Auditor Choice in Politically Connected Firms

Authors


  • Accepted by Philip Berger. We thank Najah Attig, Narjess Boubakri, Jean-Claude Cosset, Sadok El Ghoul, Mara Faccio, Clive Lennox, Pete Lisowsky, Stefan Sundgren, and especially an anonymous referee for their insights on an earlier version of this paper. Our paper has also benefited from comments from participants at the 2012 European Accounting Association Conference, the 2011 Global Finance Conference, and seminars at various universities. We appreciate financial support from Canada's Social Sciences and Humanities Research Council as well as excellent research assistance from Nabhomani Aggarwal, David Godsell, and Zeina Mehdi. Omrane Guedhami and Jeffrey Pittman gratefully acknowledge funding from the Center for International Business Education and Research at the University of South Carolina and the CMA Professorship/Chair in Corporate Governance and Transparency at Memorial University, respectively.

ABSTRACT

We extend recent research on the links between political connections and financial reporting by examining the role of auditor choice. Our evidence that public firms with political connections are more likely to appoint a Big 4 auditor supports the intuition that insiders in these firms are eager to improve accounting transparency to convince outside investors that they refrain from exploiting their connections to divert corporate resources. In evidence consistent with another prediction, we find that this link is stronger for connected firms with ownership structures conducive to insiders seizing private benefits at the expense of minority investors. We also find that the relation between political connections and auditor choice is stronger for firms operating in countries with relatively poor institutional infrastructure, implying that tough external monitoring by Big 4 auditors becomes more valuable for preventing diversion in these situations. Finally, we report that connected firms with Big 4 auditors exhibit less earnings management and enjoy greater transparency, higher valuations, and cheaper equity financing.

1. Introduction

In response to calls for research on this issue (e.g., Wang, Wong, and Xia [2008]), we estimate the importance of corporate insiders’ political connections to auditor choice.1 Our analysis contributes to extant research by isolating whether political connections affect the likelihood that public firms rely on Big 4 auditors that tightly constrain insiders’ discretion over financial reporting.2 Moreover, we examine three other research questions: (1) Are connected firms with ownership structures that leave minority investors more vulnerable to expropriation by insiders more likely to appoint Big 4 auditors?, (2) Are connected firms in countries with relatively weak governance institutions more eager to engage a Big 4 auditor?, and (3) Do connected firms benefit from hiring a Big 4 auditor?

Political connections heighten the tension that insiders in public firms experience in their financial reporting incentives. These insiders could exploit their position to deny outside investors by siphoning corporate resources that they later conceal by distorting the financial statements (e.g., Shleifer and Vishny [1994], La Porta et al. [1998]). In other words, they may manipulate accounting numbers to suppress information on actual economic performance in order to ensure that their diversionary practices, largely stemming from political cronyism and corruption, are kept hidden. In fact, Chaney, Faccio, and Parsley [2011] find that earnings quality is lower in politically connected firms. Rendering the financial statements less informative to provide cover for expropriation activities would be evident in the absence of a Big 4 auditor.

However, there are countervailing incentives pushing politically connected firms to improve disclosure. In particular, connected insiders who refrain from self-dealing would prefer higher-quality financial reporting to ensure that outside investors realize this. It follows that such politically connected firms would be more likely to appoint Big 4 auditors since investors value accounting transparency for protecting their interests (e.g., Watts and Zimmerman [1983], Dyck and Zingales [2004]). This argument reflects that more reliable financial reporting helps prevent expropriation by dominant insiders and their political patrons.

We evaluate which of these financial reporting incentives dominates by examining whether auditor choice varies systematically with political ties. Given that recent evidence implies that connected firms suffer unique agency problems that may magnify the demand for external monitoring—for example, Qian, Hongbo Pan, and Yeung [2011] find that insiders’ expropriation of minority investors in China is more severe in politically connected firms—we extend research on political connectivity and financial reporting outcomes to include the role of auditor choice.

Our analysis leads to four primary insights. First, we provide evidence that the demand for Big 4 auditors is greater for politically connected public firms relative to their nonconnected counterparts matched, using alternative techniques, on country, industry, size, ownership structure, and other characteristics. In various specifications, our coefficient estimates translate into political affiliations materially increasing the likelihood that firms will appoint a Big 4 auditor by a range of 5–8%, with all other variables assigned their mean values.3 Second, we report that connected firms with ownership structures that intensify insiders’ incentives to divert corporate resources are even more likely to appoint Big 4 auditors. Third, we find that the link between auditor choice and political ties is stronger in firms operating in countries with relatively poor governance institutions, implying that Big 4 audits become more valuable for disciplining connected insiders in these countries. Finally, our results suggest that connected firms with Big 4 auditors exhibit lower earnings management and enjoy greater transparency, higher valuations, and cheaper equity financing.

Although our core results are robust to confronting this issue various ways (e.g., alternative matching procedures, controlling for firm heterogeneity as well as an extensive set of country-level variables, and restricting the sample to firms with long auditor tenure), the potential endogeneity between auditor choice and political connections means that we cannot infer causality from this analysis. We address another important issue by isolating the incremental role that connections play in auditor choice beyond ownership characteristics, including large ultimate shareholding (e.g., Fan and Wong [2005]). In developing our predictions, we rely on prior research to motivate that political connections exacerbate agency conflicts between dominant large shareholders and outside investors (e.g., Morck, Stangeland, and Yeung [2000], Faccio [2006], Berkman, Cole, and Fu [2010], Qian, Hongbo Pan, and Yeung [2011]). Empirically, we employ several strategies to better identify the link between connections and auditor choice. First, we separately control in all regressions for the equity stakes held by the ultimate shareholder and the state, which have been shown to influence auditor choice (e.g., Fan and Wong [2005], Guedhami Pittman, and Saffar [2009]). Second, we apply several matching techniques to assemble a benchmark group of nonconnected firms against which we evaluate auditor choice in politically connected firms. Third, we conduct cross-sectional analysis to examine whether the link between political connections and auditor choice varies with corporate ownership characteristics. In all of these estimations, we find that connections affect auditor choice beyond the ownership effects.

The rest of this paper is organized as follows. Section 'Motivation' reviews prior research to develop the testable hypotheses. Section 'Data Description' outlines our data and reports descriptive statistics on the regression variables. Section 'Empirical Results' covers the empirical evidence and section 'Conclusions' concludes.

2. Motivation

2.1 THE IMPACT OF POLITICAL CONNECTIONS ON AUDITOR CHOICE

Political connections can benefit firms in many ways.4 However, another perspective holds that they can also lead to value-destroying tunneling by dominant insiders eager to at least recover the costs incurred in developing these ties (e.g., Morck, Stangeland, and Yeung [2000]). In fact, Qian, Hongbo Pan, and Yeung [2011] report that the share of earnings that insiders expropriate from outside investors in China exceeds the collective value that political connections generate for the firm. Controlling shareholders in connected firms may have more opportunity to divert corporate resources since they tend to be subject to fewer disciplinary constraints from regulators. For example, Berkman, Cole, and Fu [2010] find evidence from stock market returns in China implying that minority investors perceive that regulators will fail to protect their interests by strictly enforcing new governance standards when the firm has a controlling owner with political connections.5 Similarly, recent U.S. research suggests that politically connected firms that fraudulently exaggerate their earnings experience more lenient monitoring from regulators relative to fraudulent firms without political connections (e.g., Correia [2010], Yu and Yu [2011]).

Our research is grounded in prior research implying that Big 4 auditors supply better monitoring than non–Big 4 auditors. However, a natural question is whether the Big 4 outperform other auditors outside the United States where legal institutions governing investor protection are more benign. Some equity pricing evidence suggests that the severe exposure to civil lawsuits confronting auditors in the United States is responsible for a major fault line that separates this country from the rest of the world on differential audit quality (e.g., Khurana and Raman [2004], El Ghoul, Guedhami, and Pittman [2012]). Although both reputation incentives and litigation shape Big 4 audit quality in the United States (e.g., Baber, Kumar, and Verghese [1995], Mansi, Maxwell, and Miller [2004]), the Big 4's interest in protecting their reputations is largely behind their audits becoming economically distinct in other countries where it is harder for investors to recover damages when audit failure occurs.

In short, prior research suggests that reputation incentives are sufficient to generate an audit quality differential in countries with mild private enforcement against auditors. Consistent with theory (e.g., DeAngelo [1981], Rogerson [1983]), this evidence implies that large auditors with valuable reputations at stake provide stricter monitoring. For example, recent research on German (Weber, Willenborg, and Zhang [2008]) and Japanese (e.g., Skinner and Srinivasan [2012]) firms supports the reputation explanation for audit quality in countries that impose minimal discipline on auditors in the form of holding them liable for violating securities laws.6 Moreover, Big 4 auditors with global practices may provide uniformly high-quality assurance services worldwide to avoid undermining their reputations (e.g., Humphrey, Loft, and Woods [2009]). Altogether, this research helps justify our focus on the importance of Big 4 auditors to constraining insiders in politically connected firms against distorting their financial statements.

Politically connected firms may be reluctant to appoint Big 4 auditors to improve accounting transparency since prior research suggests that they have access to cheap loans from state-owned banks anyway (e.g., Dinç [2005], Claessens, Feijen, and Laeven [2008]). Moreover, Leuz and Oberholzer-Gee [2006] find that Indonesian firms with close connections to the state avoid raising capital from arm's length sources that insist on more transparency since they are eager to conceal transactions benefiting controlling insiders and their political backers. In motivating their research on the impact of politically charged events on the release of negative financial news by Chinese state-owned enterprises, Piotroski, Wong, and Zhang [2008, p. 3] explain that: “transparency will limit the ability of politicians and managers to consume their private benefits of control by exposing poor governance…” They find that connected firms heavily suppress information, which Piotroski, Wong, and Zhang [2008] partly attribute to these firms’ incentives to hide from minority shareholders’ expropriation-related activities stemming from political cronyism and corruption.

Similarly, Stulz [2005] concludes that the threat of public exposure has a sobering impact on whether politicians and insiders collude to extract private benefits. In evidence implying that smaller investors are marginalized in these situations, Fan, Wong, and Zhang [2007] find that Chinese firms with politically connected CEOs seldom appoint directors representing minority shareholders, which stands in sharp contrast to the large fraction of their directors who are affiliated with the largest shareholder or governments. Faccio [2006] reports some cross-country evidence from stock price reactions to announcements that politicians were joining the boards of firms in which outside investors perceived that controlling shareholders and their political allies would exploit these connections to expropriate them. Her evidence reconciles with prior research implying that politicians extract rents from the firms that they manage (e.g., Shleifer and Vishny [1994]). Politically connected firms in Canada tend to have concentrated ownership (e.g., Morck, Stangeland, and Yeung [2000]), reinforcing that outside investors have valid concerns that insiders in these firms may consume private benefits to their detriment. Accordingly, although prior research supports that political connections can add value to all shareholders, they can also engender agency conflicts between dominant insiders and outside investors.

Given their diverging accounting transparency incentives, it remains unclear whether insiders in politically connected firms turn to Big 4 auditors to lower information asymmetry. In one direction, the stricter monitoring imposed by a higher-quality auditor would reduce insiders’ discretion to distort financial reporting. Consequently, financing costs (valuations) for connected firms with Big 4 auditors will fall (rise) under this argument since accounting transparency helps outside investors identify any expropriation. In the other direction, connected insiders extracting private benefits may choose a non–Big 4 auditor to hide that they are depriving outside investors. Accompanying political ties according to this argument are strong incentives for insiders to deliberately render the financial statements less informative by hiring a lower-quality auditor to help cover their tracks.

This underlying tension in insiders’ financial reporting incentives motivates our analysis (Fan and Wong [2005]). We focus on how firms with political connections weigh the marginal benefits of appointing a Big 4 auditor (e.g., cheaper equity pricing and higher firm value) against its marginal costs (e.g., narrower scope to expropriate). Big 4 audits are a two-edged sword from the standpoint of insiders: negative when they are extracting private benefits and positive when they internalize outside investors’ best interests. The importance of political connections to the demand for Big 4 audits remains an empirical question that hinges on which financial reporting incentive dominates. We, largely for expositional convenience, predict that firms with political ties tend to hire higher-quality auditors to credibly commit to refrain from diverting corporate resources (all hypotheses are stated in the alternative):

  • H1: In comparison to other public firms, politically connected firms are more likely to appoint Big 4 auditors.

2.2 THE MEDIATING ROLE OF FIRMS’ OWNERSHIP STRUCTURES

Ownership characteristics may shape the role that political connections play in auditor choice. La Porta et al. [2002] report that ownership structures in which dominant shareholders exercise control despite owning only a small fraction of the firm's cash flow are widespread around the world.7 Mapping into our research questions, Fan and Wong [2005] find that the demand for external monitoring by a high-quality auditor in East Asian firms rises when the controlling shareholder's voting rights exceed her cash flow rights. Given that minority investors become more susceptible to expropriation as the ownership-control gap widens (e.g., Shleifer and Vishny [1997], Joh [2003], Faccio [2006]), we expect to observe that such connected firms are even more likely to appoint Big 4 auditors to reduce information asymmetry.

Similarly, firms with a single large shareholder suffer worse agency conflicts with outside investors. Pagano and Röell [1998] model that firms with multiple large shareholders—with both the ability (via their voting rights) and the incentive (via their cash flow rights) to actively cross-monitor each other—prevent insiders from accruing private benefits. In contrast, rather than requiring the consent of a coalition of large shareholders, a single dominant shareholder can unilaterally dictate corporate policy, including diversionary activities (Bennedsen and Wolfenzon [2000]). It follows that strict external monitoring by Big 4 auditors is more valuable to outside investors in connected firms when they cannot rely on committed internal monitoring by multiple major shareholders to protect their interests.

