVit CGit
Mean Median Standard Mean Median Standard Mean Median Standard Mean Median Standard
China 0.013 0.005 0.025 0.010 0.006 0.011 0.008 0.005 0.010 0.459 0.467 0.148
Hong Kong 0.018 0.005 0.051 0.009 0.005 0.010 0.010 0.004 0.031 0.584 0.590 0.126
Indonesia 0.023 0.006 0.049 0.009 0.005 0.016 0.010 0.005 0.012 0.428 0.416 0.167
India 0.015 0.006 0.032 0.008 0.005 0.010 0.007 0.004 0.008 0.522 0.517 0.127
Japan 0.021 0.006 0.063 0.006 0.004 0.014 0.011 0.005 0.027 0.544 0.538 0.140
Korea 0.050 0.011 0.020 0.015 0.009 0.020 0.017 0.009 0.023 0.546 0.535 0.141
Malaysia 0.012 0.006 0.016 0.007 0.005 0.006 0.005 0.004 0.004 0.581 0.582 0.114
The Philippines 0.043 0.006 0.211 0.010 0.005 0.013 0.013 0.004 0.015 0.524 0.549 0.174
Singapore 0.009 0.004 0.017 0.007 0.005 0.008 0.022 0.003 0.009 0.583 0.586 0.119
Thailand 0.021 0.006 0.054 0.012 0.008 0.017 0.008 0.005 0.007 0.629 0.637 0.098
Taiwan 0.009 0.004 0.016 0.007 0.005 0.007 0.009 0.006 0.012 0.542 0.554 0.120

Descriptive statistics: comparison of variables by institutional environment

Comparison of means Comparison of medians
Common law Code law Difference t-statistic Common law Code law Difference z-statistic
FERRORit 0.015 0.027 −0.012*** −3.45 0.006 0.007 −0.001* −1.59
DISPERSIONit 0.008 0.011 −0.003*** −7.77 0.005 0.007 −0.002*** −6.79
VOL_REVit 0.010 0.013 −0.003* −1.48 0.005 0.006 −0.001*** −7.26
CGit 0.560 0.520 0.040*** 8.09 0.560 0.536 0.024*** 12.36
SIZEit 11.49 12.36 −0.87*** −11.17 15.20 11.69 3.51*** 8.53
SD_ROEit 10.60 19.69 9.09*** 2.82 6.151 7.868 −1.717*** −6.47
LOSSit 0.063 0.064 −0.001 −0.12 0 0 0 −0.85
EARNINGS_SURit 0.042 0.235 −0.193 1.18 0.017 0.026 −0.009 0.33
Notes:

Provides t-statistics from the two-sample t-tests of the equality of means and z-statistics from Wilcoxon rank-sum tests of the equality of medians;

***

significant at the 1% level;

**

significant at the 5% level;

*

significant at the 10% level

Descriptive statistics: means of CGit by jurisdiction and year

2004 2005 2006 2007 2008 2009 2010 2011 2012
China 0.505 0.522 0.534 0.564 0.420 0.434 0.457 0.483 0.472
Hong Kong 0.651 0.662 0.663 0.664 0.603 0.598 0.580 0.559 0.577
Indonesia 0.484 0.491 0.477 0.496 0.374 0.413 0.402 0.391 0.439
India 0.574 0.553 0.548 0.539 0.528 0.466 0.517 0.521 0.538
Japan N/A N/A N/A N/A 0.590 0.587 0.539 0.541 0.542
Korea 0.597 0.639 0.645 0.624 0.563 0.541 0.546 0.535 0.502
Malaysia 0.614 0.645 0.644 0.650 0.582 0.582 0.510 0.553 0.575
The Philippines 0.591 0.591 0.591 0.620 0.473 0.495 0.520 0.527 0.511
Singapore 0.642 0.647 0.639 0.634 0.537 0.538 0.538 0.537 0.548
Thailand 0.629 0.633 0.643 0.648 0.617 0.656 0.660 0.660 0.615
Taiwan 0.600 0.610 0.608 0.600 0.494 0.495 0.504 0.509 0.526
Whole sample 0.603 0.603 0.596 0.607 0.530 0.529 0.528 0.536 0.534
Notes:

