J Bus Ethics DOI 10.1007/s10551-017-3495-5
ORIGINAL PAPER
Financial Reports and Social Capital Anand Jha1
Received: 25 July 2016 / Accepted: 3 March 2017 Ó Springer Science+Business Media Dordrecht 2017
Abstract I examine social capital’s impact on financial reports. Based on the social capital literature, I predict that the quality of the financial reports is higher when a firm is headquartered in a region with high social capital. Consistent with this prediction, I find that the firms that are headquartered in this type of region in the USA have a lower probability of committing fraud by misrepresenting financial information. Further, I find that the firms in regions with high social capital have lower levels of discretionary accruals and much more readable annual reports. Keywords Financial fraud Social capital Financial report Discretionary accruals Real earnings management Fog index JEL Classification M14 M41 G3
Introduction The firms’ managers are expected to report earnings truthfully and clearly. But in reality, managers often exaggerate or conceal true earnings by manipulating accruals (Healy and Wahlen 1999), altering investment decisions (Graham et al. 2005; Roychowdhury 2006), or by obfuscating financial information in their annual reports. These actions exacerbate the information asymmetry between managers and investors and are costly to shareholders (Beatty et al. 2010; Bharath et al. 2008; Bhojraj
& Anand Jha
[email protected] 1
5201 Cass Avenue, Detroit, MI 48202, USA
et al. 2009; Biddle et al. 2009; Graham et al. 2005; Francis et al. 2008). Given the cost associated with poor-quality financial reports, a large body of the literature investigates what causes this poor quality. In recent years, many of these studies have focused on how the cultural forces in the region of the firm’s headquarters affects its financial reports. This new stream of literature shows that the region’s culture permeates the firm’s culture and therefore can affect how the manager reports financial information. For example, McGuire et al. (2012), Grullon et al. (2010), and Dyreng et al. (2012) find that a firm that is located in a religious region has higher-quality financial reports. They argue that religiosity is associated with higher ethical norms, and higher ethical norms induce managers to report truthfully. My goal is to extend this stream of the literature. As in these studies, I also examine the impact of a social environment on the quality of the financial reports. But the social environment I focus on is the social capital. Although the accounting literature rarely addresses the concept of social capital, this topic is extensively studied in economics, political science, sociology, and management. The concept of social capital is based on the idea that while human beings are selfish, the degree of selfishness can vary across individuals and regions. A high-socialcapital region comprises individuals who have a high degree of altruism, a community-centric attitude, a high propensity to honor obligations, and a high degree of mutual trust (Portes 1998; Guiso et al. 2004a). These regions also have dense networks; that is, more social, civic, and political organizations per capita. Many studies in different disciplines show that the social capital of a region can impact its individuals’ decisions and the subsequent economic outcomes (Fukuyama 1995;
123
A. Jha
Putnam 2000; Guiso et al. 2004b; Chenhall et al. 2010; Laursen et al. 2012). Often the social capital’s impact reduces transaction costs. For example, in a region with high social capital, firms’ participation in the stock market is less costly as is their access to bank loans and funds from venture capitalists (Guiso et al. 2004b). These regions also have more honest bureaucrats and judges (La Porta et al. 1997) and fewer criminal activities (Buonanno et al. 2009). A recent study by Jha and Chen (2015) finds that auditors also view firms in low-social-capital regions as being less trustworthy and charge them a premium. A common theme in the studies on social capital is that the regions with high social capital have more altruistic norms and a set of networks that encourage honest behavior; both of which lower transaction costs. I build on these studies and argue that social capital also affects financial reports. I argue that, ceteris paribus, a firm headquartered in a region with high social capital has a higher-quality financial reporting. In making this argument, I assume, as in the literature (Grullon et al. 2010; Li 2008; McGuire et al. 2012), that poor financial reports are examples of misbehavior that are driven largely by the selfinterest of the managers. Also, as in those studies, I argue that the region’s culture affects the culture of the firm’s headquarters. My argument is based on the idea that the altruistic norms and the dense networks of these regions with high social capital induce the managers to report financial conditions truthfully. I argue that the managers in these regions are likely to have a higher set of ideals when deciding what degree of earnings management is acceptable and what is not. The further they deviate from this self-imposed standard, the guiltier they are likely to feel. I find that this sense of guilt is a psychological cost; and, therefore, the managers in the regions with high-socialcapital exercise more restraint when managing earnings or obfuscating information. Furthermore, because there are more civic, social, and political organizations in these regions, the external monitors have a greater chance of interacting and sharing information with each other—this sharing can lead to more effective monitoring (Wu 2008). In addition, the managers in a dense network of associations might perceive a harsher punishment for deviant behavior and thus might be encouraged to fulfill their obligations (Coleman 1990; Spagnolo 1999; Hilary and Huang 2015). This dense network also has an indirect effect. Over time, it strengthens the norm that encourages the managers to fulfill their obligations (Fukuyama 1997). To test my proposition, I exploit the variations in the social capital of counties in the USA and examine their impact on the quality of the financial reports. I use the Karpoff et al. (2012) database to measure the quality of the financial reports. This database lists the firms that were
123
prosecuted for financial misrepresentation by the Security and Exchange Commission (SEC) and the Department of Justice (DOJ). The database also lists the date the violation occurred. Based on the violation year and the firm identifier, I can establish the firm and the year it misrepresents its financial report. I consider those firm years that are not listed in Karpoff et al. (2012) as those where the financial information are presented correctly. To measure social capital, I construct an index as in Rupasingha et al. (2008). As in their study, I use the first principal component in an analysis of four variables: voter turnout in the presidential election, the census response rate, the number of social and civic associations, and the number of non-government organizations (NGO) at the county level. I match the social capital data at the county-level to the firm-level data based on the headquarters’ zip codes. I then conduct a firm-level analysis to examine the social capital’s impact on the probability of being found guilty for financial fraud. I find that the firms with a headquarters in a county with high social capital are less likely to be prosecuted by the SEC for financial fraud. Based on my model, a one standard deviation increase in the social capital is associated with a 19% decrease in the odds that the SEC will prosecute a firm. The results continue to hold for the period after the Sarbanes–Oxley Act of 2002 (SOX) and are not driven by a large sample size. Further, the results are not driven by other cultural forces that affect the quality of the financial reports, such as religiosity. In fact, this result continues to hold when controlling for a myriad of regional level controls, such as the income per capita, literacy, income inequality, and racial diversity that are highly correlated with social capital and are its possible determinants. The results are robust when using propensity score matching. I also use the change in headquarters of a firm. I find that the propensity to commit fraud is less for firms that relocate to a county with a higher level of social capital, compared to firms that relocate to a county with a low level of social capital. This result indicates that there is likely a causal effect from social capital on the propensity to commit fraud. Further, I find evidence that the probability of a CEO’s replacement is more than double in a high-socialcapital county compared to a low-social-capital county when the firm’s possible commitment of fraud becomes publicly known. This evidence further strengthens this causal link. The negative association between social capital and the probability of the SEC’s prosecution appears to be due to managers’ good behavior in high-social-capital regions rather than the lax attitude of SEC officials. If the good behavior of managers is driving the association between social capital and financial fraud, then the effect of social capital should be much stronger when external monitoring is weak. Indeed, my results show that is the case. For
Financial Reports and Social Capital
example, the effect of social capital on the propensity to commit financial fraud is stronger for weakly monitored firms, such as those with fewer institutional investors and those with less financial leverage. High-social-capital regions also have lower levels of discretionary accruals and much more readable annual reports. These two measures of financial reporting are unlikely to be affected by the SEC’s aggressive prosecution. What is more, these associations also appear stronger when external monitoring is weak. There are two separate changes in the regulations that increase the SEC’s scrutiny—SOX created a shock that altered the incentives to use discretionary accruals, and the regulation for plain English disclosures changed the incentive for writing unreadable annual reports. For example, following SOX, the incentives to use discretionary accruals has declined (Cohen et al. 2008). If the social capital does have an impact on the incentives to manipulate accruals, then it should have a much stronger impact in the pre-SOX period—that is precisely what I find. Similarly, the plain English guidelines that the SEC mandated in 1998 also created a shock to how annual reports were written, and likely increased the SEC’s scrutiny on how annual reports were written. If the social capital affects the managers’ language in the annual reports, then its effect should be much more salient before the SEC’s mandate—again, that is precisely what I find. Taken together, my results provide strong and direct evidence that social capital positively affects the quality of financial reports, ceteris paribus, and this positive effect is mainly driven by the good behavior of the managers in high-social-capital regions.
Contribution to the Literature By providing direct evidence that the social capital affects financial reporting, my study makes an important contribution to the financial reporting literature. There is a growing interest in understanding how the social environment affects financial reports. Many of these studies have specifically focused on the effect of religiosity (e.g., Dyreng et al. 2012; Grullon et al. 2010; McGuire et al. 2012). But, these studies do not analyze how the social capital affects the quality of financial reports. My results raise the possibility that the effect on this quality comes from social capital instead of religiosity; or, at the very least, that the social capital might affect the financial reports in addition to religiosity. Based on the literature and additional tests, I argue that religiosity and social capital measure different aspects of cultural norms—religiosity measures faith in God, social capital measures the norms for cooperative behavior such as honesty and the propensity to honor one’s
obligations. While religion is, arguably, a source of social capital, it is not the only one. Moreover, social capital neither mediates nor moderates the effect of religiosity and the quality of financial reports. That the effect of social capital on the quality of financial reports is independent of religiosity makes the results quite important to the literature on how the social environment affects financial reporting. Besides contributing directly to the financial reporting literature, my study makes an important extension to Jha and Chen (2015) who show that firms headquartered in a low-social-capital region pay higher fees to their external auditor. Jha and Chen (2015)’s key argument is that auditors trust the clients from low-social-capital regions less and therefore charge them a premium. Because they do not investigate the social capital’s effect on the quality of the financial reports, their study is not clear on whether the lack of trust is simply due to the auditors’ prejudice based on the client’s location, or because of the auditor’s prudence. Further, Jha and Chen (2015) also find that in addition to the firm’s social capital, the auditor’s social capital also matters, which raises the possibility that the auditor’s prejudice could be playing a role. My study finds that high fees are a reflection of the auditor’s prudence, not prejudice—firms in low-social-capital regions have poorquality reports; therefore, auditors are prudent in charging them higher fees. My study extends the literature on what affects ethical decision making. For example, Carpenter and Reimers (2005) find that the attitude of top management has a strong effect on ethical behavior. My study finds that an attitude toward misbehavior might be more restrained when the firms are in a high-social-capital region. More broadly, it contributes to the emerging strand of literature that tries to better understand the factors that influence the propensity to commit financial fraud (Albrecht et al. 2015; Chakrabarty 2015; Raval 2016; Hass et al. 2016). My study also complements a number of other research streams. Callen and Fang (2012) argue that firms in religious regions hoard bad news to a lesser extent. These authors show that the firms that are headquartered in religious regions have a lower risk of a crash in their stock price. As Li and Wang (2015) find, my research indirectly indicates that social capital might also affect the hoarding of bad news and therefore might also affect the risk of those crashes. In short, this study has implications that go beyond the social capital’s impact on financial reports. My study also complements Hasan et al. (2015) who find that banks provide loans at a lower interest rate and with less stringent terms to firms with high social capital. My results support their conjecture that managers from high-social-capital regions are less opportunistic.
123
A. Jha
My study extends Kim et al. (2012) who argue that corporate social responsibility (CSR)1 may be positively or negatively associated with the financial reporting’s quality—but finds a negative association. My study, along with Jha and Cox (2015), finds a positive association for social capital that indicates that CSR’s effect on financial reporting might be coming from the social capital of where the firm is headquartered.
Related Literature and Hypothesis Development
over a long period might foster a norm conducive to cooperation. And people internalize this norm over generations and are intrinsically less likely to act opportunistically. In short, although the concepts of ‘‘norms’’ and ‘‘networks’’ appear distinct at first glance, the distinction is not clear. The differentiation of the agent’s good behavior is difficult, if not impossible to do. Is the good behavior because of higher ethical norms or because of the fear of harsher punishment for misbehavior from strong networks? Because the effects impact financial reports in the same direction, I do not make a distinction between norms and networks, as in the literature (Guiso et al. 2004b, 2008a, b).
What is Social Capital? I follow Woolcock (2001) and define social capital as the norms and the networks that facilitate collective action. This definition is comprehensive. It incorporates what is the consensus in the social capital literature: The regions with higher social capital have norms that encourage the honoring of obligations and mutual trust, have dense networks, and have norms and networks that are related to each other. The predominant approach in economics and political science is to view social capital as a norm. Guiso et al. (2004b) define social capital as the levels of mutual trust and altruistic tendency in a society. Fukuyama (1997) defines social capital as ‘‘the existence of a certain set of informal values or norms shared among members of a group that permits cooperation among them.’’ He points out that cooperative norms arise from the iterated prisoner’s dilemma games and shared historical traditions. Portes (1998) argues that people internalize such social norms over generations and feel obligated to behave in a certain way. He goes on to define social capital as the propensity to honor obligations. These definitions share a common theme. The people in the regions with high social capital are relatively speaking, less self-centered, and more conscious of fulfilling their obligations. However, in many studies, particularly in the management literature, the researchers often view social capital as a set of networks from which benefits are derived (Payne et al. 2011). They argue that a strong set of networks is in and of itself a resource. For example, when the social networks are strong the agent might fear a greater cost for misbehavior (Coleman 1990; Spagnolo 1999). Furthermore, the monitoring might be more effective because the monitors might share information (Wu 2008). But, as Fukuyama (1997), Portes (1998), and Putnam (2001) point out, strong networks 1
CSR is a measure of how socially responsible a firm is—it is not a measure of the intrinsic norms of the managers to behave honestly and the propensity to honor obligations—social capital measures these qualities. Also, higher CSR can sometimes be a cover for misbehavior (Hemingway and Maclagan 2004; Prior et al. 2008; Petrovits 2006).
