The Association Between the Legal and Financial Reporting Environments and Forecast Performance of Individual Analysts

In civil law countries, where the demand for earnings information is reduced because of weaker corporate governance mechanisms and lower-quality financial reporting, we predict that anal

The Association Between the Legal and Financial Reporting

Environments and Forecast Performance of Individual Analysts

Ran BarnivKent State University Graduate School of Management, CBA

Kent, OH 44242Phone: 330-672-1112Fax: 330-672-2545E-mail: rbarniv@bsa3.kent.edu

Mark MyringBall State UniversityCollege of BusinessMuncie, IN 47306Phone: 765-285-5108Fax: 765-285-8024E-mail: mmyring@bsu.edu

Wayne B Thomas*

University of OklahomaMichael F Price College of Business

307 W Brooks, Room 212BNorman, OK 73019Phone: 405-325-5789Fax: 405-325-7348E-mail: wthomas@ou.edu

* Contact author

We appreciate the helpful comments of two anonymous reviewers, Patricia O’Brien (associate editor), A Amir, S Beninga, workshop participants at the universities of Cincinnati,

Pennsylvania State, and Tel Aviv, and participants in a concurrent session of the AAA annual

meeting, Hawaii, August 2003 We also acknowledge Thomson Financial for providing I/B/E/S

consistently outperform their peers because appropriate market-based incentives exist In civil law countries, where the demand for earnings information is reduced because of weaker corporate governance mechanisms and lower-quality financial reporting, we predict that analysts with superior ability will less

consistently provide superior forecasts Results are consistent with our

expectations and suggest an association between legal and financial reporting environments and analysts’ forecast behavior

Keywords Analysts’ characteristics, relative forecast performance, common law,

civil law

JEL Descriptors G38, K22, M41

The Association Between the Legal and Financial Reporting Environments and Forecast Performance of Individual Analysts

1 Introduction

We test the ability of analyst characteristics to explain relative forecast accuracy across legal origins In common law countries, where investor protectionlaws are stronger and financial reporting is generally perceived to have higher quality (La Porta, Lopez-de-Silanes, Shleifer, and Vishny 1997, 1998, 2000a; Ball, Kothari, and Robin 2000), the increased demand by investors for earnings information may create incentives for analysts to provide that information

accurately Analysts with superior characteristics (e.g., ability, effort, experience, resources, etc.) are more likely to issue a superior forecast relative to their peers

In civil law countries, weaker investor protection laws and lower-quality financialreporting may reduce the economic incentives of analysts to incur costly activities

to provide a superior earnings forecast We expect that it will be more difficult to relate individual analysts’ characteristics to relative forecast performance in civil law countries

Examining the relation between relative forecast performance and analyst characteristics across legal regimes provides evidence outside the United States, where the bulk of this research has been conducted.1 Understanding analyst behavior in other environments provides additional insight into how analysts’ efforts in accurately forecasting earnings can contribute to the informational

efficiency of financial markets (Frankel, Kothari, and Weber 2002) The results also contribute to our understanding of the relation between investors’ demands and analysts’ behavior (Defond and Hung 2002) As the value relevance of

reported earnings declines, investors may have less demand for analysts’ earnings forecasts and demand other sources of information such as cash flow forecasts.2

Our results may also be helpful in investigating other related research issues, such

as the value relevance of accounting numbers across countries Prior research has focused primarily on estimating the relation between earnings and stock prices to understand investors’ demand for accounting earnings (e.g., Ball, Kothari, and Robin 2000; Ali and Hwang 2000) We extend this literature by examining

whether the relation between analyst characteristics and relative forecast accuracydiffers across legal origins consistent with investors’ demand for earnings

information

Consistent with expectations, we find that the relation between analyst characteristics and relative forecast accuracy is stronger in common law countries.These results are consistent with analysts’ forecast behavior responding to the demand by investors for earnings information In common law countries where investor protection laws are stronger, financial reporting is higher-quality, and the demand by investors for earnings information is greater, analysts with superior abilities tend to distinguish themselves more clearly In civil law countries, it is

