WPS7254 Policy Research Working Paper 7254 The Dark Side of Disclosure Evidence of Government Expropriation from Worldwide Firms Tingting Liu Barkat Ullah Zuobao Wei Lixin Colin Xu Development Research Group Finance and Private Sector Development Team May 2015 Policy Research Working Paper 7254 Abstract This paper studies the effects of voluntary accounting infor- largely be explained by the fact that most of the countries mation disclosure through auditing on firm access to finance, in the sample are developing countries where institutions exposure to corruption, and sales growth. Relying on a are weak. The beneficial effect of disclosure increases as a data set of more than 70,000 firms in 121 countries, the country’s property rights protection improves. The qualita- analysis finds that disclosure can be a double-edged sword. tive results are robust to considerations of the endogeneity On the one hand, audited firms exhibit a slightly lower of auditing and to alternative measures of corruption and level of financial constraints than unaudited firms. On the financial constraints. The findings reveal the dark side other hand, audited firms face a significantly higher level of voluntary information disclosure: exposing firms to of corruption obstacles. The net effects of voluntary infor- government expropriation where institutions are weak. mation disclosure on firm growth are negative, which can This paper is a product of the Finance and Private Sector Development Team, Development Research Group. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at lxu1@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team The Dark Side of Disclosure: Evidence of Government Expropriation from Worldwide Firms Tingting Liu University of Georgia, Athens, GA Email: ttliu@uga.edu; Tel: (760) 308-8139 Barkat Ullah Rhode Island College, Providence, RI Email: sbullah@ric.edu; Tel: (401) 456-9528 Zuobao Wei University of Texas at El Paso, El Paso, TX Email: zwei@utep.edu; Tel: (915) 747-5381 Lixin Colin Xu * The World Bank, Washington, D.C. Email: lxu1@worldbank.org; Tel: (202) 473-4664 April 21, 2015 JEL Classification: M42, G32, D73. Key Words: Audit; information disclosure; information asymmetry; financial constraints; Corruption; institutions. * Corresponding author. We like to thank Vivek Singh (discussant) and other participants at the FMA 2013 conference for their helpful comments. We are also grateful for the FMA 2013 track chairs for providing invaluable comments in selecting our paper as a semifinalist for the Best Paper Award in corporate finance. We also thank Erik Devos, Juliet D’Souza, Sadok El Ghoul, Omrane Guehami, Bihong (Jenny) Huang, Bill Megginson, Xiaolin (Audrey) Qian, and Oscar Varela for their helpful suggestions. I. Introduction Information asymmetry has important implications in corporate decisions and as such, is among the most researched topics in the corporate finance literature (Meyer and Majluf, 1984; Stein, 2003). One particular effect of information asymmetry between the firm and its lenders is higher cost of external financing, 1 which limits a firm’s investment opportunity set and retards firm growth. Firm management, therefore, is motivated to pursue disclosure policies that reduce the information asymmetry. An oft-used strategy is for firms to have their financial statements audited by external auditors and then disclose them to the relevant outside stakeholders.2 Many scholars have examined voluntary disclosure practices and their impact on corporate policies. A key finding is that higher level of disclosure is related to lower cost of external financing (Leuz and Verrecchia, 2000; Bushman and Smith, 2003; Hughes, Liu, and Liu, 2007; Lambert, Leuz and Verrecchia, 2007) and is therefore beneficial to firm growth (Verrecchia, 2001; Bushman and Smith, 2003; Khurana, Pereira, and Martin, 2006; Leuz and Wysocki, 2008). However, information disclosure is not without cost (Verrecchia, 1983; Elliot and Jacobson, 1994; Yosha, 1995; Leuz and Wysocki, 2008). In addition to the direct costs (such as audit and clerical fees), disclosure can incur indirect or strategic costs. For example, an extended disclosure may reveal a firm’s detailed financial and operational information to interested parties, i.e. competitors, customers, suppliers, or bureaucrats. These parties may in turn use the disclosed information to compete against the firm or to extract rents, causing the firm’s competitive advantage to erode (Admati and Pfleiderer, 2000). In light of these considerations, some predict that firms pursue an optimal level of disclosure after 1 Meyer and Majluf, 1984; Demirguc-Kunt and Maksimovic, 1998; Verrecchia, 2001; Bushman and Smith, 2001; Khurana, Pereira and Martin, 2006. 2 Chow, Kramer, and Wallace (1988) hypothesize that firms have three incentives to have their financial statement audited by external auditors. The first is the information-signaling incentive (Dye, 1993): through auditing, a firm can transmit private information to external stakeholders concerning its future prospects. The second is the insurance incentive (Kellogg, 1984): through audit opinions, a firm provides a means for investors to recover investment losses. The third is the agency-cost-reduction incentive: auditing can ensure the accuracy of the reported financial statements and thereby reduce agency costs (DeFond, 1992; Beatty, 1989). The overall empirical evidence tends to support the existence of these incentives. 2 weighing the associated costs and benefits (Lang and Lundholm, 1993; Bamber and Cheon, 1998; Admati and Pfleiderer, 2000). While there have been extensive empirical studies about the benefits of disclosure such as the effect of disclosure on the cost of external financing, there have been much fewer empirical studies on the costs associated with disclosure (Leuz and Wysocki, 2008). In addition, most existing studies are based on publicly traded firms in the U.S. where governance, markets, and institutions are more developed and efficient than those in developing countries. Thus, empirical evidence that directly examines both the benefits and costs of disclosure is scarce, while large-scale cross-country empirical evidence is almost non-existent. In this study, we employ a large cross-country firm level data set, which allows us to directly examine the benefits and costs, especially indirect political costs, of disclosure in a context where variations in institutions are substantial. We analyze the World Bank Enterprise Survey (WBES) data from 2006 to 2014 for over 70,000 firms in 121 developing countries. We provide evidence that disclosure can be a double-edged sword: while audited firms exhibit a lower level of financial obstacles than unaudited firms, they also face a higher level of corruption obstacles. Furthermore, firms strategically disclose for higher net benefits: they are more likely to disclose if they have more growth opportunities measured by capital expenditure; they are less likely to disclose if they face a high level of competition. Perhaps surprisingly, auditing has an overall negative impact on firm growth. However, the negative effect can be largely explained by the fact that most of the countries in our sample are under-developed countries where institutions are weak. Thus disclosure may be detrimental to firm development if they operate in an unsound business environment with poor governance. This conjecture is supported by our further empirical analyses: the beneficial effect of disclosure indeed increases as a country’s level of property rights protection improves. We conduct extensive specification checks. In our empirical analysis, we recognize the potential endogenous nature of auditing and rely on two methods to address the non-random nature of auditing: 3 the matching method (through either direct or propensity score matching) and the two-stage least square (2SLS) method. Our results continue to hold after we deal with potential endogeneity of information disclosure. Our results also remain robust with various measurement issues. To alleviate the concern that the measures of financial obstacles are subjective responses to survey questions, we use two objective measures as our alternative proxies of financial constraints (i.e., the use of an overdraft facility and access to a line of credit) and obtain robust results. We also employ two detailed proxies of government expropriation—the obstacles in obtaining licenses and permits and the obstacles in paying taxes—and again obtain consistent results. We contribute to two strands of literature. First, we contribute to the literature on voluntary information disclosure. Our study provides novel and comprehensive cross-country empirical evidence on the benefits and costs of disclosure. We capture the benefit of disclosure by a firm’s access to finance, and the cost of disclosure by corruption obstacles faced by the firm; we further capture the overall effect of disclosure by firm growth (Khurana et al., 2006). We demonstrate in a novel way that disclosure has important costs in allowing exposure to government expropriation, and that the costs and benefits of disclosure depend on a country’s institutions.3 The finding of the dark side of exposing firms to government expropriation is novel, and not explored in the literature. 4 Moreover, the dependence of the information-disclosure effect on institutional background nicely illustrates the complementarity of information disclosure with the underlying institutional background, which is vastly under-explored in the literature but emphasized by Leuz and Wysocki (2008) in their survey. 3 The quality of governance and institutions has been shown to significantly impact on firm-level corporate policies and outcomes (La Porta, Lopez-de-Silanes, Shleifer and Vishny (LLSV), 2000; Lemmon and Lins, 2003; Doidge, Karolyi and Stulz, 2007; Barth, Lin, Lin and Song, 2009; Harrison, Lin and Xu, 2014, among others). However, the literature has not examined how it affects the costs and benefits of disclosure. 4 For instance, in a comprehensive survey, Leuz and Wysocki (2008) have not mentioned any paper that explicitly addresses the costs of auditing in exposing the firm to government expropriation. 4 Second, we add new evidence to the corruption literature. 5 We show that accounting information disclosure can be detrimental to firm development if firms operate in a corrupt business environment with low institutional quality. Such disclosure allows corrupt bureaucrats to gain access to firm-level information and use it for endogenous harassment (Fisman and Svensson, 2007). Corruption can thus indirectly discourage the adoption of good policies (such as voluntary information disclosure) that would be efficient in economies with better institutions. The rest of this paper is organized as follows. Section II describes data, variable definitions, and summary statistics. Section III examines the effect of disclosure on firm financial obstacles. Section IV examines its effect on firm corruption obstacles. Section V investigates the effect of voluntary disclosure on firm growth. Section VI analyzes how the costs and benefits of auditing hinge on property rights protection. Section VII concludes. II. Data, Variables and Summary Statistics A. Data We construct firm-level variables from World Bank Enterprise Survey (WBES) collected for 71,677 firms in 121 countries from 2006 to 2014.6 Of the 121 countries, most were surveyed once; a few, in two to three waves. See Table A1 in the Appendix for the total number of firm observations, and the number and percentage of audited and unaudited firms by country. WBES relies on standardized survey instruments in collecting firm-level data. The survey respondents are mainly business owners and/or firm top managers. The surveys focus on assessing the critical obstacles in the business environment that hinder firm growth, including access to finance and obstacles related to corruption, political instability, infrastructure, crime, competition, the labor 5 For papers and surveys on the literature of corruption, see Bardhan (1997), Cai, Fang and Xu (2011), Clarke and Xu (2004), Li, Xu and Zou (2000), Mauro (1995), Shleifer and Vishny (1993), and Svensson (2003, 2005), 6 For a literature survey of firm-level studies using the WBES data, see Xu (2011). 5 market and the legal system. The survey also contains information on firm ownership, sales, employees, top manager experience, whether a firm is an exporter, and firm age. We start with all observations in WBES, and we proceed to delete firms that do not answer the audit question: “In the last fiscal year, did this establishment have its annual financial statement checked and certified by an external auditor?” We also delete (4,330) publicly listed firms because they are required to have their annual financial statements audited. Thus, auditing is not a choice that can be made by these firms. Our final sample consists of 71,677 firms in 121 countries from 2006 to 2014.7 Among these, 47% of firms had their annual financial statements checked and certified by external auditors. A common criticism of using survey data to conduct research related to firm performance and growth is that survey data are self-reported and therefore the findings may suffer from self-reporting bias. However, accounting data are more likely to be biased than survey data as the incentives to distort data are likely to be higher in financial statements because many firm-level decisions, such as tax, financing, and managerial compensations, are in part based on variables from financial statements (Beck Demirgüç‐Kunt and Maksimovic, 2005). Furthermore, the self-reporting nature of the WBES data is not likely to be a significant source of bias. The survey aims to evaluate the business environment instead of firm performance and growth. Even though some firm performance and growth related questions were asked, the survey was specifically designed to ask those questions at the end of the interview. This reduces the respondents’ need to justify their own performance when answering the earlier questions about the business environment. Throughout our analysis, we control for country level governance and macroeconomic variables. Country level financial development data are from International Financial Statistics (IFS). Country-level governance indices (Control of Corruption and Government Effectiveness) are from the 7 The dataset includes firms from 15 different industries according to the international industry classification codes (ISIC). 