Third-Party Reaction to Hedge Fund Activism: Auditor s Perspective

Transcription

1 Third-Party Reaction to Hedge Fund Activism: Auditor s Perspective Huimin (Amy) Chen* Lally School of Management Rensselaer Polytechnic Institute chenh18@rpi.edu Bill B. Francis Lally School of Management Rensselaer Polytechnic Institute francb@rpi.edu Yinjie (Victor) Shen Lally School of Management Rensselaer Polytechnic Institute sheny5@rpi.edu Qiang Wu Lally School of Management Rensselaer Polytechnic Institute wuq2@rpi.edu *Contact Author

2 Abstract This paper investigates the relationship between the auditors and the firms that are targeted by hedge fund activists. We provide a unique perspective, that of the auditors, who play an external monitoring role for the hedge fund activists target firms. We hypothesize that auditors react to hedge fund activists by charging higher audit fees and that the reaction is due to their concerns about target firms business risk. We find that within two years after targeting, audit fees of the targeted firms increase by 22.1%. This effect is more pronounced when target firms have high business risk or have financial reporting personnel turnovers. The finding is supported by additional tests based on propensity-score matching and difference-in-difference tests. Our study identifies and confirms a strong causal relation between hedge fund activism and audit fees. 2

3 1. Introduction Hedge fund activists have played an increasingly important role in the capital market over the last decade. For example, the number of hedge fund activists has almost doubled since 2001 (Brav, Jiang and Kim 2015), and the total assets under management rose about tenfold since 2003 to around $115 billion in 2015 (PwC 2015). The rapid growth of hedge fund activisms have spurred considerable attention by regulators and academia. According to Security and Exchange Commission (SEC) Chairman Mary Jo White, hedge fund activists have undeniably changed the corporate landscape (White 2015). The extant literature has extensively examined whether hedge fund activists create or destroy value for shareholders. Most literature have shown improvement in performance in terms of stock performance, operating performance and real effects such as production (e.g. Brav, Jiang, Partnoy, and Thomas 2008; Brav, Jiang and Kim 2013). However, these studies have concentrated on shareholders. Only a few papers started to examine the impact of hedge fund activism on other stakeholders (Brav, Jiang and Kim 2015). While hedge fund activism may benefit shareholders, it may increase risk to other stakeholders such as bondholders (Klein and Zur 2011) and bank loan contractors (Li and Xu 2009) as well as extract value from employees (Brav et al. 2015). In this paper, we investigate a unique perspective, that of auditors, who play important roles as external monitors (e.g., Jensen and Meckling 1976; Becker et al. 1998; Nelson et al. 2002). Specifically, we investigate how auditors react to hedge fund activism, using audit fees as the main indicator 1. We choose auditors perspective because it may provide unique insight. First, although the prior literature has shown that hedge fund activists improve target firms performance (e.g. Brav 1 We focus on audit fees for two reasons. Firstly, this measure has variation that produces statistical power. Other measures such as going concern opinions do not have enough variation. Secondly, audit fee models are typically well-specified with R-squares exceeding 70% to 80%, which lessens concerns about correlated omitted variables. 1

4 et al. 2008), auditors may still view activists impact differently since performance is not a major concern for auditors. Auditors concern more about clients risk since the current audit methodology is risk-oriented. Second, auditors have proprietary information about the target firms, which may give them a different view, while some stakeholders such as investors may have only public information. We hypothesize that auditors view the hedge fund activists target firms as having higher business risk. Unlike shareholders of the target firms, auditors do not benefit from hedge fund activism. Instead, they are concerned about the heightened business risk in the target firms. In response to the business risk concern, auditors charge higher audit fees. We are motivated by the anecdotal evidence which suggests that auditors may have a negative view against hedge fund activists. A recent report by one CPA firm (PwC, 2016) lays out strategies for the potential target firms to stay out of the activists. For example, it states in the report that Activists go after companies with vulnerabilities, and most companies have some. If companies want to get ahead of an activist threat, they ll want to understand their potential vulnerabilities, including how an outside hedge fund activist might see things. They can start with our risk assessment tool. From this report, it indicates that auditors view the potential target firms as vulnerable. We argue that such vulnerability is the business risk that auditors concern about. Once the firm is targeted, the auditor charges higher fees to compensate their increased business risk. In implementing our analysis, we follow Brav et al. (2008) in collecting the hedge fund activists targeting events and extend the dataset to the period of We use 1,351 targeting events and investigate the target firms audit fees changes. We obtain 5,329 firm-year observations and control for variables such as auditors and going-concern opinions as well as industry fixed effect. We compare the audit fees two years after the hedge fund intervention with 2

