Journal of Accounting and Economics

Transcription

1 Journal of Accounting and Economics 52 (2011) Contents lists available at ScienceDirect Journal of Accounting and Economics journal homepage: Hometown advantage: The effects of monitoring institution location on financial reporting discretion Benjamin C. Ayers n, Santhosh Ramalingegowda 1, P. Eric Yeung 2 J.M. Tull School of Accounting, Terry College of Business, University of Georgia, Athens, GA 30602, USA article info Article history: Received 19 April 2010 Received in revised form 2 March 2011 Accepted 9 March 2011 Available online 21 March 2011 JEL classification: M41 G2 G34 abstract We examine the impact of institutional ownership on financial reporting discretion, focusing on whether the impact varies with institutions cost of acquiring monitoring information. Using geographic distance between the firm and the institutional investor as a proxy for the cost of acquiring monitoring information, we find that corporate managers are less likely to use financial reporting discretion in the presence of local monitoring institutions than distant monitoring institutions. We also find that the impact of monitoring institutions on financial reporting discretion varies with the costs and benefits of financial reporting discretion. & 2011 Elsevier B.V. All rights reserved. Keywords: Reporting discretion Institutional investors Geographic distance Corporate governance Earnings management 1. Introduction We examine the impact of institutional ownership on financial reporting discretion, focusing on whether the impact varies with institutions cost of acquiring information about financial reporting discretion (hereafter, monitoring information). We posit that geographic distance between the firm and a monitoring institutional investor impacts the institution s cost of acquiring monitoring information and predict that corporate managers are less likely to use financial reporting discretion in the presence of local monitoring institutions (our proxy for monitors with low information costs) than distant monitoring institutions. We base our expectation on two streams of research: (1) prior findings that institutional investors located in proximity to the firm are more informed than other institutional investors located far away and (2) prior research that suggests information asymmetry between the firm and stakeholders is a necessary condition for managers to engage in the use of opportunistic reporting discretion. 3,4 We argue that the close geographic n Corresponding author. Tel.: þ addresses: bayers@terry.uga.edu (B.C. Ayers), smr@terry.uga.edu (S. Ramalingegowda), yeung@terry.uga.edu (P. Eric Yeung). 1 Tel.: þ Tel.: þ Coval and Moskowitz (2001) and Baik et al. (2010) find institutional investors located in proximity to a firm are more informed than institutional investors located far away from a firm. Likewise, other studies provide robust evidence that geographical proximity to a firm reduces information acquisition costs of banks (Petersen and Rajan, 1994, 2002), acquirer-firms (Kang and Kim, 2008), and analysts (Malloy, 2005). The economic story in these studies is that direct observation of the firms operations or face-to-face meetings with managers, directors, employees, suppliers, customers, and local media provides easier access toinformation. 4 Watts and Zimmerman (1986, Chapters 8 and 9) argue that managers engage in opportunistic discretion only when it is difficult for investors to detect it (also see Trueman and Titman, 1988; Dye, 1988) /$ - see front matter & 2011 Elsevier B.V. All rights reserved. doi: /j.jacceco

2 42 B.C. Ayers et al. / Journal of Accounting and Economics 52 (2011) proximity of local monitoring institutions reduces the information asymmetry between management and outside investors that prior research argues is a necessary condition for managers to engage in opportunistic reporting discretion. Accordingly, we predict a negative association between financial reporting discretion and local monitoring institutional ownership and that this association is incrementally negative to the association between financial reporting discretion and distant monitoring institutional ownership. We also expect the relation between financial reporting discretion and local and distant monitoring institutional ownership to vary cross-sectionally with the costs of such discretion. Prior research suggests that lower opportunistic discretion in earnings disciplines managers investment decisions (McNichols and Stubben, 2008; Biddle et al., 2009; Kedia and Philippon, 2009). Given the greater potential costs of opportunistic financial reporting discretion for firms with more investment opportunities, the presence of investors with an information advantage (local monitoring institutional investors) may have a greater impact on opportunistic financial reporting for these firms as the monitoring incentives are particularly strong in this setting. Accordingly, we predict that the association between financial reporting discretion and local monitoring institutional ownership is more pronounced among firms with high investment opportunities. We define monitoring institutions as institutions that are top five shareholders in the firm for the year, independent from management (Brickley et al., 1988), and classified as dedicated or quasi-indexers by Bushee (2001). Following prior research, we classify local monitoring institutions as those located within 100 km from the firm headquarters, and distant monitoring institutions as those located beyond 100 km from the firm headquarters. We address potential endogeneity issues between financial reporting discretion and local monitoring institutional ownership by using an instrumental variables analysis whereby we use 26 geographic location indicator variables as instruments to estimate a firm s predicted values of local and distant monitoring institutional ownership. We then use the predicted values of local and distant monitoring institutional ownership to conduct empirical tests. As opportunistic financial reporting discretion can be income increasing or income decreasing, we use the absolute value of abnormal accruals as our initial proxy for financial reporting discretion (e.g., Klein, 2002; Frankel et al., 2002; Leuz et al., 2003; Cornettetal.,2008). 