6 Worldwide Governance Indicators (WGI) database by the World Bank. Country level macroeconomic variables, i.e. GDP, GDP per capita, GDP growth rate and inflation, come from the World Development Indicators (WDI) database by the World Bank. Our measure of property rights comes from the Heritage Foundation. B. Variables and Summary Statistics We first describe our dependent variables.8 The survey contains various indicators of obstacles to firm growth. We focus on two key obstacles to firm growth: financial obstacles (FinancialObstacle) and corruption obstacles (CorruptionObstacle). For additional analyses, we also examine two other firm-level obstacles, i.e. business licensing and permits-related obstacles and tax collection-related obstacles. All these firm-level obstacles take on a value based on the self-reported answers to the following WBES question: “How problematic is _________ for the current operations of a business?” The blank space represents one of the aforementioned growth obstacles. The obstacles are on the scale from 0 – 4: no obstacle (0), minor obstacle (1), moderate obstacle (2), major obstacle (3), and very severe obstacle (4). For robustness checks, we also create dummy variables for these obstacles, FinancialObstacleDummy, CorruptionObstacleDummy, LicenseDummy, and TaxRateDummy, which take a value of 1 if the respective obstacle scores equal 2 (moderate), 3 (major), or 4 (very severe), and 0 otherwise. We employ two additional, objective measures to proxy a firm’s financial constraints: overdraft (Overdraft) and line of credit (CreditLine). Both are indicator variables. Overdraft equals 1 if a firm has an overdraft facility, and 0 otherwise. CreditLine equals 1 if a firm has a line of credit in a financial institution, and 0 otherwise. Firms with an overdraft facility or line of credit are found to be less financially constrained (Lins, Servaes and Tufano, 2010). As reported in Panel B of Table 1, univariate tests show that audited firms are more likely to have an overdraft facility and line of credit, compared to unaudited firms. 8 The details of the variables, their sources and the survey data items are provided in Table A2 in the Appendix. 7 Another key dependent variable is firm sales growth. It is computed as follows: SalesGrowth = (Sales t – Sales t-2 ) / (Sales t + Sales t-2 ) (1) Here t is year.9 This growth measure is bound between -1 and 1, and thus reduces the influence of Here t is year. The growth rate thus constructed is bound by -1 and 1, which reduces the influence of outliers. To further reduce such influence, the growth rates are also winsorized at the top and bottom one percent. Now we describe the explanatory variables. Our key variable, Audit, is an indicator variable that is 1 if a firm’s annual financial statements were checked and certified by an external auditor, and 0 otherwise. As shown in Table 1, about 47% of the surveyed firms choose to have their financial statements audited. Another variable is firm size. Large firms are likely less constrained by various firm-level obstacles than smaller firms, and small firms benefit more than large firms when growth obstacles are reduced (Schiffer and Weder, 2001; Beck, Demirgüç-Kunt and Maksimovic 2005; Cull and Xu, 2005; Knack and Xu, 2015). In our regressions, therefore, we control for firm size (FirmSize), measured as the natural logarithm of firm sales in constant US dollars. As shown in Panel A of Table 1, the mean and median sales in our sample are US$18,653,000 and US$409,000, respectively, which indicates the existence of extreme outliers in terms of firm size. We thus take the logarithm to alleviate the concern of extreme outliers. An additional control variable is firm age, which is important because younger firms tend to grow faster than older firms (Dunne, Roberts and Samuelson, 1988), and are less likely to be harassed to pay bribes due to better relationships with bureaucrats and bank officers (Fisman and Svensson, 2007). We take the natural logarithm of firm age to reduce the influence of outliers. In our sample, 9 WBES does not contain information for (t-1) for sales. 8 the average firm has been in business for about 18 years and the oldest firm is 340 years old (Table 1 Panel A).10 We also control for work experience of top managers. Ceteris paribus, more experience working in the same sector entails better understanding of the business and the business environment. For instance, prior industry experience of the manager is positively related to firm performance measured by survival rate, profitability and sales growth (Bosma, van Praag, Thurik and de Wit, 2004).11 Here managerial experience (Experience) is measured as the number of years that the top manager has been working in the same sector. In our sample, the average Experience for the top manager is 17 years (Table 1 Panel A). We further control for two ownership variables. State ownership is believed and often shown to be associated with worse firm performance (Megginson and Netter, 2001; Beck, Demirguc-Kunt and Maksimovic, 2005; Harrison, Lin and Xu, 2004). We measure government ownership in our analysis with a dummy variable (Government) that equals 1 if the firm has government ownership stakes, 0 otherwise. In our sample, only 1% of all firms have government ownership stakes. In contrast, foreign ownership tends to be positively related to firm performance (Estrin, Hanousek, Kocenda and Svejnar, 2009; Harrison, Lin and Xu, 2014), partly because firms with foreign ownership have better access to markets and technical expertise than pure domestic firms (Fisman and Svensson, 2007). We therefore control for foreign ownership in our regressions with a dummy variable Foreign that equals 1 if any foreign companies or individuals have an ownership stake in the firm, and 0 otherwise. About 11% of all firms in our sample have foreign ownership. Importantly, we also control for a key determinant of firm performance: competition. Competition increases the risk of forced exits from the market, thus spurring efforts. Though competition, according to Schumpeter, may also reduce the rent from innovation and thus reduces 10 For curiosity, we look it up and the oldest firm in our sample is a food-manufacturing firm in Jamaica. 11 However, Robb and Watson (2012) find mixed effects of owner experience on firm performance. 9 the incentives for competition—there is evidence that the relationship between competition and innovation is non-linear (Aghion, Bloom, Blundell, Griffith and Howitt, 2005), and most economists remain positive about the role of competition for improving firm performance based on empirical evidence (Li, 1997; Xu, 2000; Li and Xu, 2004). There is also evidence that firms facing more fierce competition tend to experience greater financing obstacles and corruption (Beck, Demirgüç-Kunt, Maksimovic, 2005). We thus include a dummy variable (Compete) that equals 1 if the firm answered ‘Yes’ to the question: “Does this establishment compete against unregistered or informal firms?” 12 About 55 percent of our sample firms believe that they compete against informal firms (see Panel A of Table 1). We now turn to measures of country-level financial and institutional development, which play important roles in how firm-level financial information is disseminated and used by various economic agents. As a control for country-level financial/credit market development, we use the ratio of domestic banking credit to the private sector over GDP (“Priv”). There are very wide variations in this variable in our sample countries, ranging from a low of 0.9 to a high of 121.5, with a mean of 33.9 (see Panel A of Table 1). To control for country level corruption, we use corruption control (CorruptionControl) from the World Bank’s World Governance Indicators (WGI) database. This measure ranges from -2.5 (weak control) to +2.5 (strong control). In our robustness checks, we use alternative measures of corruption by controlling for a country’s government effectiveness (GovernmentEffectiveness), which measures the quality of a country’s public services, civil services, and policy formulation and implementation. GovernmentEffectiveness also ranges from -2.5 (weak) to +2.5 (strong). Both CorruptionControl and GovernmentEffectiveness vary widely across countries in our sample (see Panel A of Table 1). 12Arguably this is not the best measure of competition since it does not capture competition from the formal sector. However, informal sector accounts for a large share of the economy in developing countries (La Porta and Shleifer, 2014); moreover, the WBES does not contain better measures. 10 We shall examine how the effect of disclosure depends on the quality of a country’s institutional environment. To this end, we rely on an index from the Heritage Foundation: property rights (PropertyRights). There is a wide variation in PropertyRights across our sample countries (see Panel A, Table 1). As the basic macroeconomic environment also influences firm performance, we control for key macro indicators including GDP, GDP growth rate, GDP per capita, and inflation (as in Beck Demirgüç-Kunt and Maksimovic, 2005; Knack and Xu, 2015). We in addition control for country, industry, and year fixed effects in all of our multivariate regressions. The country dummy holds constant all country-specific factors such as geography, culture, and the basic legal system. The inclusion of industry and year dummies further holds constant all industry-specific heterogeneity and worldwide common shocks. We thus have pushed quite far in reducing the extent of omitted variable bias. C. Univariate Tests We first present univariate test results for our key variables between audited and unaudited firms. These two types of firms exhibit a similar level of corruption obstacles and similar sales growth rates. Relative to unaudited firms, audited ones face a significantly lower level of financial obstacles; they are larger, have been in business longer, and their top managers are more experienced; they are more likely to have foreign or government ownership stakes, to be exporters, and are less likely to have to compete with competitors in the informal sector. Thus, audited and unaudited firms differ greatly in basic characteristics, and it is important to control for such key characteristics. D. Determinants of Audit 11 To further shed light on how audited and unaudited firms differ in key characteristics, we now investigate how the decision of having financial statements audited is related to firm characteristics. Our regression model is specified as follows: Auditi,j = α + β1 FirmSizei,j + β2 FirmAgei,j + β3 Experiencei,j + β4 Governmenti,j + β5 Foreigni,j + β6 Exporteri,j + β7 Competei,j + β8 CapitalExpenditurei,j + θ' Macro Controlsi + εi,j (2) The dependent variable is Audit, a dummy variable that equals 1 if the firm’s annual financial statement is audited by external auditors. Subscript i and j represent firm and country, respectively. The independent variables include various firm characteristics and macro Controls (i.e., GDP, GDP per capita, GDP growth, and inflation). We also include country, industry, and year fixed effects in all specifications. Since we have a binary dependent variable with a large number of dummy variables and adding a large number of fixed effects to a traditional probit model would induce incidental parameters bias (Lancaster 2000), we follow Angrist (2001) and use a linear probability model (LPM) to estimate marginal effects instead of relying on the Probit model--linear models are not typically subject to the incidental parameters bias (Lancaster 2000). Since LPM tends to introduce heteroskedastic residuals, we use heteroskedasticity-consistent robust standard errors. Table 2 presents the OLS and the probit results. Overall, the qualitative results are similar under alternative model specifications. According to the results, auditing is more likely for firms that are larger, older, with government or foreign ownership, and for firms that export. Consistent with our expectation that disclosure is more likely where the demand for capital is higher, the coefficient of CapitalExpenditure is positive.13 The negative coefficient on Compete implies that firms are less likely 13 However, this result is subject to reverse causality, that is, it is also possible that auditing leads to more capital expenditure. 