Variable are defined as follows: FERRORit = the absolute analysts’ forecast error, calculated as actual earnings minus the initial mean analyst forecasted earnings following the annual reports, deflated by beginning-of-fiscal-year price; DISPERSIONit = the standard deviation of the initial individual analysts’ earnings forecasts following the annual reports, deflated by beginning-of-fiscal-year price; VOL_REVit = standard deviation of monthly forecast revisions over the fiscal year, deflated by beginning-of-fiscal-year price, where forecast revision is defined as current-month median forecast minus previous-month median forecast; CGit = corporate governance score provided by CLSA, divided by 100; SIZEit = the log of beginning-of-fiscal-year market value of equity in US $millions; SD_ROEit = the standard deviation of return on equity, computed over the preceding ten years; LOSSit = 1 if firm i reports negative earnings in year t, and 0 otherwise; EARNINGS_SURit = earnings surprise, calculated as the absolute value of the difference between the current year’s earnings per share and past year’s earnings per share, deflated by beginning-of-fiscal-year price

Correlation matrix

FERRORit DISPERSIONit VOL_REVit CGit SIZEit SD_ROEit LOSSit EARNINGS_SURit LEGALc ANTIDIRECTOR_RIGHTSc R_ANTIDIRECTOR_RIGHTSc
FERRORit 0.378 (0.001) 0.382 (0.001) −0.031 (0.043) −0.022 (0.078) 0.290 (0.001) 0.169 (0.001) 0.035 (0.007) −0.060 (0.001) −0.045 (0.002) 0.013 (0.385)
DISPERSIONit 0.515 (0.001) 0.538 (0.001) −0.095 (0.001) −0.065 (0.001) 0.037 (0.005) 0.283 (0.001) 0.223 (0.001) −0.121 (0.001) −0.081 (0.001) 0.020 (0.181)
VOL_REVit 0.455 (0.001) 0.462 (0.001) −0.057 (0.001) −0.042 (0.001) 0.052 (0.001) 0.115 (0.001) 0.003 (0.831) −0.019 (0.152) −0.021 (0.136) 0.001 (0.946)
CGit −0.114 (0.001) −0.142 (0.001) −0.084 (0.001) −0.069 (0.001) −0.078 (0.001) −0.026 (0.086) 0.008 (0.620) 0.132 (0.001) 0.056 (0.001) −0.028 (0.102)
SIZEit −0.028 (0.277) −0.078 (0.001) 0.002 (0.861) −0.067 (0.001) 0.005 (0.687) −0.024 (0.056) −0.016 (0.227) −0.162 (0.001) −0.400 (0.001) 0.018 (0.198)
SD_ROEit 0.139 (0.001) 0.144 (0.001) 0.259 (0.001) −0.026 (0.090) 0.001 (0.918) 0.088 (0.001) 0.019 (0.151) −0.048 (0.001) −0.052 (0.001) 0.009 (0.546)
LOSSit 0.240 (0.001) 0.181 (0.001) 0.217 (0.001) −0.025 (0.076) −0.013 (0.297) 0.134 (0.001) 0.007 (0.611) −0.002 (0.902) −0.041 (0.004) −0.001 (0.546)
EARNINGS_SURit 0.357 (0.001) 0.334 (0.001) 0.349 (0.001) −0.061 (0.920) 0.033 (0.001) 0.243 (0.001) 0.232 (0.001) −0.022 (0.117) −0.012 (0.432) −0.012 (0.416)
LEGALc −0.084 (0.001) −0.180 (0.001) −0.143 (0.001) 0.106 (0.001) −0.102 (0.001) −0.109 (0.001) −0.002 (0.902) −0.149 (0.001) 0.827 (0.001) 0.063 (0.001)
ANTIDIRECTOR_RIGHTSc −0.074 (0.001) −0.096 (0.001) −0.170 (0.001) 0.025 (0.155) −0.348 (0.001) −0.115 (0.001) −0.045 (0.002) −0.159 (0.001) 0.794 (0.001) 0.441 (0.001)
R_ANTIDIRECTOR_RIGHTSc 0.001 (0.925) 0.025 (0.099) −0.120 (0.001) −0.051 (0.003) −0.126 (0.001) −0.018 (0.198) −0.046 (0.001) −0.096 (0.001) 0.236 (0.001) 0.577 (0.001)
Notes:

