Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Economic Uncertainty and Corruption: Evidence From Public and Private Firms

Economic Uncertainty and Corruption: Evidence From Public and Private Firms E Ec co on no om miic c u un nc ce er rtta aiin ntty y a an nd d c co or rr ru up pttiio on n:: e ev viid de en nc ce e ffr ro om m p pu ub blliic c a an nd d p pr riiv va atte e ffiir rm ms s Mansoor Afzali, Gonal Colak, Mengchuan Fu P Pu ub blliic ca attiio on n d da atte e 01-12-2021 L Liic ce en nc ce e This work is made available under the CC BY 4.0 licence and should only be used in accordance with that licence. For more information on the specific terms, consult the repository record for this item. D Do oc cu um me en ntt V Ve er rs siio on n Published version C Ciitta attiio on n ffo or r tth hiis s w wo or rk k ( (A Am me er riic ca an n P Ps sy yc ch ho ollo og giic ca all A As ss so oc ciia attiio on n 7 7tth h e ed diittiio on n) ) Afzali, M., Colak, G., & Fu, M. (2021). Economic uncertainty and corruption: evidence from public and private firms (Version 2). University of Sussex. https://hdl.handle.net/10779/uos.23487545.v2 P Pu ub blliis sh he ed d iin n Journal of Financial Stability L Liin nk k tto o e ex xtte er rn na all p pu ub blliis sh he er r v ve er rs siio on n https://doi.org/10.1016/j.jfs.2021.100936 C Co op py yr riig gh htt a an nd d r re eu us se e:: This work was downloaded from Sussex Research Open (SRO). This document is made available in line with publisher policy and may differ from the published version. Please cite the published version where possible. Copyright and all moral rights to the version of the paper presented here belong to the individual author(s) and/or other copyright owners unless otherwise stated. For more information on this work, SRO or to report an issue, you can contact the repository administrators at sro@sussex.ac.uk. Discover more of the University’s research at https://sussex.figshare.com/ Journal of Financial Stability 57 (2021) 100936 Contents lists available at ScienceDirect Journal of Financial Stability journal homepage: www.elsevier.com/locate/jfstabil Economic uncertainty and corruption: Evidence from public and private firms a b, * c Mansoor Afzali , Gonül Ҫolak , Mengchuan Fu Hanken School of Economics, Department of Accounting, P.O. Box 479, FI-00101 Helsinki, Finland Hanken School of Economics, Department of Finance and Economics, P.O. Box 479, FI-00101 Helsinki, Finland Fordham University, Gabelli School of Business, 45 Columbus Avenue, Fifth Floor, New York, NY 10023, United States ARTICLE INFO ABSTRACT JEL classification: We study the influence of policy uncertainty on the moral behavior of firms. When facing uncertainty, managers D80 perceive various socioeconomic obstacles as more severe and disruptive to their business. Using data from policy F30 uncertainty spouts in 93 countries, we document that some firms engage in norm-deviant behavior by cheating G38 on taxes and paying more bribes. While private firms prefer to cheat on taxes, public firms choose bribery as a O43 favorite tool to “grease the wheels” during periods of uncertainty. Strong social capital (local trust and religi- P48 osity) breaks this link between uncertainty and corruption. Keywords: Economic policy uncertainty Private firms Corruption Bribery Cheating on taxes Trust Religiosity 1. Introduction (Bonaime et al., 2018; Nguyen and Phan, 2017), and cash holdings (Phan et al., 2019). Political and regulatory authorities generate uncertainty through Most of this research, however, focuses on publicly traded large their decision-making process, and this uncertainty can influence firms ’ corporations. We add to this literature by studying the influence of business operations (Bloom, 2009). Several recent studies have linked policy uncertainty on small firms throughout the world, most of which policy-related economic uncertainty (policy uncertainty) to various are private firms. While we analyze both private and public firms, we managerial decisions by public firms. Julio and Yook (2012), Gulen and try to shed more light on private enterprises since they make up more Ion (2016), and Jens (2017) show that policy uncertainty has a sub- than 99% of the business entities in most countries, and they have in stantial impact on firms ’ investment decisions, and Bhattacharya et al. aggregate four times as many employees, three times the revenues, and (2017) report that it affects firms ’ innovativeness. This uncertainty also twice as many assets as listed firms (Chen et al., 2011). For instance, the affects public corporations’ decisions on external financing (Ashraf and privately held firms in the United States produce 51% of the gross na- Shen, 2019; Chan et al., 2021; Colak et al., 2017; Datta et al., 2019; tional output and employ more than half of the labor force (Nagar et al., Francis et al., 2014; Gungoraydinoglu et al., 2017), acquisition activity 2011). In the European Union, small firms with less than 250 employees, This paper has been circulated also under the name “Resorting to Corruption When Facing Policy Uncertainty: Global Evidence.” The authors would like to thank the editor (Iftekhar Hasan), the two anonymous referees, and the seminar participants at Hanken School of Economics and University of Southern Denmark. * Corresponding author. E-mail addresses: mansoor.afzali@hanken.fi (M. Afzali), gonul.colak@hanken.fi (G. Ҫolak), mfu10@fordham.edu (M. Fu). The link between privately held (non-publicly trading) firms and political uncertainty has been analyzed within specific countries. Amore and Minichilli (2018), for instance, analyze the investment behavior of family firms in Italy that are exposed to economic uncertainty. An et al. (2016) analyze the investments of private vs. state-owned enterprises in China, although the term “private