Do the Market Analysts Earnings Forecast Errors Matter with Earnings Management in the U.S. Banking Industry?

Transcription

1 Min-Lee Chan Kai-Li Wang & Pin-Shiuan Chen o the Market Analysts Earnings Forecast Errors Matter with Earnings Management in the U.S. Banking Industry? (Received Sep ; First Revision Jan ; Second Revision Feb ; Accepted Mar ) Research Motivation A number of past studies focus on the relationship between earnings management and meeting or beating analysts forecasts. As Payne and Robb (2000) put it in the Wall Street Journal: While the problem of earnings manipulation is not new it has swelled in a market that is unforgiving of companies that miss their estimates.. that failed to meet its so-called numbers by one penny and lost more than six percent of its stock value in one day. The evidence that stock markets will negatively respond to news of those firms unable to meet analysts forecasts is widely supported (Brown 1993; Skinner and Sloan 2002; Kinney Burgstahler and Martin 2002). Based on this firms have motivations to manage earnings to meet analysts forecasts to avoid market penalties (Burgstahler and Eames 2006; egeorge Patel and Zeckhauser 1999; Payne and Robb 2000; Robb 1998). These studies however have their limitations based on the following points of view. Firstly these studies make no clear distinction between ex ante and ex post reported earnings that deviate from the analysts forecasts in examining the issue (Burgstahler and Eames 2006; reman and Berry 1995; Brown 1993; Payne and Robb 2000; Matsumoto 2002). As a matter of fact observing the degree to which ex ante earnings deviate from market expectations is the real reason for inducing earnings management. Then the result of beating market expectations (i.e. posting reported earnings as being equal to or higher than analysts forecasts) should be achieved as a result. Secondly most research on the issue of beating market expectations as an incentive to motivate firms to engage in earnings management mainly focuses on non-financial industries. We argue that this phenomenon will also exist in the banking industry a regulated industry. Accordingly by using data for 107 U.S. commercial banks during the period 2001 to 2004 we adopt a logit model to examine whether banks have incentives to manage earnings upward to meet market analysts earnings forecasts. Min-Lee Chan is Associate Professor of epartment of Finance Providence University. chanml@pu.edu.tw Kai-Li Wang is Professor of epartment of Finance Tunghai University. Pin-Shiuan Chen completed her master degree at epartment of Finance Providence University. We finally conclude that meeting market expectations may not be the major motivation for managing earnings in the U.S. banking industry. In addition we suggest that for those studies that examine the issue of beating market expectations as an incentive to manage earnings there is a need to be concerned about the problem of ex ante and ex post reported earnings in comparison with market forecasts. Merely examining the issue without distinguishing the ex ante and ex post earnings might not lead to sufficient caution being exercised in drawing conclusions. Literature Review Previous studies support the avoidance of negative earnings disappointments as the motivation underlying earnings management. Some research finds that there is an unusually low frequency of firms with negative small earnings surprises and a high frequency of firms with small positive earnings surprises (Burgstahler and Eames 2006; reman and Berry 1995; Brown 1993). Furthermore those firms reporting small positive earnings surprises tend to have significantly greater discretionary accruals (Payne and Robb 2000; Matsumoto 2002). Besides the firms that beat market expectations tend to manage earnings upward (Burgstahler and Eames 2006; Lee 2007). As for the direction of earnings management this has different implications for upward and downward management. Firms manage earnings upward when they slightly beat market expectations (Burgstahler and Eames 2006; Matsumoto 2002) or when real earnings are below and near the market forecasts (Kasznik 1999; Payne and Robb 2000; Lee 2007; Burgstahler and Eames 2006). In turn firms choose to manage earnings downward i.e. they adopt a big bath reporting strategy in periods when earnings are low and the market forecasts are more dispersed regarding firms expected earnings (Payne and Robb 2000). There are only a few studies that investigate banks earnings management in line with market expectations. Robb (1998) focuses on the effect of forecast dispersion on banks earnings management and finds that the management of earnings becomes most significant when market forecasts reach a high degree of consensus. We choose the banking industry Management Review Vol. 29 (Oct 2010)