Pearson (Spearman) correlations are above (below) the diagonal. The p-values for the correlation coefficient estimates are provided in parentheses; Variable are defined as follows: FERRORit = the absolute analysts’ forecast error, calculated as actual earnings minus the initial mean analyst forecasted earnings following the annual reports, deflated by beginning-of-fiscal-year price; DISPERSIONit = the standard deviation of the initial individual analysts’ earnings forecasts following the annual reports, deflated by beginning-of-fiscal-year price; VOL_REVit = standard deviation of monthly forecast revisions over the fiscal year, deflated by beginning-of-fiscal-year price, where forecast revision is defined as current-month median forecast minus previous-month median forecast; CGit = corporate governance score provided by CLSA, divided by 100; SIZEit = the log of beginning-of-fiscal-year market value of equity in US $millions; SD_ROEit = the standard deviation of return on equity, computed over the preceding ten years; LOSSit = 1 if firm i reports negative earnings in year t and 0 otherwise; EARNINGS_SURit = earnings surprise, calculated as the absolute value of the difference between the current year’s earnings per share and last year’s earnings per share, deflated by beginning-of-fiscal-year price

Pooled regressions: the impact of corporate governance on analysts’ earnings forecast characteristics (tests of H1, H2 and H3) (number of observations = 3,393)

Predicted sign Ordinary least squares estimation
(1) (2) (3)
Intercept ? 0.035* (1.60) 0.014*** (12.23) 0.010*** (6.10)
CGit −0.008*** (−3.01) −0.007*** (−5.75) −0.006*** (−3.09)
SIZEit −0.001*** (−2.52) −0.001*** (−2.88) 0.001 (0.25)
SD_ROEit + −0.001** (−2.04) −0.001 (−0.85) 0.001*** (10.99)
LOSSit + 0.065*** (18.25) 0.016*** (16.87) 0.023*** (16.87)
EARNINGS_SURit + 0.068*** (7.33) 0.033*** (15.30) 0.059*** (18.74)
Country fixed effect Yes Yes Yes
Industry fixed effect Yes Yes Yes
Year fixed effect Yes Yes Yes
Adjusted R2 33.98% 16.13% 17.25%
Notes:

FERRORit= a0+ a1CGit+ a2SIZEit+a3SD_ROEit+a4LOSSit+ a5EARNINGS_SURit+εit (1); DISPERSIONit= b0+ b1CGit+ b2SIZEit+ b3SD_ROEit+b4LOSSit+ b5EARNINGS_SURit+εit (2); VOL_REVit= c0+ c1CGit+ c2SIZEit+ c3SD_ROEit+c4LOSSit+ c5EARNINGS_SURit+εit (3); White-adjusted t-statistics for ordinary least squares estimation are in parentheses;

***

significant at the 1% level;

**

significant at the 5% level;

*

significant at the 10% level; Variables are defined as follows: FERRORit = the absolute analysts’ forecast error, calculated as actual earnings minus the initial mean analyst forecasted earnings following the annual reports, deflated by beginning-of-fiscal-year price; DISPERSIONit = the standard deviation of the initial individual analysts’ earnings forecasts following the annual reports, deflated by beginning-of-fiscal-year price; VOL_REVit = standard deviation of monthly forecast revisions over the fiscal year, deflated by beginning-of-fiscal-year price, where forecast revision is defined as current-month median forecast minus previous-month median forecast; CGit = corporate governance score provided by CLSA, divided by 100; SIZEit = the log of beginning-of-fiscal-year market value of equity in US $millions; SD_ROEit = the standard deviation of return on equity, computed over the preceding ten years; LOSSit = 1 if firm i reports negative earnings in year t and 0 otherwise; EARNINGS_SURit = earnings surprise, calculated as the absolute value of the difference between the current year’s earnings per share and past year’s earnings per share, deflated by beginning-of-fiscal-year price

Cross-country variation in the association between corporate governance and analysts’ earnings forecast characteristics (tests of H4) (number of observations = 2,865): anti-director right index: Original measure and revised measure

