Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk

J. Account. Public Policy xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

J. Account. Public Policy journal homepage: www.elsevier.com/locate/jaccpubpol

Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk q Simon Yu Kit Fung a, K.K. Raman b, Lili Sun c,⇑, Li Xu d a

Hong Kong Polytechnic University, Hong Kong University of Texas at San Antonio, United States University of North Texas, United States d Washington State University, United States b c

a b s t r a c t In recent years, firms (and lawmakers) have sought to mitigate the dysfunctional effects of incentive-based executive compensation by adopting clawbacks. However, extant clawbacks (whether firm-initiated or as mandated by the 2010 Dodd–Frank Act) do not go far enough in that they seem to allow executives to retain trading profits linked to sales of their own companies’ shares at a time of inflated earnings (Fried and Shilon, 2011). In this paper, we examine the moderating effect of insider sales on the relation between firm-initiated clawback-adoptions and fraud risk. Our results indicate that clawback-adopting firms experience a decrease in fraud risk following adoption relative to non-adopters during the same time period. However, this decrease in fraud risk for the clawback-adopting firms is materially weakened in the presence of insider trading. At this time (July 2014), the SEC is still working on rules for implementing clawbacks (one of nearly half of the rules yet to be completed under Dodd–Frank). Our findings suggest that clawback rules (as and when issued by the SEC) need to address insider sales for clawbacks to be fully effective in mitigating the risk of fraudulent financial reporting. Published by Elsevier Inc.

q

Authors’ names are listed alphabetically.

⇑ Corresponding author at: 1155 Union Circle #305219, Denton, TX 76203, United States. Tel.: +1 732 979 5090. E-mail addresses: [email protected] (S.Y.K. Fung), [email protected] (K.K. Raman), [email protected] (L. Sun), [email protected] (L. Xu). http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002 0278-4254/Published by Elsevier Inc.

Please cite this article in press as: Fung, S.Y.K., et al. Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk. J. Account. Public Policy (2015), http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002

2

S.Y.K. Fung et al. / J. Account. Public Policy xxx (2015) xxx–xxx

1. Introduction In this paper, we investigate two issues related to clawback adoptions: (1) the effectiveness of clawback adoptions (i.e., firm-initiated adoptions of excess compensation recoupment provisions) in reducing the risk of fraudulent financial reporting (fraud risk), and (2) the impact of net insider sales of company stock on the relation between clawback adoptions and fraud risk. As background, in recent years firms have sought to mitigate the dysfunctional effects of incentive-based executive compensation by adopting clawback provisions. Specifically, by allowing firms to recoup excess incentive compensation (such as bonuses) in the event of a later restatement of previously issued financial statements, these provisions aim to both (1) ex ante deter fraudulent financial reporting, and (2) ex post penalize company executives who manipulate reported earnings. Consistent with this objective, recent studies (Chan et al., 2012; Dehaan et al., 2013) suggest that firms’ voluntary adoption of clawback provisions are associated with improved financial reporting quality. In our study, we provide a more complete understanding of the effectiveness of clawback adoptions by examining a context (insider sales) where these adoptions appear to have a limited or no effect. Specifically, we examine the moderating effect of actual realized insider trading on the relation between clawbacks and financial reporting quality as measured by the risk of fraudulent financial reporting (fraud risk). Bergstresser and Philippon (2006) argue that while the purpose of equity-based compensation is to increase managers’ exposure to company stock as a way of better aligning management incentives with shareholder interests, these equity incentives may also motivate managers to intentionally inflate reported earnings in an attempt at maintaining a high stock price and boost their trading profits while strategically reducing their net holdings of company stock. Specifically, Bergstresser and Philippon (2006) use discretionary accruals to proxy for earnings management and document a positive association between discretionary accruals and insider trading. Consistent with this finding, prior research (Fried, 2008; Summers and Sweeney, 1998) suggests that managers utilize inside information including knowledge that the firm is manipulating its reported earnings to increase their personal trading gains from insider sales of company stock. In particular, Summers and Sweeney (1998) use actual instances of financial reporting fraud to proxy for earnings management and find that net insider sales are higher in years of fraud occurrence. In our study, we assess the effectiveness of clawback adoptions by examining fraud risk using the scaled logistic probability F-score developed by Dechow et al. (2011). Basically, the F-score captures the likelihood of misstated earnings resulting from intentional misstatements (as opposed to errors) identified by the SEC in its Accounting and Auditing Enforcement Releases (AAERs). Utilizing the F-score (rather than actual instances of fraud) as a proxy for financial reporting quality allows us to increase the sample size for analyses and hence the generalizability of our findings. To the extent that the F-score is a fairly powerful predictor of fraud occurrence, the positive association between insider sales and financial reporting fraud documented in Summers and Sweeney (1998) serves as a useful foundation for our study. Consistent with Bergstresser and Philippon (2006) and Summers and Sweeney (1998), we define net insider sales as net sales (i.e., dollar sales minus dollar purchases) of company stock executed by the firm’s top managers, scaled by beginning-of-year firm equity value. Separately, Fried and Shilon (2011) suggest that extant clawback provisions – whether firm-initiated or as mandated by the 2010 Dodd–Frank Act – do not appear to go far enough in that they allow executives to keep the excess compensation (trading profit) arising from the unwinding of their equity incentives (shares) at a time of inflated earnings. They also suggest that given the complexities associated with estimating what the stock price would have been absent the earnings manipulation (and in determining the amount of excess sale proceeds), it could be difficult for firms to recover this particular form of excess compensation. Consistent with this view, a recent PwC (2014) survey of extant clawback policies notes the types of compensation that may be recouped but makes no mention of excess compensation from insider sales. Relatedly, Sprangler (2013) suggests that the potential benefits from trading on insider information likely exceed the costs of such trading (such as losing a few years’ worth of deferred compensation) in the event of clawback i.e., as a practical matter clawback provisions may not be an effective deterrent against insider trading. Hence, to the extent

Please cite this article in press as: Fung, S.Y.K., et al. Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk. J. Account. Public Policy (2015), http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002

S.Y.K. Fung et al. / J. Account. Public Policy xxx (2015) xxx–xxx

3

that clawback provisions allow managers to keep the excess compensation resulting from the sale of their company shares at inflated prices, insider sales may be expected to be associated with a weakening of the effectiveness of clawback provisions in mitigating fraud risk. To date, the weakening effect (if any) of insider sales on the relation between clawback stipulations and fraud risk remains an important yet unaddressed empirical question. We investigate this important question in our study. Our analysis is based on a sample of firms that voluntarily adopted clawbacks during 2003–2012 and a control sample of non-adopting firms matched based upon propensity of clawback provision adoption. We perform differences-in-differences (D-I-D) analysis to examine whether fraud risk (following clawback adoption) is lower for the clawback-adopting firms relative to the control firms, and whether this reduction in fraud risk (if any) is observable also for companies with insider sales. The D-I-D analysis is conducted based on 414 unique clawback-adopting firms as well as their propensity score-matched non-adopting firms for which we have complete data in both the pre- and post-adoption years for a total of 5285 firm-year observations. What do we find? Our results indicate that clawback-adopting firms experience a larger decrease in fraud risk following adoption compared to the decrease in fraud risk for non-adopters during the same time period. This result is consistent with Chan et al. (2012) and Dehaan et al. (2013) in that clawbacks appear to improve financial reporting quality. However, we find that the relative (differential) reduction in fraud risk for the clawback-adopting firms is materially lower in the presence of insider sales. In other words, the deterrent effect of clawback provisions in lowering fraud risk is materially weakened or essentially disappears in the presence of insider trading. Overall, our study extends and contributes to the literature on the effectiveness of clawback adoptions in improving financial reporting quality in at least two ways. First, our study finds clawback provisions to be effective in limiting the risk of intentional financial misstatements (fraud risk). Second, although prior research (Bergstresser and Philippon, 2006; Summers and Sweeney, 1998) suggests that insider sales provide incentives for earnings management and fraudulent financial reporting, extant clawback provisions (whether firm-initiated or as mandated by the 2010 Dodd–Frank Act) appear to allow executives to retain the excess pay linked to the unwinding of their equity positions at a time of inflated earnings (Fried and Shilon, 2011). In this study, we hypothesize and find that the effectiveness of clawbacks in reducing fraud risks is weakened (or disappears) in the presence of insider sales. Our study is important because it provides evidence relating to a topic that has seen limited empirical research yet remains an important public policy issue, i.e., whether extant clawback provisions are effective in limiting fraud risk. According to Ackerman and Zibel (2014), at this time (July 2014), the Dodd–Frank Act clawback mandate is yet to be implemented because the SEC is still working on the rules (nearly half of the rules yet to be completed under the 2010 Act). The policy implication of our findings is that clawback rules (currently under consideration by the SEC) need to address insider sales to be fully effective in mitigating the dysfunctional effects of incentive-based executive compensation. Our findings also complement the growing stream of research (Bao et al., 2015; Chan et al., 2015) that examines the unintended consequences associated with the adoption of clawback provisions. The rest of the paper is structured as follows: in Section 2 we discuss the institutional background, review relevant literature and develop our hypotheses. Section 3 discusses our sample and methodology, while Section 4 reports our empirical findings. Section 5 provides concluding remarks.