5 two years before intervention. According to our baseline OLS regression model, we find that audit fees increase by 22.1% two years after intervention. We use a difference-in-differences methodology to further investigate the causal relationship between hedge fund activisms and audit fees. Using propensity score matching to find paired treatment and control firms that have similar firm characteristics, we find that there is about 32.2% increase in audit fees among treatment firms in the years subsequent to hedge fund activists targeting events as compared to control firms. The result from the difference-in-differences model is consistent with the result from the baseline regressions. Our main finding is also robust to additional control variables and alternative testing windows. The relationship we identify involves three parties: auditors, hedge fund activists and target firms. We conduct further tests to explore possible alternative explanations of our finding. First, it is possible that hedge fund activists influence the managers of the target firms to require more audit service. Following Brav et al. (2008), we focus on the director and officer changes that have shown to be the impact of hedge fund activism. We use the changes in directors, changes in chairman, changes in CEO and whether hedge fund activists successfully change target firms governance to test hedge funds influence on the firms audit fees. We find that there is no significant difference of audit fees between target firms with and without director and officer changes, indicating that the increased audit fees are less likely driven by the demand of hedge fund activisms. Secondly, it is also possible that target firms demand more audit service as defending against the hedge fund activists. To test this possibility, we use a subsample that indicates whether the target firms involve in proxy fights and examine whether such firms demand more audit service (more audit fees) compared to target firms that do not involve in proxy fights. However, we do not 3

6 find any significant difference between these two subsample, indicating our result is less likely driven by target firms demand for more audit service. Thirdly, it is possible that regulators such as SEC may pay more attention to and increase scrutiny in the target firms and causes more audit fees. Using SEC s Accounting and Auditing Enforcement Releases (AAER) letters as a proxy for regulators attention, we find that AAER has no significant impact on the increased audit fee, indicating that SEC scrutiny is less likely a factor which affects audit fees after hedge fund activisms. The audit fee literature also suggests that audit fees could be driven by audit efforts and/or audit risk. DeFond and Zhang (2014) summarize four strategies auditors use to counter risk: (1) reduce risk by increasing effort; (2) price risk by charging a premium; (3) avoid risk through client retention and acceptance; and (4) attenuate risk through lobbying. The first two strategies are related to our topic since they lead to higher audit fees. Finally, we examine whether the increased audit fees after hedge fund intervenes are due to increased efforts of auditors. Following prior literature, we use the number of days between the date of fiscal year end and the audit report signature date to measure audit efforts. Our results show that auditors do react to the business risk by increasing audit effort. Our paper makes contribution to two streams of research. Firstly, we contribute to the emerging line of research that studies the impact of hedge fund activism. Most of the prior literature focuses on shareholders of target firms and pays less attention to the possible influence on other market participants. Following the seminal paper Brav et al. (2008), vast literature has centered around the question whether hedge fund activism creates or destroys value for shareholders, while the impact on other participants is understudied. Our paper extends the literature on third-party participants, namely auditors, and examines how they are affected by hedge fund activism. 4

7 Secondly, our paper contributes to the audit fee literature. To our knowledge, our paper is the first study to document the impact of hedge fund activism on audit fees. The prior literature studies the impact of a broader group of owners. For example, Mitra, Hossain and Deis (2007) studies the relationship between institutional ownership and audit fees. In our paper, we study a specific group of investors hedge fund activists. Hedge fund activists have been viewed quite differently than other kinds of institutional investors. Contrary to the general view that firms benefit from the monitoring effect by institutional investors, the views on hedge fund activists are highly contradictory. Mitra et al. (2007) finds decrease of audit fees and attributes it to the benefits of institutions monitoring. In our paper we find increase of audit fees, which is explained by auditors reaction to increased business risk due to hedge fund activisms. The remainder of the paper is organized as follows. Section 2 provides a literature review and develops the hypotheses examined in the paper. Section 3 describes our research design, the data selection process and summary statistics. Section 4 presents the empirical results, including the baseline regression results, channel tests, propensity score matching, difference-in-differences results, robustness tests, possible alternative explanations and additional tests on auditors reaction. We provide the conclusion in Section Literature Review and Hypothesis Development In this section, we first review the hedge fund activism literature, especially the effect on operating performance and corporate governance. We then review the literature about auditors reaction to business risk. We argue that auditors react to the possible high business risk in the target firms after hedge fund activists intervention by increasing audit fees The Impact of Hedge Fund Activism 5

8 The fundamental question of hedge fund activism is whether it creates or destroys value. The extant literature has mostly concentrated on the impact of hedge fund activism on existing shareholders, while a few papers examine other stakeholders. Although many papers consistently find positive excess returns to shareholders, it s still understudied whether other stakeholders bear the risk or are expropriated the wealth. First of all, it has been shown in most literature that hedge fund activism leads to positive stock market reaction. The seminal paper by Brav et al. (2008) examines a sample of 1,059 hedge fund activism events over the period and shows that the abnormal return from 20 days before to 20 days after the announcement of activism is significantly positive. They attribute such value creation to the improvement in corporate governance and operating performance. Following Brav et al. (2008), a few papers have examined the stock market reaction using different samples collected or focusing on different types of hedge fund activism (Clifford 2008; Griffin and Xu 2009; Klein et al. 2009;). These papers consistently find significantly positive stock market return in the short run (less than one year). However, the short-term positive stock market reaction cannot easily give the conclusion that hedge fund activism is value-creation. To shed light on that question, it is worth investigating the impact of hedge fund activism on corporates in other aspects. Brav, Jiang and Kim (2013) find that the productivity of target firms increases after the hedge fund activism intervention. The work hours and wages of the employees of the target firm decrease despite of the increase of labor productivity. A different view from Clifford (2008), who also finds higher stock returns and better operating performance (ROA) in the target firms, attributes the improvement to the divestiture of under-performing assets. Secondly, despite of the positive stock price reaction that have been documented in the literature, researchers are still investigating the alternative explanations for such finding. One 6

