Managerial Horizons, Accounting Choices and Informativeness of Earnings

Transcription

1 Managerial Horizons, Accounting Choices and Informativeness of Earnings by Albert L. Nagy University of Tennessee (423) Kathleen Blackburn Norris University of Tennessee Richard A. Riley, Jr. West Virginia University February 5, 1999 We would like to thank Susan Ayers, Bruce Behn, Joe Carcello, Terry Neal, Terry Warfield and especially workshop participants at the University of Alabama for helpful comments on previous versions of this paper. Please send all correspondence to: Albert L. Nagy, Department of Accounting and Business Law, The University of Tennessee, 637 Stokely Management Center, Knoxville, TN Phone #: (423) Fax #: (423)

2 Managerial Horizons, Accounting Choices and Informativeness of Earnings Abstract. The purpose of this study is to examine the ability of managerial stock ownership relative to annual compensation to improve the alignment of managers and outside shareholders interests among low managerial ownership companies. Exploiting the theory of the firm and separation of ownership and control, managerial stock ownership is expected to align the interests of managers and shareholders, providing managers with a long-term horizon. In contrast, annual compensation, salary and bonus, tied to current-period company performance, provide managers with a short-term horizon. Prior research has demonstrated that in companies with low managerial ownership managers exploit the latitude available in generally accepted accounting principles to alleviate contractual constraints, presumably to ensure job preservation (annual salary) and maximize incentive compensation (annual bonus). The result is accounting information that is less informative in low managerial ownership companies compared to companies with higher levels of managerial ownership. We hypothesize that in companies with low managerial ownership, higher levels of stock ownership relative to annual salary and bonus mitigate managerial incentives to exploit accounting choices. Our findings are consistent with this hypothesis and suggest that greater managerial wealth tied to the long-term performance of the company through stock ownership versus short-term performance horizons (annual salary and bonus) results in more informative accounting earnings and lower discretionary accounting accrual adjustments in low managerial ownership companies. These results are robust to numerous sensitivity tests. 1

3 Managerial Horizons, Accounting Choices and Informativeness of Earnings Introduction Warfield, Wild and Wild (1995) provide evidence that the level of managerial stock ownership is positively related to the informativeness of earnings released by management and negatively associated with the magnitude of discretionary accrual adjustments. The authors conclude that their results suggest that dispersed ownership firms construct contracts to constrain and/or monitor management behavior and to the extent that these contracts are based on accounting numbers, managers exploit discretionary accruals to alleviate contract constraints and maximize incentive benefits. Because shareholders relevant horizon is in principle unlimited (Lewellen, Loderer & Martin, 1987), one implication of Warfield, Wild and Wild (1995) is that high levels of managerial stock ownership provide managers with a long-term horizon. In contrast, managers of low managerial ownership companies have a short-term horizon where managers exploit accounting choices to alleviate accounting-based contractual constraints, presumably to ensure job preservation (annual salary) and maximize incentive compensation (annual bonus). In order to reach these conclusions, the authors implicitly assume that managerial horizons other than those tied to long-term performance through stock ownership are constant across sample firms. The purpose of this study is to examine the ability of managerial stock ownership relative to annual compensation to improve the alignment of managers and outside shareholders interests among low managerial ownership companies. 1 Specifically, managerial stock ownership relative to annual compensation is measured by the ratio of the 2

4 market value of stock owned by top managers to their annual compensation (hereafter referred to as the stock/compensation ratio). This ratio reflects the amount of managers wealth tied to the long-term performance of the firm versus managerial incentives tied to short-term objectives such as job preservation (e.g., annual salary) and incentive compensation (e.g., annual bonus). 2 Anecdotal evidence from company proxy statements suggests that some companies encourage minimum levels of this ratio for their executive officers. Based on these arguments, we hypothesize that in low managerial ownership companies greater managerial wealth tied to the long-term performance of the company through stock ownership versus short-term company performance tied to annual salary and bonus results in a closer alignment of managers and outside stockholders interests. Closer alignment of the interests of managers and shareholders is expected to result in more informative accounting earnings and lower discretionary accounting accrual adjustments because managers of high stock/compensation ratio companies will be less inclined to exploit discretionary accruals to alleviate contract constraints and maximize incentive benefits. To evaluate our hypotheses we sample low managerial ownership companies. From this sample, we examine (1) the effect of the stock/compensation ratio on the returns-earnings relation and (2) the association between the stock/compensation ratio and the absolute value of abnormal accrual adjustments. Our findings suggest that increases in the stock/compensation ratio are associated with improvements in the informativeness of earnings disclosed by management and with reductions in the magnitude of accrual adjustments employed by management in low managerial ownership companies. Based on these results, we conclude that higher levels of stock ownership 3