Finally, large shareholders in business groups can exploit, for example, pyramidal ownership to secure control rights that far exceed their equity stakes, providing them with greater incentives and means to siphon corporate resources than their counterparts in independent firms (e.g., La Porta, Lopez-de-Silanes, and Shleifer [1999], Claessens et al. [2000], Bae, Kang, and Kim [2002]). Firms belonging to business groups may be more likely to appoint a Big 4 auditor to provide outside investors with more assurance that they abstain from extracting private benefits. Since the separation of cash flow rights from voting rights in firms affiliated with a business group magnifies agency costs, minority investors will particularly value the presence of a Big 4 auditor in this situation. In our second prediction, we examine whether the importance of political connections to auditor choice varies with firms’ ownership structures:

  • H2: In comparison to other connected firms, politically connected firms with ownership characteristics that worsen agency conflicts with outside investors are even more likely to appoint Big 4 auditors.

2.3 THE MEDIATING ROLE OF COUNTRY-LEVEL INSTITUTIONS

Next, we evaluate whether country-level governance institutions mediate the relation between auditor choice and political connections. Recent evidence implies that political connections are prevalent in countries with underdeveloped legal institutions and pervasive corruption (e.g., Faccio [2006, 2010]). Reflecting the importance of transparency, Djankov et al. [2010] report that public disclosure of politicians’ finances and business activities correlates with lower perceived corruption. Similarly, Johnson et al. [2000] document that the risk of insider diversion is increasing in countries’ corruption. More recently, Boubakri et al. [2012] find that political connections are more valuable in countries with weak institutional environments. Consequently, we expect that operating in countries with worse institutional infrastructure intensifies connected firms’ incentives to engage a Big 4 auditor to lend more credibility to their financial statements. In our third prediction, we isolate whether the role that political ties play in auditor choice hinges on the quality of countries’ governance institutions:

  • H3: In comparison to other public firms, politically connected firms in countries with relatively poor governance institutions are even more likely to appoint Big 4 auditors.

2.4 ECONOMIC IMPLICATIONS OF AUDITOR CHOICE IN POLITICALLY CONNECTED FIRMS

Finally, we examine several economic outcomes to help empirically clarify what is behind any evidence that political connections elicit greater demand for Big 4 auditors. Since exploiting connections to orchestrate the diversion of corporate resources requires hiding, credible financial reporting plays a natural role in protecting outside investors by lowering information asymmetry. Accordingly, we predict that connected firms intent on reducing agency costs by appointing a Big 4 auditor exhibit less earnings management, enabling them to enjoy greater transparency, lower equity financing costs, and higher valuations:

  • H4: In comparison to other politically connected firms, politically connected firms with Big 4 auditors benefit from lesser earnings management, greater transparency, higher valuations, and cheaper equity financing.

3. Data Description

3.1 THE SAMPLE

To analyze the impact of political connections on auditor choice, we hand-match data from two sources: political connections data from Faccio [2006] and auditor identity, financial statement, and ownership data from Worldscope. We confine the analysis to countries with connected firms according to Faccio [2006], which yields a sample of 1,371 (30,181) firm-year observations with (without) political connections from 28 countries covering the period from 2001 to 2005.8 Table 1 presents sample characteristics for the politically connected firms by country. It is evident from the country distribution that the sample exhibits good diversification across geographical regions, including Asia, Europe, Latin America, and North America, which is important when examining the interplay between political connectivity and country-level governance institutions. The United Kingdom and Malaysia contribute the largest share of the sample at 28.8% and 19.6%, respectively, followed by Indonesia (7.7%), Japan (7.3%), and Thailand (6.8%). All other countries comprise less than 4% of the sample.

Table 1. Politically Connected Firms’ Distribution by Country
CountryN%CountryN%
  1. This table reports the country and industry distribution for the sample of 1,371 politically connected firms from 28 countries.

Austria50.36Italy372.70
Belgium201.46Japan1007.29
Canada50.36Korea, South251.82
Chile100.73Malaysia26819.55
Denmark100.73Mexico201.46
Finland100.73Philippines181.31
France513.72Singapore543.94
Germany282.04Spain100.73
Greece50.36Sweden100.73
Hong Kong221.60Switzerland151.09
India50.36Taiwan50.36
Indonesia1057.66Thailand936.78
Ireland100.73United Kingdom39528.81
Israel100.73United States251.82
   Total1,371100

3.2 VARIABLES AND DESCRIPTIVE STATISTICS

3.2.1. Measuring Auditor Quality

We follow extensive prior research by gauging auditor quality with a dummy variable labeled BIG 4 that takes the value of one for firms with Big 4 auditors (and their predecessors), and zero otherwise.9 Given that the Worldscope database does not provide auditor history details, we ensure accuracy in coding BIG 4 by identifying auditors with five (2001–2005) of its compact discs. The descriptive statistics in table 2 reveal that connected firms rely more on Big 4 auditors compared to nonconnected firms; the 82–77% difference in market share is statistically significant at the 1% level. The Big 4 market share in our sample is slightly higher than the 71% reported in Choi and Wong [2007] for a sample of public firms from 39 countries for the period 1993–1998 and the 74% reported in Francis and Wang [2008] for a sample of firms from 42 countries for the period 1994–2004. Table A1 summarizes all variables used in the analysis, including the data sources.

Table 2. Descriptive Statistics for the Politically Connected Firms
 MeansMedians
 Connected FirmsNonconnected Firmst-statisticsConnected FirmsNonconnected Firmsz-statistics
  1. This table reports measures of central tendency for all explanatory variables according to political connection. The full sample includes 1,371 politically connected firms and 1,911 nonconnected firms. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively. Definitions and data sources for the variables are provided in table A1.

BIG 40.820.773.39***   
LARGEOWN0.270.260.570.200.170.89
COMPLEXITY2.742.603.08***3.002.002.68***
FOREIGNSALES20.4618.062.22**0.000.002.05**
FINANCING0.070.08–0.99   
CROSS-LISTING0.100.055.48***   
SIZE15.6815.422.18**15.1515.502.25**
STATEOWN0.260.062.06**0.000.001.94*
ROA0.030.031.210.040.030.89
LEVERAGE0.590.446.59***0.340.187.31***
GROWTH4.856.30–2.36**3.252.881.34
INV0.090.10–4.91***0.060.08–5.36***
MTB1.701.611.93*1.271.201.53
ACCURACY−0.01−0.010.520.000.000.70
IAS0.180.161.67*   
ANALYSTS10.918.335.63***10.006.006.62***
EM0.040.08−4.20***0.040.05−3.84***
KMED11.1511.76−0.7610.3410.83−0.38

3.2.2. Measuring Political Connections

We rely on Faccio's [2006] database to measure political connections. Faccio [2006, p. 369] explains that a firm is considered politically connected if “at least one of its large shareholders (anyone controlling at least 10% of voting shares) or one of its top officers (CEO, president, vice-president, chairman, or secretary) is a member of parliament, a minister, or is closely related to a top politician or party.” Apart from its extensive country coverage, an important upside of this database is its considerable detail on the type of connection (i.e., connection with members of parliament, a minister or the head of state, and close relationship to a top official). Faccio [2006] classifies a firm as connected through a minister or head of state when the politician or a close relative (son or daughter) holds this office and is a large shareholder or top officer. In her analysis, a firm is connected with a member of parliament when the large shareholder or the top director is a member of parliament. Relatives are not included in this classification. Close connections in the form of well-known friendships and connections are identified by several sources (The Economist, Forbes, or Fortune) and prior studies (e.g., Backman [1999], Agrawal and Knoeber [2001], Fisman [2001]). The names of top officers and shareholders are drawn from Worldscope, Extel, Lexis-Nexis, and company Web sites. Shareholder information is also collected from Claessens et al. [2000], Faccio and Lang [2002], and various stock exchanges Web sites. Applying these definitions, Faccio [2006] identifies 541 firms with political connections. We follow recent research by specifying our main test variable as a dummy variable (CONNECTED) that takes the value of one if a company is identified as politically connected in Faccio's [2006] database, and zero otherwise (e.g., Faccio, Masulis, and McConnell [2001], Boubakri, Cosset, and Saffar [2008], Boubakri et al. [2012], Faccio [2010], Chaney, Faccio, and Parsley [2011]).

3.2.3. Measuring Country-Level Governance Institutions

We rely on four widely used proxies to calibrate the quality of countries’ governance institutions. The first variable is Faccio's [2006, p. 373] regulatory score (RESTRICTIONS), which is an assessment of the stringency of “regulations that prohibit or set limits on the business activities of public officials.” This regulatory score varies from zero to six with high values indicating major restrictions on public officials. The second variable is La Porta et al.'s [1998, p. 1125] measure of the risk of expropriation (EXPROPRIATION), which is an assessment of the “risk of a modification in a contract taking the form of a repudiation, postponement, or scaling down” due to “budget cutbacks, indigenization pressure, a change in government, or a change in government economic and social priorities.” We recode the index from 0 to 10 with higher values reflecting greater risk of expropriation. The third variable is Djankov et al.'s [2008] anti-self-dealing index (ANTISELF), which focuses on the regulation of corporate self-dealing transactions in 72 countries along three dimensions: disclosure, approval procedures for transactions, and facilitation of private litigation when self-dealing is suspected. The fourth proxy is La Porta, Lopez-de-Silanes, and Shleifer's [2006] measure of investor protection (PROTECTION), which is equal to the principal component of the anti-director rights, disclosure requirements, and liability standards indices described in their database. Table A2 reports the means of these variables by country.

3.2.4. Control Variables

In our multivariate regression analysis, we attempt to isolate the role that political connections play by comprehensively controlling for two sets of variables known to affect auditor choice according to prior research (e.g., Mansi, Maxwell, and Miller [2004], Fan and Wong [2005], Lennox [2005], Choi and Wong [2007], Fortin and Pittman [2007], Wang, Wong, and Xia [2008], Guedhami, Pittman, and Saffar [2009]). The first set includes the following firm-level characteristics: firm size (SIZE), which we measure with the natural logarithm of total assets expressed in U.S. dollars;10 asset structure (INV), which we capture with the ratio of inventory to total assets; leverage (LEVERAGE), which we code as the ratio of long-term debt to total equity; growth (GROWTH), which amounts to the asset growth ratio in the past year; ownership structure variables: the equity stakes held by the largest shareholder (LARGEOWN) and the government (STATEOWN), and the percentage of voting rights belonging to the ultimate owner (CONTROLRIGHTS) obtained from Claessens, Djankov, and Lang [2000] and Faccio and Lang [2002]; complexity (COMPLEXITY), which we measure with the number of business segments based on two-digit SIC codes; foreign sales (FOREIGNSALES), which is the portion of sales from foreign operations; level of financing activities, which we capture with two variables: a dummy variable that takes the value of one if the sum of new long-term debt plus new equity exceeds 20% of total assets (FINANCING) and cross-listing in foreign markets (CROSS-LISTING); and profitability (ROA), which is the return on assets ratio. The second set of controls includes two macroeconomic variables, namely the logarithm of GDP per capita (LGDPC) and a country's foreign direct investment as a fraction of GDP (FDI), and a proxy for a country's auditor discipline infrastructure, which reflects the intensity of civil litigation against auditors (SUE AUDITOR) according to La Porta, Lopez-de-Silanes, and Shleifer [2006]. Prior evidence suggests a high correlation between the level of economic development and the demand for transparency, including auditor choice (e.g., Leuz, Nanda, and Wysocki [2003], Guedhami, Pittman, and Saffar [2009]).

Table 2 presents descriptive statistics for the regression variables. The average politically connected firm in our sample is relatively large according to its assets (SIZE = 15.7; $553 million before logarithmic transformation) and valuable (MTB = 1.7), with considerable long-term debt to equity (LEVERAGE = 59%) and moderate asset growth (GROWTH = 5%) and foreign operations (FOREIGNSALES = 20%). In comparison to nonconnected firms, we find that connected firms, on average, are larger, more leveraged, more complex, and more likely to cross-list in foreign markets. Although connected firms have a higher fraction of state ownership and foreign sales than nonconnected firms, their growth rates and inventory levels tend to be lower. These differences are consistent with prior research (e.g., Boubakri, Cosset, and Saffar [2008], Boubakri et al. [2012], Faccio [2010]).

Table A3 reports correlation coefficients between the regression variables while allowing for country and firm level clustering. Consistent with the prediction in H1, we observe a positive correlation between CONNECTED and BIG 4 that is significantly different from zero at the 1% level. In the next section, we consider whether this preliminary evidence that politically connected firms are more likely to appoint Big 4 auditors persists in a series of multivariate regressions. Finally, we find that the correlations between the controls are generally small, reducing concerns that multicollinearity is spuriously responsible for our evidence on the predictions.

4. Empirical Results

In a multivariate regression framework, we estimate the impact of political connections on the likelihood that firms will hire a Big 4 auditor to examine the prediction in H1. Next, we analyze the prediction in H2 that agency problems embedded in firms’ ownership structures strengthen the link between political connections and auditor choice. For the prediction in H3, we isolate whether the importance of political connections to auditor choice hinges on the quality of a country's governance institutions. Finally, we examine whether politically connected firms with Big 4 auditors benefit from lower earnings management, greater transparency, higher valuations, and cheaper financing under the prediction in H4.

4.1 POLITICAL CONNECTIONS AND AUDITOR CHOICE

4.1.1. Main Evidence

To evaluate the link between political connections and auditor choice, we focus on the sample of connected firms described in section 'THE SAMPLE' and a set of peer firms without political connections. In order to improve identification by confronting the threat that differences in firm characteristics, such as their ownership structures and size, are spuriously responsible for any evidence supporting our predictions, we employ two separate matching techniques to specify the benchmark group of nonconnected firms.11 The first peer group consists of nonconnected firms matched to connected firms based on country, industry, year, and decile of total assets. The second group comprises firms matched on various observable characteristics according to a propensity score matching procedure that we outline below.