5 Using a broad sample of 10,695 firm-years over , we find a negative association between the magnitude of absolute abnormal accruals and local monitoring institutional ownership. We also find that the association between absolute abnormal accruals and local monitoring institutional ownership is incrementally negative to the association between absolute abnormal accruals and distant monitoring institutional ownership. These results are consistent with the lower cost of acquiring monitoring information reducing the information asymmetry between managers and local monitoring institutions, and managers, in turn, voluntarily reducing opportunistic financial reporting discretion. We also find that the impact of local monitoring institutional ownership on financial reporting discretion is more pronounced for firms with higher market-to-book ratios or with more intensive research and development (i.e., our proxies for high investment opportunities, where monitoring incentives are particularly strong), and that the difference in the associations between absolute abnormal accruals and local and distant monitoring institutional ownership is more pronounced for these firms. Because financial reporting discretion can be used opportunistically or to convey private information (i.e., discretion is not always detrimental), we conduct three event-based tests where we have clear predictions regarding the direction and nature of financial reporting discretion. First, we examine the associations between signed abnormal accruals and local and distant institutional ownership in acquirer firms prior to takeover announcements. Prior research shows that acquirer firms use reporting discretion to increase income before stock-for-stock takeover transactions to boost stock prices and hence reduce the cost of takeovers. However, prior studies also find that acquirers overall wealth decreases around stockfor-stock acquisitions if they have used reporting discretion to increase income before the acquisitions (Louis, 2004; Gong et al., 2008). Based on this evidence and local monitoring institutions being more informed, we expect that managers are less likely to use income increasing discretion prior to stock-for-stock takeovers in the presence of local monitoring institutions and less so than in the presence of distant monitoring institutions. Results confirm our expectations. Second, we examine the associations between signed abnormal accruals and local and distant monitoring institutional ownership prior to stock splits, an event characterized by the use of pre-event income increasing discretion that is more likely to reflect private information about superior performance rather than opportunistic behavior (Louis and Robinson, 2005). Given the nature of financial reporting discretion preceding stock splits, we do not expect that the presence of local monitoring institutions deters financial reporting discretion in this setting. Consistent with expectations, in the quarter before stock splits, we find no evidence of a negative relation between signed abnormal accruals and local monitoring institutional ownership. Third, we examine whether managers are less likely to engage in financial statement fraud in the presence of local monitoring institutions. Because local monitoring institutions have easier access to information possessed by firms employees and trading partners such as customers and suppliers, managers may be less likely to engage in financial statement fraud, which often requires collusion among employees, customers, and suppliers. Consistent with expectations, we find a negative association between the probability of Security Exchange Commission (SEC) fraud enforcement action and local monitoring institutional ownership. In addition, we find that the association between the probability of fraud enforcement action and local monitoring institutional ownership is incrementally negative to the association between the probability of fraud enforcement action and distant monitoring institutional ownership. 5 Because monitoring institutions may be especially concerned with managers using financial reporting discretion to overstate earnings, we also re-estimate our analyses using signed abnormal accruals and obtain similar results (see the discussion in Section 3.5).

3 B.C. Ayers et al. / Journal of Accounting and Economics 52 (2011) One implication of our argument that managers voluntarily reduce opportunistic reporting discretion in the presence of local monitoring institutions is that managers who do engage in opportunistic reporting in the presence of local monitoring institutions are more likely to be detected and subject to penalties in a timelier fashion. We thus examine the impact of financial reporting fraud on subsequent CEO turnover as a function of the level of local monitoring institutional ownership. We focus on CEO turnover beginning the year immediately following the fraud period through the fourth year after the fraud year, which spans both the pre- and post-public announcement of fraud allegations for the vast majority of sample firms. Consistent with local monitoring investor information advantage, we find that CEO turnover in the first two years following fraud is significantly higher for firms with high local monitoring institutional ownership (relative to control no-fraud firms and fraud firms with low local monitoring institutional ownership). Our study contributes to the literature investigating financial reporting discretion by showing that the effects of outside shareholders monitoring a firm vary with their information acquisition cost. Specifically, we find evidence that the geographic distance between a firm and its monitoring institutional investors has a systematic and predictable impact on financial reporting discretion. We also demonstrate how the effect of local monitoring institutions on financial reporting discretion varies with the costs and benefits of financial reporting discretion. Among other implications, our findings highlight an important advantage associated with investing locally. While our study focuses on institutional investment in the U.S., our findings may be of particular interest for researchers examining cross-border investments. Section 2 summarizes prior studies and introduces our hypotheses. Section 3 describes our research design, data sources, sample, and results for our analyses using absolute abnormal accruals. Section 4 presents event-based tests, and Section 5 summarizes and concludes. 2. Literature review and hypothesis development 2.1. Prior research Prior research shows that opportunistic financial reporting discretion leads to significant direct costs to shareholders. For example, several studies indicate that use of opportunistic financial reporting discretion leads to over- and underinvestments (e.g., Biddle et al., 2009; Kedia and Philippon, 2009; McNichols and Stubben, 2008) and conclude that higher financial reporting quality decreases both over- and under-investment problems. Evidence also indicates that opportunistic financial reporting discretion increases litigation risk (e.g., Palmrose and Scholz, 2004; Lev et al, 2008; Ducharme et al., 2004; Gong et al., 2008). For example, Gong et al. (2008) find that pre-merger abnormal accruals are a strong determinant of post-merger lawsuits, even after controlling for post-merger abnormal returns. Finally, managers also use financial reporting discretion to extract wealth from shareholders through bonus, option, and restricted stock compensation (e.g., Healy, 1985; Holthausen et al., 1995; Cohen et al., 2008; Cheng and Warfield, 2005; Bergstresser and Philippon, 2006). In sum, opportunistic financial reporting discretion can lead to significant direct costs to shareholders. In the area of corporate governance, several studies focus on the association between good governance structures and financial reporting discretion. For example, a greater proportion of outside directors on the board of directors is associated with less SEC accounting enforcement actions (e.g., Dechow et al., 1996) and lower absolute abnormal accruals (e.g., Klein, 2002). Likewise, in the context of institutional investors, several papers argue and/or find evidence that monitoring by institutional investors constrains corporate managers behavior (e.g., Shelifer and Vishny, 1986; Brickley et al., 1988; Hartzell and Starks, 2003; Chen et al., 2007). Armstrong et al. (2010) argue that active investors, such as institutional investors, require timely and reliable information about the firm so they can monitor management s actions and participate in the firm s strategic direction. Bushee (1998) finds that institutional ownership is negatively associated with research and development (R&D) expenditure cuts to meet earnings benchmarks. Cornett et al.