5 relative to annual salary and bonus for highly dispersed ownership firms help mitigate owner-manager agency problems. Further, evidence is presented that suggests that as the stock/compensation ratio increases, managers of highly dispersed ownership companies disclose earnings of similar quality to those disclosed by companies whose managers own greater levels of stock (i.e., managerial ownership levels greater than five percent). Finally, these results are robust, though less effective, when annual compensation tied to long-term horizons (through stock options and restricted stock) versus short-term horizons (annual salary and bonus) is examined. The results of this study contribute to the extant managerial compensation and earnings quality literature in several ways. First, this study provides insight into the tradeoff between long-term horizons related to managerial ownership and short-term horizons related to annual compensation in reducing owner-manager agency costs for low managerial ownership firms. Second, the results suggest that even in low managerial ownership companies, managers with sufficient wealth tied to long-term firm performance relative to their annual compensation act as if they have larger stakes in long-term company performance. As a result, their incentives to manipulate accounting numbers for short-term compensation gains are mitigated. According to prior research, the resulting improved earnings disclosures reduce information asymmetry and hence increase liquidity in equity markets (Akerlof 1970; Welker 1995). Third, the results of Warfield, Wild and Wild (1995) are replicated for a more recent period. Finally, the results may benefit investors and practitioners in evaluating the effectiveness of managerial ownership and annual compensation on aligning the disparate interests of managers and outside shareholders. By discouraging low levels of stock ownership relative to annual 4

6 compensation and emphasizing stock-based compensation, low managerial ownership companies may encourage managers to balance their interests between the future prospects of their companies and alleviating/maximizing short-term contractual conditions. The next sections are organized as follows: Section two describes the theoretical and empirical background related to agency theory, annual compensation and managerial ownership. The informativeness of earnings and the stock/compensation ratio are discussed in Section three. Discretionary accruals and the empirical models are addressed in Sections four and five, respectively. Section six includes the sample selection, data sources and descriptive statistics. Finally, the findings and conclusions are presented in Sections seven and eight. Agency Theory, Annual Compensation and Managerial Ownership In their seminal paper, Jensen and Meckling (1976) describe how the separation of ownership and control misaligns the interests of managers and owners and results in management behavior that may be detrimental to the value of the firm. 3 Rational outside equity owners, who anticipate this detrimental management behavior, deduct the costs of the expected opportunistic management behavior when determining equity price. Management is left with the onus of assuring outside equity holders that opportunistic behavior detrimental to the firm will be restrained. These assurances often result in a nexus of contracts that restrict or monitor opportunistic management behavior in an attempt to better align the interests of management and shareholders. Jensen and Meckling (1976) state that these contracts include the establishment of incentive compensation 5

7 systems that serve to more closely identify managers interests with those of outside equity holders. Management compensation contracts often incorporate accounting-based constraints that are developed to monitor or direct managerial behavior. As discussed in Warfield, Wild and Wild (1995), the incorporation of these accounting-based constraints in managerial employment contracts encourages managers to exploit the latitude available in accepted accounting procedures to minimize the impact of the constraints or allow managers to maximize their incentive compensation. Managerial discretion related to financial reporting is not eliminated because (1) monitoring and contracting are costly and (2) managers are assumed to possess a comparative informational advantage, and elimination of their discretion may result in information that is less likely to reflect the economic substance of the company. Therefore, the market expects managers to capitalize on the latitude permitted by both contracts and accepted accounting procedures in reporting accounting numbers, and managers compensation contracts incorporate this expected behavior. Consistent with agency theory, prior researchers have suggested that annual salary and bonus induce short-term performance horizons. For example, Lewellen, Loderer and Martin (1987) observe the following: By whatever formula these (salary and current bonus) payments are established, the payoff to the executive/recipient reflects only the firm s revealed performance up to the payment date; the amounts involved are fixed when awarded, and as received are independent of the firm s subsequent performance. Moreover, bonus awards are typically based on short-term performance measures such as current-year profits or returns on equity. 6

8 Thus, managerial compensation contracts based on annual performance may result in managerial myopia. 4 This problem has been referred to in Smith and Watts (1982) as the horizon problem. 5 In contrast, managerial stock ownership is expected to undo the effects of short-term compensation horizons by inducing managers to align their interests with the long-term performance of companies that employ them. 6 Investigating the horizon problem, Dechow and Sloan (1991), examine CEO s discretionary expenditures during their final years in office. Examining companies with earnings-based performance measures, the research tests the hypothesis that such measures provide executives with incentives to focus on short-term performance. The authors examined the behavior of research and development expenditures (R&D) during the CEO s final years in office. The findings suggest that executives who are in their final years as CEO spend less on R&D. Further, the authors find that the more company stock and options executives held, the less likely they were to reduce discretionary expenditures prior to their departure. Core and Guay (1998) further examine the determinants of CEO equity incentives. The authors assume that companies have target levels of equity incentives (where the CEO s stock of incentives consists of shares and options held) for their CEO. First, the authors model the optimal level of CEO equity incentives and then determine whether current year stock and cash compensation is systematically related to deviations from those optimal levels. The findings suggest that companies use new incentives to correct deviations in the CEO s stock of incentives, and that firms substitute cash and equity compensation as part of the adjustment process. Overall, the authors conclude that 7