firms ” in the context of China encompasses those listed on the exchange that are not owned by the local or federal government. https://doi.org/10.1016/j.jfs.2021.100936 Received 22 March 2021; Received in revised form 25 August 2021; Accepted 27 August 2021 Available online 8 September 2021 1572-3089/© 2021 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 employ more than 66% of the labor force (Airaksinen et al., 2016). to engage in corruption in order to “grease the wheels” of their busi- While researchers have examined financing and growth constraints in nesses. Similarly, Husted (1999) claim that when uncertainty avoidance private firms (for instance, Beck et al., 2005; D’Souza et al., 2017; Hope (a la Hofstede, 1980) in a country is high, firms engage more in corrupt et al., 2011; Mertzanis, 2019), we do not know how these small and behavior. Rodriguez et al. (2005) note that corrupt economies are often relatively more fragile firms cope with the pressure generated by the characterized by widespread uncertainties. When faced with uncer- rising uncertainty. tainty, managers will rationalize their norm-deviant behavior and will The main obstacle in studying private firms ’ reaction to policy un- even make it part of the corporate culture (Anand et al., 2004; Ashforth certainty is the lack of data on them in many smaller countries and the and Anand, 2003). In light of these findings, we hypothesize that most unavailability of a proper policy uncertainty measure in those countries. firms from around the world will reduce their exposure to economic Using the enterprise-level data from the World Bank Enterprise Survey uncertainty shocks by engaging in corruption. (henceforth WBES), we study how policy uncertainty affects private To test this prediction, we examine two aspects of business activity firms in various countries. This data not only provides detailed infor- that pertain to corruption. These include, cheating on taxes and paying mation on the structure and financials of both private and public firms, bribes. We show that during times of high policy uncertainty, small but the survey also seeks the managers’ opinions on the main obstacles businesses cheat more on taxes and pay more bribes (both measured as (inefficiencies) in the business environment. This survey data is avail- percent of sales). All of these results are statistically significant at the 1% able for around 146,000 firms in 143 countries. level and are robust to the inclusion of several firm-level and country- The unavailability of a local measure of policy uncertainty for many specific characteristics as well as industry, country, and year fixed ef- countries around the world has been a major challenge for researchers. fects. The results are economically significant too; for instance, a one While the news-based index of Baker et al. (2016) is widely used to standard deviation increase in our economic uncertainty measure capture economic policy uncertainty, it is predominantly available for translates to 3.5% less sales reported to the government for tax purposes. the developed or the large developing nations. To address this chal- We also document an interesting divergence between private versus lenge, Ahir et al. (2019) develop a world uncertainty index for 143 public firms ’ reaction to economic uncertainty. Private firms ’ preferred countries for each quarter since 1996Q1. Their method relies on textual form of corruption is cheating on taxes, probably because they find it analysis of the Economist Intelligence Unit’s (EIU) quarterly country easier to fudge their actual sales numbers subject to taxes. Public firms, reports, and it is designed in a similar fashion to Baker et al.’s (2016) on the other hand, are relatively less likely to cheat on taxes and instead economic policy uncertainty index. prefer to cope with uncertainty by paying more bribes. This is a novel We combine this world uncertainty index (WUI) with the WBES data finding for the literatures on corruption and policy uncertainty. for the period of 2002–2018 to get a large sample of 96,769 firm-year To improve identification and to increase the reliability of our observations for small business firms from 93 countries that cover 16 findings, we use instrumental variable (IV) regressions, where election regions within Africa, Americas, Asia, and Europe. Using this sample, we closeness and the proportion of seats held by the largest opposition party assess how economic uncertainty influences the business environment in the legislature serve as instruments for uncertainty. We continue to that firms operate in. Since our data is survey based, our goal is not to find higher levels of corruption after using these proxies and IV re- establish a causal link between uncertainty and business constraints but gressions. Using a carefully balanced sample obtained by applying rather to provide evidence that uncertainty alters managerial perception propensity score matching, where we match on all firm and country- of the severity of the business constraints that are prevalent in a country. level characteristics, produces similar results. To further improve the We find that as policy uncertainty rises, the managerial opinion robustness of our results and to reduce the effects of measurement error (measured by the responses to survey questions) about the severity of in our estimations, we employ four alternative measures of economic business obstacles increases. These obstacles include, a dysfunctional policy uncertainty. First, using the voting outcomes from the national court system, high crime levels (theft and disorder), various forms of elections in each country (as in Julio and Yook, 2012), we construct a public sector corruption, high tax rates, ineffective tax administration, measure (Election closeness) to capture the uncertainty generated around