9 possibility is that the positive returns to shareholders are simply wealth transferred from other stakeholders (Brav, Jiang and Kim 2009). Greenwood and Schor (2009) argue that the positive abnormal short- and long-term returns are driven solely by targets that are acquired later. Uchida and Xu (2008) indicates that the stock price reaction is more favorable for the target firms with higher leverage. More importantly, two papers have clearly shown the wealth transfer from shareholders to other stakeholders. Klein and Zur (2011) finds that the average excess bond return is 3.9% around the initial hedge fund activists targeting, and is an additional 4.5% over the remaining year. They argue that the hedge fund managers actions in the target firms can in fact result in lower profitability, loss of collateral, or an increase in the firms debt ratios. They use Standard and Poor s (2006a,b) assessment as an example - both Heinz and Wendy s in 2006 are downgraded based partially on the aggressive financial policies instituted by Trian Fund Management. In addition, they find that the unsystematic equity risk increases due to shareholder activism. Li and Xu (2009) has similar findings on bank loan contracts. They find that hedge fund target firms pay higher spreads, face more covenant restrictions on their financial and investment policies, and have shorter loan maturities. Based on the discussion above, it appears that although shareholders may benefit from hedge fund activism, other stakeholders may bear the loss. The changes caused by hedge fund activists in the target firms include higher CEO turnover, changes in board composition, changes in operating strategies and higher payouts and debt-to-equity ratio. While these changes may benefit shareholders, they also increase risk in the target firms and other stakeholders. In this paper, we investigate the impact of hedge fund activism through the lens of external auditors. As independent external monitors, auditors have different views than shareholders. The client firms performance can not financially benefit auditors. Instead, they are more concerned 7

10 about the client firms inherent risk that may lead to their financial or reputational losses. If other stakeholders bear losses in the hedge fund activists target firms, they may sue the firms and blame auditors since they are often considered to have deep pockets. Considering these possible consequences, auditors are more likely to have a negative view on hedge fund activism rather than positive. We conjecture that auditors may view the client firms that are targeted by hedge fund activists as having higher business risk Auditors Reaction to Client Business Risk The prior literature on audit fees has well-established theory and extensive findings to explain the determinants of audit fees. Simunic (1980) provides a theoretical framework for explaining how auditors risk judgment enter into the audit pricing. The auditor s total expected cost, E(c) includes two elements: (1) the cost of resources invested in the audit (cq), and (2) expected cost arising from potential losses due to litigation and/or reputational damage. The audit pricing function is as follow: E(c) = cq + E (d a, q) * E(ø) where: E(c) = audit fees c = per-unit factor cost of audit resources to the auditor, including all opportunity costs and, therefore, a provision for a normal profit; q = quantity of resources invested by the auditor in performing the audit; a = quantity of resources invested by the auditee in operating the internal accounting system; E(d a,q) = expected present value of possible future losses that may be attributed to the audited financial statements, given a and q; and 8

11 E(ø) = expected likelihood the future losses will become the responsibility of the auditor. Auditors professional judgement plays a critical role in assessing E (d a) * E(ø) and choosing q and E(c). Conceptually, the component E (d a, q) * E(ø) represents the audit risk that auditors undertake. Audit risk is formally defined as the risk that the auditor may unknowingly fail to appropriately modify his or her opinion on financial statements that are materially misstated (American Institute of Certified Public Accountants [AICPA] 1983, AU ). In our paper, we examine one important determinant of audit risk client business risk, which is the firm s future outcome uncertainty in operating performance. In the prior literature, researchers have examined the relationship between client business risk and audit fees (e.g. Pratt and Stice 1994; Simunic and Stein 1996; Bell, Landsman and Shackelford 2001; Lyon and Maher 2005). The findings are mostly consistent in that high business risk leads to high audit fees. Managers of clients with high business risk may have the pressure to intentionally bias the financial reports smooth out earnings over the years for a persistent growth rate or conceal declining performance. Bentley, Omer and Sharp (2013) finds that firms with high growth rate and heavy investment in research and development activities are charged with higher audit fees. They argue that these firms are vulnerable to overextending their resources and increasing their risk of incurring losses. The auditors spend more effort and charge higher fees for these firms due to their risk-oriented focus and organizational instability. In our context, we argue that auditors may perceive hedge fund activists as leading to higher business risk in the target firms and react to such risk by charging higher audit fees. Based on the discussion above, we propose the two following hypotheses: HYPOTHESIS 1. Audit fees increase after clients are targeted by hedge fund activists. 9