9 companies effectively use equity incentives to reward past performance and re-optimize incentives for future performance. In summary, the prior literature suggests that dispersed ownership firms may effectively construct compensation contracts to help mitigate high agency costs, and that the individual components included in these contracts affect management horizons differently. Therefore, in order to meaningfully assess compensation influence on management behavior, the long-term horizon incentives relative to the short-term horizon incentives must be considered. The Informativeness of Earnings and the Stock/Compensation Ratio Rational managers without incentives to focus on the long-term performance of the company are expected to exploit the latitude available in accepted accounting procedures to alleviate contractual constraints in order to ensure job preservation and maximize current-period incentives. When managers use accounting information for self-interest, accounting numbers are less likely to reflect the economic substance of the company (i.e., become less informative). 7 Empirically, Warfield, Wild and Wild (1995) examine the hypothesis that the level of managerial ownership affects both (1) the informativeness of accounting earnings and (2) the magnitude of discretionary accounting adjustments. The authors examine over 1,600 corporations and nearly 5,000 annual reports from The results from their study suggest that the level of managerial ownership is positively associated with earnings explanatory power for returns, inversely related to the magnitude of accounting accrual adjustments and less important for regulated corporations. These results are consistent with economic theory that suggests low managerial ownership firms 8

10 develop monitoring and control contracts, often containing accounting-based constraints, to restrict managers value-reducing behavior. The existence of accounting-based contracts in low managerial ownership companies creates incentives for management to exploit discretionary accruals to alleviate contracts constraints or maximize incentive benefits. In turn, earnings disclosures are less informative to shareholders and include large adjustments of discretionary accruals. The results of prior analytical and empirical research suggest companies strive to achieve optimum governance mechanisms that balance managers short-term and longterm horizons. However, the results of Warfield, Wild and Wild (1995) suggest that managers of dispersed managerial ownership companies have a short-term horizon and exploit the latitude available in generally accepted accounting principles to alleviate contractual constraints, presumably to ensure job preservation (annual salary) and maximize incentive compensation (annual bonus). Warfield, Wild and Wild (1995) implicitly assume that managerial horizons other than those tied to long-term performance through stock ownership are constant across sample firms. We assert that these other managerial horizons are not constant across low managerial ownership firms, and that they should be considered when examining overall managerial horizons. We hypothesize that greater managerial wealth tied to the long-term performance of the company relative to short-term performance rewards, will better align the interest of managers and shareholders in highly dispersed managerial ownership companies. The tradeoff between long-term and short-term horizon incentives is measured by a stock/compensation ratio, which equals the market value of stock owned by top management divided by their annual salary and bonus. This ratio reflects the amount of 9

11 managers' wealth tied to the long-term value of the firm relative to managers short-term objectives: job preservation (e.g., annual salary) and incentive compensation (e.g., annual bonus). 8 Anecdotally, a recurring compensation philosophy found in a number of companies proxy statements encourages the executive team to own company stock equal in value to a minimum level of their annual compensation. 9 For example, Pfizer Inc. states in its 1995 proxy statement that executive officers are expected to own company common stock equal in value to at least three times their annual compensation. Similarly, NL Industries Inc. encourages executives to own common stock with a value between two and four times their annual compensation. Finally, General Motors encourages executive ownership of common stock equal in value to a minimum of two times their annual compensation. 10 The results of prior analytical and empirical research suggest that an optimum stock/compensation ratio exists which balances managers short-term and long-term horizons. The findings suggest that simply maximizing the stock/compensation ratio may not always be optimal. However, consistent with the findings of Warfield, Wild and Wild (1995), we argue that in low managerial ownership companies, agency costs tend to be significant and it is likely that stock/compensation ratios will fall short of optimal balances (i.e., managers will own less stock than optimal). This leads to our first hypothesis: H1: The informativeness of accounting earnings is positively associated with the stock/compensation ratio found among low managerial ownership firms. 10

12 Discretionary Accruals Compensation contracts provide managerial incentives to choose accounting techniques that are both consistent with the conditions of the contract and in the best interest of managers. To the extent these conditions are based on accounting numbers, this often results in management exploiting discretionary accruals to satisfy the contractual conditions. Since managers accounting choices are not entirely constrained, the magnitude of accounting adjustments is expected to be inversely related to managerial ownership because managers are expected to capitalize on the latitude in reported accounting numbers. 11 This leads to our second hypothesis: H2: The magnitude of adjustments in managers accounting choices is negatively associated with the stock/compensation ratio found among low managerial ownership firms. Empirical Models The models incorporated in this study are similar to those employed by Warfield, Wild and Wild (1995). Hypothesis one (H1) tests whether the stock/compensation ratio is associated with the informativeness of earnings of highly dispersed ownership companies. Consistent with previous financial studies, earnings coefficients will measure the level of earnings informativeness. The following model is employed to test H1: RETURN = b 0 + b 1 EPS + b 2 EPS*STKRATIO + e where the variables are defined as follows: RETURN. The stock return with a measurement period extending from nine months prior to the fiscal year-end through three months after the fiscal year-end. EPS. The change in earnings-per-share from the prior year. 11