12 to disclose information if they operate in a competitive environment, implying that, when making audit decisions, firms are mindful of the strategic costs of disclosure (Leuz and Wysocki, 2008). III. Audit and Financial Obstacles We now examine how auditing affects FinancialObstacle to shed light on its impact on access to finance. We employ multivariate regressions controling for relevant firm characteristics, the industry effect, the year effect, the country effect, and (time-varying) country credit market development and economic development. Our regression model is specified as follows: FinancialObstaclei,j = α + β1 Auditi,i + β2 FirmSizei,j + β3 FirmAgei,j + β4 Experiencei,j + β5 Governmenti,j + β6 Foreigni,j here> + β7 Exporteri,j + β8 Competei,j + β9 Priv,j + θ' Macro Controlsi + εi,j (3) The subscripts i and j represent firm and country, respectively. The dependent variable, FinancialObstacle, is the observed firm-level financial obstacles and is polychotomous with a natural order, i.e., from 0 (no obstacle) to 4 (severe obstacle). We mainly rely on OLS to estimate equation (3). We also report the ordered probit results and show them to be qualitatively similar. In robustness checks, we also create a dummy variable, FinancialObstacleDummy, that equals 1 if the level of financial obstacle is 2 (moderate) or above, and 0 otherwise. In all these models, we expect β1, the coefficient of Audit, to be significant and negative, as our hypothesis is that audited firms face lower financial obstacles than unaudited firms. A. OLS and Ordered Probit Regressions Columns (1) to (4) of Table 3 report the results. Models (1), (2) and (3) report the OLS results and Model (4) reports the Ordered Probit results. As shown in Table 3, the coefficients of Audit, β1, are negative and significant across all four models, consistent with our hypothesis that disclosure reduces financial obstacles. Our finding remains robust after controlling for country credit market 13 development, macroeconomic factors, industry effects, year effects and country effects. Based on column (2), firms changing from no auditing to auditing are associated with a drop in FinancialObstacle by 0.059, or 4% of its standard deviation (SD); the magnitude of the effect of auditing in terms of SDs of the dependent variable is similar when we use FinancialObstacleDummy. The coefficients of other explanatory variables also make sense. Financing obstacles are lower for larger and older firms, for firms whose top manager have more work experience and for firms with foreign ownership. The coefficients of Compete are positive and highly significant across all models, indicating that financing obstacles are higher when firms have to compete with informal firms. One explanation is that a significant number of firms in our sample are small firms (with a median sales of less than half a million US dollars) and they have to compete for the limited funds with other unregistered small firms, especially through the informal financing channel (Beck, Demirguc-Kunt and Maksimovic, 2008). So far we have assumed that a firm’s audit decision is exogenous to financial obstacles. However, the audit decision could be endogenous due to reverse causality (i.e., financing need determines voluntary information disclosure) and/or omitted variables (i.e., some omitted variables that are related to auditing also determine financing obstacles). Since the decision is voluntary (as opposed to regulatory mandate), a firm takes into account many factors when deciding whether to have its financial statements audited. For example, some financially-constrained firms may opt not to apply for external financing (hence no need to have their books audited) because they do not believe their applications would be approved due to bad credit ratings or other characteristics. On the other hand, it could be that some firms apply for credit (hence have their financial statements audited) while they are still in relatively good financial health. Thus, we have to deal with potential endogeneity associated with self-selection involved in the auditing decision. We rely on two methods: two-stage least squares regressions (2SLS) and the matching method. 14 B. 2SLS Regressions A valid instrument for Audit must meet two criteria: a strong correlation with Audit and being orthogonal to the error term. Our instrumental variable is logarithm of IndustryActivity -- the ratio of the number of audited firms in a given industry in a given year to the total number of firms in that industry. Industry level instrumental variable, commonly used in the literature (Xu, 2012), can act as a reasonable instrument because some industries have strong need for auditing, and IndustryActivity captures such industry characteristics. Indeed, we expect a firm’s audit decision to be positively related to industry-level audit activity: firms are more willing to disclose information if more firms in that industry disclose such information, for instance, due to lower strategic costs of disclosing. In the meantime, industry-year level auditing intensity is not likely to directly affect individual firm obstacles once we control for industry fixed effects and year fixed effects in the regressions. The 2SLS regression results are presented in Models (5) (i.e., the first stage) and (6) (i.e., the second stage) of Table 3. Independent variables of both regressions include the same set of firm characteristics and macro controls. Our diagnostic tests suggest that our instrumental variable is strong. As pointed out by Bound, Jaeger, and Baker (1993; 1995), the “cure can be worse than the disease” when the excluded instruments are weakly correlated with the endogenous variable, in which case the IV estimates will be inconsistent and biased in the same direction as OLS. As we can see, the F statistic exceeds the conventional rule of thumb by a large margin.14 As shown in Model (6), once being instrumented, the 2SLS estimate of the effect of auditing on FinanciaObstacleDummy becomes statistically insignificant, though the sign remains negative and the magnitude becomes even more pronounced. From the change in magnitude, it seems that Audit is positively correlated with the error term, implying that firms facing more financing obstacle tend to resort to voluntary information disclosure to a greater extent. In light of the lack of statistical 14 Since we allow for heteroskedasticity for our error term, the relevant test statistic is the Kleibergen-Paap rk F statistics. 15 significance, the 2SLS results thus cast some doubt as to whether voluntary information disclosure reduces financing obstacles. C. Matching As an alternative to 2SLS estimation, whose validity hinges on the validity of the IV, we also use two matching methods to shed light on the effect of Audit on financing obstacles: propensity score matching and direct matching. The advantage of these methods is that it is transparent, easier to understand, and intuitive; its disadvantages is that it does not handle the issue of selection on unobservables. To ensure that our key results do not rest on fragile identifying assumptions, we thus also provide the matching results; we shall draw our main conclusions based on the findings from both 2SLS and the matching results.15 To conduct propensity score matching analysis, we first use the full regression specification in Table 3 to estimate a propensity score, which is the probability that a given firm would have its financial statements audited. We then match the audited firm with an unaudited firm using the nearest neighbor method. We match with replacement (i.e., allow a control firm to be used multiple times as potential match for various treated firms) in the region of common support, which ensures that the matches do not fall outside of the range of propensity values given by the treated group. The results of propensity score matching are reported in Table 4 Panel A. The second row, labeled ‘Matched’, compares the treated firms to their counterparts based on the nearest matched non- audited firms. The results show that audited firms experience a slightly lower level of financial obstacles (i.e., the average treatment effect on the treated, or ATT) and it is only marginally significant when measured by FinancialObstacleDummy. 15 In other words, we shall assign the weight of zero for the findings associated with the OLS results since the underlying selectivity of auditing is quite clear. 16 Although propensity score matching better controls for covariates that affect the probability of treatment, it does not require an explicit match on some important independent variables that affect both the likelihood of having financial statements audited and firm obstacles. For example, we have shown that firm size is the most important determinant of auditing decision, and that size is also significantly related to firm level obstacles.16 Similarly, ‘Government’ status and ‘Foreign’ status also affect firm obstacles, growth, and the auditing decision. We thus also employ an alternative matching method: to explicitly match the treated group (audited firms) with the control group (unaudited firms) on some important aspects that affect both the decision of auditing and firm obstacles. We thus match the audited firms (the treated group) with unaudited firms (the control group) explicitly based on the following criteria. First, we require that they come from the same country, the same industry, and the same year to control for unobserved country, industry, and year factors. Second, they must have the same ‘Government’ status. Third, they have the same ‘Foreign’ status. Fourth, the size of the control firm is the closest to that of the treated firm, and it has to be within 15 percent of the size of the treated firm.17 As shown in Panel B of Table 4, the direct matching results show that audited firms experience slightly lower incidence of financing obstacles being moderate or more severe. The results are similar to those obtained through propensity score matching and are marginally significant (with a p-value of 8%). D. Alternative Measures of Financial Obstacles 16 Beck Demirgüç‐Kunt and Maksimovic, 2005 also show that size is significantly related to firm obstacles. 17 A more precise match on firm size would make the treated group and the control group more comparable. However, a stricter size matching criterion also reduces the number of observations for our matching sample. For instance, if we restrict the size range of the control firm to be 85% to 115% of the treated firm, we have 11,380 pairs of firms. In untabulated results, we also require size range to be 90% to 110% of the treated firm, our sample decreases to only 9,566 pairs of firms. Our results hold when using the alternative size range. 17 The measures of financial obstacle in Table 3 are subjective responses to survey questions. What if the subjective measures do not accurately reflect reality? To guard against this possibility, we employ two alternative, objective measures of financial constraints, i.e. overdraft facilities (Overdraft) and lines of credit (CreditLine). These dummy variables equal to 1 if a firm, respectively, has an overdraft facility or a line of credit with a financial institution, and 0 otherwise. Firms with lines of credit or an overdraft facility are found to be significantly less financially constrained (Sufi, 2009; Lins, et al., 2010). We also construct a third measure, FinancialIndex, defined as the sum of Overdraft and CreditLine. We replace the dependent variable (FinancialObstacle) in equation (3) with Overdraft, CreditLine and FinancialIndex, and estimate with the OLS model with heteroskedasticity-robust standard errors. We further address the potential endogeneity of audit using 2SLS regression analysis; the IV is the same as before. Table 5 reports the results. Since the OLS results may simply reflect selectivity, we focus on the 2SLS results. Auditing has positive effects on all three alternative, objective measures of financial constraints. The effect on Overdraft and FinancialIndex is positive and significant. Combining the 2SLS/matching results on FinancialObstacle, Financial Obstacle Dummy and the results on the three objective measures of financial access, auditing seems to have marginally beneficial effect on financial access. However, the magnitude does not look large, and the effect is not robustly statistically significant. Our results thus render a weak support to the contention that voluntary information disclosure tends to improve financial access. We also note that firms facing higher financing obstacles tend to opt for voluntary information disclosure. IV. Audit and Corruption Obstacles A. OLS and Ordered Probit Regressions 18 We have found marginal benefits of auditing to firms in reducing financial constraints. Now we investigate the costs associated with auditing in terms of exposure to government expropriation. The specification is as follows: CorruptionObstaclei,j = α + β1 Auditi,j + β2 FirmSizei,j + β3 FirmAgei,j + β4 Experiencei,j + β5Governmenti,j + β6 Foreigni,j + β7 Exporteri,j + β8 Competei,j + β9 CorruptionControlj + θ' Macro Controlsj + εij (4) We expect β1, the coefficient of Audit, to be positive and significant, as we hypothesize that disclosure increases the potential of government expropriation. The first four columns of Table 6 present the regression results. Models (1), (2) and (3) report the OLS results and Model (4) reports the Ordered Probit results. Consistent with our hypothesis, the coefficients of Audit, β1, are positive and significant at the conventional level. Our finding is robust after controlling for relevant firm characteristics, country corruption control, macroeconomic factors, industry effects, year effects, and country effects. Based on the estimate in Model (2), firms changing from no auditing to auditing are associated with an increase in CorruptionObstacle by 0.041, or 2.5% of one SD for the outcome variable. How other variables affect corruption obstacles is also interesting. Perceived corruption obstacle is higher for larger firms, perhaps because of economy of scale in collecting bribes— bureaucrats