12 HYPOTHESIS 2.1. The impact of hedge fund activists intervention on audit fees is more pronounced for firms with higher business risk. We acknowledge that in response to the high risk, auditors can either spend more audit effort or charge higher risk premium to compensate. Bell, Landsman and Shackelford (2001) finds that auditors perceive firm-level differences in business risk and obtain compensation through billing additional hours, not by raising the hourly charge. In Section 4.7 in this paper, we also examine how auditors react to the target firms business risk Financial Reporting Resources and Audit Fees A client firm exposed to high business risk is at greater risk of lacking the resources necessary for preparing reliable reports. In our context, as the target firms of hedge fund activism experience sudden improvement in performance, the financial personnel may need to process a vast amount of transactions. The lacking of resources will be more severe if the firm changes CFO or financial director. The prior literature has shown that hedge fund activists tend to exert their influence on the target firms by changing the managers. Brav et al. (2008) finds that during the year after the announcement of activism, the CEO turnover rate increases by almost 10 percentage points. As the hedge fund activists demand changes of CEOs, they may also demand changes of CFOs. Mian, S. (2001) finds that CFO turnover is preceded by a decline in operating return on assets and by high CEO turnover. Since CFOs have more decision-making power on financial reporting, we expect the CFO turnovers are the underlying mechanism on the positive association between hedge fund activism and audit fees. Particularly, Cheng, Huang and Li (2015) finds that CEO changes are not a significant driver for the improvement of accounting conservatism in hedge fund target firms. Nevertheless, CFO turnovers are associated with greater increases in conservatism 10

13 following the hedge fund intervention. We expect that the impact of hedge fund intervention on audit fees is more pronounced when the target firms CFOs change following intervention. Similarly, the financial person on the board of the firm also has the responsibility for the faithful representation and relevance of the financial report. The changes of such persons may cause the firm have less resource for financial reporting and lead to higher audit fees. Therefore, we hypothesize that: HYPOTHESIS 2.2 The impact of hedge fund activists intervention on audit fees is more pronounced for firms with CFO s or financial director s turnover. 3. Research Design and Summary Statistics 3.1. Data and Sample Selection We use a unique dataset of hedge fund activism to test our research question. Following Brav et al. (2008), we start our data collection process by gathering the entire Schedule 13D filings from the SEC s EDGAR database between year 2003 and Using the information contained in item 2 of Schedule 13D, we then exclude filers that are classified as banks, brokerage companies, regular corporations, foreign institutions, individuals, insurance companies, pension funds and trusts. After cross checking with the activist hedge funds contained in Brav et al. (2008), we use Google search to pin down a list of activist hedge funds from the remaining filers. We then exclude those Schedule 13D filings that are related to risk arbitrage, distress-financing, and M&As, or those target investment trusts or closed-end funds. The remaining Schedule 13D files filed by the list of activist hedge funds represent all the activist events where more than 5% of the target company s shares are owned by the activist hedge funds. For hedge fund activism events that are not contained in Schedule 13D (fewer than 5% of the target company s shares are owned), we 11

14 obtain the information from FactSet s SharkWatch dataset where detailed information regarding the development of the hedge fund activism campaign, such as proxy fight and exempt solicitation, is also contained. Before 2001, there are few hedge fund activism targeting events. We exclude two-year events before and after our sample period (2001, 2003, 2013 and 2014) to incorporate enough auditing and accounting data of targets firm before and after the events. In the baseline OLS regression and the difference-in-difference model, we test the difference two years before and after targeting. Starting with 4,990 hedge fund activists target events during the sample period, we only include the first-time target event for each firm and require at least one year auditing and accounting data before and after the event. In total 1,351 hedge fund activism targeting events are included in our sample. The detailed data selection process is presented in Table 1. [Insert Table 1 here] Since 2003, the number of hedge fund intervention events per year has been mostly stable, although there s a spike in 2006 and 2007 and a small decline after financial crisis. The distribution of the events across the years is summarized in Table 2. The distribution is comparable to the prior literature (Brav et al. 2015). [Insert Table 2 here] We extend the events sample to panel data to include observations across all the years. After merging with auditing data in Audit Analytics and accounting data in Compustat across, the dataset has 12,293 firm-year observations. For baseline OLS regression, we use the observations two years before and after the events. The window [-2, +2] indicates the event years t-2, t-1, t, t+1 and t+2, where t is the event year that the hedge fund activist targets the firm. In total 5,329 observations are within this 5-year event window. 12