13 STKRATIO. The stock/compensation ratio equals the market value of stock owned by the top five managers divided by the managers annual compensation (salary and bonus). The earnings coefficient is the sum of b 1 and b 2 (i.e., b 1 + b 2 ). The interaction term coefficient (b 2 ) measures the earnings coefficient contribution by STKRATIO. A significant positive b 2 coefficient would support H1. Regarding control variables for the above model, prior literature provides evidence that earnings coefficients are positively related to managerial ownership (Warfield, Wild and Wild, 1995), growth prospects (Collins and Kothari, 1989) and the persistence of earnings (Collins and Kothari, 1989; Easton and Zmijewski, 1989). In contrast, earnings coefficients are negatively associated with systematic risk (Collins and Kothari, 1989; Easton and Zmijewski, 1989), firm size (Atiase, 1985), leverage (Dhaliwal, Lee, and Fargher, 1991) and unexpected earnings variability (Lipe, 1990). Similar to Warfield, Wild and Wild (1995), these control variables are included in the above model to test the robustness of the results as follows: RETURN = b 0 + b 1 EPS + b 2 EPS*STKRATIO + b 3 EPS *OWN + b 4 EPS*SIZE + b 5 EPS*RISK + b 6 EPS*DEBT + b 7 EPS*GROWTH + b 8 EPS*VAR + b 9 EPS*PERS + e where the variables are defined as follows: 12 OWN. The percent of equity shares held by individuals who can exercise significant influence over corporate affairs. This variable is obtained from the annual proxy statements and is the Warfield, Wild and Wild (1995) measure of managerial ownership. SIZE. The natural logarithm of the market value of equity. 12

14 RISK. The firm-specific systematic risk measured by the company s beta estimated from the market model. Generally, the company-specific market model is estimated using the most recent 60 months stock returns prior to the test period. DEBT. The ratio of the company s long-term debt to total assets. GROWTH. Potential company growth measured by the market value of equity scaled by the book value of equity. VAR. The variability of earnings computed as the standard deviation of earningsper-share for the sixteen quarters prior to the test date. PERS. The persistence of earnings calculated as the first order autocorrelation in earnings for the sixteen quarters prior to the test date. For the full model, the earnings coefficient equals the sum of all of the independent variables coefficients (i.e., b 1 through b 9 ). A significant and positive b 2 would support H1. Hypothesis two (H2) is tested by examining the association between the absolute value of abnormal accruals employed by management and the stock/compensation ratio. Consistent with Warfield, Wild and Wild (1995), the absolute value of the abnormal accrual is the model s dependent variable because the hypothesis does not rely on the direction of the accrual adjustments but rather on the magnitude of the adjustments. The abnormal accrual is estimated by subtracting the expected normal accrual from the current period accrual, and then standardizing by the beginning of the year stock price. The modified Jones (1991) model is employed to estimate expected normal accruals. 13 The parameters of the modified Jones (1991) model were estimated for each firm using eight to nineteen years of firm data prior to the abnormal accrual measuring date. 14 The predicted 13

15 accrual generated from the estimated Jones (1991) model was then subtracted from the current period accrual to calculate the abnormal accounting accrual. Prior research identifies additional factors that affect management accounting choices. Specifically, the magnitude of discretionary accrual adjustments are inversely related to the level of managerial ownership (Warfield, Wild and Wild, 1995) and positively related to firm size, firm risk (Watts and Zimmerman, 1978; Hagerman and Zmijewski, 1979; Zmijewski and Hagerman, 1981) and debt level (Sweeney, 1994). The following model is employed to test H2: ACC = b 0 + b 1 OWN + b 2 STKRATIO + b 3 SIZE + b 4 RISK + b 5 DEBT + e where variables not previously discussed are defined as follows: ACC. The absolute value of the abnormal accrual estimated by the current period accrual less the expected accrual (estimated per the modified Jones (1991) model), and then standardized by the beginning of the year stock price. Table 1 summarizes the definitions of variables used in the empirical models. [Insert Table 1 about here] As a sensitivity test, we also include three control variables from the returns model to measure the robustness of the findings: firm growth (GROWTH), earnings variability (VAR) and earnings persistence (PERS). 14

16 Sample Selection, Data Sources and Descriptive Statistics Sample and Data Sources To be included in the final sample, companies had to meet the following criteria: (1) the level of managerial ownership must be less than or equal to five percent; 15 (2) listed on the New York Stock Exchange (NYSE) or the American Stock Exchange (ASE) with a fiscal year-end 12/31/94; (3) primary SIC code not in the 4000 s (regulated industries) or 6000 s (financial institutions); (4) managerial annual compensation (salary and bonus) must be available in annual proxy statements; (5) managerial ownership data must be available in proxy statements or through CD Disclosure; (6) financial data must be available on Compustat s Industrial or Full Coverage databases; and (7) stock price and returns data must be available from the CRSP (Center for Research on Stock Prices) database. The data to compute the STKRATIO (stock/compensation ratio) variable were manually obtained from annual proxy statements. The managerial ownership data (OWN) were also manually collected from annual proxy statements, supplemented with data from CD Disclosure where available. Earnings and stock returns data were drawn for NYSE and ASE companies with fiscal years ending December 31, 1994 from the CRSP and Compustat databases. 16 After considering the five percent ownership, available proxy statements, exchange, fiscal year-end, and SIC code criteria, a total of 282 companies emerged. Companies were dropped from this group for the following reasons (amounts in parentheses): missing financial statement (Compustat) data (35); missing data in proxy statements (13); and limited partnerships (26). These criteria yielded 208 observations for tests of the 15

