2012年12月3日学术报告通知.doc

Politically Connected Boards and Audit Pricing Chansog (Francis) Kim* Zhifeng Yang Oliver Zhou City University of Hong Kong October, 2012 *Corresponding author Contact: Chansog (Francis) Kim (Tel: 852-3442-7962, Email: acckim@cityu.edu.hk), Zhifeng Yang (Tel: 852-3442-4013, Email: zhifeng@cityu.edu.hk), Oliver Zhou (Tel: 852-3442-2912, Email: yuxizhou@student.cityu.edu.hk). We thank Jinghua Fu for his research assistance in data processing. Politically Connected Boards and Audit Pricing Abstract This study examines how clients’ political connections influence auditors’ assessments of audit risk that is reflected in audit fees in the United States. On the one hand, political connections may help improve the financial performances of connected firms by receiving the economic rents extracted from the connections. They also increase connected firms’ chances of being bailed out in case of financial distress. As such, clients’ political connections are predicted to reduce audit risk and thus audit fees. On the other hand, politically connected firms tend to be more opaque in financial reporting. Connected firms are also subject to the additional risk related to election outcomes and political misfortune of their backers. Hence, we predict higher audit fees for client-firms with political connections than for non-connected firms. Our empirical results for a large sample of U.S. publicly listed firms show that auditors charge higher fees to politically connected firms than to non-connected ones. We also find that this adverse effect of firms’ political connections on audit fees is more pronounced among firms with weaker corporate governance and with more complicated operational structures. Furthermore, this relation becomes more pervasive in the post-SOX era. These findings suggest that auditors perceive politically connected firms to be riskier. Accordingly, they exercise greater effort and/or charge higher fees to these connected firms. Our evidence is robust to a battery of econometric endogeneity remedies and to such exogenous events as presidential and mid-term elections and financial crisis. 1 1. Introduction U.S. firms become politically connected by inviting former politicians, e.g., former House Representatives, Senators, and Secretaries of Departments, to sit on their boards. Extant studies find that such political connections influence the financial reporting practices of connected firms.1 Ramanna and Roychowdhury (2010) provide evidence that U.S. firms with connections through political contributions are more likely to manage earnings during the 2004 elections. Yu and Yu (2011) find that firms’ lobbying activities slow down the regulatory detection of accounting fraud committed by connected firms, which enables their managers to manipulate earnings for longer periods, compared to non-lobbying ones. These findings suggest that a client’s political connections should influence auditors’ assessments of audit risk of those connected firms, due to their differential reporting behaviors—an issue that has received increasing attentions from practitioners. For example, in a recent report, Deloitte (2012) calls for more transparency about corporate political activity and the acceptance of board directors with political backgrounds. In this regard, it is warranted to examine how clients’ political connections affect auditors’ assessments of audit risk and, in turn, audit fees. To our best knowledge, this is the first study examining the effects of political connections on audit pricing in the US setting – A country with high litigation risk. Ex ante it is not clear how and to what extent political connections influence audit risk and audit fees. Political connections may help improve financial performances of connected firms by receiving economic rents including lower costs of capital and For international evidence on political connection effect on financial reporting, see Leuz and Oberholzer-Gee (2006) for income smoothing in Indonesia, and Chaney, Faccio, and Parsley (2011) for accrual quality in their cross-country study. 1 2 more government procurement contracts (Houston, et al., 2011; Karpoff et al., 1999; Goldman, et al., 2012). Such political benefits may reduce the management incentives to misreport earnings of connected firms (Gul, 2006). Moreover, politically connected firms may face a lower failure risk, because they are more likely to be bailed out by their government when distressed (Duchin and Sosyura, 2012; Faccio, Masulis, and McConnell, 2006). Consequently, it mitigates the potential class-action lawsuits by shareholders due to corporate failure. As such, auditors may charge lower fees to clients with political connections. However, the covert nature of corporate rent-seeking activities may increase the opaqueness of financial statements prepared by politically connected firms. The executives of such firms have incentives to obscure their financial statements to protect political favors, which are often of a dubious nature (Chaney et al., 2011; Fisman, 2001; Leuz and Oberholzer-Gee, 2006). These problems suggest that accounting quality could be poorer in politically connected firms, which is confirmed by prior research including studies by Chaney, Faccio, and Parsley (2011), Ramanna and Roychowdhury (2010), and Yu and Yu (2011). In addition, politically connected firms could be riskier because their performances might change significantly, due to changes in political power (Leuz and Oberholzer-Gee, 2006; Fisman, 2001). Market values of politically connected firms in the U.S. are influenced by election outcomes (Goldman et al., 2009).2 Furthermore, politically connected firms tend to commit more accounting fraud and have a higher chance of being collapsed, as in Enron and WorldCom (Yu and Yu, 2011). The aforementioned arguments lead to higher audit risks for politically 2 Fisman (2001) in the Indonesian setting and Johnson and Mitton (2003) in the Malaysian setting also provide further evidence of political risk. 3 connected clients. This perspective predicts that auditors should exercise more effort in auditing politically connected firms, and thus charge such firms higher audit fees. We also expect that the phenomenon of higher audit fees being charged to connected firms should be more pronounced in the U.S., due to its highly litigious environment. Thus, we focus on U.S. firms from 2001 to 2009 in order to assess political connections effects on audit fees. We use human-tie based political connections. A firm is defined as politically connected if one or more of its directors held political positions before sitting on the board. Our findings suggest that auditors charge significantly higher audit fees to firms with directors who have formal political experience. In addition, this finding is robust to the inclusion of other measures for political connections including political contributions and lobbying expenditures. Our findings show that auditors charge higher fees to politically connected firms than to non-connected ones. We contend that this relation between firms’ political connections and audit fees should be more pronounced among firms with weaker corporate governance and with more complicated operation structures. Any adverse effects of political connections on audit risk and audit fees should be mitigated, if corporate insiders are closely monitored by inside directors and outside institutional shareholders. However, their abilities to monitor should be decreased if the firms’ operations become more complicated due to extensive geographic or product-line diversification. We also argue that this relation should become more prevalent in the post-SOX era, because the litigation environment against auditors became more stringent after passage of the Sarbanes–Oxley Act (SOX) in 2002. We find that this relation between firms’ political connections and audit fees is more pronounced among firms with weaker internal governance (proxied by directors’ shareholdings) and external governance (proxied by institutional 4 shareholdings). In addition, we find the mentioned relation is stronger among firms with more complicated operational structures including multi-segment firms and MNCs. The relation also becomes more ubiquitous in the post-SOX period, as the litigation risk imposed on auditors is intensified. On balance, these findings suggest that auditors perceive politically connected clients to be riskier. Auditors exercise greater efforts to reduce audit risk of these connected firms. As a result, they charge higher fees to politically connected clients than to non-connected firms. However, our findings are also subject to the endogeneity problems which are common in auditing literature. The first concern is the reverse causality problem, i.e., firms that are more opaque in external reporting and/or riskier in their business natures are more likely to establish political connections.3 The other concern is the omitted correlated variable problem, i.e., the audit fee model fails to control for some unobservable firm characteristics that may simultaneously increase audit risk and the likelihood of establishing political connections. We conduct a battery of robustness tests to address these endogeneity concerns. First, we use an instrumental variable approach to address these endogeneity problems. As in Houston, et al. (2011), we use two alternative instrumental variables: (1) distance of firms’ headquarter from Washington D.C. and (2) the percentage of politically connected firms in the industry to which a firm belongs. Firms that are close to Washington D.C. may have better opportunities to develop political connections. In addition, the propensity to establish political connections can vary due to the differential values of connections in different industries (Agrawal and Knoeber, 2001). However, we contend that these two instrumental variables do not directly affect audit fees. In addition to using two stages least squares (2SLS) methods, we 3 This concern is also addressed by Chaney, Faccio and Parsley (2010). These authors show that ex ante earnings quality does not affect the chance of political connection establishment. 