15 3.2. Research Design To investigate the effect of hedge fund activism on audit fees, we test a dummy variable POST, which indicates whether year t is before or after hedge fund invention. We run an OLS regression of the natural logarithm of audit fees (LAUDIT) on POST. This regression model is our baseline. We limit our sample to two years before, two years after and the event year, totally 5 relative event years. We run regressions with alternative event windows as robustness checks. For the testing of Hypothesis 1, we interact the dummy variables with POST. We use the dummy variables to indicate whether there are significant changes in ROA and LEV. We sort the firms by the absolute value of changes in ROA from event year t-1 to event year t. We use a dummy variable ROA_CHG to indicate whether the changes are above the median of the sample. It takes the value of 1 if the firm is above the median, 0 otherwise. LEV_CHG is defined in a similar fashion. For the testing of Hypothesis 2, we use the dummy variables to indicate whether there are changes of CFO or financial person on the board of directors. CFO_CHG takes the value of 1 if there s any CFO changes within the event window [0, +2], and 0 otherwise. CFO_POST is the interaction between POST and CFO_CHG. FIN_CHG and FIN_CHG_POST are defined in a similar fashion. For sensitivity tests, we conduct propensity score matching and difference-in-differences model. The test clarifies the causality between hedge fund intervention and audit fees. We also test alternative event windows and include more control variables. 3.3 Summary Statistics 13

16 Table 2, Panel A presents the summary statistics of all the target firms for the 5 event years. In total 5,329 firm-year observations are in our sample. The continuous variables are winsorized at +/- 1 percent. We compare the summary statistics of the variables with the prior literature (eg. Bills, Cunningham and Myers 2015; Carcello et al. 2002) and find similar results. The mean of AUDITOR is 0.666, which means 66.6% of our sample have Big-N auditors. The high percentage of the Big-N auditors is because public companies tend to appoint Big-N auditors. The mean of FOREIGN is 0.396, meaning that 39.6% of our sample have foreign pre-tax income. Table 2, Panel B presents the summary statistics for the comparison between before targeting and after targeting. We include the year of being targeted in the after targeting subsample. There are 2,252 observations in the before targeting subsample and 3,077 observations in the after targeting subsample. We also conduct t test on the means of these two subsamples. The mean difference of LAUDIT before and after targeting is positively significant (t stat 6.94). This univariate result is consistent with our hypothesis that audit fees increase after hedge fund intervention. The mean difference of ROA before and after targeting is negative but not significant. This result is contradictory to the findings in the prior literature that target firms improve performance after targeting. However, the mean difference is relatively small. The variable AUDITOR is also significantly positive after intervention. This is consistent with our conjecture that auditor changes may be one channel for the impact of hedge fund activism on audit fees. 4. Empirical Results 4.1. Baseline OLS Regression (H1) 14

17 We adopt the audit fee model developed by Simunic (1980) and follow prior literature (e.g., Johnstone and Be dard 2003; Gul and Goodwin 2010; Hanlon, Krishnan, and Mills 2012; Bentley et al. 2013) in selecting the control variables. LAUDIT i,t = β 0 + β 1 POST i,t + β 2 ROA i,t + β 3 LOSS i,t + β 4 SIZE i,t + β 5 AUDITOR i,t + β 6 GCO i,t + β 7 FOREIGN i,t + β 8 RECV i,t + β 9 INVINT i,t + Industry Dummies + ε i,t (1) LAUDIT is measured as the log of audit fees. POST is a dummy variable that takes the value of one if year t is after hedge fund intervention, and zero otherwise. In this model, we limit our sample to observations two years before and after targeting. We estimate the equation using OLS regression and robust standard errors clustered at the company level. In the model, we control for firm fundamentals ROA, LOSS and SIZE. We expect that ROA has negative association with audit fees, because auditors charge less risk premium on firms with high profitability. For the same reason, we expect that LOSS has positive association with audit fees. We predict positive sign on SIZE because large-size firms need more audit effort. We also control for AUDITOR and GCO. It has been examined in the prior literature that Big-N auditors charge a premium on clients for branding image, thus we expect positive coefficient on AUDITOR. We predict a positive coefficient on GCO since firms with going concern opinions are riskier and need more audit effort. Following the prior literature, we control for audit complexity using three proxies: FOREIGN, RECV and INVINT. Firms with foreign operation need more audit effort such as traveling and translation. RECV and INVINT are proxies for audit task complexity because the auditing of these items involves more judgements and riskier. We also expect them to be positively associated with audit fees. For detailed description of the other variables, refer to Appendix A. We report the OLS regression result in Table 3. Model 1 is OLS regression without control variables and industry fixed effect. Model 2 includes control variables. Model 3 includes both 15

18 control variables and industry fixed effect. Our variable of interest POST is significantly positive across three models. The coefficient of POST in model 3 is (t stat 13.81). It indicates that the target firms audit fees increase by 22.1% after hedge fund intervention. We find that the signs of most control variables are as expected, although the coefficient of INVINT is not significant. The R-squared is also comparable to the prior audit fees studies (e.g., Simunic 1980; Stanley 2011; Bentley et al. 2013). This finding in the baseline OLS regression model supports our first hypothesis that audit fees increase after hedge fund intervention. [Insert Table 3 here] 4.2. Business Risk and Financial Reporting Resources In this section, we channel tests. Our variable of interest is the interaction term between POST and the dummy variables that indicate the level of business risk or lacking in financial resources. The results are presented in Table 4. [Insert Table 4 here] In column (1), the dummy variable ROA_CHG indicates the level of changes in ROA from event year t-1 to event year t. ROA_CHG takes the value of 1 if the changes are above median, and 0 otherwise. Our variable of interest is the interaction term ROA_CHG_POST, which is positively significant at 10% level. It means that for the firms that have high changes in ROA, audit fees increase even more after hedge fund activists targeting. In other words, auditors react to high operating performance volatility by charging higher audit fees. In column (2), the dummy variable LEV_CHG indicates the level of changes in LEV from event year t-1 to event year t. LEV_CHG takes the value of 1 if the changes are above median, and 0 otherwise. Our variable of interest is the interaction term LEV_CHG_POST, which is positively 16