the relation is strongest in the United States, followed by non-U.S common law countries The relation is weakest in the civil law countries Results within the three origins of the civil law classification (French, German, and Scandinavian) suggest that the quality of financial reporting systems plays a role in these

relations beyond the influence of investor protection laws Finally, we find some empirical support for the notion that cash flow forecasts may substitute for earnings forecasts when earnings are less relevant (Defond and Hung 2002) The relation between analysts’ characteristics and relative cash flow forecast accuracy

is stronger in civil law countries than in common law countries

The remainder of the paper is organized as follows Section 2 develops thehypotheses Section 3 outlines the research design and section 4 details the data and sample selection Section 5 reports results and section 6 provides additional analyses The paper concludes in section 7

2 Hypotheses

We provide the following rationale for our tests Common law countries are generally perceived to have stronger investor protection laws (La Porta, Lopez-de-Silanes, Shleifer, and Vishny 1997, 1998, 2000a)3 and higher-quality financial reporting (Ball, Kothari, and Robin 2000).4 In these settings, earnings information can play a more prominent role in corporate governance mechanisms and therefore have greater value relevance.5 The greater value relevance of

earnings information increases investors’ demand for that information when making decisions The increased demand by investors offers proper economic incentives for analysts to compete in providing accurate forecasts of earnings Those analysts having the ability and resources to outperform other analysts will,

on average, do so because the market-based reward structure established by investor demand offers analysts fair incentives (Schipper 1991) In other words, the rewards for making accurate forecasts fairly outweigh the cost of gathering and processing information when investor protection laws are strong and the quality of the financial reporting system is good For common law countries, we expect analysts with superior characteristics (ability, effort, experience, resources, etc.) to more consistently outperform their peers, resulting in a stronger relation between analysts’ characteristics and relative forecast accuracy

In civil law countries, financial accounting systems are generally

perceived to be of lower quality in terms of their ability to reflect accurately the underlying economic activity of the firm (Ball, Kothari, and Robin 2000;

Guenther and Young 2000; Bhattacharya, Daouk, and Welker 2003; Francis, Khurana, and Pereira 2004) Financial accounting practices in civil law countries are oriented less toward serving the needs of outside investors (O’Brien 1998; Lang, Lins, and Miller 2004) and investor protection laws are weaker (La Porta, Lopez-de-Silanes, Shleifer, and Vishny 1997, 1998, 2000a) These factors likely

economic incentives of superior analysts to outperform their peers Providing a superior earnings forecast is costly and analysts with superior abilities and resources will incur the incremental costs of gathering information only when they expect to be equitably rewarded We expect that the reduction in incentives

of superior analysts to make superior forecasts will lead to a weaker relation between analyst characteristics and relative forecast performance in civil law countries (i.e., relative forecast accuracy occurs more randomly in civil law countries)

Furthermore, among the common law countries, prior studies cited in the preceding paragraphs suggest that the United States has some of the strongest investor protection laws and higher-quality financial reporting If investor

protection laws and quality of financial reporting affect the relevance of

accounting earnings to investors, one would expect the demand for earnings information by investors and the incentives of analysts to compete and provide that information accurately to be greater in the United States than in most of the non-U.S common law countries Similarly, as discussed in the preceding

paragraphs, we expect that analysts in non-U.S common law countries will have more incentive to compete and provide more accurate forecasts relative to their peers than will analysts in civil law countries

Overall, the preceding ideas lead to our first three hypotheses

H1: Analyst characteristics better explain relative forecast accuracy in the

common law countries than in civil law countries

H2: Analyst characteristics better explain relative forecast accuracy in the

United States than in non-U.S common law countries

H3: Analyst characteristics better explain relative forecast accuracy in

non-U.S common law countries than in civil law countries

It is also interesting to consider whether the strength of investor protectionlaws or the quality of the financial reporting system offers the greater motivation