save time costs and reduce the risk of being exposed when extracting the same amount of bribes from one large firm than from several small firms. Perceived corruption obstacle is lower for firms owned by the government, perhaps because state-owned enterprises tend to have stronger political support and are thus less vulnerable. Perceived corruption obstacle is higher for exporters, which is not surprising since in developing countries they have to go through a lot of bureaucratic red tape to obtain export licenses, which increases their vulnerability to government expropriation (Shleifer and Vishny, 1993). Particularly interestingly, perceived corruption obstacle is higher for firms competing with informal businesses. This is not surprising since informal firms are largely immune 19 from government expropriation; as a result, with more informal firms, bureaucrats focus more on expropriating formal ones. B. 2SLS Regressions and Matching Reverse causality between Audit and the corruption obstacles is less of a concern than in the case of financing obstacles—due to the likely negative correlation of Audit and the prevailing level of corruption obstacles (i.e., firms would be less likely to voluntarily disclose information where corruption obstacles are more severe), the OLS estimate of the effect of Audit would be under- estimated, and a positive OLS estimate of the auditing effect would thus imply an even more pronounced positive effect of Audit. Therefore, the potential reverse causality should strengthen our empirical findings. Nevertheless, the issue of omitted variables remains. We thus employ the 2SLS regressions and the matching method to address the potential endogeneity of Audit. We present the results below. As shown in Model (6) of Table 6, after dealing with endogeneity, the coefficient of Audit remains positive and significant (at the 5% level) and much larger than in the OLS results. Based on this estimate, firms changing from no auditing to auditing are associated with an increase in CorruptionObstacle by 0.25, or about 15% of one SD for the outcome variable. The matching results in Table 7 are largely consistent with the OLS regression results for corruption obstacles. In fact, the effect of auditing on corruption obstacles becomes larger under both matching methods. Propensity score matching results show that the coefficient of Audit becomes 0.074, almost double the original coefficient size of 0.041 in Model (2) Table 6. Similarly, direct matching results show an increase in the magnitude of the effect of audit on firm corruption obstacles. C. Alternative Measures of Corruption Obstacles 20 We have used CorruptionObstacle as a summary measure of perception of corruption. Our confidence on the results would be strengthened if this measure is supported by more detailed measures related to corruption or government expropriation. Indeed, if auditing leads to more exposure to government expropriation, we are likely to see more specific harassments in various areas, such as applying for licenses and permits and dealing with tax collectors. The harassment in these areas, fortunately, is covered in WBES. We thus construct two alternative firm-level corruption obstacle indicators: the firm perceives moderate-or-above obstacles in the areas of licenses and permits (LicenseDummy) and tax collections (TaxRateDummy). We also construct the third measure, CorruptionIndex, equal to the sum of LicenseDummy and TaxRateDummy. The OLS results in Table 8 show that Audit is significantly related to higher levels of license- related obstacles and to CorruptionIndex. The 2SLS results show that after dealing with endogeneity, Audit significantly increases the level of license and tax-related obstacles. Audit is also significantly positively related to CorruptionIndex. These results lend further support for our hypothesis that information disclosure leads to higher level of corruption obstacles. V. Audit and Firm Growth A. OLS Regressions We have shown that financial disclosure by firms through auditing entails both benefits (i.e., better access to finance) and costs (i.e., exposure to “grabbing hands”). What is the net effect of auditing? We shed light on this question by examining the effect of auditing on firm development, measured by sales growth, which can be interpreted as (roughly) the net effect of auditing. To this end, we estimate the following equation: SalesGrowthi,j = α + β1 Auditi,j + β2 FirmSizei,j + β3 FirmAgei,j + β4 Experiencei,j + β5 Governmenti,j + β6 Foreigni,j + β7 Exporter i,j + β8 Competei,j + β9 CorruptionControli,j+ β10 Privi,j + θ' Macro Controlsj + εi,j (5) 21 Subscript i and j represent firm and country, respectively. The dependent variable is sales growth (SalesGrowth), as defined in equation (1). We report the results in Table 9, with the first three columns being the OLS results, and the next two being the 2SLS results. As before, we control for industry effect, year effect and country effect in all the models. Column (1) only controls for firm size and firm age, while Column (2) adds as additional controls top manager experience, government ownership, foreign ownership, exporter status, competition, and macroeconomic variables. Column (3) further adds CorruptionControl and financial market development (Priv). The OLS results in Table 9 show that the coefficients for Audit are negative and significant (at the 1% level). This indicates that disclosure has an overall detrimental effect to firm growth. The significant detrimental effect of information disclosure on sales growth may be attributable to the fact that most of the countries surveyed in the WBES are developing countries where institutions and governance are weak and corruption is rampant. In corrupt business environments, the cost of financial disclosure may outweigh its benefit. We examine this point further in section VI. We also observe that growth is faster for larger and younger firms. Growth is slower for exporters and firms with government or foreign ownership. Not related to top manager work experience, growth is faster for firms operating in competitive environments. Finally, growth is also positively related to financial depth. B. Endogeneity Checks There are two potential sources of endogeneity. First, reverse causality between Audit and firm growth is a real possibility. In general, high-growth firms need more external financing and therefore, are likely to apply for more external credits (hence decide to have their financial statements audited). Since our OLS regression results in Table 9 has an overall negative effect on firm growth, the aforementioned reverse causality between firm growth and auditing thus actually strengthen our 22 results: the unbiased estimate of the effect of Audit should remain negative and even more pronounced. Second, there may still be issues related to omitted variables. As before, we address the endogeneity of Audit using 2SLS regressions and the direct matching method. The 2SLS results in Model (5) of Table 9 show that the coefficient of Audit remains negative and significant. Moreover, the negative effect is much more pronounced than before, indicating a positive correlation between auditing and omitted factors affecting firm growth. Based on the 2SLS results, increasing Audit by one SD (0.50) is associated with a drop in sales growth by 15 percentage points, or half a SD of the outcome variable, which constitutes a large effect. The matching results in Table 10 show that auditing has a significant and negative effect on firm sales growth under both matching methods, consistent with the OLS and 2SLS results shown in Table 9. VI. The Role of Institutional Development So far we have assumed homogeneous effect of Audit on financing and corruption obstacles and on firm growth. But the effect of disclosure likely hinges on the underlying institutions. In fact, a large volume of literature has documented a direct positive relation between a country’s institutional development and economic growth at the country level as well as at the firm level (Knack and Keefer, 1995; La Porta et al. 2000; Beck, Levin and Loayza 2000; Acemoglu, Johnson and Robinson, 2001; Acemoglu and Johnson, 2005; Dyck and Zingales 2004; Beck et al. 2005, Doidge, Karolyi and Stulz 2007, Harrison, Lin and Xu, 2014, among others). To measure institutional quality, we employ the Heritage Foundation’s index of property rights protection (PropertyRights), which is an estimate of the legally protected freedom to accumulate private property and wealth by citizens of the country. Higher values for PropertyRights indicate better property rights protection and less government expropriation. As shown in Panel A of Table 1, PropertyRights varies widely in our sample countries, ranging from 0 to 90. We standardize PropertyRights for an easier interpretation of the coefficients. 23 We now examine how the effects of information disclosure on firms’ financial obstacles, corruption obstacles, and growth depend on the level of institutional development of the country in which the firm resides. We thus add to the equation PropertyRights and its interactions with Audit, and re-estimate the equation with both OLS and 2SLS regressions. The results on all three outcome variables are reported in Table 11. Since the 2SLS is more defendable in terms of filtering out selectivity, we focus on the 2SLS results. First, the negative effect of Audit on the incidence of moderate-and-more-severe financing obstacle is doubled when Property Rights increases by one SD. Thus, better property rights are associated with stronger benefits of information disclosure on access to finance. Second, the positive effect of Audit on the incidence of moderate-and-more-severe corruption obstacles is slightly smaller, though this effect is not statistically significant. Third, the negative effect of Audit on sales growth is significantly smaller. A one-SD improvement in the country’s property rights protection leads to 12.3 percentage points increase in the effect of auditing on sales growth rate; or, it reduces the negative effect of auditing on sales growth by 43 percent. The effects of property rights protection on access to finance and on sales growth are thus substantial. VII. Conclusions Using the WBES data from 2006 to 2014 for more than 70,000 firms in 121 countries, we study the effect of financial information disclosure through auditing on firm level financial and corruption obstacles and on firm growth. We find that disclosure is a double-edge sword. On the one hand, disclosure leads to a somewhat lower level of financial obstacles, though the results are not robust. On the other hand, disclosure also leads to a significantly higher level of corruption obstacles faced by the firm. Our explanation is that our sample countries surveyed are largely developing countries with underdeveloped markets and institutions, and that, once firm information is disclosed, the threat of government expropriation is widespread. Information disclosure thus allows rent-seeking 24 bureaucrats to gain access to the disclosed information and use it to extract bribes (Fisman and Svensson, 2007). In terms of the aggregate effect of auditing, we find that auditing has a negative, significant, and pronounced effect on sales growth. We further document that a country’s property rights protection plays an important role in the effect of disclosure on firm obstacles. As a country’s property rights protection improves, disclosure leads to bigger reduction in financial obstacles and less reduction in sales growth. Our paper offers a vivid illustration that an important hindrance to institutional development— here in the form of adopting information disclosure—is government expropriation. With more information about firms available, government expropriations (i.e., corruption obstacle) become more severe, especially in countries with poor property rights protection. The results are thus supportive of Acemoglu and Johnson (2005) on the overwhelming importance of constraining government expropriation in facilitating economic development.18 18 Knack and Xu (2015) offer further evidence on the overwhelming importance of property rights institutions (i.e., containing government expropriation) relative to contracting institutions (i.e., facilitating exchange between private citizens) in facilitating external finance. 25 References Acemouglu, D., and S. Johnson. 2005. "Unbundling Institutions." Journal of Political Economy 113(5), 949–95. Acemoglu, D., Johnson, S., Robinson, J., 2001. The colonial origins of comparative development: an empirical investigation. American Economic Review 91(5), 1369-1401. Admati, A., and Pfleiderer, P. 2000. Forcing firms to talk: Financial disclosure regulation and externalities. Review of Financial Studies 15, 479-519. Aghion, P., N. Bloom, R. Blundell, R Griffith, P. Howitt. 