5 also utilize a Heckman IMR method to alleviate endogeneity concerns from selection bias due to unobservables (Tucker 2011). Second, we also use a propensity score matching (PSM) strategy to mitigate endogeneity problems from selection bias due to observables. For each politically connected firm, we select a non-connected firm with a similar score on propensity for having political connections. Then, we run audit fee models using both the politically connected firms and their matching non-connected ones. The empirical results of both the instrumental variable analysis and the propensity score matching approach confirm our previous findings that political connections lead to higher audit fees, and this relation is economically and statistically significant. Third, in addition to making use of econometric remedies to endogeneity issues, we also utilize such exogenous events as mid-term and presidential elections to address endogeneity concerns. We examine whether audit fees charged to connected firms are different between election and non-election years. This test is motivated by Ramanna and Roychowdhurry (2010). These authors argue that the incentives for connected firms to manage earnings are reinforced in election years, due to the enhanced public scrutiny of such firms during the election period. As a consequence, the relation between political connections and audit fees ought to be stronger in election years. Our results ascertain that the relation between connections and fees is indeed more pronounced during the election period. As our multiple robustness checks confirm our results of the adverse political connection effect on audit fees, we conclude that our findings cannot be interpreted by the endogeneity-based arguments. There are two alternative explanations of our findings. First, one may argue that political directors may lack expertise and incentives to monitor managers. Thus, managers of such firms may have more opportunities to misreport earnings. In this 6 regard, auditors may have to exercise greater efforts in auditing firms with political directors. Second, due to the fact that political directors receive greater publicity than other directors, they demand high quality audit from auditors to protect their political reputations. To evaluate these alternative interpretations, we partition our sample of connected firms based on political directors sitting on audit committee. If our findings are indeed consistent with the views that political directors lack incentives and expertise or political directors excessively demand higher audit quality to protect their reputations, these audit fee effects of political connections should be more pronounced among firms with political directors sitting on the audit committees. However, we fail to find a significant difference in audit fees between firms with / without political directors sitting on the audit committee. Therefore, this evidence provides little support to these alternative explanations. Our work appears to be closely related to that of Gul (2006). This author examines how auditors charge audit fees to politically connected firms in Malaysia during an abnormal period, the Asian financial crisis of 1997 and 1998. In addition, Gul’s findings may not directly apply to the U.S. setting, because Malaysia and U.S. have significant difference in the strength of their economic and legal institutions (La Porta, et al., 1998). Litigation risk faced by auditors in U.S. is substantial and significant as opposed to the limited litigation risk in Malaysia (Choi, et al., 2008). Gul (2006) shows that auditors charge lower audit fees to politically connected firms in Malaysia when Malaysian government regains its capital control. As opposed to his findings, we find that auditors charge higher audit fees to connected firms than to non-connected firms in U.S. This difference can be explained by the disparity in institutional/legal environments. Our paper contributes to the literature in several aspects. First, to the best of 7 our knowledge, our paper is the first study using a large sample of 29,785 firm-year observations which are hand-collected from all listed firms in U.S. for the period of 2001 to 2009. The sample of Gul (2006) is limited to 740 firm-year observations because he focused on such a specific period as the Asian financial crisis in Malaysia. Second, we carefully control for the endogeneity issue of political connections as opposed to Gul (2006). Our results are robust to the econometric solutions and exogenous events like both presidential and mid-term elections. Our findings are also robust to financial crisis period of 2008 and 2009. Third, our study is related to the literature on how political and institutional factors affect auditors’ behaviors (Gul, 2006; Choi, Kim, Liu, and Simunic, 2008 & 2009; Wang, Wong, and Xia, 2008; Guedhami, Pittman, and Saffar, 2009 & 2011).4 Our findings suggest that auditors perceive politically connected clients as riskier. As such, they exercise greater effort in auditing these connected firms and charge higher audit fees to these connected firms than to non-connected firms. By doing so, this paper documents the additional costs of corporate political activities on top of corporate expenditure for establishing and maintaining political connections, as contrary to the conventional political connection literature which documents firms’ political benefits. Finally, this paper shows that political connections matter to auditors and they properly incorporate firms’ political connectedness in accessing audit risk of those connected firms vs. non-connected ones. Our findings are in a sharp contrast with the evidence of Gul (2006). This author show that auditors charge lower audit fees to politically-connected firms than to non-connected firms, when Malaysian government 4 Refer to the following literature on how political economy shapes the financial reporting process around the world (Bushman, Piotroski, and Smith, 2004; Bushman and Piotroski, 2006; Piotroski, Wong, and Zhang, 2011; Leuz and Oberholzer-Gee, 2006; Chaney, Faccio, and Parsley, 2010). 8 regains its capital control during the Asian financial crisis periods. In this regard, our study is in line with Jensen and Meckling (1976). These authors maintain that auditors play a governance role in mitigating agency problems. We provide evidence that the governance role of auditors extends to politically connected firms as well as to non-connected clients. Our findings are also consistent with those international studies of Wang, Wong, and Xia (2008) and of Guedhami, Pittman, and Saffar (2011) who document the corporate governance role of auditors. The remaining sections of this paper are organized as follows. In the next section, we analyze the positive and negative effects of political connections on audit risk and audit fees, and develop testable hypotheses. Section 3 introduces our political connection measures and audit fee model. Section 4 reports empirical results followed by conclusions in Section 5. 2. Hypothesis Development 2.1. Political Connections and Audit Risk Audit fees are determined by audit risk, which is in turn affected by managers’ incentives to misreport and by the probabilities of corporate failure (Simunic, 1980; Francis and Stokes, 1986; Gul, 2006; Choi, et al., 2008 & 2009). Auditors expect to exercise more auditing effort to detect potential accounting irregularities if the managers’ incentives to misreport are stronger, and this extra effort leads to higher audit fees. Furthermore, auditors may charge extra fees to clients with a higher risk of shareholder litigation, which is often triggered by corporate failure (Dye, 1993). We present below a detailed analysis of how clients’ political connections may influence audit risk and consequently audit fees. The literature shows that political connections add value to the connected 9 firms. In a cross-country study, Faccio (2006) finds that stock prices rise upon the news of firms obtaining political connections. Similar evidence is documented in the United States. For instance, Goldman, Rocholl, and So (2009) show that U.S. firms accumulate abnormal stock returns upon the nomination announcement of politicians to their boards. These findings suggest that political connections are expected to bring future economic benefits to connected firms. Consistent with this conjecture, extant literature shows that political connections benefit connected firms in various forms, such as financial subsidies from the government, government procurement contracts, and relaxed regulatory oversight. Faccio, Masulis, and McConnell (2006) provide evidence showing that politically connected firms around the world are more likely to be bailed out by governments in case of financial distress. Goldman, Rocholl, and So (2012) show that U.S. firms connected with the winning party in the 1994 mid-term and 2000 presidential elections obtain significantly more government contracts after the election. Duchin and Sosyura (2012) find that the allocation of government funds under the Troubled Asset Relief Program (TARP), in response to the subprime mortgage crisis, was biased toward politician-favored firms. Houston et al. (2011) argue that banks take into consideration potential political favors in their lending decisions. These authors show that bankers make loans at lower interest rates to those with political connections. To the extent that political connections can help improve the financial performance of connected firms via more government procurement contracts and lower costs of private debts, such connections may reduce management incentives for misreporting. If so, auditors should face lower risks in auditing clients with political connections. Furthermore, politically connected firms are more likely to be rescued by the government in case of financial distress. As corporate failure more often triggers 10 shareholder lawsuits against auditors, a lower bankruptcy risk of politically connected firms should alleviate auditors’ assessments of audit risk in these firms. This perspective predicts that auditors may charge lower audit fees to politically connected clients than to non-connected firms. Although political connections may bring lucrative benefits, it is by no means costless for firms to build and maintain such connections. Politicians seek rents from connected firms in various forms, such as political donations, lobbying expenses, and directorships. U.S. firms are required to disclose “hard money” expenditures on political activities including political contributions via political action committees (PACs) and lobbying expenditures. However, the true amount of corporate political spending including “soft money” is extremely difficult to track (Cooper, et al., (2010) and Yu and Yu (2011)). The disclosed numbers could be just the tip of the iceberg. For example, the Center of Political Accountability (CPA) Report (2006) shows that very large amounts of political donations are made via trade associations. Such donations are not visible to outside shareholders, and are not monitored by outside directors in most cases. The secret nature of corporate political expenditures and its rent-seeking activities increases the opaqueness of financial statements in politically connected firms. Executives of politically connected firms also tend to make their financial statements less transparent to protect the political favors they obtain, because such favors are often of dubious legality (Leuz and Oberholzer-Gee, 2006). Leuz and Oberholzer-Gee (2006) show that politically connected firms in Indonesia are often reluctant to issue foreign securities, as they wish to avoid the more stringent disclosure requirements imposed by foreign regulators. Chaney, Faccio, and Parsley (2011) argue that the quality of earnings reported by politically connected firms are 11 lower than those of non-connected counterparts, and this is the case for two potential reasons. First, insiders have incentives to divert benefits brought by political connections, and thus intentionally make their accounting information more opaque in order to avoid monitoring by outsiders. Second, politically connected firms have preferential access to external finance, such as state bank loans and government subsidies, which may render accounting quality less important in these firms. Thus, their managers may pay less attention to accrual quality, resulting in a lower quality of accruals. Chaney, Faccio, and Parsley (2011) show that politically connected firms around the world indeed have lower accrual quality than non-connected firms. Some researchers argue that politically connected firms in the U.S. are more likely to manage earnings than their counterparts without political connections (Ramanna and Roychowdhury, 2010). Yu and Yu (2011) also show that politically connected companies are associated with higher incidences of accounting fraud. They further indicate that the regulatory detection of fraud is significantly delayed for politically connected firms, which enables managers of such firms to manipulate earnings for longer periods. Companies including Enron and WorldCom which committed massive accounting frauds and subsequently collapsed in early 2000s were politically connected. Such connections help these firms avoid fraud detection by regulators, and thus to continue their misconduct for longer years, which may contribute in part to their eventual collapse. Kido, Petacchi, and Weber (2012) examine how political forces may affect financial reporting quality in a somewhat different setting. They show that incumbent governors tend to manipulate their state government’s accounting numbers to present a strong financial picture in election years. Such findings are consistent with the notion that political considerations tend to adversely affect accounting quality. 