to analysts to provide superior forecasts We examine whether analyst

characteristics are useful for explaining relative forecast accuracy across three groups of civil law countries Within the civil law origin, La Porta, Lopez-de-Silanes, Shleifer, and Vishny (1997, 1998) find that countries of the French origin have weaker investor protection laws than do countries of the German origin However, countries of the French origin have higher quality and more transparent financial accounting information than do countries of the German origin (Ball, Kothari, and Robin 2000; Francis, Khurana, and Pereira 2004) Thus, the

incentives for analysts to provide superior forecasts might be stronger in the German origin countries because of better investor protection laws or stronger in the French origin countries because of higher-quality financial reporting By estimating the ability of analyst characteristics to explain relative forecast

accuracy in the French versus German origins, we expect to obtain some

indication of the impact that investor protection laws, on the one hand, versus quality of financial reporting, on the other hand, has on the behavior of analysts.

Relative to other civil law origin countries, the Scandinavian origin has thebetter investor protection laws (La Porta, Lopez-de-Silanes, Shleifer, and Vishny

1997, 1998) and the higher-quality financial reporting (Ball, Kothari, and Robin 2000; Francis, Khurana, and Pereira 2004) We therefore predict that analyst characteristics will have greater explanatory power for this group of civil law countries Thus,

H4: Analyst characteristics differentially explain the relative forecast

accuracy across the three civil law origins

Table 1 summarizes the strength of investor protection laws and the quality and transparency of financial reporting and their expected impact on the relative performance of analysts across legal origins

3 Research design and determinants of relative forecast accuracy

Extending the model of Jacob, Lys, and Neale (1999), we examine the impact of analyst activity, experience, portfolio complexity, specialization, and internal environmental factors on the ability of analysts to produce superior forecasts of earnings relative to their peers One possible limitation of using their model is that it was developed in a U.S context While there could be other important analyst characteristics in other countries, we believe the United States

provides a good setting for establishing a benchmark model of the way in which analyst characteristics explain relative forecast accuracy when the demand for earnings information is high.6

We estimate the following model, where the first ten variables are those used in Jacob, Lys, and Neale (1999) and the final three represent additional international attributes of analysts and their brokerage firms

(AFEk,j,t/MAFEj,t)-1 = α0 + β1*HORIZk,j,t + β2*CHANGEk,j,t + β3*EXPk,j,t +

β4*COMPk,j,t

+ β5*SPECk,j,t + β6*FREQk,j,t + β7*B-SIZEk,j,t + β8*B-INDk,j,t + β9*PINk,j,t +

β10*POUTk,j,t + β11*C-EXPk,j,t + β12*C-SPECk,j,t + β13*B-Ck,j,t + εk,j,t

The dependent variable measures the relative forecast accuracy of analyst

k to all other analysts following company j in year t AFE is the absolute value of analyst k’s forecast error and MAFE is the mean absolute forecast error of all analysts issuing a forecast for company j in year t.7

The independent variables are defined as follows:

HORIZ = The number of calendar days between the forecast issue date and the

earnings announcement date

CHANGE = Dummy variable that takes a value of 1 (0 otherwise) when there has

been a change in the assignment of specific analyst k following company jfor a particular brokerage in year t.8

EXP = The natural log of the number of years analyst k has issued forecasts for

company j

COMP = The number of companies followed by analyst k in the calendar year in

which the forecast was issued

SPEC = Percentage of companies followed by analyst k with the same I/B/E/S

industry code as company j

FREQ = Number of forecasts issued by analyst k for company j in year t

B-SIZE = Percentile ranking of the total number of analysts employed by the

brokerage house to which analyst k belongs in the calendar year in which the forecast was issued, relative to other brokerage houses

B-IND = Percentage of analyst k’s brokerage house analysts which follows

company j’s industry in the calendar year in which the forecast was issued.PIN = Portion of new analysts that come from outside the brokerage house

relative to the total number of analysts who worked for analysts k’s brokerage house during the calendar year in which the forecast was issued.POUT = Portion of analysts who left analyst k’s brokerage house relative to the

total number of analysts who worked for analysts k’s brokerage house during the calendar year in which the forecast was issued