2005. “Competition and Innovation: An Inverted-U Relationship.” Quarterly Journal of Economics 120, 701-728. Angrist, J. D. (2001). Estimation of limited dependent variable models with dummy endogenous regressors. Journal of Business & Economic Statistics, 19(1). Bamber, L., and Cheon, Y. 1998. Discretionary management earnings forecast disclosure: Antecedents and outcomes associated with forecast specialty choices. Journal of Accounting Research 36, 167-190. Bardhan, Pranab, 1997. “Corruption and Development: A Review of Issues.” Journal of Economic Literature 35:1320–1346. Barth, J., Lin, C., Lin, P., & Song, F. (2009). Corruption in bank lending to firms: Cross-country micro evidence on the beneficial role of competition and information sharing. Journal of Financial Economics 91, 361-386. Beatty, R. 1989. Auditor reputation and the pricing of initial public offerings. The Accounting Review 64 (4): 693–709. Beck, T., Demirguc-Kunt, A., Levine, R., 2006. Bank supervision and corruption in lending. Journal of Monetary Economics 53, 2131–2163. Beck, T., Demirgüç‐Kunt, A., and Maksimovic, V. 2005. Financial and legal constraints to growth: Does firm size matter? The Journal of Finance, 60(1), 137-177. Beck, T., Demirgüç‐Kunt, A., and Maksimovic, V. 2008. Financing patterns around the world: Are small firms different? Journal of Financial Economics 89, 468-487. Beck, T., Levine, R., and Loayza, N. 2000. Finance and sources of growth. Journal of Financial Economics 58, 261-300. Bosma, N., van Praag, M., Thurik, R., and de Wit, G. 2004. The Value of Human and Social Capital Investments for the Business Performance of Startups. Small Business Economics 23(3), 227–36 Bound, J., Jaeger, D. A., & Baker, R. 1993. The cure can be worse than the disease: A cautionary tale regarding instrumental variables. NBER Technical Working Paper No. 137. Bound, J., Jaeger, D. A., & Baker, R. M. 1995. Problems with instrumental variables estimation when the correlation between the instruments and the endogenous explanatory variable is weak. Journal of the American statistical association, 90(430), 443-450. Bushman, R., and A. Smith. 2003. Transparency, Financial Accounting Information, and Corporate Governance. FRBNY Economic Policy Review, 9, 65-87. Bushman, R., and A. Smith. 2001. Financial Accounting Information and Corporate Governance. Journal of Accounting and Economics, 32, 237-333. Cai, H.B., H.M. Fang, L. C. Xu. 2011. “Eat, Drink, Firms, Government: An Investigation of Corruption from Entertainment and Travel Costs of Chinese Firms.” Journal of Law and Economics 54, 55-78. 26 Chow, C. W., Kramer, L., and Wallace, W. A. 1988. The environment of auditing. In Research Opportunities in Auditing: The Second Decade, eds. A. R. Abdel-Khalik and Solomon, 155–83. Sarasota, FL: American Accounting Association. Clarke, G., L.C. Xu. 2004. “Privatization, Competition, and Corruption: How Characteristics of Bribe Takers and Payers Affect Bribe Payments to Utilities,” Journal of Public Economics 88 (9-10), 2067-97. Cull, R., L.C. Xu. 2005. “Institutions, Ownership, and Finance: The Determinants of Investment among Chinese Firms,” Journal of Financial Economics 77, 117-146. DeFond, M. L. 1992. The association between changes in client firm agency costs and auditor switching. Auditing: A Journal of Practice & Theory 11 (1): 16–31. Demirguc-Kunt, A., and Maksimovic, V. 1998. Law, Finance, and Firm Growth. Journal of Finance, 53, 2107-2137. Doidge, C, Karolyi, G., and Stulz, R. 2007. Why do countries matter so much for corporate governance? Journal of Financial Economics 86, 1-39. Dunne, T., M. J. Roberts and L. Samuelson. 1988, ‘Patterns of Firm Entry and Exit in U.S. Manufacturing Industries’, Rand Journal of Economics 19(4), 495–515. Dyck, A., and Zingales, L. 2004. Private benefits of control: An international comparison. Journal of Finance 59(2), 537-600. Dye, R. A. 1993. Auditing standards, legal liability and auditor wealth. Journal of Political Economy 101 (5): 887–914. Elliott, R. K., and Jacobson, P. D. 1994. Costs and Benefits of Business information Disclosure. Accounting Horizons Vol. 8 No. 4, pp. 80-96. Estrin, S., J. Hanousek, E. Kocenda, and J. Svejnar, 2009. “The Effects of Privatization and Ownership in Transition Economies.” Journal of Economic Literature 47(3): 699-728. Fisman, R., and Svensson, J. 2007. Are corruption and taxation really harmful to growth? Journal of Development Economics, 83, 63–75. Harrison, A., J. Y. Lin, L. C. Xu. 2014. “Explaining Africa’s (Dis)advantage”,World Development 63, 59-77. Hughes, J.S., Liu, J., and Liu, J. 2007. Information asymmetry, diversification, and cost of capital. The Accounting Review 82, 705-729. Kellogg, R. L. 1984. Accounting activities, security prices, and class action lawsuits. Journal of Accounting and Economics 6 (3): 185–204. Khurana, K. I., Pereira, R., and Martin, X. 2006. Firm Growth and Disclosure: An Empirical Analysis. The Journal of Financial and Quantitative Analysis, Vol. 41, No. 2, pp.357-380. Knack, S., and Keefer, P. 1995. Institutional and economic performance: Cross-country tests using alternative institutional measures. Economics and Politics 7, 207-227. Knack, Stephen, Lixin Colin Xu. 2015. Unbundling Institutions for External Finance: Worldwide Firm- Level Evidence. Mimeo, World Bank. La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert Vishny. 2000. Investor protection and corporate governance. Journal of Financial Economics 58, 3-27. Lambert, R., C. Leuz, and Verrecchia, R. 2007. Accounting Information, Disclosure, and the Cost of Capital. Journal of Accounting Research 45, 385-420. Lancaster, T. 2000. The incidental parameter problem since 1948. Journal of Econometrics, 95(2), 391-413. 27 Lang, M., and Lundholm, R. 1993. Cross-sectional determinants of analyst ratings of corporate disclosures. Journal of Accounting Research 31, 246-271. Lemmon, M., and Lins, K. 2003. Ownership structure, corporate governance, and firm value: Evidence from the East Asian financial crisis. The Journal of Finance 58, 1445-1468. Leuz, C., and Verrecchia, R. 2000. The economic consequences of increased disclosure. Journal of Accounting Research 38, 91-124. Leuz, C., and P. Wysocki. 2008. “Economic Consequences of Financial Reporting and Disclosure Regulation: A Review and Suggestions for Future Research.” Working paper, University of Chicago. Li, H., L. C. Xu, H-F Zou. 2000. “Corruption, Income Distribution, and Growth,” Economics and Politics, 12(2), 155-82. Li, W. 1997. The Impact of Economic Reforms on the Performance of Chinese State-Owned Enterprises. Journal of Political Economy 105, 1080–106. Li, W, L. C. Xu. 2004. “The Impact of Privatization and Competition in the Telecommunications Sector around the World?” Journal of Law and Economics XLVII (2), 395-430. Lins, K. V., Servaes, H., and Tufano, P. 2010. What drives corporate liquidity? An international survey of cash holdings and lines of credit. Journal of Financial Economics, 98(1), 160-176. Mauro, P. 1995. Corruption and growth. The quarterly journal of economics, 110(3), 681-712. Megginson, W.L., Netter, J.M., 2001. From state to market: a survey of empirical studies on privatization. Journal of Economic Literature 39(2), 321-389. Myers, S., and N. Majluf. 1984. Corporate Financing and Investment Decisions When Firms Have Information that Investors Do Not Have. Journal of Financial Economics, 13, 187-221. Robb, A. M., and Watson, J. 2012. Gender differences in firm performance: Evidence from new ventures in the United States. Journal of Business Venturing, 27(5), 544-558. Shleifer, A., and Vishny, R. W. 1993. Corruption. The Quarterly Journal of Economics, 108(3), 599-617. Schiffer, M., and Weder, B. 2001. Firm size and the business environment: worldwide survey results (Vol. 43). World Bank Publications. Stein, J. 2003. Agency Costs, Information, and Capital Investment. In Handbook of the Economics of Finance, G. Constantinides, M. Harris, and R. Stulz, eds. Boston, MA: Elsevier/North-Holland. Stock, J. H. and Yogo, M. 2005. Testing for Weak Instruments in IV Regression, in Identification and Inference or Econometric Models: A Festschrift in Honor of Thomas Rothenberg. Donald W. K. Andrews and James H. Stock, eds. Cambridge University Press, pp.80-108. Sufi, A. 2009. Bank lines of credit in corporate finance: an empirical analysis. Review of Financial Studies 22, 1057–1088. Svensson, J. 2003. Who must pay bribes and how much? Evidence from a cross section of firms. The Quarterly Journal of Economics, 118(1), 207-230. Svensson, J. 2005. Eight questions about corruption. The Journal of Economic Perspectives, 19(3), 19-42. Verrecchia, R. 1983. Discretionary disclosure. Journal of Accounting and Economics 5, 179-194. Verrecchia, R. 2001. Essays on Disclosure. Journal of Accounting and Economics 32, 97-180. Xu, J. 2012. Profitability and capital structure: Evidence from import penetration. Journal of Financial Economics, 106(2), 427-446. 28 Xu, L. C. 2000. “Control, Incentives, and Competition: The Impact of Reform in Chinese State-Owned Enterprises.” Economics of Transition 8(1), 151–73. Xu, L. C. 2011. "The Effects of Business Environments on Development: A Survey of New Firm-Level Evidence,” World Bank Research Observer 26 (2), 310-340, August. Yosha, O. 1995. Information disclosure costs and the choice of financing source. Journal of Financial Intermediation 4 (1): 3–20. 29 Table 1: Summary Statistics This table reports the summary statistics of our sample. Panel A presents the summary statistics for our full sample and Panel B reports the summary statistics for audited and unaudited firms in the sub-samples. Audit is a dummy variable that equals 1 if the firm’s annual financial statements were audited by an external auditor, and 0 otherwise. FinancialObstacle and CorruptionObstacle are survey responses for firm-level obstacles as specified in the survey questionnaire. They take values between 0 and 4, where 0 indicates no obstacle and 4 indicates a very severe obstacle. FinancialObstacleDummy and CorruptionObstacleDummy are dummy variables that equals 1 if firm-level obstacles equal 2 (moderate), 3 (major), or 4 (very severe), and 0 otherwise. SalesGrowth is the average growth rate between year t and (t-2), with t being the survey year. Sales is the sales at the end of year t. We convert the sales in local currency into U.S. Dollars. FirmAge is the firm’s actual age. Experience is the firm’s top manager’s years of working experience in this sector. Government and Foreign are dummies that equal 1 if the firm has government or foreign ownership, respectively, and 0 otherwise. Exporter is a dummy variable that equals 1 if the firm is an exporter, 0 otherwise. Compete is a dummy variable that equals 1 if the firm answered ‘Yes’ to the question: “Does this establishment compete against unregistered or informal firms?” CapitalExpenditure is a dummy variable that equals 1 if the firm purchased fixed assets, such as machinery, vehicles, equipment, land or buildings at year t, and 0 otherwise. Overdraft is a dummy variable that equals 1 if the firm has an overdraft facility, and 0 otherwise. CreditLine is a dummy variable that equals 1 if the firm has a line of credit or loan from a financial institution, and 0 otherwise. Priv is the ratio of domestic banking credit to the private sector divided by GDP. CorruptionControl and GovernmentEffectiveness are country-level governance estimates by World Bank’s Worldwide Governance Indicators (WGI), and range from -2.5 (weak) to 2.5 (strong governance). GDP is the log of GDP in current millions of U.S. dollars. GDPperCapita is the log of real GDP per capita in U.S. dollars. GDPGrowth is the real GDP growth rate. Inflation is log difference of consumer price indices. The Heritage Foundation freedom index, PropertyRights, is the foundation’s estimates of a country’s property rights protection. The column “T-test” reports the two-tail t-statistics of two-sample t-tests comparing the means of audited and unaudited firms. Detailed variable definitions and sources are given in Table A2 in the Appendix. ***, **, and * indicate statistical significances at the 1%, 5%, and 10% levels, respectively. Panel A: Summary Statistics for Full Sample N Mean Median SD Min Max Firm level variables Audit 71,677 0.47 0 0.50 0 1 FinancialObstacle 68,377 1.61 2 1.36 0 4 FinancialObstacleDummy 68,377 0.51 1 0.50 0 1 CorruptionObstacle 69,446 1.77 2 1.51 0 4 CorruptionObstacleDummy 69,446 0.53 1 0.50 0 1 SalesGrowth (%) 53,504 0.14 0.11 0.30 -0.76 0.98 Sales (US Dollar in thousands) 63,828 18,653 409 119,048 0.024 1,363,314 FirmAge (years) 68,279 17.87 13 15.65 3 340 Experience (years) 70,217 17.24 15 11.11 0 75 Government 70,340 0.01 0 0.10 0 1 Foreign 70,325 0.11 0 0.31 0 1 Compete 55,083 0.55 1 0.50 0 1 Exporter 71,491 0.21 0 0.41 0 1 CapitalExpenditure 71,336 0.50 1 0.50 0 1 Overdraft 67,745 0.42 0 0.49 0 1 CreditLine 69,344 0.39 0 0.49 0 1 30 N Mean Median SD Min Max Country level variables Priv 149 33.90 25.92 26.41 0.92 121.49 CorruptionControl 152 -0.48 -0.61 0.65 -1.53 1.38 GovernmentEffectiveness 152 -0.31 -0.39 0.60 -1.72 1.48 GDP 152 24.23 23.97 1.99 19.14 28.36 GDPperCapita 152 7.30 7.39 1.12 4.49 9.94 GDPGrowth 152 0.05 0.05 0.03 -0.06 0.23 Inflation 152 0.09 0.07 0.17 0.00 1.91 Heritage freedom index: PropertyRights 136 36.95 30 17.55 0 90 Panel B: Summary Statistics for Audited versus Unaudited Firms (1) Audited (2) Unaudited T-test N Mean N Mean Difference (1-2) T-value FinancialObstacle 33,115 1.49 35,262 1.72 -0.23*** -22.31 FinancialObstacleDummy 33,115 0.48 35,262 0.55 -0.07*** -19.35 CorruptionObstacle 32,980 1.77 36,466 1.77 0.00 0.08 CorruptionObstacleDummy 32,980 0.53 36,466 0.53 0.00 -0.40 SalesGrowth (%) 25,885 0.14 27,619 0.14 0.00 -1.01 Sales (US Dollar in thousands) 30,388 20,209 33,440 6,145 14,064*** 28.51 FirmAge (years) 32,555 20.44 35,724 15.53 4.91*** 41.46 Experience (years) 33,067 18.23 37,150 16.35 1.88*** 22.45 Government 33,231 0.02 37,109 0.01 0.01*** 13.63 Foreign 33,220 0.16 37,105 0.06 0.11*** 45.70 Compete 26,452 0.53 28,631 0.56 -0.03*** -6.55 Exporter 33,790 0.29 37,701 0.14 0.16*** 51.80 CapitalExpenditure 33,697 0.57 37,639 0.43 0.14*** 37.45 Overdraft 32,121 0.54 35,624 0.32 0.22*** 60.09 CreditLine 32,749 0.48 36,595 0.31 0.17*** 47.13 31 Table 2: Determinants of Audit This table reports OLS and Probit analyses of the determinants of audit. The dependent variable, Audit, is a dummy variable that equals 1 if the firm’s annual financial statements were audited by an external auditor, and 