Insider Ownership and Analyst Forecast Properties

Abstract

In this article, we investigate the effect of insider ownership on analyst forecast properties and find a significant nonlinear relationship between the two. Specifically, in the low to medium range, a rise in insider ownership improves analyst forecast properties (making them more accurate and less dispersed), but a further rise in insider ownership beyond moderate levels leads to deteriorating forecasts. We also find that this nonlinear relationship is attenuated for firms in countries with better investor protection. We interpret our findings as evidence that the role of insider ownership as an interest alignment or entrenchment mechanism is diminished in these countries due to their stronger investor protection.

Keywords

insider ownership analyst forecast properties information environment corporate governance investor protection

Introduction

In this article, we investigate the effect of insider ownership on analyst forecast properties and find a nonlinear relationship between the two. A firm’s level of insider ownership by managers/directors and other major shareholders may affect the information environment through at least two channels. First, the level of insider ownership, especially that by a controlling shareholder who also plays a significant managerial role, may influence a firm’s disclosure incentives and practices. As the level of insider ownership initially rises, the closer interest alignment between insiders and outside investors may improve a company’s disclosure quality. Consistent evidence has been provided by researchers such as Warfield, Wild, and Wild (1995), who find that managerial ownership is related positively to earnings informativeness and negatively to discretionary accruals. However, as the level of insider ownership reaches beyond moderate levels, insiders can pursue private benefits of control, without much risk of being removed due to their high ownership, at the expense of outside shareholders. To hide their self-serving activities, the now entrenched insiders have incentives to withhold information from the public. Thus, the entrenchment effect may inhibit disclosure. Ruland, Tung, and George (1990); Nagar, Nanda, and Wysocki (2003); Karamanou and Vafeas (2005); and Baik, Kang, and Morton (2010), among others, provide consistent empirical evidence that firms with high insider ownership are less likely to make management earnings forecasts, and that when they do, the forecasts are less precise and less accurate. Furthermore, the earnings of these firms are less informative (Fan & Wong, 2002) and are more likely to be manipulated (Leuz, Nanda, & Wysocki, 2003).

Second, the level of insider ownership may affect financial analysts’ information intermediary and/or independent information production. The nature of the effect may depend on whether the primary role of financial analysts is information intermediary or alternative information production. The first view holds that analysts primarily play the role of information intermediary in the sense that they receive information disclosed by a firm, process it, and transmit it to the capital market. Supportive evidence for this view has been provided by, for example, Lang and Lundholm (1996), Hope (2003a, 2003b), and Lang, Lins, and Miller (2003). However, the second view posits that analysts primarily play the role of information provider (Bhushan, 1989; Diamond, 1985). In this setting, the external demand by outside investors for analysts’ services is greater for firms with poorer disclosure, with analysts playing the substitutional role in providing independent information to outside investors. Ahn, Cai, Hamao, and Ho (2005); Boubaker and Labegorre (2008); Chung, McInish, Wood, and Wyhowski (1995); Lehavy, Li, and Merkley (2011); and Lobo, Song, and Stanford (2012) provide supportive empirical evidence for this view.

The preceding discussions suggest that the relationship between insider ownership and analyst forecast properties may be nonlinear and complex. Suppose there is a nonlinear relationship between insider ownership and disclosure quality, driven by the interest alignment effect and entrenchment effect. If the primary role of financial analysts is information intermediary, making disclosure quality and the quality of financial analyst services complementary to each other, we can expect analyst forecast properties to improve as insider ownership initially rises, and then to deteriorate as insider ownership rises above moderately high levels. However, if the primary role of analysts is independent information production, then corporate disclosure quality and the quality of financial analyst services are substitutional for each other. In this setting, financial analysts’ incentives to provide independent information vary inversely with corporate disclosure quality, as such independent information about a firm with poorer disclosure resulting from high insider ownership may be more valuable to outside investors.¹ If so, the nonlinear relationship just described may be attenuated or even reversed.

However, as explained earlier, the literature offers inconclusive evidence as to which primary role dominates, suggesting that the two roles played by financial analysts are not mutually exclusive. Regardless, the forecast properties are the joint outcome of corporate disclosure, information processing, and independent information production by financial analysts. With the premise that analyst forecasts are influenced by insider ownership irrespective of the dominant role of analysts, the nature of the relationship between insider ownership and forecast properties is ultimately an empirical issue.

We conduct our analyses in an international setting, which provides us with much more variation in the levels of insider ownership than do developed countries such as the United States and the United Kingdom (La Porta, Lopez-de-Silanes, & Shleifer, 1999). In addition, it is well established in the finance and accounting literature that the level of legal protection a country provides to shareholders plays a significant role in explaining both the level of insider ownership (e.g., La Porta et al., 1999; Shleifer & Vishny, 1997) and disclosure quality (Leuz et al., 2003). Therefore, we also perform tests to determine whether and how the relationship between insider ownership and analyst forecast properties differs across countries with different levels of investor protection.

Our study is one of the first to provide fresh evidence for the assertion made by Frankel, Kothari, and Weber (2006) that insider ownership may affect the properties of analyst forecasts.² We briefly summarize our major findings as follows. Consistent with analysts’ primary role being that of information intermediary, we find that analyst forecast properties improve (forecasts become more accurate and less dispersed) as insider ownership initially rises and then deteriorate as insider ownership rises beyond the 40% to 65% range. Furthermore, we find that the nonmonotonic relationship is attenuated for firms in countries with stronger investor protection and stricter law enforcement. In addition, the inflection points that trigger the turning of the relationship are generally much higher in these countries, suggesting it is harder for insiders in these countries to become entrenched.

A key variable in our study is the level of insider ownership. Our main tests rely on ownership data retrieved from Thomson Reuters’ Worldscope database, which provides the sum of percentage of shares held by each immediate owner holding more than 5%. The immediate insiders identified in Worldscope are often nominee accounts, trusts, or other corporations. Tracing each immediate insider to the ultimate owner to differentiate a controlling insider (e.g., the CEO or board chairman) from a passive blockholder (e.g., a pension fund or mutual fund) is a task beyond the scope of our article. However, it is controlling shareholders, not passive blockholders, who may derive private benefits of control, and hence have the incentives and ability to exert influence to alter a firm’s information environment.

As a sensitivity test, we rerun our main tests using a smaller sample of firms for which controlling insiders are clearly identified (Carney & Child, 2013). Our inferences remain qualitatively the same, although the statistical significance levels are generally reduced.

Our article joins a growing line of research on the effect of corporate governance on analyst forecast properties. For example, Ackert and Athanassakos (2003) find that the level of institutional ownership is positively related to analyst forecast accuracy; Ali, Chen, and Radhakrishnan (2007) report that family firms have more accurate and less dispersed analyst forecasts; Clement, Rees, and Swanson (2003) report that analyst forecast accuracy is less dependent on analyst ability and the reputation of the analyst’s employer in code law countries than in common law countries; Byard, Li, and Weintrop (2006) show that analyst forecast accuracy is positively related to corporate governance quality as measured by the independence of the board of directors; Nowland (2008) reports that stricter governance regulations improve analyst forecast accuracy in eight East Asian countries; and Verriest (2014) shows that analyst forecast accuracy is positively related to auditing quality and a country’s securities regulations and enforcement.

The rest of this article is organized as follows. In the following section, we provide a brief literature review. We then discuss the data and research designs in the “Data and Method” section. We present and discuss our empirical results in the “Empirical Results” section, and we conclude our article in the “Conclusion” section.

Literature Review

Our article is most directly related to four recent studies. The first is a study by Taylor (2007), who finds based on a sample of U.K. firms that analyst forecast accuracy improves initially as the ownership by inside directors increases and then deteriorates as the insider ownership rises further beyond 18%. Our article differs from Taylor’s in several important ways. First, ownership in the United Kingdom is considered widely dispersed without much ownership concentration (La Porta et al., 1999). In addition, insider ownership in Taylor’s study is confined to ownership by board directors and top officers, without capturing ownership by other types of insiders such as founding families, who do not hold board or top management positions but yet may still extract private benefits of control through manipulative disclosure. Third, as alluded to earlier, our international setting allows us to address the important interplay between ownership structure and a country’s investor protection, which cannot be done in a single-country setting.

The second study is one by Haw, Ho, Hu, and Wu (2010), who investigate the relationship between the divergence of a controlling shareholder’s control rights in excess of cash flow rights and analyst forecast properties. However, our approaches and findings differ significantly from those of Haw et al. First, they focus on the largest (controlling) shareholder only, without considering the role of other insider shareholders in information production. In addition, regardless of the level of divergence between control and cash flow rights, insiders with high enough ownership may have the ability and incentives to distort information flows to the public (Ball, Kothari, & Robin, 2000; Bhushan, 1989). Second, although the sample period considered by Haw et al. spans from 1990 to 1996, the control-cash flow divergence measure they use is based on a single year due to limited data availability.³ Third, whereas we hypothesize and find a nonlinear relationship between insider ownership and forecast properties, Haw et al. test a monotonic, linear relationship between forecast properties and the control-cash flow divergence without finding any significant evidence. The inverted U-shaped relation in our article may explain the lack of a significant linear relationship in the study by Haw et al.