2. Background and hypotheses development 2.1. Background As discussed by Fried and Shilon (2011), a portion of the compensation that managers receive is in the form of incentive pay, i.e., bonuses and equity awards that are tied to achieving performance metrics such as the company’s earnings targets. In turn, incentive pay can result in excess pay in two separate ways. First, bonuses and equity (shares and options) may be awarded because the performance Please cite this article in press as: Fung, S.Y.K., et al. Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk. J. Account. Public Policy (2015), http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002

4

S.Y.K. Fung et al. / J. Account. Public Policy xxx (2015) xxx–xxx

metrics (e.g., reported earnings) were inflated. Absent the earnings manipulation, there could have been no (or fewer) shares and options awarded. Thus, the incentive pay (or a portion thereof) based upon inflated performance metrics may qualify as excess pay. Second, the payoff to executives from the unwinding of their equity incentives may be linked to inflated earnings at the time of the sale. For instance, a manager may obtain trading gains from selling stock (awarded in current or prior periods including those obtained from option exercise) in the current year. However, a portion (or all) of the trading profits may qualify as excess pay to the extent that the trading is at higher prices linked to inflated (i.e., manipulated) price-affecting metrics such as earnings.1 Nonetheless, it would be ‘‘difficult’’ to determine the amount of excess stock sale proceeds because it would involve the ‘‘complicated’’ task of determining what the stock price would have been absent the manipulation (Fried and Shilon, 2011, pp. 749–750). Section 304 of the 2002 Sarbanes–Oxley Act (SOX) enables the SEC to recover not only excess pay but all of the incentive pay (i.e., bonuses and equity awards) received by executives in the 12 months following the issuance of financial statements that are subsequently restated, provided the restatement is the result of misconduct. These SOX recovery provisions appear to be quite punitive in that the SEC could potentially recover any profits from the sale of stock within 12 months of the misleading financial statements. However, as noted by Fried and Shilon (2011), experience has shown that the SOX clawback provisions are very unlikely to be deployed for at least two reasons, substantially lowering their ex ante deterrent and ex post recovery effects. First, under SOX section 304, the clawback provision can be invoked only by the SEC and not by the firm’s board or shareholders. Second, the recovery is contingent on executive misconduct and it can be very expensive for the SEC to demonstrate such misconduct. Consequently, given its resource constraints, the SEC has deployed clawbacks only in rare instances such as where the executive has actually been convicted of fraud. Different from SOX, section 954 of the 2010 Dodd–Frank Act requires all publicly owned firms to have clawback provisions that can be invoked by the firm itself and not just the SEC.2 Further, under Dodd–Frank, recovery of excess pay can be triggered even in the absence of managerial misconduct, although a restatement of previously issued financial statements is required. However, Dodd–Frank does not appear to require clawback of excess pay (proceeds) linked to net insider sales of stock at a time of inflated earnings (Fried and Shilon, 2011, p. 747). Consequently, under Dodd–Frank ‘‘executives will still have the incentive to manipulate earnings before they dispose of large amounts of stock’’ (p. 749). 2.2. The impact of insider sales on the relation between clawback adoptions and fraud risk Despite the reforms mandated by SOX and Dodd–Frank, managers continue to retain significant discretion over the timing of the sale of their equity incentives including the shares obtained via exercise of their options. As noted by Fried (2008), the ability to unwind substantial amounts of equity over a short period creates a temptation (and the opportunity) to inflate the short-term stock price and boost trading profits. In fact, there is a substantial body of evidence to suggest that managers utilize inside information (including their inside knowledge that the firm is manipulating earnings) around stock sales to raise trading profits (e.g., Bartov and Mohanram, 2004; Beneish, 1999; Burns and Kedia, 2006; Carpenter and Remmers, 2001; Fried, 1998; Gimein, 2002; Ke et al., 2003; Summers and Sweeney, 1998). As examples, the CEO of Global Crossing and executives at Qwest purportedly sold more than $700 million and $2 billion of shares, respectively, in the year their firms were allegedly overstating revenues and earnings in an attempt at bolstering the stock price (Lerach, 2002; Sender and Blumenstein, 2002).3 Current regulations do attempt to restrain executives’ trading profits from their equity incentives. However, these regulations – the prohibition on insider trading (Rule 10b-5), SOX’s stock-trading 1 Note that excess pay (if not subsequently recovered) can harm shareholders in two ways: first, by diverting value from shareholders to the managers, and second, by destroying value as a by-product of the accounting manipulations aimed at generating the excess pay. 2 As of July 2014, the SEC is still working on the rules for implementing section 954. Further, this is only one of nearly half the rules yet to be completed in implementing Dodd–Frank. 3 For a discussion of executives’ ‘‘good luck’’ in trading their company stock, see Pulliam and Barry (2012, 2013).

Please cite this article in press as: Fung, S.Y.K., et al. Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk. J. Account. Public Policy (2015), http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002

S.Y.K. Fung et al. / J. Account. Public Policy xxx (2015) xxx–xxx

5

disclosure requirements, firms’ and executives’ voluntary restrictions on insider trading (trading windows and the so-called ‘‘10b5-1’’ plans) – are all porous and of limited effectiveness.4 Consequently, extant regulations fail to deter managers from using their private information (including knowledge that the firm is manipulating earnings) for extracting excess pay while unwinding their equity incentives (Fried, 2008). As noted previously, neither firm-initiated clawback stipulations nor section 954 of the 2010 Dodd–Frank Act appear to require the clawback of the excess pay linked to net insider sales of stock at a time of inflated earnings. Further, it would be difficult for firms to recover this particular form of excess compensation, given the complications associated with estimating what the stock price would have been absent the earnings manipulation and in determining the amount of excess share-sale proceeds (Fried and Shilon, 2011). A potential unintended consequence of omitting excess pay from insider sales of stock at a time of inflated earnings from clawback is that it could potentially hinder clawback provisions’ effectiveness in constraining managers’ earnings manipulation behavior prior to disposing off their shares and/or indeed increase managerial incentives for such manipulation. As a test of this potential unintended consequence, we analyze the moderating effect of insider sales on the relation between clawback adoptions and fraud risk. Prior clawback research (Chan et al., 2012; Dehaan et al., 2013) has examined the effect of clawback adoptions on financial reporting quality using various proxies including financial restatements, earnings response coefficients (as a proxy for investors’ perception of earnings quality), frequency of meeting or beating analyst forecasts, and auditor response. Their findings suggest that clawback adoption is associated with fewer restatements, larger earnings response coefficients, reduced meet-or-beat behavior, as well as reduced audit effort implied by lower fees and shorter audit lags. However, prior studies did not consider the role of insider sales in mitigating the effect of clawbacks on financial reporting quality. Further, among the various earnings quality proxies examined in the prior literature, restatement is perhaps the most direct measure of factual earnings quality. Although all restatements indicate the presence of material misstatements (and therefore poor financial reporting quality), restatements due to fraud or intentional misstatements (as opposed to error or unintentional misstatements) are often associated with more severe consequences (Hennes et al., 2008). Further, restatements are incomplete in the sense that they only include misstatements that are both discovered and admitted to by companies (Srinivasan et al., 2015). Also, restatements are indicative of ‘‘black and white’’ (i.e., discrete) reporting failures and do not capture subtle variations in financial reporting quality (DeFond and Zhang, 2014). For these reasons, in our study we utilize the FSCORE which represents a continuous measure of the risk of fraudulent financial reporting (Dechow et al., 2011), as our metric to empirically examine the relation between clawback adoptions and financial reporting quality. In our first hypothesis H1 (stated below in the alternative form), to be consistent with prior clawback research

4 First, Rule 10b-5’s ability to prevent executives from trading on insider knowledge is limited in practice by the courts’ high threshold of materiality. Further, the rule is difficult to enforce and is aggravated by the fact that the SEC has limited resources most of which are not allocated to preventing executives’ insider trading. Second, although SOX now requires managers to report to the SEC by the end of the second day following a trade, this still allows executives to sell secretly for two days prior to revealing their trades and before any correction to the stock price following the disclosure of their trades. Third, firms have adopted ‘‘trading windows’’ which restrict the times an executive can buy or sell shares. For example, managers may be permitted to trade only during a two- or three-week window after the quarterly earnings announcement. Nonetheless, these rules allow executives to profit from inside information during the time when the trading window is open. In other words, if executives anticipate selling considerable amounts of stock during an upcoming window, they do have the opportunity to manipulate reported earnings prior to the opening of the trading window. Finally, the SEC has a safe harbor plan (‘‘10b5-1’’ plan) which allows the executive to transact on material nonpublic information if the trade is effected according to a plan drawn-up at a time when the executive did not possess material nonpublic information. However, managers often have knowledge pertinent to the stock price long before the information emerges, and it would be difficult for the SEC to prove that the executive had the information at the time the plan was drawn-up months before the information became public. Further, as noted previously, the materiality threshold can be quite high. Potentially, managers intending to sell on inside information could even set up 10b5-1 plans to do so, in order to camouflage their trading on private information and to reduce the risk of legal liability. In any event, managers who have already scheduled large share sales under these plans, can attempt to manipulate earnings to obtain a higher price for their stock (Fried, 2008; Siconalfi and Eaglesham, 2013).

Please cite this article in press as: Fung, S.Y.K., et al. Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk. J. Account. Public Policy (2015), http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002

6

S.Y.K. Fung et al. / J. Account. Public Policy xxx (2015) xxx–xxx

(Chan et al., 2012; Dehaan et al., 2013), we do not consider the role of insider sales in the relation between clawbacks and financial reporting quality: H1. There is a negative relation between clawback adoptions and the risk of fraudulent financial reporting. As discussed previously, the excess pay from insider sales of stock at a time of inflated earnings does not appear to be subject to recoupment under extant clawback provisions (Fried and Shilon, 2011; PwC, 2014). In other words, extant clawback provisions are unlikely to deter earnings manipulations to the extent that executives can continue to earn excess pay via abnormal trading profits on insider sales at a time of inflated price-affecting metrics such as earnings. As such, we expect the deterrent effect of clawback provisions on fraud risk to be weakened (or disappear) in the presence of insider sales. In our second hypothesis H2, also stated in the alternative form, we take into account the role of insider sales in the relation between clawbacks and financial reporting quality: H2. The magnitude of the negative relation between clawback adoptions and the risk of fraudulent financial reporting is smaller in the presence of insider sales.