12 Politically connected firms could also be inherently riskier because their performance “might vary dramatically over time depending on the political fortunes of their backers” and “political connections can lose their value overnight when the government fails to win an election” (Leuz and Oberholzer-Gee, 2006: pp. 2–3). Fisman (2001) show that the stock prices of Indonesian firms connected to the then-president Suharto moved up and down upon a string of rumors concerning his health. Similar evidence is documented in the United States. Goldman, Rocholl, and So (2009) find that after the 2000 presidential election, firms connected with the Republican Party gained positive stock returns, whereas those connected with the Democratic Party lost value. Corroborating with these findings, Goldman, Rocholl, and So (2012) show that after the 1994 mid-term election and the 2000 presidential election (both elections leading to power shifts from Democratic to Republican Party), companies connected with the Republican Party obtained significantly more procurement contracts than before, while those connected with Democratic Party lost such contracts. After reviewing the role of political connections in the rise and fall of five Fortune 500 companies—Enron, Global Crossing, WorldCom, Qwest, and Westar Energy—the CPA report (2005) concludes that corporate political activities were often associated with “significant issues of tax and accounting problems, fraud, bribery, conspiracy, and other illegal actions” and that could “expose companies to significant legal, reputational and financial risk” (p.11). The report concludes that a combination of these problems eventually “led to their ignominious downfall at the expense of their shareholders” (p.5). To the extent that political connections increase information opaqueness and business risk, auditors may exercise greater audit effort to discover potential 13 accounting fraud, and/or charge higher fees to cover the greater legal liability costs from potential shareholder lawsuits. It is also possible that political connections in the United States have no impact on audit risk or audit fees. The United States is commonly believed to have strong institutions, such as an independent jurisdiction system and extensive media networks for monitoring both business and politicians. These institutions may be quite effective in curbing rent-seeking through political connections. As a result, political connections may not generate a significant value to connected firms. Public firms employing politicians may simply be harnessing the expertise and knowledge their business requires. Fisman, et al. (2006) support this view by providing evidence that connections with then Vice-President Cheney did not add value to the companies concerned. Bertrand, et al. (2007) also document similar evidence that politicians in France, an industrialized and democratic country much like the U.S., deliver only very limited favors to their connected firms. 2.2. Hypotheses Development The foregoing discussion suggests that the way political connections may affect audit risks and consequently audit fees is an empirical issue. However, we argue that audit fees increase with firms’ political connections for the following two major reasons. First, Acemoglu, et al. (2010) argue that the probability of bailouts, and the size of bailouts as a percentage of GDP have been, at least until recently, much lower in developed countries such as U.S. This suggests that any reductions in audit risk for politically connected U.S. firms due to potential government bailouts may not be as great as that for their counterparts in developing countries such as Malaysia (Gul, 2006). Second, auditors in the United States face the highest litigation risk in the world, which may incentivize auditors to exert more effort in auditing 14 politically connected clients, which are usually more visible and risky. The value of the Wingate litigation index, developed by Wingate (1997) and used to measure auditors’ country-level litigation risks (Choi, et al., 2008), is 15 for the United States, but only 3.67 for Malaysia. This suggests that the litigation risk facing auditors is much higher in the U.S. than in Malaysia. Reynolds and Francis (2001) also provide evidence showing that the threat of shareholder lawsuits encourages auditors to be more conservative in auditing relatively risky clients such as large public companies. Thus, we state our first hypothesis as follows. HYPOTHESIS 1: Auditors charge higher audit fees to politically connected firms than to non-connected ones, ceteris paribus. Insiders’ agency problems are mitigated in well-governed firms. Any adverse effects of political connections on audit risk and audit fees should be mitigated, if corporate insiders are closely monitored by inside directors and outside institutional shareholders. The strength of such monitoring is in turn affected by the monitors’ incentives and ability. Numerous theoretical and empirical studies show that the incentives and abilities of directors or institutional investors to monitor company managers increases with their equity ownership (e.g., Jensen and Meckling, 1976; Beasley, 1996; Hartzell and Starks, 2003). However, their ability to monitor is reduced if the firms’ operations are more complicated due to extensive geographic or product-line diversification (Bushman, et al., 2004). These discussions lead to the following two collaborating hypotheses. HYPOTHESIS 2a: The adverse effect of political connections on audit fees is mitigated by director and/or institutional shareholdings. HYPOTHESIS 2b: The adverse effect of political connections on audit fees is more pronounced among firms with more complicated operations. The litigation environment against auditors became more stringent after 15 passage of the Sarbanes–Oxley Act (SOX) in 2002. Moreover, many of the massive accounting frauds and corporate failures involved politically connected firms. This should make auditors be more alert to the corresponding risk involved in auditing politically connected firms. Thus, the adverse effect of political connections on audit fees should become more pronounced in the post-SOX era. The last hypothesis is stated as follows. HYPOTHESIS 3: The adverse effect of political connections on audit fees becomes more pronounced in the post-SOX era than in the pre-SOX era. 3. Research Design 3.1. Definition of Politically Connected Firms We hand-collected biographic information about board directors from SEC filings, including Def-14A, 10-K, and 8-K filings. A firm is defined as politically connected if it has one or more directors who held political positions before sitting on the board. As in Goldman, Rocholl and So (2009), we define political positions as follows: President, presidential candidate, Senator, member of the House of Representatives, (assistant) secretary, deputy secretary, deputy assistant secretary, undersecretary, governor, director (CIA, FEMA), deputy director (CIA, OMB), commissioner (IRS, NRC, SSA, CRC, FDA, SEC), representative to the United Nations, ambassador, staff (White House, President, presidential campaign), chairman of Party Caucus, chairman or staff of a presidential election campaign, and chairman or member of the president’s committee/council. In addition to using a binary variable (PC Dummy) to indicate that a firm is politically connected, we also apply several other measures that indicate not only the presence of political connections but also the strength of those connections. The basic idea is that if political connections matter to audit risk, then their effect should 16 increase with the strength of political connections. These additional measures include: PC Directors, which equals the number of political directors in the board, PC Freshness, which is determined by the number of elapsed years since the most recent political position held by those political directors, PC Tenure, which is the number of years the political directors served in the government, and PC Rank, which is determined based on the strength of political power associated with political positions held by these connected directors.5 For these additional measures, a greater value corresponds to a higher degree of the strength to a firm’s political connections. The assigned value for these variables in non-connected firms is zero. The detailed calculation procedure for these variables in politically connected firms can be found in the Appendix. Such a human-tie-based definition of political connections is commonly used in international studies, such as those of Fisman (2001), Faccio (2006), Gul (2006), Leuz and Oberholzer-Gee (2006) and Fan, Wong, and Zhang (2007). A similar definition is also used in U.S. studies, such as those of Agrawal and Knoeber (2001), Goldman, Rocholl, and So (2009 & 2012), and Houston, et al. (2011).6 Some prior research uses campaign contributions or lobbying expenditures to proxy for political connections (e.g., Cooper, Gulen, and Ovtchinnikov, 2010). However, Aggarwal, Meschke, and Wang (2012) argue that campaign contributions or lobbying expenditures tend to be transactional and short-term, and that such contributions vary over time. These authors further suggest that political connections result from personal ties, and represent long-term relationships. For these reasons, we adopt a 5 Most of these measures are also used in prior research (e.g., Houston, et al., 2011). 6 Several other studies including Chaney, Faccio, and Parsley (2010), Faccio, Masulis, and McConnell (2006), Chen, Ding, and Kim (2010) and Guedhami, Pittman, and Saffar (2011), make use of Faccio’s (2006) cross-country data for human-tie based political connection. 17 human-tie-based definition of political connections. However, we do not deny that political donations or lobbying expenditures may also capture political connections to some degree. Thus, we also conduct a robustness test in Section 4, by controlling for political contributions and lobbying expenditures. 3.2. Audit Fee Model Building on the work of Simunic (1980), various researchers have modeled audit fees based on audit complexity and risk. Client characteristics such as size, leverage, asset liquidity, growth, profitability, or business and geographic diversifications are related to audit complexity and risk, and thus are found to influence audit fees (e.g., Chaney, Jeter, and Shivakumar, 2004; Gul 2006; Choi et al., 2008 & 2009; Gul and Goodwin, 2010). Therefore we include a number of variables related to these firm characteristics. We also control for peak pricing of audit services by including an indicator variable that equals 1 for firms with a December fiscal year-end and 0 otherwise (Gul and Goodwin, 2010). Prior auditing research (e.g., Craswell, Francis, and Taylor, 1995; Ferguson, Francis, and Stokes, 2003; Francis, Reichelt, and Wang, 2005; Gul and Goodwin, 2010) suggests that Big N auditing firms, or auditors that are national or city-specific industry leaders, may provide higher-quality audits and thus charge fee premiums. Following these studies, we control for dummy variables that indicate Big 4 auditors, national industry leaders, or city-specific industry leaders. Year and industry effects are also included. The baseline audit fee model is as the following. A F E E = AFEE = natural logarithm of audit fees; 18 PCON = one of the political connection measures including PC Dummy, PC Directors, PC Freshness, PC Tenure, PC Rank; SIZE = natural logarithm of total assets; LEV = total liabilities / total assets; INVREC = sum of inventory and receivables / total assets; QUICK = (current assets–inventory) / total assets; CURRENT = current assets / total assets; ROA = net income / total assets; LOSS = 1 for firms with negative net income, and 0 otherwise; NGS = number of geographic segments; NBS = number of business segments; FOREIGN = foreign sales / total sales; YE = 1 for firms with fiscal year end of December, and 0 otherwise; OPINION = 1 for firms receiving a going-concern opinion, and 0 otherwise; BIG4 = 1 for firms audited by Big 4 auditors, and 0 otherwise; NLEADER = 1 if the audit firm is the national leader for the client affiliated industry (two-digit SIC code) for that year, and 0 otherwise; CLEADER = 1 if the audit firm is the city-specific industry leader for the city where the client is headquartered for that year, and 0 otherwise; VOLATILITY = standard deviation of daily stock returns in that year; MB = natural logarithm of the ratio of market equity over book equity; YEAR = year dummies; INDUSTRY = industry dummies. 