The third study is one by Liu (2016), who finds based on a sample of listed firms in China that analyst forecast properties are related to the level of state ownership, but not related consistently to the levels of ownership by other types of shareholders, including foreign investors, managers, and institutional investors. We first note that state ownership is generally insignificant in many other countries around the world (La Porta et al., 1999). Similar to Taylor (2007), Liu conducts analyses to determine the effects of each ownership type, while our focus is on the effect of aggregate ownership by all insiders, with each owning 5% or more. Therefore, Liu’s analyses may not be appropriate if the objective is to capture the effect of insider ownership on a firm’s information environment. For example, foreign investors together as a group may have a significant level of ownership, but individually each foreign investor may own a very small stake, and as such they are unlikely to exert much influence on the firm’s information production.

Finally, our article is also related to a study by Ali et al. (2007), who report that analyst forecasts are more accurate and less dispersed for family firms in the Standard & Poor’s 500 than for nonfamily firms. However, according to Ali et al., a significant equity ownership by a family is not always required for a firm to be classified as a family firm—A firm is a family firm if the founder and/or his or her heirs hold top management or board positions or are among the company’s largest shareholders. Furthermore, ownership by other insiders (e.g., directors who are not related to the founding family) is not considered. Therefore, their findings do not speak directly to the general effect of insider ownership on analyst forecast properties. In addition, investor protection in the United States is considered one of the best in the world (La Porta, Lopez-de-Silanes, Shleifer, & Vishny, 1998), limiting insiders’ ability to influence a firm’s information environment.

Data and Method

Analyst Forecast Properties

We obtain analyst forecast data from the Institutional Broker’s Estimate System (I/B/E/S) Detail File. We follow Hope (2003b) and Chang, Khanna, and Palepu (2000) by using analyst forecasts of annual earnings per share (EPS) instead of quarterly EPS, as quarterly reporting was not required in most of our sample countries during our sample period. Following Hope’s (2003b) approach to computing analyst forecast properties, we use the current fiscal year’s annual EPS forecasts in months 4 to 12 following the end of the previous fiscal year to allow sufficient time for the publication of the annual reports of the previous fiscal year. For example, for firms whose fiscal year-end is December 2000, we use forecasts made from April to December 2000. As we are interested in testing the effect of insider ownership, there is no single time during the year when one would expect insider ownership to have the greatest influence on analyst forecasts. Forecasts made over the 8-month period instead of in a single month therefore better capture the disclosure agency problems associated with insider ownership over the entire year.⁴

Following the literature (e.g., Haw et al., 2010), we focus on two dimensions of analyst forecast properties in this study: accuracy and dispersion.⁵ They are measured as follows.

Accuracy

Following Lang and Lundholm (1996) and Hope (2003b), analyst forecast accuracy is computed as the negative of the absolute difference between the actual annual EPS and the mean of individual annual EPS forecasts, scaled by the beginning-of-fiscal-year stock price, and then multiplied by 100 as follows to facilitate exposition (Barniv & Myring, 2015; Tan, Wang, & Welker, 2011)⁶:

Accuracy = - \frac{| A c t u a l E P S - M e a n F o r e c a s t e d E P S |}{B e g i n n i n g - o f - f i s c a l - y e a r s t o c k p r i c e} \times 100,

where Actual EPS refers to the actual EPS for fiscal year t, and Mean Forecasted EPS is the average of the annual EPS forecasts.

Dispersion

Consistent with the literature (e.g., Hope, 2003a), we measure analyst forecast dispersion based on the standard deviation of the annual EPS forecasts, deflated by the beginning-of-fiscal-year stock price and then multiplied by 100 as follows⁷:

Dispersion = \frac{S t a n d a r d d e v i a t i o n o f E P S f o r e c a s t s}{B e g i n n i n g - o f - f i s c a l - y e a r s t o c k p r i c e} \times 100 .

Following Hope (2003a), we keep the observation only if three or more analysts provide forecasts for the firm to minimize measurement problems.

Insider Ownership and Investor Protection Variables

We use the data item “Percentage of Shares Closely Held,” retrieved from Thomson Reuters’ Worldscope database, to measure insider ownership for each firm each year. Worldscope defines closely held shares as shares held by insiders, including senior corporate officers and directors and their immediate families; shares held in trusts/nominee accounts; shares held by another corporation; and shares held by individuals who hold 5% or more of the shares outstanding. Following many other international ownership studies such as those by Dahlquist, Pinkowitz, Stulz, and Williamson (2003); Doidge, Karolyi, and Stulz (2007); and Giannetti and Koskinen (2010) to mitigate the potential effect of noisy data and the skewed nature of ownership data, we use the decile rank in our baseline regressions, where a higher rank indicates a higher level of insider ownership.⁸

As mentioned in the “Introduction” section, our use of insider ownership data from Worldscope presents a potential concern, as these data do not distinguish controlling insiders such as CEO or board chairman, who may derive private benefits of control from passive blockholders such as pension/mutual funds that enjoy only the cash flow rights.⁹ Although passive blockholders are unlikely to influence a firm’s disclosure practices to their benefit, they may still be viewed as “insiders” in the sense that they have much better access to information about a firm than do small outside investors. In any case, we note that the inclusion of these passive blockholders is likely to bias against us finding significant results.

As a sensitivity test, we rerun our main tests using a smaller sample of firms for which controlling insiders are clearly identified. In particular, we use the data compiled by Carney and Child (2013) for the largest companies in nine East Asian economies (Hong Kong, Indonesia, Japan, Malaysia, the Philippines, Singapore, South Korea, Taiwan, and Thailand). Based on Carney and Child’s classifications of ultimate insiders, we define controlling insiders as those who also play significant management roles.

Our measures of investor protection variables are based on the recent law and finance literature pioneered by La Porta et al. (1998). We include these country-level governance variables for two reasons. First, as reported by La Porta et al. (1999), corporate ownership varies greatly from country to country and the key explanatory variables include a country’s legal origin and the level of legal protection for outside investors. It is thus important to control for the underlying factors that may influence our key independent variable of interest: the level of insider ownership. Second, a country’s investor protection is significantly related to analyst forecast properties, as reported by Chang et al. (2000) and Barniv, Myring, and Thomas (2005).

As all of the investor protection variables are taken from studies by La Porta et al. (1998); La Porta, Lopez-de-Silanes, and Shleifer (2006); and Bhattacharya and Daouk (2002), who describe them in more detail, we define them briefly here:

Antidirector rights, a measure of the ease with which shareholders exercise their voting rights and legal rights in suing directors, obtained from La Porta et al. (2006).

Legal origin, measured by a dichotomous variable classifying a country’s legal origin from either the common law or code law tradition, as outlined by La Porta et al. (1998).

Good government, measured by the sum of three variables from a study by La Porta et al. (1998): (1) government corruption index; (2) the risk of expropriation by the government, meaning outright confiscation or forced nationalization; and (3) the level of repudiation of contracts by the government. Studies have found that this variable is important in deterring risk arbitrage (Morck, Yeung, & Yu, 2000).

Rule of law, as a proxy for the overall quality of country’s legal system, is taken from La Porta et al. (1998).

Judicial efficiency, an assessment of the efficiency and integrity of the legal environment, is also taken from La Porta et al. (1998).

Insider trading law enforcement, a measure of a country’s insider trading law enforcement developed by Bhattacharya and Daouk (2002). We use this measure because stricter enforcement of insider trading laws may reduce the incentives for corporate insiders to withhold information.

Disclosure requirement, a measure of laws mandating disclosure, taken from La Porta et al. (2006). A higher score of this variable indicates a better information environment for outside investors and financial analysts alike.

Liability standard, a measure of the extent to which private enforcement is facilitated through liability rules, which could reduce the uncertainties and cost of private litigation, developed by La Porta et al. (2006).

Public enforcement, a public enforcement index taken from La Porta et al. (2006).

As many of these investor protection variables are highly correlated with one another, and hence cannot enter into a regression model simultaneously, we perform a factor analysis using maximum likelihood estimation procedures. Significant factor(s) based on this analysis can be viewed as the encompassing factor(s) capturing the overall effect of investor protection as represented by the preceding nine individual factors. Our factor analysis identifies one significant factor with an eigenvalue greater than 1.

The Appendix presents the results of our factor analysis. The associated coefficients with these nine raw factors are presented in the first column. Following the methodology used by Bushman, Piotroski, and Smith (2004), we rotate the factor using the varimax rotation technique and the standardized coefficients with the varimax rotation are presented in the second column. The three highest loadings are for the indices on Good government, Rule of law, and Judicial efficiency. We label this significant factor Investor Protection Level as an overall measure of a country’s external institutional environment.

Control Variables

To isolate the effect of insider ownership on analyst forecasts, we must include a set of variables that have previously been found to be important determinants for both our dependent and independent variables of interest.