(2008) and Rajgopal and Venkatachalam (1997) find a negative relation between absolute abnormal accruals and institutional ownership. These results are consistent with corporate managers, on average, being less likely to use financial reporting discretion in the presence of institutional investors. An implicit assumption in studies examining the relation between institutions monitoring and financial reporting discretion is that institutions (with equal incentives to monitor a firm) are equally informed about financial reporting discretion and its effect on firm value. At the least, prior studies assume that proxies for monitoring by institutions, such as ownership level, capture the joint effect of monitoring incentives and monitoring information Testable hypotheses As monitoring information acquisition is costly, it is likely to vary across institutions, even among those with similar incentives to monitor. Because not all institutions have equal incentives to monitor managers behavior, we follow prior studies (e.g., Shleifer and Vishny, 1986; Brickley et al., 1988; Chen et al., 2007) and classify institutions that have concentrated shareholdings, are independent from corporate management, and have long investment horizons as institutions with strong incentives to monitor managers. 6 We then focus on the frictions created by cross-sectional differences in monitoring 6 Non-monitoring institutions, because of their short investment horizon, fragmented ownership, or/and fear of losing business with the firm, are less likely to be concerned about the consequences of opportunistic reporting or about penalizing managers who engage in opportunistic reporting.

4 44 B.C. Ayers et al. / Journal of Accounting and Economics 52 (2011) information cost. We posit that geographic distance between the firm and a monitoring institutional investor impacts the institution s cost of acquiring monitoring information and predict that corporate managers are less likely to use financial reporting discretion in the presence of local monitoring institutions (our proxy for monitors with low information costs) than distant monitoring institutions. Specifically, in the absence of the necessary information asymmetry between managers and local monitoring institutions, we predict that managers voluntarily reduce opportunistic financial reporting discretion: H1. Financial reporting discretion is negatively associated with local monitoring institutional ownership and this association is incrementally negative to the association between financial reporting discretion and distant monitoring institutional ownership. We also expect the strength of monitoring from institutions varies with the costs of opportunistic financial reporting discretion. As discussed in Section 2.1, prior research suggests that lower opportunistic discretion in earnings deters overand under-investments (e.g., Biddle et al., 2009; Kedia and Philippon, 2009; McNichols and Stubben, 2008). As managers of firms with more investment opportunities likely have more opportunities and incentives to undertake investments, the costs of financial reporting discretion should be greater among such firms. Given the potentially greater costs of opportunistic financial reporting discretion (and thus, increased monitoring incentives in this setting), we expect that the presence of investors with an information advantage (local monitoring institutional investors) has a greater impact on opportunistic financial reporting for firms with more investment opportunities relative to firms with less investment opportunities 7 : H2. The negative association between financial reporting discretion and local monitoring institutional ownership is more pronounced among firms with more investment opportunities. Because financial reporting discretion can be used opportunistically or to convey private information (i.e., discretion is not always detrimental), we test three event-based hypotheses where we have clear predictions regarding the direction and nature of financial reporting discretion. Prior research finds that acquirer firms use reporting discretion to increase income before stock-for-stock takeover announcements to lower takeover costs (Erickson and Wang, 1999; Louis, 2004; Gong et al., 2008). However, prior studies also find that an acquirer s pre-acquisition use of reporting discretion is value destroying for its shareholders. For example, Louis (2004) finds during the three-year period after acquisition announcements, the average stock-for-stock acquirer suffers a cumulative negative abnormal return of 17.49%, which is partly attributable to the reversal of the effects of the acquirer s pre-merger use of reporting discretion. 8 Likewise, Gong et al. (2008) link post-acquisition lawsuits with the acquirer s pre-acquisition abnormal accruals and find that acquisitions in which the acquiring firm is subsequently sued by shareholders, on average, earn zero abnormal return during the year prior to takeover announcements but suffer a 77% return during the four years after the merger announcement. Overall, these studies suggest that acquirers overall wealth decreases around stock-for-stock acquisitions if they have used reporting discretion to increase income before the acquisitions. Based on this evidence and local monitoring institutions being more informed, we expect that managers are less likely to use income increasing discretion prior to stock-for-stock takeovers in the presence of local monitoring institutions: H3. Income increasing financial reporting discretion prior to stock-for-stock takeovers is negatively associated with local monitoring institutional ownership and this association is incrementally negative to the association between financial reporting discretion and distant monitoring institutional ownership. Financial reporting discretion can sometimes be beneficial to shareholders e.g., by signaling private information about future superior performance, which benefits shareholders by alleviating information asymmetry between the firm and outside shareholders and increases price efficiency and influences resource allocation (e.g., Watts and Zimmerman, 1986; Guay et al., 1996; Subramanyam, 1996; Louis and Robinson, 2005). Recent research identifies stock splits as a setting where financial reporting discretion is less opportunistic and conveys private information about future superior performance. Specifically, prior studies suggest that managers split stocks when they have favorable private information (e.g., Ikenberry et al., 1996; Ikenberry and Ramnath, 2002) but the market under-reacts to stock split announcements (e.g., Ikenberry and Ramnath, 2002). Louis and Robinson (2005) posit that abnormal accruals and stock splits complement and reinforce each other as signals of future superior performance. Consistent with their expectation, they find that firms report positive abnormal accruals prior to stock splits and that there is a positive association between abnormal accruals prior to stock splits and stock returns at the split announcement dates. Based on the evidence in Louis and Robinson (2005), we anticipate that the presence of local monitoring institutions does not deter financial reporting discretion preceding stock splits. Specifically, the private information that local monitoring institutions have should allow these 7 Managers of firms with more investment opportunities also may have more incentives to utilize financial reporting discretion than managers of other firms. However, under the assumption that the costs of being detected exceed the benefits of financial reporting discretion, greater incentives to manage earnings suggests that local monitoring institutions should have a greater impact on opportunistic financial reporting for firms with more investment opportunities (i.e., by mitigating what otherwise would be more pronounced financial reporting discretion in the (a) presence of stronger incentives to manage earnings and (b) absence of informed local monitoring institutions). 8 In our sample, we find that the stock-for-stock acquirers earn zero abnormal return over the year prior to the merger announcements and 53% over three years after the announcement.

5 B.C. Ayers et al. / Journal of Accounting and Economics 52 (2011) institutions to understand the signaling nature of financial reporting discretion preceding stock splits. Accordingly, we test the following hypothesis: H4. Income increasing financial reporting discretion prior to stock splits is not negatively associated with local monitoring institutional ownership. Finally, we examine whether managers are less likely to engage in financial statement fraud in the presence of local monitoring institutions. Because local monitoring institutions have easier access to information possessed by firms employees and trading partners such as customers and suppliers, managers may be less likely to engage in financial statement fraud in the presence of local monitoring institutions since fraud often requires collusion among employees, customers, and/or suppliers. Accordingly, we test the following hypotheses: H5. Financial statement fraud is negatively associated with local monitoring institutional ownership and this association is incrementally negative to the association between financial statement fraud and distant monitoring institutional ownership. 3. Research design and results Abnormal accrual analyses 3.1. Proxies for local and distant monitoring institutional ownership To identify local and distant monitoring institutions, we first define monitoring institutions following Chen et al. (2007) who define monitoring institutions as institutions that meet the following three criteria: (1) top five institutions in a firm for the year in terms of the size of stock holdings, (2) independent from corporate management as defined by Brickley et al. (1988), and (3) classified as dedicated or quasi-indexer institutions by Bushee (2001). We focus on top five institutions because concentrated holdings enhance investors influence over managers and increase their share of the gain from monitoring (e.g., Shleifer and Vishny, 1986; Hartzell and Starks, 2003; Chen et al., 2007). We gauge the independence of institutions based on potential business relationships with their investee firms following Brickley et al. (1988), who classify investment companies, independent investment advisors, and public pension funds as independent institutions because these types have no business relationships with their investee firms. Chen et al. (2007) further argue that institutions classified as dedicated and quasi-indexers have a long-term focus and therefore serve a monitoring role. We follow Bushee s (2001) classification, which classifies institutions into three groups based on their portfolio turnover and diversification, namely, dedicated, transient, and quasi-indexing institutions. Dedicated institutions exhibit low turnover and low diversification, consistent with a long-term strategy of holding large stakes in fewer companies. Quasi-indexing institutions exhibit low turnover and high diversification, consistent with a passive long-term buy-and-hold strategy in a wide set of firms. Transient institutions are characterized by high turnover and high diversification, consistent with short trading horizons and fragmented investments in a large number of companies. Once we identify monitoring institutions as per Chen et al. (2007), we split monitoring institutions into two types based on their geographic proximity to the firm. We classify monitoring institutions located within 100 km of the firm as local monitoring institutions and monitoring institutions located beyond 100 km as distant monitoring institutions. While we follow prior research to classify local and distant investors (e.g., Coval and Moskowitz, 2001; Gasper and Massa, 2007), we note that our results are robust to alternative cutoffs. 9 Non-monitoring institutions include institutions that trade aggressively to maximize short-term trading profit (proxied by Bushee s transient institutions), institutions that are not top five institutions for the year in terms of the size of stock holdings, and institutions that are top five institutions but are not independent from management (proxied by Bushee s dedicated institutions that are also classified as not independent by Brickley et al., 1988). These types of non-monitoring institutions have weaker monitoring incentives than monitoring institutions because of their short investment horizon, fragmented ownership, or/and fear of losing business with the firm Endogeneity and instrumental variable approach Endogeneity is a common concern in corporate governance studies because financial reporting discretion and monitoring institutions ownership can be potentially jointly driven by unobservable firm characteristics or factors that are inadequately controlled for in a linear regression. We address endogeneity concerns through a standard instrumental variable approach. Good instrumental variables should satisfy the conditions of (1) strong correlation with the instrumented regressor and (2) orthogonality with the residuals (i.e., unobservable factors correlated with the dependent variable). Following Gasper and Massa (2007), we use as exogenous instruments 25 firm location indicator variables representing major 25 metropolitan areas of firm headquarters domicile, plus a remote city dummy for a firm more than 250 km away from any of the 25 major cities. Specifically, for each sample year t we regress firms local monitoring institutional ownerships on the 26 firm location indicator variables. We then use the estimated coefficient for firm i s 9 Using a cutoff of 100 km to distinguish local and distant monitoring ownership results in low local monitoring institutional ownership. We observe higher local monitoring institutional ownership when we use alternate cutoffs of 150 and 200 km and find similar results.