anti-competitive behavior of informal businesses, reduced access to contested elections. Second, we construct the orthogonalized measure of finance, and political instability. policy uncertainty by running a regression of country-level index This relationship between uncertainty and managerial opinion is to (country-specific WUI) on the aggregate global uncertainty (available in be expected because individuals (firm managers) tend to perceive Ahir et al., 2019) and use the residuals from this regression as our un- increased uncertainty as a threat to their livelihood (Hogg, 2007). As certainty measure in the second stage. Third, we calculate stock market studies have shown that uncertainty amplifies people’s reactions to volatility by estimating the standard deviation in returns on the main negative events (Arenas et al., 2006), managers are likely to react more stock index in each country. Fourth, we use the news-based index of negatively to existing market frictions and business obstacles. That is, Baker et al. (2016) for countries with available data on electronic news the unpredictability of future economic and political outcomes induces articles. These four alternative measures do not alter our qualitative an environment of uncertainty and inefficiency. Leff (1964), Acemoglu conclusions. and Verdier (2000), and Mendoza et al. (2015) argue that when eco- Policy uncertainty can directly affect corruption by inducing external nomic agents think that a country’s institutions are ineffective, they tend conditions that incentivize norm-deviant behavior in firms. It can also affect corruption indirectly by changing managerial perception regarding obstacles facing their businesses that in turn can trigger corrupt managerial actions. To test this indirect affect, we perform a See https://www.enterprisesurveys.org/en/data. formal mediation analysis (Baron and Kenny, 1986). Our mediation tests As of January 2020, the measure was available for only 23 countries: show that worsened perception of managers regarding business obsta- Australia, Brazil, Canada, Chile, China, Colombia, France, Germany, Greece, cles explains 4.2% of the relationship between economic uncertainty Hong Kong, India, Ireland, Italy, Japan, Mexico, Netherlands, Russia, and cheating on taxes. The indirect effect of paying bribes explains about Singapore, South Korea, Spain, Sweden, the United Kingdom, and the United 18.5% of the total effect. Thus, economic uncertainty affects corrupt States. behavior both directly and indirectly through the mediating role of In this study, we use the concepts of economic uncertainty and policy un- managerial perception. certainty as close substitute for each other. While conceptually there can be We next explore how cross-country heterogeneity affects the rela- certain differences between these uncertainties, due to lack of separate country- tionship between policy uncertainty and corruption. We first look at the level measures for each type of uncertainty, we work with the WUI, which is constitutional framework of the country. A parliamentary system is referred to by its creators (Ahir et al., 2019) with the generic name of “world uncertainty index.” characterized by simultaneous changes in the control of the 2 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 government’s executive branch and the legislative branch that thereby trust in others and religiosity reduce such corruption. Overall, these increases the freedom of politicians when trying to consolidate de- results are consistent with the idea that uncertainty induces an envi- mocracy (Stepan and Skach, 1993). A presidential system, on the other ronment where norm-deviant behavior prevails, and a country’s social hand, is characterized by a high degree of checks and balances that often capital ameliorates this negative effect. minimize policy fluctuations and imposes constraints on the adoption of new laws and regulations. Julio and Yook (2012) and Gungoraydinoglu 2. Theoretical motivation and hypotheses development et al. (2017) argue that in general, presidential systems have less po- litical uncertainty; that is, parliamentary systems are more likely to Studies like Bernanke (1983), Bloom (2009), and (Pastor and Ver- experience greater policy-related economic uncertainty. We therefore onesi, 2012) have set the theoretical foundations of the literature on the predict the relationship between policy uncertainty and corruption to be consequences of economic policy uncertainty (Pastor and Veronesi, more significant in countries with a parliamentary system of govern- 2012). As predicted by these studies, empirical tests confirm that eco- ment. To test this prediction, we interact our main uncertainty measure nomic policy uncertainty has broad macroeconomic and financial im- (WUI) with an indicator variable for a country with a parliamentary plications (Petersen, 2009; Bernanke, 1983; Bloom et al., 2007; Leahy system of government. We find that firms located in these countries and Whited, 1996). Similarly, reports prepared after the 2008 financial cheat more on taxes. However, we do not find any statistically signifi - crisis show that the period of high economic uncertainty that followed cant differences between the levels of bribes under the two systems. We the recession created business obstacles and impeded growth and re- also explore whether this uncertainty-corruption relationship is influ - covery (International Monetary Fund, 2012). In particular, policy un- enced by the economic conditions in a country. We predict that this certainty lowers investment and delays the implementation of projects relationship will be stronger in low-income regions due to the general (Dixit and Pindyck, 2012; Drobetz et al., 2018), reduces employment lack of oversight and transparency. Consistent with our prediction, growth (Bentolila and Bertola, 1990), boosts interest rates (Ashraf and private firms located in low-income countries have a greater tendency to Shen, 2019), and suppresses asset valuations (Brogaard and Detzel, engage in cheating on taxes and paying bribes when economic uncer- 2015; Kelly et al., 2016; Kim et al., 2012; Pastor and Veronesi, 2012, tainty increases. 