0 otherwise. Model (1) reports coefficients estimated by OLS/Linear Probability Model (LPM). Models (2) and (3) present coefficients and marginal effects estimated by Probit regressions. The independent variables include various firm characteristics. FirmSize is the natural logarithm of the firm’s sales in U.S. Dollars. FirmAge is the firm’s actual age. Experience is the firm’s top manager’s years of working experience in this sector. Government and Foreign are dummies that equal 1 if the firm has government or foreign ownership, respectively, and 0 otherwise. Exporter is a dummy variable that equals 1 if the firm is an exporter, 0 otherwise. Compete is a dummy variable that equals 1 if the firm answered ‘Yes’ to the question: “Does this establishment compete against unregistered or informal firms?” CapitalExpenditure is a dummy variable that equals 1 if the firm purchased fixed assets, such as machinery, vehicles, equipment, land or buildings at year t, and 0 otherwise. Macro Controls include GDP, GDP per capita, GDP growth, and inflation. Each regression includes a separate (unreported) intercept. Industry, year, and country fixed effects are included in all specifications. Heteroskedasticity-robust standard errors are reported in parentheses. Detailed variable definitions and sources are given in Table A2 in the Appendix. ***, **, and * indicate statistical significances at the 1%, 5%, and 10% levels, respectively. Dep. Variable: Audit OLS/LPM Probit (1) (2) (3) Coefficient Coefficient Marginal Effect FirmSize 0.064*** 0.217*** (0.001) (0.004) 0.046 FirmAge 0.031*** 0.103*** (0.003) (0.011) 0.022 Experience -0.000 -0.001 (0.000) (0.001) -0.000 Government 0.180*** 0.593*** (0.018) (0.068) 0.167 Foreign 0.099*** 0.367*** (0.007) (0.025) 0.091 Exporter 0.061*** 0.191*** (0.006) (0.019) 0.043 Compete -0.016*** -0.052*** (0.004) (0.014) -0.011 CapitalExpenditure 0.046*** 0.150*** (0.004) (0.014) 0.032 Macro Controls Yes Yes Industry Fixed Effects Yes Yes Year Fixed Effects Yes Yes Country Fixed Effects Yes Yes Observations 44,684 44,684 R2/Pseudo R2 0.297 0.250 32 Table 3: Audit and Firm Financial Obstacle This table reports the impact of audit on a firm’s financial obstacles. The dependent variable for Model (1), (2), and (4) is FinancialObstacle, which takes values between 0 and 4, where 0 indicates no financial obstacle and 4 indicates a very severe financial obstacle. The dependent variable for Model (3) and (6) is FinancialObstacleDummy, a dummy variable that equals 1 if firm-level obstacles equal 2 (moderate), 3 (major), or 4 (very severe), and 0 otherwise. Model (1), (2), and (3) report OLS regression results. Model (4) reports Ordered Probit regression results. Model (5) and (6) report 2SLS regression using an instrumental variable to address the endogeneity concern. Model (5) is the first stage regression where the dependent variable is Audit. Model (6) is the second-stage regression of financial obstacle on the fitted value of Audit and the control variables. Audit is a dummy variable that equals 1 if the firm’s annual financial statements were audited by an external auditor, and 0 otherwise. The independent variables also include various firm characteristics. FirmSize is the natural logarithm of the firm’s sales in U.S. Dollars. FirmAge is the firm’s actual age. Experience is the firm’s top manager’s years of working experience in this sector. Government and Foreign are dummies that equal 1 if the firm has government or foreign ownership, respectively, and 0 otherwise. Exporter is a dummy variable that equals 1 if the firm is an exporter, 0 otherwise. Compete is a dummy variable that equals 1 if the firm answered ‘Yes’ to the question: “Does this establishment compete against unregistered or informal firms?” Priv is the ratio of domestic banking credit to the private sector divided by GDP. Our instrumental variable is IndustryActivity, the ratio of the number of audited firms in a given industry in a given year to the total number of firms in that industry. Macro Controls include GDP, GDP per capita, GDP growth, and inflation. Each regression includes a separate (unreported) intercept. Industry, year, and country fixed effects are included in all specifications. Heteroskedasticity-robust standard errors are reported in parentheses. Detailed variable definitions and sources are given in Table A2 in the Appendix. ***, **, and * indicate statistical significances at the 1%, 5%, and 10% levels, respectively. 33 Ordered OLS/LPM Probit 2SLS First Second Stage Stage Financial Financial Financial Obstacle Financial Obstacle Dep. Var. Obstacle Dummy Obstacle Audit Dummy (1) (2) (3) (4) (5) (6) Audit -0.151*** -0.059*** -0.025*** -0.048*** -0.082 (0.011) (0.014) (0.005) (0.012) (0.122) FirmSize -0.049*** -0.014*** -0.042*** 0.065*** -0.010 (0.003) (0.001) (0.003) (0.001) (0.008) FirmAge -0.058*** -0.023*** -0.048*** 0.029*** -0.021*** (0.010) (0.004) (0.009) (0.003) (0.005) Experience -0.001* -0.001** -0.001** -0.000 -0.001** (0.001) (0.000) (0.001) (0.000) (0.000) Government -0.040 -0.007 -0.058 0.172*** 0.003 (0.056) (0.021) (0.052) (0.020) (0.030) Foreign -0.210*** -0.071*** -0.188*** 0.097*** -0.065*** (0.021) (0.008) (0.019) (0.007) (0.014) Exporter 0.040** 0.006 0.039*** 0.065*** 0.010 (0.016) (0.006) (0.014) (0.006) (0.010) Compete 0.242*** 0.074*** 0.215*** -0.012*** 0.074*** (0.013) (0.005) (0.011) (0.004) (0.005) Priv 0.009* 0.001 0.008** 0.005*** 0.001 (0.005) (0.002) (0.004) (0.002) (0.002) Log(IndustryActivity) 0.118*** (0.013) Macro Controls No Yes Yes Yes Yes Yes Industry Fixed Effects Yes Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes Yes Country Fixed Effects Yes Yes Yes Yes Yes Yes Observations 68,346 42,389 42,389 42,389 42,361 42,361 R2/Pseudo R2 0.131 0.170 0.130 0.0588 0.285 0.127 Weak identification test: Kleibergen-Paap rk Wald F statistic 81.53 Stock-Yogo weak ID test critical values: 10% maximal IV size 16.38 34 Table 4: Audit and Firm Financial Obstacle: Matching Analysis This table reports the results of matching analysis to address the endogeneity concern. The outcome variables are FinancialObstacle and FinancialObstacleDummy. Panel A reports the results of propensity score matching analysis and Panel B reports the results of direct matching analysis. The first stage in the propensity score matching computes a propensity score which is the probability that a given sample firm would have its financial statements audited. The second stage matches the firm that have its financial statements audited (the treated group) with a sample firm that did not have its financial statements audited (the control group). This process is followed for every firm with replacement to ensure the closest possible characteristic match. In the results below, the Unmatched sample computes the simple average of financial obstacles for audited firms versus all other firms. The Matched sample compares the treated firms to their counterparts based on the nearest matched non-audited firms who are in the region of common support. The direct matching process matches the audited firms (the treated group) with unaudited firms (the control group) explicitly on the following criteria: (1) the treated firm and matched firm are from the same country, the same industry, and the same year, and (2) the treated firm and control firm must have the same ‘Government’ status, and (3) the treated firm and control firm must have the same ‘Foreign’ status, and (4) the control firm size is the closest compared to the treated firm size and has to be within a range of 0.85*treated size and 1.15*treated size. FinancialObstacle is a survey response for firm-level obstacles as specified in the survey questionnaire. It takes values between 0 and 4, where 0 indicates no obstacle and 4 indicates a very severe obstacle. FinancialObstacleDummy is a dummy variable that equals 1 if firm-level obstacles equal 2 (moderate), 3 (major), or 4 (very severe), and 0 otherwise. Government and Foreign are dummies that equal 1 if the firm has government or foreign ownership, respectively, and 0 otherwise. Size is proxied by Sales, which is the sales revenue in U.S. Dollars. Detailed variable definitions and sources are given in Table A2 in the Appendix. ***, **, and * indicate statistical significances at the 1%, 5%, and 10% levels, respectively. Panel A: Propensity Score Matching Sample Treated Controls Difference S.E. T-Value FinancialObstacle Unmatched 1.480 1.736 -0.255 0.013 -19.82 Matched 1.480 1.520 -0.039 0.029 -1.38 FinancialObstacleDummy Unmatched 0.471 0.555 -0.084 0.005 -17.67 Matched 0.471 0.489 -0.018* 0.011 -1.69 Panel B:Direct Matching N Treated Control Difference T-value P-value FinancialObstacle 11,380 1.569 1.575 -0.003 -0.18 0.86 FinancialObstacleDummy 11,380 0.493 0.505 -0.011* -1.74 0.08 35 Table 5: Audit and Firm Financial Obstacle: Alternative Measures This table reports the impact of audit on a firm’s financial obstacles using alternative measures. The dependent variables are Overdraft, CreditLine, and FinancialIndex. Overdraft is a dummy variable that equals 1 if the firm has an overdraft facility, and 0 otherwise. CreditLine is a dummy variable that equals 1 if the firm has a line of credit or loan from a financial institution, and 0 otherwise. FinancialIndex is the sum of Overdraft and CreditLine. Models (1) - (3) report OLS regression results. The main independent variable is Audit, a dummy variable that equals 1 if the firm’s annual financial statements were audited by an external auditor, and 0 otherwise. The independent variables include various firm characteristics. FirmSize is the natural logarithm of the firm’s sales in U.S. Dollars. FirmAge is the firm’s actual age. Experience is the firm’s top manager’s years of working experience in this sector. Government and Foreign are dummies that equal 1 if the firm has government or foreign ownership, respectively, and 0 otherwise. Exporter is a dummy variable that equals 1 if the firm is an exporter, 0 otherwise. Compete is a dummy variable that equals 1 if the firm answered ‘Yes’ to the question: “Does this establishment compete against unregistered or informal firms?” Priv is the ratio of domestic banking credit to the private sector divided by GDP. Models (4) - (6) report 2SLS regressions using IndustryActivity as an instrumental variable. IndustryActivity is the ratio of the number of audited firms in a given industry in a given year to the total number of firms in that industry. The dependent variable of the first stage regressions is Audit, a dummy variable that equals 1 if the firm’s annual financial statements were audited by an external auditor, and 0 otherwise. The first stage regressions also control for all of the second-stage variables (except for Audit). For brevity, first stage regression results are not reported. Models (4) - (6) are the second-stage regressions of financial obstacles (measured by Overdraft, CreditLine, and FinancialIndex) on the fitted value of Audit and the control variables. Macro Controls include GDP, GDP per capita, GDP growth, and inflation. Each regression includes a separate (unreported) intercept. Industry, year, and country fixed effects are included in all specifications. Heteroskedasticity-robust standard errors are reported in parentheses. Detailed variable definitions and sources are given in Table A2 in the Appendix. ***, **, and * indicate statistical significances at the 1%, 5%, and 10% levels, respectively. OLS 2SLS Dep. Var. Overdraft CreditLine FincialIndex Overdraft CreditLine FincialIndex (1) (2) (3) (4) (5) (6) Audit 0.080*** 0.062*** 0.139*** 0.429*** 0.064 0.477*** (0.005) (0.005) (0.008) (0.116) (0.107) (0.180) FirmSize 0.044*** 0.049*** 0.093*** 0.021*** 0.049*** 0.070*** (0.001) (0.001) (0.002) (0.008) (0.007) (0.012) FirmAge 0.022*** 0.004 0.025*** 0.012*** 0.004 0.016** (0.003) (0.003) (0.005) (0.005) (0.005) (0.007) Experience 0.000 0.000** 0.001 0.000 0.000** 0.001 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Government -0.040** -0.045** -0.080** -0.101*** -0.045* -0.141*** (0.020) (0.020) (0.032) (0.029) (0.027) (0.045) Foreign -0.025*** -0.107*** -0.130*** -0.059*** -0.107*** -0.163*** (0.007) (0.008) (0.012) (0.014) (0.013) (0.021) Exporter 0.048*** 0.062*** 0.111*** 0.026*** 0.062*** 0.089*** (0.006) (0.006) (0.009) (0.009) (0.009) (0.015) Compete 0.012*** 0.032*** 0.041*** 0.017*** 0.032*** 0.046*** (0.004) (0.004) (0.007) (0.005) (0.005) (0.007) Priv -0.008*** -0.002 -0.015*** -0.008*** -0.002 -0.015*** (0.002) (0.002) (0.003) (0.002) (0.002) (0.003) Macro Controls Yes Yes Yes Yes Yes Yes Industry/Yes/Country Effects Yes Yes Yes Yes Yes Yes Observations 43,640 44,324 43,332 43,611 44,295 43,303 R-squared 0.311 0.236 0.352 0.223 0.235 0.322 36 Table 6: Audit and Firm Corruption Obstacle This table reports the impact of audit on a firm’s corruption obstacles. The dependent variable for Model (1), (2), and (4) is CorruptionObstacle, which takes values between 0 and 4, where 0 indicates no corruption obstacle and 4 indicates a very severe corruption obstacle. The dependent variable for Model (3) and (6) is CorruptionObstacleDummy, a dummy variable that equals 1 if firm-level obstacles equal 2 (moderate), 3 (major), or 4 (very severe), and 0 otherwise. Model (1), (2), and (3) report OLS regression results. Model (4) reports Ordered Probit regression results. Model (5) and (6) report 2SLS regression using an instrumental variable to address the endogeneity concern. Model (5) is the first stage regression where the dependent variable is Audit. Model (6) is the second-stage regression of corruption obstacle on the fitted value of Audit and the control variables. Audit is a dummy variable that equals 1 if the firm’s annual financial statements were audited by an external auditor, and 0 otherwise. The independent variables also include various firm characteristics. FirmSize is the natural logarithm of the firm’s sales in U.S. Dollars. FirmAge is the firm’s actual age. Experience is the firm’s top manager’s years of working experience in this sector. Government and Foreign are dummies that equal 1 if the firm has government or foreign ownership, respectively, and 0 otherwise. Exporter is a dummy variable that equals 1 if the firm is an exporter, 0 otherwise. Compete is a dummy variable that equals 1 if the firm answered ‘Yes’ to the question: “Does this establishment compete against unregistered or informal firms?” CorruptionControl is a country-level governance estimate by World Bank’s Worldwide Governance Indicators (WGI), and ranges from -2.5 (weak) to 2.5 (strong governance). Our instrumental variable is IndustryActivity, the ratio of the number of audited firms in a given industry in a given year to the total number of firms in that industry. Macro Controls include GDP, GDP per capita, GDP growth, and inflation. Each regression includes a separate (unreported) intercept. Industry, year, and country fixed effects are included in all specifications. Heteroskedasticity-robust standard errors are reported in parentheses. Detailed variable definitions and sources are given in Table A2 in the Appendix. ***, **, and * indicate statistical significances at the 1%, 5%, and 10% levels, respectively. 