4. Empirical Results 4.1. Sample and Data The sample period starts from 2001 (which is the first year for U.S. firms to publicly disclose audit fees) and ends in 2009. We begin with 44,985 firm-years, and retain 29,785 firm-years in the final sample, after excluding those in a financial sector (9,141), a utility sector (1,032), and firms lacking required data (5,027). Table 1 reports the descriptive statistics for the firms’ political connections and financial characteristics that may affect audit fees. This table shows that about 19 11% of sample firm-year observations are politically connected. Politically connected firms have on average 1.3 politicians on their boards. The median number of years in government service for political directors is 5, and the median political rank is 3. The distributions of values for firm characteristics are reasonable, and are not discussed in detail for brevity. [Insert Table 1 about here] Table 2, Panel A provides the Pearson correlations between various political connection measures, and shows that these measures are highly correlated. Panel B shows the Pearson correlations between PC Dummy and financial characteristics, which suggest that politically connected firms are significantly larger, use more debt financing, and are more profitable than non-connected firms. These findings are consistent with Goldman et al. (2009). [Insert Table 2 about here] 4.2. Main Analyses Table 3 reports the results for testing Hypothesis 1 that auditors charge higher audit fees to politically-connected firms than to non-connected clients. The coefficient on PCON is 0.114 when political connections are measured by PC Dummy, which suggests that the audit fees charged to politically connected firms are 12% more than those charged to non-connected counterparts (). Audit fees also increase with the strength of political connections. For example, the coefficient on PCON is 0.087 when PC Directors is considered, suggesting that the extra audit fees are 9% () for politically connected firms with one political director, and 18% for those with two political directors. These findings support H1 that auditors charge higher audit fees to politically-connected firms. Results on control variables are consistent with prior audit 20 pricing research (e.g., Gul and Goodwin, 2010). [Insert Table 3 about here] We next test whether the political connection effect on audit fees is mitigated by corporate governance. We use institutional and director shareholdings to measure the quality of external and internal governance, respectively. Firms are sorted evenly into ten groups every year based on their levels of institutional ownership (director equity ownership). GOV equals the decile rank for firms, based on their director shareholdings (or institutional shareholdings). For this variable, we rank 9 for the top decile, and 0 for the bottom decile. Both GOV and GOV*PCON are added to the baseline audit fee model. If the coefficients on GOV*PCON are significantly negative, H2a is supported. Panels A and B of Table 4 report the results based on director shareholdings and institutional shareholdings, respectively. The results reveal that the coefficients on GOV*PCON are significantly negative in most columns. These results suggest that although auditors charge higher fees to politically connected clients, they tend to reduce fees to those with relatively strong internal or external monitoring. These results support Hypothesis 2a that the adverse effect of political connections on audit fees is mitigated by director and/or institutional shareholdings. [Insert Table 4 about here] For our next measure we use two indexes—the revenue-based industrial Herfindahl index, (in which a lower value suggests a higher degree of product-line diversification), and the revenue-based geographic Herfindahl index (in which a lower value suggests a higher degree of geographic diversification). Both indexes were developed by Bushman, Chen, Engel, and Smith (2004) to proxy for operational complexity. We also define a new variable, COMP, which equals 1 for firms whose 21 value of the industrial (geographic) Herfindahl index is below the sample median, and 0 for others. We include COMP*PCON to the baseline model, to determine whether the political connection effect on audit fees is stronger among firms with greater industrial (geographic) diversification. We expect that the positive effects of political connections on audit fees should be stronger among firms with more complicated operations. The results, reported in Table 5, show that the coefficients on COMP*PCON are significantly positive for all specifications, and thus support Hypothesis 2b that the adverse effect of political connections on audit fees is more pronounced among firms with more complicated operations.7 [Insert Table 5 about here] Table 6 reports the results for testing H3. Prior studies (e.g., Raghunandan et al., 2009 and DeZoort et al., 2008) consider 2002 and 2003 as the transition period from the pre-SOX era to the post-SOX era, and thus exclude these years when analyzing the economic consequence of SOX. Following these prior studies, we exclude years 2002 and 2003 from the test. Thus, the pre-SOX era includes only year 2001, and the post-SOX era starts from year 2004. SOX takes a value of one for firm-years since 2004, and 0 otherwise. We add SOX and SOX*PCON to the baseline model, and report the results from this new test in Table 6. The results suggest that auditors have charged higher fees to politically connected firms in both pre- and post-SOX periods. However, the positive effect of political connections on audit fees is much stronger after the passage of SOX.8 These results support H3 that the adverse 7 The baseline audit fee model already controls for NGS, a proxy for geographic diversification, and NBS, a proxy for business diversification. The results are qualitatively similar if COMP is defined based on either NGS or NBS. 8 We also examine whether the effect of political connections on audit fees varies with firm-specific litigation risk. We follow Shu’s (2000) approach to measure firm-specific litigation risk. Untabulated results show that the political connection effect on audit fees is more pronounced among firms with higher firm-specific litigation risk. 22 effect of political connections on audit fees becomes more pronounced in the post-SOX era than in the pre-SOX era. [Insert Table 6 about here] 4.3. Endogeneity Issues Our interpretation for the empirical results documented so far is that auditors perceive politically connected firms to be riskier than non-connected firms, and thus exert more effort and charge higher fees to connected firms. This interpretation, however, is subject to two related problems. The first of these is the reverse causality problem, that is, it is possible that firms with more opaque information or higher risk are more likely to establish political connections. The second possible problem is that there may be some unobservable characteristics of firms that simultaneously increase audit risks and the likelihood of establishing political connections. Both of these scenarios could cause the positive associations between audit fees and political connections. The endogeneity issue is a common theme in studies on the economic consequences of political connections (e.g., Houston, et al., 2011; Goldman, et al., 2009). This issue, however, is in part mitigated by Chaney, et al., (2011). These authors find that ex ante earnings quality does not affect the likelihood of establishing political connections. To further address the endogeneity concern, we perform four additional tests. We first conduct an instrumental variable analysis by using a two-stage test. In the 1st stage of this test we identify two instrumental variables, namely, DISTANCE (defined as the distance, in thousands of kilometers, between a firm’s headquarter and Washington D.C.), and PROBPC (defined as the percentage of politically connected firms in the industry each firm belongs to). Because most political positions 23 considered in this study are at the Federal government level, DISTANCE captures a measure of the opportunities for firms to approach to federal politicians, and thus the likelihood of gaining political directors. Agrawal and Knoeber (2001) find that firms from different industries have different incentives to build political connections. PROBPC captures the propensity for building political connections arising from industry characteristics. Neither DISTANCE nor PROBPC is expected to affect audit fees. The dependent variable in the 1st stage model is one of the political connection measures, and the independent variables include two instruments and all control variables of the 2nd stage structural model. The predicted values of PCON from the 1st stage model are used to measure political connections in the 2nd stage’s audit fee model. Results are reported in Table 7. The results from the 1st stage model (Panel B) show that the coefficients on both instrumental variables are significant at a conventional level of 1%. The coefficients on DISTANCE are significantly negative, suggesting that the likelihood of having political connections is reduced for firms far from Washington D.C. The coefficients on PROBPC are positive, suggesting that the likelihood of building political connections is affected by industry characteristics. Importantly, the results of the 2nd stage model show that the effect of political connections on audit fees remains significantly positive across all different measures of corporate political connectedness. [Insert Table 7 around here] The second remedy for endogeneity problems makes use of a Heckman’s IMR method as presented in Table 8. Regression results on the relationship between political connections and audit fees are presented in Table 8. As can be seen, the coefficient of political connections (PCON) remains positive and significant at a 24 conventional level of 1%. The coefficient of Mill’s lambda is negative and significant for the audit fee model suggesting that the corresponding OLS coefficient of the PCON variable would be inconsistent. Correcting for self-selection bias of corporate connectedness, the evidence continues to support the view that auditors charge higher audit fees to politically-connected firms than to non-connected ones. [Insert Table 8 around here] The third test relies on the propensity-score matching approach. Specifically, we run a Probit model with PC Dummy as the dependent variable. The independent variables include the two instrumental variables, DISTANCE and PROBPC, and the control variables of the baseline audit fee model. The predicted value from this model is the propensity score for establishing political connections. Politically connected firms are matched with non-connected firms based on their propensity scores. We adopt both