C-EXP = Dummy variable that takes a value of 1 (0 otherwise) when analyst k

has issued forecasts for more than three years for any company in a country.9

C-SPEC = Percentage of companies followed by analyst k in the same country

where the analyst has issued forecasts for company j in year t

B-C = Percentage of analyst k’s brokerage house analysts which follow company

j’s country in the calendar year in which the forecast was issued

The variables HORIZ and FREQ represent analyst activity, EXP and EXP stand for experience, COMP represents portfolio complexity, while SPEC and C-SPEC correspond to specialization Finally, internal environmental factors include B-SIZE, B-IND, B-C, PIN, and POUT Consistent with prior research, wesubtract the mean of the independent variable for each year for the empirical tests (Clement 1999; Jacob, Lys, and Neale 1999) For the common lawsample, we expect HORIZ, COMP, PIN, and POUT to have a positive relation with relative forecast error and CHANGE, EXP, SPEC, FREQ, B-SIZE, B-IND, C-EXP, C-SPEC, and B-C to have a negative relation We expect the significance of the relations to be higher for the U.S sample than for the non-U.S common law

C-sample For our civil law sample, we predict that the significance of the relations will be further reduced or become insignificant.

4 Data and sample selection

The data are obtained from I/B/E/S for the period 1984-2001 (for

companies with fiscal year-end between January 1984 and December 2000) We

use the International edition and U.S edition of the I/B/E/S Detail History files.10

Summary statistics for the final sample used in our study are reported in Table 2 The results are aggregated for 12 common law countries and 21 civil law

countries We base our aggregation of the results on legal origins (La Porta, Lopez-de-Silanes, Shleifer, and Vishny 1997, 1998, 2000a) The common law countries are further separated into the United States and non-U.S common law countries.11 We further classify the civil law countries into French origin (11), German origin (6), and Scandinavian origin (4)

Including only the analyst’s most recent annual forecast for each

company-year, the database includes 1,038,329 annual earnings forecasts issued

by 30,966 analysts for 27,379 companies We first exclude observations for countries not included in our data, that are team forecasts, and that do not have actual annual earnings available This reduces the sample to 28,738 analysts in 1,321 brokerage firms who provide 1,012,189 company-year forecasts for 25,933

companies in the 33 countries We further exclude observations where there are less than three analysts following the company in that year.12

Panel A of Table 2 reports summary statistics for the final sample The final sample consists of 673,817 annual forecasts for 15,220 companies, issued by27,082 analysts who work for 1,151 brokerage houses The majority of forecasts (79%) were for companies in common law countries.13 The average number of analyst following per firm-year is greatest in the U.S (10.55), followed by non-U.S common law countries (8.65) and civil law countries (8.23).14 Panel B of Table 2 shows additional descriptive statistics regarding firm characteristics for each of the legal regimes For our sample, we report that civil law companies are,

on average, larger than common law companies This is important because an overweighting of larger firms in the common law sample could bias results in favor of our hypotheses The enhanced information environment of larger firms increases demand for their securities (Merton 1987), offering greater economic incentives to superior analysts to outperform their peers In this case, civil law countries (i.e., larger firms in our study) rather than common law countries would

be expected to exhibit a stronger association between relative forecast accuracy and analyst characteristics, absent any effects of the legal origin (e.g., investor protection laws, quality of accounting) We also show that the common law firms have higher earnings to price ratio and greater analyst forecast errors than the civillaw firms Forecast dispersion among analysts is approximately the same across

origins We further report descriptive statistics for U.S versus non-US companiesand for companies in the three civil law origins

5 Results

Univariate statistics

Table 3 shows the means and medians for variables representing analyst characteristics for each legal origin The reported unadjusted amounts provide descriptive statistics for the independent variables and comparisons between (1) common law and civil law countries, (2) the U.S and non-U.S common law countries, (3) non-U.S common law countries and civil law countries, and (4) pairs of the three civil law origin countries.15 In general, the common law analystsprovide significantly shorter forecast horizons, present fewer changes in analysts following a company, have more experience, follow fewer companies, specialize more in the same industry, and provide more forecasts for each company than do the analysts in the civil law countries They are employed in larger brokerage houses with larger percentages of analysts following companies in the same industry, and these houses have smaller portions of new and outgoing analysts than do civil law analysts Finally, common law analysts have more country-specific experience, greater specialization in a particular country, and greater brokerage house specialization within a country.16