Table 3 presents the results from estimating several pooled multivariate logistic regressions to analyze the impact of political connections on auditor choice worldwide. All standard errors are clustered at the firm and country level, and adjusted for heteroskedasticity. We rely on two-sided tests to gauge statistical significance in all estimations. To better assess the economic impact of the key test variable (CONNECTED), we also report marginal effects in square brackets. In these regressions, we control for ownership structure in alternative ways. In models 1 and 2, we control for the equity stake of the largest shareholder (LARGEOWN), rather than their ultimate ownership (CONTROLRIGHTS), to reduce data attrition. For example, narrowing our focus to firms with ultimate ownership data would lower the countries under study from 28 to 19, preventing us from examining firms from countries such as the United States and Mexico that have a considerable number of political connections according to table 1. However, we replace LARGEOWN with CONTROLRIGHTS in models 3 and 4 and with both CONTROLRIGHTS and CASHFLOWRIGHTS in models 5 and 6 to better isolate the incremental importance of political connections to auditor choice beyond the ultimate ownership effects that Fan and Wong [2005] document.

Table 3. Political Connections and the Auditor Choice
 Full SampleUltimate Ownership Sample
 Connected Firms Matched on Country, Industry, Year, and Decile of Total AssetsPropensity Score MatchingConnected Firms Matched on Country, Industry, Year, and Decile of Total AssetsPropensity Score MatchingConnected Firms Matched on Country, Industry, Year, and Decile of Total AssetsPropensity Score Matching
  1. This table reports pooled logit estimation results for auditor choice of politically connected firms and nonconnected firms. The sample consists of 1,371 politically connected firms and the set of peer firms without political connections matched to connected firms based on country, industry, year, and decile of total assets (models 1, 3, and 5) and various observable characteristics according to a propensity score matching procedure (models 2, 4, and 6). Beneath each estimate is reported the robust z-statistic clustered at both the country and the firm level. The table also reports the marginal effect of the variable CONNECTED. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively, based on two-sided tests. The definitions and data sources for the variables are outlined in table A1.

Models123456
CONNECTED0.41**0.32*0.54**0.36**0.53**0.36**
 (2.52)(1.78)(2.48)(2.06)(2.45)(2.05)
LARGEOWN0.67*−0.17    
 (1.78)(−0.47)    
CONTROLRIGHTS  −1.30***0.24−1.22**0.31
   (−2.66)(0.62)(−2.34)(0.49)
CASHFLOWRIGHTS    −0.19−0.09
     (−0.42)(−0.15)
COMPLEXITY−0.11*−0.050.02−0.010.02−0.01
 (−1.76)(−0.78)(0.23)(−0.15)(0.23)(−0.14)
FOREIGNSALES0.01***0.000.01**0.000.01**0.00
 (3.02)(0.82)(2.36)(0.23)(2.33)(0.22)
FINANCING0.01−0.220.15−0.270.15−0.27
 (0.07)(−1.17)(0.55)(−1.49)(0.54)(−1.49)
CROSS-LISTING0.620.440.83**0.490.82**0.48
 (1.47)(0.95)(2.08)(1.05)(2.05)(1.05)
SIZE0.15***0.21***0.10***0.20***0.10***0.20***
 (6.25)(6.87)(3.39)(6.44)(3.40)(6.34)
ROA2.08***0.74*1.95***0.611.97***0.61
 (4.09)(1.75)(2.71)(1.46)(2.75)(1.46)
STATEOWN−0.05***−0.06**−0.02−0.06**−0.02−0.06**
 (−3.60)(−2.30)(−0.40)(−2.38)(−0.41)(−2.38)
LEVERAGE−0.05−0.100.01−0.080.01−0.08
 (−0.46)(−0.87)(0.09)(−0.70)(0.07)(−0.70)
GROWTH−0.01***−0.00−0.00−0.00−0.00−0.00
 (−2.84)(−0.16)(−0.21)(−0.55)(−0.23)(−0.56)
INV−1.19*−0.96−1.93**−0.75−1.92**−0.75
 (−1.69)(−1.37)(−2.47)(−1.07)(−2.46)(−1.07)
FDI0.07***0.07***0.010.07***0.010.07***
 (3.41)(2.97)(0.27)(2.82)(0.30)(2.82)
LGDPC0.33***0.42***0.17**0.45***0.17*0.45***
 (4.58)(5.41)(2.03)(5.61)(1.91)(5.62)
SUE AUDITOR1.05***1.12***1.66***1.10***1.64***1.09***
 (2.73)(3.32)(3.43)(3.18)(3.40)(3.17)
Intercept−4.99***−6.43***−2.80**−6.81***−2.71**−6.82***
 (−5.91)(−6.42)(−2.43)(−6.45)(−2.31)(−6.45)
[Marginal effect of CONNECTED in %][6.11][5.37][7.69][6.28][7.60][6.25]
Pseudo R20.100.110.090.110.090.10
χ2153.8132.190.37119.390.57119.4
N3,2822,7422,6902,2662,6902,266

Model 1 reveals that the coefficient for the dummy variable identifying politically connected firms (CONNECTED) loads positively at the 5% level, implying that these firms are more likely to hire Big 4 auditors relative to nonconnected peers in the same country, industry, year, and decile of total assets.12 Reflecting its material economic impact, the coefficient estimate for CONNECTED translates into political affiliations increasing the likelihood of appointing a Big 4 auditor by 6%, with all other variables assigned their mean values. This result is consistent with the prediction in H1 that politically connected firms are associated with greater demand for Big 4 auditors.13

Next, we examine the demand for a Big 4 auditor between connected firms and nonconnected peers matched according to propensity scores derived as follows (e.g., Boubakri et al. [2012]). First, we require that the candidate firm for the matching share the same country, year, and industry class as the connected firm. Second, among the potential control sample firms, we select the optimal match based on the nearest neighbor technique of the propensity score matching procedure. We follow Rosenbaum and Rubin [1983] and Heckman, Ichimura, and Todd [1997, 1998] in relying on this procedure in an attempt to control for differences in characteristics between connected and nonconnected firms. To calculate the propensity score, we analyze a comprehensive set of firm characteristics that should capture the likelihood that a given firm will be politically connected according to prior research. More specifically, we consider size, leverage, the largest shareholder's ownership stake, state ownership, and cross-listing as connected firms are likely to be different from nonconnected peers along these characteristics (see, e.g., Faccio [2006, 2010], Leuz and Oberholzer-Gee [2006], Boubakri, Cosset, and Saffar [2008], and Bunkanwanicha and Wiwattanakantang [2009]).

This matching procedure translates into a sample of 2,742 firm-year observations equally distributed by country, industry, and year between politically connected and nonconnected firms. Despite the major data attrition that accompanies constructing a matched sample using propensity scores in our setting, an upside of applying this technique here is the large number of potential matches (30,181 nonconnected observations), ensuring that the connected and nonconnected samples have extremely close propensity scores. In model 2, we find that connected firms are significantly more likely to appoint a Big 4 auditor than their propensity score matched peers. The coefficient estimate for CONNECTED in model 2 implies that a 5% increase in the probability of engaging a Big 4 auditor accompanies a political connection. In unreported tests, we exploit the large number of close matches available in our data set by implementing one-to-five and one-to-ten matching, and find that CONNECTED loads positively at the 5% and 1% levels, respectively, in these larger samples.

To better isolate the impact of political connections on auditor choice, we replace in models 3 and 4 the variable LARGEOWN with the ultimate ownership variable CONTROLRIGHTS after Fan and Wong [2005]. In both regressions, we continue to estimate a positive and statistically significant (at the 5% level) coefficient on CONNECTED. Corroborating our earlier evidence, these results imply that connected firms are more likely to engage a Big 4 auditor compared to their peer group of nonconnected firms. Economically, the coefficient estimates for CONNECTED suggest that firms become 8% (model 3) and 6% (model 4) more likely to hire a Big 4 auditor in the presence of a political connection, with all other variables set to their mean values. We complement the propensity score matching evidence in model 4 by conducting one-to-five and one-to-ten matching to take advantage of the deep pool of control observations available in our data set. In these unreported estimations, we find that CONNECTED is positive and statistically significant at the 1% level in both cases. In models 5 and 6, we continue to find that CONNECTED loads positively at the 5% level when we add the cash flow rights of the ultimate owner to models 3 and 4. Consistent with Fan and Wong [2005], CASHFLOWRIGHTS is statistically insignificant in these regression models, while CONTROLRIGHTS only enters negatively in model 5. It is important to note that these ownership variables are highly correlated in our data (ρ = 0.77, p < 0.001).

In other unreported analysis, we follow Fan and Wong [2005] by specifying control concentration with a dummy variable (CONTROLRIGHTS ≥ 30%) assigned the value one if CONTROLRIGHTS is at least 30%, and zero otherwise. Additionally, in successive regressions, we replace CONTROLRIGHTS with the separation between the voting and cash-flow rights belonging to the ultimate owner (WEDGE), and a dummy variable set equal to one if WEDGE is at least 20%, and zero otherwise. We find that our core evidence on CONNECTED holds at the 5% level or better in these re-estimations, reinforcing that the importance of political connections to auditor choice is incremental to the impact of the ultimate ownership characteristics that Fan and Wong [2005] examine.14

Collectively, the results in table 3 using different matching techniques suggest that the demand for a Big 4 auditor is higher in politically connected firms relative to their nonconnected peers. Importantly, our evidence persists after controlling for a comprehensive set of firm-level determinants of auditor choice. In particular, we find that the role that political connections play in auditor choice extends beyond the impact of the ownership variables capturing the equity stakes held by the ultimate shareholder and the state. Interestingly, the coefficient for state ownership (STATEOWN) is generally negative and statistically significant in table 3. This result suggests that the demand for Big 4 auditors is decreasing in state ownership, supporting that the evidence in Guedhami, Pittman, and Saffar [2009] for the specific case of newly privatized firms generalizes to public firms worldwide. More generally, this evidence squares with prior research on the political economy and financial reporting transparency (e.g., Bushman, Piotroski, and Smith [2004], Bushman and Piotroski [2006]). Among the other firm-level determinants, we find that firm size, profitability, and foreign sales are positively related to auditor choice; only FINANCING and LEVERAGE have no perceptible impact on this decision in any of the table 3 regressions. Additionally, the three country-level controls (FDI, LGDPC, and SUE AUDITOR) generally enter positively and significantly at the 5% level or better, implying that the demand for high-quality auditors is higher in countries with more foreign investor involvement, more developed economies, and more discipline imposed on auditors through civil litigation institutions.

4.1.2. Sensitivity Analyses

In this section, we evaluate whether our earlier evidence on the importance of political connections to auditor choice persists when we tackle potential omitted variables bias and endogeneity, re-estimate our regressions on alternative matched samples, and respecify auditor quality. To preview, the results that we report in table 4 lend additional support to the prediction in H1 that public firms with political connections are more likely to appoint high-quality auditors. Except when examining whether our evidence holds when we apply alternative matching approaches, we rely on the sample of connected firms and nonconnected peers matched based on country, industry, year, and decile of total assets (model 3 in table 3) as our baseline in these sensitivity tests.

Table 4. Political Connections and the Auditor Choice: Sensitivity Tests
 Random EffectsAdditional Control VariablesExclude Short Auditor TenureConnected Firms Matched on Country, Industry, Year, and Decile of the Largest Shareholder's Voting RightsConnected Firms Matched on Country, Industry, Year, and Largest Owner Identity
  1. This table reports pooled logit estimation results for auditor choice of politically connected firms and nonconnected firms using the specification in model 3 in table 3 as the baseline regression. Model 1 uses panel random effects to estimate our baseline specification. Model 2 includes OPENNESS, NEWSPAPER, DISCLOSE, JUDICIAL, RULEOFLAW, PREDATION, and BUSINESSGROUP as additional control variables. Model 3 considers the sample of connected firms that neither upgrade from a non–Big 4 to a Big 4 auditor nor downgrade from a Big 4 to a non–Big 4 auditor between 1998 and 2005 (i.e., the model excludes firms with short auditor tenure). Model 4 considers a sample of connected firms matched on country, industry, year, and decile of the largest shareholder's voting rights. Model 5 considers a sample of connected firms matched on country, industry, year, and the largest shareholder's identity. Beneath each estimate is reported the robust z-statistic clustered at both the country and the firm level. The table also reports the marginal effect of the variable CONNECTED. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively, based on two-sided tests. The definitions and data sources for the variables are outlined in table A1.

Models12345
CONNECTED1.74**0.59**1.00***0.56**0.54**
 (2.44)(2.13)(3.07)(2.36)(2.48)
CONTROLRIGHTS−4.92***−0.50−1.02*−1.21**−1.30***
 (−2.93)(−0.87)(−1.74)(−2.08)(−2.66)
COMPLEXITY−0.000.03−0.01−0.020.02
 (−0.00)(0.37)(−0.16)(−0.25)(0.23)
FOREIGNSALES0.010.01*0.01**0.01*0.01**
 (1.24)(1.92)(2.32)(1.71)(2.36)
FINANCING1.61**0.300.120.460.15
 (1.99)(1.01)(0.34)(1.46)(0.55)
CROSS-LISTING2.35*0.87**1.49**1.09*0.83**
 (1.83)(2.15)(2.19)(1.91)(2.08)
SIZE0.27***0.23***0.10**0.12***0.10***
 (2.89)(4.44)(2.51)(3.45)(3.39)
STATEOWN−0.01−0.020.010.01−0.02
 (−0.04)(−0.49)(0.14)(0.09)(−0.40)
ROA3.24*2.08***1.80**2.49***1.95***
 (1.66)(2.83)(2.06)(2.66)(2.71)
LEVERAGE0.030.030.24−0.120.01
 (0.10)(0.24)(1.46)(−0.76)(0.09)
GROWTH−0.00−0.00−0.00−0.00−0.00
 (−0.33)(−1.06)(−0.62)(−0.22)(−0.21)
INV−7.96***−0.65−2.40***−1.86*−1.93**
 (−3.50)(−0.76)(−2.61)(−1.92)(−2.47)
FDI−0.10**−0.04*0.010.020.01
 (−2.25)(−1.95)(0.35)(0.61)(0.27)
LGDPC0.44−0.570.030.21**0.17**
 (1.62)(−1.34)(0.26)(2.04)(2.03)
SUE AUDITOR8.37***2.16**2.06***1.51**1.66***
 (4.80)(2.40)(3.39)(2.40)(3.43)
OPENNESS 0.01***   
  (2.60)   
NEWSPAPER 0.36   
  (1.08)   
DISCLOSE 1.71   
  (1.57)   
JUDICIAL −0.14   
  (−0.95)   
RULEOFLAW 0.14   
  (0.48)   
PREDATION −0.95***   
  (−4.53)   
BUSINESSGROUP −0.32   
  (−1.10)   
Intercept−4.78−0.51−1.47−3.31**−2.80**
 (−1.34)(−0.27)(−1.00)(−2.43)(−2.43)
[Marginal effect of CONNECTED in %][8.16][9.84][7.98][7.69]
Pseudo R20.140.110.090.09
χ2116.167.8967.3590.37
N2,6902,5342,1132,1442,690

4.1.2.1. Omitted Variables Bias and Endogeneity

We begin the sensitivity analysis by confronting whether endogeneity or omitted variables bias explains the evidence in table 3. Although our research design mitigates these issues since we control for an extensive set of firm- and country-level variables and implement various matching procedures, we address lingering concerns with three techniques.