the nearest neighbor and the Kernel-based matching approaches. Politically connected firms and their matched non-connected firms are then pooled to estimate the audit fee model. The results, reported in Table 9, show that politically connected firms pay higher fees than comparable non-connected firms, suggesting that political connections indeed matter to audit pricing. [Insert Table 9 around here] The fourth test contrasts the effects of political connections on audit fees between election and non-election years. This test is motivated by Ramana and Roychowdhury (2010), who argue that the incentives for earnings management in politically connected firms are heightened during election years. This happens because politically connected firms receive more public scrutiny during election years, and thus their managers have incentives to obscure financial statements in order to 25 protect themselves and their connected politicians. In this case, the audit risk associated with politically connected client-firms increases during election years, which may encourage auditors to spend more effort and charge higher fees accordingly. For this test we consider both general and mid-term elections. ELECTION takes the value of 1 for firm-years in which the company’s fiscal year-end falls within a calendar year having general or mid-term elections. Other firm-years take a value of 0. Both ELECTION and ELECTION*PCON are added to the baseline audit fee model. Results are reported in Table 10. The coefficients on ELECTION*PCON are significantly positive, as are those on PCON. These results suggest that political connections positively affect audit fees, and their positive effect is more pronounced during election years. [Insert Table 10 around here] The empirical results in this subsection do not support the endogeneity argument. Instead, they show that political connections indeed affect auditors’ assessments of audit risk, and subsequently audit fees. 4.4. Political Directors on Audit Committees as a Factor in Audit Fees One may argue that directors with political backgrounds often lack the required expertise or incentives to monitor managers, thus giving managers more opportunities to misreport earnings for their own benefit. As auditors recognize the expertise or incentive problems associated with political directors, they may exert more effort in auditing clients having political directors, in order to detect or prevent accounting irregularities. This explanation is consistent with prior research showing that agency problems affect the incidences of accounting fraud (Beasley, 1996) and 26 therefore influence audit pricing (Gul and Tsui, 1998; Gul and Goodwin, 2010). Alternatively, one may argue that political directors have greater a public profile than other directors, and they face a greater reputational loss if their affiliated firms are discovered to have accounting irregularities. Hence, political directors may request more effort from external auditors in hope of minimizing financial misreporting. This demand-side argument similarly predicts a positive association between political connections and audit fees. To evaluate these alternative explanations, we divide politically connected firms into two groups—one consisting of firms with at least one political director on their audit committee and the other consisting of those without political directors on their audit committee. The above alternative explanations are supported if the effect of political connections on audit fees is larger for the first group than for the second group. The empirical results as reported in Table 11, however, show that the difference in audit fees between these two groups is insignificant, and thus lend little support to either of these alternative explanations. [Insert Table 11 around here] 4.5. Alternative Measures for Political Connections Next we conduct additional analyses by including campaign contributions and lobbying expenditures to the baseline model, to determine the robustness of our findings. CONTRIBUTION is the log amount of political contributions made via Political Action Committees (PACs), and LOBBYING is the log amount of lobbying expenditures. We first replace PCON with CONTRIBUTION and LOBBYING, respectively, and then include both CONTRIBUTION and LOBBYING along with PCON, to estimate audit fee models. The results are reported in Table 12. 27 The results show that the coefficient on CONTRIBUTION or LOBBYING is significantly positive if the other two measures are not included, suggesting that both CONTRIBUTION and LOBBYING capture political connections to some degree.9 However, when all three measures are included in the same model, the coefficient on CONTRIBUTION becomes insignificant. This result could be consistent with the argument by Goldman, Rocholl, and So (2009), which states that political donations may represent the political preferences of a given industry more than political connections. The coefficients on lobbying expenditures are significantly positive, suggesting that firms spending more on political lobbying activity are perceived to have higher risk, and thus are charged higher fees by auditors. The coefficients on human-tie-based political connection variables remain significantly positive after controlling for the effects of political donations and lobbying expenditures. [Insert Table 12 around here] 4.6. Alternative Sample Most politically connected firms are significantly larger than non-connected firms, as demonstrated by the correlations presented in Table 2. Prior auditing research also shows that audit pricing is different to some extent between large and small firms (e.g., Simunic, 1980; Francis and Stokes, 1986). To determine whether our findings are driven by some unobserved firm characteristics correlated with firm size, we next run the baseline model based on S&P 500 firms that are relatively large and comparable in size, and report the results in Table 13. Despite reducing the sample size by 90%, the coefficients on PCON from this test are significantly positive, 9 Untabulated results indicate that the Pearson correlation between PC Dummy and CONTRIBUTION is 0.23, and that between PC Dummy and LOBBYING is 0.28. Both are significant at the conventional levels. 28 based on all but one of the political connection measures (PC Rank). [Insert Table 13 around here] 4.7. Distress Risk, Political Connections, and Audit Fees Prior literature shows that politically connected firms in developing countries are more likely to be rescued by the governments in case of distress (Faccio, Masulis, and McConnell, 2006). In Section 2, however, we argue that this type of benefits tends to be small in the U.S., because the size and frequency of bailouts in the U.S. have been minor (Acemoglu, et al., 2010). One way to evaluate this argument is to examine the effect of political connections on audit fees for firms close to distress. We assume that those firms showing financial losses are close to distress, and we modify the baseline model by adding LOSS*PCON. If political connections significantly increase the probability of being bailed out by the government in case of financial distress, then the coefficients on LOSS*PCON should be significantly negative. The results reported in Table 14, however, show that the coefficients on LOSS*PCON are not significant. This supports our hypothesis that the net effect of political connections on audit fees should be positive, because political connections do not bring much benefit in bailouts to politically connected firms in the U.S.10 [Insert Table 14 around here] 4.8. Audit Opinion Analysis We have shown that auditors charge higher fees to politically connected clients, implying that such clients are riskier. Prior auditing studies (e.g., Reynolds and Francis, 2001) show that auditors are more conservative toward clients that may impose higher auditing risks (e.g., relatively large clients). The auditors therefore tend 10 The results are similar if we use other measures for financial distress, including Altman’s (1968) Z-score, or the presence of negative operating cash flows. 29 to lower their threshold for issuing modified opinions on such firms. Prior research suggests that auditors should be more likely to issue going-concern opinions to politically connected clients, ceteris paribus, if they indeed perceive such clients to be riskier. This discussion motivates our final analysis. For our last test, we run an audit opinion model with a dummy variable that indicates reception of a going-concern opinion as the dependent variable. Following Francis and Yu (2009), we control for financial characteristics that proxy for audit risk, variables that measure auditor independence, and indicator variables for audit quality (such as Big N, national leader, and city-specific leader). We run this audit opinion model for both the whole sample, and for the sub sample consisting of firms in financial distress. Results are reported in Table 15. The results show that in both the whole sample and the distress sub-sample, auditors are more likely to issue going-concern opinions to politically connected clients. The coefficients on PCON, based on most political connection measures, are significantly positive. These results suggest that auditors are more conservative toward politically connected clients. [Insert Table 15 around here] 4.9. Pre- and Post-Financial Crisis Table 16 presents the audit-fee effects of corporate political connections for one-year period prior to financial crisis of October 2008 and for the corresponding one-year period of the post-financial crisis. Auditors charge higher audit fees to connected firms than to non-connected ones for both pre- and post-financial crisis periods, evidenced by the significant and positive coefficients of political connections variables for both periods. Table 16 also tests the difference of the adverse audit-fee effects of political connections between pre- and post-financial crisis of 2008. The 30 difference between pre- and post-financial crisis is not statistically significant. These results do not support the view that auditors’ perceived audit risk for firms with political connections has changed by financial crisis of 2008. [Insert Table 16 around here] 5. Conclusions Political connections significantly affect accounting quality and business risk. They may help connected firms improve financial performance, which would reduce those firms’ incentives to misreport earnings. Moreover, politically connected firms may be less likely to fail, because they have a higher chance of being rescued by the government in case of financial distress. These arguments suggest that auditors should expect lower risks in auditing clients with political connections than those without. However, managers inside politically connected firms may have greater incentives to obscure financial statements in order to hide the politicians’ rent-seeking activities, and/or to protect their obtained political benefits from being revealed to outsiders. Politically connected firms also tend to have more volatile performances, which result from the changes in political landscape from time to time. This perspective predicts a higher risk associated with auditing clients with political connections. Audit risk determines the level of audit effort, and consequently audit fees. We thus examine empirically how clients’ political connections affect audit fees to help resolve the issue of how political connections may affect audit risk. This issue is also important because political connections are common among public firms in the United States. We find that about 11% of U.S. firms have politicians on their boards. Focusing on the period from 2001 to 2009, we find that auditors charged higher fees to clients with political connections. This finding 31 suggests that overall, political connections increase audit risk and consequently audit fees in the U.S. We attribute this result to two major factors. First, one of the important benefits brought by political connections, namely a greater likelihood of being bailed out by the government, is relatively small in the U.S. Second, the highly litigious environment facing auditors in the U.S. motivates them to exert more effort in auditing politically connected clients, which are more visible and risky. These institutional factors enable a positive effect of political connections on audit fees, and diminish any negative effect. We also find that the effect of political connections on audit risk and audit fees is constrained by internal and external monitoring mechanisms and more pronounced among firms with more complicated operations. Moreover, the political connection effect on audit fees has grown stronger since the passage of SOX which increases the litigation risk facing auditors. We conduct a battery of additional tests to address the endogeneity issue, and evaluate other explanations. Overall, our findings are robust to these additional tests, and to the inclusion of other measures for political connections and alternative samples. Finally, we show that auditors are more likely to issue modified opinions to politically connected firms. To the extent that auditors tend to set a lower threshold for issuing modified opinions to risky clients (Reynolds and Francis, 2001), this finding helps confirm that auditors perceive politically connected clients to be riskier than non-connected firms. 