19 significant at 1% level. It means that for the firms that have high changes in LEV, audit fees increase even more after hedge fund activists targeting. In other words, auditors react to high leverage volatility by charging higher audit fees. The findings in column (1) and (2) support our second hypothesis. Business risk is the underlying channel for the positive relationship between hedge fund activists intervention and audit fees. In column (3) and (4), we test the third hypothesis about the effect of financial reporting resources. We use financial executive turnovers or financial director turnovers to measure lack of financial reporting personnel. Our variables of interest are the interaction terms CFO_CHG_POST and FIN_CHG_POST. In column (3), the coefficient of CFO_CHG_POST is positively significant at 1% level. It means that auditors react to hedge fund activists intervention by charging more audit fees when the target firms are lacking experienced CFO. The result in column (4) can be interpreted in the same way. The findings in column (3) and (4) support our third hypothesis Identification Propensity Score Matching One may argue that auditors reaction to hedge fund activism may result from the firm characteristics associated with a high likelihood of becoming a hedge fund target firm. To mitigate this concern, we identify matched target and non-target firms using propensity score technique and test with difference-in-differences model. Before matching, we identify a pool of candidate match firms as the public firms that have not been targeted by hedge fund activists during our sample period. In total 89,092 firm-year observations are in the pool. We follow recent studies (e.g., Hasan et al. 2014) and use propensity 17

20 score matching to identify one match firm for each treatment firm. These two groups of firms ideally have the same firm characteristics except that one group of firms have been targeted by the hedge fund activists and that the other group of firms haven t. We identify a non-target control firm with the closest propensity score in event year t-1. We use the probit model the estimate the probability of being targeted by hedge fund activists to obtain the propensity score. We modify the probit model used by Brav et al. (2008) by including LAUDIT as control. The model can be presented as follows: D_Target i,t = β 0 + β 1 LAUDIT i,t 1 + β 2 MV i,t 1 + β 3 Q i,t 1 + β 4 GROWTH i,t 1 + β 5 ROA i,t 1 + β 6 LEV i,t 1 + β 7 DIVYLD i,t 1 + β 8 RND i,t 1 + β 9 HHI i,t 1 + β 10 ANALYST i,t 1 + β 11 INST i,t 1 + Industry Dummies + Year Dummies + ε i,t (3) D_TARGET equals one if the firm is a hedge fund activism target in the year t and zero otherwise. MV is the log of market capitalization. Q is Tobin s Q. GROWTH is the growth rate of sales over the previous year. DIVYLD is the dividend per share. RND is R&D scaled by total asset. HHI is the Herfindahl-Hirschman index of sales in different business segments as reported by COMPUSTAT. ANALYST is the number of analysts covering the firm. INST is the institutional ownership. Industry dummies are defined using 2-digit SIC industries. For detailed description of the other variables, refer to Appendix A. The result of the probit model is presented in Panel A, Table 5. The results are mostly consistent with Brav et al. (2008). Most variables except ANALYST have the same signs of coefficients as in Brav et al. (2008). Target firms tend to be low growth firms, but are significantly more profitable. The variables for growth such as q and GROWTH have negative coefficients. The variable for profitability ROA has positive coefficient. The negative coefficient on DIVYLD indicates that target firms dividend payout is relatively lower than peer firms. They are also 18

21 relatively less competitive (coefficient of HHI is , significant at 1% level). Target firms tend to have higher institutional ownership (coefficient of INST is 0.135, significant at 5% level). The interpretations of these variables are all consistent with the findings in the prior literature. [Insert Table 5 here] In the next step, we obtain the predicted value of the probit model as the propensity score. Without replacement, we match each hedge fund target firm with a control firm that has the closest propensity score in the same year. Firms that are matched in prior years will be excluded from the pool of candidate match firms. To ensure that there are no significant differences between treatment firms and match firms, following Hasan et al. (2014), we use the caliper matching method, in which caliper refers to the difference in the predicted probabilities between treatment and match firms. By matching within a caliper of 1%, we are able to identify 812 treatment-match pairs from this propensity score matching method. We compare the statistics of the treatment and match firms. The comparison is presented in Panel B, Table 5. For most variables, the t statistics for the mean difference between the treatment firms and the match firms are not significant. It means that our matching method gives balanced treatment and match samples. After matching, we expand the sample to include the proceeding and succeeding firm-year observations for each treatment and match firm Difference-in-Differences Model One could argue that economy-wide shocks contemporaneous with hedge fund activists targeting events could also cause changes in audit fees of target firms. To address this concern, we use the following difference-in-differences model to capture the causal effect of hedge fund activists targeting events on subsequent audit fees changes: 19