2 as our research sample to examine the issue of beating market expectations as the motivation underlying earnings management. The model used to estimate earnings management in the banking industry differs from that for non-financial firms. That is the widely-accepted accounting items used to estimate the management of earnings are loan loss provisions( LLP ) and security purchases/sales and most studies find that the former lead to better estimation results (Robb 1998). To be specific explanatory variables are used in the OLS model to explain the nondiscretionary component of loan loss reserves and the predicted residual term is calculated as the estimate of the discretionary component of loan loss reserves as a proxy for earnings management. The positive value of the predicted residual implies downward management of earnings because LLP is expensed at a higher rate than the expected LLP necessary for nondiscretionary parts. By contrast the upward management of earnings is inferred by the negative value of the predicted residual terms. Research esign Establishment of Hypotheses According to our literature review firms will significantly present greater discretionary accruals to manage earnings upward when their actual earnings are below market forecasts (Payne and Robb 2000; Kasznik 1999; Lee 2007; Burgstahler and Eames 2006). Thus we propose our first hypothesis as follows: H1: Banks tend to manage earnings upward when they observe that their first half-year earnings deviate significantly from the average forecasts of those market analysts. To measure the gap between the firm s ex ante premanaged earnings and market forecasts we use first half-year earnings as a proxy for actual earnings in order to truly measure the gap. Then we argue that the motivation for meeting/beating market expectations for earnings management can not be confirmed until we are able to provide evidence that the forecast error after upward earnings management should be decreased or is more likely to beat market expectations. Thus we propose our second hypothesis as follows: H2: Other things being equal the probability of meeting/beating market forecasts will increase after upward earnings management. Meeting/beating market forecasts is measured by whether the market average earnings forecast of analysts is greater or less than reported annual earnings. Both hypotheses are intended to examine whether the purpose of banks upward earnings management is to avoid earnings disappointment. Research Sample and Source of ata We use 107 U.S. banks covering the period as our research sample. The bank information such as the loan loss reserves loan loss provisions charge-offs and capital adequacy ratios is collected from OSIRIS and the analysts earnings forecasting information including EPS forecasts and the dispersion of EPS among those analysts is obtained from I/B/E/S. The other information comes from Compustat. Empirical Model We follow Robb (1998) to estimate the earnings management for banks. Our model is as follows: LLPit 0 1 LLRit 1 2 WOit 3 WOit 1 it (1) where LLP it represents the ratio of loan loss provisions to total loans for bank i in period t LLP it 1 is the ratio of loan loss reserves to total loans in the prior period t-1 and WO it and WO it 1are the ratios of bank charge-offs to total loans in period t and t+1 respectively. We expect the sign of the coefficient estimate to be negative and the signs of and to be positive. The predicted residual terms are calculated by the actual value of LLP it minus predicted LLP it which we use as a proxy for earnings management. The positive value of the predicted residual terms implies downward earnings management since the actual LLP it is expensed more than the expected LLP it necessary for the nondiscretionary component leading to a reduction in earnings. Conversely a negative value for the predicted residual terms implies upward earnings management. Next we will examine the first hypothesis using the following Logit model. EM FE _ up EPS S it 0 1 it 2 it 3 it Size Capital LLR 4 it 5 it 6 it Non _ perf Int 7 it 8 it it (2) where EM it represents a binary variable which is coded 1 for upward earnings management and 0 otherwise FE _ up it represents the gap between the first half-year earnings and market average forecasts EPS it is the earnings per share of bank i in period t S it is the dispersion of EPS it forecasts among analysts Size it is the natural log of total assets Capital it is the bank capital adequacy ratio LLR it refers to the loan loss reserves Non _ perf it refers to the non-performing loans of banks and Int it is the interaction term between FE _ up it and S it. We expect that the greater the deviation of the bank s first half-year earnings from the average forecast of EPS it the more likely it is that the bank will manage earnings upward and a negative sign of the estimate γ1 is predicted (we use the SAS software package to estimate the model and the coefficient is used to explain the likelihood of EM it 0 ). 148 Management Review October 2010