3. Sample, models and variables 3.1. Sample and data Our sample period spans the post-SOX time period 2003–2012. We obtain firms’ clawback adoption status from GMI ratings, and manually identify the year of adoption by reading proxy statements. We require our sample firms to have data available for the construction of the FSCORE, insider sales, and the control variables in our analysis. We obtain financial accounting data from Compustat, insider sales and purchases data from Thomson Reuters, and corporate governance data from RiskMetrics. We obtain data on restatements (used as a control variable) from Professor Andrew Leone (http://sbaleone.bus.miami.edu) and Audit Analytics. Because accruals is an important element in the calculation of the FSCORE, we exclude financial services firms since the nature of their accruals is different from that of non-financial firms. We also exclude utilities firms because their accruals could be different from other firms due to regulatory reasons. These procedures yield a sample of 414 firms with voluntary clawback adoptions (2968 firm-year observations) during 2003–2012. Panel A of Table 1 provides clawback adopters’ sample distribution by adoption year. Very few firms adopt clawbacks in the early years (2003–2005) of our sample period and the number of first-year adopters range from 35 to 83 in the later years (2006–2012). Table 1 Panel B reports industry membership based on SIC classifications for the clawback adopting firms in reference to all Compustat firms in the sample period. Industry composition for clawback firms is fairly similar to that of the Compustat population with some variations. For instance, the Computers industry constitutes the highest percentage of clawback adopting firms as well as the general Compustat population. In all our regression analyses (discussed below), we control for industry effects. 3.2. Construction of propensity score-matched control sample Since clawback adoption during our study period (2003–2012) is voluntary, we use propensity score-matching to identify a control sample of non-adopting firms with a similar propensity (as the adopting firms) to adopt clawback provisions. As noted by Lawrence et al. (2011), this technique is useful in controlling for endogenous firm characteristics that may potentially affect the relationship

Please cite this article in press as: Fung, S.Y.K., et al. Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk. J. Account. Public Policy (2015), http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002

7

S.Y.K. Fung et al. / J. Account. Public Policy xxx (2015) xxx–xxx Table 1 Sample distribution. Number clawback-adopters

% Sample

Panel A: Clawback-adopters by clawback adoption year 2003 2 2004 0 2005 4 2006 38 2007 35 2008 79 2009 72 2010 83 2011 44

0.48 0.00 0.97 9.18 8.45 19.08 17.39 20.05 10.63

2012

57

13.77

Total

414

100.00

SIC

Number

% Sample

Panel B: Clawback-adopters by industry membership (excluding utilities and financial industries) NOT ASSIGNED 0 0.00 MINING/CONSTRUCTION 1000–1999 0 0.00 FOOD 2000–2111 14 3.38 TEXTILES/PRINT/PUBLISH 2200–2780 22 5.31 CHEMICALS 2800–2824; 23 5.56 2840–2899 PHARMACEUTICALS 2830–2836 25 6.04 EXTRACTIVE 2900–2999; 17 4.11 1300–1399 MANF_RUBBER/GLASS/ETC. 3000–3299 5 1.21 MANF_METAL 3300–3499 19 4.59 MANF_MACHINERY 3500–3599 31 7.49 MANF_ELECTRICALEQPT 3600–3699 22 5.31 MANF_TRANSPORTEQPT 3700–3799 13 3.14 MANF_INSTRUMENTS 3800–3899 33 7.97 MANF_MISC 3900–3999 3 0.72 78 18.84 COMPUTERS 3570–3579; 3670–3679; 7370–7379 TRANSPORTATION 4000–4899 10 2.42 RETAIL_WHOLESALE 5000–5199 25 6.04 RETAIL_MISC 5200–5999 31 7.49 RETAIL_RESTAURANT 5800–5899 6 1.45 7000–8999 SERVICES 8.94 37 TOTAL

414

100.00

% Compustat 1.54 3.25 2.71 4.96 2.67 7.55 4.78 2.02 2.84 3.59 5.16 2.33 7.03 1.08 18.48

8.14 3.95 5.69 1.66 10.57 100.00

The underlined terms represent sum numbers or the last numbers to be summed.

between the dependent variable (FSCORE) and the independent variables of interest (POSTCB, and POSTCB  INSDSALE) discussed later in the paper. See Appendix A for variable definitions. We estimate a clawback-adoption Probit model based on several factors known to be associated with clawback adoptions (Addy et al., 2014; Chan et al., 2012): firm size, leverage, return on assets, financial distress measured by a revised version of Altman Z-score following Graham et al. (2008), market to book ratio, restatement (measured over a five-year window preceding the clawback period), board of director independence, board size, audit committee size, presence of interlocking director, absolute value of performance adjusted abnormal accruals calculated following Kothari et al. (2005), CEO duality, number of business segments and number of geographical segments, and

Please cite this article in press as: Fung, S.Y.K., et al. Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk. J. Account. Public Policy (2015), http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002

8

S.Y.K. Fung et al. / J. Account. Public Policy xxx (2015) xxx–xxx

institutional holdings.5,6 Appendix B provides the estimation results of the clawback adoption model. All non-adopters with an estimated propensity to adopt within a 20% distance of the estimated propensity of clawback adopters are included in the control sample.7 We allow a control firm to be matched with multiple clawback-adopting firms (i.e., we match with replacement).8 Each non-adopter is assigned an assumed adoption date identical to its matched adopter’s actual adoption date. Firm-year observations prior to (after) the assumed adoption dates are considered as pre-adoption (post-adoption) observations. This procedure yields a propensity score-matched control sample consisting of 1321 non-adopter firm-year observations for the pre-adoption period and 996 non-adopter observations for the post-adoption period, representing 326 unique non-adopter firms. 3.3. Model for testing Hypothesis H1 To test hypothesis H1 which examines the effect of clawback adoptions on fraud risk, we conduct a differences-in-differences (D-I-D) analysis. Our model specification primarily follows Chan et al. (2012) and is as follows:

FSCORE ¼ b0 þ b1 CB þ b2 POSTCB þ b3 POST þ ak Controls þ Year and Industry Dummies þ error

ð1Þ

where ak Controls = b4 LNSIZE + b5 MTB + b6 ZSCORE + b7 LEV + b8 ROA + b9 INDLIT + b10 BDIND + b11 INTERLOCK + b12 CEODUAL + b13 RESTATE + b14 ANALYST + b15 INSTHLD + b16 NBSEG + b17NGSEG. In model (1), the dependent variable FSCORE is the fraud risk index developed by Dechow et al. (2011). The FSCORE is a scaled logistic probability metric that incorporates variables including the accruals measure from Richardson et al. (2005), change in receivables, change in inventory, percent of soft assets, change in cash sales, change in return on assets, and actual issuance of securities. Appendix A provides details of FSCORE calculation. The FSCORE can range from 0 to 4+, with a higher FSCORE representing a higher likelihood of financial reporting fraud (Dechow et al., 2011, p. 63).9 As noted previously, using the FSCORE allows us to focus on the effectiveness of clawback provisions in constraining intentional misstatements rather than muddying the waters with unintentional misstatements (errors). It also allows us to examine the risk of intentional misstatements rather than just those misstatements that were discovered and admitted to by firms (Srinivasan et al., 2015). Further, the 5 The clawback adoption model is estimated using a sample of 4216 firm-year observations of 635 clawback-adopting firms and 3522 firm-year observations of 553 control firms, all with the required data for our sample period (see Appendix B). Results suggest that the propensity for clawback adoption is positively associated with firm size, board independence, and board size. The model’s Pseudo R2 is 14.7%, comparable to what is reported in prior literature. For instance, Addy et al. (2014)’s model has an R2 of 18%. Potentially, future research could address other factors influencing firms’ voluntary clawback adoption decisions. 6 Addy et al. (2014) develop a corporate governance index based on 13 components including CEO duality, family business, newly appointed CEO following a restatement, insider ownership, dual class of ownership, and several other variables, and find that the Index is positively associated with the likelihood of voluntary clawback adoption. In our propensity matching model, we included a number of the variables suggested in Addy et al. (2014), such as CEO duality, board independence, and the presence of interlocking director. In a robustness test, we construct a similar index based on a smaller set of index components (we do not have data to include three out of the 13 components, including an indicator on the presence of founders, an indicator of family business, and an indicator of firms with secret ballot). We include this index in the propensity matching model and redo the analyses, and we find that our inferences remain the same as those reported later in the paper. We do not include this index in our main analyses because about 52% of our sample firms do not have the necessary data to compute the governance index, and the smaller sample materially lowers the power of our tests. 7 In alternative analyses, we use a 10% distance and obtain results (untabulated) similar to what is reported in the paper, although the stricter matching criterion yields a much smaller sample size. Separately, prior studies (Chan et al., 2012; Dehaan et al., 2013) select control firms based on the closest predicted probability of clawback adoption. We perform additional analysis by forming the control sample based on the closest propensity and our inferences remain the same. 8 The ‘‘with replacement’’ control sample more closely resembles the treatment sample but biases standard errors which we address by clustering standard errors by firm. As discussed later in the paper, for all our analyses we cluster standard errors by firm and also by year. Further, our inferences remain unchanged when we match ‘‘without replacement’’ although the sample size is smaller, i.e., several clawback firms have to be dropped from the analysis because no proper matches are available for these treatment firms under the ‘‘without replacement’’ matching rule. 9 As an example, Enron’s FSCORE for year 2000 was 2.76. We note that the mean (median) FSCORE for our sample is 1.06 (0.98) and Enron’s FSCORE is at the 95th percentile of our sample distribution.