37 Ordered OLS/LPM Probit 2SLS First Second Stage Stage Corruption Corruption Corruption Obstacle Corruption Obstacle Dep. Var. Obstacle Dummy Obstacle Audit Dummy (1) (2) (3) (4) (5) (6) Audit 0.064*** 0.041*** 0.012** 0.036*** 0.251** (0.011) (0.015) (0.005) (0.013) (0.116) FirmSize 0.010*** 0.004*** 0.010*** 0.066*** -0.012 (0.004) (0.001) (0.003) (0.001) (0.008) FirmAge 0.001 -0.001 0.002 0.029*** -0.008 (0.010) (0.004) (0.009) (0.003) (0.005) Experience 0.001 0.000 0.001 -0.000 0.000 (0.001) (0.000) (0.001) (0.000) (0.000) Government -0.128** -0.050** -0.118** 0.185*** -0.095*** (0.059) (0.020) (0.058) (0.019) (0.030) Foreign -0.020 -0.012* -0.008 0.098*** -0.036*** (0.022) (0.007) (0.018) (0.007) (0.014) Exporter 0.053*** 0.014** 0.050*** 0.064*** -0.001 (0.017) (0.006) (0.014) (0.006) (0.010) Compete 0.312*** 0.090*** 0.270*** -0.013*** 0.094*** (0.013) (0.005) (0.011) (0.004) (0.005) CorruptionControl -0.449** -0.187*** -0.357** 0.229*** -0.243*** (0.196) (0.064) (0.155) (0.062) (0.072) Log(IndustryActivity) 0.116*** (0.012) Macro Controls No Yes Yes Yes Yes Yes Industry Fixed Effects Yes Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes Yes Country Fixed Effects Yes Yes Yes Yes Yes Yes Observations 69,420 43,855 43,855 43,855 43,825 43,825 R2/Pseudo R2 0.225 0.253 0.216 0.0949 0.295 0.176 Weak identification test: Kleibergen-Paap rk Wald F statistic 88.65 Stock-Yogo weak ID test critical values: 10% maximal IV size 16.38 38 Table 7: Audit and Firm Corruption Obstacle: Matching Analysis This table reports the results of matching analysis to address the endogeneity concern. The outcome variables are CorruptionObstacle and CorruptionObstacleDummy. Panel A reports the results of propensity score matching analysis and Panel B reports the results of direct matching analysis. The first stage in the propensity score matching computes a propensity score which is the probability that a given sample firm would have its financial statements audited. The second stage matches the firm that have its financial statements audited (the treated group) with a sample firm that did not have its financial statements audited (the control group). This process is followed for every firm with replacement to ensure the closest possible characteristic match. In the results below, the Unmatched sample computes the simple average of corruption obstacles for audited firms versus all other firms. The Matched sample compares the treated firms to their counterparts based on the nearest matched non-audited firms who are in the region of common support. The direct matching process matches the audited firms (the treated group) with unaudited firms (the control group) explicitly on the following criteria: (1) the treated firm and matched firm are from the same country, the same industry, and the same year, and (2) the treated firm and control firm must have the same ‘Government’ status, and (3) the treated firm and control firm must have the same ‘Foreign’ status, and (4) the control firm size is the closest compared to the treated firm size and has to be within a range of 0.85*treated size and 1.15*treated size. CorruptionObstacle is a survey response for firm-level obstacles as specified in the survey questionnaire. It takes values between 0 and 4, where 0 indicates no obstacle and 4 indicates a very severe obstacle. CorruptionObstacleDummy is a dummy variable that equals 1 if firm-level obstacles equal 2 (moderate), 3 (major), or 4 (very severe), and 0 otherwise. Government and Foreign are dummies that equal 1 if the firm has government or foreign ownership, respectively, and 0 otherwise. Size is proxied by Sales, which is the sales revenue in U.S. Dollars. Detailed variable definitions and sources are given in Table A2 in the Appendix. ***, **, and * indicate statistical significances at the 1%, 5%, and 10% levels, respectively. Panel A: Propensity Score Matching Sample Treated Controls Difference S.E. T-Value CorruptionObstacle Unmatched 1.674 1.678 -0.004 0.014 -0.26 Matched 1.674 1.600 0.074** 0.032 2.33 CorruptionObstacleDummy Unmatched 0.500 0.505 -0.004 0.005 -0.86 Matched 0.500 0.480 0.021** 0.011 1.96 Panel B: Direct Matching N Treated Control Difference T-value P-value CorruptionObstacle 11,374 1.747 1.688 0.060*** 3.55 0.00 CorruptionObstacleDummy 11,374 0.519 0.500 0.019*** 3.35 0.00 39 Table 8: Audit and Corruption Obstacle: Alternative Measures This table reports the impact of audit on a firm’s corruption obstacles using alternative measures. The dependent variables are LicenseDummy, TaxRateDummy, and CorruptionIndex. LicenseDummy and TaxRateDummy are dummy variables that equal 1 if firm-level obstacles equal to 2 (moderate), 3 (major), or 4 (very severe), and 0 otherwise. CorruptionIndex is the sum of LicenseDummy and TaxRateDummy. Models (1) - (3) report OLS regression results. The main independent variable is Audit, a dummy variable that equals 1 if the firm’s annual financial statements were audited by an external auditor, and 0 otherwise. The independent variables include various firm characteristics. FirmSize is the natural logarithm of the firm’s sales in U.S. Dollars. FirmAge is the firm’s actual age. Experience is the firm’s top manager’s years of working experience in this sector. Government and Foreign are dummies that equal 1 if the firm has government or foreign ownership, respectively, and 0 otherwise. Exporter is a dummy variable that equals 1 if the firm is an exporter, 0 otherwise. Compete is a dummy variable that equals 1 if the firm answered ‘Yes’ to the question: “Does this establishment compete against unregistered or informal firms?” GovernmentEffectiveness is a country-level governance estimate by World Bank’s Worldwide Governance Indicators (WGI), and ranges from -2.5 (weak) to 2.5 (strong governance). Models (4) - (6) report 2SLS regressions using IndustryActivity as an instrumental variable. IndustryActivity is the ratio of the number of audited firms in a given industry in a given year to the total number of firms in that industry. The dependent variable of the first stage regressions is Audit, a dummy variable that equals 1 if the firm’s annual financial statements were audited by an external auditor, and 0 otherwise. The first stage regressions also control for all of the second-stage variables (except for Audit). For brevity, first stage regression results are not reported. Models (4) - (6) are the second-stage regressions of financial obstacles (measured by LicenseDummy, TaxRateDummy, and CorruptionIndex) on the fitted value of Audit and the control variables. Macro Controls include GDP, GDP per capita, GDP growth, and inflation. Each regression includes a separate (unreported) intercept. Industry, year, and country fixed effects are included in all specifications. Heteroskedasticity-robust standard errors are reported in parentheses. Detailed variable definitions and sources are given in Table A2 in the Appendix. ***, **, and * indicate statistical significances at the 1%, 5%, and 10% levels, respectively. OLS 2SLS License TaxRate Corruption License TaxRate Corruption Dep. Var. Dummy Dummy Index Dummy Dummy Index (1) (2) (3) (4) (5) (6) Audit 0.011** 0.003 0.015* 0.270** 0.367*** 0.642*** (0.005) (0.005) (0.008) (0.114) (0.116) (0.187) FirmSize 0.006*** 0.008*** 0.014*** -0.011 -0.016** -0.028** (0.001) (0.001) (0.002) (0.008) (0.008) (0.013) FirmAge -0.007* -0.000 -0.007 -0.014*** -0.011** -0.025*** (0.004) (0.004) (0.006) (0.005) (0.005) (0.008) Experience -0.001*** 0.001*** -0.000 -0.001*** 0.001*** -0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Government -0.032* -0.104*** -0.130*** -0.080*** -0.172*** -0.249*** (0.018) (0.020) (0.029) (0.030) (0.030) (0.049) Foreign -0.003 -0.032*** -0.032*** -0.028** -0.068*** -0.093*** (0.007) (0.008) (0.012) (0.013) (0.014) (0.022) Exporter 0.014** 0.004 0.018* -0.003 -0.020** -0.024 (0.006) (0.006) (0.009) (0.010) (0.010) (0.016) Compete 0.067*** 0.062*** 0.131*** 0.071*** 0.067*** 0.140*** (0.005) (0.005) (0.007) (0.005) (0.005) (0.008) GovernmentEffectiveness -0.501*** -0.356*** -0.839*** -0.361*** -0.169 -0.503*** (0.104) (0.102) (0.168) (0.118) (0.120) (0.194) Macro Controls Yes Yes Yes Yes Yes Yes Industry/Year/Country Effects Yes Yes Yes Yes Yes Yes Observations 43,837 44,755 43,607 43,808 44,725 43,578 R-squared 0.125 0.180 0.193 0.072 0.083 0.078 40 Table 9: Audit and Firm Growth This table reports the impact of audit on firm growth. The dependent variable is SalesGrowth, the average growth rate between year t and (t-2), with t being the survey year. Model (1), (2), and (3) report OLS regression results. Model (4) and (5) report 2SLS regression using an instrumental variable to address the endogeneity concern. Model (4) is the first stage regression where the dependent variable is Audit. Model (5) is the second-stage regression of firm sales growth on the fitted value of Audit and the control variables. Audit is a dummy variable that equals 1 if the firm’s annual financial statements were audited by an external auditor, and 0 otherwise. The independent variables also include various firm characteristics. FirmSize is the natural logarithm of the firm’s sales in U.S. Dollars. FirmAge is the firm’s actual age. Experience is the firm’s top manager’s years of working experience in this sector. Government and Foreign are dummies that equal 1 if the firm has government or foreign ownership, respectively, and 0 otherwise. Exporter is a dummy variable that equals 1 if the firm is an exporter, 0 otherwise. Compete is a dummy variable that equals 1 if the firm answered ‘Yes’ to the question: “Does this establishment compete against unregistered or informal firms?” CorruptionControl is a country-level governance estimate by World Bank’s Worldwide Governance Indicators (WGI), and ranges from -2.5 (weak) to 2.5 (strong governance). Priv is the ratio of domestic banking credit to the private sector divided by GDP. Our instrumental variable is IndustryActivity, the ratio of the number of audited firms in a given industry in a given year to the total number of firms in that industry. Macro Controls include GDP, GDP per capita, GDP growth, and inflation. Each regression includes a separate (unreported) intercept. Industry, year, and country fixed effects are included in all specifications. Heteroskedasticity-robust standard errors are reported in parentheses. Detailed variable definitions and sources are given in Table A2 in the Appendix. ***, **, and * indicate statistical significances at the 1%, 5%, and 10% levels, respectively. 