3 Table 1 escriptive Statistics Sample size = 226 Variable Name Mean Standard eviation Minimum Maximum LLP LLR WO EM FE_ up EPS S Size Capital Non _ perf FE EPS _ 5yr EPS _5yrsd Period Notes: LLP represents the ratio of the loan loss provision to total loan LLR is the ratio of loan loss reserves to total loans and WO are the ratios of the bank charge-offs to total loans respectively. EM is the predicted value of the residual term FE _ up represents the gap between the first half-year earnings and market average forecasts EPS is the earnings per share S is the dispersion of EPS forecasts among analysts Size is the natural log of total assets Capital is the bank capital adequacy ratio Non _ perf refers to the non-performing loans of banks FE is the market average of EPS forecasts minus reported earnings EPS _ 5yr is the growth rate of EPS for the past five years EPS _ 5yrsd is the variability of EPS _ 5yr and Period is the forecast period ranging from 1 to 3 years. Variable Name Table 2 Estimation of Earnings Management Coefficient Estimates Intercept (< ) *** LLRit (< ) *** WO t (< ) *** WOt (< ) *** N = 226 R-Square = Adj R-Sq = ependent variable LLR represents the ratio of loan loss provision to total loans for bank LLRit 1 is the ratio of loan loss reserves to total loans in the prior period t-1 and WO it and WOit 1 are the ratios of bank charge-offs to total loans in periods t and t+1 respectively. Notes: 1. The values in the parentheses ( ) are the p values of the t statistic. 2. *** ** and * represent significance at the 10% 5% and 1% levels respectively. Whether the bank s upward earnings management is motivated by avoiding the negative earnings surprise we should further observe that the likelihood of beating the market expectations will increase after managing earnings upward. We now present our second hypothesis using logit analysis in the following model: FEit 0 1 EMit 2 EPSit 3 Sit Size EPS _5 yr EPS _5yrsd 4 it 5 it 6 it 7 Periodit it (3) where FE it is a binary variable that is coded 1 when the analysts forecasted yearly EPS is greater than the firm s reported yearly EPS and 0 when reported earnings beat market forecasts. EM EPS it S it and Size it are defined as in model (2). EPS _5yr it is the growth rate of EPS for the past five years EPS _5yrsd it is the variability of EPS _5yr it and Period is the forecast period ranging from 1 to 3 years. it Bank s Earnings Management Analysts Forecasts 149

4 Table 3 Logit Analysis of Earnings Management Variable Name Parameter Estimation Intercept (0.9949) FE_ up (0.0131)** EPS (0.0346)** S (0.7290) Size (0.7315) Capital (0.7895) LLR (0.0868)* Non _ perf (0.1839) Int (0.7261) Likelihood ratio: Chisquare= df=8 Pr>Chisquare=0.016 Wald: Chisquare= df=8 Pr>Chisquare= ependent variable EM represents a binary variable which is coded 1 for upward earnings management 0 otherwise FE _ up represents the gap between first half-year earnings and market average forecasts EPS is the earnings per share of bank S is the dispersion of EPS forecasts among analysts Size is the natural log of total assets Capital is the bank capital adequacy ratio LLR is the loan loss reserves Non _ perf refers to the non-performing loans of banks and Int is the interaction term between FE _ up and S. Notes: 1. The values in the parentheses ( ) are the p values of the Chi square statistic. 2. *** ** and * represent significance at the 10% 5% and 1% levels respectively. Table 4 Logit analysis of forecast errors Variable name Parameter estimation Intercept (<0.0001) *** EM (<0.0039) *** EPS (<0.0001) *** S (<0.0055) *** Size (<0.0005) *** EPS _ 5yr (<0.4591) EPS _5yrsd (<0.2005) Period (<0.0295) *** Likelihood ratio: Chisquare=55.59 df=7 [Pr>Chisquare]< Wald: Chisquare=44.25 df=7 [Pr>Chisquare]< ependent variable FE is a binary variable which is coded 1 when the analysts forecast of yearly EPS is greater than the firm s reported yearly EPS and 0 when reported earnings beat market forecasts. EM EPS is the earnings per share of bank S is the dispersion of EPS forecasts among analysts and Size is the natural log of total assets. EPS _ 5yr is the growth rate of EPS for the past five years EPS _5yrsd is the variability of EPS _ 5yr and Period is the forecast period ranging from 1 to 3 years. Notes: 1. The values in the parentheses ( ) are the p values of the Chi square statistic. 2. *** ** and * represent significance at the 10% 5% and 1% levels respectively. We anticipate that the likelihood of beating the market forecast will increase after upward earnings management that is a positive coefficient of β1is expected. If β1 is shown to be significantly positive we can conclude that the bank s upward earnings management will be to avoid the negative earnings surprise. Empirical Results Empirical Results of Earnings Management Table 1 one table 2 demonstrate the descriptive statistics for each variable and the estimation result of earnings management. As shown in Table 3 we find that banks are more likely to manage earnings significantly upward when ex ante premanaged earnings have greater deviations from the market average forecast of EPS. Our results support hypothesis H1 indicating a motivation for meeting/beating market expectations for upward earnings management a result that is consistent with past studies (Kasznik 1999; Lee 2007; Burgstahler and Eames 2006). However this motivation can not be strongly confirmed until there is evidence that the likelihood of beating the market forecast after upward earnings management will increase. Thus we further test our second hypothesis H2 in the next section. 150 Management Review October 2010