32 References Acemoglu, D., Johnson, S., Kermani, A., Kwak, J., Mitton, T., 2010. The Value of Political Connections in the United States. Working Paper. Aggarwal, R.K., Meschke, F., Wang, T.Y., 2012. Corporate Political Donations: Investment or Agency? Working Paper. Agrawal, A., Knoeber, C.R., 2001. Do Some Outside Directors Play a Political Role? Journal of Law and Economics 44, 179-198. Altman, E.I., 1968. Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy. The Journal of Finance 23, 589-609. Beasley, M.S., 1996. An Empirical Analysis of the Relation between the Board of Director Composition and Financial Statement Fraud. Accounting Review 71, 443-465. Betrand, M., Kramarz, F., Schoar, A., Thesmar, D., 2005. Politically Connected CEOs and Corporate Outcomes: Evidence from France. Working Paper. Bushman, R., Piotroski, J., 2006. Financial Reporting Incentives for Conservative Accounting: The Influence of Legal and Political Institutions. Journal of Accounting & Economics 42, 107-148. Bushman, R.M., Piotroski, J.D., Smith, A.J., 2004. What Determines Corporate Transparency? Journal of Accounting Research 42, 207-252. Center for Political Accountability (CPA). 2005. The Green Canary: Alerting Shareholders and Protecting Their Investments. Available at www.politicalaccountability.net. Center for Political Accountability (CPA). 2006. Hidden Rivers: How Trade Associations Conceal Corporate Political Spending, Its Threat to Companies, and What Shareholders Can Do. Available at www.politicalaccountability.net. Chaney, P.K., Faccio, M., Parsley, D., 2011. The Quality of Accounting Information in Politically Connected Firms. Journal of Accounting & Economics 51, 58-76. Chen, C., Ding Y., Kim, C.F., 2010. High-Level Politically Connected Firms, Corruption, and Analyst Forecast Accuracy Around the World, Journal of International Business Studies 41 (9), 1505–1524. Choi, J.H., Kim, J.B., Liu, X.H., Simunic, D.A., 2008. Audit Pricing, Legal Liability Regimes, and Big 4 Premiums: Theory and Cross-Country Evidence. Contemporary Accounting Research 25, 55-99. Choi, J.H., Kim, J.B., Liu, X.H., Simunic, D.A., 2009. Cross-Listing Audit Fee Premiums: Theory and Evidence. Accounting Review 84, 1429-1463. Cooper, M.J., Gulen, H., Ovtchinnikov, A.V., 2010. Corporate Political Contributions and Stock Returns. Journal of Finance 65, 687-724. Craswell, A.T., Francis, J.R., Taylor, S.L., 1995. Auditor Brand Name Reputations and Industry Specializations. Journal of Accounting & Economics 20, 297-322. Deloitte, 2012. Political Contribution Disclosures and Oversight, Hot Topics. available at http://www.corpgov.deloitte.com/binary/com.epicentric.contentmanagement.serv let.ContentDeliveryServlet/USEng/Documents/Deloitte%20Periodicals/Hot%20 Topics/Political%20Contributions_Deloitte%20Hot%20Topics_April%202012.p df DeZoort, F.T., Hermanson, D.R., Houston, R.W., 2008. Audit Committee Member Support for Proposed Audit Adjustments: Pre-Sox Versus Post-Sox Judgments. Auditing-a Journal of Practice & Theory 27, 85-104. Duchin, R., Sosyura, D., 2012. The Politics of Government Investment. Journal of 33 Financial Economics, Forthcoming. Dye, R.A., 1993. Auditing Standards, Legal Liability, and Auditor Wealth. Journal of Political Economy 101, 887-914. Faccio, M., Masulis, R.W., McConnell, J.J., 2006. Political Connections and Corporate Bailouts. Journal of Finance 61, 2597-2635. Fan, J.P.H., Wong, T.J., Zhang, T.Y., 2007. Politically Connected CEOs, Corporate Governance, and Post-IPO Performance of China's Newly Partially Privatized Firms. Journal of Financial Economics 84, 330-357. Ferguson, A., Francis, J.R., Stokes, D. 2003. The Effects of Firm-Wide and Office-Level Industry Expertise on Audit Pricing. The Accounting Review 78(2), 429-448. Fisman, D., Fisman, R., Galef, J., Khurana, R., 2006. Estimating the Value of Connections to Vice-President Cheney. Working Paper. Fisman, R., 2001. Estimating the Value of Political Connections. American Economic Review 91, 1095-1102. Francis, J.R., Reichelt, K., Wang, D., 2005. The Pricing of National and City-Specific Reputations for Industry Expertise in the Us Audit Market. Accounting Review 80, 113-136. Francis, J.R., Stokes, D.J., 1986. Audit Prices, Product Differentiation, and Scale Economies - Further Evidence from the Australian Market. Journal of Accounting Research 24, 383-393. Francis, J.R., Wang, D, 2008. The Joint Effect of Investor Protection and Big 4 Audits on Earnings Quality around the World. Contemporary Accounting Research 25(1), 157-191. Goldman, E., Rocholl, J., So, J., 2009. Do Politically Connected Boards Affect Firm Value ? Review of Financial Studies 22, 2331-2360. Goldman, E., So, J., Rocholl, J., 2012. Political Connections and the Allocation of Procurement Contracts. Working Paper. Guedhami, O., Pittman, J., Saffar, W., 2011. Auditor Choice in Politically Connected Firms. Working Paper. Guedhami, O., Pittman, J., Saffar, W., 2009. Auditor Choice in Privatized Firms: Empirical Evidence on the Role of State and Foreign Owners. Journal of Accounting & Economics 48, 151-171. Gul, F.A., 2006. Auditors' Response to Political Connections and Cronyism in Malaysia. Journal of Accounting Research 44, 931-963. Gul, F.A., Goodwin, J., 2010. Short-Term Debt Maturity Structures, Credit Ratings, and the Pricing of Audit Services. Accounting Review 85, 877-909. Gul, F.A., Tsui, J.S.L., 1997. A Test of the Free Cash Flow and Debt Monitoring Hypotheses: Evidence from Audit Pricing. Journal of Accounting & Economics 24, 219-237. Hartzell, J.C., Starks, L.T., 2003. Institutional Investors and Executive Compensation. Journal of Finance 58, 2351-2374. Houston, J.F., Jiang, L., Lin, C., Ma, Y., 2012. Political Connections and the Cost of Borrowing. Working Paper. Huang, H.W., Raghunandan, K., Rama, D., 2009. Audit Fees for Initial Audit Engagements Before and After SOX. Auditing-a Journal of Practice & Theory 28, 171-190. Jensen, M.C., Meckling, W.H., 1976. Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure. Journal of Financial Economics 3, 305-360. 34 Johnson, S., Mitton, T., 2003. Cronyism and Capital Controls: Evidence from Malaysia. Journal of Financial Economics 67, 351-382. Karpoff, J.M., Lee, D.S., Vendrzyk, V.P., 1999. Defense Procurement Fraud, Penalties, and Contractor Influence. Journal of Political Economy 107, 809-842. Kido, N., Petacchi, R., Weber, J., 2012. The Influence of Elections on the Accounting Choices of Governmental Entities. Journal of Accounting Research 50, 443-476. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R.W., 1998. Law and Finance. Journal of Political Economy 106, 1113-1155. Larcker, D.F., Rusticus, T.O., 2010. On the Use of Instrumental Variables in Accounting Research. Journal of Accounting & Economics 49, 186-205. Leuz, C., Oberholzer-Gee, F., 2006. Political Relationships, Global Financing, and Corporate Transparency: Evidence from Indonesia. Journal of Financial Economics 81, 411-439. Piotroski, J., Wong, T.J., Zhang, T.Y., 2011. Political Incentives to Suppress Negative Financial Information: Evidence from Chinese Listed Firms. Working Paper. Ramanna, K., Roychowdhury, S., 2010. Elections and Discretionary Accruals: Evidence from 2004. Journal of Accounting Research 48, 445-475. Simunic, D.A., 1980. The Pricing of Audit Services - Theory and Evidence. Journal of Accounting Research 18, 161-190. Tucker, J.W., 2011. Selection Bias and Econometric Remedies in Accounting and Finance Research. Journal of Accounting Literature Forthcoming. Wang, Q., Wong, T.J., Xia, L.J., 2008. State Ownership, the Institutional Environment, and Auditor Choice: Evidence from China. Journal of Accounting & Economics 46, 112-134. Wingate, M., 1997. An Examination of Cultural Influence on Audit Environment. Research in Accounting Regulation 11, 129-148. Yu, F., Yu, X., 2011. Corporate Lobbying and Fraud Detection. Journal of Financial and Quantitative Analysis 46, 1865-1891. 35 Appendix A: Definitions of Political Connection Measures Variable PC Dummy PC Directors PC Rank PC Tenure PC Freshness Definition Equals 1 for a firm with one or more directors who held political positions before sitting on the board and 0 otherwise. Following Goldman, Rocholl and So (2009), we consider the following political positions: president, presidential candidate, senator, member of the House of Representative, (assistant) secretary, deputy secretary, deputy assistant secretary, undersecretary, associate director, governor, director (CIA, FEMA), deputy director (CIA, OMB), commissioner (IRS< NRC, SSA, CRC, FDA, SEC), representative to the United Nations, ambassador, mayor, staff (White House, president, presidential campaign), chairman of the Party Caucus, chairman or staff of the presidential election campaign, and chairman or member of the president’s committee/council. Total number of board directors who held political positions before sitting on the board. Political power of PC director’s former political position ranging from 1 to 5, the larger values of this variable means the stronger political power; (5 for president, vice president, (vice) presidential candidates, etc.; 4 for secretary of important departments (e.g., General Attorney, Secretaries of State, Treasury, and Defense, White House Executives, SEC commissioners); 3 for governors and other secretaries of departments, and Senators; 2 for House representatives, 1 for Assistant/deputy secretaries of all departments and ambassadors Total number of years of political positions held by all connected directors in a firm For a politically connected firm, the value of this measure is determined based on the following formula: PC Freshness = Ln(45) – Ln(1 + # of elapsed years). # of elapsed years is the number of years between current year and the most recent political position held by any connected directors. The largest value of # of elapsed years within our sample is 44. The value of PC Freshness decreases with the number of elapsed years for connected firms, and set to be 0 for non-connected firms. 