2013). Studies have indicated that cultural and sociological norms in the The above findings imply that policy uncertainty can disrupt firms’ country can mitigate norm-deviant behavior. For instance, Kanagar- regular business activities and in turn can create new impediments or etnam et al. (2018) use a sample of firms from 25 countries to show that can exacerbate existing frictions in the economy. Thus, we first analyze societal trust is negatively correlated with corporate tax avoidance. how policy uncertainty affects managers’ perceptions about the efficient Aghion et al. (2010) claim that trust is vital when country-level gover- functioning of the regulatory authorities, the governmental agencies, nance institutions are weak. Similarly, studies have shown that religi- and various other socioeconomic institutions within a country. The osity affects corporate outcomes (Callen and Fang, 2015; Jiang et al., research in different studies indicates that individuals perceive various 2018; McGuire et al., 2012). Therefore, we examine how societal trust types of uncertainties as a threat to themselves (Hogg, 2007). Uncer- and religious norms affect the link between uncertainty and corruption. tainty further intensifies people’s reactions to negative events (Arenas To measure trust and civic norms (social capital), we follow Knack and et al., 2006; Van Den Bos et al., 2007). Thus, managers should react Keefer (1997) and use the World Values Survey (WVS). Specifically, we more negatively to existing economic frictions and business obstacles define social capital as the average of trust in others and norms of civic when the uncertainty about their firms’ future increases. On the basis of cooperation. Our religiosity measure is an indicator variable equal to the above arguments, we formulate our first hypothesis: one if more than 70% of the country’s residents replied “yes” when Hypothesis 1. Policy uncertainty worsens managers’ perception about the asked, “Is religion important in your daily life?” in a Gallop Poll in 2009, severity of the business obstacles prevalent in a country. and zero otherwise. Using these measures, we show that both societal norms and religiosity mitigate the uncertainty-corruption relationship. If business obstacles are perceived to be more severe by the firm We contribute to the literature in several ways. First, we extend the managers during periods of high uncertainty, then these managers literature on policy uncertainty to private firms. Due to lack of data on should take precautionary action to ameliorate the effects of such ob- private firms, most of the literature about policy uncertainty focuses on stacles on their businesses. Put differently, the unpredictability of eco- publicly traded firms. For instance, Nagar et al. (2019) assert that policy nomic and political outcomes can create an uncertain environment in uncertainty contributes to increasing information asymmetry between which firms constantly look for ways to mitigate their exposure to these public firms and their investors. Boutchkova et al. (2012) claim that uncertainties. One such remedy is engaging in corruption and other local policy uncertainty translates into higher systematic (stock market) illegal practices. Husted (1999) and Rodriguez et al. (2005) argue that risk, while global shocks result in higher firm-specific return volatility. firms engage in corruption to reduce the uncertainty in their business Politically motivated local (state level) fiscal shocks also tend to reduce dealings with government officials. Using data from India, Collins et al. the productivity and the output of local public firms (Cohen et al., 2011). (2009) show that corruption can be costly to the society but vital in Furthermore, heightened policy debates slow down the public firm’s reducing regulatory uncertainty for firms, especially when the managers ability to adjust to the optimal capital structure by using stock issuances perceive corruption as increasing their firms’ survival chances. Man- (Colak et al., 2018). We contribute to this line of research by showing agers then rationalize their norm-deviant behaviors, and corruption that policy uncertainty can induce corrupt business practices such as becomes part of the corporate culture (Anand et al., 2004; Ashforth and bribery and tax cheating in both public and private firms. Anand, 2003). We are the first to link policy uncertainty to corruption in private Consistent with this view, we argue that uncertainty increases the firms. While some studies have linked uncertainty to corruption (e.g., unpredictability of future economic and political outcomes that explains Braun and Di Tella, 2004; Goel and Ram, 2013; Treisman, 2007), they why firms engage in shady behavior to reduce their exposure to the largely confine their empirical evidence to cross-country analyses using potentially detrimental consequences of uncertainty. For instance, high macro-level corruption indices. We fill this gap by studying how policy policy uncertainty can induce firms to engage in more aggressive po- uncertainty effects the behavior of small firms. During times of high litical lobbying activities (Shang et al., 2021). Kaufmann and Wei (2000) economic uncertainty, firms face more severe obstacles that impede their business, and they respond by engaging in norm-deviant behavior. While private firms cheat through underreporting sales and paying lower taxes, the public firms prefer to “grease the wheels” through bribery. Further, our results also