17 magnitude of discretionary accruals. Missing stock return (CRSP) data (4) resulted in a sample of 204 observations for tests of the informativeness of earnings. Descriptive Statistics Descriptive statistics, univariate correlations and industry data are presented in Table 2. The mean managerial ownership amount for the sample of 208 companies is 2.03 percent of common shares outstanding, while median ownership is 1.8 percent of common shares outstanding. [Insert Table 2 about here] The mean stock/compensation ratio (STKRATIO) for the sample companies is 8.92, with a median ratio of The rather large mean of this variable suggests that despite the low percentage of outstanding stock owned by sample companies management, a significant portion of management s wealth is associated with their employer company s stock price. Table 2b presents univariate correlations for the independent variables included in the models. Interestingly, our variable of interest (STKRATIO) has a positive correlation of only.077 with the Warfield, Wild and Wild (1995) ownership variable (OWN). This insignificant correlation suggests that the stock/compensation ratio (STKRATIO) is measuring different effects than the Warfield, Wild and Wild (1995) ownership variable for low managerial ownership companies. 17 The STKRATIO variable is significantly correlated with the SIZE (.436) and GROWTH (.180) variables. 18 None of the remaining independent variables are highly collinear with STKRATIO. Tables 2c and 2d present some descriptive statistics for the stock/compensation ratio (STKRATIO) and ownership (OWN) variables by industry. The majority of sample 16

18 observations are represented by the manufacturing industry (168 companies (81%)). Other industries represented include construction and mining (SIC codes ), wholesale and retail (SIC codes ) and services (SIC codes ). At the one-digit SIC code level (Table 2c), except for the product manufacturing industry (SIC codes ), the means of the stock/compensation ratio across industries are comparable to the overall sample mean. The means of the ownership variable (OWN) across industries are also comparable to the overall sample mean. In Table 2d, two-digit SIC code industries are presented where the two-digit industry has at least 10 members in the sample. This analysis suggests that the stock/compensation ratio (STKCOMP) is higher than the overall sample in the chemical and allied products ( ), petroleum refining and related industries ( ) and electrical equipment other than computers ( ) industries. These descriptive statistics suggest that industry effects may have an influence on our results. We consider this possibility in sensitivity tests later in the paper. Empirical Results Explanatory Power of Earnings Conditional on the Stock/Compensation Ratio The first hypothesis predicts that the informativeness of earnings is positively related to the stock/compensation ratio. One measure of the informativeness of earnings is its explanatory power for returns. Table 3 presents the results for the returns regression analyses. 19 The regression of stock returns on changes in earnings-per-share is presented in column 3a. 20 Consistent with prior research, the results suggest that changes in earningsper-share are significantly associated with stock returns. Column 3b presents the results of our replication of Warfield, Wild and Wild (1995). Specifically, stock returns are 17

19 regressed on changes in earnings-per-share and the interaction of changes in earnings-pershare and levels of managerial ownership (i.e., EPS*OWN). Consistent with Warfield, Wild and Wild (1995), the interaction between changes in earnings-per-share and levels of managerial ownership ( EPS*OWN) is positive and mildly significant (p-value = 0.07). This result suggests that, even for low ownership level firms, earnings are more informative as ownership levels increase. Column 3c presents the results of the model when the interaction of changes in earnings-per-share and levels of managerial ownership ( EPS*OWN) is replaced with the interaction of changes in earnings-per-share and the stock/compensation ratio ( EPS*STKRATIO). The results suggest that the informativeness of earnings significantly increases (p-value <0.00) when the stock/compensation ratio (STKRATIO) increases, which supports H1. [Insert Table 3 about here] To examine the robustness of these results, we perform additional testing. First, the interaction of changes in earnings-per-share and levels of managerial ownership ( EPS*OWN) may proxy for the same effect as the interaction between changes in earnings-per-share and the stock/compensation ratio ( EPS*STKRATIO). As previously discussed, the low correlation between these variables indicates that they are not measuring the same effect. However, to further address this possibility, column 3d presents the results of the model that regresses stock returns on both interactions and changes in earnings-per-share. The interaction between changes in earnings-per-share and the stock/compensation ratio ( EPS*STKRATIO) continues to be highly significant (p-value <0.00) while the interaction between changes in earnings-per-share and levels of 18