36 Table 1: Descriptive Statistics. PC Dummy PC Directors PC Rank PC Tenure PC Freshness AFEE SIZE LEV INVREC ROA LOSS NGS NBS YE OPINION BIG4 NLEADER CLEADER VOLATILITY MB FOREIGN QUICK CURRENT # of Obs. 29,785 3,327 3,327 3,327 3,327 29,785 29,785 29,785 29,785 29,785 29,785 29,785 29,785 29,785 29,785 29,785 29,785 29,785 29,785 29,785 29,785 29,785 29,785 Mean 0.112 1.315 2.466 8.715 1.589 6.241 5.735 0.484 0.253 0.029 0.393 2.275 2.016 0.716 0.040 0.790 0.043 0.600 0.161 2.872 0.001 2.469 0.509 Std Dev 0.315 0.741 1.239 8.140 0.862 1.333 2.055 0.362 0.189 0.257 0.488 2.507 1.640 0.451 0.196 0.407 0.202 0.490 0.107 4.498 0.043 2.912 0.256 P10 0.000 1.000 1.000 1.000 0.533 4.594 3.122 0.155 0.038 -0.245 0.000 0.000 1.000 0.000 0.000 0.000 0.000 0.000 0.063 0.582 0.000 0.586 0.146 37 P25 0.000 1.000 1.000 2.000 0.995 5.216 4.238 0.268 0.098 0.000 0.000 0.000 1.000 0.000 0.000 1.000 0.000 0.000 0.089 1.149 0.000 0.944 0.312 Median 0.000 1.000 3.000 5.000 1.431 6.142 5.619 0.456 0.219 0.097 0.000 2.000 1.000 1.000 0.000 1.000 0.000 1.000 0.133 2.011 0.000 1.516 0.512 P75 0.000 1.000 3.000 14.000 2.219 7.130 7.094 0.632 0.363 0.158 1.000 4.000 3.000 1.000 0.000 1.000 0.000 1.000 0.200 3.537 0.000 2.789 0.715 P90 1.000 2.000 4.000 22.000 2.730 8.039 8.480 0.799 0.527 0.222 1.000 5.000 4.000 1.000 0.000 1.000 0.000 1.000 0.293 6.214 0.044 5.257 0.858 Table 2: Correlation Matrix Panel A: Correlations among PC measures [1] PC Dummy [2] PC Directors [3] PC Rank [4] PC Tenure [5] PC Freshness [1] [2] [3] [4] 0.86 0.88 0.71 0.87 0.83 0.70 0.82 0.76 0.83 0.74 Panel B: Correlations among control variables [1] AFEE [2] PC Dummy [3] SIZE [4] LEV [5] INVREC [6] ROA [7] LOSS [8] NGS [9] NBS [10] YE [11] OPINION [12] BIG4 [13] NLEADER [14] CLEADER [15] VOLATILITY [16] MB [17] FOREIGN [18] QUICK [19] CURRENT [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] 0.20 0.79 0.31 0.00 0.29 -0.24 0.38 0.22 0.07 -0.10 0.36 0.16 0.20 -0.40 0.09 0.32 -0.22 -0.26 0.22 0.11 -0.05 0.06 -0.05 0.06 0.08 0.05 -0.02 0.07 0.07 0.06 -0.09 0.06 0.09 -0.07 -0.11 0.32 -0.09 0.43 -0.36 0.30 0.24 0.07 -0.20 0.38 0.15 0.25 -0.44 0.05 0.30 -0.27 -0.45 0.10 0.08 0.01 0.03 0.13 0.08 0.13 0.06 0.04 0.13 -0.08 -0.07 0.03 -0.68 -0.35 0.18 -0.17 0.20 0.12 -0.18 -0.03 -0.08 -0.04 -0.01 -0.05 -0.10 0.11 -0.20 0.35 -0.70 0.12 0.13 -0.03 -0.23 0.12 0.04 0.10 -0.40 0.20 0.26 -0.19 -0.28 -0.11 -0.13 0.05 0.23 -0.10 -0.04 -0.09 0.44 -0.16 -0.30 0.09 0.16 0.19 -0.02 -0.09 0.17 0.08 0.06 -0.12 0.04 0.34 0.01 0.05 -0.02 -0.05 0.05 0.03 0.08 -0.15 -0.06 0.08 -0.15 -0.16 0.02 0.05 0.02 0.02 0.06 0.02 -0.02 0.00 -0.11 -0.09 -0.02 -0.04 0.20 -0.08 -0.09 -0.14 -0.01 0.08 0.20 -0.15 0.05 0.11 0.01 -0.06 0.17 -0.05 0.01 0.08 -0.03 -0.05 -0.13 0.00 0.05 -0.10 -0.11 -0.10 -0.23 0.16 0.25 0.11 0.14 0.13 -0.01 -0.03 0.54 38 Table 3: Political Connections and Audit Fees PCON + SIZE + LEV + INVREC + ROA - LOSS + NGS + NBS + YE + OPINION + BIG4 + NLEADER + CLEADER + VOLATILITY + MB + FOREIGN + QUICK - CURRENT + Constant Industry effects Year effects Observations R-squared (1) PC Dummy (2) PC Directors (3) PC Rank (4) PC Tenure (5) PC Freshness 0.114*** [6.41] 0.499*** [98.30] 0.167*** [6.75] 0.170*** [3.20] -0.250*** [-9.26] 0.118*** [10.08] 0.040*** [12.39] 0.040*** [10.25] 0.128*** [8.33] 0.227*** [9.84] 0.230*** [15.06] 0.304*** [11.60] 0.031*** [2.61] 0.203*** [4.20] 0.002** [2.32] 0.814*** [6.45] -0.042*** [-15.10] 0.482*** [11.63] 1.904*** [25.82] 0.087*** [6.89] 0.498*** [97.68] 0.167*** [6.75] 0.167*** [3.14] -0.248*** [-9.18] 0.118*** [10.07] 0.040*** [12.36] 0.039*** [10.10] 0.128*** [8.38] 0.227*** [9.84] 0.231*** [15.09] 0.305*** [11.66] 0.030*** [2.58] 0.204*** [4.22] 0.002** [2.29] 0.800*** [6.38] -0.042*** [-15.10] 0.481*** [11.61] 1.914*** [25.90] 0.042*** [6.34] 0.499*** [98.25] 0.167*** [6.75] 0.170*** [3.21] -0.251*** [-9.33] 0.117*** [10.05] 0.040*** [12.40] 0.040*** [10.30] 0.128*** [8.36] 0.227*** [9.84] 0.230*** [15.05] 0.303*** [11.60] 0.031*** [2.62] 0.203*** [4.19] 0.002** [2.32] 0.807*** [6.39] -0.042*** [-15.07] 0.482*** [11.63] 1.904*** [25.90] 0.007*** [4.83] 0.501*** [98.80] 0.168*** [6.74] 0.170*** [3.20] -0.254*** [-9.43] 0.118*** [10.09] 0.040*** [12.36] 0.041*** [10.38] 0.128*** [8.36] 0.229*** [9.88] 0.229*** [14.95] 0.306*** [11.64] 0.030** [2.53] 0.202*** [4.17] 0.003** [2.49] 0.819*** [6.50] -0.042*** [-15.00] 0.482*** [11.59] 1.898*** [25.66] 0.064*** [6.76] 0.499*** [98.34] 0.167*** [6.75] 0.169*** [3.18] -0.250*** [-9.29] 0.118*** [10.12] 0.040*** [12.38] 0.040*** [10.28] 0.128*** [8.34] 0.227*** [9.83] 0.231*** [15.13] 0.304*** [11.60] 0.030** [2.56] 0.202*** [4.17] 0.002** [2.34] 0.818*** [6.48] -0.042*** [-15.03] 0.481*** [11.60] 1.905*** [25.86] Yes Yes 29,785 0.82 Yes Yes 29,785 0.82 Yes Yes 29,785 0.82 Yes Yes 29,785 0.82 Yes Yes 29,785 0.82 All continuous variables are winsored at 1% and 99% levels. All t-statistics reported in the brackets are based on standard errors clustered at the firm level. ***, **, and * denote two-tailed significance levels of 0.01, 0.05, and 0.10, respectively. 39 Table 4: Corporate Governance, Political Connections, and Audit Fees Panel A: Director Shareholdings PCON + GOV ? GOV*PCON - Control variables Industry effects Year effects Observations R-squared (1) PC Dummy (2) PC Directors (3) PC Rank (4) PC Tenure (5) PC Freshness 0.147*** [3.94] -0.006* [-1.75] -0.017*** [-2.58] 0.103*** [5.11] -0.006* [-1.86] -0.010** [-2.56] 0.052*** [3.86] -0.006** [-1.98] -0.005** [-2.23] 0.008*** [2.73] -0.008** [-2.45] -0.001 [-1.55] 0.083*** [4.06] -0.006** [-1.97] -0.008** [-2.37] Yes Yes Yes 11,940 0.78 Yes Yes Yes 11,940 0.78 Yes Yes Yes 11,940 0.78 Yes Yes Yes 11,940 0.78 Yes Yes Yes 11,940 0.78 (1) PC Dummy (2) PC Directors (3) PC Rank (4) PC Tenure (5) PC Freshness 0.182*** [5.91] 0.013*** [5.18] -0.013** [-2.53] 0.115*** [4.72] 0.012*** [5.04] -0.005 [-1.43] 0.061*** [5.26] 0.013*** [5.13] -0.004* [-1.94] 0.010*** [3.65] 0.012*** [5.05] -0.001 [-1.36] 0.096*** [5.56] 0.013*** [5.11] -0.006** [-2.15] Yes Yes Yes 29,775 0.82 Yes Yes Yes 29,775 0.82 Yes Yes Yes 29,775 0.82 Yes Yes Yes 29,775 0.82 Yes Yes Yes 29,775 0.82 Panel B: Institutional Shareholdings PCON + GOV ? GOV*PCON - Control variables Industry effects Year effects Observations R-squared All continuous variables are winsored at 1% and 99% levels. All t-statistics reported in the brackets are based on standard errors clustered at the firm level. ***, **, and * denote two-tailed significance levels of 0.01, 0.05, and 0.10, respectively. Table 5: Operation Complexity, Political Connections, and Audit Fees Panel A: Product Line Diversification PCON + COMP + COMP *PCON + Control variables Industry effects Year effects Observations R-squared (1) PC Dummy (2) PC Directors (3) PC Rank (4) PC Tenure (5) PC Freshness 0.056* [1.90] 0.111*** [6.63] 0.114*** [2.95] 0.052** [2.08] 0.114*** [6.85] 0.060** [1.96] 0.015 [1.36] 0.113*** [6.80] 0.043*** [3.03] 0.003 [1.37] 0.118*** [7.17] 0.007** [2.14] 0.031** [2.10] 0.113*** [6.84] 0.063*** [3.12] Yes Yes Yes 25,826 0.82 Yes Yes Yes 25,826 0.82 Yes Yes Yes 25,826 0.82 Yes Yes Yes 25,826 0.82 Yes Yes Yes 25,826 0.82 (1) PC Dummy (2) PC Directors (3) PC Rank (4) PC Tenure (5) PC Freshness 0.063** [2.22] 0.231*** [12.07] 0.100** [2.57] 0.054*** [2.61] 0.233*** [12.33] 0.063** [2.46] 0.024** [2.28] 0.234*** [12.35] 0.028** [1.98] 0.003 [1.45] 0.235*** [12.53] 0.006** [2.21] 0.031** [2.07] 0.230*** [12.20] 0.065*** [3.24] Yes Yes Yes 23,675 0.82 Yes Yes Yes 23,675 0.83 Yes Yes Yes 23,675 0.82 Yes Yes Yes 23,675 0.82 Yes Yes Yes 23,675 0.82 Panel B: Geographic Diversification PCON + COMP + COMP *PCON + Control variables Industry effects Year effects Observations R-squared All continuous variables are winsored at 1% and 99% levels. All t-statistics reported in the brackets are based on standard errors clustered at the firm level. ***, **, and * denote two-tailed significance levels of 0.01, 0.05, and 0.10, respectively. 41 Table 6: SOX, Political Connections, and Audit Fees PCON + SOX + SOX*PCON + Control variables Industry effects Observations R-squared (1) PC Dummy (2) PC Directors (3) PC Rank (4) PC Tenure (5) PC Freshness 0.058** [2.05] 0.942*** [75.78] 0.106*** [3.59] 0.067*** [3.82] 0.950*** [77.40] 0.033* [1.69] 0.024** [2.39] 0.946*** [77.14] 0.032*** [2.95] 0.004 [1.58] 0.947*** [78.46] 0.005** [2.00] 0.043*** [2.92] 0.947*** [77.44] 0.047*** [2.92] Yes Yes 22,452 0.82 Yes Yes 22,452 0.82 Yes Yes 22,452 0.82 Yes Yes 22,452 0.81 Yes Yes 22,452 0.81 All continuous variables are winsored at 1% and 99% levels. All t-statistics reported in the brackets are based on standard errors clustered at the firm level. ***, **, and * denote two-tailed significance levels of 0.01, 0.05, and 0.10, respectively. 42 Table 7: Political Connections and Audit Fees (Instrumental Variables Using 2SLS) Panel A: The 2nd Stage Regression: Political Connections and Audit Fees PCON + Control variables Industry effects Year effects Observations Over-identification test (p-value) 1st-stage F-test (p-value) Hausman test R-squared (1) PC Dummy (2) PC Directors 0.902*** [3.70] Yes Yes Yes 29,785 0.3533 0.0000 0.0000 0.79 (3) PC Rank (4) PC Tenure (5) PC Freshness 0.655*** [3.38] 0.361*** [3.56] 0.095*** [3.25] 0.510*** [3.66] Yes Yes Yes 29,785 0.7080 0.0000 0.0000 0.79 Yes Yes Yes 29,785 0.3323 0.0000 0.0000 0.79 Yes Yes Yes 29,785 0.2351 0.0000 0.0000 0.74 Yes Yes Yes 29,785 0.2612 0.0000 0.0000 0.77 Panel B: The 1st Stage Model: The Probability of Establishing Political Connections DISTANCE - PROBPC + SIZE + LEV ? INVREC ? ROA ? LOSS ? NGS + NBS + YE ? OPINION ? BIG4 ? FOREIGN ? QUICK ? (1) PC Dummy (2) PC Directors (3) PC Rank (4) PC Tenure (5) PC Freshness -0.000*** [-2.75] 0.846*** [7.84] 0.048*** [14.48] 0.010 [1.17] -0.033 [-1.33] -0.090*** [-5.16] 0.002 [0.36] -0.001 [-0.65] 0.011*** [3.52] 0.007 [0.83] 0.027** [2.24] -0.016** [-2.05] 0.091 [0.94] -0.000 [-0.35] -0.000*** [-3.21] 1.181*** [6.92] 0.075*** [11.71] 0.005 [0.48] -0.045 [-1.22] -0.136*** [-5.57] 0.008 [0.78] -0.001 [-0.12] 0.023*** [3.38] 0.002 [0.17] 0.038** [2.31] -0.030*** [-2.79] 0.294* [1.66] -0.002 [-0.95] -0.000** [-2.29] 2.372*** [7.64] 0.139*** [13.11] 0.018 [0.88] -0.115* [-1.65] -0.206*** [-4.35] 0.022 [1.29] -0.005 [-0.92] 0.028*** [2.79] 0.007 [0.27] 0.086*** [2.64] -0.047** [-2.27] 0.445 [1.54] -0.003 [-1.05] -0.001* [-1.88] 8.063*** [4.99] 0.472*** [10.55] 0.014 [0.15] -0.710** [-2.52] -0.619*** [-2.89] 0.066 [0.85] -0.021 [-0.83] 0.098** [2.26] 0.087 [0.79] 0.243* [1.75] -0.181** [-1.99] 1.202 [0.92] -0.032** [-2.25] -0.000** [-2.44] 1.300*** [7.03] 0.084*** [13.28] 0.005 [0.36] -0.082* [-1.88] -0.134*** [-3.93] -0.000 [-0.04] -0.001 [-0.27] 0.019*** [3.20] 0.011 [0.73] 0.057** [2.57] -0.046*** [-3.32] 0.071 [0.38] -0.003 [-1.49] 43 Table 7 (Cont’d): Political Connections and Audit Fees (Instrumental Variables Using 2SLS) CURRENT ? NLEADER ? CLEADER ? VOLATILITY ? MB ? Constant Industry effects Year effects Observations R-squared 0.059*** [2.69] 0.021 [1.05] -0.003 [-0.51] 0.016 [0.73] 0.003*** [4.02] -0.282*** [-10.12] 0.089*** [2.62] 0.028 [0.79] -0.003 [-0.31] 0.025 [0.82] 0.004*** [3.78] -0.442*** [-8.82] 0.176*** [2.71] 0.095 [1.55] -0.011 [-0.60] 0.094 [1.53] 0.007*** [3.82] -0.845*** [-10.42] 0.871*** [2.97] 0.262 [0.93] 0.068 [0.79] 0.291 [1.07] 0.016** [2.01] -2.857*** [-7.65] 0.114*** [2.96] 0.050 [1.25] 0.002 [0.16] 0.078* [1.90] 0.004*** [3.64] -0.480*** [-9.61] Yes Yes Yes Yes Yes Yes 27,393 0.10 Yes 27,393 0.10 Yes 27,393 0.11 Yes 26,356 0.06 Yes 26,356 0.04 All continuous variables are winsored at 1% and 99% levels. All t-statistics reported in the brackets are based on standard errors clustered at the firm level. ***, **, and * denote two-tailed significance levels of 0.01, 0.05, and 0.10, respectively. 44 Table 8: Political Connections and Audit Fees (Heckman IMR method) Panel A: The 2nd Stage Regression: Political Connections and Audit Fees PC Dummy PCON + MILLS ? 0.100*** [5.70] -0.135*** [-3.08] Variance-inflation-factors (VIFs) PCON MILLS 1.11 16.54 Control variables Industry effects Year effects Observations R-squared Yes Yes Yes 27,393 0.83 Panel B: The 1st Stage Model: The Probability of Establishing Political Connections PC Dummy DISTANCE - PROBPC + SIZE + LEV ? INVREC ? ROA ? LOSS ? NGS + NBS + YE ? OPINION ? BIG4 ? FOREIGN ? QUICK ? 