123
Why Might a Place with Social Capital Constrain the Financial Misreporting of the Firms Headquartered There? In this subsection, I consider the firms’ managers to have high social capital when the firm is located in a region with high social capital.2 I argue, as summarized in Fig. 1, that there are two channels by which social capital affects the quality of financial reports: an altruistic norm and a dense network. The managers of firms in regions with high social capital are likely to have similar norms because the social norms of the firm’s location and that of the headquarters are congruent. In other words, they are less self-centric, take their obligations more seriously, and face greater intrinsic costs when deviating from their ideals. Such norms are likely to constrain opportunistic behavior and therefore reduce the likelihood of financial misrepresentation. The idea that the values shaped by social norms affect individual decisions is not new (Arrow 1979; Sen 1987). More than 2000 years ago, in his book Nicomachean Ethics, Aristotle discussed how in civilized societies, people take ethics into consideration when making decisions (Aristotle 2004). More recently, in his presidential address to the American Economic Association, Akerlof argued that for a better understanding of why agents make certain decisions, researches should take into account the opportunity cost of deviating from one’s ideal (Akerlof 2007). In an experimental setting, Karlan (2005) shows that some people are more trustworthy than others, and the trustworthiness of a person is associated with good behavior, such as a lower default rate on loan re-payments. 2
These are some of the recent studies (e.g., Grullon et al. 2010; Hilary and Hui 2009; McGuire et al. 2012; Jha and Chen 2015; Jha and Cox 2015) that examine the role of either religiosity or social capital in financial decisions. In each of these studies the culture of where the firm is headquartered is used as measure for the culture of the managers. These studies provide detailed discussions on how the culture of the place of domicile and that of a firm’s manager are congruent. For brevity, I refer to these papers and do not present that argument in this paper.
Financial Reports and Social Capital
Fig. 1 Rationale for how social capital affects the quality of financial reports. This figure summarizes the channels via which high social capital can increase the quality of financial reports
A high-density network is a key characteristic of regions with high social capital that also encourages the good behavior. There are three ways in which a dense network can lead to higher-quality financial reports. First, a highdensity network means that individuals interact more with each other. This interaction means that the stakeholders, such as institutional investors, bankers, and managers, are more likely to interact regularly with each other. The more frequent interaction among these parties leads to greater information exchange. The information exchanged is also more reliable. The higher quality and quantity of information shared among stakeholders translates into more effective monitoring (Wu 2008). Second, when the network’s density is high, then the managers interact more in the community and therefore they perceive a higher cost for unethical behavior (Coleman 1990; Spagnolo 1999). Managers are more likely to take seriously whatever checks that stakeholders have put in place for effective monitoring. Third, a dense network, over time, incorporates the cooperative norms in the community, and such norms encourage managers to behave honestly (Fukuyama 1997; Portes 1998; Putnam 2001). Thus, the punishment for deviant behavior is harsher because the standard to behave honestly is higher.
Can High Social Capital Lead to Poor-Quality Financial Reports? An argument could be made that in a high-social-capital region because the expectation of unethical behavior is lower, monitoring might be lax and managers might be able to misbehave more easily. This possibility is unlikely for two reasons. First, this argument assumes a one-shot game between the regulator and the firm. In fact, the game is a multiple-shot game with multiple iterations. If managers of certain regions misbehave in one round, they will no longer enjoy high trust. So, at the equilibrium, a situation of higher trust and more misbehavior should not exist. Overall based on the discussion in ‘‘Related literature and hypothesis development’’ section, I hypothesize the following: Hypothesis Financial misrepresentation is less likely when a firm is located in a region with high social capital.
Research Design and Data Empirical Model To test the hypothesis, I use the following empirical model:
123
A. Jha
FinclFraud ¼ b0 þ b1 SocialCapital þ b2 lnðMarketValueÞ þ b3 Analysts þ b4 ReturnOnAssets þ b5 DebtToAssets þ b6 Big4 þ b7 MarketToBook þ b8 Loss þ b9 VolatilityOfCashflow þ b10 Rural þ b11 Religiosity þ b12 IncomePerCapita þ b13 lnðDistancefromSECÞ þ b14 lnðPopulationÞ þ b15 PopulationDensity þ FamaFrench Industry Indicators þ Year Indicators þ e:
ð1Þ
Each variable is at the firm year level. For expositional ease, I suppress the subscripts i and t. The dependent variable, FinclFraud, is a binary variable that equals one for the firm year in which the firm commits its financial fraud or zero otherwise. The key research variable, SocialCapital, measures the social capital of the county where the firm is headquartered. Because the dependent variable is a binary variable, I use the logit model for analysis. The firm-level control variables are based on McGuire et al. (2012) who examine the association between the lawsuit and the firm’s religiosity. Because the dependent variable in my study is closely related to their dependent variable, I mimic their control variables. Specifically, I control for the natural logarithm of the market value of the firm (ln(MarketValue)); the number of analysts following (Analyst); the return on assets (ReturnOnAssets); the ratio of the debt to the total assets (DebtToAssets), an indicator variable that equals one if the auditor is one of the Big4 auditing firms and zero otherwise (Big4); the ratio of the book value to the market value of equity (MarketToBook), an indicator variable to measure if the firm is in loss (Loss); and the volatility of the cash flow (VolatilityOfCashflow). Next, I control for the county-level characteristics that are also largely based on McGuire et al. (2012). As in their study, I control for whether the county is rural or not (Rural), its religiosity (Religiosity), and income per capita (IncomePerCapita). I also control for the distance from the SEC (ln(DistancefromSEC)) because Kedia and Rajgopal (2011) find that firms located further away from a SEC branch are less likely to be prosecuted by the SEC. In addition, I also control for the population (ln(Population)) and its density (PopulationDensity). A more detailed description of the control variables is in ‘‘Appendix 1’’ section. I also add the 48 industry indicators based on Fama and French’s 48-digit industry classifications and year indicators. The propensity to commit financial fraud can be higher in certain industries, and the SEC and the DOJ
123
might be more aggressive in certain years in pursuing firms for financial misrepresentation.3 Further, I cluster the standard errors at the county level. This clustering method automatically controls for the autocorrelation at the firm level (Bertrand et al. 2004).4,5 The expected signs of the control variables are as follows: I predict the variables Analysts, DebtToAssets, and Big4 will have a negative association with FinclFraud because more analysts following, a greater proportion of debt, and a reputable auditor are all associated with betterquality monitoring and hence can constrain managerial opportunism. Because the research finds that financial fraud is more likely when the firm is in financial distress (Bell et al. 1991), I expect FinclFraud to be negatively associated with ReturnOnAssets, and positively associated with Loss and VolatilityOfCashflow. How ln(MarketValue) is related to FinclFraud is unclear. On the one hand, firms with a large market value attract greater scrutiny from external monitors such as analysts and investors; hence, they might have a lower likelihood of financial fraud. On the other hand, the benefits of misbehavior, if undetected, might also be greater for managers of large firms. Nevertheless, ln(MarketValue) is an important control because it is a good measure for size as large firms can be quite different from small firms. How MarketToBook is related to FincFraud is not obvious, but because firms with greater market-to-book are quite different from those with low values of market-to-book, we use it as a control variable. As for county-level controls, we know that religious norms can constrain managerial opportunism (McGuire et al. 2012), but that greater distance from the SEC’s regional office can encourage fraud (Kedia and Rajgopal 2011). Therefore, I predict a negative association between Religiosity and FinclFraud, and a positive association between ln(DistancefromSEC) and FinclFraud. As for the other control variables, how they are related to FinclFraud is not obvious.
3
I do not control for the literacy, income inequality, and racial diversity, because these variables are determinants of the SocialCapital. However, as robustness tests I verify that the results continue to hold even when I control for these three regional characteristics. 4 I check the inflation factor (VIF) for multicollinearity among the independent variables in each OLS test on the main hypotheses. The largest VIF, 3.96, is much \10. This VIF shows that multicollinearity is not a problem. 5 I also verify that the results are robust when standard errors are double clustered in the following ways: clustering by firm and county, firm and industry, firm and year, county and year, and county and industry.
Financial Reports and Social Capital
misconduct they identify. Third, these databases omit a large fraction—up to 62%—of the events they purport to capture. Fourth, between 15.7% and 84.6% of the events in these databases are duplicates in the sense that they identify the same underlying cases of misconduct.
Measuring Key Variables Dependent Variable I obtain the dependent variable in Eq. (1) from the Federal Securities Regulation (FSR) database compiled by Karpoff et al. (2012). The database consists of the universe of federal enforcement actions for the misrepresentation in books, records, and internal controls. The database comprises all of the firms that are guilty of violating one or more of the following three ‘‘13(b)’’ provisions of the Securities and Exchange Act of 1934: 1.
2.
3.
Section 13(b) (2) (a), a.k.a. 15 U.S.C. §§ 78 m (b) (2) (A)—it requires firms to keep and maintain books and records that accurately reflect all transactions. Section 13(b) (2) (b), a.k.a. 15 U.S.C. §§ 78 m (b) (2) (B)—it requires firms to devise and maintain a system of internal accounting controls. Section 13(b) (5), a.k.a. 15 U.S.C. §§ 78 m (b) (5)—it prohibits knowingly circumventing or failing to implement a system of internal accounting controls, or knowingly falsifying any book, record, or account.
To construct the FSR database, Karpoff et al. (2012) hand-collect data from seven different sources: (1) the SEC Web site; (2) the Department of Justice Web site; (3) the Wolters Kluwer Law and Business Securities (Federal) electronic library; (4) Lexis-Nexis; (5) the PACER database that contains the lawsuit-related information from the federal appellate, district, and bankruptcy courts; (6) the SEC’s Electronic Data Gathering, Analysis, and Retrieval (EDGAR) system; and (7) Lexis-Nexis’ All News and the Dow Jones’ Factiva news sources that provide the civil suits and regulatory scrutiny. These sources group the related events into ‘‘cases’’ and read all of the related events in these databases to avoid double counting, and to identify the correct dates. This database has several advantages compared to the data from the four traditional sources: Government Accountability Office (GAO), Audit Analytics (AA), Securities Class Action Clearinghouse (SCAC), and the Securities and Exchange Commission’s Accounting and Auditing Enforcement Releases (AAERs). Karpoff et al. (2012) note that the four traditional data sets suffer from a number of flaws: First, the dates provided by these databases [the four databases mentioned above] lag the initial revelation of the financial misconduct by an average of 150 to 1017 calendar days, depending on the database. Second, the events in these databases capture, on average, only 6% to 36% of the value-relevant announcements associated with the cases of
In other words, Karpoff et al. (2012)’s FSR database is a clean database with which to examine the factors that might affect the propensity to commit fraud.6 A number of studies on financial fraud use an earlier version of the FSR database (Giannetti and Wang 2014; Khanna et al. 2015; Files et al. 2012; Karpoff et al. 2008a, b). This database spans from 1978 to 2011 and provides the date the violation began. I use the year the violation began and the firm identifier, also available in the database, to identify which firms misrepresent the financial information in their reports. If a firm year does not appear in this database, then I consider the firm to have represented its financial information correctly. The FSR database lists 721 unique cases (firm years) of fraud spanning from 1990 to 2009. When I limit the firms available in COMPUSTAT to those that at least list their total assets, the number is 580. Then when I limit the sample to only those firms that are nonfinancial, nonregulated, and have the firm-level variables that are used in Eq. (1), the number of firm years with fraud cases becomes 408. Measuring Social Capital I construct a social capital index for each county following Rupasingha et al. (2008). I use the data set that they compile and use their approach.7 They use two measures of altruistic norms and two measures of network density. They also conduct a principal component analysis to construct an index for each county. As far as I know, the Rupasingha et al. (2008) approach to measuring social capital is the most comprehensive measure at the county level. Many authors in different disciplines have used Rupasingha et al. (2008) index or have followed their approach to construct their own (Chetty et al. 2014; Putnam 2007; Deller and Deller 2010). Following Rupasingha et al. (2008), the two measures for norms that I use are voter turnout in the presidential election and the census response rate. A higher value in both represents higher social capital. Also following 6
The fact that the GAO, AA, SCAC, and the AAERs omit a large faction of events they purport to capture and double count some of them creates type 1 and type 2 errors—this problem might compromise the validity of any tests using these data sources to examine the propensity to commit fraud. 7 The data and descriptions on how social capital indexes are constructed are available at: http://aese.psu.edu/nercrd/community/ tools/social-capital.
123
A. Jha
Rupasingha et al. (2008), the two measures for the networks are the number of social and civic associations that include physical fitness facilities, public golf courses, religious organizations,8 sports clubs, managers and promoters, political organizations, professional organizations, business associations, and labor organizations in the county normalized by the population of the county; and the number of non-government organizations (NGO) that excludes the NGOs with an international focus, normalized by the population in the county. I then conduct a principal component analysis to construct an index for each county for the years 1990, 1997, 2005, and 2009 and consider the first component as a measure of the social capital.9 To maintain consistency between the years, I only use the first component for each year and consider it the social capital index. I then linearly interpolate the data to fill in the years 1990–1997, 1998–2004, and 2006–2008. That is, I assume a constant change from year to year between the two points as in Hilary and Hui (2009). I present the variation in the social capital at the county level in Fig. 2 for the year 2005. The map shows that the counties with high social capital are concentrated mainly in the Midwest and the Northeast. The correlation between the social capital index of 1990 and that of 2009 is 0.72. This is consistent with the idea that unlike physical and human capital, social capital is ‘‘sticky’’ (Anheier and Gerhards 1995). In this sample the top three counties with the highest social capital are Fairfax, VA, Polk, Iowa and Hardin, Iowa, respectively. The three counties with lowest social capital are Queens, New York; followed by Kings, New York, and Walker, Texas. When constructing the social capital index, I assume that all of the association memberships increase the general altruism and propensity to honor obligations in the society. I acknowledge that not all researchers agree with this view. In the social capital context, the common approach is to view associations as one of two types: those that act like ‘‘bridges’’ between groups, such as religious organizations, civic and social associations, bowling centers, physical fitness facilities, public golf courses, and sports clubs; and those that strengthen the ‘‘bonding’’ between the members of the groups such as professional, labor, and political organizations and business associations, all of which have more exclusive memberships. Although the bridging associations clearly can increase the general trust in the society, whether the bonding
associations increase the general trust in the society is unclear. Some researchers argue that a bonding association creates trust within the network but might create distrust with those outside the network (Burt 1999, 2000). Some researchers disagree: Putnam (2007) argues that viewing bridging and bonding social capitals as inversely related in a kind of zero-sum game is flawed. He goes on to show that people who have strong bonding social capital also appear to have more bridging social capital. Citing an example to prove this point, Putnam (2007) notes that the research shows that whites who have non-white friends also have more white friends. Brewer (1999) also finds that the higher trust inside a group does not mean a lack of trust outside of the group. I take the view of Putnam (2007) and Brewer (1999) and consider both types as increasing trust. Data The sample I use to test the hypothesis consists of 85,743 firm years, 10,168 firms, 741 counties that span 1990–2009. Following the literature, the sample excludes firms in the financial industry (SIC: 6000–6999) and the regulated industry (SIC: 4000–4999). It comprises all of the firm years for which there are non-missing observations for the SocialCapital and the control variables that are used in Eq. (1). The sample selection process is as follows: I start with, 112,379 firm years, all of the nonfinancial and unregulated firms in Compustat between 1990 and 2009 that are located in the USA and that have non-missing values for their total assets. I remove 565 firm years because the social capital of the county they are located in is not available. I remove 15,487 firm years because ln(MarketValue) is unavailable, 1176 because ReturnOnAssets is unavailable, 191 because DebtToAssets is unavailable, 430 because the name of the auditor is unavailable, 42 because the MarketToBook is unavailable; 67 because Loss is unavailable, 7325 because VolatilityOfCashflow is unavailable, and 709 because religiosity is unavailable. Next, I remove 644 observations because they belong to industries where there was not a single case of fraud during the sample period.10 To remove the effect of outliers, I winsorize all continuous variables at the 1st and the 99th percentile. The summary statistics and the correlations of the variables that are used to test the hypothesis are presented in Panels A and B of Table 1, respectively. Based on the summary statistics, about 0.5% of the firm years in the
8
The results continue to hold if I remove religious organization. The eigenvalues of the first components for these years are 1.80, 2.06, 1.94, and 1.80 respectively. The eigenvalues of the other components are less than one except in 2009 when the second component has an eigenvalue of 1.03.