Please cite this article in press as: Fung, S.Y.K., et al. Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk. J. Account. Public Policy (2015), http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002

S.Y.K. Fung et al. / J. Account. Public Policy xxx (2015) xxx–xxx

9

F-score metric is developed by examining AAER firm-years with overstated earnings (Dechow et al., 2011, p. 19), consistent with our intent of examining the efficacy of clawbacks in the presence of net insider sales associated with inflated earnings. Finally, the FSCORE can be calculated for a large sample of firms to assess the likelihood of an intentional misstatement (Dechow et al., 2011, p. 22). By contrast, focusing on firms with actual restatements linked to intentional misstatements can significantly limit sample size. In model (1), CB equals 1 for clawback adopters, and 0 for non-adopters; POST equals 1 for firm-year of adoption and later years, and 0 for firm-years prior to adoption.10 The study variable of interest, POSTCB, equals 1 for clawback adopters in firm-years when they have clawback provisions in place, 0 otherwise. The coefficient on CB represents the difference in FSCORE for clawback adopters and matched non-adopters in the pre-adoption period. The coefficient on POSTCB measures the difference in the change in FSCORE (from the pre- to the post-adoption time period) for the adopters relative to the change in the FSCORE for the matched non-adopters. A negative coefficient for POSTCB would provide support for H1, that is, relative to the non-adopters, clawback adopters experience a larger reduction in fraud risk following adoption. Also in model (1), the control variables are drawn from the prior literature on fraud and misreporting. These variables include ZSCORE (a measure of bankruptcy risk based on Altman, 1968 and revised by Graham et al., 2008; the higher the ZSCORE, the lower the risk of bankruptcy), ROA (return on assets), LEV (leverage), and MTB (market to book ratio). Poor financial performance and solvency problems incentivize managers to commit fraud (Beasley, 1996; Loebbecke et al., 1989; Summers and Sweeney, 1998). Hence, as leverage (LEV) increases or ZSCORE decreases, the risk of fraudulent financial reporting is expected to increase. Since stock price declines following negative earnings surprises can increase litigation risk (Ali and Kallapur, 2001; Bowen et al., 1995; Francis et al., 1994), the higher the litigation risk INDLIT (an industry-based litigation dummy) to begin with, the greater the risk of fraud. Weak governance allows more opportunity for fraud and increases fraud risk (Beasley, 1996). We employ several proxies to capture strength of corporate governance: BDIND (dummy variable indicating an independent board), INTERLOCK (dummy variable indicating an interlocked board), CEODUAL (CEO duality), ANALYST (analyst following), and INSTHOLD (institutional holdings). We expect BDIND, ANALYST, and INSTHOLD to be negatively associated with fraud risk and INTERLOCK and CEODUAL to be positively associated with fraud risk. Because accounting complexity can provide opportunities for intentional earnings manipulation, we expect fraud risk to be higher for firms with complex operations measured by the number of business segments (NBSEG) and the number of geographical segments (NGSEG). We also control for firm size (LNSIZE), profitability (ROA), growth (MTB), and prior restatements (RESTATE), although we do not predict the sign of the relation with the dependent variable FSCORE. Finally, model (1) – as well as model (2) below – includes industry dummies (defined as the two-digit SIC code) and year dummies to control for fixed industry and year effects. Standard errors for coefficients in all regression models included in our main and additional analyses are corrected for serial correlation within firms and for cross-sectional dependence across firms through clustering by firm and year, respectively (Petersen, 2009). 3.4. Model for testing Hypothesis H2 To test H2 (which hypothesizes a weaker or no relation between clawback adoptions and the FSCORE in the presence of insider sales), we augment model (1) by adding an insider sales (INSDSALE) variable and its interactions with CB, POST, and POSTCB:

FSCORE ¼ b0 þ b1 CB þ b2 POSTCB þ b3 POST þ b4 INSDSALE þ b5 POST  INSDSALE þ b6 CB  INSDSALE þ b7 POSTCB  INSDSALE þ ak Controls þ Year and Industry Dummies þ error

ð2Þ

Consistent with Bergstresser and Philippon (2006), INSDSALE is defined as net insider sales (i.e., top managers’ dollar sales less dollar purchases of company stock) scaled by the firm’s beginning-of-year market value of equity, multiplied by 1000. Top managers refer to the five most highly compensated 10

As noted previously, each control firm is assigned an artificial adoption year that is identical to its matched study firm.

Please cite this article in press as: Fung, S.Y.K., et al. Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk. J. Account. Public Policy (2015), http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002

10

S.Y.K. Fung et al. / J. Account. Public Policy xxx (2015) xxx–xxx

Table 2 Descriptive statistics. CB = 0

CB = 1

Mean difference test

Median difference test

Mean

Median

Mean

Median

Difference

Difference

FSCORE INSDSALE LNSIZE MTB ZSCORE LEV ROA INDLIT BDIND INTERLOCK CEODUAL RESTATE ANALYST INSTHLD NBSEG

1.040 0.690 6.980 2.060 2.060 0.150 0.068 0.280 0.840 0.004 0.570 0.240 11.300 0.790 1.860

0.940 0.330 6.800 1.730 2.040 0.110 0.069 0.000 1.000 0.000 1.000 0.000 9.000 0.840 1.950

1.090 0.570 7.820 1.960 2.090 0.190 0.067 0.250 0.910 0.007 0.640 0.220 13.900 0.810 2.050

1.030 0.170 7.690 1.630 2.000 0.180 0.066 0.000 1.000 0.000 1.000 0.000 12.000 0.840 2.300

0.050*** 0.120*** 0.840*** 0.100*** 0.030 0.040*** 0.001 0.030* 0.070*** 0.004 0.070*** 0.020 2.600*** 0.020** 0.190***

0.090*** 0.160*** 0.890*** 0.100*** 0.040*** 0.070*** 0.003* 0.000 0.000 0.000 0.000 0.000 3.000*** 0.000 0.350***

NGSEG

2.140

2.300

2.230

2.300

0.090***

0.000

# OF OBSERVATIONS

2317

5285

5285

2968

Note: The table reports univariate analysis results for 2968 clawback-adopter firm-year observations (CB = 1) and 2317 propensity score-matched non-clawback-adopter firm-year observations (CB = 0). Differences in means and medians are tested based on t-tests and Wilcoxon rank tests, respectively. Variables are defined in Appendix A. The underlined terms represent sum numbers or the last numbers to be summed. * Two-tailed significance at 10% level. ** Two-tailed significance at 5%, level. *** Two-tailed significance at 1% level.

executives in the Execucomp database. The prior literature finds that net insider sales are positively associated with earnings management and the likelihood of fraudulent financial reporting (Bergstresser and Philippon, 2006; Summers and Sweeney, 1998), consistent with the notion that equity incentives can incentivize insiders to fraudulently inflate reported earnings in an attempt at maintaining a high stock price while strategically reducing their net holdings of company stock. Hence, the expected sign for INSDSALE in model (2) is positive. Further, the coefficient for POSTCB (b2) captures the effectiveness of clawback adoptions in the absence of insider sales. A negative b2 would be consistent with the expectation that in the absence of insider sales, clawback adopters experience a larger reduction in fraud risk than non-adopters. The test variable POSTCB  INSDSALE examines whether insider sales affect the relation between clawback adoptions and fraud risk. A positive coefficient on POSTCB  INSDSALE (b7) would be consistent with H2, our hypothesis that insider sales weaken the effectiveness of clawback adoptions in mitigating fraud risk.11 The sum of the coefficients b2 and b7 indicates whether in the presence of insider sales, the clawback adopters experience a differential amount of change in post-adoption fraud risk relative to the non-adopters. The control variables in model (2) are the same as those in model (1) discussed previously. 4. Empirical results 4.1. Univariate results Table 2 provides descriptive statistics for the clawback-adopter firm-years (N = 2968) and the propensity score-matched control firm-years (N = 2317). We find that clawback firm-years are on average larger (LNSIZE), more leveraged (LEV), more complex (NBSEG and NGSEG), less litigious 11 Clawback adoption could be part of a broader plan and a larger commitment on the part of the board to improve financial reporting integrity (Denis, 2012). To the extent that a stronger governance structure reduces the risk of fraudulent reporting, we could observe clawback adopting firms to be associated with a lower risk of fraudulent reporting regardless of the level of insider trading. This would bias against us finding results.

Please cite this article in press as: Fung, S.Y.K., et al. Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk. J. Account. Public Policy (2015), http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002

11

S.Y.K. Fung et al. / J. Account. Public Policy xxx (2015) xxx–xxx Table 3 Pairwise correlations.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

FSCORE CB POSTCB POST INSDSALE LNSIZE MTB ZSCORE LEV ROA INDLIT BDIND INTERLOCK CEODUAL RESTATE ANALYST INSTHLD NBSEG NGSEG

10 11 12 13 14 15 16 17 18 19

ROA INDLIT BDIND INTERLOCK CEODUAL RESTATE ANALYST INSTHLD NBSEG NGSEG

1

2

3

4

5

6

7

8

9

1.00 0.05*** 0.03* 0.02 0.02 0.09*** 0.13*** 0.05*** 0.11*** 0.04** 0.07*** 0.02 0.01 0.02 0.02 0.01 0.13*** 0.06*** 0.12***

1.00 0.49*** 0.01 0.07*** 0.31*** 0.04** 0.01 0.13*** 0.00 0.04* 0.10*** 0.02 0.07*** 0.07*** 0.02 0.12*** 0.04** 0.15***

1.00 0.65*** 0.10*** 0.27*** 0.12*** 0.01 0.10*** 0.01 0.03 0.15*** 0.03* 0.05*** 0.10*** 0.00 0.01 0.06*** 0.10***

1.00 0.11*** 0.17*** 0.14*** 0.03* 0.07*** 0.02 0.02 0.19*** 0.04** 0.13*** 0.11*** 0.04** 0.07*** 0.07*** 0.05***

1.00 0.14*** 0.23*** 0.14*** 0.14*** 0.23*** 0.09*** 0.09*** 0.02 0.01 1.00 0.00 0.02 0.09*** 0.09***

1.00 0.12*** 0.12*** 0.31*** 0.04** 0.07*** 0.08*** 0.01 0.13*** 0.06*** 0.46*** 0.03 0.20*** 0.16***

1.00 0.24*** 0.28*** 0.52*** 0.08*** 0.05*** 0.01 0.01 0.08*** 0.24*** 0.01 0.17*** 0.05***

1.00 0.34*** 0.56*** 0.02 0.09*** 0.01 0.01 0.05*** 0.06*** 0.06*** 0.04** 0.13***

1.00 0.24*** 0.17*** 0.04** 0.01 0.06*** 0.03 0.07*** 0.00 0.11*** 0.07***

10

11

12

13

14

15

16

17

18

19

1.00 0.00 0.05** 0.02 0.00 0.06*** 0.17*** 0.08*** 0.08*** 0.01

1.00 0.02 0.03* 0.01 0.08*** 0.14*** 0.05*** 0.17*** 0.14***

1.00 0.07*** 0.01 0.06*** 0.04** 0.14*** 0.02 0.06***

1.00 0.04** 0.01 0.01 0.07*** 0.04** 0.02

1.00 0.05*** 0.06*** 0.07*** 0.09*** 0.02

1.00 0.03* 0.05*** 0.01 0.05***

1.00 0.06*** 0.08*** 0.04**

1.00 0.11*** 0.00

1.00 0.20***

1.00

Note: This table reports Pearson correlations for the sample of 5285 firm-year observations used in the differences-in-differences analysis. Variables are defined in Appendix A. * Two-tailed significance at 10% level. ** Two-tailed significance at 5% level. *** Two-tailed significance at 1% level.