Firm size is negatively correlated with insider ownership, as it requires more wealth to own 1% of a bigger firm than a smaller one. Firm size has also been found to affect analyst forecast properties (e.g., Duru & Reeb, 2002), as there is more information available on larger firms in general. Firm size is measured based on the natural logarithm of market capitalization (in US$ million). We also control for leverage, as it is easier for insiders to own 1% of the equity of a firm with higher leverage, ceteris paribus. Furthermore, both Myers (1984) and Jensen (1986) argue that higher leverage is associated with better governance because higher debt financing reduces the cash flows available to insiders. In addition, earnings for firms with high leverage are more volatile due to the fixed nature of interest expenses on debt, making it harder for analysts to forecast future earnings (Chen, Ding, & Kim, 2010). Leverage is measured based on the ratio of a firm’s total debt over its total assets. Studies have found that the number of analysts following a firm is positively related to analyst forecast properties (e.g., Lys & Soo, 1995). We take the natural logarithm of the number of analysts following the firm plus one as our measure.

Studies have found that earnings and their attributes are related to forecast properties. In particular, we control for earnings levels as measured by return on assets (ROA; for example, Eames, Glover, & Kennedy, 2002); a loss dummy if earnings are negative (Brown, 2001; Hwang, Jan, & Basu,1996); change in earnings as measured by the absolute value of the change in EPS over the previous year, scaled by the previous year’s EPS (Hope 2003b; Lang & Lundholm, 1996); historical earnings volatility as measured by the standard deviation of ROA for the previous 5-year period (F. Gu & Wang, 2005); and earnings skewness as measured by the skewness of EPS over the past 5 years (Z. Gu & Wu, 2003). In addition to financial performance, stock market performance has been associated with analyst forecast properties. We control for the market-to-book ratio (Barniv & Myring, 2015; Tan et al., 2011), annual stock returns in the previous year (Tan et al., 2011), stock return volatility as measured by the annualized standard deviation of weekly stock returns (Barniv & Myring, 2015), and an American depository receipt (ADR) dummy if the shares are cross-listed in the United States (Tan et al., 2011). Finally, intangible assets may be harder to value than physical or financial assets, compromising analysts’ ability to make accurate forecasts (F. Gu & Wang, 2005). We use the ratio of intangible assets to total assets as our last control variable (F. Gu & Wang, 2005; Tan et al., 2011).

Regression Framework

We test our hypotheses in a multivariate regression model as follows.

\begin{matrix} F o r e c a s t P r o p e r t y_{i, t} = α + β_{1} O w n e r s h i p_{i, t} + β_{2} {O w n e r s h i p}_{i, t}^{2} \\ + β_{3} O w n e r s h i p_{i, t} \times I n v e s t o r P r o t e c t i o n + β_{4} {O w n e r s h i p}_{i, t}^{2} \\ \times I n v e s t o r P r o t e c t i o n + β_{5} I n v e s t o r P r o t e c t i o n_{i, t} + β_{6} S i z e_{i, t} \\ + β_{7} L e v e r a g e_{i, t} + β_{8} \ln {(A n a l y s t + 1)}_{i, t} + β_{9} R O A_{i, t} + β_{10} L o s s_{i, t} \\ + β_{11} Δ E a r n_{i, t} + β_{12} E v o l a t_{i, t} + β_{13} S k e w n e s s_{i, t} \\ + β_{14} M / B_{i, t} + β_{15} S t o c k R e t_{i, t - 1} + β_{16} R e t u r n V o l_{i, t - 1} + β_{17} A D R_{i, t} \\ + β_{18} I n t a n g i b l e A s s e t s_{i, t} + ε_{i, t} . \end{matrix}

We run the preceding regression model for each of the two analyst forecast properties: Accuracy and Dispersion. Following Stulz (1988) and McConnell and Servaes (1990), this quadratic model specification (Ownership and Ownership²) allows for a nonlinear relationship between analyst forecast properties and insider ownership.¹⁰ Following common practice in the literature (e.g., Dittmar & Smith, 2007; Gordon & Pound, 1993; Wang & Yu, 2015), we use a dichotomous version to measure investor protection: Investor Protection takes the value of 1 if the principal factor Investor Protection Level, as determined in the “Insider Ownership and Investor Protection Variables” section, is greater than the sample median of the 39 countries and 0 otherwise.¹¹ In addition to all of the control variables defined earlier, we control for the industry effect at the two-digit Standard Industrial Classification (SIC) level, although their coefficients are not reported in the article to save space.

Our base sample consists of around 50,000 firm-year observations from 39 countries for the period 1990-2005, resulting in an unbalanced panel. We follow the procedure proposed by Petersen (2009), that jointly controls for both firm and year clustering. The standard errors and, consequently, the reported t statistics for the coefficients are therefore robust.

Empirical Results

Descriptive Statistics and Correlation Matrix

Table 1 presents the mean values of the key variables by country. Consistent with Chang et al. (2000), Lang et al. (2003), and Hope (2003b), there is considerable variation across our sample countries in terms of forecast properties, analyst coverage, and insider ownership. The mean forecast accuracy is −2.9, representing an average forecast error of 2.9% of the beginning-of-fiscal-year stock price, comparable with the mean of 4.1% reported by Hope (2003b). The mean forecast dispersion is 2.3, representing 2.3% of the stock price, significantly higher than the mean of 1.1% for the 22 developed countries in the study by Hope (2003b). The mean insider ownership around the world is 35.4%, significantly higher than the mean insider ownership of 3.1% for the United Kingdom reported by Taylor (2007).¹² The average number of analysts following each firm is 5.9, comparable with the findings reported by Lang et al. (2003) and Haw et al. (2010) in a similar international setting. The value of the principal factor that captures the level of investor protection effectiveness for each country, Investor Protection Level, is also displayed in Table 1. In general, developed economies have higher scores than emerging economies.

Table 1.

Basic Summary Statistics by Country.

					Analyst Forecast Properties
Country	n	Analyst Coverage	Ownership	Investor Protection Level	Accuracy	Dispersion
Argentina	164	7.3	55.68	−1.165	−7.4	4.1
Australia	1,255	5.1	36.10	0.748	−4.1	2.2
Austria	145	5.2	52.04	0.993	−3.7	2.6
Belgium	296	7.7	50.42	0.893	−3.8	2.2
Brazil	345	5.8	49.49	−0.540	−6.3	7.1
Chile	262	3.6	60.98	−0.614	−3.7	2.3
Denmark	391	5.0	33.11	1.092	−3.4	2.8
Finland	443	5.1	35.02	1.089	−4.4	3.2
France	1,708	7.3	52.60	0.781	−3.1	2.6
Germany	1,727	9.9	50.17	0.992	−3.6	3.0
Greece	245	3.6	61.12	−0.400	−2.7	1.9
Hong Kong	1,234	8.9	54.60	0.519	−6.1	3.6
India	274	3.9	52.84	−0.873	−2.9	1.3
Indonesia	544	6.0	63.16	−1.611	−9.1	7.2
Ireland	236	3.3	29.99	0.774	−5.7	2.3
Israel	69	3.4	49.85	0.196	−5.5	2.8
Italy	603	9.3	51.76	0.161	−4.9	3.2
Japan	8,101	3.6	43.18	0.825	−2.5	2.2
Korea	511	3.9	31.66	−0.396	−9.2	5.2
Malaysia	1,233	7.7	50.80	−0.055	−5.2	2.4
Mexico	71	7.7	60.74	−0.914	−10.7	11.6
The Netherlands	599	12.1	39.66	1.169	−3.0	2.3
New Zealand	385	4.7	51.31	1.185	−3.8	1.5
Norway	311	5.6	38.38	1.206	−6.3	4.3
Pakistan	97	3.3	57.27	−1.747	−2.6	3.4
Peru	147	3.9	59.24	−1.492	−10.3	7.2
The Philippines	358	7.2	72.27	−2.005	−11.8	8.5
Portugal	130	5.1	54.74	0.278	−5.5	3.6
Singapore	913	8.9	54.97	0.607	−4.1	2.9
South Africa	550	3.8	49.22	−0.145	−4.6	2.1
Spain	465	10.4	41.81	0.319	−3.7	3.0
Sri Lanka	33	3.4	33.07	−1.328	−4.7	1.7
Sweden	665	5.1	35.76	1.100	−3.3	2.3
Switzerland	718	8.0	39.76	1.229	−3.8	2.4
Taiwan	504	2.7	25.31	0.235	−2.8	1.6
Thailand	479	4.6	51.10	−0.771	−6.8	7.9
Turkey	502	5.2	63.79	−1.056	−5.6	8.4
The United Kingdom	4,014	5.3	23.55	1.004	−2.1	1.9
The United States	19,069	6.2	23.25	0.862	−1.7	1.5
Total	49,796	5.9	35.76	0.667	−2.9	2.3

Note. This table reports the mean values of the key variables by country. Accuracy and Dispersion are analyst forecast accuracy and dispersion, respectively, as defined in Equations 1 and 2 in the text. Analyst Coverage is the average number of I/B/E/S analysts for each firm in a country. Ownership is the average percentage of insider ownership in a country. Investor Protection Level is the value of the first principal factor, as identified in the Appendix; n is the number of firm-year observations in each country. I/B/E/S = Institutional Broker’s Estimate System.