9
123
10
The industries that I remove have the following Fama–French 48-digit Industry classifications: 3 (Soda: Candy and Beer), 5 (Smoke: Tobacco Products), 26 (Guns: Defense), and 29 (Coal).
Financial Reports and Social Capital
Fig. 2 Social capital by counties. This figure presents the distribution of social capital by county for the year 2005. The higher values represent higher social capital
sample represent financial fraud, and the rest represent no fraud. The correlations show that FinclFraud and SocialCapital are, as expected, negatively correlated. The correlation is significant at the 5% level. The SocialCapital, as expected, is also positively and significantly correlated with Religiosity, Rural, and IncomePerCapita.
Main Results Higher Social Capital is Associated with a Lower Propensity of Committing Fraud Consistent with the hypothesis, the multivariate logit model shows that the firms in regions with high social capital are less likely to commit financial fraud. The results are presented in Panel C of Table 1. Column 1 presents the coefficients when the specification of Eq. (1), the main model, is used. The coefficient for SocialCapital is -0.242 and significant at the 1% level. The coefficient retains its significance when I remove the county-level control variables (column 2), or when I use only the firm’s size, industry, and year as control variables (column 3). The coefficient for SocialCapital continues to be significant, although at a lower level, in the pre-SOX and post-SOX periods (columns 4 and 5). A large sample size also does not appear to drive the results. To address this possible concern, I collapse the data so that each firm has only one observation. I consider that a firm commits financial fraud if for any of the years between
1990 and 2009, the firm committed financial fraud. For all of the other variables, I take the median of the sample. The sample size is reduced to 10,168 firms, but the results continue to be similar. The coefficient for SocialCapital continues to be negative and significant. The coefficients for the other variables are also similar (column 6).11,12 The economic significance of the results is also quite large. Based on the coefficient in column 1, the factor change in a one standard deviation increase in SocialCapital is 0.81 (exp (-0.242 9 0.850)) (see p. 296, Hardy and Bryman 2004). In other words, this change is associated with 19% lower odds of committing financial fraud. The SocialCapital appears to better explain the propensity to commit financial fraud than religiosity— although negative, the coefficient for Religiosity is not
11
Although untabulated, I also verify that the results continue to hold when I conduct a county-year regression. That is, I calculate the median of all variables at the county-year level and test if the results are robust. 12 In the main model, I do not control for ethnic diversity, literacy, and income inequality from the county where the firm is headquartered because these variables are highly correlated with social capital and are also its possible determinants. Whether these variables have a direct effect on financial reporting is also unclear, but quite likely they affect social capital, which in turn affects financial reporting. Therefore, the inclusion of these variables can likely take away from what might be the effect of social capital. Regardless, even when I control for these variables, I continue to find that, ceteris paribus, higher social capital has an association with better quality financial reports. However, as expected the significance levels are slightly lower. For brevity, I do not report these results.
123
A. Jha Table 1 Main result: Higher social capital is associated with a lower propensity of committing fraud N
Mean
SD
Median
p25
p75
Panel A: Summary statistics FinclFraud
85,743
0.005
0.069
0.000
0.000
0.000
SocialCapital
85,743
-0.601
0.850
-0.561
-1.240
0.036
ln(MarketValue)
85,743
4.493
2.395
4.475
2.820
6.176
Analysts
85,743
3.083
5.318
0.000
0.000
4.000
ReturnOnAssets
85,743
-0.086
0.684
0.091
-0.045
0.159
DebtToAssets
85,743
0.651
0.992
0.485
0.285
0.682
Big4
85,743
0.650
0.477
1.000
0.000
1.000
MarketToBook
85,743
2.765
6.495
1.803
0.895
3.436
Loss
85,743
0.585
0.493
1.000
0.000
1.000
VolatalityOfCashFlow
85,743
0.198
0.508
0.071
0.039
0.144
Rural
85,743
0.121
0.326
0.000
0.000
0.000
Religiosity
85,743
0.595
0.141
0.601
0.479
0.693
IncomePerCapita
85,743
36,581.957
13,709.540
33,682.000
26,553.000
43,649.000
ln(DistancefromSEC)
85,743
4.778
1.535
5.251
3.617
5.883
ln(Population)
85,743
13.659
1.138
13.713
13.126
14.301
PopulationDenisty
85,743
4374.467
12,846.379
1426.613
660.591
2261.205
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
Panel B: Pearson correlations [1]
FinclFraud
[2]
SocialCapital
[3]
1.00 -0.01
1.00
ln(MarketValue)
0.04
0.06
1.00
[4]
Analysts
0.02
0.02
0.65
[5]
ReturnOnAssets
0.01
0.05
0.32
0.18
1.00
[6]
DebtToAssets
-0.01
-0.03
-0.25
-0.10
-0.67
1.00
[7]
Big4
0.01
0.05
0.39
0.25
0.22
-0.18
1.00
[8]
MarketToBook
0.02
-0.00
0.17
0.08
0.08
-0.18
0.03
1.00
[9]
Loss
-0.01
-0.08
-0.43
-0.31
-0.31
0.17
-0.16
-0.01
[10]
VolatalityOfCashFlow
-0.00
-0.05
-0.25
-0.15
-0.68
0.52
-0.21
-0.04
0.22
1.00
[11]
Rural
-0.01
0.15
-0.02
-0.02
0.01
0.00
-0.01
-0.02
-0.03
-0.02
1.00
[12]
Religiosity
0.01
0.19
0.01
-0.03
0.03
0.01
-0.03
-0.01
-0.05
-0.03
-0.14
[13]
IncomePerCapita
-0.00
0.27
0.12
0.08
-0.12
0.06
-0.03
0.02
0.08
0.10
-0.21
0.17
1.00
[14]
ln(DistancefromSEC)
0.00
-0.04
0.09
0.09
0.05
-0.04
0.09
-0.00
-0.04
-0.04
0.10
-0.22
-0.30
1.00
[15]
ln(Population)
0.00
-0.52
0.01
0.01
-0.05
0.02
0.00
0.02
0.08
0.05
-0.41
0.07
0.15
-0.23
1.00
[16]
PopulationDenisty
0.00
0.21
0.00
-0.01
-0.04
0.04
-0.03
0.01
0.02
0.04
-0.11
0.27
0.56
-0.44
0.15
(1)
1.00
(2)
(3)
1.00
1.00
1.00
(4)
(5)
(6)
Year B 2002
Year [ 2002
One obs. per firm
DV = FinclFraud
Panel C: Higher social capital is associated with a lower propensity to commit fraud SocialCapital ln(MarketValue) Analysts ReturnOnAssets DebtToAssets Big4 MarketToBook
123
-0.242***
-0.194***
-0.206***
-0.209**
-0.343**
-0.197**
(0.004)
(0.003)
(0.002)
(0.034)
(0.046)
(0.016)
0.331***
0.327***
0.263***
0.358***
0.293***
0.423***
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
-0.025**
-0.026***
-0.021*
-0.041**
-0.042***
(0.011)
(0.009)
(0.052)
(0.049)
(0.001)
0.194
0.205
0.238
0.180
0.225
(0.166)
(0.146)
(0.166)
(0.553)
(0.216)
-0.096
-0.089
0.002
-0.321
0.152*
(0.184)
(0.224)
(0.986)
(0.141)
(0.052)
-0.115
-0.124
0.085
-0.806***
-0.205
(0.381)
(0.348)
(0.583)
(0.005)
(0.143)
0.008
0.008
-0.003
0.023*
-0.017
Financial Reports and Social Capital Table 1 continued (1)
(2)
(3)
(4)
(5)
(6)
Year B 2002
Year [ 2002
One obs. per firm
DV = FinclFraud
Loss VolatalityOfCashFlow Rural
(0.312)
(0.314)
(0.782)
(0.056)
(0.128)
0.162
0.150
0.163
0.193
0.441***
(0.153)
(0.190)
(0.181)
(0.417)
(0.002)
0.401***
0.406***
0.395***
0.304*
0.098
(0.000)
(0.000)
(0.001)
(0.094)
(0.624)
-0.409**
-0.493**
-0.132
-0.450**
(0.034)
(0.022)
(0.748)
(0.027)
-0.125
-0.138
-0.333
0.271
(0.708)
(0.732)
(0.660)
(0.437)
IncomePerCapita
-0.000
-0.000
-0.000
-0.000**
(0.188)
(0.246)
(0.805)
(0.017)
ln(DistancefromSEC)
0.010
-0.029
0.136**
0.036
(0.754)
(0.500)
(0.043)
(0.325)
Religiosity
ln(Population) PopulationDenisty
-0.102*
-0.150**
0.051
-0.112**
(0.050)
(0.012)
(0.682)
(0.039)
0.000**
0.000**
0.000
0.000***
(0.012)
(0.039)
(0.227)
(0.001)
Industry dummies
Yes
Yes
Yes
Yes
Yes
Yes
Year dummies
Yes
Yes
Yes
Yes
Yes
Yes
Observations
85,743
85,743
85,743
53,290
29,947
10,168
Pseudo R2
0.0753
0.0734
0.0687
0.0757
0.0808
0.0663
Prob [ v2
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
Low social capital (matched group)
High social capital (treated)
(N = 42,864)
(N = 42,864)
Mean
Median
SD
Mean
Median
t stat
p value
Mean in low social capital/mean in high social capital
SD
Panel D: Propensity score matching: Firms in high-social-capital regions are less likely to commit financial fraud FinclFraud
0.0055
0.0000
0.0737
0.0042
0.0000
0.0643
2.77
0.006
1.31
ln(MarketValue)
4.5418
4.5435
2.3681
4.5859
4.5642
2.3917
-2.71
0.007
0.99
Analysts ReturnOnAssets
3.0779
0.0000
5.2995
3.1375
0.0000
5.2071
-1.66
0.097
0.98
-0.0693
0.0960
0.6550
-0.0581
0.0992
0.6382
-2.52
0.012
1.19
DebtToAssets
0.6516
0.4771
1.0096
0.6390
0.4911
0.9408
1.89
0.059
1.02
Big4
0.6647
1.0000
0.4721
0.6678
1.0000
0.4710
-0.94
0.346
1.00
MarketToBook
2.6712
1.8176
6.2422
2.6927
1.7949
6.1225
-0.51
0.611
0.99
Loss
0.5558
1.0000
0.4969
0.5501
1.0000
0.4975
1.67
0.095
1.01
VolatalityOfCashFlow
0.1893
0.0727
0.4672
0.1789
0.0649
0.4767
3.22
0.001
1.06
(1)
(2)
DV = CEOReplaced Social capital [ = Median
Social capital \ Median
Panel E: The punishment for committing fraud is harsher is a high-social-capital region FraudTriggerYear ln(MarketValue) Analysts ReturnOnAssets
1.328***
0.567**
(0.000)
(0.026)
-0.139***
-0.107***
(0.000)
(0.000)
0.017***
0.007*
(0.000)
(0.056)
-0.457**
-0.349
123
A. Jha Table 1 continued (1)
(2)
DV = CEOReplaced Social capital [ = Median
DebtToAssets Big4 MarketToBook Loss VolatalityOfCashFlow County-level controls
Social capital \ Median
(0.039)
(0.225)
0.400***
0.491***
(0.000)
(0.005)
-0.081
0.163**
(0.295)
(0.032)
-0.006
0.005
(0.207)
(0.285)
0.318***
0.330***
(0.000)
(0.000)
0.371
-0.444
(0.386)
(0.326)
Yes
Yes
Diff in Coeff of FraudTriggerYear p value
(0.048)
Industry dummies
Yes
Year dummies
Yes
Yes
Observations
10,433
10,453
Pseudo R2
0.039
0.038
Prob [ v2
0.000
0.000
(1)
(2)
DV = FinclFraud DebToAssets \ median
Yes
(3)
(4)
DV = FinclFraud DebToAssets C median
Inst. Investor \ median
Inst. Investor C Median
Panel F: The effect of social capital on the propensity to commit financial fraud is stronger when the external monitoring is weak SocialCapital ln(MarketValue) Analysts ReturnOnAssets DebtToAssets Big4 MarketToBook Loss VolatalityOfCashFlow County-level Controls
-0.277**
-0.238**
-0.253*
(0.019)
(0.032)
(0.071)
-0.208** (0.026)
0.271***
0.345***
0.390***
0.219***
(0.000)
(0.000)
(0.000)
(0.000)
-0.014
-0.033**
-0.119*
0.006
(0.376)
(0.028)
(0.085)
(0.725)
0.109
0.232
0.276
-0.174
(0.708)
(0.182)
(0.131)
(0.546)
0.865
-0.295*
-0.216*
0.310**
(0.146)
(0.054)
(0.063)
(0.019)
-0.321*
0.127
-0.351**
0.221
(0.054)
(0.475)
(0.044)
(0.208)
0.022**
0.000
0.009
-0.001
(0.013)
(0.993)
(0.304)
(0.906)
0.392**
-0.043
0.297*
-0.094
(0.024)
(0.808)
(0.089)
(0.571)
0.338**
0.504***
0.361***
0.414
(0.014)
(0.003)
(0.003)
(0.106)
Yes
Yes
Yes
Yes
Diff in Coeff of SocialCapital p value
(0.013)
(0.048)
Industry dummies
Yes
Yes
Yes
Yes
Year dummies
Yes
Yes
Yes
Yes
Observations
39,012
39,151
40,208
41,615
Pseudo R2
0.0632
0.0993
0.1001
0.0817
Prob [ v2
0.000
0.000
0.000
0.000
123
Financial Reports and Social Capital Table 1 contiued (1) DV = FinclFraud Panel G: Propensity to commit fraud is lower for firms that relocate to a county with higher social capital compared to those that locate to one with lower social capital Post
0.004 (0.166)
SocialCapital_Increasing_Move
0.010 (0.116)
Post * SocialCapital_Increasing_Move
-0.012* (0.052)
ln(MarketValue)
0.002*** (0.005)
Analysts
-0.000 (0.845)
ReturnOnAssets
0.001 (0.747)
DebtToAssets
-0.000 (0.653)
Big4
-0.002 (0.210)
MarketToBook
0.000 (0.248)
Loss
-0.001 (0.589)
VolatalityOfCashFlow
0.004 (0.284)
County-level controls
Yes
Industry fixed effect
X
Year fixed effect
X
Observations
9000
R2
0.012
This table reports the data summary and regression results of the main test: Does social capital affect the propensity to commit financial fraud? The table consists of Panels A, B, C, D, E, F, and G. Panel A reports the summary statistics of the sample. Panel B reports the Pearson correlations—the ones in bold are significant at the 5% level. Panel C reports the coefficients for the logit model with the dependent variable FinclFraud. In column 6 of Panel C, I report results of a collapsed sample. I collapse the data so that each firm only has one observation. I consider the firm to have committed financial fraud if it does so in any of the years between 1990 and 2009. For all other variables, I take the median of the sample. Panel D shows that the probability of the CEO being replaced after a committing fraud is higher in a high-social-capital region. Panel E shows that the effect of social capital on reducing the propensity to commit fraud appears stronger when external monitoring is weak. The two measures for external monitoring are the debt to assets ratio and the percentage of institutional investors. Panel F reports the results of a propensity score matching. Panel G reports the results of an analysis involving firms that change headquarters. This panel shows that the propensity to commit fraud become lower when a firm moves to a higher-social-capital county. The coefficients reported are that of a linear probability model. The p values are in the brackets. They are based on the robust standard errors clustered at the county level. The ***, **, and * represent significance at the 1, 5, and 10% levels, respectively. The variable descriptions are in ‘‘Appendix 1’’ section. All continuous variables are winsorized at the 1st and the 99th percentile
statistically significant (p value is 0.708). When I remove SocialCapital from the specification, I still do not find this variable to be significant. I discuss this issue in more detail in ‘‘Social Capital Measures Different Aspects of Social Norm Than Religiosity’’ section. The signs of the control variables are largely in line with our prediction, albeit not always significantly. As expected the sign of Analysts is negative and significant that means external monitoring might constrain fraud. Also, as expected, the coefficient for VolatalityOfCashFlow is positive and significant that means the volatile cash flow
might be an indication of turbulence in the firm, which is associated with a greater proclivity to misbehave.13 The Main Results are Robust When Using Propensity Score Matching I also use the propensity matching technique as in Rosenbaum and Rubin (1983), which is used in a number of studies 13
Yu (2008) finds a negative association between analysts following and discretionary accruals, and Jiang et al. (2010) find a positive association between cash flow volatility and discretionary accruals.