indicate that societal norms such as 3 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 claim that firms engage in corruption to reduce various inefficiencies. delay their decision to raise new equity capital. However, the rest of the At the country level, there is some evidence that macroeconomic un- private firms proceed with their IPO events and are willing to accept certainty has a positive association with corruption; Braun and Di Tella lower price-to-value ratios for their equity. (2004) and Treisman (2007) find that countries with high variability in In view of the differences in the nature of public and private firms, inflation have higher levels of corruption. Cohen and Malloy (2014) the assumption that they may adopt different norm-deviant behavior show that political favoritism (a form of corruption) is prevalent within when faced with high economic uncertainty is a reasonable one. Typi- US Senate as alumni’s network connections to powerful Senate com- cally, compared with public firms, private firms have less pressure from mittee chairmen help those ex-senators secure additional discretionary the capital market and are therefore more likely to actively reduce taxes earmark funding for local constituents. by deliberately reporting lower financial earnings (Slemrod, 2007). Many studies have pointed to the costs and benefits of corruption Conversely, because public firms need to disclose their activities through when widespread inefficiency prevails and creates serious economic financial reports or discussions with analysts, the perceived benefits of frictions (Huntington, 1968; Meon and Sekkat, 2005; Shleifer and tax evasion may be less than the perceived costs of putting themselves in Vishny, 1993). For instance, corruption can facilitate trade and improve a disadvantaged position (Cloyd et al., 1996). Additionally, due to the efficiency by allowing private sector agents to circumvent cumbersome high concentration of ownership, private firms generally have lower regulations (Bardhan, 1997; Leff, 1964; Mauro, 1995). Furthermore, costs for financial reporting that are associated with reducing taxes and Leff (1964), Acemoglu and Verdier (2000), Meon and Weill (2010), and are more likely to conduct transactions that generate tax savings Mendoza et al. (2015) theorize that when a country’s institutions are (Klassen, 1997). Therefore, we predict that private firms will be more ineffective, corruption is a useful tool to “grease the wheels”, especially willing to cheat through taxation to deal with economic uncertainties. at the small firm level. Although both tax evasion and bribery are corrupt practices, their We build on this literature and argue that a country’s economic penalty functions are different. In the process of tax evasion, the firm uncertainty increases the severity of business constraints for all firms may be held liable for the fraudulent acts of its internal accountant. But (Hypothesis 1). We further predict that these conditions induce an in the process of bribery, individuals may be liable (e.g., imprisonment) environment of inefficiency. To reduce this inefficiency, some managers if involved in any corrupt payment scheme. As the ownership of private can engage in paying bribes and cheating on taxes. This idea is largely firms is more concentrated, their managers will increase their efforts to consistent with the theoretical models of Aguilera and Vadera (2008) in control bribery to avoid responsibility (Clarke and Xu, 2002). The scale which they argue that corruption occurs when there is a context or of bribery can also vary with firm ’s bargaining power (Svensson, 2003). environment that makes such actions feasible and there are factors that Given that public firms and public officials may have a closer relation- encourage agents to act in such ways. ship, bribery by public firms may be more cost effective than that by private firms. Our next hypothesis focuses on such differences between Hypothesis 2. Higher policy uncertainty is associated with more severe public and private firms: corruption in the form of cheating on taxes or paying more bribes by firms. Hypothesis 3. High policy uncertainty induces a different norm-deviant Studies have shown well that policy uncertainty can influence a behavior in private firms than in public firms; private firms prefer cheating manager’s decisions at the firm level. For instance, Julio and Yook through taxes and public firms prefer paying more bribes. (2012), Gulen and Ion (2016), and Jens (2017) show that policy un- certainty influences a firm ’s investment decisions. Political uncertainty The social norm theory argues that social constructs such as trust and can also affect the financing of projects by delaying the decision to go civic norms reduce norm-deviant behavior among all economic agents. public (Colak et al., 2017) or by impeding equity or bond issuance Based on the findings of Putnam et al. (1994), La Porta et al. (1997), and (Colak et al., 2018; Francis et al., 2014; Gungoraydinoglu et al., 2017; Fisman and Miguel (2007), we argue that societal trust is a reliable Chan et al., 2021; Datta et al., 2019). Policy uncertainty is also shown to deterrent of corruption. For instance, Kanagaretnam et al. (2018) show affect technological innovation (Bhattacharya et al., 2017) and cash that societal trust is negatively associated with corporate tax avoidance. holdings (Phan et al., 2019). However, almost everything we know Consistent with Adelopo and Rufai (2020), we argue that trust inhibits about the relationship between policy uncertainty and managerial de- corruption. Furthermore, following the arguments in Aghion et al. cisions is based on the behavior of publicly listed firms that are in the US. (2010), we predict that trust becomes vital when country-level gover- In this study, we focus on private firms