20 managerial ownership ( EPS*OWN) is no longer significant (p-value =0.22). This result suggests that after controlling for ownership levels (OWN), the interaction between changes in earnings-per-share and the stock/compensation ratio ( EPS*STKRATIO) continues to provide explanatory power for returns of companies whose ownership levels are less than five percent. Similar to Warfield, Wild and Wild (1995) several control variables are added to the model presented in column 3d. 21 The results generated from this model are presented in column 3e. The variable of interest, the interaction between changes in earnings-per-share and the stock/compensation ratio ( EPS*STKRATIO), continues to provide explanatory power for returns of companies whose ownership levels are less than five percent. In addition, the coefficient for the changes in earnings-per-share is positive and significant, and the coefficients for the control variables risk (RISK), debt levels (DEBT) and the variability of earnings (VAR) are at least marginally significant in the predicted directions. As a final sensitivity, we consider the influence of industry effects on our results (not presented in tabular form). In the first industry analysis, dummy variables equaling one for industry representation (e.g., ), zero otherwise, are created to control for industry effects. The industry dummy variables are interacted with changes in earnings per share and are included in the full model presented in column e of Table 3. The variable of interest, the stock/compensation ratio interaction ( EPS*STKRATIO) continues to be highly significant (p-value <0.00). 22 Based on the sensitivity analyses, we conclude our results supporting H1 are robust. 19

21 Managers Accounting Choices Conditional on the Stock/Compensation Ratio Hypothesis two predicts a negative relation between the absolute value of abnormal accruals and the stock/compensation ratio observed in low managerial ownership companies. Referring to Table 4, column 4a presents the results for the model regressing abnormal accruals on managerial ownership (OWN), the stock/compensation ratio (STKRATIO), size (SIZE), market model beta (RISK) and debt (DEBT). The coefficient on STKRATIO is significant and negative, which supports H2 and suggests that increases in the stock/compensation ratio are associated with lower discretionary accrual adjustments made by management. Consistent with the results of the returns regressions, when the stock/compensation ratio (STKRATIO) is included in the model, the effect of the ownership level (OWN) is not statistically significant for 0 5 percent ownership companies. The control variables RISK and DEBT are significant in the predicted direction. [Insert Table 4 about here] As a sensitivity test, we add three additional control variables from the returns model to help ensure the above results are robust. The three control variables are company growth (GROWTH), the variability of earnings (VAR) and the persistence of earnings (PERS). The results of this model are presented in column 4b and are qualitatively unchanged from the restricted model. The primary variable of interest STKRATIO continues to be significant and inversely related to the absolute value of abnormal accruals. 20

22 Additional Analyses Warfield, Wild and Wild (1995) provide evidence that the level of managerial stock ownership is positively related to the informativeness of earnings. Included in their analysis are the correlations between earnings and returns as well as earnings/returns regression coefficients for various managerial ownership levels (e.g., 0% 5% managerial ownership). As an additional analysis, we provide evidence related to the correlations between earnings and returns as well as earnings/returns regression coefficients for various groups of stock/compensation ratios (e.g., stock/compensation ratios between 0 3). These results are presented in Table 5, panel a. 23 Given the positive association between returns and the interaction between stock/compensation ratios and changes in earnings-pershare ( EPS*STKRATIO), the correlations between returns and changes in earnings-pershare ( EPS) are expected to increase as the stock/compensation ratio increases. In Table 5, panel a, low managerial ownership companies with a stock/compensation ratio of less than three exhibit a correlation between returns and earnings-per-share changes ( EPS) of approximately 5%. This correlation generally increases until the stock/compensation ratio group, where the correlation is 38.1%, and then decreases to 21.9% for the >25 stock/compensation ratio group. Consistent with prior literature, these results suggest that the stock/compensation ratio has an optimum level; low managerial ownership companies with a stock/compensation ratio greater than 25 exhibit a relatively weak correlation of 21.9%. 24 [Insert Table 5 about here] 21

23 An important question is whether companies with low managerial ownership and high stock/compensation ratios release earnings that are comparable in informativeness to companies with managerial ownership greater than five percent. To examine this issue, companies with managerial ownership levels greater than five percent were sampled and the returns/earnings changes relationship was examined for the years These results are presented in Table 5, panel b. Consistent with Warfield, Wild and Wild (1995), generally, the correlations increase as managerial ownership levels increase. 25 Similar to Warfield, Wild and Wild (1995), the correlation for companies with managerial ownership levels greater than 55% declines slightly. The focal group of this study, 0% - 5% ownership companies, has a correlation of 13.8% (Table 5, bottom of panel a). Comparing the correlations and coefficients in panel a to those in panel b suggests that the stock/compensation ratio appears to be reasonably effective in influencing managers of low managerial ownership firms to disclose earnings of similar quality to managers of high managerial ownership firms. The average correlation between returns and earnings-per-share changes for companies with stock/compensation ratios between 6 and 15 (categories 6 10 and 10 15) is approximately 20%. This average correlation is greater than the correlation for companies whose managerial ownership is between 5% and 15% (17.5%) and only slightly below that of companies whose managerial ownership is between 15%-25% (24.3%) and 25%-35% (22.8%). The comparison of regression coefficients reveal similar results. Overall, consistent with the analyses presented in Tables 3 and 4, these findings suggest that in low managerial ownership companies, increases in stock/compensation ratios help mitigate owner-manager agency problems. 22