Table 2 displays the correlation coefficients matrix for the variables in our study. Insider ownership is negatively associated with forecast accuracy but positively associated with dispersion, both significant at the 1% level. Consistent with Chang et al. (2000) and Barniv et al. (2005), the correlation matrix in Table 2 also shows that the investor protection factor is associated positively with accuracy but negatively with dispersion, both again significant at the 1% level. Consistent with the evidence of La Porta et al. (1999), insider ownership is negatively associated with the investor protection factor, significant at the 1% level. As for the control variables, they are generally correlated with the forecast properties in the predicted fashion.

Table 2.

Pearson Correlations.

Dependent variable	(a)	(b)	(c)	(d)	(e)	(f)	(g)	(h)	(i)	(j)	(k)	(l)	(m)	(n)	(o)	(p)	(q)
(a) Accuracy	1
(b) Dispersion	−.713(<.0001)	1
(c) Ownership	−.063(<.0001)	.086(<.0001)	1
(d) Investor Protection	.135(<.0001)	−.169(<.0001)	−.262(<.0001)	1
(e) Size	.199(<.0001)	−.218(<.0001)	−.244(<.0001)	.134(<.0001)	1
(f) Leverage	−.147(<.0001)	.158(<.0001)	.020(<.0001)	−.105(<.0001)	.012(<.0001)	1
(g) Analyst	.086(<.0001)	−.074(<.0001)	−.179(<.0001)	.046(<.0001)	.624(<.0001)	.006(.050)	1
(h) ROA	.252(<.0001)	−.256(<.0001)	.055(<.0001)	.035(<.0001)	.225(<.0001)	−.109(<.0001)	.129(<.0001)	1
(i) Loss	−.317(<.0001)	.328(<.0001)	−.009(.005)	.041(<.0001)	−.255(<.0001)	.107(<.0001)	−.141(<.0001)	−.448(<.0001)	1
(j) ΔEarn	−.002(.6705)	.005(.1867)	.001(.7195)	.005(.1645)	−.008(.0195)	.014(<.0001)	−.001(.9619)	−.002(.647)	.004(.1943)	1
(k) Evolat	−.039(<.0001)	.039(<.0001)	.129(<.0001)	−.088(<.0001)	−.201(<.0001)	−.059(<.0001)	−.139(<.0001)	−.066(<.0001)	.067(<.0001)	−.001(.8584)	1
(l) Skewness	.147(<.0001)	−.137(<.0001)	.002(.471)	−.054(<.0001)	.153(<.0001)	−.119(<.0001)	.089(<.0001)	.190(<.0001)	−.237(<.0001)	.001(.8682)	−.016(<.0001)	1
(m) M/B	.088(<.0001)	−.109(<.0001)	−.093(<.0001)	.026(<.0001)	.245(<.0001)	−.100(<.0001)	.115(<.0001)	.056(<.0001)	−.005(.254)	−.006(.167)	.107(<.0001)	.090(<.0001)	1
(n) Stock Return	.076(<.0001)	−.093(<.0001)	−.036(<.0001)	−.018(<.0001)	.081(<.0001)	−.085(<.0001)	.002(.627)	.112(<.0001)	−.081(<.0001)	−.007(.126)	.058(<.0001)	.074(<.0001)	.159(<.0001)	1
(o) Return Volatility	−.137(<.0001)	.141(<.0001)	.031(<.0001)	−.167(<.0001)	−.265(<.0001)	−.104(<.0001)	−.141(<.0001)	−.278(<.0001)	.276(<.0001)	.001(.916)	.272(<.0001)	−.077(<.0001)	.081(<.0001)	.106(<.0001)	1
(p) ADR	−.006(.1461)	.018(<.0001)	−.017(<.0001)	.122(<.0001)	.301(<.0001)	.037(<.0001)	.273(<.0001)	.037(<.0001)	−.034(<.0001)	−.001(.940)	−.065(<.0001)	.016(.001)	.010(.019)	.001(.843)	−.070(<.0001)	1
(q) Intangible Assets	.019(<.0001)	−.038(<.0001)	−.124(<.0001)	.034(<.0001)	.105(<.0001)	.074(<.0001)	.108(<.0001)	−.071(<.0001)	.024(<.0001)	−.004(.311)	.040(<.0001)	−.006(.210)	.057(<.0001)	.044(<.0001)	.045(<.0001)	−.021(<.0001)	1

Note. This table presents the correlation matrix for the variables used in our analysis. The values in parentheses are the p values and are the two-tailed t statistics. Accuracy (Dispersion) is the analyst forecast accuracy (dispersion). Ownership is the decile rank of the percentage of share ownership by all major insiders (each owning at least 5%). Investor Protection is equal to 1 if Investor Protection Level is above the sample median and 0 otherwise. Size is ln(market capitalization in millions). Leverage is the debt asset ratio. Analyst is ln(number of analyst following + 1). ROA is return on assets. Loss is equal to 1 if a loss is reported and 0 otherwise. ΔEarn is the absolute change in earnings from the previous year scaled by the previous year’s earnings. Evolat is the standard deviation of ROA for the previous 5 years. Skewness is the skewness on the time series of earnings over the past 5 years. M/B is the market-to-book ratio. Stock Return is the annual stock return for firm j at year t− 1, adjusted for the contemporaneous annual market return. Return Volatility is the standard deviation of weekly stock returns for firm j at year t− 1. ADR is an indicator variable that equals 1 if firm i in year t trades American depository receipt (ADR) in the United States. Intangible Assets is the ratio of intangible assets to total assets for firm i at the beginning of the year.

Baseline Regression Results

Next, we use multivariate regression analyses to test our hypotheses on the relationship between analyst forecast properties and insider ownership, after controlling for other determinants previously identified in the literature. As discussed earlier, because our data are cross-sectional time series in nature, we use Petersen’s (2009) methodology to correct for the potential correlations in the residuals, both across firms within a year and over time within a firm. Table 3 displays the regression results.

Table 3.

Determinants of Analyst Forecast Properties.

	Without interaction terms		With interaction terms
Dependent variable	Accuracy	Dispersion	Accuracy	Dispersion
Ownership	0.539*** (6.03)	−0.219*** (−5.54)	0.687*** (6.76)	−0.274*** (−6.36)
Ownership ²	−0.048*** (−5.72)	0.023*** (6.10)	−0.061*** (−6.36)	0.029*** (6.80)
Ownership×Investor Protection			−0.656*** (−3.30)	0.260*** (2.83)
Ownership ²× Investor Protection			0.059*** (3.21)	−0.027*** (−3.07)
Investor Protection	1.219*** (7.03)	−0.412*** (−5.11)	2.541*** (5.70)	−0.862*** (−4.14)
Size	1.277*** (21.81)	−0.711*** (−27.00)	1.270*** (21.62)	−0.708*** (−26.87)
Leverage	−6.958*** (−15.49)	3.789*** (18.32)	−6.958*** (−15.47)	3.797*** (18.35)
Analyst	−1.084*** (−10.88)	0.746*** (15.63)	−1.061*** (−10.60)	0.736*** (15.36)
ROA	7.964*** (6.38)	−2.293*** (−4.30)	7.975*** (6.39)	−2.302*** (−4.31)
Loss	−7.344*** (−28.02)	4.095*** (31.56)	−7.345*** (−28.03)	4.097*** (31.58)
ΔEarn	−0.041 (−0.93)	0.020 (1.05)	−0.041 (−0.93)	0.020 (1.05)
Evolat	−0.124 (−0.58)	0.177 (1.62)	−0.125 (−0.59)	0.180 (1.63)
Skewness	0.562 (0.91)	−0.199 (−1.27)	0.564 (0.99)	−0.199 (−1.38)
M/B	0.027 (1.07)	−0.026*** (−2.54)	0.027 (1.07)	−0.027*** (−2.58)
Stock Return	0.403*** (3.96)	0.003 (0.27)	0.406*** (3.99)	0.003 (0.28)
Return Volatility	−2.479*** (−6.95)	1.113*** (7.31)	−2.480*** (−6.95)	1.109*** (7.28)
ADR	−1.085*** (−4.28)	0.705*** (5.88)	−1.130*** (−4.41)	0.715*** (5.92)
Intangible Assets	−0.024 (−0.05)	−0.449*** (−2.59)	−0.029 (−0.07)	−0.459*** (−2.64)
Two-digit SIC dummies	Included	Included	Included	Included
Adjusted R²	.1670	.2058	.1671	.2060
Sample size	49,796	49,796	49,796	49,796

Note. This table provides the regression results for assessing the effect of insider ownership on analyst forecast properties. The dependent variables are Accuracy, which is the negative of the absolute value of the difference between actual annual EPS and the mean forecasted EPS deflated by the beginning-of-fiscal-year stock price, and Dispersion, which is the standard deviation of annual EPS forecasts deflated by the beginning-of-fiscal-year stock price. Ownership is the decile rank of the percentage of share ownership by all major insiders (each owning at least 5%). Investor Protection is equal to 1 if Investor Protection Level is above the sample median and 0 otherwise. Size is ln(market capitalization in millions). Leverage is the debt asset ratio. Analyst is ln(number of analyst following + 1). ROA is the return on assets. Loss is equal to 1 if a loss is reported and 0 otherwise. ΔEarn is the absolute change in earnings from the previous year scaled by the previous year’s earnings. Evolat is the standard deviation of ROA for the previous 5 years. Skewness is the skewness on the time series of earnings over the past 5 years. M/B is the market-to-book ratio. Stock Return is the annual stock return for firm j at year t− 1, adjusted for contemporaneous annual market return. Return Volatility is the standard deviation of weekly stock returns for firm j at year t− 1. ADR is an indicator variable that equals 1 if firm i in year t trades American depository receipt in the United States. Intangible Assets is the ratio of intangible assets to total assets for firm i at the beginning of the year. Industry indicator variables are included in all of the model specifications. The robust standard errors method of Petersen’s (2009) two-way clustering is used, and t statistics are reported in parentheses (two-tailed). EPS = earnings per share; SIC = Standard Industrial Classification.

*, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

In the first two columns, we report the regression results without interacting the insider ownership variables with the investor protection variable to establish the baseline findings. The results from the first column in Table 3 suggest strong evidence of a curvilinear relation between insider ownership and analyst forecast accuracy—an inverted U-shaped relation in which accuracy first increases, then decreases as insider ownership rises. The results from the second column in Table 3 also suggest a curvilinear relationship between insider ownership and dispersion—a U-shaped relation in which dispersion declines as insider ownership rises initially, but increases as insider ownership rises beyond a high level. Furthermore, analyst forecasts are more accurate and less dispersed in countries with stronger investor protection, consistent with studies by Chang et al. (2000) and Hope (2003b).

Taken together, these results are consistent with the notion that as insider ownership initially rises, the converging interest alignment between insiders and outside investors improves a firm’s incentives for timely and accurate disclosure, resulting in more accurate and less dispersed analyst forecasts. Equivalently, analysts following the firm may find it easier to gather and process information due to more and/or higher quality disclosure. However, as insider ownership rises beyond a high level of ownership, insiders may become entrenched and want to withhold information from the public, resulting in less accurate and more dispersed forecasts. The deteriorating analyst forecast properties for firms whose shares are heavily concentrated in the hands of a few insiders are also consistent with the decreased demand for financial analysts’ services due to limited trading opportunities for outside investors (Bhushan, 1989) or the lack of disclosed information jeopardizing analysts’ ability to make accurate forecasts.

Based on the regression coefficients on Ownership and Ownership² in column 1 on forecast accuracy, the inflection point in the inverted U-shaped curve is 5.62, representing an insider ownership of 42.9%.¹³ Similarly, the inflection point for the U-shaped curve between forecast dispersion and insider ownership is 4.8, representing an insider ownership of 37.8%. Each inflection point is computed by first setting the partial derivative on Ownership to 0, solving for the rank value of Ownership, and then converting the rank value to a percentage based on our sample data.

To assess the economic significance of the effect of insider ownership on forecast accuracy (dispersion), we compute the change in accuracy (dispersion) from the first decile of insider ownership to the inflection point and then to the last decile of insider ownership, holding all other variables in the regression at their means. Forecast accuracy would improve by 1.02 percentage points from the first decile of insider ownership to the inflection point, and deteriorate by 0.92 percentage points from the inflection point to the last decile of insider ownership,¹⁴ both quite significant in relation to the mean forecast accuracy of 2.9 percentage points as reported in Table 1. Similarly, dispersion would decrease by 0.32 percentage points from the first decile of insider ownership to the inflection point, and increase by 0.63 percentage points from the inflection point to the last decile of insider ownership,¹⁵ both significant as well in relation to the mean forecast dispersion of 2.3 percentage points as reported in Table 1.

Most of the control variables have the predicted signs, consistent with earlier findings in the literature. For example, the coefficients on Size and ROA are significantly positive, indicating that analyst forecasts are more accurate for firms that are bigger in size and more profitable, a finding consistent with prior studies (e.g., Barniv & Myring, 2015; Tan et al., 2011). The coefficients on Leverage, Loss, Return Volatility, and ADR are negative and significant, indicating it is relatively more difficult for analysts to forecast earnings for firms with higher leverage, reporting losses, higher stock return volatility, and ADR listing, consistent with the findings of Lim, Lim, and Lobo (2013); Barniv and Myring (2015); and Tan et al. (2011). Other control variables exhibit less explanatory power. One notable deviation from the literature is the effect of analyst following, which is negatively related to forecast accuracy in our regression results, although the correlation between the two variables is significantly positive as shown in Table 2. We note the very high correlation of .624 between firm size and analyst following (see Table 2), which may explain the sign change on the coefficient of the latter when both are simultaneously included in a regression. We confirm this by removing firm size from the regressions, which has the effect of reversing the sign on analyst following.

The Effects of Investor Protection

In columns 3 and 4 of Table 3, we present the regression results with the interaction terms between investor protection and insider ownership included, to evaluate whether the relationship varies significantly with the level of investor protection. As can be seen, both of the interaction terms are significantly different from 0. In particular, the nonlinear relationship between insider ownership and forecast accuracy and dispersion is much less pronounced for firms in countries with more effective investor protection, as the magnitudes of the coefficients on Ownership and Ownership² are significantly smaller for those firms.

Specifically, for the accuracy regression in column 3, the coefficients on Ownership and Ownership² are 0.687 and −0.061, respectively, for countries with weaker investor protection, and 0.031 (= 0.687 − 0.656) and −0.002 (= −0.061 + 0.059), respectively, for countries with stronger investor protection. Although the inverted U-shaped pattern is also true for countries with stronger investor protection, the smaller magnitudes of the coefficients noted previously suggest that the role of insider ownership as an interest alignment mechanism (at the low to medium range) and as an entrenchment mechanism (above a high level) is significantly diminished in those countries. This is plausible because internal governance (insider ownership in our case) and investor protection mechanisms are natural substitutes.

Based on the regression coefficients on Ownership and Ownership² in column 3, the inflection points in the inverted U-shaped curve are 5.665 and 8.736, respectively, representing insider ownership of 43.7% and 68.9% for countries with weaker and stronger investor protection, respectively. We interpret the higher inflection point for firms in countries with stronger investor protection as evidence that entrenchment in these countries requires a far higher level of insider ownership, because entrenchment is harder in these countries.

Column 4 in Table 3 reports the results with analyst forecast dispersion as the dependent variable. Briefly, the U-shaped relationship between insider ownership and dispersion is less pronounced for countries with stronger investor protection. As can be seen in column 4, the coefficients on Ownership and Ownership² are −0.274 and 0.029, respectively, for countries with weaker investor protection, and −0.014 (= −0.274 + 0.26) and 0.002 (= 0.029 − 0.027), respectively, for countries with stronger investor protection. The flatter relationship between insider ownership and dispersion for firms in these countries is further evidence that the role of insider ownership, as an interest alignment or entrenchment mechanism, is reduced due to the more effective investor protection. We also compute the inflection points for the relationship between forecast dispersion and insider ownership in the same manner as for accuracy; they are 4.793 (representing 38.3% ownership) for firms in countries with weaker investor protection, and 6.187 (44.4% ownership) for firms in countries with stronger investor protection.

In summary, analyst forecast properties are significantly related to insider ownership in a nonlinear manner. At low to moderate levels, a rise in insider ownership represents better alignment of interests between corporate insiders and outside investors, resulting in analyst forecasts that are more accurate and less dispersed. However, beyond moderate levels of ownership, a further rise in insider ownership encourages entrenchment, which reduces forecast accuracy and increases forecast dispersion. Both the interest alignment and entrenchment effects are less pronounced for firms in countries with stronger investor protection, consistent with the view that internal and external governance mechanisms are natural substitutes (La Porta et al., 1998).

Robustness Tests

In this section, we perform additional tests to check the robustness of our findings. First, we repeat our analysis by replacing the decile rank of insider ownership with the actual insider ownership level. To save space, the regression results are not tabulated here. Our findings remain qualitatively the same using the actual insider ownership levels instead of their ranks.