First, in unreported analysis, we estimate a treatment effects model since it is plausible that some unobserved determinants of auditor choice may also affect political connections, causing our reported results to be biased and inconsistent. After Chaney, Faccio, and Parsley [2011] and Chen, Chen, and Wei [2011], we rely on CAPITAL—a dummy variable set equal to one if the firm is located in the capital city, and zero otherwise—as an instrument for political connectivity given prior evidence on its role in facilitating the formation of political connections (e.g., Roberts [1990], Agrawal and Knoeber [2001], Bertrand et al. [2007]). Importantly, the correlation between CAPITAL and BIG 4 is small in our data set (ρ = 0.05), helping to justify the validity of this exclusion restriction (e.g., Larcker and Rusticus [2010], Lennox, Francis, and Wang [2012]). In the first stage, we perform a logit regression of political ties on the dummy variable CAPITAL and the set of independent variables included in model 3 of table 3. The first stage fitted values for political connections are then used in the second stage logit regression explaining auditor choice. Similar to Boubakri, Cosset, and Saffar [2008] and Chaney, Faccio, and Parsley [2011], the first stage regression results include that the firm's location is a good predictor of political connections: the coefficient for CAPITAL loads positively at the 1% level, implying that the incidence of connections is higher for firms located in the capital. Consistent with the prediction in H1, we find in the second stage logit estimation that the coefficient of the (predicted) political connection variable is positive and statistically significant at the 1% level.15

Second, we exploit the panel nature of our data by estimating a random effects model. In the results reported in model 1 of table 4, we continue to find that appointing a Big 4 auditor becomes more likely in the presence of a political connection. This is constructive for mitigating the concern that omitted variables are spuriously behind the evidence supporting the prediction in H1. Third, we control for several additional country-level factors that capture: openness to international trade (OPENNESS); the level of freedom of the press (NEWSPAPER), which measures the extent of public opinion pressure according to Dyck and Zingales [2004]; the risk of state expropriation (PREDATION) using Durnev and Fauver's [2009] predation index; and three proxies for the quality of legal institutions derived from La Porta, Lopez-de-Silanes, and Shleifer [2006], namely, accounting standards (DISCLOSE), the efficiency of the judicial system (JUDICIAL), and the rule of law (RULEOFLAW). Prior research implies that these country-level variables matter to the demand for accounting transparency as well as the presence and value of political connections (e.g., Faccio [2006], Choi and Wong [2007], Guedhami, Pittman, and Saffar [2009], Boubakri et al. [2012]). Besides these country-level variables, we control for business group membership (BUSINESSGROUP), which may affect group-wide auditor choice. Despite that adding these controls leads to some sample attrition, we find in model 2 that the positive relation between CONNECTED and auditor choice remains in this regression. Although these additional tests provide some assurance that endogeneity is not responsible for the link that we observe between political connections and auditor choice, we stress that we still cannot infer causality from the analysis.

It also would be premature to conclude that our evidence reflects that politically connected firms are more likely to appoint a Big 4 auditor to provide stricter monitoring of the financial reporting process unless we consider two other competing explanations for our evidence. First, a Big 4 auditor can protect its valuable reputation and avoid litigation by refusing to accept engagements from clients that are more apt to resort to manipulating their financial statements to hide underlying performance, or resign from engagements when the ensuing risk reaches an intolerable level (e.g., DeFond, Ettredge, and Smith [1997], Shu [2000], Johnstone and Bedard [2004]). Second, insiders intending to materially distort their firms’ financial statements by, for example, overstating earnings to conceal their diversionary activities may prefer to hire a non–Big 4 auditor in order to fly under the radar when they begin accumulating private benefits at the expense of outside investors (e.g., Fan and Wong [2005], Guedhami, Pittman, and Saffar [2009]).

We follow Lennox and Pittman [2010] in considering the potential roles that screening by auditors and selection by their clients play by evaluating whether our core results hold when we isolate firms with long auditor tenure. The intuition for this analysis is that endogeneity is worse when the duration between auditor choice and the decision to deliberately exaggerate earnings is shorter since a firm planning to misreport would require less lead time to cover its tracks by switching to a non–Big 4 auditor, while its Big 4 auditor may shed clients that have become riskier (e.g., Jones and Raghunandan [1998], Johnstone and Bedard [2004]).16 Consequently, we restrict the sample to firms that neither upgrade from a non–Big 4 to a Big 4 auditor nor downgrade from a Big 4 to a non–Big 4 auditor between 1998 and 2005—this timeframe covers the five years (2001–2005) under study in the rest of our analysis and the three preceding years—since we are on more solid ground in treating BIG 4 as predetermined when auditor tenure is longer (Myers, Myers, and Omer [2003], Caramanis and Lennox [2008], Chang, Dasgupta, and Hilary [2009]). In this smaller sample, we find support at the 1% level in model 3 for the prediction in H1, reducing concern that screening or selection phenomena are spuriously responsible for our earlier evidence. Our results are almost identical when we apply the same five-year tenure breakpoint as Lennox and Pittman [2010].

4.1.2.2. Alternative Samples

In recent cross-country research, Guedhami, Pittman, and Saffar [2009] examine auditor choice in state-owned enterprises and during their transition to private ownership. In particular, they report a negative relation between the extent of government ownership and the likelihood of selecting a Big 4 auditor. They also find that privatization corrects the distortions in auditor choice that Wang, Wong, and Xia [2008] stress, although continued government ownership lowers the probability of appointing a Big 4 auditor after privatization. To alleviate the concern that our evidence on the role that political connections play in auditor choice reflects the government's ownership, we control for its equity stake in all regressions. In unreported analysis, we better isolate the importance of political connections to this decision by excluding privatized firms. The positive and statistically significant relation at the 1% level between political connections and auditor choice persists in this regression.

Our empirical strategy in table 3 involves applying two forms of matching to improve identification. Next, we expand this analysis to consider whether our core evidence is robust to implementing alternative matching techniques. First, in model 4, we compare connected firms to their nonconnected peers matched based on country, industry, year, and decile of the largest shareholder's voting rights. Second, in model 5, we select nonconnected firms based on country, industry, year, and the largest shareholder's identity—the specific types reflect ownership by widely held nonfinancial institutions, widely held financial institutions, families, the state, and dispersed shareholders according to data available from Claessens et al. [2000] and Faccio and Lang [2002]. Most relevant to our purposes, CONNECTED remains positive and statistically significant at the 5% level in both regressions, reinforcing our earlier evidence on the link between political connection and auditor choice.

4.1.2.3. Alternative Dependent Variables

There are reasons to doubt that BIG 4 is the relevant construct for all of the countries represented in our sample. For example, although table 2 reports that the Big 4 audit nearly 82% of the politically connected firms in our sample, data inspection reveals wide variation in their market shares across countries. Consequently, we respecify auditor choice by coding a dummy variable, LARGEST FIVE, one when the auditor is among the five largest in the country according to client assets, and zero otherwise. In unreported analysis, we find that CONNECTED remains positive and statistically significant at the 5% level when we replace BIG 4 with this alternate dependent variable. We find almost identical evidence when we calibrate market share with the number of clients rather than their assets (e.g., Wang, Wong, and Xia [2008]), or identify large auditors as those that hold at least a 5% market share in the country (e.g., DeFond, Wong, and Li [2000]).17

4.2 POLITICAL CONNECTIONS AND AUDITOR CHOICE: THE MEDIATING ROLE OF FIRMS’ OWNERSHIP STRUCTURES

In table 5, we examine in panel A whether the importance of political connections to auditor choice varies systematically with the extent of agency costs embedded in firms’ ownership structures. Initially, we estimate whether the relation between auditor choice and political ties is sensitive to the presence of any wedge between the controlling shareholder's voting rights and cash flow rights. To ensure that we draw valid inferences concerning the interaction, we also report the corrected mean interactive effect of the interaction terms using the methodology proposed by Ai and Norton [2003] for nonlinear models. In model 1, we find that the OWNERSHIPWEDGE*CONNECTED interaction has no perceptible impact on auditor choice, inconsistent with the prediction in H2 that control diverging from ownership explains demand for high-quality auditors in connected firms. Next, we analyze in model 2 whether connected firms are more likely to hire a Big 4 auditor when the control rights of the ultimate owner exceed 50% and find supportive results.

Table 5. Political Connections and Auditor Choice
Panel A: The mediating role of the ownership structure
 OWNERSHIP WEDGECONTROL RIGHTS > 50SINGLE LARGEOWNNUMBER OF LARGE OWNERSBUSINESS GROUP
Models12345
OWNERSHIPWEDGE × CONNECTED0.13    
 (0.29)    
CONTROLRIGHTS > 50 × CONNECTED 1.12*   
  (1.88)   
SINGLELARGEOWN × CONNECTED  1.25**  
   (2.25)  
NUMBER OF LARGE OWNERS × CONNECTED   −0.84** 
    (−2.02) 
BUSINESSGROUP × CONNECTED    1.76***
     (2.88)
Intercept−2.92**−2.60**−2.37**−3.07**−2.46*
 (−2.54)(−2.07)(−1.96)(−2.44)(−1.92)
[Marginal effect of #VARIABLES × CONNECTED in %][1.88][12.09][14.59][−12.83][14.86]
Mean interactive effect of #VARIABLES × CONNECTED0.020.130.17**−0.12*0.14*
 (0.42)(1.58)(1.96)(−1.87)(1.74)
Pseudo R20.090.090.110.110.13
χ291.8494.19106.8106.8102.0
N2,6902,6902,4432,4432,625
Panel B: The mediating role of the firm characteristics
 SIZENEG- EARNINGSCAPEXXDOPSBANKRUPTCY PROBABILITYCURRENT ASSETSCROSS- LISTINGLEVERAGEFINANCING
Models123456789
SIZE × CONNECTED0.14**        
 (2.03)        
NEG_EARNINGS × CONNECTED −0.76**       
  (−2.28)       
CAPEX × CONNECTED  0.09**      
   (2.06)      
XDOPS × CONNECTED   0.73**     
    (2.16)     
BANKRUPTCY PROBABILITY × CONNECTED    0.02*    
     (1.86)    
CURRENT ASSETS × CONNECTED     0.35***   
      (2.59)   
CROSS-LISTING × CONNECTED      0.62  
       (0.60)  
LEVERAGE × CONNECTED       −0.01 
        (−1.48) 
FINANCING × CONNECTED        −0.17
         (−0.31)
Intercept−2.72**−2.76**−2.56**−2.71**−3.15**−2.73**−2.78**−2.88**−2.81**
 (−2.31)(−2.38)(−2.22)(−2.35)(−2.54)(−2.35)(−2.41)(−2.49)(−2.44)
[Marginal effect of #VARIABLES × CONNECTED in %][2.07][−13.99][1.33][9.81][0.33][5.26][7.73][−0.18][−2.71]
Mean interactive effect of #VARIABLES × CONNECTED0.01−0.11*0.01*0.10*0.000.05**0.01−0.001−0.03
 (0.96)(−1.90)(1.85)(1.80)(1.62)(2.22)(0.06)(−1.20)(−0.39)
Pseudo R20.090.090.090.090.100.090.090.090.09
χ293.0693.0696.6692.4289.6493.5790.9491.8890.36
N2,6902,6902,6902,6902,3952,6902,6902,6902,690
Panel C: The mediating role of the country-level variables
 NEWSPAPERPOLITICALOPPELECTIONBANKDEPFINSYSDEP
  1. This table reports results on the role of ownership characteristics (panel A), other firm characteristics (panel B), and country-level variables (panel C) in conditioning the impact of political connections on auditor choice. In all specifications, we use model 3 in table 3 as the baseline regression. In panel A, the ownership variables interacted with CONNECTED are the wedge between the controlling shareholder's voting rights and cash flow rights (OWNERSHIPWEDGE) in model 1, a dummy variable for whether control rights of the ultimate owner exceed 50% (CONTROLRIGHTS>50) in model 2, the presence of a single large shareholder (SINGLELARGEOWN) in model 3, the number of multiple major shareholders (NUMBER OF LARGE OWNERS) in model 4, and a dummy variable indicating business group affiliation (BUSINESSGROUP) in model 5. In panel B, the firm characteristics interacted with CONNECTED are firm size (SIZE) in model 1, a dummy variable for loss firms (NEG_EARNINGS) in model 2, capital expenditures (CAPEX) in model 3, a dummy variable for firms reporting extraordinary items or discontinued operations (XDOPS) in model 4, the change in bankruptcy probability (BANKRUPTCY PROBABILITY) in model 5, current assets ratio (CURRENT ASSETS) in model 6, cross-listing status (CROSS-LISTING) in model 7, leverage (LEVERAGE) in model 8, and financing demand (FINANCING) in model 9. In panel C, the country-level variables interacted with CONNECTED are the circulation of daily newspapers divided by population (NEWSPAPER) in model 1, the strength of political opposition (POLITICALOPP) in model 2, a dummy variable indicating whether it is a national election year (ELECTION) in model 3, and the total value of bank (financial system) deposits to GDP in model 4 (model 5). Beneath each estimate is reported the robust z-statistic clustered at both the country and the firm level. The table also reports the marginal effect of each interaction term and the mean interactive effect using the methodology proposed by Ai and Norton [2003]. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively, based on two-sided tests. The definitions and data sources for the variables are outlined in table A1. Results for the control variables are not reported for the sake of brevity.