(INDLIT), have more independent boards (BDIND), larger analyst following (ANALYST) and higher institutional holdings (INSTHLD) when compared to the control group, but they are also associated with lower growth opportunities (MTB), greater CEO duality (CEODUAL) and lower insider sales (INSDSALE). Interestingly, FSCORE is on average higher for firms with clawback provisions than the control firms. Non-parametric test results for the median difference are by and large consistent with those for the mean difference. Table 3 presents Pearson correlations for the variables in our D-I-D analysis. Although several of these correlations are significant (greater than 0.35), the VIF scores suggest that collinearity is not an issue in the interpretation of our multivariate regression results discussed below.12 4.2. Multivariate test results for H1 and H2 based on differences-in-differences analysis Table 4 presents the results for models (1) and (2) using a differences-in-differences (D-I-D) approach with a sample of 5285 firm-year observations over the ten-year period 2003–2012. The 12 In the D-I-D analysis (Table 4), the highest VIF for any independent variable was only 2.02 (for ROA). In the time series analysis (Table 5), although the highest VIF for any independent variable was a relatively high 9.85 (for the LAMBDA ratio), excluding LAMBDA from the regression does not change our inferences.

Please cite this article in press as: Fung, S.Y.K., et al. Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk. J. Account. Public Policy (2015), http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002

12

S.Y.K. Fung et al. / J. Account. Public Policy xxx (2015) xxx–xxx

Table 4 Regression results: differences-in-differences analysis. Variables

Expected sign

CB POSTCB POST INSDSALE POST  INSDSALE CB  INSDSALE POSTCB  INSDSALE LNSIZE MTB ZSCORE LEV ROA INDLIT BDIND INTERLOCK CEODUAL RESTATE ANALYST INSTHLD NBSEG NGSEG INTERCEPT OBSERVATIONS ADJ R-SQUARED F-test of sum of coefficients: POSTCB + POSTCB  INSDSALE

? H1 () ? + ? ? H2 (+) ? ?  + ? +  + + ?   + +

Model (1)

Model (2)

Coefficient

t-Stat

Coefficient

t-Stat

0.002 0.050** 0.037*

0.07 2.30 1.66

0.064*** 0.046*** 0.011 0.295*** 0.192 0.127*** 0.014 0.042 0.033 0.004 0.004*** 0.140*** 0.039* 0.01 0.700***

5.22 5.18 0.64 3.14 1.28 2.77 0.43 0.77 1.62 0.19 3.61 2.70 1.92 0.47 5.18

0.031 0.090*** 0.045 0.041** 0.001 0.046* 0.066*** 0.064*** 0.049*** 0.013 0.304*** 0.134 0.119*** 0.012 0.039 0.032 0.004 0.004*** 0.129** 0.039* 0.01 0.670***

0.93 3.67 1.44 2.06 0.03 1.92 3.03 5.43 5.35 0.76 3.27 0.89 2.58 0.38 0.71 1.59 0.18 3.58 2.51 1.92 0.44 5.02

5285 0.24

5285 0.25 p-value = 0.322

Model (1)

FSCORE ¼ b0 þ b1 CB þ b2 POSTCB þ b3 POST þ ak Controls þ Year and Industry Dummies þ error Model (2)

FSCORE ¼ b0 þ b1 CB þ b2 POSTCB þ b3 POST þ b4 INSDSALE þ b5 POST  INSDSALE þ b6 CB  INSDSALE þ b7 POSTCB  INSDSALE þ ak Controls þ Year and Industry Dummies þ error Note: The differences-in-differences analysis examines the change in the FSCORE for clawback-adopters – relative to the change in the FSCORE for non-adopters – from the pre- to the post-adoption time periods, for our sample of 2968 firm-year observations of 414 adopters and 2317 firm-year observations of 326 propensity score-matched non-adopters. All variables are defined in Appendix A. Year dummies and industry dummies are included in each specification but are not reported for brevity. All p-values are based on standard errors adjusted for firm-level and year-level clustering. * Two-tailed significance at 10% level. ** Two-tailed significance at 5% level. *** Two-tailed significance at 1% level.

sample includes 2968 and 2317 firm-year observations for 414 clawback adopters and 326 propensity score-matched non-adopters, respectively.13 As reported in Table 4 column 1, variable CB is not significant indicating that there is no significant difference in the FSCORE between the treatment firms and the control firms during the pre-clawback-adoption time period. By contrast, the coefficient on POSTCB is significantly negative (0.050, t-stat = 2.30), which suggests that following clawback adoption the FSCORE of the treatment firms is significantly lower (after controlling for the change in FSCORE of the control firms over the same time period). This decrease in fraud risk following clawback adoption is 4.7% of the sample

13 The final sample used for the D-I-D analysis is smaller than that used for the clawback adoption model (Appendix B) due to different data requirements and loss of clawback firms for which we cannot identify proper matches.

Please cite this article in press as: Fung, S.Y.K., et al. Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk. J. Account. Public Policy (2015), http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002

S.Y.K. Fung et al. / J. Account. Public Policy xxx (2015) xxx–xxx

13

average FSCORE, which is economically significant.14 Moreover, the sum of the coefficients for CB and POSTCB is significantly negative [(0.002) + (0.050) = 0.052, p-value < 0.05], which suggests that the clawback adopting firms have a lower FSCORE (i.e., a lower likelihood of fraudulent financial reporting) than the control firms in the post-adoption period. Collectively, our results suggest that clawback adoptions lower the risk of fraudulent financial reporting, consistent with our hypothesis H1. As discussed previously, to test hypothesis H2 we augment model (1) by including the variable net insider sales (INSDSALE) and its interaction terms with CB, POST and POSTCB in model (2). The results for model (2) are presented in column (2) of Table 4. Here, the coefficient on POSTCB continues to be negative and significant (0.090, t-stat = 3.67), consistent with the prediction of hypothesis H1 discussed previously. In addition, the coefficient on INSDSALE is positive and significant (0.041, t = 2.06) indicating that in the pre-clawback-adoption time period there is a positive and significant association between INSDSALE and FSCORE. This particular finding is consistent with prior research (Bergstresser and Philippon, 2006; Summers and Sweeney, 1998) which documents that insider sales are associated with higher levels of earnings management. In Table 4 column (2), the coefficient on POSTCB  INSDSALE captures the association between INSDSALE and FSCORE in the post-clawback-adoption time period in a D-I-D setting. The results show that the coefficient on POSTCB  INSDSALE is significant and positive (0.066, t-stat = 3.03), suggesting that the positive association between INSDSALE and FSCORE is stronger after clawback-adoption. Moreover, this finding suggests that the negative association between POSTCB and FSCORE is significantly weaker (i.e., of smaller magnitude) for firms with high INSDSALE. To assess this finding, we test the sum of coefficients for POSTCB and POSTCB  INSDSALE (0.090 + 0.066 = 0.024) and find that it is not statistically significant (p-value for F-test = 0.322). This F-test result suggests that the efficacy of clawback adoptions in mitigating the risk of fraudulent reporting basically disappears for firms that have high net insider sales.15 In economic terms, for a clawback-adopting firm with no insider sales (INSDSALE = 0), the differential decrease in FSCORE following adoption is 8.4% of the sample mean FSCORE, nearly double the 4.6% differential decrease in FSCORE for a clawback adopter with sample-average insider sales (INSDSALE = 0.623). Further, when insider sales reach the 90th percentile of the sample distribution (INSDSALE = 1.95), the clawback-adopting firms experience an increase (rather than a decrease) in the FSCORE of about 3.6% of the sample-average FSCORE.16 4.3. Additional analyses 4.3.1. Time series analysis The results reported in Table 4 are based on a differences-in-differences (D-I-D) approach which examines the differential change, i.e., the difference between the change in the FSCORE following clawback-adoption for the adopting firms and the change in the FSCORE for the propensity score-matched control firms over the same pre–post time period. Since there is no triggering event (i.e., clawback adoption) for the control sample during our sample period, the maintained assumption underlying this approach is that the differential change we observe in the FSCORE between the two groups of firms (the adopters and the non-adopters) can be attributed to clawback-adoption. Although it is possible that some of the observations included in our control sample experience a change in the FSCORE for reasons that are not properly controlled for in our empirical analysis, the propensity score-matching method we use for constructing our control sample is basically designed to mitigate this concern (Lennox et al., 2012). 14