Jurisdiction Original measure (La Porta et al., 1998) Revised measure (Djankov et al., 2008)
China NA NA
Hong Kong 5 5
Indonesia 2 5
India 5 5
Japan 4 4.5
Korea 2 4.5
Malaysia 4 5
The Philippines 3 4
Singapore 4 5
Thailand 2 4
Taiwan 3 3
Note:

From Spamann (2010) Table I on page 475

Cross-country variation in the association between corporate governance and analysts’ earnings forecast characteristics (tests of H4) (number of observations = 2,865): regression results

Measure of ENVIRONMENTc LEGALc ANTIDIRECTOR_RIGHTSc R_ANTIDIRECTOR_RIGHTSc
Sign (4) (5) (6) (4) (5) (6) (4) (5) (6)
Intercept ? 0.016* (1.42) 0.023*** (11.20) 0.014*** (4.76) 0.034* (1.42) 0.020*** (4.50) −0.002 (−0.41) 0.007 (0.18) −0.001 (−0.01) 0.017* (1.63)
CGit −0.014* (1.38) 0.013* (−1.36) 0.010* (1.50) −0.031** (−2.20) 0.013*** (2.55) 0.015** (2.02) −0.025*** (−2.43) 0.016* (1.49) 0.030** (1.83)
CGit*ENVIRONMENTc −0.011*** (−2.82) −0.008*** (−3.17) −0.004** (−2.27) −0.006*** (−2.84) −0.006*** (−4.44) −0.005*** (−2.36) −0.003*** (−5.25) −0.005** (−2.26) −0.007** (−1.94)
ENVIRONMENTc ? −0.010 (−1.07) 0.001 (0.73) 0.006** (2.24) −0.007 (−1.06) 0.001 (0.27) 0.007*** (3.64) 0.002 (0.23) 0.006*** (4.12) −0.001 (−0.31)
SIZEit −0.001 (−0.13) −0.001*** (−6.95) −0.001*** (−4.95) −0.001 (−0.16) −0.001*** (−6.11) −0.001*** (−4.47) −0.001 (−0.14) −0.001*** (−6.14) −0.001*** (−4.46)
SD_ROEit + −0.001** (−2.30) −0.001** (−1.72) 0.001 (0.83) −0.001** (−2.22) −0.001** (−1.69) 0.001 (0.54) −0.001** (−2.23) −0.001* (−1.64) 0.001 (0.56)
LOSSit + 0.017*** (3.63) 0.016*** (17.03) 0.018*** (13.78) 0.017*** (3.04) 0.015*** (14.69) 0.020*** (12.75) 0.017*** (3.05) 0.015*** (14.58) 0.020*** (12.70)
EARNINGS_SURit + 0.469*** (41.78) 0.032*** (15.00) 0.059*** (18.80) 0.502*** (40.48) 0.030*** (12.79) 0.057*** (16.57) 0.502*** (40.46) 0.030*** (12.92) 0.058*** (16.66)
Country fixed effect Yes Yes Yes Yes Yes Yes Yes Yes Yes
Industry fixed effect Yes Yes Yes Yes Yes Yes Yes Yes Yes
Year fixed effect Yes Yes Yes Yes Yes Yes Yes Yes Yes
Adjusted R2 (%) 36.45 22.68 23.94 38.69 21.02 23.71 38.68 21.24 23.06%
Notes:

FERRORit= d0+ d1CGit+ d2CGit*ENVIRONMENTc+ d3ENVIRONMENTc+ d4SIZEit+d5SD_ROEit+d6LOSSit+ d7EARNINGS_SURit+εit (4); DISPERSIONit= e0+ e1CGit+ e2CGit*ENVIRONMENTc+ e3ENVIRONMENTc+ e4SIZEit+ e5SD_ROEit+e6LOSSit+ e7EARNINGS_SURit+εit (5); VOL_REVit= f0+ f1CGit+ f2CGit*ENVIRONMENTc+ f3ENVIRONMENTc+ f4SIZEit+ f5SD_ROEit+f5LOSSit+ f6EARNINGS_SURit+εit (6); White-adjusted t-statistics for ordinary least squares estimation are in parentheses;

***

significant at the 1% level;

**

significant at the 5% level;