22 LAUDIT i,t =β 0 + β 1 TREAT i,t + β 2 POST i,t + β 3 TREAT*POST i,t + β 4 ROA i,t + β 5 LOSS i,t + β 6 SIZE i,t + β 7 AUDITOR i,t + β 8 GCO i,t + β 9 FOREIGN i,t + β 10 RECV i,t + β 11 INVINT + Industry Dummies + ϵ i,t i,t (4) In the model above, TREAT indicates whether the firm is in the treatment group (target firms) or control group (non-target firms). POST indicates whether it is the year after the hedge fund intervention. Consistent with our baseline regression model, we use [-2, +2] event window (5-year window) for this difference in differences model. Our variable of interest is TREAT*POST, which is the interaction term between TREAT and POST. Given that treatment firms have higher audit fees after hedge fund intervention, a positive (negative) coefficient on this interaction term indicates that hedge fund activists targeting leads to increases (decreases) in audit fees. We use the same set of control variables in the difference in differences model. Consistent with our baseline models, model 1 doesn t include control variables and industry fixed effect; model 2 includes controls variables but no industry fixed effect; model 3 includes both control variables and industry fixed effect. The results are reported in Table 6. [Insert Table 6 here] The coefficients of TREAT are not significant across the three models. It indicates that our treatment firms and match firms are not significantly different. The coefficients of POST are still significantly positive, which is consistent with our baseline regression model. Mostly importantly, our variable of interest in this difference in differences model is significantly positive at 5% level. The coefficient of TREAT*POST in model 2 is (t stat 2.477). It means that the hedge fund target firms audit fees increase by 7.7% after the intervention, compared with match firms that 20

23 are not targeted by hedge fund activists. The coefficient of TREAT*POST in model 3 is (t stat 2.349). These findings confirm our results in the baseline regression model. The model provides strong evidence on the causal effect of hedge fund intervention on audit fees Robustness Tests The results above support our first hypothesis that target firms audit fees increase after hedge fund intervention. In this section, we conduct sensitivity tests to provide more solid evidence Other Control Variables One may argue that a few observable variables that are correlated with our variable of interest are not controlled in our model. To mitigate this concern, we include more control variables in our model. One may argue that hedge fund activists cause higher audit fees by demanding changes of auditors. We include AUDITOR_CHG to control for such effect. To control for any financial constraints and capital market valuation, we include leverage (LEV) and book-to-market ratio (BM). We also include a set of variables to control for auditors characteristics. BUSY indicates whether the audit report deadline is within auditors busy season. We include this variable to proxy auditors time constraint. One line of auditing literature argues that auditors independence has a significant effect on audit pricing. We use TENUE, LNFEE_NON and DUMMY_NONAUD to proxy for auditors independence. After including these control variables, we rerun the baseline OLS regression model and difference-in-differences model. To save space, the results are not tabulated in this paper. The coefficient of POST is (t stat 15.67), which is comparable to our baseline 21

24 result (t stat 13.81). The coefficient of TREAT*POST in difference-in-differences model is also significantly positive at 10% level. These findings are consistent with the prior results Alternative Rolling Windows We use different rolling windows to test the sensitivity of our empirical results. While we use a 2- year rolling window in our main analyses in this paper, we test the robustness of the results by using both a 1-year and 3-year window around the hedge fund intervention, that is, a total period of 3 year and 7 years, respectively. Since we define the entire calendar year of being targeted as the event year t and include it into the post-targeting sample, it may confound the effect of hedge fund activism on audit fees. To mitigate this effect, we test the result by excluding the event year and test the 1-year, 2-year and 3-year window again. We find that the variables of interest in the baseline regression models and the difference-in-differences model are still statistically significant and have the same signs of coefficients. These results suggest that the evidence is robust to the choice of observation period Possible Alternative Explanations In this section, we rule out other possible explanations for the positive relation between hedge fund activism and audit fees. The relationship we examine involves three parties: hedge fund activists, target firms and auditors. We run tests in section and to rule out the possible confounding effects from hedge fund activists and target firms. We also concern that the effect can be due to regulators high attention on hedge fund activism. We test for it in section Hedge fund activists influence on corporate governance 22

25 The prior literature has shown that hedge fund activists have significant impact on the target firms in various aspects. In particular, hedge fund activists aim to change target firms capital structure, change business strategy, sell target firms, or improve governance. By changing the target firms in these aspects, the hedge fund activists have been proved to create value to shareholders. The seminal paper Brav et al. (2008) attributes the value creation mostly to the improvement in governance. The improvement in governance includes hiring more incentivized managers, change the board chairman, having more independent board members and requesting more information disclosure. In our context, one can argue that hedge fund activists demand changes in target firms and cause increases in audit fees. For example, the hedge fund activists may hire the new CEO or change the chairman of the board, who may have higher standards for audit quality and request more auditors effort. As a result, audit fees may increase. We test the possibility of this explanation and show the results in Table 7. We focus on the changes in governance and run four tests, including the changes in directors, changes in chairperson, changes in CEO and changes in governance. In column (1), DIRECTOR_CHG indicates whether there s any director changes within two years (event window [0, +2]) after hedge fund targeting. In column (2), CHAIR_CHG indicates any chairman changes within two years after hedge fund targeting. In column (3) indicates any CEO changes within two years after hedge fund targeting. In column (4), we select a small subsample where the hedge fund activists declare their goals of targeting the firms. GOV_CHG takes the value of 1 if the hedge fund activists successfully change the target firms governance, and 0 otherwise. In each column, we interact the variable with POST. For example, DIRECTOR_CHG_POST is the interaction between DIRECTOR_CHG and POST. The interaction terms are our variables of interest. In each column, we find no significance in this interaction term. In column (1), the interaction term 23