Further analyses of the descriptive statistics show significant differences between all characteristics of analysts in the United States and analysts in non-U.S common law countries For example, U.S analysts have more experience and follow more companies in the same industry, but they provide fewer forecastsfor each company than analysts in the non-U.S common law countries Similarly, all characteristics of analysts in the non-U.S common law countries differ

significantly from those of the civil law countries For instance, analysts in the civil law countries are less experienced, have less country-specific experience, provide less forecasts for each company, and have a greater forecast horizon Finally, the comparisons between the three civil law origin countries show

significant t-statistics and Wilcoxon Z-statistics for all of the 13 independent variables

Test of hypothesis 1

Table 4 presents OLS results for the common law countries and the civil law countries.17 Eleven coefficients are statistically significant in the predicted direction for the common law sample, but only six coefficients are significant in the predicted direction for the civil law sample The coefficients for the forecast horizon (HORIZ) and the analyst-broker turnover variables (PIN and POUT) are positive (as predicted) and they are significant in the common law and the civil law countries The estimated coefficient for portfolio complexity (COMP) is

positive and statistically significant for the common law countries, suggesting thatthe larger number of companies followed by an analyst reduces forecast accuracy This coefficient is not significant for the civil law countries The coefficients for frequency (FREQ) are significant and negative (as predicted) for both origins Thecoefficients for analyst specialization (SPEC), broker size (B-SIZE), and broker-industry specialization (B-IND) are significantly negative (as predicted) for the common law countries, while the coefficients for B-SIZE and SPEC are not significant (at p <0.05) and B-IND is significant in the unpredicted direction for the civil law countries These results suggest that internal environment factors in brokerage houses in civil law countries do not have the predicted impact on analysts’ relative forecast accuracy as they do in common law countries The regression coefficients for experience (EXP) are not statistically significant acrossthe two groups of countries.

The coefficients for the three international variables are statistically

significant for the common law countries The coefficients for C-SPEC and B-C are significantly negative (as predicted) for both origins The coefficient for C-EXP, however, is significantly positive (not negative as predicted) for the

common law origin and insignificant for the civil law origin This result indicates that analysts in common law countries may have reasons to extend their

forecasting activity to other countries (e.g., they may possess private knowledge)

The R2 for common law countries is 0.1491 and the R2 for civil law countries is 0.0811, suggesting that analyst characteristics better explain relative forecast accuracy in the common law countries than in civil law countries The Chow test (F = 237.6) suggests that model coefficients are significantly different across origins.18The t-statistics for tests of differences in individual coefficients are also reported in Table 4.19 The results show that portfolio complexity,

specialization, analyst activity, and internal environment factors, but not

experience, more significantly explain (in the predicted directions) analysts’ relative forecast accuracy in the common law countries than in the civil law countries Eight of the estimated coefficients in the common law countries have significantly greater magnitudes, in the predicted directions, than do the

corresponding coefficients in the civil law countries For example, relative

forecast accuracy increases by 0.37 percent per day as the forecast age decreases

in the common law countries, which is significantly faster (t = 36.3) than the 0.23 percent increase per day in the civil law countries, controlling for other variables

in the model This result suggests that common law analysts provide more effort and compete more aggressively in utilizing recent information than do civil law analysts The relative forecast accuracy increases by 3.97 percent for each

additional forecast issued by analysts (FREQ) in the common law countries, which is significantly greater (t = 4.97) than the 1.56 percent increase in relative forecast accuracy in the civil law countries The differences in the coefficients are

not significant for the three international variables This result suggests that domestic characteristics have a greater impact on the relative forecast accuracy in the common law countries than in the civil law countries, but the international characteristics provide similar effects on the relative forecast accuracy in both origins.