41 OLS 2SLS First Stage Second Stage Dep. Var. SalesGrowth Audit SalesGrowth (1) (2) (3) (4) (5) Audit -0.028*** -0.025*** -0.024*** -0.309*** (0.003) (0.003) (0.003) (0.078) FirmSize 0.028*** 0.029*** 0.029*** 0.067*** 0.048*** (0.001) (0.001) (0.001) (0.001) (0.005) FirmAge -0.053*** -0.050*** -0.051*** 0.031*** -0.043*** (0.002) (0.003) (0.002) (0.004) (0.003) Experience -0.000 -0.000 -0.000 -0.000 (0.000) (0.000) (0.000) (0.000) Government -0.040*** -0.040*** 0.178*** 0.011 (0.013) (0.013) (0.020) (0.020) Foreign -0.031*** -0.030*** 0.096*** -0.003 (0.005) (0.005) (0.007) (0.009) Exporter -0.019*** -0.018*** 0.066*** 0.001 (0.004) (0.004) (0.006) (0.006) Compete 0.004 0.005* -0.012*** 0.002 (0.003) (0.003) (0.005) (0.003) CorruptionControl 0.188*** 0.175** 0.240*** (0.052) (0.069) (0.051) Priv 0.006*** 0.003** 0.007*** (0.001) (0.002) (0.001) Log(IndustryActivity) 0.121*** (0.013) Macro Controls No Yes Yes Yes Yes Industry Fixed Effects Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes Country Fixed Effects Yes Yes Yes Yes Yes Observations 52,865 39,364 38,988 38,970 38,970 R2/Pseudo R2 0.160 0.179 0.182 0.304 0.024 Weak identification test: Kleibergen-Paap rk Wald F statistic 90.22 Stock-Yogo weak ID test critical values: 10% maximal IV size 16.38 42 Table 10: Audit and Firm Growth: Direct Matching Analysis This table reports the results of matching analysis to address the endogeneity concern. The outcome variable is SalesGrowth. Panel A reports the results of propensity score matching analysis and Panel B reports the results of direct matching analysis. The first stage in the propensity score matching computes a propensity score which is the probability that a given sample firm would have its financial statements audited. The second stage matches the firm that have its financial statements audited (the treated group) with a sample firm that did not have its financial statements audited (the control group). This process is followed for every firm with replacement to ensure the closest possible characteristic match. In the results below, the Unmatched sample computes the simple average of sales growth for audited firms versus all other firms. The Matched sample compares the treated firms to their counterparts based on the nearest matched non-audited firms who are in the region of common support. The direct matching process matches the audited firms (the treated group) with unaudited firms (the control group) explicitly on the following criteria: (1) the treated firm and matched firm are from the same country, the same industry, and the same year, and (2) the treated firm and control firm must have the same ‘Government’ status, and (3) the treated firm and control firm must have the same ‘Foreign’ status, and (4) the control firm size is the closest compared to the treated firm size and has to be within a range of 0.85*treated size and 1.15*treated size. SalesGrowth is the average growth rate between year t and (t-2), with t being the survey year. Government and Foreign are dummies that equal 1 if the firm has government or foreign ownership, respectively, and 0 otherwise. Size is proxied by Sales, which is the sales revenue in U.S. Dollars. Detailed variable definitions and sources are given in Table A2 in the Appendix. ***, **, and * indicate statistical significances at the 1%, 5%, and 10% levels, respectively. Panel A: Propensity Score Matching Sample Treated Controls Difference S.E. T-Value SalesGrowth Unmatched 0.127 0.141 -0.014*** 0.003 -4.56 Matched 0.127 0.161 -0.033*** 0.007 -4.82 Panel B: Direct Matching N Treated Control Difference T-value P-value SalesGrowth 8,689 0.151 0.170 -0.024*** -4.56 0.00 43 Table 11: Audit, Obstacles, Growth, and Institutional Developments This table reports regressions analyzing how audit affects a firm’s financial obstacle, corruption obstacle, and firm growth under different levels of institutional development. The dependent variable for Model (1) and (4) is FinancialObstacleDummy, a dummy variable that equals 1 if firm-level obstacles equal 2 (moderate), 3 (major), or 4 (very severe), and 0 otherwise. The dependent variable for Model (2) and (5) is CorruptionObstacleDummy, a dummy variable that equals 1 if firm-level obstacles equal 2 (moderate), 3 (major), or 4 (very severe), and 0 otherwise. The dependent variable for Model (3) and (6) is SalesGrowth, the average growth rate between year t and (t-2), with t being the survey year. The main independent variable is Audit, a dummy variable that equals 1 if the firm’s annual financial statements were audited by an external auditor, and 0 otherwise. It is interacted with the institutional variable: PropertyRights. PropertyRights is an estimate of the legally protected freedom to accumulate private property and wealth by workers and investors. We standardize PropertyRights for an easier interpretation of the coefficients. Models (1) - (3) report OLS regression results. Firm Level Controls include FirmSize, FirmAge, Experience, Government, Foreign, Exporter, and Compete. FirmSize is the natural logarithm of the firm’s sales in U.S. Dollars. FirmAge is the firm’s actual age. Experience is the firm’s top manager’s years of working experience in this sector. Government and Foreign are dummies that equal 1 if the firm has government or foreign ownership, respectively, and 0 otherwise. Exporter is a dummy variable that equals 1 if the firm is an exporter, 0 otherwise. Compete is a dummy variable that equals 1 if the firm answered ‘Yes’ to the question: “Does this establishment compete against unregistered or informal firms?” Models (4) - (6) report 2SLS regressions. Audit is instrumented by IndustryActivity. Audit interacted with the institutional variable is instrumented by IndustryActivity interacted with the institutional variable. IndustryActivity is the ratio of the number of audited firms in a given industry in a given year to the total number of firms in that industry. For brevity, first stage regression results are not reported. Macro Controls include GDP, GDP per capita, GDP growth, and inflation. Each regression includes a separate (unreported) intercept. Industry, year, and country fixed effects are included in all specifications. Heteroskedasticity-robust standard errors are reported in parentheses. Detailed variable definitions and sources are given in Table A2 in the Appendix. ***, **, and * indicate statistical significances at the 1%, 5%, and 10% levels, respectively. OLS 2SLS Financial Corruption Financial Corruption Sales Sales Obstacle Obstacle Obstacle Obstacle Growth Growth Dep. Var. Dummy Dummy Dummy Dummy (1) (2) (3) (4) (5) (6) Audit -0.035*** 0.007 -0.024*** -0.096 0.252** -0.286*** (0.006) (0.005) (0.004) (0.130) (0.126) (0.083) Audit × PropertyRights 0.013*** -0.005 0.003 -0.082* -0.065 0.123*** (0.005) (0.005) (0.003) (0.047) (0.046) (0.033) PropertyRights -0.055** 0.035* -0.045*** 0.010 0.087** -0.138*** (0.023) (0.020) (0.015) (0.042) (0.039) (0.029) Firm Level Controls Yes Yes Yes Yes Yes Yes Macro Controls Yes Yes Yes Yes Yes Yes Industry Fixed Effects Yes Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes Yes Country Fixed Effects Yes Yes Yes Yes Yes Yes Observations 39,867 40,990 36,777 39,855 40,978 36,766 R-squared 0.132 0.217 0.130 0.121 0.172 -0.044 44 Appendix: Table A1: Audited and Unaudited Firms by Country Country N Audited Unaudited Country N Audited Unaudited Observations Observations N % N % N % N % Afghanistan 790 293 37 497 63 Croatia 545 237 43 308 57 Albania 276 90 33 186 67 Czech Republic 226 119 53 107 47 Angola 718 87 12 631 88 DRC 687 156 23 531 77 Antigua and Barbuda 150 79 53 71 47 Dominica 150 72 48 78 52 Argentina 2,037 1,434 70 603 30 Dominican Republic 353 320 91 33 9 Armenia 342 62 18 280 82 Ecuador 950 501 53 449 47 Azerbaijan 292 112 38 180 62 El Salvador 991 892 90 99 10 Bahamas 143 85 59 58 41 Eritrea 171 142 83 29 17 Bangladesh 2,839 1,171 41 1,668 59 Estonia 210 159 76 51 24 Barbados 148 123 83 25 17 Ethiopia 625 446 71 179 29 Belarus 511 208 41 303 59 Fiji 150 135 90 15 10 Belize 150 103 69 47 31 FYR Macedonia 341 218 64 123 36 Benin 141 81 57 60 43 Gabon 150 60 40 90 60 Bhutan 240 124 52 116 48 Gambia 173 55 32 118 68 Bolivia 916 702 77 214 23 Georgia 650 228 35 422 65 Bosnia and Herzegovina 274 133 49 141 51 Ghana 491 194 40 297 60 Botswana 582 407 70 175 30 Grenada 152 92 61 60 39 Brazil 1,723 368 21 1,355 79 Guatemala 1,001 626 63 375 37 Bulgaria 1,250 505 40 745 60 Guinea 223 16 7 207 93 Burkina Faso 364 180 49 184 51 Guinea Bissau 158 13 8 145 92 Burundi 270 36 13 234 87 Guyana 158 144 91 14 9 Cameroon 352 240 68 112 32 Honduras 707 428 61 279 39 Cape Verde 117 36 31 81 69 Hungary 289 217 75 72 25 Central African Republic 147 79 54 68 46 Indonesia 1,383 195 14 1,188 86 Chad 134 77 57 57 43 Iraq 736 309 42 427 58 Chile 1,928 959 50 969 50 Ivory Coast 499 104 21 395 79 China 2,603 1,838 71 765 29 Jamaica 348 264 76 84 24 Colombia 1,903 1,095 58 808 42 Kazakhstan 492 137 28 355 72 Congo 126 76 60 50 40 Kenya 650 452 70 198 30 Costa Rica 512 308 60 204 40 Kosovo 256 48 19 208 81 45 Country N Audited Unaudited Country N Audited Unaudited Observations Observations N % N % N % N % Kyrgyz Republic 354 107 30 247 70 Samoa 102 73 72 29 28 Lao PDR 625 124 20 501 80 Senegal 506 124 25 382 75 Latvia 267 191 72 76 28 Serbia 297 173 58 124 42 Lesotho 140 94 67 46 33 Sierra Leone 146 40 27 106 73 Liberia 139 30 22 109 78 Slovak Republic 261 141 54 120 46 Lithuania 267 92 34 175 66 Slovenia 240 94 39 146 61 Madagascar 443 229 52 214 48 South Africa 930 697 75 233 25 Malawi 141 102 72 39 28 Sri Lanka 567 365 64 202 36 Mali 804 260 32 544 68 St. Kitts and Nevis 146 100 68 46 32 Mauritania 237 40 17 197 83 St. Lucia 150 71 47 79 53 Mauritius 386 241 62 145 38 St. Vincent and Grenadines 150 118 79 32 21 Mexico 2,646 1,188 45 1,458 55 Suriname 152 79 52 73 48 Micronesia 67 17 25 50 75 Swaziland 286 209 73 77 27 Moldova 276 53 19 223 81 Tajikistan 275 57 21 218 79 Mongolia 332 263 79 69 21 Tanzania 414 220 53 194 47 Montenegro 102 52 51 50 49 Timor Leste 146 29 20 117 80 Mozambique 479 195 41 284 59 Togo 148 78 53 70 47 Namibia 324 269 83 55 17 Tonga 147 75 51 72 49 Nepal 843 674 80 169 20 Trinidad and Tobago 364 298 82 66 18 Nicaragua 705 292 41 413 59 Turkey 1,117 689 62 428 38 Niger 146 75 51 71 49 Uganda 1,152 536 47 616 53 Nigeria 1,887 305 16 1,582 84 Ukraine 705 207 29 498 71 Pakistan 869 220 25 649 75 Uruguay 1,176 434 37 742 63 Panama 771 600 78 171 22 Uzbekistan 273 98 36 175 64 Paraguay 934 284 30 650 70 Vanuatu 126 55 44 71 56 Peru 1,473 485 33 988 67 Venezuela 678 523 77 155 23 Philippines 1,171 1,067 91 104 9 Vietnam 1,005 343 34 662 66 Poland 405 136 34 269 66 Yemen 469 149 32 320 68 Romania 467 170 36 297 64 Zambia 481 341 71 140 29 Russia 4,497 1,058 24 3,439 76 Zimbabwe 532 285 54 247 46 Rwanda 446 214 48 232 52 46 Table A2: Variables and Sources Original Variable Definition - t is the survey year Source Dummy variable that equals 1 if firm’s annual financial statement was Audit checked and certified by an external auditor (WBES data item ‘k21’), 0 WBES otherwise. “How problematic is access to finance for the current operations of a Financial business?” No Obstacle =0, Minor Obstacle =1, Moderate WBES Obstacle Obstacle=2, Major Obstacle =3, and Very Severe Obstacle=4 (WBES data item ‘k30’). Financial Dummy variable that equals 1 if financial obstacles equal to 2 WBES ObstacleDummy (moderate), 3 (major), or 4 (very severe), and 0 otherwise. “How problematic is corruption for the current operations of a Corruption business?” No Obstacle =0, Minor Obstacle =1, Moderate WBES Obstacle Obstacle=2, Major Obstacle =3, and Very Severe Obstacle=4 (WBES data item ‘j30f’). Corruption Dummy variable that equals 1 if corruption obstacles equal to 2 WBES ObstacleDummy (moderate), 3 (major), or 4 (very severe), and 0 otherwise. The average difference of sales in year (t) (WBES data item ‘d2’) and SalesGrowth WBES sales in year (t-2) (WBES data item ‘n3’). FirmSize Logarithm of firm’s sales at the end of year (t-1) (WBES data item ‘d2’). WBES Logarithm of a firm’s actual age, age=survey year – firm founding year FirmAge WBES (WBES data item ‘b5’). “How many years of experience working in this sector does the top Experience WBES manager have?” (WBES data item ‘b7’) Dummy variable that equals 1 if firm is owned by government/ state Government WBES (WBES data item ‘b2c’), 0 otherwise. Dummy variable that equals 1 if any foreign company or individual has Foreign a financial stake in the ownership of the firm (WBES data item ‘b2b’), 0 WBES otherwise. Dummy variable that equals 1 if the firm answered ‘Yes’ to the Compete question: “Does this establishment compete against unregistered or WBES informal firms?” Exporter Dummy variable that equals 1 if the firm is an exporter, 0 otherwise. WBES (continued) 47 Table A2-Continued Original Variable Definition - t is the survey year Source Private credit by deposit money banks to GDP, calculated using the following deflation method: {(0.5)*[Ft/P_et + Ft-1/P_et- Priv IFS 1]}/[GDPt/P_at] where F is credit to the private sector, P_e is end-of period CPI, and P_a is average annual CPI. Corruption Country-level corruption estimate of Worldwide Governance Indicators (WGI) and ranges from approximately -2.5 (weak) to 2.5 (strong) WGI Control governance performance Country-level estimate of Worldwide Governance Indicators (WGI) for the quality of public services, the quality of the civil service and the degree Government of its independence from political pressures, the quality of policy WGI Effectiveness formulation and implementation, and the credibility of the government's commitment to such policies. It ranges from approximately -2.5 (weak) to 2.5 (strong) governance performance GDP Logarithm of GDP in current US$, the average over year (t-3), (t-2) WDI and (t-1). GDP per Logarithm of per capita in US$, the average real GDP per capita over WDI capita year (t-3), (t-2) and (t-1). GDPGrowth Real growth rate of GDP, the average over year (t-3), (t-2) and (t-1). WDI Log difference of consumer prices, the average over year (t-3), (t-2) and Inflation WDI (t-1). Dummy variable equal to 1 if firm has an overdraft facility at year t Overdraft WBES (WBES data item ‘k7’), and 0 otherwise. Line of Dummy variable equal to 1 if firm has a line of credit or loan from a WBES Credit financial institution at year t (WBES data item ‘k8’), and 0 otherwise. Capital Dummy variable that equals 1 if firm purchased fixed assets, such as machinery, vehicles, equipment, land or buildings at year (t-1), (WBES WBES Expenditure data item ‘k4’) and 0 otherwise. “How problematic are business licensing and permits for the current License operations of a business?” No Obstacle =0, Minor Obstacle =1, WBES Moderate Obstacle=2, Major Obstacle =3, and Very Severe Obstacle=4 (WBES data item ‘j30c’). “How problematic are tax rates for the current operations of a Tax business?” No Obstacle =0, Minor Obstacle =1, Moderate WBES Obstacle=2, Major Obstacle =3, and Very Severe Obstacle=4 (WBES data item ‘j30b’). The Property An estimate of the legally protected freedom to accumulate private Heritage Rights property and wealth by workers and investors. Foundation * Sources of Data: WBES = World Bank Enterprise Survey (WBES); WDI = World Development Indicators, World Bank; WGI = Worldwide Governance Indicators, World Bank; IFS = International Financial Statistics. 48