Second, instead of using McConnell and Servaes’s (1990) quadratic specification of the nonlinear effect of insider ownership, we run a piecewise linear model following Morck, Shleifer, and Vishny (1988). Specifically, we define the piecewise segments as follows¹⁶:

O w n e r s h i p 0 t o 5 = {\begin{matrix} I n s i d e r o w n e r s h i p i f i n s i d e r o w n e r s h i p < 0.05 \\ 0.05 i f i n s i d e r o w n e r s h i p \geq 0.05 \end{matrix},

O w n e r s h i p 5 t o 50 = {\begin{matrix} 0 i f i n s i d e r o w n e r s h i p < 0.05 \\ (I n s i d e r o w n e r s h i p - 0.05) i f 0.05 \leq i n s i d e r o w n e r s h i p < 0.5 \\ 0.45 i f i n s i d e r o w n e r s h i p \geq 0.5 \end{matrix},

O w n e r s h i p O v e r 50 = {\begin{matrix} 0 i f i n s i d e r o w n e r s h i p < 0.5 \\ (I n s i d e r o w n e r s h i p - 0.5) i f i n s i d e r o w n e r s h i p \geq 0.50 . \end{matrix}

We rerun the regressions as specified in Equation 4 replacing Ownership and Ownership² with Owership 0 to 5, Owernship 5 to 50, and Owernship Over 50. The resulting piecewise regression results, again not tabulated to save space, are consistent with the findings using the quadratic specifications. For example, for the accuracy regression model, the coefficient on insider ownership is significantly positive for the range up to 5%, significantly positive with a smaller magnitude for the range of 5% to 50%, and significantly negative for ownership greater than 50%. All of the coefficients are attenuated for firms in countries with better investor protection, similar to the findings that use the quadratic form as reported in Table 3.

Third, to mitigate concerns about the effect of outliers in computing accuracy, we replace the mean forecast with the median forecast to compute the forecast accuracy, and then rerun the accuracy regressions. We also check whether our findings are affected by the staleness of forecasts by using only the last forecast per analyst-year to compute accuracy and dispersion. The results of these sensitivity tests are displayed in Panel A of Table 4. (All of the control variables are omitted to save space.) These alternative definitions do not lead to any material changes in our inferences as reported previously.

Table 4.

Determinants of Analyst Forecast Properties—Sensitivity Tests.

Panel A: Forecast Properties Computed With the Median and/or Last Forecasts.
	Without interaction terms			With interaction terms
Dependent variable	ACCURACY_Med	ACCURACY_Last	DISPERSION_Last	ACCURACY_Med	ACCURACY_Last	DISPERSION_Last
Ownership	0.395*** (5.39)	0.383*** (6.65)	−0.167*** (−5.24)	0.520*** (6.15)	0.472*** (7.08)	−0.197*** (−5.51)
Ownership ²	−0.037*** (−5.37)	−0.038*** (−7.02)	0.019*** (6.19)	−0.047*** (−5.91)	−0.046*** (−7.23)	0.022*** (6.35)
Ownership×Investor Protection				−0.320*** (−3.30)	−0.291*** (−3.07)	0.106* (1.88)
Ownership ²×Investor Protection				0.034*** (2.93)	0.034*** (2.89)	−0.015** (−2.09)
Investor Protection	0.726*** (5.09)	0.385*** (3.47)	−0.211*** (−3.16)	1.922*** (5.21)	1.199*** (4.07)	−0.436** (−2.53)
Adjusted R²	.2281	.2151	.2709	.2293	.2153	.2710
Sample size	49,796	49,796	49,796	49,796	49,796	49,796

Panel B: Firms From the United States, the United Kingdom, and Japan Removed.
	Accuracy	Dispersion
Ownership	0.346*** (3.10)	−0.169** (−2.18)
Ownership ²	−0.028** (−2.14)	0.018** (2.31)
Ownership×Investor Protection	−0.028* (−1.93)	0.117* (1.76)
Ownership ²×Investor Protection	0.009* (1.73)	−0.015* (−1.89)
Investor Protection	3.438*** (5.34)	−1.173*** (−3.42)
Adjusted R²	.1714	.2132
Sample size	18,612	18,612

Panel C: Separate Regressions for Subsamples Divided by Investor Protection.
	Accuracy		Dispersion
	High investor protection countries	Low investor protection countries	High investor protection countries	Low investor protection Countries
Ownership	0.181** (2.05)	0.294* (1.78)	−0.064* (−1.66)	−0.063* (−1.76)
Ownership ²	−0.016* (−1.86)	−0.045** (−2.38)	0.006 (1.54)	0.023*** (2.82)
Adjusted R²	.1520	.1998	.2166	.2179
Sample size	28,914	20,882	28,914	20,882

Panel D: Separate Regressions for Subsamples Divided Into “Outsider” or “Insider” Economies as Identified by Leuz, Nanda, and Wysocki (2003).
	Accuracy		Dispersion
	Outsider economies	Insider economies	Outsider economies	Insider economies
Ownership	0.217** (2.03)	0.351* (1.77)	−0.116** (−2.49)	−0.084* (−1.93)
Ownership ²	−0.013 (−1.14)	−0.035** (−2.16)	0.012** (2.55)	0.020*** (2.64)
Adjusted R²	.1612	.1830	.1918	.2071
Sample size	28,029	21,767	28,029	21,767

Note. This table reports the results of additional robustness tests. Ownership is the decile rank of the percentage of share ownership by all major insiders (each owning at least 5%). Investor Protection is equal to 1 if Investor Protection Level is above the sample median and 0 otherwise. To save space, none of the coefficients of the control variables is tabulated. The sample period is 1990-2005. The robust standard errors method of Petersen’s (2009) two-way clustering is used, and t statistics are reported in parentheses (two-tailed). Panel A presents the regression results using forecast properties computed with the median/last forecasts. ACCURACY_Med is the negative of the absolute value of the difference between the actual annual EPS and median forecasted EPS; ACCURACY_Last is the negative of the absolute value of the difference between the actual annual EPS and last forecasted EPS; and DISPERSION_Last is the standard deviation of the last annual EPS forecasts. Panel B presents the regression results with firms from the United States, the United Kingdom, and Japan removed. Accuracy is the negative of the absolute value of the difference between the actual annual EPS and mean forecasted EPS; and Dispersion is the standard deviation of annual EPS forecasts. Panel C presents the separate regression results for each subsample divided by Investor Protection. Panel D presents the separate regression results for each subsample divided into “Outsider Economies” or “Insider Economies” as identified by Leuz et al. (2003, p. 519). Please see their discussion and Table 3 for details. EPS = earnings per share.

*, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Fourth, as there are disproportionate numbers of firms from the United States, the United Kingdom, and Japan, our baseline analyses may overweigh them, as they have large numbers of observations. We rerun the regressions after these countries are removed, with the results reported in Panel B of Table 4. (All of the control variables are omitted to save space.) As can be seen, our inferences are very much intact, although the significance levels for the coefficients are generally lower than those in Table 3.

Furthermore, we run separate regressions for subsamples divided into strong and weak investor protection, with the results displayed in Panels C and D of Table 4. (Again, none of the control variables is tabulated.) In Panel C, our sample firms are divided by the dichotomous value of Investor Protection. In Panel D, our sample is divided using the country clusters identified by Leuz et al. (2003). In particular, we follow their grouping of countries into two subsamples: A group of “outsider economies,” characterized by developed stock markets, low ownership concentration, extensive minority shareholder rights, high corporate transparency, and effective legal enforcement; and another group of “insider economies” with markedly less developed stock markets, high ownership concentration, weaker minority shareholder rights, low corporate transparency, and weaker legal enforcement.¹⁷ As can be seen from Panels C and D, although the statistical significance levels for the coefficients on insider ownership are generally lower for the subsample regressions than for those in Table 3, all except one remain significant at the 10% level or better, and thus our inferences remain qualitatively similar as before. We also run country fixed-effect regressions. The untabulated results are consistent with those in Table 3, but the significance levels are generally lower. Therefore, caution should be exercised in interpreting our evidence, as we cannot rule out the possibility that our results may be driven by a country-specific variable, not insider ownership. However, our regressions in Table 3 already take into account the effect of a “catch-all” variable, that is, Investor Protection, which represents many facets of a country’s legal and financial institutions. Such a country-specific variable would create an omitted-variable bias only if it were correlated with Investor Protection and/or insider ownership.

Finally, we repeat our analyses for a subsample of the nine East Asian economies using the data from a study by Carney and Child (2013). Starting from Worldscope’s detailed immediate insider ownership data for 2008, Carney and Child trace out each ultimate insider using many other data sources such as Bureau van Dijk’s OSIRIS, Gale’s Major Companies of Asia and Australasia 2008, annual reports, company websites, business reports, and newspaper articles. The ultimate insiders are then grouped into the following categories: family/individual, state (both domestic and foreign), widely held financial institutions, and widely held corporations. Most importantly, Carney and Child also identify whether a member of the family or an employee of the state/widely held financial institution or corporation is the CEO, board chairman, honorary board chairman, or vice board chairman. A firm with such an arrangement is not only owned but also managed by the controlling shareholder. Accordingly, for this sensitivity test, insider ownership is set to the level of ownership by the controlling shareholder for firms with such an arrangement and 0 otherwise.