123
A. Jha
in the finance and accounting literature (see for, e.g., Fang et al. 2014; Hilary and Huang 2015). This technique ensures that my results are not sensitive to the relation I assume between the covariates and the dependent variable. Thus, I divide the sample into two groups based on the median level of social capital: the group with high social capital (treated group) and that with low social capital (control group). For each of the observations in the treated and control groups, I use a logit model to calculate the propensity score—the probability of belonging to a high-social-capital region. I use the following firm-level variables: ln(MarketValue), Analysts, ReturnOnAssets, DebtToAssets, Big4, MarketToBook, Loss, VolatilityOfCashflow, and an industry fixed effect when constructing the propensity scores.14 Then for each observation from the treated sample, I find the nearest neighbor. This match comes from the control group for which the absolute value of the difference in propensity scores is the minimum. I match with a replacement. I call this sample the matched sample. Next, I test whether there is a statistically significant difference in FinclFraud between firms in the treated group and those in the matched group. The findings remain the same. The propensity to commit fraud for firms in a low-social-capital region are quite high compared to the firms in a high-social-capital region. The results are reported in Panel D of Table 1. The table compares the summary statistics between the treated group and the matched sample constructed from the control group. The mean of FinclFraud in the treated group (firms in a highsocial-capital region) is 0.0042, compared to a mean of 0.0055 in the matched sample (firms from a low-social-capital region). Put differently, the mean of FinclFraud is about 34% higher for firms in a low-social-capital region. The difference is significant at 1%. The comparison also shows that other firm-level variables are in most cases virtually the same. The Punishment for Committing Fraud is Harsher is a High-Social-Capital Region A key argument I make is that in a high-social-capital region the norms and networks are such that there is a greater punishment for deviant behavior. Stakeholders of the firm are more likely to harshly punish managers that misbehave. To examine whether this is the case, I collect data on the CEOs’ identity from Execucomp and use the date on which a firm’s possible fraud was first disclosed to the public to construct two indicator variables.15 The first
variable, CEOReplaced equals one if the CEO is replaced in the following 2 years after the disclosure. The second variable, FraudTriggerYear, equals one for the year in which investors were made aware of the firm’s possible fraud. Next, I split the sample into two groups based on the median level of social capital and use a logit model to examine the association between FraudTriggerYear and CEOReplaced. I report the results in panel E of Table 1. Consistent with the idea that punishment is harsher in a high-social-capital region, I find that the odds of a CEO being replaced within 2 years in a high-social-capital county are double those in a low-social-capital county (exp(1.382)/exp(0.567) = 2.26). The differences in the coefficients for FraudTriggerYear are statistically significant. Because I use a logit model, the comparison of the coefficients between the two groups is not straightforward (Ai and Norton 2003; Allison 1999). Allison (1999) points out that a Wald Chi-square test is misleading if residual variation differs across the groups. Williams (2009) acknowledges this complexity in testing the difference across groups in a binary regression such as a logit model and suggests a likelihood ratio (LR) test to address Allison’s concern. I use the approach outlined by Williams and conduct the LR tests. The p value for the difference is 0.048.16 The Effect of Social Capital on Reducing the Propensity to Commit Fraud Appears Stronger When the External Monitoring is Weak I also investigate if the effect of the SocialCapital is stronger when the external monitoring is weak. Given that the internal norms are a major component of the social capital and that higher network density encourages ethical norms, the effect of social capital should be much stronger when the monitoring is weak.17 To do so, I split the sample at the median between those that have weak or strong external monitoring. My first measure of the external monitoring is the proportion of debt. A lower proportion of debt means weaker external monitoring (Ang et al. 2000). As expected, I find that the coefficient for SocialCapital is much more negative when the financial leverage is low. My second measure of the monitoring is the percentage of institutional investors. A smaller percentage of institutional investors indicate lower 16
14
Matching based on propensity scores constructed using these firm level variables produces a matched sample with the least bias—that is, the treated and the matched sample are the most alike. However, the results are robust when I construct propensity scores based on the control variable specified in Eq. 1. 15 These dates are available in the FSR database compiled by Karpoff et al. (2012).
123
In untabulated results, I also use a linear probability model (LPM) (i.e., an OLS) and verify that the results are similar—the p values are slightly lower but still significant at 10%. 17 Consider the following example: in the presence of a strict teacher all students are likely to behave. But in the absence of a strict teacher, naughty children are more likely to misbehave. Put differently, the effects of naughtiness (intrinsic nature) are more salient when the disciplinarian (external monitoring) is weak.
Financial Reports and Social Capital
external monitoring. I find that the effect of the SocialCapital is stronger when there are fewer institutional investors, which is consistent with norms mattering more when the external monitoring is weak. Panel F of Table 1 contains these results. Overall, the differences in the coefficients for the SocialCapital are statistically significant. I use LR tests and find that the p value for the difference is 0.013 when I split the sample based on the ratio of the debt to the assets; the p value for the differences is 0.048 when I split the sample based on the ratio of institutional investors.18,19,20,21 Propensity to Commit Fraud is Lower for Firms that Relocate to a County with Higher Social Capital I also conduct a diff-in-diff analysis of firms that relocate to higher-social-capital counties to examine whether relocation is associated with a lower propensity to commit fraud. I find a positive association between this relocation and a lower propensity to commit fraud. To conduct this test, I follow Hasan et al. (2015).22 Following their study, I limit the sample of firms that changed headquarters and construct two new variables: Post and SocialCapital_Increasing_Move. Post equals one for the firm years after the firm relocates and zero otherwise. SocialCapital_Increasing_Move equals one for the 18
In untabulated results, I also use a linear probability model (LPM) (i.e., the OLS) and verify that the results are similar—the p values are slightly lower. The LPM avoids the pitfalls of the logit model when examining the difference in the coefficients across groups. For brevity, I do not report these results. 19 I also examine if the effect of social capital is stronger for firms that are further away from SEC. The idea is that the SEC’s enforcement might be weaker for firms headquartered further away from its office. To do so, I split the sample into two groups based on the median distance from the SEC. The results are qualitatively consistent with the idea that when external monitoring is weak, the effect of social capital is stronger. I find that the coefficient for social capital is negative and significant for firms that are further away, but nonsignificant for those close. However, there is no statistical difference. For brevity, I do not report the results. 20 I also find that qualitatively the effect of social capital is stronger for firms that have a higher g-index (i.e., those that are poorly monitored by their board of directors). I conduct this analysis by dividing the sample into two groups based on the median level of the g-index. I find that for the group with a high g-index the effect of social capital is large, negative, and significant but for those with a low value for the g-index, the coefficient is not significant. For brevity, I do not report the results. 21 The results are also similar when I split the sample based on the median level of analysts. For firms that have no analysts following, the effect of social capital on financial fraud is much stronger, as expected. It is nonsignificant for the group that has analysts following. 22 They conduct a similar analysis to examine if the social capital reduces the price of bank loan contracting for firms that move to a higher social capital region, compared to those that do not.
firm years when the firm moves to a county with higher social capital, and zero otherwise. Next, I conduct a regression analysis where the dependent variable is Fraud, and the control variables are the same as specified in the equation except that I replace SocialCapital with Post, SocialCapital_Increasing_Move, and the interaction term Post * SocialCapital_Increasing_Move. If moving to a higher-social-capital county is associated with a lower propensity to commit fraud, then the coefficient for Post * SocialCapital_Increasing_Move should be negative and statistically significant. Panel G of Table 1 shows that indeed the coefficient for this interaction term is negative with a p value of 0.052.
Additional Tests A natural question to ask is whether the other methods of obscuring the financial reports are also related to social capital. In this section, I discuss such tests. These tests involve the construction of new variables. In ‘‘Appendix 1’’ section, I describe in detail how these variables are constructed. Firms in High-Social-Capital Regions Have Lower Levels of Discretionary Accruals One method to obscure financial statements is the manipulation of accruals. There are many reasons why managers manipulate accruals, but the manipulation often is for personal benefit and at the cost of the firm’s long-term economic performance (Healy and Wahlen 1999). The managers themselves consider this type of management to be unethical (Graham et al. 2005). Because the regions with high social capital have altruistic norms and dense networks and because both characteristics discourage selfish behavior, it is reasonable to expect that the level of discretionary accruals will be lower for firms that are headquartered in these regions. Therefore, I test whether this expectation holds or not. As expected, my results show that the firms headquartered in counties with high social capital have lower levels of discretionary accruals. The results are in Table 2. The dependent variable in columns 1–3 is |DiscAccrual|, which is the absolute value of discretionary accruals calculated by using the modified Jones model. The control variables are based on McGuire et al. (2012). The coefficient for SocialCapital in column 1 is -0.015 and significant at the 1% level. The economic significance is also quite large: A one standard deviation increase in the social capital is associated with about a 5.7% lower |DiscAccrual| than the mean level of the discretionary accruals (-0.015 9 0.850/
123
A. Jha Table 2 Additional tests: firms in high-social-capital counties have lower levels of discretionary accruals (1)
(2)
(3)
(4)
DV = |DiscAccrual|
(5)
(6)
DV = DechowDichevAQ
DiscAccrual [ 0
DiscAccrual \ 0
One Obs. per firm
-0.017***
-0.012***
-0.015***
One Obs. per firm
SocialCapital
-0.015*** (0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
ln(MarketValue)
0.000 (0.939)
-0.005*** (0.000)
0.005*** (0.000)
-0.005*** (0.002)
-0.005*** (0.000)
-0.007*** (0.000)
Analysts
-0.002***
-0.002***
-0.002***
-0.000
-0.001***
-0.000
(0.000)
(0.000)
(0.000)
(0.315)
(0.000)
(0.320)
-0.075***
0.034***
-0.121***
-0.139***
-0.008*
-0.070***
(0.000)
(0.001)
(0.000)
(0.000)
(0.089)
(0.000)
0.073***
0.075***
0.070***
0.072***
0.025***
0.019***
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.007)
-0.043***
-0.036***
-0.045***
-0.035***
-0.025***
-0.026***
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
0.001*
0.001***
0.000
-0.001
0.001***
0.001
(0.081)
(0.006)
(0.815)
(0.402)
(0.000)
(0.246)
0.018***
0.016***
0.045***
0.009*
0.016***
0.015***
(0.000)
(0.000)
(0.000)
(0.093)
(0.000)
(0.000)
VolatalityOfCashFlow
0.109***
0.132***
0.096***
0.081***
0.124***
0.085***
(0.000) -0.021***
(0.000) -0.016***
(0.000)
(0.000)
(0.000)
(0.000)
Benchmark
-0.030***
-0.005
-0.008***
-0.015
(0.000)
(0.000)
(0.000)
(0.693)
(0.000)
(0.249)
ln(AuditorTenure)
-0.005***
-0.003
-0.007***
-0.009**
-0.005***
-0.010***
(0.004)
(0.137)
(0.001)
(0.040)
(0.000)
(0.002)
-0.001
0.002
-0.005***
-0.009***
-0.001
-0.003
(0.441)
(0.110)
(0.000)
(0.004)
(0.176)
(0.319)
0.059***
0.037***
0.078***
0.082***
0.027***
0.061***
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
0.061***
0.060***
0.026**
0.134***
0.003
0.064***
(0.000)
(0.000)
(0.049)
(0.000)
(0.699)
(0.000)
0.026***
0.035***
0.009
0.013
0.006*
0.007
(0.000)
(0.000)
(0.168)
(0.118)
(0.063)
(0.122)
County-level controls
Yes
Yes
Yes
Yes
Yes
Yes
Industry dummies
Yes
Yes
Yes
Yes
Yes
Yes
Observations
82,148
45,014
37,134
9992
57,635
7052
0.259
0.157
0.386
0.524
0.345
0.438
ReturnOnAssets DebtToAssets Big4 MarketToBook Loss
ChangeInGDP Investment NetOperatingAssets Rural
2
R
-0.007***
-0.009***
This table reports the coefficients from the OLS. The DV refers to the dependent variables. The county-level controls are the same as in Table 1. The p values are in the brackets. They are based on the robust standard errors clustered at the county level. The ***, **, and * represent significance at the 1, 5, and 10% levels, respectively. The variable descriptions are in ‘‘Appendix 1’’ section. All continuous variables are winsorized at the 1st and the 99th percentile
0.224). To put this number in perspective, a one standard deviation increase in the number of analysts is associated with about 4.8% lower discretionary accruals (-0.002 9 5.318/0.224). The coefficient for SocialCapital continues to be significant when I limit the sample to either those firms that manage earnings upwards (column 2) or downwards
123
(column 3).23 I continue to find that the coefficient for SocialCapital is negative when I collapse the data to only one 23
I use the absolute value of the discretionary accruals rather than the signed value because the managers have an incentive to manage earnings upwards as well as downwards. However, in unreported tests I verify that the coefficient for SocialCapital is also significant when I use the signed value of the discretionary accruals.