across the world and their nance institutions are weak. Similarly, society’s religiosity can reduce managerial decisions under uncertainty. We contrast their behavior to corporate misconduct such as financial reporting fraud (McGuire et al., the behavior of public firms. How exactly small private firms cope with 2012). Mensah (2014) also argues that religion reduces the levels of uncertainty is still unclear. There is some scant evidence that indicates perceived corruption in the country. Thus, we hypothesize that societal private firms react somewhat differently to policy uncertainty than trust and religiosity will break the link between policy uncertainty and public firms. A study by Amore and Minichilli (2018), for instance, fo- corruption for all firms: cuses on a single country (Italy) and shows that family firms are more Hypothesis 4. Societal norms such as trust and religiosity will ameliorate likely to invest during periods of high uncertainty. Colak et al. (2017) the corruptive effects of higher policy uncertainty on managerial behavior in shows that when facing policy uncertainty, about 15% of private firms all firms. Kaufmann and Wei (2000) also provide evidence that when firms pay more bribes, they spend more, not less time, with the governmental authorities. In untabulated results, we find results consistent with this argument. The literature on private firms ’ decision-making is growing. Caballero et al. (1995), Cooper and Haltiwanger (2006), and Asker et al. (2011) claim that Different dimensions of social capital can have different effects on the level private firms ’ investment policies are different than public firms. Michaely and of corruption (Pena Lopez ´ and S´ anchez Santos, 2014). In particular, strong Roberts (2012) show that private firms smooth dividends less than public firms. bonds between individuals can positively relate to corruption. On the other Gao and Li (2015) report that private firms ’ CEO compensation is less sensitive hand, linking and bridging social capital can inhibit individuals or groups from to performance; however, Edgerton (2012) shows that public firms overuse engaging in corruption. Recent evidence, however, shows that both civic norms corporate jets compared to private firms. Financing options, growth constraints, and dense social networks can mitigate norm deviance and reduce transaction and capital structure policies are also different in private firms (Beck et al., costs (Hasan et al., 2017a, 2017b). In our analysis, we measure trust using the 2005; Brav, 2009; D’Souza et al., 2017; Hope et al., 2011; Mertzanis, 2019; civic norm (linking and bridging) component of social capital and expect a Saunders and Steffen, 2011). negative link between trust and corruption. 4 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 3. Research design and sample selection perceptions about the seriousness of the obstacle. In this section, we describe the measures for economic policy un- 3.3. Sample selection certainty, the details of our empirical design, and the sample selection procedure. We also present the summary statistics. We start with the raw firm-level data from the WBES. This initial sample has 104,638 firm-years of which 4897 (or 4.7% of the sample) belong to publicly listed firms. Since in WBES all values of sales are in 3.1. World economic uncertainty index local currency units, we convert these values to US dollars using the exchange rate at the end of the month in which the survey was con- We define economic policy uncertainty using the newly developed ducted. Consistent with the “grease the wheels” hypothesis (Acemoglu world uncertainty index (WUI) of Ahir et al. (2019). The WUI utilizes the and Verdier, 2000; Mendoza et al., 2015), we focus on small and Economist Intelligence Unit’s (EIU) quarterly country reports to capture medium-level firms. We, thus, screen out the very large firms, as well as the economic, political, and financial trends in a country by using a the firms whose sales are missing or negative. These screening only five-step comprehensive process that is developed by experienced ana- affects about 1.9% of our original sample of firms. lysts within each country. Ahir et al. (2019) construct the index for each We then match the data from this survey with the uncertainty index. country by counting the frequency of the word “uncertainty” and its This index is constructed for each quarter, and we therefore take the variants in the EIU reports. To make the index comparable, they divide timing into consideration when matching. Using the date of the inter- the word count by the total number of words used in the report. The view we are able to precisely match the firm data (WBES) to the WUI uncertainty index does well in capturing important global events and is measure using country, year, and quarter. We remove any country positively correlated with the index for economic uncertainty and the with fewer than 50 observations per year (affects only 184 firm-years). volatility in the stock market. To further smooth the idiosyncratic usage Our final sample contains 103,738 firm-year observations from 93 of the word “uncertainty” in a particular quarterly report and to remove countries that cover the period from 2002 to 2018. idiosyncratic spikes in economic uncertainty that may not last long Table 1 provides the sample composition with the number of ob- enough to change managers’ behavior, we use the three-quarter servations for each country, the number of surveys of each country, and weighted moving average of this index as our main uncertainty mea- the mean uncertainty index. The sample is highly inclusive with gross sure. To reduce potential measurement problems related to the WUI, we domestic product (GDP) per capita ranging from $225 in Burundi to use three alternative measures of uncertainty in a series of robustness $27,698 in Greece. The largest number of observations comes from India checks (see Section 5). These alternative measures are available for a (10,322), followed by the Russian Federation (3004), and Chile (2721). smaller set of countries. The mean policy uncertainty index is higher in the Central Asian countries, the Middle East, and Africa. For majority of our sample 3.2. World Bank’s Enterprise Surveys (WBES) countries, the survey is conducted two or more times. We can therefore use country and year fixed effects that should increase the identification The World Bank’s Enterprise Survey (WBES) covers a wide range of strength of our tests. topics related to a country’s business environment, such as access to The survey data from WBES are well populated for the question(s) financing, gender, corruption, infrastructure, innovation, competition, related to bribery and are used to create our left-hand side variable that informality, and performance indicators. The survey asks a series of measures the percentage of sales paid as bribes during a given year. questions and allows business owners and top managers to express their However, the survey data related to firms’ cheating on taxes is available views on the business environment, growth opportunities, financial and for 68 out of 93 countries in our sample. Thus, when analyzing the tax legal barriers, and corruption issues that help to identify the obstacles cheating behavior of firms, our sample size is reduced to 30,032 firm- that hinder the performance and growth of firms. It is considered to have year observations. the most comprehensive firm-level data on emerging markets and The data on macro-level variables for each country are obtained from developing economies from all over the world. While this survey is conducted in several waves, the World Bank conveniently provides combined datasets for several countries for the periods of 2002–2005 According to WBES guidelines, negative sales values indicate that the firm and 2006–2019. Despite some differences, the two datasets are largely 8 did not report their sales during the survey. We also exclude all firms with sales the same and compatible with each other. equal to zero since they are likely to be micro firms with no activity during the The financial accounting studies have used the WBES data before. year. To focus on small firms, we place an upper limit of $100 million in sales. They have used it to study growth (Beck et al., 2005; D’Souza et al., WBES does not provide the exact date of the interview for the 2002–2005 2017), financing constraints (Hope et al., 2011; Mertzanis, 2019; sample. We therefore assume that the interview was conducted in the fourth Wellalage et al., 2019), and product innovation (Lederman, 2010; Xie quarter of the year. In robustness checks, we change this to the first, second, et al., 2019) for firms as well as the association between auditing and third quarters. Our baseline results remain qualitatively unchanged. infrastructure and bribery mitigation (Khalil et al., 2015), recovery of If countries appeared only once in the sample, our results could simply SMEs from a financial crisis (Lawless et al., 2015), and the link between reflect the cross-country differences. For instance, firms in countries with higher uncertainty are also likely to face more obstacles. In our analyses, this investment efficiency and financial reporting quality (Chen et al., 2011). issue is of little concern since we apply country and year fixed effects. We also From the WBES data we identify eight major obstacles (in- control for several other macro-level variables which further reduce the prob- efficiencies) for which all private firms have sufficient data. These ob- ability that we capture mere cross-country differences. stacles are the court system, crime, corruption, tax rate, tax Information on cheating on taxes is available for Albania, Angola, administration, anticompetitive behavior of the informal sector, Argentina, Armenia, Belarus, Benin, Bolivia, Bosnia-Herzegovina, Botswana, financing, and political instability. The severity of an obstacle is Brazil, Bulgaria, Burundi, Cambodia, Chile, Colombia, Croatia, Czech Republic, measured by the survey, and thus it reflects managers’ opinions and Dominican Republic, Ecuador, El Salvador, Georgia, Ghana, Greece, Guatemala, Guinea, Honduras, Hungary, India, Indonesia, Jamaica, Jordan, Kazakhstan, Kenya, Kyrgyzstan, Latvia, Lebanon, Liberia, Lithuania, Madagascar, Malawi, We merge the two datasets on all required identification variables. The Mali, Mauritania, Mexico, Moldova, Mongolia, Namibia, Nicaragua, Panama, industries are defined slightly differently in the two datasets, and so we Paraguay, Peru, Poland, Republic of North Macedonia, Romania, Russian manually adjust the industry names in the 2002–2005 dataset to correspond to Federation, Rwanda, Senegal, Slovakia, Slovenia, South Africa, Tajikistan, the 2006–2019 dataset. After combining the datasets, we remove the duplicates Tanzania, Turkey, Uganda, Ukraine, Uruguay, Uzbekistan, Vietnam, and based on the unique identification number and year of the firm. Zambia. 