6 46 B.C. Ayers et al. / Journal of Accounting and Economics 52 (2011) location in year t to predict local monitoring institutional ownership for firm i, year t. We follow the same procedure to generate the predicted distant monitoring institutional ownership for firm i, year t. 10,11 These variables represent ideal instruments because, notwithstanding the impact of a firm s proximity to investors, a firm s location in one of the 25 major metropolitan areas (or in a remote city) is unlikely to affect its financial reporting discretion. 12 We use the predicted local and distant monitoring institutional ownerships as our instrumental variables to test our hypotheses. Results are similar when we use the actual values of local and distant monitoring institutional ownerships to test our hypotheses Regression model We use the following ordinary least square (OLS) model to test hypotheses 1 and 2: 9Aacc9 i,t þ 1 ¼ a 0 þa 1 plocal_mon i,t þa 2 pdistant_mon i,t þa k Governance i,t þa j Firm i,t þe 1i,t ð1þ Following prior literature (e.g., Klein, 2002; Frankel et al., 2002; Leuz et al., 2003; Cornett et al., 2008), we use the absolute abnormal accruals to proxy for financial reporting discretion in a general setting. Consistent with these studies, we believe that good corporate governance aims to prevent both income increasing and income decreasing opportunistic discretion (e.g., cookie jar reserves). Therefore, we use the absolute value of abnormal accruals to test our hypotheses 1 and 2, where 9Aacc9 i,t þ1 represents the absolute value of abnormal accruals of firm i for year tþ1, estimated using the modified Jones model (Dechow et al., 1995) for each year and two-digit SIC-code industry. 13,14 plocal_mon i,t is the predicted local monitoring institutional ownership for firm i in year t based on the estimated coefficients from the year t regression of local monitoring institutional ownership on the 26 firm location dummy variables, while pdistant_mon i,t is the predicted distant monitoring institutional ownership based on the estimated coefficients from the year t regression of distant monitoring institutional ownership on the 26 firm location dummy variables (see the discussion in Section 3.2). We expect a 1 o0 and a 1 oa 2 (H1) because, on average, managers are less likely to use opportunistic financial reporting discretions in the presence of local monitoring institutions. H2 predicts a 1 to be incrementally negative for firms with a greater investment opportunity set relative to firms with a lower investment opportunity set. Governance i,t represents a vector of variables that control for variations in the governance structures of the firms. We include: CEO i,t, the Chief Executive Officer s (CEO) ownership; Exec i,t, the sum of non-ceo executives ownership; IndDir i,t, the sum of all independent directors ownership; Indp i,t, the proportion of independent directors; and Gov i,t,a governance index derived from Gompers et al. (2003). We measure each of these governance control variables as of the end of year t. We also control for a vector of firm variables (Firm i,t ) that may cause variations in abnormal accruals. These variables include total assets at the end of year t, cash flow volatility measured from year t 10 through year t 1, sales volatility measured from year t 10 through year t 1, a loss dummy for year t, the operating cycle of year t (defined as the log of the sum of (360/(sales/average accounts receivable))þ(360/(cost of goods sold/average inventory))), market-to-book ratio at the end of year t (e.g., Hribar and Nichols, 2007; Doyle et al., 2007), and firm performance as measured by return on assets of year t. 15 Finally, we include the number of analyst following because security analysts may affect firms financial reporting discretion and are correlated with the level of institutional ownership Data sources and sample Our primary data source is the CRSP-Compustat Merged database from Our sample period begins in 1996 because RiskMetrics begins reporting governance variables in Our initial sample includes all firms with common 10 We find average adjusted R 2 s of 14% and 10% based on 26 firm location indicators in the models that predict local and distant institutions ownership, respectively. We consider the estimated R 2 s quite high for a broad cross-sectional regression. For example, Gasper and Massa (2007) find a R 2 of 6% for abnormal local mutual fund ownership. In addition, the F-values are 335 and 257 in these two models, respectively, which are substantially higher than 10, the value considered as a weak set of instruments (Staiger and Stock, 1997). 11 The Durbin Wu Hausman test for endogeneity (Wooldridge, 2002, p. 118) indicates the existence of endogeneity between the absolute value of abnormal accruals and local and distant monitoring institutional ownership. However, when absolute abnormal accruals is regressed on predicted value of local and distant monitoring institutional ownership, control variables, and endogenous residuals, the F-test fails to reject the null of no endogeneity (F¼1.13, p¼0.29). 12 In concurrent research, Kedia and Rajgopal (2008) find an association between financial misreporting and the firm s proximity to an SEC office having jurisdiction of the firm. Although ten of the 25 metropolitan locations we consider have an SEC office, the remaining 16 location indicators are exogenous. Nevertheless, as a robustness check, we include the distance between the firm s headquarters and the SEC office having jurisdiction of the firm as an additional control variable in our analyses. All our inferences remain unaltered. 