Financial Reports and Social Capital
observation (column 4). The coefficient for SocialCapital is also negative and significant in columns 5 and 6. In these columns, instead of the modified Jones model, I use DechowDichevAQ, as a measure of the quality of the accruals. This measure is based on Dechow and Dichev (2002) and is constructed as in Francis et al. (2005).24 Firms in High-Social-Capital Regions Have More Readable Annual Reports Managers opportunistically manage not only the content of the balance sheet and income statement but also the readability of their annual reports. For example, Li (2008) shows that when the current performance is poor, then the annual reports are difficult to read. The readability is frequently used as an alternative measure for a financial report’s quality (Biddle et al. 2009; Callen et al. 2012; Lee 2012; Lehavy et al. 2011). Making financial reports less readable comes at a cost to other stakeholders. It makes extracting information time consuming and mentally taxing (Bloomfield 2002). Even the experts have difficulty: Lehavy et al. (2011) find that firms with less readable annual reports have more analysts following, and the analysts’ estimates are more dispersed. Poor readability can lead to suboptimal decisions. Lee (2012) finds that when the 10-K annual reports of the firms are less readable, less of the earnings information is embedded in the stock prices. Lawrence (2011) finds that the individual investors are wary of investing in the firms with less readable annual reports—and when the individual investors do invest in these firms, they perform poorly. I expect that firms with high social capital have more readable annual reports because the managers can choose whether to make the annual reports readable and because investors lose when the annual reports are less readable. Therefore, I use three measures of readability. The first two are the fog index and the number of words in the annual report from Li (2008). For the third measure, I use the size of the file in megabytes obtained from Loughran and Mcdonald (2014). These authors show that the size of the file is a better measure of the readability for a sample after 2005. Therefore, following their method, I only use data up to 2005 when using the fog index and the number of words. The results are in Table 3 and are consistent with the idea that the readability of annual reports is better when the firm is headquartered in a county with high social capital. The coefficient for SocialCapital is negative and significant at the 1% level for all three different measures of readability. The significance does not appear to be due to the large sample size. They continue to hold if I take the 24
The results are similar when I use the performance adjusted discretionary accruals as suggested in Kothari et al. (2005).
median of all variables so that there is only one observation per firm and repeat the tests. Exploiting Shocks to Assess How Social Capital Affects Readability and Discretionary Accruals Two changes in the regulations allow me to conduct additional tests that alleviate this concern even further. The first shock is the plain English disclosure guidelines that the SEC issued in October 1998. These guidelines require that the managers of public firms should comply with six basic principles of presenting information in annual reports. These principles include writing short sentences and using concrete, definitive, and everyday language. If the social capital does capture the manager’s propensity to write less foggy annual reports, then the effect of the social capital should be much stronger before 1999—in the weaker legal regime. To test this idea, I split the sample into a pre-1999 group and a post-1999 group and examine if SocialCapital is stronger before 1999. I find that this is indeed the case—the effect of the social capital on the fogginess of the annual reports is stronger before 1999. The results in Table 4 show that the coefficient for SocialCapital is -0.152 with a p value of \0.001 before 1999 (column 1) and only -0.068 with a p value of 0.011 after 1999 (column 2). The Wald Chi-square test shows that the difference is significant at the 1% level. The second regulation is the SOX of 2002 that abruptly changed the incentives to manage the accruals. The research shows that before SOX, the use of accruals was more extensive. After SOX, because of greater scrutiny from regulators and greater fear of prosecution, managers shifted from accrual management to real earnings management. This shock means that the pre-SOX period can be considered as a weak legal regime (Cohen et al. 2008). Therefore, the impact of the ethical norms such as the social capital is likely to be stronger before 2002. Thus, I split the sample into pre-2002 and post2002 and examine the effect of social capital on the absolute value of the discretionary accruals. I find that the coefficient for SocialCapital is indeed much stronger in the pre-SOX period. The difference between the coefficients for SocialCapital between the two groups is significant at the 1% level. The results are reported in columns 3 and 4 of Table 4.
Discussion Social Capital Measures Different Aspects of Social Norm Than Religiosity A natural concern is whether I only capture the effect of a religious norm that is operating through social capital or capture a norm that is largely independent of the religious norm. This skepticism is compounded by studies that show
123
A. Jha Table 3 Additional tests: Higher social capital is associated with more readable annual reports (1)
(2)
DV = ln(SizeMegabytes)
(3)
(4)
DV = FogIndex
One obs. per firm SocialCapital
(5)
(6)
DV = NumofWords One obs. per firm
One obs. per firm
-0.025**
-0.115***
-0.092***
-1632.529***
-0.100***
-1690.512***
(0.020)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
ln(MarketValue)
0.086*** (0.000)
0.170*** (0.000)
0.007 (0.370)
1909.564*** (0.000)
0.020** (0.040)
2377.604*** (0.000)
ln(NumberOfSubs)
0.070***
0.113***
-0.053***
1123.623***
-0.074***
473.970
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.123)
-0.003**
-0.020***
-0.002
-76.657*
-0.010**
-229.473***
Analysts ReturnOnAssets DebtToAssets
(0.016)
(0.000)
(0.417)
(0.058)
(0.012)
(0.000)
-0.034***
-0.069**
-0.126***
-1068.312***
-0.043
-1810.267**
(0.010)
(0.021)
(0.000)
(0.009)
(0.525)
(0.015)
0.051***
0.052**
0.021
3863.117***
0.045
3342.543***
(0.000)
(0.016)
(0.273)
(0.000)
(0.265)
(0.000)
0.109***
-0.070**
0.054**
1397.316***
0.034
798.422*
(0.000)
(0.011)
(0.049)
-0.001
(0.312)
(0.087)
-0.004***
-0.012***
0.002
-76.748***
0.011**
-165.995***
(0.000)
(0.000)
(0.248)
(0.001)
(0.038)
(0.001)
-0.006
0.190***
-0.028
-147.622
-0.029
112.340
(0.804)
(0.000)
(0.543)
(0.789)
(0.553)
(0.844)
County-level controls
Yes
Yes
Yes
Yes
Yes
Yes
Industry Dummies
Yes
Yes
Yes
Yes
Yes
Yes
Year Dummies
Yes
Yes
Yes
Yes
Yes
Yes
Observations
40,854
7011
35,573
35,573
6221
6221
R2
0.593
0.363
0.075
0.083
0.106
0.129
Big4 MarketToBook Rural
This table reports the coefficients from the OLS. The DV refers to the dependent variables. The county-level controls are the same as in Table 1. The p values are in the brackets. They are based on the robust standard errors clustered at the county level. The ***, **, and * represent significance at the 1, 5, and 10% levels, respectively. The variable descriptions are in ‘‘Appendix 1’’ section. All continuous variables are winsorized at the 1st and the 99th percentile
that religiosity affects managerial decisions—including decisions pertaining to the financial reports’ quality. A closer look at the recent literature on social capital and religiosity and additional empirical tests indicates that the effect of social capital goes beyond that of religiosity and that these two measures capture different aspects of the social norm. Religiosity measures the faith in God and the attendance of religious activities. In contrast, social capital measures the propensity to honor obligations and other norms that facilitate cooperation, such as honesty and mutual trust. The suggestion that a religious norm is the main basis of social capital in effect means that religious individuals mainly have a greater propensity to honor their obligations and to place greater trust in each other. This view is at odds with the current theoretical and empirical research (Berggren and Bjørnskov 2011). Berggren and Bjørnskov (2011), based on their cross-country research, show that ‘‘those who do not believe literally in the bible are clearly
123
more trusting than believers.’’ This finding means that extremely religious individuals might actually be less trusting of each other. True, studies view religious activities as a source of social capital (deTocqueville 1835; Coleman 1988; Smidt 1999, 2003), but they are not the only source of social capital and possibly today only a minor source. The empirical studies support this idea. In an international setting, McCleary and Barro (2006) find no statistical association between trust and religious beliefs such as heaven, hell, and the afterlife. Berggren and Bjørnskov (2011) find a negative association. Specifically, they note: We make use of new data from the Gallup World Poll for 109 countries and 43 U.S. states. Our empirical results indicate a robust, negative relationship between this measure of religiosity and trust, both internationally and within the U.S.
Financial Reports and Social Capital Table 4 Exploiting shocks to assess how social capital affects readability and discretionary accruals
(1)
(2)
DV = FogIndex
(3)
(4)
DV = |DiscAccrual|
Year \ 1999
Year C 1999 & B2005
Year \ 2002
Year C 2002
-0.152***
-0.068**
-0.019***
-0.006
(0.000)
(0.011)
(0.000)
(0.145)
Firm-level controls
Yes
Yes
Yes
Yes
County-level controls
Yes
Yes
Yes
Yes
SocialCapital
Industry dummies
Yes
Yes
Yes
Yes
Year dummies
Yes
Yes
No
No
Diff in Coeff of SocialCapital p value (0.0080) Diff in Coeff of SocialCapital p value
(0.0057)
Observations
12,419
23,154
51,649
30,499
R2
0.091
0.072
0.201
0.327
The purpose of this table is to show that the effect of social capital is stronger in a weak legal regime. The period before 1999 is identified as a weak legal regime when it comes to writing foggy annual reports. Similarly, the pre-SOX period is identified as a weak legal regime for managing accruals. Columns 1 and 2 report the regression coefficients when the dependent variable is FogIndex. Column 1 reports the results for the sample period before 1999 (1994, 1995, 1996, 1997, and 1998). Column 2 reports the results for the sample period from 1999 to 2005. The control variables (i.e., firm-level, county-level, industry dummies, and year dummies) are as in Table 3. Columns 3 and 4 report the regression coefficients when the dependent variable is |DiscAccrual|. Column 3 reports the results the sample period before 2002 (1990–2001). Column 4 reports the results for the sample period from 2002 to 2009. The control variables (i.e., firm-level, county-level, industry dummies, and year dummies) are the same as in Table 2. The p values are in the brackets. They are based on the robust standard errors clustered at the county level. The ***, **, and * represent significance at the 1, 5, and 10% levels, respectively. The variable descriptions are in ‘‘Appendix 1’’ section. All continuous variables are winsorized at the 1st and the 99th percentile
In fact, Smidt (1999), who argues that religiosity is a source of social capital, finds that religious faith only moderately affects social capital. The key idea in the literature is not that the religious norms that are preached affect social capital but instead that religious congregation provides its members an opportunity to organize and cooperate. The idea is that repeated meetings and organizing hones the skills to cooperate and develops cooperative norms that can extend beyond religious activities (Smidt 1999). But in the last half century, religiosity in general and religious adherence in particular has declined (Chaves 2011). There are many organizations besides churches and temples where people meet and develop the ideals and the skills of high social capital. More recent research corroborates these findings. For example, Alesina and La Ferrara (2002) find that denominational membership has no appreciable impact on trust. Welch et al. (2007) use a multivariate framework of English-speaking residents of the USA who are 18 years and older and find that religiosity, even when measured by membership and the degree of activity in the church, is not associated with how much a person trust others—the key metric for social capital. Specifically, the authors note:
Although several religious measures initially display statistically significant net relationships to one’s trust in acquaintances, none of these relationships remain significant once all measures are included in the model. Only age, educational level, political ideology and race display any statistically significant net relationship to the measure of trust One way to empirically establish that Religiosity and SocialCapital do not capture similar norms in my context is to test whether in a multivariate framework, the coefficient for SocialCapital alters significantly depending on whether Religiosity is included in the specification. When I rerun my main model [specified in Eq. (1) and presented in column 1 of Panel C in Table 1] without Religiosity, the coefficient for SocialCapital hardly changes—this finding indicates that Religiosity and SocialCapital measure different aspects of the propensity to commit financial fraud.25 Similarly, when I keep Religiosity but remove 25 Jha and Chen (2015) conduct a similar exercise with audit fees and find that in a multivariate framework, the effect of social capital hardly changes when religiosity is removed as one of the control variables in a regression where the dependent variable is the natural logarithm of the audit fees.
123
A. Jha
SocialCapital from the specification, I do not find that Religiosity has a significantly negative association with FinclFraud. This result further confirms that social capital and religiosity have different sets of norms. These results are reported in columns 1–3 of Panel A in Table 5.26 Social Capital Does Not Significantly Moderate the Effect of Religiosity I have discussed why social capital and religiosity measure different aspects of norms. This discussion suggests that social capital does not mediate the effect of religiosity. But arguably, social capital might also moderate the effect of religiosity. That is, the effect of religiosity might be stronger in counties with high social capital compared to counties with low social capital: The counties with higher social capital have dense networks in which the deviations from the religious norms might mean harsher punishment. One way to test the degree of punishment is to divide the sample between the two types of firms and compare whether the coefficients for religiosity are different between the two groups. When I do so, I find that the coefficient for Religiosity is 0.715 with a p value of 0.124 when the social capital is low, and a coefficient of -0.724 with a p value of 0.203 when the social capital is high. The fact that the coefficient switches signs makes it appear that the effect of Religiosity might be different in counties with high social capital than in counties with low social capital. But the difference is not statistically significant.27 This significance indicates that the moderating effect is not significant. One reason might simply be that the net effect of Religiosity on FinclFraud, if any, might be quite low.28 Panel B in Table 5 has the results. The Results are not Driven by the Political Activism of the County in Which the Firm is Headquartered Two recent studies raise the possibility that my results could be driven largely by political activism instead of social capital. Bonaparte and Kumar (2012) find that in politically active regions, measured by the electoral participation rate, people are more likely to participate in the stock market, which indirectly suggests that these regions also have more investment clubs. Chhaochharia et al. (2012) find that when there are more local investment clubs, the earnings 26
Column 1 presents the coefficients when both SocialCapital and Religiosity are included (it is the same as Column 1 of Panel C in Table 1). I present it here again to make it easier to compare. 27 Neither a likelihood ratio test as in Allison (1999) nor a test using an interaction term for a dummy of the high social capital and religiosity in a LPM shows that the difference is statistically significant. 28 The control variables include SocialCapital. But the results are similar if I exclude SocialCapital.