Models12345
NEWSPAPER × CONNECTED0.02    
 (0.07)    
POLITICALOPP × CONNECTED 2.27*   
  (1.82)   
ELECTION × CONNECTED  0.53*  
   (1.93)  
BANKDEP × CONNECTED   −0.01* 
    (−1.82) 
FINSYSDEP × CONNECTED    −0.01*
     (−1.86)
Intercept−1.71−2.39**−2.78**−1.01−0.99
 (−1.46)(−2.05)(−2.41)(−0.78)(−0.76)
      
[Marginal effect of #VARIABLES × CONNECTED in %][0.25][35.20][6.86][0.13][0.13]
Mean interactive effect of #VARIABLES × CONNECTED−0.050.230.06*−0.001*−0.001*
 (−1.58)(1.05)(1.77)(−1.75)(−1.78)
      
Pseudo R20.110.090.090.100.10
χ2109.797.2294.4883.3583.86
N2,5442,5442,6901,9881,988

Consistent with the prediction in H2, we find evidence in model 3 that connected firms with a single large shareholder are more likely to appoint a Big 4 auditor, implying that they rely more heavily on external monitoring by high-quality auditors in the absence of internal monitoring by multiple large shareholders. This evidence persists in model 4 when we replace the dummy variable representing the absence of multiple major shareholders with a continuous version of this conditioning variable, NUMBER OF LARGE OWNERS, which is the natural logarithm of one plus the number of shareholders holding at least a 10% equity stake. These results reinforce that committed monitoring by several large shareholders obviates the disciplinary role that Big 4 audits play in connected firms.

Similarly, the interaction between BUSINESSGROUP and CONNECTED loads positively in model 5, implying that Big 4 appointments become more likely when connected firms belong to a business group. We also estimate a positive and significant mean interactive effect (reported at the bottom of model 5). Our results square with Faccio's [2006] evidence that cumulative abnormal returns surrounding politicians joining firms’ boards are negative when the connected firm has a pyramidal ownership structure that increases agency costs since expropriation becomes more lucrative for the controlling shareholder as the gap between their voting and cash flow rights widens (e.g., Johnson et al. [2000], La Porta et al. [2002], Fan and Wong [2005]).

Collectively, the evidence in panel A of table 5 generally supports the intuition that connected firms are even more eager to engage high-quality auditors when their ownership structures leave minority investors more vulnerable to expropriation by dominant shareholders. Economically, our coefficient estimates translate into connected firms (i) with the ultimate owner's control rights exceeding 50%; (ii) with a single large shareholder; and (iii) affiliated with a business group becoming 12%, 15%, and 15% more likely to appoint a Big 4 auditor, respectively, with the rest of the regression variables set to their mean values.

In panel B of table 5, we complement the cross-sectional results involving ownership structure by analyzing whether the link between political connections and auditor choice hinges on other firm-level characteristics. More specifically, we follow prior research by examining the role that client size, profitability, complexity, financial constraints, and risk play in shaping connected firms’ demand for Big 4 auditors (e.g., Fan and Wong [2005], Hay, Knechel, and Wong [2006], Chang, Dasgupta, and Hilary [2009], Sankaraguruswamy, Whisenant, and Willenborg [2013]). Consistent with expectations, we report in models 1–6 evidence that connected firms that are larger have positive earnings in the past year, have larger capital expenditures, are more complex according to the presence of extraordinary items or discontinued operations, exhibit a greater change in bankruptcy probability, and have more current assets—firms with higher levels of inventory and receivables are harder to audit—are more likely to rely on a Big 4 auditor.18 Except for models 1 and 5, the mean interactive effects reported at the bottom of panel B are statistically significant. In contrast, we find no evidence in model 7 supporting that connected firms that are cross-listed tend to prefer higher-quality auditors. Similarly, in untabulated results that shift the focus from foreign financing to foreign operations, this interaction remains statistically insignificant when we replace cross-listing with the amount of foreign sales to gauge firms’ international orientation. In models 8 and 9, we also fail to find that connected firms with more debt in their capital structures or requiring more external financing exhibit greater demand for Big 4 auditors; both LEVERAGE and FINANCING are irrelevant to auditor choice according to table 3.

In the first three regressions in panel C of table 5, we examine the role of public scrutiny in shaping auditor choice in politically connected firms. This analysis is rooted in the intuition that connections increase the exposure of insiders to public scrutiny from the press and political opponents, reducing their ability and incentives to extract private benefits (Stulz [2005]). In short, subjecting connected firms to tough public scrutiny constrains their diversionary instincts. Consequently, these firms will prefer to appoint Big 4 auditors since they have nothing to hide when they refrain from self-dealing. In successive regressions, we interact CONNECTED with three variables reflecting the extent of media and political oversight, namely the circulation of daily newspapers divided by population (NEWSPAPER) from Dyck and Zingales [2004], the strength of political opposition (POLITICALOPP) obtained from the Database of Political Institutions (Beck et al. [2001]), and a dummy variable indicating whether it is a national election year (ELECTION) from the Database of Political Institutions. The results generally indicate that these variables condition the link between political connections and auditor choice in the predicted directions. More specifically, the interaction terms between CONNECTED and the variables measuring the strength of political opposition (model 2) and the presence of an election year (model 3) load positively, although we fail to find supportive results when we focus on the role of the press in model 1. The positive mean interactive effects reported in models 2 and 3 reinforce our conclusions. Next, we evaluate whether the countries’ debt market development mediates the link between political ties and auditor choice given Chaney, Faccio, and Parsley's [2011] evidence on the importance of political connections to firms’ borrowing costs. We measure debt market development with the total value of bank (financial system) deposits to GDP in model 4 (model 5). In both regressions, we find evidence implying that connected firms in countries with more developed debt markets are less likely to hire Big 4 auditors; the mean interactive effects are also negative and statistically significant.

4.3 DOES COUNTRY-LEVEL GOVERNANCE AFFECT THE LINK BETWEEN POLITICAL CONNECTION AND AUDITOR CHOICE?

Given prior research that the value of political connections is higher in weaker institutional environments, it follows that insiders resisting the temptation to expropriate outside investors will have even more incentive to improve external monitoring by hiring a Big 4 auditor in these countries. Consequently, we examine the prediction in H3 that the relation between political connections and auditor choice is stronger in countries with relatively lax governance institutions by re-estimating the regression in model 3 in table 3 after bisecting the sample into countries with weak versus strong governance according to the median rating of the various proxies described in section 'VARIABLES AND DESCRIPTIVE STATISTICS'. 19

Table 6 presents the results after dividing the sample using the four proxies for country-level governance, RESTRICTIONS, EXPROPRIATION, ANTISELF, and PROTECTION. We find across all proxies that the coefficient on CONNECTED is positive and statistically significant in the subsample of firms located in countries with weak governance institutions (models 1, 3, 5, and 7), suggesting that political connections magnify the demand for high-quality audits in these situations.20 However, in stark contrast, we find that the coefficient on CONNECTED is statistically indistinguishable from zero in the subsample of countries with strong governance institutions (models 2, 4, 6, and 8). The difference in the CONNECTED coefficients between the samples of weak and strong institutional environments is statistically significant for two out of the four country-level governance variables; the exceptions are the comparisons involving RESTRICTIONS and EXPROPRIATION. In unreported regressions, we control for the largest shareholder's equity stake (LARGEOWN) rather than their control rights (CONTROLRIGHTS) in order to recover observations by improving the country coverage in table 6 to 28 from 19, and find that the CONNECTED coefficients are significantly different between the countries with weak and strong governance institutions in all four comparisons. These results lend support to the intuition that the incentives of politically connected firms to appoint a Big 4 auditor are stronger in countries with relatively poor governance institutions.

Table 6. Political Connections and Auditor Choice: The Mediating Role of Country-Level Institutions
VariableRESTRICTIONS LowRESTRICTIONS HighEXPROPRIATION HighEXPROPRIATION LowANTISELF LowANTISELF HighPROTECTION LowPROTECTION High
  1. For subsamples of weak versus strong governance countries, this table reports pooled logit estimation results for auditor choice for politically connected firms and nonconnected peers using model 3 in table 3 as the baseline regression. Models 1 and 2 report the results for subsamples of firms from countries with low and high regulatory scores (RESTRICTIONS), respectively. Models 3 and 4 report the results for subsamples of firms from countries with high and low risk of expropriation (EXPROPRIATION), respectively. Models 5 and 6 report the results for subsamples of firms from countries with low and high scores on the anti-self-dealing index (ANTISELF), respectively. Models 7 and 8 report the results for subsamples of firms from countries with low and high investor protection scores (PROTECTION), respectively. Beneath each estimate is reported the robust z-statistic clustered at both the country and the firm level. The table also reports the marginal effect of the variable CONNECTED and the p-value for the difference in the CONNECTED coefficient between the weak and strong institution subsamples. The standard errors for the differences between weak and strong institutions regressions are computed with a seemingly unrelated regression (SUR) system that estimates both groups jointly. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively, based on two-sided tests. The definitions and data sources for the variables are outlined in table A1.

Models12345678
CONNECTED0.60*0.420.71**0.180.87**−0.110.97***−0.23
 (1.80)(1.35)(2.08)(0.63)(2.53)(−0.37)(2.85)(−0.71)
CONTROLRIGHTS2.70*−2.00***−1.37**0.59−0.81−1.19*−1.64**−1.32*
 (1.73)(−3.66)(−2.32)(0.54)(−0.95)(−1.67)(−2.05)(−1.78)
COMPLEXITY0.15−0.08−0.100.050.050.040.050.01
 (1.28)(−0.78)(−1.01)(0.36)(0.50)(0.27)(0.47)(0.10)
FOREIGNSALES0.010.01**0.010.01**0.010.01−0.000.01
 (1.47)(1.99)(1.54)(2.11)(1.18)(1.64)(−0.06)(1.63)
FINANCING−0.220.390.64−0.250.320.390.270.36
 (−0.48)(0.99)(1.47)(−0.62)(0.70)(1.03)(0.63)(0.97)
CROSS-LISTING1.09*0.390.81*1.110.680.560.550.69
 (1.73)(0.84)(1.79)(1.26)(1.54)(0.49)(1.10)(0.59)
SIZE0.080.22***0.14*0.050.10**0.44***0.16***0.48***
 (1.47)(3.32)(1.94)(1.13)(2.14)(5.22)(3.03)(5.47)
STATEOWN−0.010.030.03−0.06−0.13*0.01−0.16***0.00
 (−0.12)(0.49)(0.55)(−0.79)(−1.90)(0.10)(−2.66)(0.05)
ROA3.20**1.76**2.20***2.36*1.202.63***1.132.41***
 (2.51)(2.07)(2.69)(1.82)(0.93)(2.83)(0.80)(2.70)
LEVERAGE0.10−0.170.28*−0.120.20−0.270.19−0.35
 (0.50)(−0.90)(1.67)(−0.59)(1.24)(−1.17)(1.23)(−1.57)
GROWTH0.00−0.00−0.010.00−0.00−0.01−0.00−0.01
 (0.16)(−0.66)(−1.25)(0.46)(−0.20)(−1.14)(−0.25)(−1.35)
INV−0.36−2.10*−1.88*−0.67−1.32−0.27−1.10−0.43
 (−0.27)(−1.94)(−1.70)(−0.52)(−1.06)(−0.21)(−0.82)(−0.34)
FDI−0.07**0.040.01−0.04−0.08**−0.00−0.05*−0.00
 (−2.28)(1.17)(0.28)(−1.46)(−2.52)(−0.11)(−1.81)(−0.04)
LGDPC0.52***−0.09−0.390.29**0.29**0.73***0.22*0.73***
 (3.51)(−0.40)(−0.31)(2.01)(2.56)(2.96)(1.84)(2.90)
SUE AUDITOR0.222.12***3.76***−0.452.82***−1.060.041.57
 (0.07)(3.23)(3.63)(−0.74)(3.31)(−0.87)(0.06)(0.51)
Intercept−6.30*−1.361.36−2.13−5.23***−10.15***−3.82**−12.28***
 (−1.87)(−0.53)(0.12)(−1.23)(−2.90)(−3.72)(−2.09)(−2.86)
[Marginal effect of CONNECTED in %][8.02][5.72][8.93][2.78][12.93][−1.17][14.33][−2.36]
p-value for difference in CONNECTED coefficient[0.69][0.23][0.03]**[0.01]***
Pseudo R20.090.160.170.060.140.160.120.17
χ254.6275.7288.7631.7276.4872.1770.5374.39
N1,1871,5031,6511,0391,4261,2641,4431,247

4.4 ECONOMIC OUTCOMES STEMMING FROM AUDITOR CHOICE IN POLITICALLY CONNECTED FIRMS

The analysis above suggests that politically connected firms are more likely to engage a Big 4 auditor, especially when they suffer more severe agency problems and operate in countries with relatively lax country-level governance institutions. This evidence naturally raises another question: why are these firms more eager to engage a Big 4 auditor? Extant research detects that firms with better auditors enjoy higher-quality earnings (e.g., Becker et al. [1998]), higher valuations (e.g., Fan and Wong [2005]), and lower cost of capital (e.g., Mansi, Maxwell, and Miller [2004]). However, there is evidence that connected firms have preferential access to credit (e.g., Khwaja and Mian [2005]) and are more likely to be bailed out by governments (Faccio, Masulis, and McConnell [2006]). In other words, some research implies that the upside of tough external monitoring by Big 4 auditors may be minimal for connected firms. In the other direction, the presence of a Big 4 auditor protects outside investors by constraining insiders’ discretion over the financial reporting process, which may translate into connected firms benefiting from lower information asymmetry. In this section, we contribute to empirically settling this issue by testing the prediction in H4 that politically connected firms appointing a Big 4 auditor practice less earnings management, enjoy greater transparency, exhibit higher valuations, and attract cheaper financing.