13 More specifically, we first obtain the estimated coefficients from the following cross-sectional regression for each year and each two-digit SIC industry with at least ten observations in Compustat: Tacc it ¼l 0 þl 1 DSales it þl 2 PPE it þe it where Tacc is total accruals (annual Compustat data item #123 #308), DSales it is the change in the firm s sales (#12), and PPE is the firm s gross property, plant, and equipment (#7). We scale all variables by total assets at the end of year t 1. We then subtract nondiscretionary accruals from total accruals for the observations in our sample to calculate abnormal accruals: Aacc it ¼Tacc it [l 0 þl 1 (DSales it DAR it )þl 2 PPE it ], where DAR it is the change in accounts receivable (#302). 14 Performance does not cause biases in unsigned accrual tests because, for example, small and large absolute values of accruals consist of both good and poor performers. Nevertheless, in robustness analyses, we re-estimate our analyses using performance-matched abnormal accruals and find similar results. 15 Following prior earnings management literature (e.g., Kothari et al., 2005), we use return on assets to control for performance. Results are robust to using year t size-adjusted stock returns as an alternative control variable for performance.

7 B.C. Ayers et al. / Journal of Accounting and Economics 52 (2011) stocks in CRSP with share codes 10 or 11. To construct measures of ownership by geographic distance, we identify the locations of the firm s headquarters and of the institutions that hold the firm s shares. We obtain firms headquarters locations (city and states) from Compustat and the corresponding latitude and longitude coordinates from the US Census Bureau s Gazetteer Place and Zip Code Database. We obtain institutional stockholdings data from Thomson Reuters S34 Institutional Holdings database. To identify an institution s location, we manually match the name of the S34 institution with the names in the Nelson Directory of Investment Managers, which lists the locations of institutional investors. We use the US Census Bureau s Gazetteer Place and Zip Code Database to obtain the corresponding latitude and longitude coordinates of each institution. Following prior research (e.g., Coval and Moskowitz, 2001; Gasper and Massa, 2007), we calculate the distance between firm i and institution j to determine whether the institution is local or distant as follows: d i,j ¼ð2pr=360Þarccosfcosðlat i Þcosðlon i Þcosðlag j Þcosðlon j Þþcosðlat i Þsinðlon i Þcosðlag j Þsinðlon j Þþsinðlat i Þsinðlat j Þg ð2þ where d i,j is the kilometric distance between firm i and institution j, lat and lon are institution and firm latitudes and longitudes, and r is the radius of the earth (approximately 6,378 km). Institution j is defined as local with respect to firm i if the institution is located within a 100 km radius of firm i s headquarters (i.e., d i,j r100). Institution j is defined as distant with respect to firm i if the institution is located outside 100 km radius of firm i s headquarters (i.e., d i,j 4100). We identify 65,214 firm-year observations for our sample period (i.e., ) without missing firm or institution location data. After requiring one-year-ahead abnormal accruals data (9Aacc9 i,tþ1 ), we retain 64,559 firm-year observations over We then merge this dataset with the governance variables available from Execcomp and RiskMetrics, which results in 12,507 observations. 16 After deleting observations with missing data from Compustat and I/B/E/S to calculate firm control variables, our sample includes 10,695 firm-year observations. Table 1 shows descriptive statistics of key variables in our analysis of abnormal accruals. 17 We find that average institutional ownership for our sample is 59.3%. Average local monitoring institutional ownership is 1.4%, which is small compared to average distant monitoring institutional ownership of 18.4%. A relatively low local monitoring institutional ownership is not surprising, because the institution has to be within 100 km of the firm s headquarters. If none of the monitoring institutions reside within 100 km of the firm s headquarters, we set the value of Local_mon to zero. We also find that average transient institutional ownership (Trans) is 13.5%. We are interested in isolating the effects of transient institutions, because the short-term focus of these institutions may exert pressure on firms to manage earnings (Bushee, 1998; Graham et al., 2005). We use the variable Other to capture non-monitoring, non-transient institutional ownership, and the average value of Other in our sample is 25.9%. The distributions of our instrumental variables (plocal_mon and pdistant_mon) are similar to those of Local_mon and Distant_mon Regression results Table 2 presents the results of testing H1. We conduct all OLS regression tests using standard errors clustered at year and firm level. Consistent with H1, column (1) reports negative coefficients for local monitoring institutional ownership (¼ 0.268, po0.01, one-tailed), and the difference between local and distant monitoring institutional ownership is statistically significant (F¼8.52, po0.01, one-tailed). This evidence is consistent with our expectation that the lower cost of acquiring monitoring information reduces the information asymmetry between managers and local monitoring institutions, and managers, in turn, voluntarily reducing opportunistic financial reporting discretion. In column (2), we add transient institutional ownership and other institutional ownership to the regression model for completeness. Column (2) reports that the coefficient for plocal_mon is significantly negative (¼ 0.246, po0.01, one-tailed) and the difference between local and distant monitoring institutional ownership is statistically significant (F¼7.28, po0.01, one-tailed). To further strengthen our inferences that the presence of monitoring institutions deters financial reporting discretion that may be detrimental to the firm, we test whether the absolute abnormal accruals mitigated by monitoring institutions is negatively associated with two-year-ahead return on assets. Following Bowen et al. (2008), we run the following regression to disentangle whether observed absolute abnormal accruals in our sample reflect opportunism or managerial