123
management is lower because the institutional investors can better monitor. Because the electoral participation rate is one of the four components that I use in constructing SocialCapital, the concern is that I might be capturing the effect of political activism instead of the social capital. To address this concern, I construct another measure of social capital that this time excludes the election participation variable from the principal component analysis. I label the variable the SocialCapital(NoElection). The results continue to hold—I find that when firms are located in a county with high social capital, the propensity to commit fraud and manage accruals is lower, and the propensity to make annual reports readable is much higher. The results are in Table 6. Whether the Effect of Social Capital is Stronger for Less Geographically Dispersed Firms is Unclear Some extent of the earnings management also occurs at the subsidiaries, not just the head office (Gunny 2010). Moreover, for firms that are less geographically dispersed, their location’s social norm and the managers’ norm are much more congruent. This congruence means that the impact of social capital should be more pronounced for firms that are less geographically dispersed. To test this idea by using the approach of McGuire et al. (2012), I divide the sample into two groups based on the median number of subsidiaries and test if the effect of SocialCapital differs between the groups. The results in Table 7 are mixed. Except when FinclFraud is the dependent variable, the effect of the social capital based on the point estimates appears stronger; however, the Wald Chi-square test shows that only in the case of the discretionary accruals are the differences statistically significant. Managers Switch Jobs to Places with Similar Social Capital One of the implicit assumptions in my study is that the social capital of the upper management is similar to the social capital of the place where the firm is headquartered. This assumption is strengthened if I show that the social capital plays a role in the selection of the CEO. To do so, I examine the 39 instances in my sample where a CEO switched jobs. I test whether the previous county’s social capital is related to that of the new county. I find that to be the case. These results are in Table 8. The dependent variable in these regressions is the NewSocialCapital, the new county’s social capital; and the key research variable is the OldSocialCapital, the old county’s social capital. I find that the NewSocialCapital is positively associated with the OldSocialCapital, which means that a CEO is likely to move to a county with a similar level of social capital. The results are similar to the findings of Hilary and Hui (2009).
Financial Reports and Social Capital Table 5 Relative effects of social capital and religiosity on the propensity to commit financial fraud (1)
(2)
(3)
DV = FinclFraud Panel A: Social capital measures the different aspects of social norms and religiosity—evidence from a multivariate framework SocialCapital Religiosity
-0.242***
-0.248***
(0.004)
(0.003)
-0.125
-0.363
(0.708)
(0.290)
Firm-level controls
Yes
Yes
Yes
County-level controls
Yes
Yes
Yes
Industry dummies
Yes
Yes
Yes
Year dummies
Yes
Yes
Yes
Observations
85,743
85,743
85,743
Pseudo R2
0.0753
0.0753
0.0737
Prob [ v2
(0.000)
(0.000)
(0.000)
(1)
(2)
DV = FinclFraud SocialCapital B median
SocialCapital [ median
Panel B: Social capital does not moderate the effect of religiosity on the propensity to commit financial fraud Religiosity
0.715
-0.724
(0.124)
(0.203)
SocialCapital
-0.377**
-0.183
(0.025)
(0.437)
Firm-level controls
Yes
Yes
County-level controls
Yes
Yes
Industry dummies
Yes
Yes
Year dummies
Yes
Yes
Difference in the coefficient of religiosity between the two groups (p value)
(0.428)
Observations
41,843
39,655
Pseudo R2
0.072
0.067
Prob [ v2
(0.000)
(0.000)
This table consists of two panels. Columns 1 and 2 of Panel A report the results of a logit model with and without Religiosity as the control variable, respectively. Column 3 removes SocialCapital from the specification. The control variables (i.e., firm-level, county-level, industry dummies, and year dummies) are the same as in Panel C of Table 1. They are not reported for economy of space. The p values are in the brackets. They are based on robust standard errors clustered at the county level. The ***, **, and * represent significance at the 1, 5, and 10% levels, respectively. The variable descriptions are in ‘‘Appendix 1’’ section. All continuous variables are winsorized at the 1st and the 99th percentile
The Norms Appear to Have a Stronger Impact Than the Network In ‘‘Measuring Social Capital’’ section, I discuss why the difficulty in distinguishing between the norms and the network in social capital. Now, I examine which has a stronger effect in constraining financial fraud. To this end, I construct Norm, the first component of a principal component analysis that uses Electoral participation to measure the participation in a presidential election and Census response rate that measures the percentage of people that return filled-out census forms. I also construct Network that is the first component of a principal component analysis of the following two variables: # of NGOs, the number of non-government
organizations normalized by the population in the county; and # of Associations, the number of associations normalized by the population. The result, which I report in Panel A of ‘‘Appendix 2’’ section, shows that based on our measures, norms might play a stronger role than the network. The Common Component of the Four Measures of Social Capital has a Stronger Effect on Financial Fraud I also estimate the main model by using each component of social capital, Electoral participation, Census response rate, # of NGOs, and # of Associations to assess which elements are driving the results. The results reported in Panel B of
123
A. Jha Table 6 Results are not driven by the political activism in the county in which the firm is headquartered
SocialCapital(NoElection)
(1) FinclFraud
(2) |DiscAccrual|
(3) DechowDichevAQ
(4) ln(SizeMegabytes)
(5) FogIndex
(6) NumofWords
-0.248*
-0.016***
-0.008***
-0.216***
-0.123***
-1522.176*** (0.000)
(0.057)
(0.000)
(0.000)
(0.000)
(0.000)
Firm-level controls
Yes
Yes
Yes
Yes
Yes
Yes
County-level controls
Yes
Yes
Yes
Yes
Yes
Yes
Industry dummies
Yes
Yes
Yes
Yes
Yes
Yes
Year dummies
Yes
No
No
Yes
Yes
Yes
Observations (Pseudo) R2
85,743 0.0746
82,148 0.258
57,635 0.345
40,854 0.378
35,573 0.079
35,573 0.104
Prob [ v2
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
The purpose of this table is to show that the social capital’s effect on the financial reports’ quality is not driven by electoral participation in the county where the firm is located. The table reports the results of a regression analysis. The key independent variable SocialCapital is replaced by SocialCapital(NoElection), a measure of social capital constructed without using the participation rate in the election. The control variables (i.e., firm-level, county-level, industry dummies, and year dummies) in column 1 are the same as in column 1 of Panel C in Table 1; those in columns 2 and 3 they are the same as in Table 2; and those in columns 4, 5, and 6 are the same as in Table 3. The p values are in the brackets. They are based on the robust standard errors clustered at the county level. The ***, **, and * represent significance at the 1, 5, and 10% levels, respectively. The variable descriptions are in ‘‘Appendix 1’’ section. All continuous variables are winsorized at the 1st and the 99th percentile Table 7 It is unclear whether social capital has a stronger effect for less geographically dispersed firms (1)
(2)
(3)
DV = FinclFraud
(4)
(5)
DV = |DiscAccrual|
(6)
DV = DechowDichevAQ
Subsd B 2
Subsd [ 2
Subsd B 2
Subsd [ 2
Subsd B 2
Subsd [ 2
-0.226
-0.240***
-0.019***
-0.010**
-0.010***
-0.005***
(0.122)
(0.006)
(0.000)
(0.015)
(0.000)
(0.005)
Panel A SocialCapital Coeff. of SocialCapital differs between the two groups (p value)
(0.721)
(0.030)
(0.062)
Control variables
Yes
Yes
Yes
Yes
Yes
Yes
Observations
44,246
37,841
44,787
37,361
28,582
29,053
(Pusedo) R2
0.064
0.107
0.331
0.173
0.371
0.317
(1)
(2)
ln(SizeMegabytes)
(3)
(4)
FogIndex
(5)
(6)
NumofWords
Subsd B 2
Subsd [ 2
Subsd B 2
Subsd [ 2
Subsd B 2
Subsd [ 2
-0.027**
-0.016
-0.111***
-0.070**
-1614.037***
-1401.003***
(0.014)
(0.237)
(0.000)
(0.014)
(0.000)
(0.000)
Panel B SocialCapital Coeff. of SocialCapital differs between the two groups (p value)
(0.451)
(0.182)
(0.690)
Control variables Observations
Yes 15,009
Yes 25,845
Yes 16,364
Yes 19,209
Yes 16,364
Yes 19,637
R2
0.553
0.554
0.091
0.077
0.100
0.103
This table consists of two panels. Columns 1 and 2 of Panel A represent the results of a logit model. The rest of the results are based on the OLS. The control variables (i.e., firm-level, county-level, industry dummies, and year dummies) correspond to the respective controls variables based on the dependent variable. In Panel A, the control variables in columns 1 and 2 are the same as in column 1 of Panel C in Table 1; those in columns 3, 4, 5, and 6 they are the same as in Table 2. In Panel B, the control variables are the same as in Table 3. The p values are in the brackets. They are based on the robust standard errors clustered at the county level. The ***, **, and * represent significance at the 1, 5, and 10% levels, respectively. The variable descriptions are in ‘‘Appendix 1’’ section. All continuous variables are winsorized at the 1st and the 99th percentile
123
Financial Reports and Social Capital Table 8 CEOs are likely to move to a firm headquartered in a location with a similar level of social capital (1) (2) DV = NewSocialCapital OldSocialCapital
(3)
0.592***
0.767**
1.030**
(0.005)
(0.028)
(0.012)
OldReligiosity
0.711 (0.624)
0.447 (0.756)
Oldln(Population)
0.073
0.164
OldIncomePerCapita OldPrctMinority OldLiteracy OldMedianAgeOfResident
(0.721)
(0.473)
-0.000
-0.000
(0.973)
(0.293)
-1.302
-2.379
(0.472)
(0.250)
-0.011
-0.064
(0.833)
(0.273)
-0.082
-0.042
(0.372)
(0.642)
NewReligiosity
-1.772
Newln(Population)
-0.489*
NewIncomePerCapita
(0.087) 0.000*
(0.370)
(0.096) NewPrctMinority
3.133 (0.246)
NewLiteracy
0.087 (0.128)
NewMedianAgeOfResident
-0.082 (0.290)
Observations
39
39
39
R2
0.194
0.252
0.500
This table reports the regression coefficients where the dependent variable is the NewSocialCapital, which is the social capital of the new county the CEO moves to. The prefix Old in the control variable refers to the county the CEO moves from. The Religiosity is the religiosity of the county, ln(Population) is the natural logarithm of the population in the county, IncomePerCapita is the income per capita, PrctMinority is the percentage of minorities in the county, Literacy is the literacy of the county, and the MedianAgeOfResident is the median age of the residents. The ***, **, and * represent significance at the 1, 5, and 10% levels, respectively. The variable descriptions are in ‘‘Appendix 1’’ section. All continuous variables are winsorized at the 1st and the 99th percentile
‘‘Appendix 2’’ section show, as expected, that all four measures have coefficients with negative signs. However, only the coefficient for Electoral participation is statistically significant. Thus, this variable is the common component for all of the measures and therefore captures the social capital in a county the best (Rupasingha et al. 2008).
State-Fixed Effects There is limited variation in the social capital within a state. And the inclusion of state-fixed effects in the model can take away a large part of social capital’s effect on manager’s behavior. Regardless, in Panel C of ‘‘Appendix 2’’ section, I report the results from adding state-fixed effects. As expected, the significance level decreases, but the signs of SocialCapital remain negative.
Conclusion I investigate whether the social capital of the region in which the firm is headquartered affects the quality of its financial reporting. My research question is motivated by two independent streams of the literature. One, the social capital literature shows that the regions with high social capital have more altruistic norms and denser networks, both of which induce people to fulfill their obligations. Two, the managerial decision literature finds that the culture of the place where the firm is headquartered affects managerial decisions. Based on these two streams of literature, I propose that the quality of the financial reports is higher when firms are headquartered in US counties with high social capital. As expected, the results confirm my hypothesis. These firms have a lower propensity to be prosecuted for financial fraud. I find that firms in counties with high social capital also have lower levels of discretionary accruals and much more readable annual reports. These results are robust to an extensive set of control variables and alternative measurements. Additional tests also indicate that the relation between the social capital and the financial reports might be causal and not just correlational. To my knowledge, this is the first paper to conduct a comprehensive study that examines the association between the social capital of the region where the firm is headquartered and the quality of its financial report. These results have implications that go beyond earnings management. Other sorts of corporate decisions, particularly those that could be considered managerial misbehavior, are likely to be lower for firms that are headquartered in regions with high social capital. I leave it to future research to examine these questions. Acknowledgements I thank Steven Dellaportas (editor) and two anonymous referees for their valuable feedback. I thank the authors who have been kind enough to share their hand-collected data: Jonathan Karpoff, Allison Koeste, Scott Lee, and Gerald Martin for sending me their financial misconduct data; and Feng Li and Bill McDonald for sharing their data on the readability of annual reports on their Web site. I thank Yu Chen, Siddharth Shankar, George Clarke, and Christopher Boudreaux for their valuable feedback. I also thank Jonathan Moore for copyediting my manuscript. Funding This study was not funded by any source.
123
A. Jha Compliance with Ethical Standards Conflict of interest There are no conflicts of interest for any authors. Ethical approval This article does not contain any studies with human participants performed by any of the authors.