In sum, the empirical evidence supports our first hypothesis that analyst characteristics better explain the relative forecast accuracy in common law

countries than in civil law countries It seems that analysts in common law

countries compete to provide superior forecasts more than do analysts in civil law countries, and those analysts with superior abilities and resources tend to

outperform their peers

Test of hypothesis 2

Table 4 also reports the results for comparing the regression model betweenthe United States and non-U.S common law countries In the United States, ten coefficients are statistically significant in the predicted direction One coefficient (C-EXP) is significant in the unpredicted direction.20 Eight coefficients are

significant in the predicted directions for the non-U.S common law countries The coefficient for COMP is significant in the unpredicted direction

The R2s are 0.1494 for the United States and 0.1271 for the non-U.S

significant (F = 75.5).21 Furthermore, of the ten coefficients significant in the predicted direction for the United States, eight have a significantly greater

magnitude than the corresponding coefficients for the non-U.S common law sample For example, the magnitude of the coefficient for B-SIZE is three times greater for the U.S sample than for the non-U.S common law sample (t = –4.70).The coefficients for international specialization (C-SPEC) and broker-country factor (B-C) are significantly negative (as predicted) The coefficient for B-C is significantly more negative for the United States than for the non-U.S common law countries (t = –4.09) The coefficients for C-SPEC, however, are notsignificantly different The coefficient for country-specific experience (C-EXP) is significantly positive (not negative as predicted) for the United States, and it is notsignificantly different from the coefficient for the non-U.S common law

countries In sum, our results support the second hypothesis In non-U.S commonlaw countries, where investor protection laws and financial reporting are

somewhat weaker than in the United States, we find less evidence that superior analysts distinguish themselves

Test of hypothesis 3

On the final right column of Table 4, we report results for comparing the regression model between the non-U.S common law countries and the civil law countries The R2 for the non-U.S common law countries (0.1271) is greater than

that of the civil law countries (0.0811) The significant Chow test (F = 76.0) suggests that model coefficients are not equal across origins.

Four of the variables have a significantly greater magnitude with the predicted signs for the non-U.S common law countries than for the civil law countries The estimated coefficient for B-C is the only one that is more

significant in the civil law countries Overall, the results support the third

hypothesis, indicating that analyst characteristics have less ability to distinguish superior analysts in the civil law countries than in the non-U.S common law countries

The results in Table 4 support the first three hypotheses We conclude that analyst characteristics better explain relative forecast accuracy in countries with stronger investor protection laws and higher-quality financial reporting In the United States, where investor protection laws and financial reporting are generallyconsidered the strongest, we find the most evidence that analyst characteristics distinguish superior analysts Furthermore, the evidence is stronger for non-U.S common law countries than for civil law countries

Test of hypothesis 4

Analyses within the civil law countries provide an interesting setting to test the relative influence of investor protection laws and the quality of financial

have weaker investor protection laws (La Porta, Lopez-de-Silanes, Shleifer, and Vishny 1997) but higher-quality financial reporting (Ball, Kothari, and Robin 2000; Francis, Khurana, and Pereira 2004) Which environment provides superior analysts with stronger incentives to distinguish themselves? If analyst forecast behavior is related more to quality of financial reporting (strength of investor protection laws), we would then expect superior analysts to distinguish

themselves better in French (German) origin countries Therefore, tests within French and German origin classifications provide some insights into the relative influence investor protection laws and quality of financial reporting can have on investors’ demand for earnings information and analysts’ forecast activities Since Scandinavian countries have stronger investor protection laws and higher-quality financial reporting systems (relative to other civil law countries), we expect this group to provide the strongest evidence of analysts’ differential forecast ability.Table 5 reports regression results by the three civil law origins The R2s are 0.0980, 0.0881, and 0.0711, respectively, for countries of the Scandinavian, French, and Germanic origins Chow tests are significant, suggesting differences

in model coefficients across origins Six (five) regression coefficients are

significant for French (German) origin countries in the predicted direction.22 Of the six significant coefficients for the French sample, three have a significantly greater magnitude in the predicted direction than those for the German sample For example, the estimated positive coefficient for HORIZ in the French origin