In table 7, we report the results of this analysis, which involves interacting auditor choice with the variable CONNECTED to isolate whether politically connected firms benefit more from becoming better known by hiring a Big 4 auditor. In focusing on the interaction term between CONNECTED and auditor choice, we examine an extensive set of firm-level outcome variables motivated by prior research, namely earnings management (e.g., Becker et al. [1998], Chaney, Faccio, and Parsley [2011]); analyst forecast coverage and accuracy (e.g., Lang, Lins, and Miller [2004], Lang, Lins, and Maffett [2012], Chen, Ding, Kim [2010]); accounting standard choice (e.g., Lang and Maffett [2011], Lang, Lins, and Maffett [2012]); valuation (e.g., Fan and Wong [2005]); and equity pricing (e.g., Khurana and Raman [2004]).

Table 7. Analysis of the Importance of Auditor Choice to Earnings Management, Transparency, Market Valuation, and Equity Pricing
(1) Earnings Management (EM)(2) Analyst Following (ANALYSTS)(3) Forecast Accuracy (ACCURACY)
Variable Variable Variable 
  1. This table reports results from regressing earnings management (model 1), the natural logarithm of one plus the number of analysts following the firm (model 2), analyst forecast accuracy (model 3), accounting standard choice (model 4), valuation (model 5), and equity pricing (model 6) on firm-level and country-level variables. The sample includes connected and nonconnected firms analyzed in table 3. Except for model 2, which includes year, industry, and country effects, all other models include industry and year effects. Beneath each estimate is reported the robust t/z-statistic clustered at both the country and the firm level. Model 4 reports the marginal effect of the interaction term and the mean interactive effect using the methodology proposed by Ai and Norton [2003]. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively, based on two-sided tests. The definitions and data sources for the variables are outlined in table A1.

PBIG4*CONNECTED−0.12***PBIG4*CONNECTED1.62**PBIG4*CONNECTED0.13***
 (−2.67) (2.44) (3.06)
PBIG40.05PBIG41.04PBIG40.01
 (1.45) (1.45) (0.37)
CONNECTED0.09***CONNECTED−1.42**CONNECTED−0.11***
 (2.74) (−2.52) (−2.98)
ROA−0.10***CROSS-LISTING0.13CROSS-LISTING−0.00
 (−4.14) (1.37) (−0.85)
SIZE−0.00SIZE0.27***SIZE−0.00*
 (−1.46) (10.20) (−1.74)
GROWTH−0.00LARGEOWN−0.04ANALYSTS0.00
 (−0.81) (−0.20) (1.20)
LEVERAGE0.01STATEOWN0.02NEG_EARNINGS0.08**
 (1.34) (1.59) (2.34)
LARGEOWN0.01GROWTH0.00CHANGE_EARNINGS−0.00
 (0.57) (1.32) (−0.66)
CROSS-LISTING−0.00VARIANCE0.23CIFAR−0.00
 (−0.43) (1.36) (−0.96)
STATEOWN−0.00EARNINGS_SURPRISE−1.69*ANTISELF−0.00
 (−0.36) (−1.82) (−0.17)
LAWORDER0.00Intercept−3.20***EARNINGS_MGN−0.00
 (0.17) (−5.88) (−0.64)
SECREG−0.00  LISTED FIRMS0.00
 (−0.34)   (0.31)
ANTISELF−0.05***  IDV−0.00
 (−2.96)   (−0.41)
STOCKTRAD−0.00  UAI0.00
 (−0.42)   (1.05)
Intercept0.14***  Intercept0.04
 (5.10)   (0.90)
Adjusted R20.09Adjusted R20.53Adjusted R20.09
N1,661N1,126N842
(4) Accounting Standard Choice (IAS)(5) Market Valuation (MTB)(6) Cost of Equity (KMED)
Variable Variable Variable 
PBIG4 × CONNECTED5.10**PBIG4*CONNECTED1.38**PBIG4*CONNECTED−5.97***
 (2.40) (2.23) (−2.73)
PBIG40.39PBIG40.72**PBIG43.15*
 (0.22) (2.26) (1.83)
CONNECTED−4.01**CONNECTED−1.01**CONNECTED4.77***
 (−2.25) (−2.05) (2.59)
LARGEOWN0.53SIZE−0.03**SIZE−0.32***
 (1.20) (−2.14) (−6.41)
FOREIGNSALES0.00GROWTH0.02***VARIANCE7.33***
 (1.14) (9.00) (10.24)
CF−0.00LARGEOWN−0.19FBIAIS9.32**
 (−0.50) (−1.16) (2.37)
FINANCING−0.05CROSS-LISTING0.13MTB−0.39***
 (−0.19) (0.87) (−8.37)
CROSS-LISTING0.15FOREIGNSALES0.00**DISCLOSE−1.49*
 (0.45) (2.22) (−1.82)
SIZE−0.25***CAPEX−0.00MACVAR8.55***
 (−4.01) (−0.50) (4.40)
LEVERAGE0.48***ANTISELF−0.12LAWORDER−0.22*
 (3.98) (−0.70) (−1.69)
TURNOVER−0.26*Intercept2.08***INFLATION0.45***
 (−1.82) (5.88) (2.97)
GROWTH0.01*  Intercept11.54***
 (1.72)   (4.77)
LGDPC0.26*    
 (1.81)    
ANTISELF−4.50***    
 (−7.71)    
STOCKTRAD0.21    
 (0.82)    
CIFAR0.07***    
 (3.47)    
Intercept−16.46***    
 (−6.87)    
[Marginal effect of PBIG4 × CONNECTED in %][44.71]    
Mean interactive effect of0.53*    
PBIG4 × CONNECTED(1.80)    
Pseudo R20.25Adjusted R20.13Adjusted R20.36
N2,993N3,119N1,124

We initially examine the impact of auditor choice on earnings management since providing evidence that the presence of a Big 4 auditor reduces information asymmetry evident in accounting transparency is arguably a necessary condition for proceeding to analyze, for example, whether Big 4 clients enjoy higher valuations and lower equity financing costs. After Fan and Wong [2005], we implement a two-stage estimation procedure. In the first stage, we predict, for each firm-year, the probability of choosing a Big 4 auditor using model 1 in table 3. In the second stage, we regress earnings management (EM)—specified after Leuz, Nanda, and Wysocki [2003] as the absolute value of accruals over total assets where accruals are calculated as: (Δtotal current assets – Δcash) – (Δtotal current liabilities – Δshort-term debt – Δtaxes payable) – depreciation expense (see table A1 for more details)—on the predicted probability of Big 4 auditors (PBIG4). Importantly, relying on Leuz, Nanda, and Wysocki [2003] specification minimizes data attrition in our sample. In addition to our test variables, we include several firm- and country-level controls in this regression: firm size (SIZE), cross-listing (CROSS-LISTING), profitability (ROA), firm growth (GROWTH), firm leverage (LEVERAGE), the ownership stake of the largest shareholder (LARGEOWN), state ownership (STATEOWN), a composite securities regulation index (SECREG), law and order (LAWORDER), stock market development (STOCKTRAD), and investor protection (ANTISELF). All variables are defined in table A1. The results in model 1 include that the coefficient on the interaction between PBIG4 and CONNECTED is negative and statistically significant at the 1% level, implying that earnings management is lower in politically connected firms with Big 4 auditors. In this regression, CONNECTED is positive and statistically significant at the 1% level, suggesting that earnings management is worse in politically connected firms. This result reconciles with Chaney, Faccio, and Parsley's [2011] evidence that politically connected firms exhibit lower earnings quality. Lending support to the prediction in H4, our evidence implies that this effect is less pronounced in politically connected firms that appoint a Big 4 auditor.

Table A1. Variables, Definitions, and Sources
VariableDefinitionSource
Firm-Level Variables
CONNECTEDA dummy variable equal to one for politically connected firms, and zero otherwise.Faccio [2006]
BIG 4A dummy variable equal to one for firms with Big 4 auditors, and zero otherwise.Worldscope
EMEqual to |Accruals|/Total Assets.Worldscope
 We compute the accrual component of earnings as Accrualsit = (ΔCAitΔCashit) – (ΔCLitΔSTDitΔTPit) − Depit, where ΔCAit = change in total current assets, ΔCashit = change in cash/cash equivalents, ΔCLit = change in total current liabilities, ΔSTDit = change in short term debt included in current liabilities, ΔTPit = change in income taxes payable, and Depit = depreciation and amortization expense for firm i in year t. If a firm does not report information on taxes payable or short-term debt, then the change in both variables is assumed to be zero. 
LARGEOWNThe largest shareholder's equity stake.Worldscope
CONTROLRIGHTSThe voting rights of the largest shareholder.Claessens et al. [2000] and Faccio and Lang [2002]
CASHFLOWRIGHTSThe cash flow rights of the largest shareholder.Claessens et al. [2000] and Faccio and Lang [2002]
SINGLELARGEOWNA dummy variable equal to one if there is only a single large shareholder that controls at least 10% of the voting rights, and zero otherwise.Claessens et al. [2000] and Faccio and Lang [2002]
BUSINESSGROUPA dummy variable equal to one if the firm belongs to a business group, and zero otherwise.Claessens et al. [2000], Faccio and Lang [2002], and authors’ calculations
OWNERSHIPWEDGEA dummy variable equal to one if the largest shareholder's share of control rights is higher than his share of ownership rights, and zero otherwise.Claessens, Djankov, and Lang [2000] and Faccio and Lang [2002]
CONTROLRIGHTS>50A dummy variable equal to one if the voting rights of the largest shareholder is greater than 50%, and zero otherwise.Claessens et al. [2000] and Faccio and Lang [2002]
Log(1+NUMBER OF LARGE OWNERS)The natural logarithm of one plus the number of shareholders that control more than 10% of the voting rights.Claessens et al. [2000] and Faccio and Lang [2002]
STATEOWNThe state's equity stake in percentage.Worldscope
MTBMarket value of common equity plus book value of debt divided by total assets at the end of year.Worldscope
PBIG4Predicted probability that the auditor is a Big 4 public accounting firm.Authors’ calculations
CAPEXThe ratio of capital expenditures divided by total assets.Worldscope
GROWTHAsset growth in the past year.Worldscope
LEVERAGEThe ratio of long-term debt divided by total equity.Worldscope
SIZEThe natural logarithm of total assets denominated in US dollars.Worldscope
COMPLEXITYThe number of business segments based on two-digit SIC codes.Worldscope
FOREIGNSALESThe fraction of sales from foreign operations.Worldscope
FINANCINGA dummy variable that takes the value of one if the sum of new long-term debt plus new equity exceeds 20% of total assets, and zero otherwise.Worldscope
CROSS-LISTINGA dummy variable that is equal to one if the firm is cross-listed in the United States, and zero otherwise.Worldscope
INVTotal inventory divided by total assets.Worldscope
ROANet income divided by total assets.Worldscope
XDOPSA dummy variable that takes the value of one if the company reports extraordinary items or discontinued operations, and zero otherwise.Worldscope
BANKRUPTCYOne-year change in Zmijewski's probability of bankruptcy score.Worldscope
PROBABILITYz = –4.336–4.513*(net income/total assets)+5.679*(total liabilities/total assets) + 0.004*(current assets/current liabilities). 
CURRENT ASSETSThe ratio of current assets to total assets.Worldscope
ANALYSTSThe number of analysts averaged over the fiscal year.Authors’ calculations based on I/B/E/S
NEG_EARNINGSAn indicator variable for loss firms.Worldscope
CHANGE_EARNINGSThe absolute value of the change in earnings over the previous year scaled by the previous year's earnings.Worldscope
EARNINGS_SURPRISEThe absolute value of the difference between current earnings per share and earnings per share from the prior year, divided by the firm's stock price.Authors’ calculations based on CRSP data.
CFAnnual net cash flow from operating activities divided by end-of-year total assets.Worldscope
TURNOVERSales divided by end-of-year total assets.Worldscope
FBIAISSigned forecast error defined as the difference between the one-year-ahead consensus earnings forecast and realized earnings deflated by beginning of period assets per share.Authors’ calculations based on I/B/E/S
VARIANCEVolatility of stock returns over the previous 12 months.Authors’ calculations based on Compustat, CRSP, and CFRMC data
IASA dummy variable equal to one for firms that adopted international accounting standards, and zero otherwise.Authors’ calculations based on Compustat data
ACCURACYThe negative of the absolute difference between actual EPS and analysts’ forecasts scaled by stock price.Authors’ calculations based on I/B/E/S
KGLSImplied cost of equity capital estimated from the Gebhardt, Lee, and Swaminathan [2001] model 10 months after the fiscal year-end.Authors’ calculations based on I/B/E/S and Compustat data
KCTImplied cost of equity capital estimated from the Claus and Thomas [2001] model 10 months after the fiscal year-end.Authors’ calculations based on I/B/E/S and Compustat data
KOJNImplied equity premium capital estimated from the Ohlson and Juttner-Nauroth [2005] model 10 months after the fiscal year-end.Authors’ calculations based on I/B/E/S and Compustat data
KMPEGImplied cost of equity capital estimated from the Easton [2004] model 10 months after the fiscal year-end.Authors’ calculations based on I/B/E/S and Compustat data
KMEDThe median of KGLS, KCT, KOJN, and KMPEG.Authors’ calculations based on I/B/E/S and Compustat data
Country-Level Variables
INFLATIONRealized inflation rate over the next year.World Development Indicators [2008]
OPENNESSImports plus exports as fraction of GDP.World Development Indicators [2008]
NEWSPAPERCirculation of daily newspapers divided by population.Dyck and Zingales [2004]
ELECTIONA dummy variable equal to one for election years, and zero otherwise.Database of Political Institutions
POLITICALOPPOThe strength of the opposition. High values reflect strong opposition.Database of Political Institutions
PROTECTIONThe principal component of the indices for anti-director rights, disclosure requirements, and liability standards.La Porta, Lopez-de-Silanes, and Shleifer [2006]
ANTISELFAverage of ex-ante and ex-post private control of self-dealing.Djankov et al. [2008]
SECREGStrength of securities regulation. Equals the arithmetic mean of: (i) disclosure index, (ii) liability standard index, and (iii) public enforcement index.La Porta, Lopez-de-Silanes, and Shleifer [2006]
LAWORDERThe law and order in the country.International Country Risk Guide
STOCKTRADStock market total value traded divided by GDP.Beck, Demirguc-Kunt, and Levine [2009]
BANKDEPThe total value of demand, time, and saving deposits at domestic deposit money banks as a share of GDP.International Financial Statistics
FINSYSDEPDemand, time, and saving deposits in deposit money banks and other financial institutions as a share of GDP.International Financial Statistics
JUDICIALAssessment of the efficiency and integrity of the legal environment as it affects business, particularly foreign firms, produced by the country risk rating agency Business International Corp. It may be taken to represent investors’ assessment of conditions in the country in question. Average between 1980 and 1983. The scale ranges from 0 to 10 with lower scores representing lower efficiency levels.La Porta, Lopez-de-Silanes, and Shleifer [2006]
RESTRICTIONSA regulatory score constructed based on regulations that prohibit or set limits on the business activities of public officials.Faccio [2006]
CIFARIndex created by examining and rating companies’ 1995 annual reports on their inclusion or omission of 90 items. These items fall into seven categories: general information, income statements, balance sheets, funds flow statement, accounting standards, stock data, and special items. A minimum of three companies in each country were studied. See Bushman, Piotroski, and Smith [2004].La Porta, Lopez-de-Silanes, and Shleifer [2006]
RULEOFLAWMeasures the extent to which agents have confidence in and abide by the rules of society. These include perceptions of the incidence of both violent and nonviolent crime, the effectiveness and predictability of the judiciary, and the enforceability of contracts. See Kaufmann, Kraay, and Mastruzzi [2008].La Porta et al. [2006]
LISTED FIRMSThe number of domestic listed firms divided by population in 2000. A measure of the importance of the stock market.La Porta, Lopez-de-Silanes, and Shleifer [2006]
MACVARMACVAR is the first principal component of four proxies for macroeconomic variability: (i) the country-year median standard deviation of annual earnings per share over the last five years scaled by total assets per share, (ii) the country-year median standard deviation of accounting returns on equity over the last five years, (iii) the standard deviation of the residuals from a regression of annual gross domestic product growth rates on a time index over the sampling period, and (iv) the coefficient of variation of yearly average exchange rates (US$ to local currency) over the sampling period.Authors’ calculations
IDVA preference for a loosely knit social fabric or an independent, tightly knit fabric.Hofstede [2001]
UAIThe degree to which people feel uncomfortable with ambiguity and an uncertain future.Hofstede [2001]
PREDATIONAn index that incorporates multiple attributes capturing the effectiveness of institutional and political systems in curbing government extortion.Durnev and Fauver [2009]
EXPROPRIATIONAssessment of the risk of a modification in a contract taking the form of a repudiation, postponement, or scaling down due to budget cutbacks, indigenization pressure, a change in government, or a change in government economic and social priorities. This variable is recoded to vary between 0 to 10 with higher scores indicating greater risk of expropriation.La Porta et al. [1998]
EARNINGS_MGNAggregate earnings management score: the average rank across four measures, EM1–EM4. EM1 is the country's median ratio of the firm-level standard deviations of operating income and operating cash flow (both scaled by lagged total assets). EM2 is the country's Spearman correlation between the change in accruals and the change in cash flow from operations (both scaled by lagged total assets). EM3 is the country's median ratio of the absolute value of accruals and the absolute value of the cash flow from operations. EM4 is the number of ‘‘small profits’’ divided by the number of ‘‘small losses’’ for each country.Leuz, Nanda, and Wysocki [2003]
DISCLOSEAn assessment of disclosure requirements relating to: (i) prospectus; (ii) compensation of directors and key officers; (iii) ownership structure; (iv) inside ownership; (v) contracts outside the ordinary course of business; and (vi) transactions between the issuer and its directors, officers, and/or large shareholders. The index ranges from 0 to 1, with higher values indicating more extensive disclosure requirements.La Porta, Lopez-de-Silanes, and Shleifer [2006]
SUE AUDITORIndex of the procedural difficulty in recovering losses from the auditor in a civil liability case for losses due to misleading statements in the audited financial information accompanying the prospectus. Equals one when investors are only required to prove that the audited financial information accompanying the prospectus contains a misleading statement. Equals two-thirds when investors must also prove that they relied on the prospectus and/or that their loss was caused by the misleading accounting information. Equals one-third when investors must also prove that the auditor acted with negligence. Equals zero if restitution from the auditor is either unavailable or the liability standard is intent or gross negligence.La Porta, Lopez-de-Silanes, and Shleifer [2006]
LGDPCThe natural logarithm of the country's GDP per capita denominated in US dollars.World Development Indicators [2008]
FDIForeign direct investment divided by GDP.World Development Indicators [2008]
Table A2. Means for Country Level Institutional Infrastructure Variables
CountryRESTRICTIONSEXPROPRIATIONANTISELFPROTECTION
  1. This table reports the means for the four conditioning variables capturing countries’ governance institutions.