Appendix 1
Dependent variables FinclFraud
This is an indicator variable that equals one if it is included in the FSR database compiled by Karpoff et al. (2012). To be included in this database, the firm needs to be guilty of financial fraud. Source: Karpoff et al. (2012)
|DiscAccrual|
This variable is the absolute value of the discretionary accruals calculated by using the modified Jones (1991) model. To calculate the discretionary accruals, I first calculate the total accruals (ibc - (oancf - xidoc)) by using the cash flow approach as suggested by Hribar and Collins (2002). To remove the effect of the outliers, I remove the observations where the ratio of the total accruals to the assets is more than two standard deviations away. Following the steps of the modified Jones (1991) model, I use this regression for each industry year: TAit Assetsit1
DSalesit PPEit 1 ¼ b0 Assets þ b1 Assets þ b2 Assets þ eit it1 it1 it1
The industry classification is based on the two-digit SIC code, and I require that there are at least 15 observations in each industry year. I also require that the firms be headquartered in the USA. The TA is the total accruals; Assets is the total assets (at); Sales is the total sales (sale); and the PPE is the gross plant, property, and equipment (ppegt). In the second step, I use the coefficient for the regression in the first step to calculate the discretionary accruals as follows: 1 DSalesit DReceivablesit ^ PPEit þ b^1 þ b2 NDAit ¼ b^0 Assetsit1 Assetsit1 Assetsit1 where the Receivables are the total TAit DISC ACCit ¼ NDAit Assetsit1 receivables (rect). Source: COMPUSTAT DechowDichevAQ
This variable is calculated following Francis et al. (2005). As in their study, I first calculate the residuals from the following regression for each industry year CFOit1 CFOit CFOitþ1 TCAit ¼ b^0 þ b^1 þ b^2 þ b^3 Assetsit2 Assetsit1 Assetsit DSales PPE it it þ b^4 þ b^4 þ eit : Assetsit1 Assetsit1 The TCA is the total accruals calculated as follows: ((act-L.act) - (lct-L.lct) - (che-L.che) ? (dlc-L.dlc))/ (L.at)). The CFO is calculated as follows: (ibc-TA), where TA is calculated as ((act-L.act) - (lctL.lct) - (che-L.che) ? (dlc-L.dlc) - (dp)). The Sales is the total sales (sale); the PPE is the gross plant, property, and equipment (ppegt); and the Assets is the total assets (at). Before running this regression and to remove the effect of the outliers, I remove the observations where the ratio of the total accruals to the assets is more than two standard deviations away. The industry classification is based on the two-digit SIC code, and I require that there be at least 15 observations in each industry-year. I also require that the firms be headquartered in the USA I then calculate the standard deviations in the residuals of the current year and the past 4 years. If this cannot be calculated due to missing values, then I replace the missing values with the standard deviations of the residuals of the current year and the past 3 years. Higher values represent higher accruals management. Source: COMPUSTAT
RealErngMgmt1
123
Following the literature (Cohen and Zarowin 2010), this variable is the sum of the abnormal discretionary expense (AB_DISEXP) and the abnormal level of production (AB_PROD). It is constructed so that the higher values represent greater real earnings management. The industry classification is based on the twodigit SIC code, and I require that there be at least 15 observations in each industry year. I also require that the firms be headquartered in the USA
Financial Reports and Social Capital
The AB_DISEXP is the residual (actual-predicted) of the following regression: DIS EXPit Assetsit1
DSalesit 1 ¼ b0 Assets þ b1 Assets þ eit ; it1 it1
where DIS_EXP is the total discretionary expenses (xsga ? xad ? xrd). Consistent with the research (Cohen et al. 2008), the advertising (xad) and the R&D expense (xrd) are set to zero if the SG&A expense (xsga) is available. The Assets variable is the total assets (at), and the Sales is the total sales (sales). To remove the effect of the outliers, I remove the observations where the ratio of the discretionary expense to the assets is more than two standard deviations away. Following their studies, I standardize the residuals The AB_PROD is the standardized value of the abnormal level of production. I first estimate the residuals (actual-predicted) of the following regression: Salesit1 DSalesit 1 it1 ¼ b0 Assets þ b1 Assets þ b2 Assets þ b3 DSales Assetsit1 þ eit where the PROD is the sum of the costs of it1 it1 it1 the goods sold and the change in inventory (cogs ? (invt - L.invt)), Assets is the total assets (at), and the Sales is the total sales (sales). To remove the effect of the outliers, I remove the observations where the ratio of the PROD to the assets is more than two standard deviations away. The residual is multiplied by -1 so that the higher values represent higher earnings management. I then standardize it. Source: COMPUSTAT
PRODit Assetsit1
RealErngMgmt2
This variable is the sum of the abnormal discretionary expenses (AB_DISEXP) and the abnormal level of cash flows (AB_CASH). The higher values represent greater real earnings management calculated as in Cohen and Zarowin (2010). The industry classification is based on the two-digit SIC code, and I require that there be at least 15 observations in each industry year. I also require that the firms be headquartered in the USA The AB_DISEXP is calculated as described above. The AB_CASH is the residual (actual-predicted) of the following regression: Salesit1 DSalesit 1 it1 ¼ b0 Assets þ b2 Assets þ b3 Assets þ b4 DSales Assetsit1 þ eit where the CFO is the cash flow from it1 it1 it1 operations (oancf-xidoc), Assets is the total assets (at), and the Sales is the total sales (sales). To remove the effect of the outliers, I remove the observations where the ratio of the CFO to the assets is more than two standard deviations away. The residual is multiplied by -1 so that the higher values represent higher earnings management. I then standardize the residuals. Source: COMPUSTAT
CFOit Assetsit1
FogIndex
This variable measures the readability of the annual reports as in Li (2008). It is constructed by Feng Li and available on his website. Source: http://webuser.bus.umich.edu/feng/
NumberOfWords
This variable measures the number of words in the annual report. It is constructed by Feng Li and available on his Web site. Source: http://webuser.bus.umich.edu/feng/
ln(SizeMegabytes)
The data is obtained from Bill McDonald’s Web site. Source: http://www3.nd.edu/*mcdonald/Word_Lists. html
Main variable of interest SocialCapital
This measures the social capital at the county level and is constructed as in Rupasingha et al. (2008). Source: Northeast Regional Center for Rural Development (NERCRD), Rupasingha et al. (2008)
Demographic controls Rrural
This is an indicator variable that takes the value of one if a firm is in a county that does not fall under the 100 most populated and zero otherwise. Core Based Statistical Area (CBSA). CBSA is a US geographic area defined by the Office of Management and Budget that consists of one or more counties (or equivalents) anchored by an urban center of at least 10,000 people. Source: Census Bureau
Religiosity
This variable is the percentage of religious adherents in the county. Source: Association of Religion Data Archive (ARDA)
ln(Population)
This variable is the natural log of the county’s population. Source: Census
IncomePerCapita
This variable is the income per capita in the county. Source: BEA
ln(DistancefromSEC)
This variable is the natural log of one plus the distance to the closest SEC branch. As in Kedia and Rajgopal (2011), the SEC branches considered are the SEC headquarters in Washington D.C and the regional offices located in New York City, NY; Miami, FL; Chicago, IL; Denver, CO; and Los Angeles, CA. As in their studies, the Haversine formula that uses the longitude and latitude of the two points is used to calculate the distance. Source: BEA
PopulationDensity
This variable is the population of the county divided by the land area of the county. Source: Census Bureau
Firm-level controls ln(MarketValue)
This variable is the market value of equity calculated by multiplying the stock’s closing price in the calendar year and the number of common shares outstanding (prcc_c * csho). Source: COMPUSTAT
123
A. Jha
Analysts
This variable is the number of analysts for the latest consensus forecast (numest). If this number is not available for a firm, then the number of analysts following is assumed to be zero. Source: I/B/E/S unadjusted summary file.
ReturnOnAssets
This variable is the ratio of earnings before the interest tax depreciation and amortization to the total assets. Source: COMPUSTAT
DebtToAssets
This variable is the ratio of the total debt to the total assets (lt/at). Source: COMPUSTAT
Big4
This variable is a binary variable that equals one if the auditor is one of the Big4 and zero otherwise. Sources: COMPUSTAT This variable is the market-to-book value of the firm. It is the ratio of the common equity to the market value (ceq/(prcc_c * csho)). Source: COMPUSTAT
MarketToBook Loss
This is an indicator variable that equals one if the income before the extraordinary items (ibc) is less than zero for the current year or the last 2 years. Source: COMPUSTAT
VolatilityOfCashflow
This variable measures the volatility of the cash flows to the total assets (oancf/at) for the current year and the last 4 years. If the lag values for all 4 years are unavailable, I construct the volatility measure using the current year and the last 3 years. Source: COMPUSTAT
Benchmark
This variable is a binary variable that equals one if (1) the net income to the lag of assets (ni/at) is greater than or equal to zero but is \0.01 or (2) the change in the net income divided by the assets from the previous year is greater than or equal to zero but \0.01, and zero otherwise. Source: COMPUSTAT
ln(AuditorTenure)
This variable is the natural log of one plus the number of years the auditor has been with the firm. Sources: COMPUSTAT
ChangeInGDP
This variable is the percentage growth rate of the real GDP compared to the previous year. Source: BEA
Investment
This variable is the ratio of the capital expenditure in year t to the net property, plant, and equipment at the end of year t - 1 (capxv/L.ppent). Source: COMPUSTAT
NetOperatingAssets
This variable is the net operating assets. It is the sum of the stockholder’s equity less marketable securities and the total debt that is scaled by the total assets ((seq - che ? lt)/at). Source: COMPUSTAT
ln(NumberOfSubs) CEOReplaced
This variable is the natural logarithm of the total number of the subsidiaries. Source: COMPUSTAT This is an indicator variable that equals one if the CEO is replaced in the following 2 years after the firm’s possible fraud was first revealed to the public. Source: Execucomp
FraudTriggerYear
This is an indicator variable that equals one for the firm year in which investors were told of possible fraud in the firm. Source: Execucomp
Post
Post is an indicator variable that equals one for the firm years after the firm relocates and zero otherwise. Source: Karpoff et al. (2012) This variable equals one for the firm year after the firm moved to a county with higher social capital, and zero otherwise. Source: Northeast Regional Center for Rural Development (NERCRD), Rupasingha et al. (2008).
SocialCapital_Increasing_Move
SocialCapital(NoElection)
This variable measures the social capital without the influence of political activism. It is calculated in the same way as the SocialCapital except that one of the norm variables (the votes cast in the presidential elections divided by the population above 18) is omitted. Source: NERCRD
Electoral participation
This variable measures participation in a presidential election
Census response rate
This variable measures the percentage of people that return filled-out census forms
# of NGOs
This variable measures the number of non-government organizations normalized by the population in the county
# of associations
This variable measures the number of associations normalized by the population
Norm
This is the first component of a principal component analysis that uses Electoral participation and Census response rate
Network
This is the first component of a principal component analysis that uses # of NGOs and # of Associations
Industry dummies
This variable is a set of binary variables constructed based on the Fama–French 48-industry grouping. Source: COMPUSTAT
123
Financial Reports and Social Capital
Appendix 2: Additional Tests
(1)
(2)
(3)
DV = FinclFraud Panel A: The norms appear to have a stronger impact than the network Norms
-0.221***
-0.211***
(0.002)
(0.003)
Network
-0.157
-0.090
(0.216)
(0.453)
Firm-level controls
Yes
Yes
Yes
County-level controls
Yes
Yes
Yes
Industry dummies
Yes
Yes
Yes
Year dummies
Yes
Yes
Yes
Observations
85,743
85,743
85,743
(Pseudo) R2
0.076
0.074
0.076
(0.000)
(0.000)
(0.000)
Prob [ v
2
(1)
(2)
(3)
(4)
DV = FinclFraud Panel B: The common component of the four measures of social capital has a stronger effect on financial fraud Electoral participation
-2.050*** (0.004)
Census response rate
-1.448 (0.151)
# of NGOs
-0.007 (0.136)
# of associations
-0.424 (0.127)
Firm-level controls
Yes
Yes
Yes
Yes
County-level controls
Yes
Yes
Yes
Yes
Industry dummies
Yes
Yes
Yes
Yes
Year dummies
Yes
Yes
Yes
Yes
Observations
85,743
85,743
85,743
85,743
(Pseudo) R2
0.0750
0.0741
0.742
0.741
Prob [ v2
(0.000)
(0.000)
(0.000)
(0.000)
(1)
(2)
(3)
(4)
(5)
year B 2002
year [ 2002
DV = FinclFraud
Panel C: State-fixed effects SocialCapital
-0.175
-0.173*
-0.177*
-0.093
-0.623**
(0.209)
(0.079)
(0.069)
(0.567)
(0.028)
State-fixed effect
Yes
Yes
Yes
Yes
Yes
Firm-level controls
Yes
Yes
No
Yes
Yes
123
A. Jha
(1)
(2)
(3)
(4)
(5)
year B 2002
year [ 2002
DV = FinclFraud
County-level controls
Yes
No
No
Yes
Yes
Industry dummies
Yes
Yes
Yes
Yes
Yes
Year dummies
Yes
Yes
Yes
Yes
Yes
Observations
84,896
84,896
84,896
51,982
25,996
(Pseudo) R2 Prob [ v2
0.0826 (0.000)
0.0812 (0.000)
0.0766 (0.000)
0.0841 (0.000)
0.0979 (0.000)
The DV refers to the dependent variables. The county-level controls are the same as in Table 1. The p values are in the brackets. They are based on the robust standard errors clustered at the county level. The ***, **, and * represent significance at the 1, 5, and 10% levels, respectively. The variable descriptions are in ‘‘Appendix 1’’ section. All continuous variables are winsorized at the 1st and the 99th percentile