Carney and Child’s (2013) ownership data were compiled at the end of 2008, the year of a major global financial crisis, and this may make earnings more difficult to forecast due to the significant jumps in volatility that year. In addition, analyst forecast data for many of the sample firms are not available, with only 465 of the original 1,296 sample firms covered by I/B/E/S. Given that an ownership structure is rather stable over time, we run regressions using 2007, 2008, and 2009 to overcome these problems and to increase the power of our empirical tests. Table 5 reports the regression results, and they are consistent with our baseline results in Table 3, although the significance levels are generally lower.¹⁸ For example, the inflection points in the inverted-U shaped relationship between accuracy and controlling ownership are 4.41 and 6.23 (representing ownership levels of 33.7% and 45.4%, respectively) for firms in countries with weaker and stronger investor protection, respectively. Similarly, the inflection points in the U-shaped relationship between dispersion and controlling ownership are 5.24 and 6.71 (representing ownership levels of 37.9% and 49.8%, respectively) for firms in countries with weaker and stronger investor protection, respectively. These inflection points are comparable with those reported in Table 3.

Table 5.

Determinants of Analyst Forecast Properties—With Carney and Child’s (2013) Controlling Insider Ownership Data.

	Without interaction terms		With interaction terms
Dependent variable	Accuracy	Dispersion	Accuracy	Dispersion
Cont_Ownership	0.301* (1.70)	−0.269* (−1.70)	0.374 (1.65)	−0.555* (−1.72)
Cont_Ownership ²	−0.061** (−1.99)	0.019 (1.57)	−0.042* (−1.72)	0.053* (1.65)
Cont_Ownership×Investor Protection			−0.143*** (−3.05)	0.338** (2.35)
Cont_Ownership ²×Investor Protection			0.024*** (4.28)	−0.037** (−2.66)
Investor Protection	7.606** (34.91)	−1.028*** (−5.12)	0.943 (1.58)	−1.746*** (−5.13)
Size	0.364 (1.34)	−0.080* (−1.68)	0.327 (1.31)	−0.074 (−1.59)
Leverage	−3.882*** (−2.21)	0.031*** (3.56)	−3.977** (−2.37)	0.032*** (3.74)
Analyst	−0.509*** (−13.21)	0.192*** (7.32)	−0.537*** (−12.16)	0.165*** (5.53)
ROA	−1.102 (−0.59)	9.771 (1.55)	−1.382 (−0.81)	9.888 (1.58)
Loss	−4.564** (−2.40)	1.995*** (3.98)	−4.599*** (−2.61)	2.014*** (4.01)
ΔEarn	−0.101 (−1.53)	0.343** (2.56)	−0.119 (−1.51)	0.344** (2.66)
Evolat	−0.091*** (−4.31)	0.242*** (3.79)	−0.082*** (−3.55)	0.241*** (3.76)
Skewness	0.435 (1.07)	0.328** (2.08)	−0.327 (−1.10)	0.323** (2.07)
M/B	0.141*** (3.52)	−0.402*** (−4.34)	0.167*** (3.16)	−0.409*** (−4.61)
Stock Return	0.001*** (0.01)	0.001 (0.01)	0.001 (0.01)	0.001 (0.01)
Return Volatility	−0.343 (−1.12)	1.572* (1.68)	−0.359** (−2.13)	1.615* (1.70)
ADR	−0.636** (−2.37)	−0.244 (−0.59)	−0.598** (−2.50)	−0.169 (−0.43)
Intangible Assets	1.756 (1.03)	1.031 (0.48)	2.065 (1.17)	1.140 (0.54)
Two-digit SIC dummies	Included	Included	Included	Included
Adjusted R²	.2775	.2379	.4013	.2410
Sample size	1,392	1,392	1,392	1,392

Note. This table provides additional regression results using the controlling ownership data provided by Carney and Child (2013). The dependent variables are Accuracy and Dispersion. Cont_Ownership is the decile rank of the percentage of share ownership by insiders who also hold significant management or board positions (e.g., serving as the CEO, board chairman, or board vice chairman). Investor Protection is equal to 1 if Investor Protection Level is above the sample median and 0 otherwise. Size is ln(market capitalization in millions). Leverage is the debt asset ratio. Analyst is ln(number of analyst following + 1). ROA is the return on assets. Loss is equal to 1 if a loss is reported and 0 otherwise. ΔEarn is the absolute change in earnings from the previous year scaled by the previous year’s earnings. Evolat is the standard deviation of ROA for the previous 5 years. Skewness is the skewness on the time series of earnings over the past 5 years. M/B is the market-to-book ratio. Stock Return is the annual stock return for firm j at year t− 1, adjusted for the contemporaneous annual market return. Return Volatility is standard deviation of weekly stock returns for firm j at year t− 1. ADR is an indicator variable that equals 1 if firm i in year t trades American depository receipt (ADR) in the United States. Intangible Assets is the ratio of intangible assets to total assets for firm i at the beginning of the year. Industry indicator variables are included in all of the model specifications. The robust standard errors method of Petersen’s (2009) two-way clustering is used, and t statistics are reported in parentheses (two-tailed).

*, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Conclusion

In this article, we document a significant nonlinear relationship between insider ownership and analyst forecast properties. Specifically, in the low to medium range, a rise in insider ownership improves forecasts; however, a further rise in insider ownership beyond moderate levels of ownership leads to deteriorating forecasts.

Our results suggest that insider ownership is a significant determinant of analyst forecast properties. In particular, a rise in insider ownership, in the low to medium range, represents a better alignment of interests between insiders and outside investors, which is conducive to timely and accurate disclosure. However, a rise in insider ownership beyond moderate levels of ownership makes entrenchment more likely, increasing the potential incentives to withhold or manipulate the information disclosed to the public.

We also find that the nonlinear relationship between insider ownership and analyst forecast properties is attenuated in countries with stronger investor protection. We interpret this as evidence that the role of insider ownership as an interest alignment or entrenchment mechanism is less important in these countries due to the presence of more effective external forces that mitigate the potential conflict between insiders and outside minority shareholders.

Our study adds to the growing literature on the effect of corporate governance quality on analyst forecast properties. By focusing on the level of insider ownership and its interplay with the level of investor protection in a country, our characterization of corporate governance extends the measures used in the literature such as board independence, CEO compensation schemes, and institutional ownership.

Footnotes

Appendix

	Investor protection level
	Factor pattern	Factor pattern: Varimax rotation
Antidirector Rights	0.48429	0.05292
Legal Origin	0.46491	0.3474
Good Government	0.77325	0.96577
Rule of Law	0.68585	0.88454
Efficiency of Judicial System	0.72012	0.79165
Insider Trading	0.36829	0.24919
Disclosure Requirement	0.61383	0.12845
Liability Standard	0.47584	0.10894
Public Enforcement	0.22631	−0.25396
Eigenvalue	2.8276406

Note. This appendix table provides the factor analysis loading results of the nine country-level governance variables obtained from La Porta, Lopez-de-Silanes, Shleifer, and Vishny (1998); La Porta, Lopez-de-Silanes, and Shleifer (2006); and Bhattacharya and Daouk (2002). Legal Origin takes a value of 1 if a country is of common-law origin and 0 otherwise; Antidirector Rights Index ranges from 0 to 6; Rule of Law ranges from 0 to 10; Judicial Efficiency ranges from 0 to 10; and Good Government ranges from 0 to 30, equal to the sum of a country’s corruption index, risk of expropriation index, and risk of contract repudiation index, based on a study by La Porta et al. (1998). Insider Trading is the insider trading law enforcement index taken from Bhattacharya and Daouk. Disclosure Requirement, Liability Standard, and Public Enforcement, each ranging from 0 to 1, are taken from La Porta et al. (2006).

Acknowledgements

We thank the anonymous referees, Steven Cahan, and Theodore Sougiannis (Associate Editor) for their many comments, which have greatly improved the paper. We also thank Yuan Huang, Nancy Su, Donghui Wu, and workshop participants at City University of Hong Kong and the 2014 Annual IFMA Conference for their constructive discussions and comments.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Financial support from the Hong Kong Polytechnic University is gratefully acknowledged.

Notes

References

Ackert

Athanassakos

(1997). Prior uncertainty, analyst bias, and subsequent abnormal returns. Journal of Financial Research, 20, 263-273.

Ackert

Athanassakos

(2003). A simultaneous equations analysis of analysts’ forecast bias, analyst following, and institutional ownership. Journal of Business Finance & Accounting, 30, 1017-1041.

Ahn

H.-J.

Cai

Hamao

(2005). Adverse selection, brokerage coverage, and trading activity on the Tokyo Stock Exchange. Journal of Banking & Finance, 29, 1483-1508.

Ali

Chen

T. Y.

Radhakrishnan

(2007). Corporate disclosure by family firms. Journal of Accounting & Economics, 44, 238-286.

Baik

Kang

Morton

(2010). Why are analysts less likely to follow firms with high managerial ownership? Journal of Accounting, Auditing & Finance, 25, 171-200.

Ball

Kothari

S. P.

Robin

(2000). The effect of international institutional factors on properties of accounting earnings. Journal of Accounting & Economics, 29, 1-51.

Barniv

Myring

M. J.

(2015). How would the differences between IFRS and U.S. GAAP affect U.S. analyst performance?Journal of Accounting and Public Policy, 34, 28-51.

Barniv

Myring

M. J.

Thomas

W. B.

(2005). The association between the legal and financial reporting environments and forecast performance of individual analysts. Contemporary Accounting Research, 22, 727-758.