In model (1), we estimate economic significance by dividing b2 by 1.068, the (untabulated) sample mean for FSCORE. A possible explanation for the finding is window-dressing, i.e., for some firms, clawback adoption may be more symbolic than substantive. In other words, for these firms, the adoption of a clawback provision is no guarantee that it will be effective, i.e., for these firms clawback adoption is likely to exhibit limited or no effect on financial reporting quality. This phenomenon of window-dressing has been observed in other corporate contexts such as stock repurchase programs (Westphal and Zajac, 2001), CEO long-term incentive plans (Westphal and Zajac, 1994), CEO compensation side-benefits from assets securitization gains (Dechow et al., 2010), and companies’ short-term borrowing disclosure (SEC, 2010). 16 In model (2), we estimate economic significance by dividing (b2 + b7 INSDSALE) by 1.068, the (untabulated) sample mean for FSCORE. When INSDSALE equals 0.623, economic significance = (0.09 + 0.066  0.623)/1.068 = 4.6%. When INSDSALE equals 1.95, economic significance = (0.09 + 0.066  1.95)/1.068 = 3.6%. 15

Please cite this article in press as: Fung, S.Y.K., et al. Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk. J. Account. Public Policy (2015), http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002

14

S.Y.K. Fung et al. / J. Account. Public Policy xxx (2015) xxx–xxx

In any event, to further mitigate the concern that our findings are driven by some factor other than clawback-adoption, in this additional analysis we use an alternative within-sample time-series analysis research design. We remove the control group and examine the change in FSCORE for the clawback-adopting firms before and after the adoption. Essentially, this time-series analysis utilizes the individual clawback-adopting firm’s pre-adoption year as its own control. Since the sample now contains only CB = 1 firms, the variable CB and its related interaction variables are no longer pertinent and we estimate the following models (3) and (4) to test our hypotheses H1 and H2, respectively:

FSCORE ¼ b0 þ b1 POST þ ak Controls þ Year and Industry Dummies þ error

ð3Þ

FSCORE ¼ b0 þ b1 POST þ b2 INSDSALE þ b3 POST  INSDSALE þ ak Controls þ Year and Industry Dummies þ error

ð4Þ

In models (3) and (4) we employ the same control variables as those in models (1) and (2) plus an additional control variable LAMBDA intended to correct for self-selection bias (if any) because the clawback sample is not generated randomly. LAMBDA is the inverse Mills ratio estimated from a Probit choice model where the choice being estimated is clawback adoption.17 The time-series analysis is performed using a sample of 3617 firm-year observations for 585 clawback adopting firms only, and includes 2191 pre-adoption and 1426 post-adoption observations.18 The results of this additional analysis are reported in Table 5. Consistent with the results reported previously in Table 4 column (1), the model (3) results in Table 5 column (1) show that the coefficient on POST is negative and significant (0.028, t-stat = 1.68), suggesting that the FSCORE of the adopters is significantly lower following clawback-adoption. Similarly, the model (4) results in Table 5 column (2) show that the coefficient on POST is significant and negative (0.055, t-stat = 2.29), while the coefficient on POST  INSDSALE is significant and positive (0.054, t-stat = 2.63). Further, the sum of the coefficients for POST + (POST  INSDSALE) (0.055 + 0.054 = 0.001) is not statistically significant (p-value for F-stat = 0.655). Consistent with our inferences based on the earlier differences-in-differences approach (Table 4), the results of the time-series analyses in Table 5 suggest that the negative relation between clawback adoption and the risk of fraudulent financial reporting disappears for firms with high net insider sales of stock. Thus, the findings reported in Table 5 provide further support for our hypotheses H1 and H2. 4.3.2. An analysis of components of F-score In essence, FSCORE is a scaled logistic probability metric that incorporates variables including the accrual quality measure (rsst_acc) from Richardson et al. (2005), change in receivables (ch_rec), change in inventory (ch_inv), percent of soft assets (soft_assets), change in cash sales (ch_cs), change in return on assets (ch_roa), and actual issuance of securities (issue). By incorporating both accruals and other relevant firm characteristics in estimating the FSCORE, Dechow et al. (2011) note that the FSCORE is likely to provide incremental information beyond accruals for researchers investigating earnings management. We examine whether the moderating effect of insider sales on the relation between firm-initiated clawback-adoptions and fraud risk is discernible only at the level of the total FSCORE or also discernible at the accruals and non-accruals sub-set (component) levels of the FSCORE.19 Following Dechow et al. (2011), we categorize the following variables in the FSCORE model as proxies for accrual quality: rsst_acc, ch_rec, ch_inv, and soft_assets, and the remaining variables (ch_cs, ch_roa, and issue) as non-accruals proxies primarily measuring performance and growth.20 We 17 The choice model is the same as the one used for constructing the propensity score-matched control sample discussed previously in Section 3.2. 18 Note that the prior D-I-D analysis (models 1 and 2) required us to find propensity score-matched control firms (CB = 0) for each of our treatment firms (CB = 1). Without this restriction in the time-series analysis (models 3 and 4), our sample of treatment (CB = 1) firms is now somewhat larger. 19 We thank an anonymous reviewer for motivating this additional analysis. 20 In order to develop fraud prediction models, Dechow et al. (2011) explore a broad range of variables measuring accrual quality, financial performance, nonfinancial performance, off-balance-sheet attributes, and stock market performance. Based on a backward elimination technique, Dechow et al. select subsets of variables to develop three FSCORE models: model (1) using only financial statement variables, model (2) adding non-financial-statement and off-balance-sheet variables, and model (3) adding market-based variables. In our study, FSCORE is computed based on Dechow et al. (2011) model (1) – the best-performing model in their assessment.

Please cite this article in press as: Fung, S.Y.K., et al. Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk. J. Account. Public Policy (2015), http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002

15

S.Y.K. Fung et al. / J. Account. Public Policy xxx (2015) xxx–xxx Table 5 Regression results: additional (time-series) analysis. Variables

POST INSDSALE POST  INSDSALE LNSIZE MTB ZSCORE LEV ROA INDLIT BDIND INTERLOCK CEODUAL RESTATE ANALYST INSTHLD NBSEG NGSEG LAMBDA INTERCEPT OBSERVATIONS ADJ R-SQUARED F-test of sum of coefficients: POST + POST  INSDSALE

Expected sign

H1 () + H2 (+) ? ?  + ? +  + + ?   + + ?

Model (3)

Model (4)

Coefficient

t-Stat

Coefficient

t-Stat

0.028*

1.68

0.017 0.042*** 0.012 0.094 0.032 0.140*** 0.105*** 0.091** 0.028 0.052*** 0.006*** 0.066* 0.022 0.004 0.403*** 1.699***

0.71 4.28 1.25 1.36 0.22 4.46 2.59 2.57 1.61 2.7 5.37 1.78 1.16 0.22 2.80 4.74

0.055** 0.005 0.054*** 0.013 0.043*** 0.013 0.099 0.021 0.139*** 0.102** 0.089** 0.026 0.052*** 0.006*** 0.065* 0.024 0.003 0.390*** 1.656***

2.29 0.42 2.63 0.56 4.42 1.35 1.44 0.15 4.42 2.52 2.48 1.53 2.71 5.35 1.77 1.24 0.16 2.71 4.61

3617 0.40

3617 0.40 p-value = 0.655

Model (3)

FSCORE ¼ b0 þ b1 POST þ ak Controls þ Year and Industry Dummies þ error Model (4)

FSCORE ¼ b0 þ b1 POST þ b2 INSDSALE þ b3 POST  INSDSALE þ ak Controls þ Year and Industry Dummies þ error Note: The additional (time-series) analysis examines the change in FSCORE from the pre- to the post-clawback adoption period for clawback-adopting firms based upon 3617 firm-year observations of 585 clawback adopters. All variables are defined in Appendix A. Year dummies and industry dummies are included in each specification but are not reported for brevity. All pvalues are based on standard errors adjusted for firm-level and year-level clustering. * Two-tailed significance at 10% level. ** Two-tailed significance at 5% level. *** Two-tailed significance at 1% level.

calculate the accruals component of FSCORE by retaining the accruals proxies and their original coefficients in FSCORE model and dropping non-accruals proxies from the model. Similarly, we calculate the non-accruals component of FSCORE by retaining the non-accruals proxies and their coefficients and dropping all accruals proxies. We rerun the analyses by replacing FSCORE with the accruals and non-accruals components of FSCORE separately. Untabulated results indicate that in the regression with the accruals component of FSCORE as the dependent variable, POSTCB is negatively significant (p < 0.01), the interaction term POSTCB  INSDSALE is significantly positive (p < 0.05), and the combined coefficient for (POSTCB + POSTCB  INSDSALE) is insignificant. In the regression with the non-accruals component of FSCORE as the dependent variable, none of the variables of interest is significant. Taken together, these results suggest that the moderating effect of insider sales on the relation between firm-initiated clawback-adoptions and fraud risk is discernible not only at the level of the total FSCORE but also at the accruals sub-set (component) level of the FSCORE. 4.3.3. Variation in clawback provisions As pointed out in the prior literature (Chan et al., 2012; Dehaan et al., 2013; Iskandar-Datta and Jia, 2013), an important cross-sectional variation in clawback provisions relates to the trigger for Please cite this article in press as: Fung, S.Y.K., et al. Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk. J. Account. Public Policy (2015), http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002