signaling: ROA i,t þ 2 ¼ b 0 þb 1 9Aacc9_Mon i,t þ 1 þb 2 9Aacc9_Other i,t þ 1 þb 3 LagROA i,t þ 1 þb 4 Logsale i,t þ 2 þb 5 Stdroa i,t þ 1 þe 2i,t ð3þ where ROA i,tþ 2 is the two-year-ahead return on assets, 9Aacc9_Mon i,tþ 1 is the predicted one-year-ahead absolute abnormal accruals impacted by monitoring institutions and estimated as the sum of the product of the predicted local and distant monitoring institutions ownership for firm i at the end of year t and the estimated coefficients for plocal_mon i,t and pdistant_mon i,t in column (1) of Table 2, 9Aacc9_Other i,tþ 1 is the predicted one-year-ahead absolute abnormal accruals associated with governance and estimated as the sum of the product of the estimated coefficients for CEO i,t, Exec i,t, IndDir i,t, Indep i,t,andgov i,t in column (1) of Table 2 and their corresponding variable values for firm i and year t, Logsale i,tþ2 is the log of sales, and Stdroa i,tþ1 is the standard deviation of return on assets. Consistent with the maintained assumption that financial 16 Requiring governance control variables substantially reduces the sample size. Results are stronger for the larger sample that does not require governance controls. Further, if we exclude the governance variables in the final sample of 10,695 observations, the results are similar to those reported in the tables, suggesting that the governance variables do not influence our analyses. 17 Throughout our analysis, we winsorize continuous variables at the 1% and 99% levels.

8 48 B.C. Ayers et al. / Journal of Accounting and Economics 52 (2011) Table 1 Descriptive statistics of 10,695 observations over Mean Std. dev. Percentile 5th 25th Median 75th 95th Abnormal accruals 9Aacc Aacc Institutional ownership (%) Total_inst Local_mon Distant_mon Total_mon Trans Other plocal_mon pdistant_mon Governance characteristics CEO Exec IndDir Indp Gov Firm characteristics Logasset Stdcfo Stdsale Neg Logcycle Mb Roa Analyst Aacc9 is the absolute value of abnormal accruals for year tþ1(t¼ ). Aacc is signed abnormal accruals. Total_inst is the total institutional ownership (as a percentage of shares outstanding) as of the end of year t. Local_mon is the ownership (as a percentage of shares outstanding) of monitoring institutions located within 100 km of the firm s headquarters, where monitoring institutions are top-five, long-term, independent, and dedicated/quasi-indexer institutions. Distant_mon is the ownership (as a percentage of shares outstanding) of monitoring institutions located beyond 100 km of the firm s headquarters. Total_mon represents total monitoring institutional ownership and equals the sum of Local_mon and Distant_mon. Trans is the ownership (as a percentage of shares outstanding) of transient institutions. Other is the ownership (as a percentage of shares outstanding) of non-monitoring and non-transient institutions. plocal_mon and pdistant_mon are the predicted values of Local_mon and Distant_mon, based on a regression including firm headquarters location indicator variables. CEO is the CEO ownership (as a percentage of shares outstanding) as of the end of year t per Execcomp. Exec is the sum of non-ceo executives ownership (as a percentage of shares outstanding) as of the end of year t per Execcomp. IndDir is the sum of all independent directors ownership (as a percentage of shares outstanding) as of the end of year t per RiskMetrics. Indp is the proportion of independent directors as of the end of year t per RiskMetrics. Gov is the governance index, defined as (24-Gindex)/24, where Gindex is the Gompers et al. (2003) governance index as of the end of year t per RiskMetrics. Logasset is the log of total assets at the end of year t. Stdcfo is the standard deviation of cash flows (scaled by assets) during years t 10 to t 1. Stdsale is the standard deviation of sales (scaled by assets) during years t 10 to t 1. Neg is the proportion of years with reported loss during years t 10 to t 1. Logcycle is the log of operating cycle of year t, estimated as 360/(sales/average AR)þ360/(cost of goods sold/average inventory). Mb is the market to book ratio at the beginning of year t. Roa is return on assets for year t. Analyst is the average number of analyst following for year t. reporting discretion leads to managerial rent extraction (opportunism), we find a significant negative association between predicted abnormal accruals eliminated in the presence of monitoring institutions and future firm performance. Likewise, when we estimate the regression separating the abnormal accruals mitigated in the presence of local institutions and distant institutions, we find a negative and significant association between predicted abnormal accruals mitigated in the presence of local monitoring institutions and future firm performance and a negative but insignificant association between predicted abnormal accruals mitigated in the presence of distant monitoring institutions and future firm performance. Consistent with Bowen et al. (2008), we do not find a negative coefficient for 9Aacc9_Other, suggesting that abnormal accruals associated common governance variables do not predict poorer future performance. As mentioned in Section 3.3, we follow prior literature that examines reporting discretion in a general setting and use absolute abnormal accruals as a proxy for financial reporting discretion. However, one may argue that overstating earnings may be more detrimental to the firm than understating earnings i.e., that monitoring investors are particularly concerned with income increasing reporting discretion. 18 To test this possibility, in columns (3) and (4) of Table 2, we present results 18 Also, Butler et al. (2004) argue that inferences from studies using absolute abnormal accruals may differ depending on whether the large absolute abnormal accruals are especially of one particular sign or symmetric. They suggest the use of signed abnormal accruals as a complementary measure of discretion.