45 -0.001*** [-5.89] 4.412*** [8.72] 0.263*** [32.04] 0.076** [2.51] -0.164 [-1.82] -0.475*** [-7.87] 0.014 [0.47] -0.017*** [-3.51] 0.045*** [7.08] 0.072*** [2.75] 0.169*** [2.72] -0.039 [-1.24] 0.068 [0.26] -0.003 [-0.51] Table 8 (Cont’d): Political Connections and Audit Fees (Heckman IMR method) CURRENT ? NLEADER ? CLEADER ? VOLATILITY ? MB ? Constant Industry effects Year effects Observations R-squared 0.259*** [3.80] 0.012 [0.25] -0.023 [-0.98] -0.059 [-0.43] 0.013*** [5.50] -3.698*** [-23.67] Yes Yes 27,393 0.13 All continuous variables are winsored at 1% and 99% levels. All t-statistics reported in the brackets are based on standard errors clustered at the firm level. ***, **, and * denote two-tailed significance levels of 0.01, 0.05, and 0.10, respectively. 46 Table 9: Political Connections and Audit Fees (Propensity Score Matching) Panel A AFEE (mean value) PCON=1 Near Neighbor Kernel-based Kernel-based (# of obs. 3212) (n=1) (Gaussian) (Epanechnikov) 7.03 6.93 6.76 6.91 3.95*** 16.12*** 7.31*** t-stat of differences Panel B Near Neighbor (n=1) Whole sample 0.089*** 0.078*** [4.41] [4.33] Control PSCORE Portfolio Dummies No Yes Control variables Yes Yes Industry effects Yes Yes Year effects Yes Yes Observations 5,868 27,393 R-squared 0.87 0.83 PCON + Panel A presents the differences of audit fees between PC firms and matched outcomes calculated by nearest-neighbor and Kernel-matching methods respectively. Panel B presents the results of OLS regressions on audit fees. We construct 20 portfolio based on propensity matching score (PSCORE). Then we assign dummy variables to indicate each portfolio and include them into regression to control for the effects of PC probability. All continuous variables are winsored at 1% and 99% levels. All t-statistics reported in the brackets are based on standard errors clustered at the firm level. ***, **, and * denote two-tailed significance levels of 0.01, 0.05, and 0.10, respectively. 47 Table 10: Elections, Political Connections, and Audit Fees PCON + ELECTION + ELECTION*PCON + Control variables Industry effects Year effects Observations R-squared (1) PC Dummy (2) PC Directors (3) PC Rank (4) (5) PC Tenure PC Freshness 0.077*** [4.23] 0.031*** [4.31] 0.045*** [3.83] 0.061*** [5.02] 0.032*** [4.42] 0.029*** [3.97] 0.028*** [4.22] 0.031*** [4.34] 0.017*** [4.05] 0.005*** [3.47] 0.033*** [4.65] 0.002** [2.26] 0.043*** [4.49] 0.031*** [4.38] 0.025*** [3.70] Yes Yes Yes 29,659 0.82 Yes Yes Yes 29,659 0.82 Yes Yes Yes 29,659 0.82 Yes Yes Yes 29,659 0.82 Yes Yes Yes 29,659 0.82 All continuous variables are winsored at 1% and 99% levels. All t-statistics reported in the brackets are based on standard errors clustered at the firm level. ***, **, and * denote two-tailed significance levels of 0.01, 0.05, and 0.10, respectively. 48 Table 11: Connected Directors on Audit Committee and Audit Fees (1) PC Dummy (2) PC Directors PC Rank (5) PC PC Tenure Freshness 0.114*** [3.67] 0.114*** [5.70] 0.081*** [4.74] 0.093*** [5.77] 0.038*** [3.80] 0.045*** [5.73] 0.005** [2.49] 0.008*** [4.79] 0.064*** [4.04] 0.064*** [5.87] PCON_AUD = PCON_NUAD (p-value) 0.986 0.566 0.538 0.170 0.987 Control variables Industry effects Year effects Observations R-squared Yes Yes Yes 29,785 0.82 Yes Yes Yes 29,785 0.82 Yes Yes Yes 29,785 0.82 Yes Yes Yes 29,785 0.82 Yes Yes Yes 29,785 0.82 PCON_AUD + PCON_NAUD + (3) (4) PCON_AUD equals 1 for firms with at least one political director sitting on audit committee and 0 otherwise. PCON_NAUD equals 1 for those firms which are politically connected but do not have a political director on audit committee and 0 otherwise. All continuous variables are winsored at 1% and 99% levels. All t-statistics reported in the brackets are based on standard errors clustered at the firm level. ***, **, and * denote two-tailed significance levels of 0.01, 0.05, and 0.10, respectively. 49 Table 12: Political Contributions, Lobbying Expenditure, and Audit Fees (1) PCON + CONTIBUTION + LOBBYING + Control variables Industry effects Year effects Observations R-squared 0.008*** [2.84] Yes Yes Yes 29,785 0.82 (2) (3) PC Dummy (4) PC Directors (5) PC Rank (6) PC Tenure (7) PC Freshness 0.010*** [6.88] 0.097*** [5.42] -0.000 [-0.13] 0.009*** [6.22] 0.075*** [6.02] -0.001 [-0.31] 0.009*** [6.15] 0.035*** [5.35] -0.000 [-0.09] 0.009*** [6.16] 0.005*** [3.93] 0.000 [0.07] 0.009*** [6.44] 0.055*** [5.80] -0.001 [-0.17] 0.009*** [6.22] Yes Yes Yes 29,785 0.82 Yes Yes Yes 29,785 0.82 Yes Yes Yes 29,785 0.82 Yes Yes Yes 29,785 0.82 Yes Yes Yes 29,785 0.82 Yes Yes Yes 29,785 0.82 Continuous variables are winsored at 1% and 99% levels. All t-statistics reported in the brackets are based on standard errors clustered at the firm level. ***, **, and * denote two-tailed significance levels of 0.01, 0.05, and 0.10, respectively. 50 Table 13: Political Connections and Audit Fees for S&P 500 Firms PCON Control variables Industry effects Year effects Observations R-squared + (1) PC Dummy (2) PC Directors (3) (4) (5) PC Rank PC Tenure PC Freshness 0.082** [2.01] 0.057*** [3.19] 0.020 [1.54] 0.004* [1.81] 0.045** [2.49] Yes Yes Yes 2,724 0.78 Yes Yes Yes 2,724 0.78 Yes Yes Yes 2,724 0.78 Yes Yes Yes 2,724 0.78 Yes Yes Yes 2,724 0.78 All the continuous variables are winsored at 1% and 99% levels. All t-statistics reported in the brackets are based on standard errors clustered at the firm level. ***, **, and * denote two-tailed significance levels of 0.01, 0.05, and 0.10, respectively. 51 Table 14: Political Connections and Audit Fees for Financially Distressed Firms PCON + LOSS + PCON*LOSS ? Control variables Industry effects Year effects Observations R-squared (1) PC Dummy (2) PC Directors (3) PC Rank (4) PC Tenure (5) PC Freshness 0.058** [2.32] 0.093*** [6.34] -0.011 [-0.32] 0.048*** [3.15] 0.092*** [6.37] -0.005 [-0.22] 0.015 [1.63] 0.091*** [6.28] 0.002 [0.17] 0.003 [1.64] 0.092*** [6.42] -0.000 [-0.09] 0.025* [1.88] 0.093*** [6.44] -0.007 [-0.39] Yes Yes Yes 29,785 0.73 Yes Yes Yes 29,785 0.73 Yes Yes Yes 29,785 0.73 Yes Yes Yes 29,785 0.73 Yes Yes Yes 29,785 0.73 All the continuous variables are winsored at 1% and 99% levels. All t-statistics reported in the brackets are based on standard errors clustered at the firm level. ***, **, and * denote two-tailed significance levels of 0.01, 0.05, and 0.10, respectively. 52 Table 15: Political Connections and Audit Opinions Panel A: Whole sample PCON + SIZE - DACC + LEV + ROA - LOSS + BIG4 - LIT_SHU + OFFICESIZE + INFLUENCE + NLEADER + CLEADER + TENURE + REPORTLAG + MB + VOLATILITY + PRIOGC + Constant Industry effects Year effects Observations Pseudo R2 (1) (2) PC Dummy PC Directors (3) PC Rank (4) (5) PC Tenure PC Freshness 0.239*** [3.22] -0.247*** [-10.14] 0.008 [0.73] 0.214* [1.93] -1.107*** [-12.33] 0.858*** [11.57] -0.128* [-1.80] 0.222*** [8.27] 0.050** [2.30] 0.026 [0.33] -0.198 [-1.28] 0.044 [0.89] 0.006 [0.11] 0.004*** [6.53] -0.000 [-0.41] 0.865*** [5.25] 1.745*** [19.93] -0.923** [-2.03] 0.112** [2.55] -0.245*** [-10.07] 0.008 [0.72] 0.216* [1.94] -1.110*** [-12.40] 0.859*** [11.62] -0.129* [-1.81] 0.222*** [8.28] 0.051** [2.36] 0.029 [0.36] -0.201 [-1.31] 0.042 [0.86] 0.008 [0.15] 0.004*** [6.53] -0.000 [-0.39] 0.865*** [5.25] 1.745*** [19.97] -0.938** [-2.06] 0.085*** [3.23] -0.247*** [-10.13] 0.008 [0.71] 0.215* [1.94] -1.110*** [-12.36] 0.857*** [11.56] -0.128* [-1.80] 0.222*** [8.26] 0.050** [2.32] 0.026 [0.33] -0.200 [-1.29] 0.043 [0.87] 0.007 [0.13] 0.004*** [6.53] -0.000 [-0.40] 0.868*** [5.26] 1.743*** [19.94] -0.930** [-2.05] 0.015*** [2.81] -0.245*** [-10.11] 0.008 [0.75] 0.214* [1.93] -1.111*** [-12.42] 0.857*** [11.59] -0.128* [-1.80] 0.223*** [8.29] 0.050** [2.33] 0.027 [0.34] -0.196 [-1.28] 0.041 [0.83] 0.009 [0.17] 0.004*** [6.53] -0.000 [-0.39] 0.866*** [5.25] 1.746*** [19.99] -0.929** [-2.04] 0.122*** [2.97] -0.247*** [-10.13] 0.008 [0.72] 0.215* [1.94] -1.105*** [-12.29] 0.860*** [11.57] -0.124* [-1.74] 0.222*** [8.27] 0.050** [2.30] 0.024 [0.30] -0.196 [-1.28] 0.042 [0.85] 0.006 [0.12] 0.004*** [6.53] -0.000 [-0.39] 0.864*** [5.23] 1.743*** [19.94] -0.917** [-2.02] Yes Yes 22,387 0.48 Yes Yes 22,387 0.48 Yes Yes 22,387 0.48 Yes Yes 22,387 0.48 Yes Yes 22,387 0.48 53 Table 15 (Cont’d): Political Connections and Audit Opinions Panel B: Financially distressed sub-sample PCON Control variables Industry effects Year effects Observations Pseudo R2 + (1) (2) PC Dummy PC Directors (3) PC Rank (4) (5) PC Tenure PC Freshness 0.270*** [3.46] 0.130*** [2.63] 0.094*** [3.28] 0.016*** [2.83] 0.140*** [3.18] Yes Yes Yes 8,528 0.37 Yes Yes Yes 8,528 0.37 Yes Yes Yes 8,528 0.37 Yes Yes Yes 8,528 0.37 Yes Yes Yes 8,528 0.37 This table reports the results of the Probit model with an indicator variable for going-concern opinions as the dependent variable. Among independent variables, LIT_SHU is firm-specific litigation risk based on Shu (2000), OFFICESIZE is the log value of audit fee revenues for the city-based practice office that provides audit services for a sample firm, INFLUENCE is the percentage of total fees (including both audit and non-audit fees) charged to a sample firm relative to the total revenue for the city-based practice office that provides audit services for the firm, TENURE equals 1 if auditor tenure is 3 years or shorter and 0 otherwise, REPORTLAG equals the number of days between the fiscal year-end and the earnings announcement date, and VOLATILITY is the standard deviation of monthly stock returns for twelve months of the current fiscal year. PRIORGC equals 1 if the company got Going-Concern audit opinion in last year and 0 otherwise. DACC is the unsigned discretionary accruals estimated from modified cross-sectional Jone’s model. All other variables are defined in Section 3.2. All the continuous variables are winsored at 1% and 99% levels. Industry and year fixed effects are controlled. All statistics are adjusted by clustering within firms, that is, standard errors are estimated by maximum likelihood under the assumption that residuals are clustered within firms. 54 Table 16: Political Connections and Audit Fees (Before and After Financial Crisis) PC Dummy One year before Oct, 2008 PCON Difference (p-Value) Control variables Industry effects Observations R-squared + 0.062** [2.30] One year after Oct, 2008 0.072*** [2.79] 0.010 (0.25) Yes Yes 3,288 0.81 Yes Yes 3,020 0.83 All continuous variables are winsored at 1% and 99% levels. All t-statistics reported in the brackets are based on standard errors clustered at the firm level. ***, **, and * denote two-tailed significance levels of 0.01, 0.05, and 0.10, respectively. 55 Table 17: Political Connections and Audit Fees (Heckman IMR method) PC firms SIZE LEV INVREC ROA LOSS NGS NBS YE OPINION BIG4 NLEADER CLEADER VOLATILITY MB FOREIGN QUICK CURRENT Mills Constant Industry effects Year effects Observations R-squared Actual Fee-E(Alt. Fee) + + + + + + + + + + + + + + + ? Non-PC Firms Coeff. t-stat Coeff. t-stat 0.520*** 0.251*** 0.531*** -0.330*** 0.107*** 0.039*** 0.029*** 0.189*** 0.098 0.125*** 0.326*** 0.099*** 0.345** 0.002 1.228*** -0.035*** 0.440*** -0.015 5.204*** 20.44 3.69 4.42 -4.18 3.67 5.02 3.38 4.66 1.47 3.14 5.66 3.49 2.26 1.02 4.13 -4.54 5.08 -0.14 22.47 0.453*** 0.131*** 0.124** -0.165*** 0.106*** 0.048*** 0.034*** 0.127*** 0.205*** 0.265*** 0.254*** 0.035*** 0.213*** 0.000 0.730*** -0.043*** 0.458*** -0.159*** 5.503*** 39.03 5.57 2.13 -4.98 8.43 13.34 6.83 7.73 7.85 16.49 8.58 2.77 4.12 0.30 4.96 -14.29 9.81 -3.23 47.45 Yes Yes 3,212 0.89 Yes Yes 3,212 0.89 Yes Yes 24,181 0.80 Yes Yes 24,181 0.80 0.104*** 11.73 -0.070*** -18.84 This table presents the differences between audit fees actually paid and assumedly paid. We estimated coefficients separately for PC and non-PC auditees, including Mill’s Ratio from stage one equation. E(Alt. Fee) takes the value of predicted audit fees paid if firms belong to alternative (PC/non-PC) group. 56