Austria2.000.310.210.10
Belgium0.000.370.540.07
Canada2.000.330.640.96
Chile2.002.500.630.61
Denmark1.000.330.460.36
Finland1.000.330.460.47
France2.000.350.380.47
Germany2.000.100.280.00
Greece4.002.880.220.32
Hong Kong1.001.710.960.85
India0.002.250.580.77
Indonesia0.002.840.650.51
Ireland4.000.330.790.48
Israel4.001.750.730.59
Italy0.000.650.420.20
Japan0.000.330.500.42
Korea, South1.001.690.470.36
Malaysia0.002.050.950.73
Mexico0.002.710.170.10
Philippines6.004.780.220.81
Singapore3.000.701.000.77
Spain3.000.480.370.55
Sweden1.000.600.330.39
Switzerland2.000.020.270.30
Taiwan0.000.880.570.55
Thailand3.002.580.810.37
United Kingdom2.000.290.950.78
United States4.000.020.651.00
Table A3. Correlations Between the Main Regression Variables
 BIG 4CONNECTEDLARGEOWNSTATEOWNCOMPLEXITYFOREIGNSALESFINANCINGCROSS-LISTINGSIZEROALEVERAGEGROWTHINVFDILGDPC
  1. This table reports correlations for the regression variables while allowing for country and firm level clustering for a sample of 1,371 politically connected firms and 1,911 nonconnected firms from 28 countries. Boldface indicates statistical significance at the 1% level. The definitions and data sources for the variables are outlined in table A1.

CONNECTED0.06              
LARGEOWN0.000.01             
STATEOWN–0.13–0.040.15            
COMPLEXITY–0.020.050.020.08           
FOREIGNSALES0.110.040.050.05–0.02          
FINANCING–0.04–0.020.01–0.030.060.00         
CROSS-LISTING0.090.100.11–0.020.000.13–0.03        
SIZE0.110.040.060.110.150.160.130.16       
ROA0.110.020.080.010.040.040.040.020.11      
LEVERAGE0.020.110.00–0.030.040.030.070.050.200.05     
GROWTH0.05–0.040.020.030.080.000.42–0.020.000.240.00    
INV–0.040.090.040.050.090.080.060.070.100.090.14–0.04   
FDI0.060.030.02–0.030.090.190.080.030.390.03–0.030.070.02  
LGDPC0.150.230.270.230.080.19–0.030.060.190.06–0.040.070.050.08 
SUE AUDITOR0.070.080.240.010.060.05–0.02–0.01–0.03–0.020.15–0.040.080.010.16

We extend our analysis to consider three measures of firm-level transparency examined in recent research (e.g., Lang and Maffett [2011], Lang, Lins, and Maffett [2012]). This includes analyzing in model 2 the number of analysts following a firm (ANALYSTS). Lang, Lins, and Maffett [2012] hold that greater analyst coverage and forecast accuracy are likely to reflect greater transparency in the firm's information environment. Besides our test variables, we include a set of controls motivated by prior research (e.g., Lang, Lins, and Miller [2003, 2004]): firm size (SIZE); cross-listing (CROSS-LISTING); the ownership stake of the largest shareholder (LARGEOWN); state ownership (STATEOWN); firm growth (GROWTH); the standard deviation of monthly returns over the previous year (VARIANCE); earnings surprise (EARNINGS_SURPRISE); as well as industry, year and country effects. We find a negative and statistically significant (at the 5% level) coefficient on CONNECTED, indicating that connected firms have lesser analyst coverage. In additional evidence supporting the prediction in H4, we provide evidence at the 5% level that this effect subsides for politically connected firms that are Big 4 clients.

In model 3, we complement the evidence in model 2 by analyzing forecast accuracy (ACCURACY), defined as the negative absolute value of the difference between the median forecast and the actual earnings per share (EPS) deflated by the stock price after Lang and Lundholm [1996] and Hope [2003]. We follow Hope's [2003] cross-country research by including these control variables in the regression: firm size (SIZE); analyst following (ANALYSTS); cross-listing (CROSS-LISTING); an indicator variable for loss firms (NEG_EARNINGS); the absolute value of the change in earnings over the previous year (CHANGE_EARNINGS); as well as country-level measures for earnings management (EARNINGS_MGN), the quality of financial reporting (CIFAR), domestic listed firms (LISTED FIRMS), investor protection (ANTISELF), Hofstede's [2001] cultural dimensions uncertainty avoidance (UAI) and individualism (IDV), along with industry and year effects. Consistent with Chen, Ding, and Kim. [2010], the results in model 3 suggest that political connections adversely affect analyst forecast accuracy. Importantly for our purposes, we find that the coefficient on the interaction term PBIG4*CONNECTED is positive and statistically significant at the 1% level, suggesting that connected firms that appoint a Big 4 auditor benefit from more accurate analyst earnings forecasts.

In model 4, we develop another conditioning variable by identifying whether a firm adopts international accounting standards (IAS). Accounting quality is generally higher for these firms according to Barth, Landsman, and Lang [2008] and Lang and Maffett [2011]. In this regression, we follow prior research (e.g., Hope, Jin, and Kang [2006], Barth, Landsman, and Lang [2008], Kim and Shi [2012]) by controlling for firm size (SIZE), the ownership stake of the largest shareholder (LARGEOWN), firm growth (GROWTH), firm leverage (LEVERAGE), cash flows from operating activities (CF), cross-listing (CROSS-LISTING), sales turnover (TURNOVER), level of financing activities (FINANCING), foreign sales (FOREIGNSALES), the logarithm of GDP per capita (LGDPC), investor protection (ANTISELF), stock market development (STOCKTRAD), and a country's quality of financial reporting (CIFAR) as well as industry and year fixed effects. Reinforcing our earlier evidence, the results indicate that the coefficient on CONNECTED is significantly negative while the mean interactive effect of PBIG4*CONNECTED loads positively at the 5% level. Collectively, these results support that the presence of a high-quality auditor mitigates the lower transparency of politically connected firms.

We consider in model 5 the role that auditor choice plays in shaping firm value, measured by the market-to-book ratio (MTB). In this regression, we include other factors shown in prior research to affect firm value (e.g., Durnev and Kim [2005]): firm size (SIZE), cross-listing (CROSS-LISTING), firm growth (GROWTH), the ownership stake of the largest shareholder (LARGEOWN), foreign sales to total assets (FOREIGNSALES), capital expenditures (CAPEX), and investor protection (ANTISELF) in addition to year and industry fixed effects. We report that the coefficient on the interaction between PBIG4 and CONNECTED loads positively at the 5% level, implying that connected firms that appoint a Big 4 auditor are valued at a premium. This finding extends evidence in Fan and Wong [2005] that the economic consequences of the presence of a high-quality auditor are larger for East Asian firms with highly concentrated control.

Recent research suggests that accounting transparency at the country- (e.g., Hail and Leuz [2006]) and firm-level (e.g., El Ghoul, Guedhami, and Pittman [2012]) is associated with a lower cost of equity capital. Accordingly, we extend our analysis in model 6 by gauging the links among political connections, auditor choice, and equity pricing. We follow extensive prior research by using analysts’ earnings forecasts and stock prices to derive the ex ante cost of equity (e.g., Hail and Leuz [2006], Dhaliwal, Heitzman, and Li [2006], Chen, Chen, and Wei [2011]). This approach constitutes a useful alternative given the failure of asset pricing models to proxy for the cost of equity (e.g., Fama and French [1997], Pástor, Sinha, and Swaminathan [2008]). More specifically, we adopt four implied cost of equity capital models, namely those developed by Gebhardt, Lee, and Swaminathan ([2001]; KGLS), Claus and Thomas ([2001]; KCT), Ohlson and Juettner-Nauroth ([2005]; KOJN), and Easton ([2004]; KES). To reduce concerns that the results are driven by the assumptions underpinning any particular model, we specify as the dependent variable the median implied cost of equity obtained from the four models (KMED). Moreover, to alleviate the concern that optimism inherent in analysts’ earnings forecasts admits bias that inflates the equity premium estimates, we follow Hail and Leuz [2006] by running a weighted least square regression that assigns less (more) weight to inaccurate (precise) forecasts. In this regression, the weight is equal to the forecast error (absolute value one-year ahead earnings forecast minus realized earnings deflated by assets per share).

In addition to our test variables, we include several control variables: the natural logarithm of total assets (SIZE), the market value of common equity plus book value of debt scaled by total assets (MTB), forecast error defined as the difference between the one-year-ahead earnings forecast and realized earnings deflated by beginning of the period assets per share (FBAIS), the volatility of stock returns over the previous 12 months (VARIANCE), the realized inflation rate over the next year (INFLATION), disclosure standards (DISCLOSE), law and order (LAWORDER), and macroeconomic variability (MACVAR). In model 6, which also includes year and industry fixed effects, we observe a negative and statistically significant (at the 1% level) coefficient on the interaction between PBIG4 and CONNECTED, implying that connected firms that appoint a Big 4 auditor attract cheaper equity financing.