*

significant at the 10% level; Variables are defined as follows: FERRORit = the absolute analysts’ forecast error, calculated as actual earnings minus the initial mean analyst forecasted earnings following the annual reports, deflated by beginning-of-fiscal-year price; DISPERSIONit = the standard deviation of the initial individual analysts’ earnings forecasts following the annual reports, deflated by beginning-of-fiscal-year price; VOL_REVit = standard deviation of monthly forecast revisions over the fiscal year, deflated by beginning-of-fiscal-year price, where forecast revision is defined as current-month median forecast minus previous-month median forecast; CGit = corporate governance score provided by CLSA, divided by 100; Environmentc is measured by three alterative country-level variables. The first variable is LEGALc, which is the legal origin (1 for a common law origin, and 0 for a code law origin) of the country where the company is domiciled in. The second variable is ANTIDIRECTOR_RIGHTSc, which is the anti-director rights index of the country where the company is domiciled in. The third variable is R_ANTIDIRECTOR_RIGHTSc, which is revised anti-director rights index of the country where the company is domiciled in; CGit*Environmentc = the interaction between CGit and Environmentc; SIZEit = the log of beginning-of-fiscal-year market value of equity in US $millions; SD_ROEit = the standard deviation of return on equity, computed over the preceding ten years; LOSSit = 1 if firm i reports negative earnings in year t and 0 otherwise; EARNINGS_SURit = earnings surprise, calculated as the absolute value of the difference between the current year’s earnings per share and last year’s earnings per share, deflated by beginning-of-fiscal-year price

Sensitivity tests: using lagged values of corporate governance score (number of observations = 3,393): simple correlations between CGt and CGt-1 by jurisdiction and whole sample

Country No. of observations Pearson correlation coefficient Spearman correlation coefficient
China 528 0.330*** 0.286***
Hong Kong 335 0.323*** 0.309***
Indonesia 208 0.359*** 0.279***
India 543 0.277*** 0.230***
Japan 369 0.127*** 0.129***
Korea 255 0.233*** 0.158***
Malaysia 191 0.410*** 0.391***
Philippines 152 0.402*** 0.347***
Singapore 272 0.496*** 0.399***
Thailand 188 0.271*** 0.240***
Taiwan 352 0.356*** 0.285***
Whole sample 3,393 0.320*** 0.265***

Sensitivity tests: using lagged values of corporate governance score (number of observations = 3,393): regression results

Predicted sign (1) (2) (3)
FERRORit DISPERSIONit VOL_REVit
Intercept ? 0.014* (1.41) 0.025*** (12.86) 0.019*** (6.46)
CGit-1 −0.006*** (−2.87) −0.005*** (−3.64) −0.003* (−1.56)
SIZEit −0.001 (−0.62) −0.001*** (−8.13) −0.001*** (−4.66)
SD_ROEit + −0.001*** (−2.41) −0.001** (−2.19) 0.001 (0.30)
LOSSit + 0.017*** (3.25) 0.016*** (16.47) 0.019*** (12.60)
EARNINGS_SURit + 0.455*** (40.37) 0.033*** (15.53) 0.056*** (17.11)
Country fixed effect Yes Yes Yes
Industry fixed effect Yes Yes Yes
Year fixed effect Yes Yes Yes
Adjusted R2 35.89% 23.05% 21.79%
Notes:

White adjusted t-statistics from ordinary least squares regressions are in parentheses;

***

significant at 1% level;

**

significant at 5% level;

*

significant at 10% level; Variables are defined as follows: FERRORit = change in the absolute analysts’ forecast error, calculated as actual earnings minus the initial mean analyst forecasted earnings following the annual reports, deflated by beginning-of-fiscal-year price; DISPERSIONit = change in the standard deviation of the initial individual analysts’ earnings forecasts following the annual reports, deflated by beginning-of-fiscal-year price; VOL_REVit = standard deviation of monthly forecast revisions over the fiscal year, deflated by beginning-of-fiscal-year price, where forecast revision is defined as current-month median forecast minus previous-month median forecast; CGit−1 = the lagged value of corporate governance score provided by CLSA, divided by 100; SIZEit = the log of beginning-of-fiscal-year market value of equity in US $millions; SD_ROEit = the standard deviation of return on equity, computed over the preceding ten years; LOSSit = 1 if firm i reports negative earnings in year t and 0 otherwise; EARNINGS_SURit = earnings surprise, calculated as the absolute value of the difference between the current year’s earnings per share and last year’s earnings per share, deflated by beginning-of-fiscal-year price

Notes

1.