24 Because 1994 was a somewhat anomalous year in the stock market, we employ the regression models presented in Table 3 (columns 3c and 3d) using data from the period Specifically, the 607 observations noted in Table 5 for low managerial ownership companies (i.e., with five percent or less managerial ownership) were used for analysis. The results from employing regression model 3c using data for the three year period are virtually identical to the previously presented results using only 1994 data. Similarly, when using three-year period data for model 3d, the stock/compensation ratio interaction ( EPS * STK RATIO) remains highly significant (p-value <0.00). These results suggest that the findings for 1994 presented in Table 3 hold under an expanded time frame. Alternative Measure of the Stock/Compensation Ratio Lewellen, Loderer and Martin (1987) note that an effective pay package will minimize the costs of the agency relationship between owners and managers (see also Jensen and Meckling, 1979; and Lambert and Larcker, 1985). However, the amount of current year executive compensation related to long-term performance horizons is typically based upon short-term performance measures. In addition, the amounts of such compensation are unknown until after the current performance period has been completed. Thus, it is unclear how current year incentives initiated to provide managers with a longterm performance horizon in future years affects current period management behavior. As a result, such current year incentives are excluded from the stock/compensation ratio. However, to the extent that long-term horizon incentives are emphasized in managerial compensation contracts, an alternative measure of the stock/compensation ratio is the ratio of current period compensation rewards tied to long-term company performance 23

25 in the form of restricted stock and stock options to annual salary and bonus. Such a measure is consistent with anecdotal observations in company proxy statements where companies suggest that top executives should own minimum company stock amounts relative to their annual compensation and where such companies design compensation packages to achieve that goal. 27 This alternative measure crudely estimates the current period influence on management s horizon resulting from the executives remuneration package. As stated above, it is unclear how current period compensation promoting longterm horizons affects current period management behavior. Assuming that equity based compensation (e.g., stock options and restricted stock) immediately increases the long-term horizon of management, we expect a positive relation between the alternative measure of the stock/compensation ratio and earnings informativeness. [Insert Table 6 about here] The results from the regressions using the alternative measurement are presented in Table 6, where the column headings (6c-6e) are designed to parallel those in Table 3 (i.e., 3c-3e). The alternative stock/compensation ratio measurement ( EPS*STKRATIO) is statistically significant in all three models. The managerial ownership interaction ( EPS*OWN) is significant in model 6d (p-value =.04) and mildly significant in model 6e (p-value =.09). These results are similar to those presented in Table 3, and suggest that current period compensation tied to long-term performance horizons (restricted stock and stock options) relative to annual salary and bonus is positively associated with the informativeness of earnings. Overall, the results from the additional analyses including the alternative measure of the stock/compensation ratio suggest that the results presented in Table 3 are robust to a number of alternative specifications and considerations. 24

26 Summary and Conclusions Consistent with agency cost literature, Warfield, Wild and Wild (1995) provide evidence that the level of managerial ownership is positively related to the informativeness of earnings disclosed by management and negatively associated with the magnitude of discretionary accrual adjustments. The authors conclude that these results suggest that dispersed ownership firms construct contracts to constrain and/or monitor management behavior and to the extent that these contracts are based on accounting numbers, managers exploit discretionary accruals to alleviate contract constraints or maximize incentive benefits. In order to reach these conclusions, the authors implicitly assume that compensation objectives and characteristics other than those tied to stock ownership are constant across sample firms. This study examines the ability of managerial stock ownership relative to annual compensation (i.e., the stock/compensation ratio) to mitigate the effects of low managerial ownership and related agency problems within the sample of firms that contained the lowest levels of informative earnings in the Warfield, Wild and Wild (1995) study. Specifically, we test (1) the effect of the stock/compensation ratio on the returns-earnings relation and (2) the association between the stock/compensation ratio and the absolute value of abnormal accruals for companies whose managerial ownership is less than five percent. Our results suggest that even for low managerial ownership companies, the amount of stock ownership relative to annual compensation better aligns the interests of managers and owners. This better alignment is reflected in both informative earnings 25

27 disclosures and lower discretionary accrual adjustments for dispersed managerial ownership companies. Our study is subject to several limitations. First, as noted in Agrawal and Mandelker (1987), Jensen and Meckling (1976) and others, mechanisms to control manager/shareholder agency problems include not only managerial stock ownership and annual compensation but also market mechanisms (takeover and managerial labor markets) and contractual schemes other than compensation (e.g., debt contracts). To the extent that such mechanisms are not captured as control variables, the results presented in this paper could be misstated. However, the effects of any omitted variables would have to be highly correlated with the stock/compensation ratio and such a high correlation is unlikely. Further, the stock/compensation ratio is designed to reflect the amount of managers wealth tied to the value of the firm (long-term performance horizon) versus the amount of managers wealth tied to short-term goals such as job preservation (e.g., annual salary) and incentive compensation (e.g., annual bonus). This ratio is a barometer mentioned in several compensation philosophy sections of companies annual proxy statements and has been used in prior research, which lends some support toward using it in our empirical models. However, this proxy may not adequately measure the relative balance between long-term and short-term horizons for low managerial ownership companies. Another limitation of our study is that our sample of less than five percent ownership firms comes only from New York Stock Exchange and American Stock Exchange companies. This restriction may limit the generalizability of our results, and thus they should be interpreted with caution. 26