Bhattacharya

Daouk

(2002). The world price of insider trading. Journal of Finance, 57, 75-108.

10.

Bhushan

(1989). Firm characteristics and analyst following. Journal of Accounting & Economics, 22, 255-274.

11.

Boubaker

Labegorre

(2008). Ownership structure, corporate governance and analyst following: A study of French listed firms. Journal of Banking & Finance, 32, 961-976.

12.

Brown

L. D.

(2001). A temporal analysis of earnings surprises: Profits versus losses. Journal of Accounting Research, 39, 221-242.

13.

Bushman

Piotroski

Smith

(2004). What determines corporate transparency?Journal of Accounting Research, 42, 207-252.

14.

Byard

Weintrop

(2006). Corporate governance and the quality of financial analysts’ information. Journal of Accounting and Public Policy, 25, 609-625.

15.

Carney

R. W.

Child

T. B.

(2013). Changes to the ownership and control of East Asian corporations between 1996 and 2008: The primacy of politics. Journal of Financial Economics, 107, 494-513.

16.

Chang

Khanna

Palepu

(2000, January). Analyst activity around the world (Working paper, HBS Strategy Unit Working Paper No. 01-061). Retrieved from http://ssrn.com/abstract=204570

17.

Chen

Ding

Kim

(2010). High-level politically connected firms, corruption, and analyst forecast accuracy around the world. Journal of International Business Studies, 41, 1505-1524.

18.

Chung

K. H.

McInish

T. H.

Wood

Wyhowski

(1995). Production of information, information asymmetry, and the bid-ask spread: Empirical evidence from analysts’ forecasts. Journal of Banking & Finance, 19, 1025-1046.

19.

Claessens

Djankov

Lang

(2000). The separation of ownership and control in East Asian corporations. Journal of Financial Economics, 58, 81-112.

20.

Clement

M. B.

Rees

Swanson

E. P.

(2003). The influence of culture and corporate governance on the characteristics that distinguish superior analysts. Journal of Accounting, Auditing & Finance, 18, 593-618.

21.

Dahlquist

Pinkowitz

Stulz

R. M.

Williamson

(2003). Corporate governance and home bias. Journal of Financial and Quantitative Analysis, 38, 87-110.

22.

Diamond

D. W.

(1985). Optimal release of information by firms. Journal of Finance, 40, 1071-1094.

23.

Dittmar

Smith

J. M.

(2007). Corporate governance and the value of cash holdings. Journal of Financial Economics, 83, 599-634.

24.

Doidge

Karolyi

G. A.

Stulz

R. M.

(2007). Why do countries matter so much for corporate governance?Journal of Financial Economics, 86, 1-39.

25.

Duru

Reeb

D. M.

(2002). International diversification and analysts’ forecast accuracy and bias. The Accounting Review, 77, 415-433.

26.

Eames

Glover

Kennedy

(2002). The association between trading recommendations and broker analysts’ earnings forecasts. Journal of Accounting Research, 40, 85-104.

27.

Faccio

Lang

(2002). The ultimate ownership of Western European corporations. Journal of Financial Economics, 65, 365-395.

28.

Fan

Wong

T. J.

(2002). Corporate ownership structure and the informativeness of accounting earnings in East Asia. Journal of Accounting & Economics, 33, 401-425.

29.

Frankel

Kothari

S. P.

Weber

(2006). Determinants of the informativeness of analyst research. Journal of Accounting & Economics, 41, 29-54.

30.

Giannetti

Koskinen

(2010). Investor protection, equity returns, and financial globalization. Journal of Financial and Quantitative Analysis, 45, 135-168.

31.

Gordon

L. A.

Pound

(1993). Information, ownership structure, and shareholder voting: Evidence from shareholder-sponsored corporate governance proposals. Journal of Finance, 48, 697-718.

32.

Wang

(2005). Intangible assets, information complexity, and analysts’ earnings forecasts. Journal of Business Finance & Accounting, 32, 1673-1702.

33.

J. S.

(2003). Earnings skewness and analyst forecast bias. Journal of Accounting & Economics, 35, 5-29.

34.

Haw

(2010). Analysts’ forecast properties, concentrated ownership and legal institutions. Journal of Accounting, Auditing & Finance, 25, 235-259.

35.

Hope

O.-K.

(2003a). Accounting policy disclosures and analysts’ forecast. Contemporary Accounting Research, 20, 295-321.

36.

Hope

O.-K.

(2003b). Disclosure practices, enforcement of accounting standards, and analysts’ forecast accuracy: An International study. Journal of Accounting Research, 41, 235-272.

37.

Hwang

Jan

Basu

(1996). Loss firms and analysts’ earnings forecast errors. Journal of Financial Statement Analysis, 1, 18-30.

38.

Jensen

M. C.

(1986). Agency cost of free cash flow, corporate finance, and takeovers. American Economic Review, 76, 323-329.

39.

Karamanou

Vafeas

(2005). The association between corporate boards, audit committees, and management earnings forecasts: An empirical analysis. Journal of Accounting Research, 43, 453-486.

40.

La Porta

Lopez-de-Silanes

Shleifer

(1999). Corporate ownership around the world. Journal of Finance, 54, 471-517.

41.

La Porta

Lopez-de-Silanes

Shleifer

(2006). What works in securities laws?Journal of Finance, 61, 1-32.

42.

La Porta

Lopez-de-Silanes

Shleifer

Vishny

R. W.

(1998). Law and finance. Journal of Political Economy, 106, 1113-1155.

43.

Lang

M. H.

Lins

K. V.

Miller

D. P.

(2003). ADRs, analysts, and accuracy: Does cross listing in the United States improve a firm’s information environment and increase market value?Journal of Accounting Research, 41, 317-345.

44.

Lang

M. H.

Lundholm

R. J.

(1996). Corporate disclosure policy and analysts behavior. The Accounting Review, 71, 467-492.

45.

Lehavy

Merkley

(2011). The effect of annual report readability on analyst following and the properties of their earnings forecasts. The Accounting Review, 86, 1087-1115.

46.

Leuz

Nanda

Wysocki

P. D.

(2003). Investor protection and earnings management: An international comparison. Journal of Financial Economics, 69, 506-527.

47.

Lim

C. Y.

Lim

Lobo

G. J.

(2013). IAS 39 reclassification choice and analyst earnings forecast properties. Journal of Accounting and Public Policy, 32, 342-356.

48.

Liu

(2016). Ownership structure and analysts’ forecast properties: A study of Chinese listed firms. Corporate Governance, 16, 54-78.

49.

Lobo

G. J.

Song

Stanford

(2012). Accruals quality and analyst coverage. Journal of Banking & Finance, 36, 497-508.

50.

Lys

Soo

(1995). Analysts’ forecast precision as a response to competition. Journal of Accounting, Auditing & Finance, 10, 751-763.

51.

McConnell

Servaes

(1990). Additional evidence on equity ownership and corporate value. Journal of Financial Economics, 27, 595-612.

52.

Morck

Shleifer

Vishny

(1988). Managerial ownership and market valuation. Journal of Financial Economics, 20, 293-315.

53.

Morck

Yeung

(2000). The information content of stock markets: Why do emerging markets have synchronous stock price movements?Journal of Financial Economics, 58, 215-260.

54.

Myers

S. C.

(1984). The capital structure puzzle. Journal of Finance, 39, 575-592.

55.

Nagar

Nanda

Wysocki

(2003). Discretionary disclosure and stock-based incentives. Journal of Accounting & Economics, 34, 283-309.

56.

Nowland

(2008). The effect of national governance codes on firm disclosure practices: Evidence from analyst earnings forecasts. Corporate Governance: An International Review, 16, 475-491.

57.

Petersen

M. A.

(2009). Estimating standard errors in finance panel data sets: Comparing approaches. Review of Financial Studies, 22, 435-480.

58.

Ruland

Tung

George

(1990). Factors associated with the disclosure of managers’ forecasts. The Accounting Review, 65, 710-721.

59.

Shleifer

Vishny

(1997). A survey of corporate governance. Journal of Finance, 52, 737-783.

60.

Stulz

(1988). Managerial control of voting rights: Financial policies and the market for corporate control. Journal of Financial Economics, 20, 25-54.

61.

Tan

Wang

Welker

(2011). Analyst following and forecast accuracy after mandated IFRS adoptions. Journal of Accounting Research, 49, 1307-1357.

62.

Taylor

(2007, August). Ownership structure and analysts’ earnings forecasts: UK evidence (Working paper). Retrieved from http://ssrn.com/abstract=922727

63.

Verriest

(2014). Auditor Governance, institutions and analyst forecast properties: International evidence. Journal of International Accounting Research, 13, 1-32.

64.

Wang

W. J.

(2015). The information content of stock prices, legal environments, and accounting standards: International evidence. European Accounting Review, 24, 471-493.

65.

Warfield

T. D.

Wild

J. J.

Wild

K. L.

(1995). Managerial ownership, accounting choices, and informativeness of earnings. Journal of Accounting & Economics, 20, 61-91.