Sell-side analysts’ forecasts have been used as a proxy for investor prediction of future earnings in accounting and finance research.

2.

La Porta et al. (1998) document that common law countries offer shareholders stronger legal protections against managerial entrenchment. As reported in Table 6 of Hung (2001), the Pearson (Spearman) correlation coefficient for legal system and anti-director rights index is 0.83 (0.79), both significant at 0.01 level.

3.

CLSA started to provide corporate governance ratings for a sample of emerging market economies in 2001 and 2002. The corporate governance ratings are unavailable in 2003 and starting from 2004, CLSA have only provided ratings for Asian-Pacific economies.

References

Aboody, D., Barth, M. and Kasznik, R. (2006), “Revaluations of fixed assets and future firm performance: evidence from UK”, Journal of Accounting and Economics, Vol. 26 Nos 1/3, pp. 149-178.

Ali, A., Klasa, S. and Yeung, E. (2007), “Corporate disclosures by family firms”, Journal of Accounting and Economics, Vol. 44 Nos 1/2, pp. 238-286.

Ariff, A., Steven, C. and Emanuel, D. (2014), “Institutional environment, ownership, and disclosure of intangibles: evidence from East Asia”, Journal of International Accounting Research, Vol. 13 No. 1, pp. 33-59.

Barniv, R. and Myring, M. (2006), “An international analysis of historical and forecast earnings in accounting-based valuation models”, Journal of Business Finance Accounting, Vol. 33 Nos 7/8, pp. 1087-1109.

Barniv, R., Myring, M. and Thomas, W. (2005), “The association between the legal and financial reporting environments and forecast performance of individual analysts”, Contemporary Accounting Research, Vol. 22 No. 4, pp. 727-758.

Barron, O., Byard, D., Kile, C. and Riedl, E. (2002), “High-technology intangibles and analysts’ forecasts”, Journal of Accounting Research, Vol. 40 No. 2, pp. 289-312.

Barth, M., Landsman, W. and Rendleman, R. (1998), “Option pricing-based bond value estimates and a fundamental components approach to account for corporate debt”, The Accounting Review, Vol. 73 No. 1, pp. 73-102.

Bartov, E., Mohanram, P. and Nissim, D. (2007), “Managerial discretion and the economic determinations of the disclosure volatility parameter for valuing ESOs”, Review of Accounting Studies, Vol. 12 No. 1, pp. 155-179.

Bebchuk, L., Cohen, A. and Ferrell, A. (2009), “What matters in corporate governance?”, Review of Financial Studies, Vol. 22 No. 2, pp. 783-827.

Behn, B., Choi, J. and Kang, T. (2008), “Audit quality and properties of analyst’ earnings forecasts”, Accounting Review, Vol. 83 No. 2, pp. 327-349.

Bhat, G., Hope, O.K. and Kang, T. (2006), “Does corporate governance transparency affect the accuracy of analyst forecasts?”, Accounting & Finance, Vol. 46 No. 5, pp. 715-732.

Brown, P., Wendy, B. and Peter, V. (2011), “Corporate governance, accounting and finance: a review”, Accounting & Finance, Vol. 51 No. 1, pp. 96-172.

Byard, D., Li, Y. and Weintrop, J. (2006), “Corporate governance and the quality of financial analysts’ information”, Journal of Accounting and Public Policy, Vol. 25 No. 5, pp. 609-625.

Chau, G. and Gray, S. (2010), “Family ownership, board independence and voluntary disclosure: evidence from Hong Kong”, Journal of International Accounting, Auditing and Taxation, Vol. 19 No. 2, pp. 93-109.

Chen, K., Chen, Z. and Wei, K. (2009), “Legal protection of investors, corporate governance, and the cost of equity Capital”, Journal of Corporate Finance, Vol. 15 No. 3, pp. 273-289.

Chen, C., Ding, Y. and Kim, C. (2010), “High-level politically connected firms, corruption, and analyst forecast accuracy around the world”, Journal of International Business Studies, Vol. 41 No. 9, pp. 1505-1524.