Notwithstanding that the implied cost of capital approach has been widely used in the accounting and finance literature (see, e.g., Hail and Leuz [2006], Chen, Chen, and Wei [2011]), it is beset by another limitation apart from deviations between analysts’ and investors’ earnings expectations. This limitation relates to the sensitivity of the implied cost of equity estimates to the assumptions of long-term growth rates beyond analysts’ forecast horizons (e.g., Easton [2004]). In particular, the cost of equity estimates derived from the models of Claus and Thomas [2001] and Ohlson and Juettner-Nauroth [2005] assume that the perpetual growth rate is equal to the future (realized) inflation rate. To verify that our findings in model 6 are robust to addressing this concern, we follow Hail and Leuz [2006] by re-estimating the median cost of equity capital (KMED) after applying these alternative growth assumptions for the models of Claus and Thomas [2001] and Ohlson and Juettner-Nauroth [2005]: (i) a constant long-run growth of 3% and (ii) a perpetual growth rate equal to the annual real GDP growth plus long-run inflation rate, respectively. The untabulated results include that the PBIG4*CONNECTED coefficient is negative and statistically significant at the 5% level, corroborating our earlier evidence that politically connected firms with Big 4 auditors enjoy equity financing costs that are closer to the risk-free rate. Overall, the evidence reported in this section provides empirical support for our fourth hypothesis that politically connected firms with Big 4 auditors exhibit lower earnings management and greater transparency, and benefit from higher valuations as well as cheaper equity financing costs.

5. Conclusions

In response to calls for research on this issue (e.g., Wang, Wong, and Xia [2008]), we examine the importance of corporate insiders’ political connections, which exacerbate the expropriation of minority investors according to recent evidence (e.g., Faccio [2006], Qian, Hongbo Pan, and Yeung [2011]), to auditor choice in public firms worldwide. The tension that connected insiders experience in deciding whether to appoint a Big 4 auditor to constrain their discretion over financial reporting motivates our analysis. In one direction, insiders eager to persuade outside investors that they are not exploiting their political connections to divert corporate resources may rely on a Big 4 auditor to strengthen external monitoring. In the other direction, connected insiders depriving outside investors may prefer to hire a non–Big 4 auditor to help conceal their diversion by rendering the financial statements less informative about underlying firm performance. Using a unique data set on political connections around the world constructed by Faccio [2006], we analyze which financial reporting incentive dominates by estimating the role that political connections play in auditor choice. Our evidence that public firms with political connections are more likely to appoint a Big 4 auditor lends support to the intuition that these firms respond to the serious agency problems that connections engender by improving accounting transparency evident in auditor choice.

Next, we separately isolate whether connected firms with ownership structures conducive to self-dealing by insiders or operating in countries with relatively poor institutional infrastructure are even more eager to reduce information asymmetry by engaging a Big 4 auditor. In a set of results consistent with these predictions, we find that the link between auditor choice and political connections is stronger when firms' ownership characteristics lead to severe agency conflicts and country-level governance institutions are worse. This evidence implies that Big 4 auditors in these situations are more valuable for protecting outside investors by disciplining insiders against diverting corporate resources.

Finally, we consider some economic outcomes stemming from auditor choice in politically connected firms. This analysis contributes to resolving what is behind our evidence that connected firms have greater demand for Big 4 auditors. We begin by documenting that connected firms that appoint Big 4 auditors exhibit lower earnings management and greater transparency evident in analyst coverage and forecast accuracy as well as the adoption of international accounting standards. Additional evidence implies that the capital markets reward connected firms that are Big 4 clients with higher valuations and cheaper equity financing costs. Importantly, despite that our core results persist when we tackle this issue with several standard techniques—including various matching procedures, estimating a treatment effects model, controlling for firm heterogeneity in random effects models, and narrowing our analysis to firms with long auditor tenure—we stress that we cannot dismiss endogeneity as a competing explanation for our evidence.

  1. 1

    Reinforcing the significance of our research, politically connected firms account for almost 8% of the world's stock market capitalization (Faccio [2006]).

  2. 2

    Although Arthur Andersen dissolved during our study period, we follow convention by labeling the large brandname public accounting firms and their predecessors as the Big 4. Choi and Wong [2007] review worldwide evidence implying that the Big 4 supply higher-quality audits, reinforcing the U.S. evidence that Francis [2004] comprehensively surveys.

  3. 3

    To provide some perspective on the economic magnitude of our evidence, Fan and Wong [2005] estimate that raising the fraction of voting (cash flow) rights belonging to the ultimate owner one standard deviation from its mean value leads to hiring a Big 4 auditor becoming 5% (2%) more likely.

  4. 4

    These political benefits include preferential treatment in the form of access to credit from state-owned banks (e.g., Dinç [2005]); the receipt of government contracts (e.g., Agrawal and Knoeber [2001]); corporate bailouts in the event of financial distress (e.g., Faccio, Masulis, and McConnell [2006]); lower tax burdens (e.g., Adhikari, Derashid, and Zhang [2006]); lax regulatory enforcement (e.g., Berkman, Cole, and Fu [2010]); and greater allocation of government investment during financial crises (e.g., Duchin and Sosyura [2012]).

  5. 5

    However, evidence from a single country may not generalize elsewhere. For example, although there is extensive research on political connections in Indonesia, Faccio, Lang, and Young [2001] caution against extrapolating inferences from there given its weak capital market institutions and poor corporate transparency. Rather than the unique conditions in play when focusing on a single country, we analyze the importance of connections worldwide to auditor choice to justify more general insights.

  6. 6

    Additionally, Fan and Wong (2005) find that Big 4 auditors improve corporate governance in East Asia, a region where there is hardly any implicit insurance coverage available to investors in the event of audit failure. In another study on this low-litigation region, Mitton [2002] finds evidence corroborating that investors perceive that Big 4 auditors improve accounting transparency in East Asian firms.

  7. 7

    The collapse of Parmalat, the Italian dairy-and-food conglomerate, in 2003 illustrates the interplay between ownership structure, political connections, and insiders disguising their diversion of corporate resources by manipulating financial reporting. The company, which remained closely controlled by the Tanzi family through a pyramidal ownership structure after it went public in 1990, was accused of deliberately exaggerating its earnings. The company's founder and CEO, Calisto Tanzi, who admitted to diverting about $640 million from Parmalat to his family's companies (Sylvers [2004]), was eventually convicted and sentenced. Other insiders later confessed to depriving outside investors by tunneling “commissions” to offshore companies that they controlled (e.g., Gumbel [2004]). After recounting how Mr Tanzi developed connections with politicians, Ferrarini and Giudici [2005, p. 12] argue that, “Parmalat reveals some features common to firms that have faced catastrophic financial failures: massive growth, questionable accounting and accountants, poor underlying performance, political connections, a dominating shareholder, complex corporate structures and operational mystery.”

  8. 8

    However, all of our inferences hold when we extend our analysis to cover the entire set of 47 countries studied in Faccio [2006]. In this expanded sample, politically connected firms represent almost 2% of the total firm-year observations, which is comparable to the 2.7% reported in Faccio's [2006] full data set.

  9. 9

    After prior cross-country research analyzing years surrounding its collapse (e.g., Francis and Wang [2008]), we include Arthur Andersen in our main specification of the Big 4 auditors. However, our core results remain when we remove all Andersen clients from the estimations. Similarly, none of our inferences are sensitive to also excluding the successor auditors (whether Big 4 or non–Big 4) to Andersen after its dissolution in 2002. Another issue is the impact on our evidence of potential Worldscope auditor coding errors, which might be evident in low Big 4 audit market shares in some of the countries represented in our sample. We help mitigate this concern by removing in successive regressions firms from countries in which the Big 4 market share is below 25% (this involves excluding firms from India and Greece) and 50% (this involves excluding firms from India, Greece, France, and Mexico). Our evidence supporting the prediction in H1 holds at the 5% level or better in these smaller samples. Also, we perform spot checks on 10% of the Indian, Greek, French, and Mexican firms in our sample, which confirms without exception that auditor identity in Worldscope matches that in firms’ annual reports. In any event, auditor coding errors, which inject noise into the analysis, would likely work against our tests rejecting the null hypotheses.

  10. 10

    None of our core results are materially sensitive to replacing SIZE with the natural logarithm of sales or market capitalization. Given that both BIG 4 and POLITICAL are positively correlated with SIZE according to table A3, we help reduce the concern that firm size is spuriously responsible for our evidence as follows. After Feltham, Hughes, and Simunic [1991] and Pittman and Fortin [2004], we re-estimate our regressions after isolating a size-truncated sample comprised of all firms ranging between the smallest Big 4 client according to assets and the largest non–Big 4 client. In this restricted sample, which helps address whether our results reflect pervasive economic phenomena rather than very large firms dominating the analysis, the evidence on our predictions persists at the 5% level or better.

  11. 11

    In another strategy for isolating the importance of political connections to auditor choice, we assemble a third peer group consisting of nonconnected peers matched to connected firms based on country, industry, and year. In unreported estimations, we find evidence at the 1% level supporting the prediction in H1 when we apply this matching technique.

  12. 12

    Reflecting the importance of political connections to auditor choice, the incremental R2 for CONNECTED is 0.037 relative to 0.027 for SIZE.

  13. 13

    Although we follow extensive prior research in primarily specifying auditor quality with the presence or absence of a Big 4 auditor, we also examine whether our results are materially sensitive to replacing BIG 4 with firms’ audit fees. Extant research implies that audit fees are higher when auditors expend more effort on an engagement, translating into better audits (e.g., Dye [1993], Davis, Ricchiute, and Trompeter [1993], Whisenant, Sankaraguruswamy, and Raghunandan [2003], Caramanis and Lennox [2008]). Consequently, politically connected firms will incur higher audit fees under the intuition for the prediction in H1. However, given poor audit fee data availability in Worldscope, the sample shrinks to 1,756 firm-year observations compared to the 3,282 observations under study in model 1 in table 3. Nevertheless, we find in unreported results that CONNECTED continues to load positively at the 1% level despite the loss in power in this smaller sample, suggesting that connected firms pay higher audit fees than their nonconnected peers matched by country, industry, year, and size.

  14. 14

    Our empirical design primarily diverges from Fan and Wong's [2005] in two ways. First, we control for a larger set of firm-level determinants of auditor choice. In fact, the 14 controls for firm characteristics in our models include the four in their main specifications. Second, we apply various matching techniques to improve identification in our setting. After restricting our sample to the eight Asian countries analyzed in Fan and Wong [2005] and no longer examining matched samples, we find a positive and statistically significant coefficient on CONTROLRIGHTS, helping to reconcile our evidence to theirs. More generally, our results are consistent with El Ghoul et al. [2013], who find that Fan and Wong's [2005] evidence for East Asia does not extend to Western Europe where investor protection institutions are typically better.

  15. 15

    Although the treatment effects model has the advantage of analyzing the entire sample (e.g., Chaney, Faccio, and Parsley [2011]), an alternative approach to addressing endogeneity in order to improve identification is to analyze political connection formation over time and observe the ensuing effects on auditor choice. However, Chaney, Faccio, and Parsley [2011] stress that a major limitation of this approach is the small sample of firms for which the date of political connection formation can be determined, preventing researchers from using formation dates to draw meaningful inferences. In our setting, we could only pinpoint the date of formation of the political connection for 44% of our sample. Regrettably, the vast majority of these connections were forged before 2001 when our sample period begins, precluding us from reliably examining the link between changes in political connections and auditor choice. Moreover, Faccio's [2006] database does not disclose the connected member, which complicates tracking shifts in political affiliations over our 2001–2005 sample period.

  16. 16

    Consistent with Bockus and Gigler's [1998] theory, Shu [2000] reports that riskier clients are less likely to retain another large auditor when their incumbent auditor resigns from the engagement. Although prior research implies that large audit firms in the United States are more eager to avoid riskier clients since they have more to lose in the form of reputational and litigation costs in the event of audit failure, this evidence is sensitive to shifts in the auditor litigation liability landscape there (e.g., Jones and Raghunandan [1998], Francis and Reynolds [2003], Choi, Doogar, and Ganguly [2004]).

  17. 17

    Another issue is that the Big 4 are prohibited from operating in some countries unless they form affiliations with local auditors. It follows that audit quality falls when the country requires the Big 4 to arrange these local affiliations to access the market. After identifying these cases using Worldscope data, we narrow our BIG 4 specification to strictly firms from countries that do not require local affiliations; that is, we treat Big 4 auditors with local affiliations as equivalent to non–Big 4 auditors. In this re-specification, we continue to find evidence at the 5% level supporting the prediction in H1 that connected firms are more likely to prefer higher-quality auditors.

  18. 18

    We continue to find supportive evidence when we replace the dummy variable for whether the firm has incurred a loss in a prior year with its return on equity to capture profitability. Similarly, our results in table 5 are robust to measuring client size with revenues rather than assets. We resort to alternative proxies for some constructs when poor data availability means that we cannot examine more standard measures. For example, we follow Whisenant, Sankaraguruswamy, and Raghunandan [2003] and Sankaraguruswamy, Whisenant, and Willenborg [2013] in gauging firm complexity with the presence of extraordinary items or discontinued operations, rather than the number of subsidiaries, the audit report lag, or the presence of pension or other postretirement plans. However, the interaction with CONNECTED becomes statistically insignificant when we measure client complexity with the number of business segments.

  19. 19

    This split sample design with respect to country-level variables follows extensive prior research (e.g., Lang, Lins, and Miller [2004], Pinkowitz, Stulz, and Williamson [2006], Wang [2010]). In an important advantage, this approach avoids multicollinearity complications arising from the high correlations between the test variables and their interaction terms, especially when the interaction involves a dummy variable and country-level time invariant variables. Indeed, in our analysis, the Belsley collinearity test indicates a score above the threshold of 30 when we include the interaction terms.

  20. 20

    Economically, our evidence in this section implies that the importance of political connections to auditor choice is generally larger when we isolate firms from countries with relatively poor governance institutions. For the full sample, the regression in model 3 in table 3 suggests that the probability of hiring a Big 4 auditor rises 8% in the presence of a political connection. In comparison, the coefficient estimates for CONNECTED translate into an 8% impact in model 1 in table 6, 9% in model 3, 13% in model 5, and 14% in model 7.

Ancillary