28 Endnotes 1 Consistent with Warfield, Wild and Wild (1995), low managerial ownership is defined as stock ownership by officers and directors that is less than five percent of the outstanding shares of common stock. 2 An implicit assumption of Warfield, Wild, and Wild (1995) is that the ratio of stock ownership (market value) to annual salary and bonus is proportional to the percentage of outstanding common shares owned by influential individuals. We contend that the short-term horizon incentives (annual salary and bonus) are not necessarily proportional to the long-term horizon incentives (stock ownership), and that the stock/compensation ratio is a reasonable proxy for the relative longterm and short-term horizon influences. This contention is supported by the low correlation between Warfield, Wild and Wild s (1995) ownership variable and the stock/compensation ratio (.077; See table 2b). 3 The fundamental role of the separation of ownership and control to the contemporary theory of the firm is also central to Berle and Means (1932), Williamson (1964), Fama (1980) and Fama and Jensen (1983). 4 Despite these negative ramifications, prior literature suggests that accounting-based constraints are incrementally useful in executive compensation contracts because they help shield executive compensation from market-wide fluctuations in equity values (Lambert and Larcker, 1987; Lewellen, Loderer and Martin, 1987; Sloan, 1993). 5 Prior researchers recognizing this horizon problem have examined issues related to the relative use of current versus deferred compensation and cash-based versus stock-related compensation. Prior researchers include Lewellen, Loderer and Martin (1987), Lambert and Larcker (1987), Jensen and Murphy (1990), Sloan (1993), Ramanan and Balachandran (1993) and more recently, Guay (1998) and Core and Guay (1998) among others. 6 Alternatively, executive compensation packages heavily weighted toward stock price, a long-term performance horizon, may create different manager-shareholder agency problems. Significant stock ownership often results in a significant percentage of managers personal wealth contingent on the stock value of their companies. Lewellen, Loderer and Martin (1987) observe that managers may not be able to easily diversify away the specific risks associated with companies that employ them. Thus, executive compensation heavily weighted toward stock value may make management more reluctant to choose high-variance investment projects because such projects increase their risk exposure. 7 Related to compensation schemes, Abdel-Khalik, Chi and Ghicas (1987) observe that executives cannot consistently generate rewards by managing the measurement of accounting profits irrespective of changes on the wealth of the firm. However, the concern presented in this paper is that outside equity holders of low managerial ownership companies are unable to determine periods of income manipulation versus periods when income is reflective of changes in the wealth of the company. As a result, because managers of low managerial ownership companies are more likely to exploit the latitude in accepted accounting procedures, market participants are less likely to value accounting information to the 27

29 same extent as in companies where the accounting information is believed to be more reflective of company wealth changes. 8 The stock/compensation ratio has been employed in empirical research. Lambert and Larcker (1987) use a variable similar to this ratio that proxies for the degree that manager s wealth is tied to the firm s stock price. Agrawal and Mandelker (1987) incorporate the stock/compensation ratio as a representation of the number of years of total annual compensation that a manager has as an investment in stock. 9 Managerial stock ownership is affected not only by the companies compensation policy but also by personal ownership decisions of managers. However, Lewellen (1971) finds that managerial liquidations of their company s stock from personal portfolios to be infrequent. Lewellen, Loderer and Martin (1987) suggest that this result is because such liquidations may result in adverse signaling which results in reductions of the value of the managers human capital or because of implicit contracts between managers and shareholders that company stock will not be liquidated. Consistent with these arguments, empirically, Ofek and Yermack (1997) find that executives sell 0.16 shares of stock for each new stock option granted. 10 Although just three examples are presented, proxy statements of many companies explicitly encourage minimum levels of the ratio of the market value of stock owned by top managers to annual compensation in the compensation philosophy section of the annual proxy statements. 11 This hypothesis is consistent with the findings of Warfield, Wild and Wild (1995). However, if either accountingbased constraints mitigate managers accounting choices or higher ownership results from difficulties in accounting numbers measuring performance, a positive relation would be expected. 12 Consistent with theoretical formulation and prior research, the coefficients b 4, b 5, b 6, and b 8, are expected to be negative while the coefficients b 3, b 7, and b 9, are expected to be positive. 13 Dechow, Sloan and Sweeney (1995) examine the ability of five models used in prior research to detect discretionary accruals. The authors conclude that the modified version of the Jones (1991) model exhibits the most power in detecting discretionary accruals. 14 The modified Jones (1991) model involves the use of the following expectations model to estimate the firm-specific parameters relating to non-discretionary accruals: TA t = a 1 + a 2 ( REV t - REC t ) + a 3 (PPE t ) + ν t where, REV t = revenues in year t less revenues in year t-1; REC t = net receivables in year t less net receivables in year t-1; PPE t = gross property plant and equipment in year t; ν t = the residual and is the estimated discretionary accrual adjustment for year t. TA t = ( CA t - CL t - Cash t + STD t Dep t )/(A t-1 ), 28