Claessens, S. and Fan, J. (2002), “Corporate governance in asia: a survey”, International Review of Finance, Vol. 3 No. 2, pp. 71-103.

Davidson, R., Goodwin-Stewart, J. and Kent, P. (2005), “Internal governance structures and earnings management”, Accounting and Finance, Vol. 45 No. 2, pp. 241-267.

Djankov, S., La Porta, R., Lopez-de-Silanes, F. and Shleifer, A. (2008), “The law and economics of self-dealing”, Journal of Financial Economics, Vol. 88 No. 3, pp. 430-465.

Duru, A. and Reeb, D. (2002), “International diversification and analysts’ forecast accuracy and bias”, Accounting Review, Vol. 77 No. 2, pp. 415-433.

Eng, L. and Mak, Y. (2003), “Corporate governance and voluntary disclosure”, Journal of Accounting and Public Policy, Vol. 22 No. 4, pp. 325-345.

Fama, E. and MacBeth, J. (1973), “Risk, return, and equilibrium: Empirical tests”, Journal of Political Economy, Vol. 81 No. 3, pp. 607-636.

Fan, J. and Wong, T. (2002), “Corporate ownership structure and the informativeness of accounting earnings in east Asia”, Journal of Accounting and Economics, Vol. 33 No. 3, pp. 401-425.

Frankel, R., Kothari, S. and Weber, J. (2006), “Determinants of the informativeness of analyst research”, Journal of Accounting and Economics, Vol. 41 Nos 1/2, pp. 295-327.

Giroud, X. and Mueller, H. (2011), “Corporate governance, product market competition, and equity prices”, The Journal of Finance, Vol. 66 No. 2, pp. 563-600.

Gow, I., Ormazabal, G. and Taylor, D. (2010), “Correcting for cross-sectional and time-series dependence in accounting research”, The Accounting Review, Vol. 85 No. 2, pp. 483-512.

Granger, C. (1969), “Investigating causal relations by economic models and cross-spectral methods”, Econometrica, Vol. 37 No. 3, pp. 424-438.

Gul, F., Hutchinson, M. and Lai, K. (2013), “Gender-diverse board and properties of analyst earnings forecasts”, Accounting Horizons, Vol. 27 No. 3, pp. 511-538.

Haw, I.M., Ho, S. and Li, Y. (2011), “Corporate governance and earnings management by classification shifting”, Contemporary Accounting Research, Vol. 28 No. 2, pp. 517-553.

Ho, S. and Wong, K. (2001), “A study of the relationship between corporate governance structures and the extent of voluntary disclosure”, Journal of International Accounting, Auditing and Taxation, Vol. 10 No. 2, pp. 139-156.

Hope, O.K. (2003a), “Accounting policy disclosures and analysts’ forecasts”, Contemporary Accounting Research, Vol. 20 No. 2, pp. 235-272.

Hope, O.K. (2003b), “Disclosure practices, enforcement of accounting standards, and analysts’ forecast accuracy”, Journal of Accounting Research, Vol. 41 No. 2, pp. 235-272.

Hwang, L., Jan, C. and Basu, S. (1996), “Loss firms and analysts’ earnings forecast errors”, Journal of Financial Statement Analysis, Vol. 1, pp. 18-31.

Januszewski, S., Koke, J. and Winter, J. (2002), “Production Competition, corporate governance and firm performance: an empirical analysis for Germany”, Research in Economics, Vol. 56 No. 3, pp. 299-332.

Johnson, S., Boone, P., Breach, A. and Friedman, E. (2000), “Corporate governance in the Asian financial crisis”, Journal of Financial Economics, Vol. 58 Nos 1/2, pp. 141-186.

Klapper, L. and Love, I. (2004), “Corporate governance, investor protection, and performance in emerging markets”, Journal of Corporate Finance, Vol. 10 No. 5, pp. 703-728.

Klein, A. (2002), “Audit committee, board of director characteristics, and earnings management”, Journal of Accounting and Economics, Vol. 33 No. 3, pp. 375-400.

La Porta, R., Lepoz-de-Silanes, F., Shleifer, A. and Vishny, R. (1998), “Law and finance”, Journal of Political Economy, Vol. 106 No. 6, pp. 1113-1155.