16

S.Y.K. Fung et al. / J. Account. Public Policy xxx (2015) xxx–xxx

compensation recoupment. Consistent with the prior literature (Brown et al., 2015; Dehaan et al., 2013), we use a classification code provided in the Corporate Library’s clawbacks database to classify clawback firms into one of the following four categories: (1) misconduct clawbacks, triggered by executive misconduct/fraud; (2) robust clawbacks, triggered by incorrect financials due to error or fraud; (3) non-compete clawbacks, triggered by an executive’s violation of non-compete clause; and (4) a general catch-all category, triggered say by an executive’s sudden departure. Although all types of clawback provisions may have a deterrent effect on managers’ misreporting behavior, we perform additional analyses to take into account the four categories of clawback firms. First, we drop 181 clawback adopters that belong to the non-financial categories (3) and (4) and re-do the analysis. The resulting reduced sample is similar to that in Dehaan et al. (2013) consisting only of restatement-related clawbacks. For this reduced sample, our results (untabulated for brevity) are similar to our main results (POSTCB coefficient = 0.105, p-value < 0.01; POSTCB  INSDSALE coefficient = 0.096, p-value < 0.01). Next, we analyze separately the category (1) and category (2) clawback firms (129 for category (1) and 104 adopters for category (2)). Our results for the category (2) robust clawbacks analysis are similar to those for our full sample (POSTCB coefficient = 0.139, p-value < 0.01; POSTCB  INSDSALE coefficient = 0.178, p-value < 0.01). Weaker results are observed for category (1) misconduct clawback firms (POSTCB coefficient = 0.076, p-value < 0.1; POSTCB  INSDSALE coefficient = 0.040, insignificant). Similar to the results of our main analyses, in all of the above sub-sample analyses, the combined coefficient of (POSTCB + POSTCB  INSDSALE) is insignificant. Collectively, the findings from the additional analyses suggest that our inferences hold for the various clawback provision categories. 4.3.4. Insider sales versus bonuses subject to recoupment To provide further evidence on the magnitude of the realized trading profits through insider sales, we perform additional analyses to compare the magnitude of insider trading profits with that of bonus payments subject to clawback provisions. We measure insider trading profits based on net insider sales (INSDSALE) (i.e., net sales or dollar sales minus dollar purchases of company stock executed by the firm’s top managers, scaled by beginning-of-year firm equity value), assuming that the cost of stock or option acquisition (which is not easily tractable) is either negligible or a sunk cost for the managers. We measure the bonus payments that are subject to clawback provisions (BONUS) as the aggregated bonus payments for the top managers, scaled by beginning-of-year equity value. We first conduct a univariate analysis to compare clawback firms’ mean (median) of INSDSALE and BONUS. Based upon untabulated results, we find that the mean (median) INSDSALE is 0.629 (0.209) and the mean (median) BONUS is 0.272 (0.036), and the differences are statistically significant (p value < 0.01 for tests on differences both in means and in medians). These results suggest that the insider trading profits in general dominate the bonus payments that are subject to clawback provisions. To further examine whether our multivariate results are affected by cases with very large bonus payments that could dominate the incentives for insider trading, we remove from our sample the firm-years that adopt clawback provision but with BONUS larger than INSDSALE. We also remove those observations with no information on BONUS. With this reduced sample (N = 4544), we re-estimate model (2) and find results similar to the findings we reported in the paper. Specifically, we find the coefficient POSTCB to be negative and significant (0.098, p-value < 0.01, two-tailed) and the coefficient POSTCB  INSDSALE to be positive and significant (0.062, p-value < 0.1). This suggests that our results and inferences hold after removing observations with little incentive to engage in insider trading after adoption of clawback provisions. 4.3.5. Management turnover To examine whether our results are sensitive to changes in management team, we conduct additional analysis by removing observations from our sample that experience CEO changes in the preand post-CB adoption periods.21 Based on this reduced sample of 4323 firm-year observations, we 21

We focus on CEO turnover because normally it is the CEO who shapes the strategies and actions of the management team.

Please cite this article in press as: Fung, S.Y.K., et al. Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk. J. Account. Public Policy (2015), http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002

S.Y.K. Fung et al. / J. Account. Public Policy xxx (2015) xxx–xxx

17

re-estimate our results and find that POSTCB is significantly negative (coefficient = 0.075, p-value < 0.01), and POSTCB  INSDSALE is significantly positive (coefficient = 0.050, p-value < 0.01). The combined coefficient of (POSTCB + POSTCB  INSDSALE) is insignificant. These results (untabulated for brevity) are similar to the main findings reported in the paper, which alleviate the concern that results reported in this study are driven by changes in management.

5. Concluding remarks In this study, we provide evidence relating to a topic that has seen relatively little empirical research yet remains an important public policy issue, i.e., the effectiveness of clawback provisions in limiting fraud risk. In particular, while prior research has focused on the relation between the voluntary adoption of clawback provisions and financial reporting quality, we examine a different yet critical issue, i.e., the impact of insider sales on the effectiveness of clawbacks in lowering the risk of fraudulent financial reporting. The purpose of clawbacks is to mitigate the dysfunctional effects of incentive-based executive compensation, i.e., ex ante deter as well as ex post penalize executives who manipulate reported earnings. Specifically, these provisions seek to recover excess pay tied to performance metrics (such as reported earnings) that are inflated. However, neither the firm-initiated provisions nor the stipulations in the 2010 Dodd–Frank Act seem to require the clawback of another important form of excess pay, i.e., the excess stock sale proceeds linked to the sale by executives of their equity holdings at a time of inflated price-affecting metrics such as earnings. With alternative routes to excess pay now subject to clawback, an unintended consequence of extant clawback provisions could be to further incentivize managers to attempt to extract excess pay by manipulating reported earnings prior to unloading their shares. Consequently, the impact of insider sales on the efficacy of clawback adoptions in lowering fraud risk remains an important, yet unanswered, empirical question. In our study, we examine a sample of firm-initiated clawback adoptions during 2003–2012 as well as a control sample of propensity score-matched non-clawback-adopting firms. Our findings (based on the differences-in-differences research method) suggest that insider sales materially weaken the effectiveness of extant clawbacks in lowering the risk of fraudulent financial reporting.22 At this time (July 2014), the SEC is still working on rules for implementing clawbacks mandated by Dodd–Frank (one of nearly half the rules yet to be completed under the 2010 financial reform Act). Our findings suggest that clawback rules (as and when issued by the SEC) need to address insider sales for clawbacks to be fully effective in lowering the risk of fraudulent financial reporting.

Acknowledgements We gratefully acknowledge helpful comments and suggestions of Martin Loeb (editor), two anonymous reviewers, and participants at the 2013 Journal of Contemporary Accounting and Economics (JCAE) Symposium. K. K. Raman acknowledges funding from the Ramsdell Memorial Chair for Accounting at the University of Texas at San Antonio.

22 However, given the complexities associated with estimating what the stock price would have been absent the earnings manipulation, it would be challenging to determine the amount of excess sale proceeds and to recoup this particular form of excess compensation. A possible solution, albeit punitive, would be to recover the entire (or a large percentage of the) amount of insider trading profits since the start of the period in which the financial misstatements first occurred. We thank a reviewer for this suggestion.

Please cite this article in press as: Fung, S.Y.K., et al. Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk. J. Account. Public Policy (2015), http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002

18

Variable definitions. Variable FSCORE

Definition =

CB POST

= =

POSTCB INSDSALE

= =

LNSIZE MTB ZSCORE LEV ROA INDLIT

= = = = = =

BDIND

=

BDSIZE AUDITSIZE INTERLOCK ABSDAROA CEODUAL RESTATE ANALYST INSTHLD NBSEG NGSEG

= = = = = = = = = =

The fraud risk score calculated based on Dechow et al.’s (2011) model (1). FSCORE ¼



eY 1þeY



 0:0037 where

eY ¼ 7:893 þ 0:790  rsst acc þ 2:518  ch rec þ 1:191  ch inv þ 1:979  soft assets þ 0:171  ch cs  0:932  ch roa þ 1:029  issue rsst_acc is working capital accruals calculated following Richardson et al. (2005), which equals (DWC + DNCO + DFIN)/average total assets (AT), where DWC = D [(ACT  CHE)  (LCT  DLC)], DNCO = D [(AT  ACT  IVAO)  (LT  LCT  DLTT)], DFIN = D [(IVST + IVAO)  (DLTT + DLC + PSTK)]. ch_rec = change in accounts receivables (RECT) scaled by average total assets (AT). ch_inv = change in inventory (INVT) scaled by average total assets (AT). soft_assets = [total assets (AT)  property, plant, and equipment (PPENT)  cash and cash equivalent (CHE)]/total assets (AT). ch_cs = percentage change in cash sales (cash sales = SALE  DRECT), ch_roa = change in return on assets (IB/average AT), issue = 1 if a firm issued securities (SSTK > 0 or DLTIS > 0) during the year, and 0 otherwise 1 for firms voluntarily adopting clawback provisions for executive compensation, and 0 otherwise 1 for firm-years of clawback adoption and all following years for both clawback adopters and control firms, and 0 otherwise. Each control firm is assigned an artificial adoption year that is identical to its matched adopter 1 for firm-years of clawback adopters when clawback provisions are in place, and 0 otherwise Net sales by top five insiders, measured as the natural logarithm of (1 + netpurchase) during the year based on Thomson Reuters database. Netpurchase is calculated as top five managers’ net share sales in dollars (i.e., share sales minus share purchases) scaled by the firm’s beginning-of-year market value of equity, multiplied by 1000. Top five insiders are defined as the five most highly compensated executives The natural logarithm of total assets (AT) Market value of equity (PRCC_F ⁄ CSHO)/book value of equity (CEQ) (1.2 ⁄ WCAP + 1.4 ⁄ RE + 3.3 ⁄ PI + 0.999 ⁄ SALE)/AT. This is a revised version of ZSCORE (Altman, 1968) calculated following Graham et al. (2008) [long-term debt (DLTT) + debt in current liabilities (DLC)]/total assets (AT) Income before extraordinary items (IBC)/lagged assets (AT) 1 if firm operates in a high-litigation industry (Pharmaceutical/biotechnology SIC 2833–2836 and 8731–8734; Computer 3570–3577 and 7370–7374; Electronic 3600–3674; Retail 5200–5961), and 0 otherwise 1 if more than 60% of the directors on the board are independent, and 0 otherwise. Independent directors refer to those who are not corporate executives and have no business relationship with the company based upon RiskMetrics database The natural logarithm of board size, measured as the number of directors serving on the board, obtained from RiskMetrics database The natural logarithm of auditor committee size, measured as the number of directors serving on the audit committee, obtained from RiskMetrics database 1 if a firm has an interlocking relationship as defined by the SEC involving members of the board of directors, and 0 otherwise Absolute value of performance adjusted abnormal accruals calculated following Kothari et al. (2005) 1 if CEO is also the chairman of the board of directors, and 0 otherwise 1 if firm has restated its financial statement in the past five years, and 0 otherwise The number of analysts following the firm, measured as the number of analysts issuing one or more annual earnings forecasts during the year based on I/B/E/S The percentage of common shares held by institutional investors, obtained from the CDS/Spectrum database The natural logarithm of (1 + the number of business segment) during the year based on COMPUSTAT segment file The natural logarithm of (1 + the number of geographic segment) during the year based on COMPUSTAT segment file