9 B.C. Ayers et al. / Journal of Accounting and Economics 52 (2011) Table 2 The relation between absolute (or signed) abnormal accruals and institutional investors. (1) (2) (3) (4) 9Aacc9 9Aacc9 Aacc Aacc Constant (6.781) nnn (6.624) nnn (3.027) nnn (2.916) nnn plocal_mon ( 3.763) nnn ( 3.444) nnn ( 1.933) nn ( 1.894) nn pdistant_mon ( 4.032) nnn ( 3.661) nnn (0.942) (0.745) Trans (1.233) ( 2.390) nn Other ( 4.042) nnn (2.072) nn CEO (1.875) n (1.486) ( 1.084) ( 1.085) Exec (0.483) (0.207) (0.513) (0.498) IndDir ( 0.334) ( 0.167) (1.314) (1.240) Indep ( 0.965) ( 0.844) ( 0.821) ( 0.823) Gov ( 2.065) nn ( 2.366) nn ( 2.068) nn ( 1.817) n Logasset ( 3.174) nnn ( 2.590) nnn ( 0.668) ( 1.056) Stdcfo (4.881) nnn (4.814) nnn Stdsale (3.982) nnn (3.974) nnn Neg (1.400) (0.762) ( 4.764) nnn ( 4.102) nnn Logcycle ( 0.033) (0.095) Mb (0.891) (1.003) (0.680) (0.659) Roa ( 0.053) (0.028) (0.302) (0.396) Analyst ( 0.312) ( 0.032) ( 1.858) n ( 1.839) n SalesGrowth ( 7.994) nnn ( 7.620) nnn Debt (4.612) nnn (4.765) nnn plocal_monopdistant_mon (0.00) (0.01) (0.01) (0.01) Observations 10,695 10,695 10,675 10,675 Adjusted R Aacc9 is the absolute value of abnormal accruals for year tþ1 (t¼ ). Aacc is the signed value of abnormal accruals for year tþ1. Local_mon is the ownership (as a percentage of shares outstanding) of monitoring institutions located within 100 km of the firm s headquarters, where monitoring institutions are top-five, long-term, independent, and dedicated/quasi-indexer institutions. Distant_mon is the ownership (as a percentage of shares outstanding) of monitoring institutions located beyond 100 km of the firm s headquarters. plocal_mon and pdistant_mon are the predicted values of Local_mon and Distant_mon, based on a regression including firm headquarters location indicator variables. Trans is the ownership (as a percentage of shares outstanding) of transient institutions. Other is the ownership (as a percentage of shares outstanding) of non-monitoring and non-transient institutions. CEO is the CEO ownership (as a percentage of shares outstanding) as of the end of year t per Execcomp. Exec is the sum of non-ceo executives ownership (as a percentage of shares outstanding) as of the end of year t per Execcomp. IndDir is the sum of all independent directors ownership (as a percentage of shares outstanding) as of the end of year t per RiskMetrics. Indp is the proportion of independent directors as of the end of year t per RiskMetrics. Gov is the governance index, defined as (24-Gindex)/24, where Gindex is the Gompers et al. (2003) governance index as of the end of year t per RiskMetrics. Logasset is the log of total assets at the end of year t. Stdcfo is the standard deviation of cash flows (scaled by assets) during years t 10 to t 1. Stdsale is the standard deviation of sales (scaled by assets) during years t 10 to t 1. Neg is the proportion of years with reported loss during years t 10 to t 1. Logcycle is the log of operating cycle of year t, estimated as 360/(sales/average AR)þ360/(cost of goods sold/average inventory). Mb is the market to book ratio at the beginning of year t. Roa is return on assets for year t. Analyst is the average number of analyst following for year t. SalesGrowth is the sales of year t minus sales of year t 1, scaled by sales of year t 1. Debt is the ratio of long term debt to total assets at the beginning of year t. We calculate the t-statistics in parentheses using standard errors clustered at year and firm level. nnn, nn, and n indicate statistical significance at the 1%, 5%, and 10% levels, respectively. p-values are one-tailed for statistical tests with directional predictions (i.e., for plocal_mono0 and plocal_monopdistant_mon) and two-tailed otherwise. using signed abnormal accruals as the dependent variable. We expect that corporate managers are more likely to reduce income increasing opportunistic financial reporting discretion in the presence of local monitoring institutions than distant monitoring institutions. Consistent with this expectation, we find significant negative coefficients for plocal_mon in both