La Porta, R., Lepoz-de-Silanes, F., Shleifer, A. and Vishny, R. (2000), “Investor protection and corporate governance”, Journal of Financial Economics, Vol. 58 Nos 1/2, pp. 3-27.

Lang, M. and Lundholm, R. (1996), “Corporate disclosure policy and analyst behavior”, The Accounting Review, Vol. 71 No. 4, pp. 467-493.

Lang, M., Lins, K. and Miller, D. (2003), “ADRS”, Journal of Accounting Research, Vol. 41 No. 2, pp. 317-363. “Analysts and accuracy: Does cross listing in the United States improve a firm’s information environment and increase market value?”.

Lang, M., Lins, K. and Miller, D. (2004), “Concentrated control, analyst following, and valuation: do analysts matter most when investors are protected least?”, Journal of Accounting Research, Vol. 42 No. 3, pp. 589-624.

Lehavy, R., Li, F. and Kenneth, M. (2011), “The effect of annual report readability on analyst following and the properties of their earnings forecasts”, The Accounting Review, Vol. 86 No. 3, pp. 1087-1115.

Leone, A., Miguel, M. and Wasley, C. (2013). “Influential observations and inference in accounting research”, Working paper, University of Miami and University of Rochester, Rochester.

Malik, M. 2011. “Can Governance mitigate real earnings management?”, Working paper, Boston University, Boston.

Nagar, V., Nanda, D. and Wysocki, P. (2003), “Discretionary disclosure and stock-based incentives”, Journal of Accounting and Economics, Vol. 34 Nos 1/3, pp. 283-309.

Petersen, M. 2005. “Estimating standard errors in finance panel data sets: comparing approaches”, Working Paper, Northwestern University, Evanston, IL.

Shen, C. and Chih, H. (2007), “Earnings management and corporate governance in Asia’s emerging markets”, Corporate Governance: An International Review, Vol. 15 No. 5, pp. 999-1021.

Song, C., Thomas, W. and Yi, H. (2010), “Value relevance of Fars no. 157 fair value hierarchy information and the impact of corporate governance mechanisms”, The Accounting Review, Vol. 85 No. 4, pp. 1375-1410.

Spamann, H. (2010), “The ‘antidirector rights index’ revisited”, Review of Financial Studies, Vol. 23 No; 2, pp. 467-486.

Waymire, G. (1986), “Additional evidence on the accuracy of analyst forecasts before and after voluntary management earnings forecasts”, Accounting Review, Vol. 59 No. 1, pp. 129-142.

Xu, L. and Tang, A. (2012), “Internal control material weakness, analysts’ accuracy and bias, and brokerage reputation”, Review of Quantitative Finance and Accounting, Vol. 39 No. 1, pp. 27-53.

Yu, M. (2010), “Analyst following and corporate governance: emerging-market evidence”, Accounting Research Journal, Vol. 23 No. 1, pp. 69-93.

Further reading

Boubakri, N., Cosset, J. and Guedhami, O. (2005), “Postprivatization corporate governance”, The Role of Ownership Structure and Investor Protection, Vol. 76 No. 2, pp. 237-470.

Mitton, T. (2002), “A cross-firm analysis of the impact of corporate governance on the East Asian financial crisis”, Journal of Financial Economics, Vol. 64 No. 2, pp. 215-241.

Yeo, G., Tan, P., Ho, K. and Chen, S. (2002), “Corporate ownership structure and the Informativeness of earnings”, Journal of Business, Finance and Accounting, Vol. 29 No. 7, pp. 1023-1046.

Acknowledgements

The authors thank Credit Lyonnais Securities Asia (CLSA) for providing corporate governance ratings data. They appreciate valuable inputs from our discussant, Luminita Enache and other conference participants at American Accounting Association’s 2014 Mid-Atlantic Region Meeting as well as conference participants at 2014 Accounting Conference at Temple University. Minna Yu acknowledges the summer research support from Business Council of Leon Hess Business School. In addition, this work was supported, in part, by a Creativity and Research Grant from Monmouth University.

Corresponding author

Minna Yu can be contacted at: miyu@monmouth.edu