S.Y.K. Fung et al. / J. Account. Public Policy xxx (2015) xxx–xxx

Please cite this article in press as: Fung, S.Y.K., et al. Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk. J. Account. Public Policy (2015), http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002

Appendix A

19

S.Y.K. Fung et al. / J. Account. Public Policy xxx (2015) xxx–xxx

Appendix B Clawback adoption propensity score matching model. Variables LNSIZE LEV ZMI_SCORE MTB RESTATE INTERLOCK BDSIZE AUDITSIZE BDIND ABSDAROA CEODUAL NBSEG NGSEG ROA INSTHLD INTERCEPT OBSERVATIONS MODEL CHI-SQUARE LOG LIKELIHOOD PSEUDO R-SQUARED

Expected sign + +  ? +  + + +   + + + +

Coefficient ⁄⁄⁄

0.274 0.376 0.034 0.027 0.041 0.207 0.767⁄⁄⁄ 0.056 0.405⁄⁄⁄ 0.522 0.004 0.124 0.129 0.019 0.084 4.513⁄⁄⁄ 7738 227.5 4497 0.147

t-Stat 6.66 1.35 0.69 0.73 0.53 1.03 3.88 0.87 4.57 1.18 0.06 1.56 1.52 0.05 0.52 9.71

Note: In the clawback adoption model, the dependent variable is CB, equal to 1 for firms voluntarily adopting clawback provisions, and 0 otherwise. The independent variables consist of determinants of clawback adoptions (based on Addy et al. (2014) and Chan et al. (2012)), and are defined in Appendix A. The model is estimated using a sample of 4216 firm-year observations of 635 clawback adopting firms and 3522 firm-year observations of 553 control firms, all with the required data for our sample period.

References Ackerman, A., Zibel, A., 2014. Dodd–Frank law still far from done. Wall Street J. 20 July, C6. Addy, N., Chu, X., Yoder, T., 2014. Voluntary adoption of clawback provisions, corporate governance, and interlock effects. J. Account. Public Policy 33 (2), 167–189. Ali, A., Kallapur, S., 2001. Securities price consequences of the Private Securities Litigation Reform Act of 1995 and related events. Account. Rev. 76 (3), 431–460. Altman, E.I., 1968. Financial ratios, discriminant analysis and the prediction of corporate bankruptcy. J. Financ. 23 (4), 589–609. Bao, D., Fung, S.Y.K., Su, L., 2015. Can Shareholders be at rest after adopting Clawback Provisions? Evidence from Stock Price Crash Risk. Available at SSRN: . Bartov, E., Mohanram, P., 2004. Private information, earnings manipulations and executive stock option exercises. Account. Rev. 79 (4), 889–920. Beasley, M.S., 1996. An empirical analysis of the relation between the board of director composition and financial statement fraud. Account. Rev. 71 (4), 443–465. Beneish, M., 1999. Incentives and penalties related to earnings overstatements that violate GAAP. Account. Rev. 74 (4), 425–457. Bergstresser, D., Philippon, T., 2006. CEO incentives and earnings management. J. Financ. Econ. 80 (3), 511–529. Bowen, R.M., DuCharme, L., Shores, D., 1995. Stakeholders’ implicit claims and accounting method choice. J. Account. Econ. 20 (3), 255–295. Brown, A.B., Davis-Friday, P.Y., Guler, L., 2015. M&A performance and the voluntary adoption of clawback provisions in executive compensation contracts. J. Bus. Financ. Account. 42 (1–2), 237–271. Burns, N., Kedia, S., 2006. The impact of performance based compensation on misreporting. J. Financ. Econ. 79 (1), 35–67. Carpenter, J., Remmers, B., 2001. Executive stock option exercises and inside information. J. Bus. 74 (4), 513–534. Chan, L., Chen, K., Chen, T., Yu, Y., 2012. The effects of firm-initiated clawback provisions on earnings quality and auditor behavior. J. Account. Econ. 54 (2–3), 180–196. Chan, L., Chen, K., Chen, T., Yu, Y., 2015. Substitution between real and accruals-based earnings management after voluntary adoption of compensation clawback provisions. Account. Rev. 90 (1), 147–174. Dechow, P.M., Myers, L.A., Shakespeare, C., 2010. Fair value accounting and gains from asset securitizations: a convenient earnings management tool with compensation side-benefits. J. Account. Econ. 49 (1), 2–25.

Please cite this article in press as: Fung, S.Y.K., et al. Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk. J. Account. Public Policy (2015), http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002

20

S.Y.K. Fung et al. / J. Account. Public Policy xxx (2015) xxx–xxx

Dechow, P., Ge, W., Larson, C., Sloan, R., 2011. Predicting material accounting misstatements. Contemp. Account. Res. 28 (1), 17– 82. DeFond, M., Zhang, J., 2014. A review of archival audit research. J. Account. Econ. 58 (2–3), 275–326. Dehaan, E., Hodge, F., Shevlin, T., 2013. Does voluntary adoption of a clawback provision improve financial reporting quality? Contemp. Account. Res. 30 (3), 1027–1062. Denis, D.K., 2012. Mandatory clawback provisions, information disclosure, and the regulation of securities markets. J. Account. Econ. 54 (2–3), 197–200. Francis, J., Philbrick, D., Schipper, K., 1994. Shareholder litigation and corporate disclosures. J. Account. Res. 32 (2), 137–164. Fried, J., 1998. Reducing the profitability of corporate insider trading through pretrading disclosure. South. Calif. Law Rev. 71 (January), 305–392. Fried, J., 2008. Hands-off options. Vanderbilt Law Rev. 61 (2), 453–474. Fried, J., Shilon, N., 2011. Excess-pay clawbacks. J. Corp. Law 36 (4), 721–751. Gimein, M., 2002. You bought, they sold: since 1999 hundreds of greedy executives at America’s worst performing companies have sold sixty-six billion dollars worth of stock. Fortune September 2, 64–70. Graham, J.R., Li, S., Qiu, J., 2008. Corporate misreporting and bank loan contracting. J. Financ. Econ. 89 (1), 44–61. Hennes, K., Leone, A., Miller, B., 2008. The importance of distinguishing errors from irregularities in restatement research: the case of restatements and CEO/CFO turnover. Account. Rev. 83 (6), 1487–1519. Iskandar-Datta, M., Jia, Y., 2013. Valuation consequences of clawback provisions. Account. Rev. 88 (1), 171–198. Ke, B., Huddart, S., Petroni, K., 2003. What insiders know about future earnings and how they use it: evidence from insider trades. J. Account. Econ. 35 (3), 315–346. Kothari, S.P., Leone, A.J., Wasley, C.E., 2005. Performance matched discretionary accrual measures. J. Account. Econ. 39 (1), 163– 197. Lawrence, A., Minutti-Meza, M., Zhang, P., 2011. Can big 4 versus non-big 4 differences in audit-quality proxies be attributed to client characteristics? Account. Rev. 86 (1), 259–286. Lennox, C., Francis, J., Wang, Z., 2012. Selection models in accounting research. Account. Rev. 87 (2), 589–616. Lerach, W., 2002. Plundering America: how American investors got taken for trillions by corporate insiders. Stanf. J. Law Bus. Financ. 8 (Autumn), 1–45. Loebbecke, J.K., Eining, M.M., Willingham, J.J., 1989. Auditors’ experience with material irregularities: frequency, nature, and detectability. Audit: J. Pract. Theory 9 (1), 1–28. Petersen, M.A., 2009. Estimating standard errors in finance panel data sets: comparing approaches. Rev. Financ. Stud. 22 (1), 435–480. Pulliam, S., Barry, R., 2012. Executives’ good luck in trading own stock. Wall Street J. Nov 27, B1. Pulliam, S., Barry, R., 2013. Executives hit sweet spot on stock sales: Corporate executives give favorable stock guidance, sell shares, then disclose bad news. Wall Street J. April 24, A1. PwC, 2014. Executive Compensation: Clawbacks—2013 Proxy Disclosure Study. PwC Human Resources Services. Richardson, S., Sloan, R., Soliman, M., Tuna, I., 2005. Accrual reliability, earnings persistence, and stock prices. J. Account. Econ. 39 (3), 437–485. Securities and Exchange Commission, 2010. SEC Proposes Measures to Enhance Short-term Borrowing Disclosure to Investors (September 17). Sender, H., Blumenstein, R., 2002. Questioning the books: global crossing creditors review sales, swaps. Wall Street J. February 25, A6. Siconalfi, M., Eaglesham, J., 2013. SEC is pressed to revamp executive trading plans. Wall Street J. May 9, B1. Sprangler, T., 2013. Clawbacks becoming Fashionable on Wall Street. Forbes, 28 May. . Srinivasan, S., Wahid, A., Yu, G., 2015. Admitting mistakes: home country effect on the reliability of restatement reporting. Account. Rev. 90 (3), 1201–1240. Summers, S., Sweeney, J., 1998. Fraudulently misstated financial statements and insider trading: an empirical analysis. Account. Rev. 73 (1), 131–146. Westphal, J.D., Zajac, E.J., 1994. Substance and symbolism in CEOs’ long-term incentive plans. Adm. Sci. Q. 39 (3), 367–390. Westphal, J.D., Zajac, E.J., 2001. Decoupling policy from practice: the case of stock repurchase programs. Adm. Sci. Q. 46 (2), 202– 229.

Please cite this article in press as: Fung, S.Y.K., et al. Insider sales and the effectiveness of clawback adoptions in mitigating fraud risk. J. Account. Public Policy (2015), http://dx.doi.org/10.1016/j.jaccpubpol.2015.04.002