References Ai, C., & Norton, E. C. (2003). Interaction terms in logit and probit models. Economics Letters, 80(1), 123–129. Akerlof, G. A. (2007). The missing motivation in macroeconomics. American Economic Review, 97(1), 5–36. Albrecht, C., Holland, D., Malaguen˜o, R., Dolan, S., & Tzafrir, S. (2015). The role of power in financial statement fraud schemes. Journal of Business Ethics, 131(4), 803–813. Alesina, A., & La Ferrara, E. (2002). Who trusts others? Journal of Public Economics, 85(2), 207–234. Allison, P. D. (1999). Comparing logit and probit coefficients across groups. Sociological Methods & Research, 28(2), 186–208. Ang, J. S., Cole, R. A., & Lin, J. W. (2000). Agency costs and ownership structure. The Journal of Finance, 55(1), 81–106. Anheier, H. K., & Gerhards, J. (1995). Forms of capital and socialstructure in cultural fields—Examining Bourdieu social topography. American Journal of Sociology, 100(4), 859–903. Aristotle (2004). The Nicomachean ethics (Further rev. ed., Penguin classics). London, Eng.; New York, NY: Penguin Books. Arrow, K. (1979). Business codes and economic efficiency (Ethical theory and business). Prentice Hall, NJ: Prentice Hall. Beatty, A., Liao, S., & Weber, J. (2010). Financial reporting quality, private information, monitoring, and the lease-versus-buy decision. Accounting Review, 85(4), 1215–1238. doi:10.2308/accr. 2010.85.4.1215. Bell, T., Szykowny, S., & Willingham, J. J. (1991). Assessing the likelihood of fraudulent financial reporting: A cascaded logit approach. Unpublished Manuscript. Berggren, N., & Bjørnskov, C. (2011). Is the importance of religion in daily life related to social trust? Cross-country and cross-state comparisons. Journal of Economic Behavior & Organization, 80(3), 459–480. Bertrand, M., Duflo, E., & Mullainathan, S. (2004). How much should we trust differences-in-differences estimates? Quarterly Journal of Economics, 119(1), 249–275. Bharath, S. T., Sunder, J., & Sunder, S. V. (2008). Accounting quality and debt contracting. Accounting Review, 83(1), 1–28. Bhojraj, S., Hribar, P., Picconi, M., & Mcinnis, J. (2009). Making sense of cents: An examination of firms that marginally miss or beat analyst forecasts. Journal of Finance, 64(5), 2361–2388. Biddle, G. C., Hilary, G., & Verdi, R. S. (2009). How does financial reporting quality relate to investment efficiency? Journal of
123
Accounting and Economics, 48(2–3), 112–131. doi:10.1016/j. jacceco.2009.09.001. Bloomfield, R. J. (2002). The ‘‘incomplete revelation hypothesis’’ and financial reporting. Accounting Horizons, 16(3), 233–243. Bonaparte, Y., & Kumar, A. (2012). Political activism, information costs, and stock market participation. Journal of Financial Economics. Brewer, M. B. (1999). The psychology of prejudice: Ingroup love or outgroup hate? Journal of Social Issues, 55(3), 429–444. Buonanno, P., Montolio, D., & Vanin, P. (2009). Does social capital reduce crime? Journal of Law and Economics, 52(1), 145–170. Burt, R. S. (1999). Entrepreneurs, distrust, and third parties (Shared cognition in organizations). Hillsdale, NJ: Lawrence Erlbaum. Burt, R. S. (2000). The network structure of social capital. Research in Organizational Behavior, 22, 345–423. Callen, J. L., & Fang, X. (2012). Religion and stock price crash risk. Journal of Financial and Quantitative Analysis (JFQA), Forthcoming. Callen, J. L., Khan, M., & Lu, H. (2012). Accounting quality, stock price delay, and future stock returns*. Contemporary Accounting Research. Carpenter, T. D., & Reimers, J. L. (2005). Unethical and fraudulent financial reporting: Applying the theory of planned behavior. Journal of Business Ethics, 60(2), 115–129. Chakrabarty, S. (2015). The influence of unrelated and related diversification on fraudulent reporting. Journal of Business Ethics, 131(4), 815–832. Chaves, M. (2011). American religion: Contemporary trends. Princeton, NJ: Princeton University Press. Chenhall, R. H., Hall, M., & Smith, D. (2010). Social capital and management control systems: A study of a non-government organization. Accounting, Organizations and Society, 35(8), 737–756. doi:10.1016/j.aos.2010.09.006. Chetty, R., Hendren, N., Kline, P., & Saez, E. (2014). Where is the land of opportunity? The geography of intergenerational mobility in the United States. Forthcoming. Chhaochharia, V., Kumar, A., & Niessen-Ruenzi, A. (2012). Local investors and corporate governance. Journal of Accounting and Economics. Cohen, D. A., Dey, A., & Lys, T. Z. (2008). Real and accrual-based earnings management in the pre- and post-Sarbanes–Oxley periods. Accounting Review, 83(3), 757–787. Cohen, D. A., & Zarowin, P. (2010). Accrual-based and real earnings management activities around seasoned equity offerings. Journal
Financial Reports and Social Capital of Accounting and Economics, 50(1), 2–19. doi:10.1016/j. jacceco.2010.01.002. Coleman, J. S. (1988). Social capital in the creation of human capital. American Journal of Sociology, 94, S95–S120. Coleman, J. S. (1990). Foundations of social theory. Cambridge, MA: Belknap Press of Harvard University Press. Dechow, P. M., & Dichev, I. D. (2002). The quality of accruals and earnings: The role of accrual estimation errors. The Accounting Review, 77(s-1), 35–59. Deller, S. C., & Deller, M. A. (2010). Rural crime and social capital. Growth and Change, 41(2), 221–275. deTocqueville, A. (1835). Democracy in America. London: Saunders and Otley. Dyreng, S. D., Mayew, W. J., & Williams, C. D. (2012). Religious social norms and corporate financial reporting. Journal of Business Finance & Accounting, 39(7–8), 845–875. Fang, V. W., Tian, X., & Tice, S. (2014). Does stock liquidity enhance or impede firm innovation? The Journal of Finance, 69(5), 2085–2125. Files, R., Martin, G., & Rasmussen, S. (2012). The monetary benefit of cooperation in regulatory enforcement actions for financial misrepresentation. Available at SSRN. Francis, J., LaFond, R., Olsson, P., & Schipper, K. (2005). The market pricing of accruals quality. Journal of Accounting and Economics, 39(2), 295–327. Francis, J., Nanda, D., & Olsson, P. (2008). Voluntary disclosure, earnings quality, and cost of capital. Journal of Accounting Research, 46(1), 53–99. Fukuyama, F. (1995). Trust: The social virtues and the creation of prosperity. New York: Free Press. Fukuyama, F. (1997). Social capital and the modern capitalist economy: Creating a high trust workplace. Stern Business Magazine, 4(1), 1–16. Giannetti, M., & Wang, T. Y. (2014). Corporate scandals and household stock market participation. Graham, J. R., Harvey, C. R., & Rajgopal, S. (2005). The economic implications of corporate financial reporting. Journal of Accounting and Economics, 40(1–3), 3–73. doi:10.1016/j. jacceco.2005.01.002. Grullon, G., Kanatas, G., & Weston, J. (2010). Religion and corporate (mis)behavior. http://ssrn.com/abstract=1472118. Guiso, L., Sapienza, P., & Zingales, L. (2004a). Does local financial development matter? Quarterly Journal of Economics, 119(3), 929–969. Guiso, L., Sapienza, P., & Zingales, L. (2004b). The role of social capital in financial development. American Economic Review, 94(3), 526–556. Guiso, L., Sapienza, P., & Zingales, L. (2008a). Alfred Marshall lecture—Social capital as good culture. Journal of the European Economic Association, 6(2–3), 295–320. Guiso, L., Sapienza, P., & Zingales, L. (2008b). Trusting the stock market. Journal of Finance, 63(6), 2557–2600. doi:10.1111/j. 1540-6261.2008.01408.x. Gunny, K. A. (2010). The relation between earnings management using real activities manipulation and future performance: Evidence from meeting earnings benchmarks. Contemporary Accounting Research, 27(3), 855–888. doi:10.1111/j.1911-3846.2010.01029.x. Hardy, M. A., & Bryman, A. (2004). Handbook of data analysis. Beverley Hills, CA: Sage. Hasan, I., Hoi, C.-K. S., Wu, Q., & Zhang, H. (2015). Social capital and debt contracting: Evidence from bank loans and public bonds. Journal of Financial and Quantitative Analysis (JFQA), Forthcoming. Hass, L. H., Tarsalewska, M., & Zhan, F. (2016). Equity incentives and corporate fraud in China. Journal of Business Ethics, 138(4), 723–742. doi:10.1007/s10551-015-2774-2.
Healy, P. M., & Wahlen, J. M. (1999). A review of the earnings management literature and its implications for standard setting. Accounting Horizons, 13, 365–383. Hemingway, C. A., & Maclagan, P. W. (2004). Managers’ personal values as drivers of corporate social responsibility. Journal of Business Ethics, 50(1), 33–44. Hilary, G., & Huang, S. (2015). Trust and contracting. Available at SSRN 2604974. Hilary, G., & Hui, K. W. (2009). Does religion matter in corporate decision making in America? Journal of Financial Economics, 93(3), 455–473. doi:10.1016/j.jfineco.2008.10.001. Hribar, P., & Collins, D. W. (2002). Errors in estimating accruals: Implications for empirical research. Journal of Accounting Research, 40(1), 105–134. Jha, A., & Chen, Y. (2015). Audit fees and social capital. Accounting Review, 90(2), 611–639. doi:10.2308/accr-50878. Jha, A., & Cox, J. (2015). Corporate social responsibility and social capital. Journal of Banking & Finance. Jiang, J., Petroni, K. R., & Wang, I. Y. (2010). CFOs and CEOs: Who have the most influence on earnings management? Journal of Financial Economics, 96(3), 513–526. doi:10.1016/j.jfineco. 2010.02.007. Jones, J. J. (1991). Earnings management during import relief investigations. Journal of Accounting Research, 29(2), 193–228. Karlan, D. S. (2005). Using experimental economics to measure social capital and predict financial decisions. American Economic Review, 95(5), 1688–1699. doi:10.1257/0002828057750 14407. Karpoff, J., Koester, A., Lee, D. S., & Martin, G. (2012). A critical analysis of databases used in financial misconduct research. Available at SSRN 2112569. Karpoff, J. M., Lee, D. S., & Martin, G. S. (2008a). The cost to firms of cooking the books. Journal of Financial and quantitative Analysis, 43(3), 581–611. Karpoff, J. M., Scott Lee, D., & Martin, G. S. (2008b). The consequences to managers for financial misrepresentation. Journal of Financial Economics, 88(2), 193–215. Kedia, S., & Rajgopal, S. (2011). Do the SEC’s enforcement preferences affect corporate misconduct? Journal of Accounting and Economics, 51(3), 259–278. doi:10.1016/j.jacceco.2011.01.004. Khanna, V., Kim, E., & Lu, Y. (2015). CEO connectedness and corporate fraud. The Journal of Finance, 70(3), 1203–1252. Kim, Y., Park, M. S., & Wier, B. (2012). Is earnings quality associated with corporate social responsibility? Accounting Review, 87(3), 761–796. doi:10.2308/Accr-10209. Kothari, S. P., Leone, A. J., & Wasley, C. E. (2005). Performance matched discretionary accrual measures. Journal of Accounting and Economics, 39(1), 163–197. doi:10.1016/j.jacceco.2004.11.002. La Porta, R., Lopez-De-Silanes, F., Shleifer, A., & Vishny, R. W. (1997). Trust in large organizations. American Economic Review, 87(2), 333–338. Laursen, K., Masciarelli, F., & Prencipe, A. (2012). Regions matter: How localized social capital affects innovation and external knowledge acquisition. Organization Science, 23(1), 177–193. doi:10.1287/orsc.1110.0650. Lawrence, A. (2011). Individual investors and financial disclosure. Toronto: University of Toronto. Lee, Y. J. (2012). The effect of quarterly report readability on information efficiency of stock prices*. Contemporary Accounting Research. Lehavy, R., Li, F., & Merkley, K. (2011). The effect of annual report readability on analyst following and the properties of their earnings forecasts. The Accounting Review, 86(3), 1087–1115. Li, F. (2008). Annual report readability, current earnings, and earnings persistence. Journal of Accounting and Economics, 45(2), 221–247.
123
A. Jha Li, X., & Wang, X. (2015). Social capital and stock price crash risk: Evidence from China. Paper presented at the Financial Management Association Annual Meetings. Loughran, T., & McDonald, B. (2014). Measuring readability in financial disclosures. The Journal of Finance. McCleary, R. M., & Barro, R. J. (2006). Religion and economy. The Journal of Economic Perspectives, 20(2), 49–72. doi:10.1257/ jep.20.2.49. McGuire, S. T., Omer, T. C., & Sharp, N. Y. (2012). The impact of religion on financial reporting irregularities. The Accounting Review, 87(2), 645–673. Payne, G. T., Moore, C. B., Griffis, S. E., & Autry, C. W. (2011). Multilevel challenges and opportunities in social capital research. Journal of Management, 37(2), 491–520. doi:10. 1177/0149206310372413. Petrovits, C. M. (2006). Corporate-sponsored foundations and earnings management. Journal of Accounting and Economics, 41(3), 335–362. Portes, A. (1998). Social capital: Its origins and applications in modern sociology. Annual Review of Sociology, 24, 1–24. Prior, D., Surroca, J., & Tribo´, J. A. (2008). Are socially responsible managers really ethical? Exploring the relationship between earnings management and corporate social responsibility. Corporate governance: An international review, 16(3), 160–177. Putnam, R. D. (2000). Bowling alone: The collapse and revival of American community. New York: Simon & Schuster. Putnam, R. D. (2001). Social capital: Measurement and consequences. Isuma: Canadian Journal of Policy Research, 2, 41–51. Putnam, R. D. (2007). E pluribus unum: Diversity and community in the twenty-first century the 2006 Johan Skytte Prize Lecture. Scandinavian Political Studies, 30(2), 137–174. doi:10.1111/j. 1467-9477.2007.00176.x. Raval, V. (2016). A disposition-based fraud model: Theoretical integration and research agenda. Journal of Business Ethics. doi:10.1007/s10551-016-3199-2.
123
Rosenbaum, P. R., & Rubin, D. B. (1983). The central role of the propensity score in observational studies for causal effects. Biometrika, 70(1), 41–55. Roychowdhury, S. (2006). Earnings management through real activities manipulation. Journal of Accounting and Economics, 42(3), 335–370. doi:10.1016/j.jacceco.2006.01.002. Rupasingha, A., Goetz, S. J., & Freshwater, D. (2008). US Countylevel social capital data, 1990–2005. The Northeast Regional Center for Rural Development, Penn State University. http:// nercrd.psu.edu/Social_Capital/index.html. Sen, A. (1987). On ethics and economics (The Royer lectures). Oxford: B. Blackwell. Smidt, C. (1999). Religion and civic engagement: A comparative analysis. The Annals of the American Academy of Political and Social Science, 565(1), 176–192. Smidt, C. E. (2003). Religion as social capital: Producing the common good. Waco, Texas: Baylor University Press. Spagnolo, G. (1999). Social relations and cooperation in organizations. Journal of Economic Behavior & Organization, 38(1), 1–25. Welch, M. R., Sikkink, D., & Loveland, M. T. (2007). The radius of trust: Religion, social embeddedness and trust in strangers. Social Forces, 86(1), 23–46. Williams, R. (2009). Using heterogeneous choice models to compare logit and probit coefficients across groups. Sociological Methods & Research, 37(4), 531–559. Woolcock, M. (2001). The place of social capital in understanding social and economic outcomes. Canadian Journal of Policy Research, 2, 11–17. Wu, W. (2008). Dimensions of social capital and firm competitiveness improvement: The mediating role of information sharing. Journal of Management Studies, 45(1), 122–146. Yu, F. (2008). Analyst coverage and earnings management. Journal of Financial Economics, 88(2), 245–271. doi:10.1016/j.jfineco. 2007.05.008.