26 DIRECTOR_CHG_POST has t statistic 0.755, which is not significant at 10% level. This means that director changes are not the underlying channel for positive relationship between hedge fund activism and audit fees. In column (2), the insignificance in CHAIR_CHG_POST means that chairman changes cannot explain the effect on audit fees either. The findings in column (3) and column (4) can be interpreted in the similar way. Based on these findings, we conclude that the influence that hedge fund activists have on the target firms, especially improvement in governance, does not lead to changes in audit fees. [Insert Table 7 here] Target firms demand for high audit quality In this section, we examine the possibility that the target firms demand higher audit quality as a self-defense to hedge fund activism. The managers of the potential target firms may change the firms to deter any hedge fund activists. In the process of changing the firms, the managers may use auditors as a mechanism to identify any potential problems, especially in the financial reporting. For example, the managers may request the auditors identify any accounting information disclosure problem, which the hedge fund activists are also able to identify and aim to change in the target firms. We test this alternative explanation and show the results in Table 8. [Insert Table 8 here] We select a subsample in our data, where the target firms disclose whether there s a proxy fight against hedge fund activists. The variable of interest is FIGHT, which takes the value of 1 if the firm discloses a proxy fight, and 0 if the firm discloses Exempt Solicitation or no publicly disclosed activism in the 13D filings. Our variable of interest is the interaction term FIGHT_POST between FIGHT and POST. We find no statistical significance in FIGHT_POST, which means 24

27 that the target firms that attempt to fight against the hedge fund activists do not have significantly higher audit fees than those do not fight. In other words, we find no evidence that target firms request more audit effort for the purpose of self-defensing Regulators increased attention The last alternative explanation that we concern about is regulators attention. The recent rise of hedge fund activism has caught the regulators attention. We are concerned that the securities and exchange commission (SEC) may increase scrutiny on the target firms such as initiating inspections in the target firms to detect any financial reporting problems. To test this possibility, we use the Accounting and Auditing Enforcement Releases (AAER) as our measure. AAERs are financial reporting related enforcement actions concerning civil lawsuits brought by the Commission in federal court and notices and orders concerning the institution and/or settlement of administrative proceedings 2. We obtained the AAERs dataset that is organized by Center for Financial Reporting and Management at University of California Berkeley. We use two dummy variables to indicate whether there is any AAER issued. AAER_ANN takes the value of 1 if SEC issues an AAER for the annual financial report, and 0 otherwise. AAER_QTR takes the value of 1 if SEC issues an AAER for the quarterly financial report, and 0 otherwise. We include each dummy variable as a control into the regression model. [Insert Table 9 here] The results are presented in Table 9. In column (1), the coefficient of AAER_ANN is not significant, meaning that the SEC inspection to annual report does not have an effect on the audit fees. The coefficient of POST is unchanged compared to the baseline regression in Table 3. This 2 Refer to SEC AAER website: 25

28 finding means that the positive relation between hedge fund activism and audit fees is not due to the SEC inspections to annual report. In column (2), we include AAER_QTR as a control. The results are similar to column (1) and can be interpreted in the same way How do auditors respond risk premium or audit effort Based on our finding that auditors react to the target firms business risk, we further examine specifically how auditors react. DeFond and Zhang (2014) summarize four strategies auditors use to counter risk: (1) reduce risk by increasing effort; (2) price risk by charging a premium; (3) avoid risk through client retention and acceptance; and (4) attenuate risk through lobbying. The first two strategies are related to our topic since they lead to higher audit fees. In this section, we try to distinguish these two strategies, although it has been difficult to do so in archival research. We do not have data on audit effort such as working hours. We do not have measures for risk premium either. However, the prior literature has used audit report lag as a proxy for audit effort. We also test whether misstatement risk has any impact on the relation between hedge fund activism and audit fees. The results are presented in Table 10. [Insert Table 10 here] In column (1), the dependent variable EFFORT is measured as the number of days between the date of fiscal year end and the audit report signature date. The variable of interest POST is statistically significant at 1% level. This means that auditors do react to the business risk by increasing audit effort. For the test of risk premium, we estimate the misstatement risk using p score, which is developed by Dechow et al. (2011). It captures each firm s probability of misstatement. To better interpret the results, we use a dummy variable P_SCORE to indicate the level of misstatement risk. It takes the value of 1 if a firm s probability of misstatement is above 26