5 Empirical Results of Forecast Error As shown in Table 4 we find that banks are unlikely to beat market expectations after managing earnings upward a finding inconsistent with H2. This implies that beating market expectations as a motivation for upward earnings management is not strongly supported. We suggest that past studies that support this motivation only through the first hypothesis H1 need to examine the issue more carefully before jumping to conclusions. The other results show the positive influence of bank profitability and the negative influence of both forecast dispersion and bank size in beating the market forecast. Conclusion We examine whether beating market forecasts has any involvement in upward earnings management for the U.S. banking industry between 2001 and We contribute to the literature by documenting that the premanaged (ex ante) earnings or ex post earnings after management should be distinguished in investigating the relationship between earnings management and beating market forecasts. The past literature supports the view that avoiding negative earnings surprises will motivate firms to manage earnings upward (Kasznik 1999; Payne and Robb 2000; Lee 2007; Burgstahler and Eames 2006). However this conclusion needs to be interpreted with caution in distinguishing the ex ante earnings from the ex post earnings after management in measuring the forecast error. We do find however that the larger gap between premanaged earnings and the market forecast will make banks more likely to manage earnings upward although we can not obtain evidence that banks will be more likely to beat the market forecast after managing earnings upward. Accordingly we should not jump to conclusions that beating the market forecast or the avoidance of negative earnings disappointment will motivate banks to engage in earnings management. Kinney William avid Burgstahler and Roger Martin (2002) Earnings Surprise Materiality as Measured by Stock Returns Journal of Accounting Research 40(5) Lee Jimmy (2007) Earnings Management to Just Meet Analysts Forecast Working paper Kellogg Graduate School of Management Northwestern University. Matsumoto awn (2002) Management s Incentives to Avoid Negative Earnings Surprises The Accounting Review 77(3) Payne Jeff L. and Sean Robb (2000) Earnings Management: The Effect of Ex Ante Earnings Expectations Journal of Accounting Auditing & Finance 15(4) Robb Sean W. G. (1998) The Effect of Analysts Forecasts on Earnings Management in Financial Institutions The Journal of Financial Research 21(3) Skinner ouglas J. and Richard G. Sloan (2002) Earnings Surprises Growth Expectations and Stock Returns or on t Let an Earnings Torpedo Sink Your Portfolio Review of Accounting Studies 7(2-3) REFERENCES Brown Lawrence. (1993) Earnings Forecasting Research: Its Implications for Capital Markets Research International Journal of Forecasting Burgstahler avid and Michael Eames (2006) Management of Earnings and Analysts Forecasts to Achieve Zero and Small Positive Earnings Surprises Journal of Business Finance and Accounting egeorge Francois Jayendu Patel and Richard Zeckhauser (1999) Earnings Management to Exceed Thresholds Journal of Business 72(1) reman avid N. and Michael A. Berry (1995) Analyst Forecasting Errors and Their Implications for Security Analysis Financial Analysts Journal 51(3) Kasznik Ron (1999) On the Association between Voluntary isclosure and Earnings Management Journal of Accounting Research 37(1) Bank s Earnings Management Analysts Forecasts 151

6 152 Management Review October 2010