5 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 1 Sample composition. Country Obs. No. of surveys Uncertainty index Country Obs. No. of surveys Uncertainty index Afghanistan 425 2 0.15 Liberia 258 2 0.06 Albania 609 3 0.07 Lithuania 693 4 0.04 Angola 647 2 0.00 Madagascar 525 2 0.05 Argentina 2654 3 0.15 Malawi 584 3 0.08 Armenia 711 3 0.02 Malaysia 708 1 0.08 Azerbaijan 402 2 0.00 Mali 930 4 0.08 Bangladesh 2537 2 0.07 Mauritania 327 2 0.00 Belarus 986 4 0.02 Mexico 2646 2 0.09 Benin 380 3 0.07 Moldova 705 4 0.09 Bolivia 980 3 0.07 Mongolia 602 3 0.08 Bosnia and Herzegovina 642 3 0.07 Morocco 239 1 0.06 Botswana 541 2 0.03 Mozambique 1031 2 0.12 Brazil 2575 2 0.09 Namibia 604 2 0.02 Bulgaria 1312 4 0.07 Nepal 682 2 0.05 Burundi 403 2 0.04 Nicaragua 1433 4 0.11 Cambodia 506 2 0.08 Niger 175 2 0.08 Cameroon 576 2 0.10 Nigeria 1953 1 0.08 Chad 266 2 0.06 Pakistan 532 1 0.02 Chile 2721 3 0.04 Panama 638 2 0.01 Colombia 2619 3 0.06 Paraguay 1077 3 0.06 Croatia 889 3 0.05 Peru 2309 3 0.06 Czech Republic 620 3 0.04 Philippines 1763 2 0.07 Cote ˆ d’Ivoire 580 2 0.08 Poland 1230 4 0.08 Dominican Republic 626 3 0.03 Macedonia 660 3 0.03 Ecuador 1558 4 0.10 Romania 1031 3 0.08 Egypt 2441 2 0.08 Russian Federation 3004 3 0.05 El Salvador 1945 4 0.07 Rwanda 361 2 0.05 Eritrea 117 1 0.00 Senegal 1047 3 0.04 Ethiopia 857 2 0.00 Slovakia 454 3 0.04 Georgia 534 3 0.04 Slovenia 661 3 0.06 Ghana 942 2 0.07 South Africa 1502 2 0.10 Greece 959 2 0.05 Sri Lanka 460 1 0.04 Guatemala 1602 4 0.09 Tajikistan 521 3 0.03 Guinea 275 2 0.22 Tanzania 990 3 0.03 Honduras 1277 4 0.04 Thailand 768 1 0.15 Hungary 830 3 0.09 Togo 224 2 0.09 India 10,322 2 0.04 Turkey 2606 4 0.10 Indonesia 2696 3 0.04 Uganda 1023 2 0.02 Jamaica 305 2 0.05 Ukraine 1368 3 0.07 Jordan 901 2 0.01 Uruguay 1179 3 0.05 Kazakhstan 1054 3 0.01 Uzbekistan 846 4 0.07 Kenya 2047 3 0.13 Venezuela 173 1 0.03 Kyrgyzstan 580 4 0.08 Viet Nam 2138 3 0.04 Lao PDR 642 3 0.01 Yemen 456 2 0.02 Latvia 570 3 0.05 Zambia 1219 3 0.10 Lebanon 493 2 0.19 Zimbabwe 1123 2 0.22 Lesotho 126 1 0.08 This table presents the sample composition with number of observations, years of survey, and the average uncertainty index for each country. Uncertainty is the country-level World Uncertainty Index (WUI) developed by Ahir et al. (2019) using the country reports prepared by Economist Intelligence Unit (EIU). the World Bank. The country-level governance data is from Kaufmann where Obstacle indicates the perceived severity of one of the eight ob- et al. (2009). These data consist of several hundred individual variables stacles facing private firms. Management’s responses range from zero that measure six different governance categories: political stability, (not an obstacle) to four (a very severe obstacle). Uncertainty is the government effectiveness, regulatory quality, law enforcement, and country-level measure of policy uncertainty (WUI). Subscripts i, j, and t corruption as well as the extent to which a country’s citizens are able to indicate firm, country, and time. Since Obstacle is a count variable be- participate in selecting their government. We follow Kaufmann et al. tween zero and four, we run count regressions that assume a Poisson (2009) and Beltratti and Stulz (2012) and create a composite index (by distribution. Assuming a negative binomial distribution or implement- extracting the principal component) of these six variables for each ing a pooled OLS does not alter our qualitative conclusions. country. Following other studies (e.g., Beck et al., 2005), we add the firm ’s size (Size) that is defined as the natural logarithm of total sales and its 3.4. Baseline model age (Age) that is defined as the natural logarithm of one plus the dif- ference between the survey year and the year of incorporation; and our To assess the relationship between the severity of business obstacles control variables are an indicator for firms that export more than 10% of and economic uncertainty, we estimate the following model: their sales (Exporter), indicators for when the government (Government) or a foreign entity (Foreign) own more than 10% of the firm, gross do- Obstacle = α+ β Uncertainty + β Size + β Age + β Exporter i,j,t i,j,t 1 j,t 2 3 i,j,t 4 i,j,t mestic product per capita (GDP) and its growth rate (ΔGDP), and + β Government + β Foreign + β GDP + β ∆GDP + β Inflation i,j,t j,t j,t 5 6 i,j,t 7 8 9 j,t country-level inflation (Inflation ). Additionally, we control for ∑ ∑ ∑ + β Country Governance + Industry+ Year+ Country+ ε j,t i,j,t country-level governance as in Beltratti and Stulz (2012) by using the k t j first principal component of six indicators of governance in the country (1) (Country Governance). Following Hope et al. (2011), we also add 6 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 industry, year, and country fixed effects. The standard errors are the composition of the legislative branches in each country in a given heteroscedasticity-robust and clustered at the industry level. All vari- year from the Database of Political Institutions. We then divide the total ables are defined in Table 13 of the Appendix. seats of the largest opposition party by the total seats in the house to To examine how economic uncertainty affects managerial behavior obtain the proportion of seats held by the largest opposition party. The and corruption, we look at two measures. First, we calculate the fraction higher the proportion of opposition seats, the more likely is the gridlock. of sales not reported to the tax authorities (Cheating on taxes). The survey To formally test the quality of our instruments, we apply Stock and specifically asks the interviewee what percentage of sales were reported Yogo’s (2005) weak instrument test. We also report the F-statistics from for taxation. Although the variable is not available for all countries, it is the test of under-identification (instruments are irrelevant) and follow well populated for vast majority of the countries (68 out of 93). This is a the instrument assessment guidelines proposed in Staiger and Stock direct measure of corruption as firms are required to report all sales for (1997). Formally, our instrumental variable (IV) regression is summa- tax purposes. Our second measure, Bribes, is the fraction of sales paid as rized as: “unofficial payments” to the government authorities. This measure Uncertainty = δ + δ Instruments for Corrupt + δ Controls + μ 0 1 2 i,j,t j,t i,j,t i,j,t captures corrupt behavior on the part of private firms as well as the (3a) government authorities. Following these definitions, we modify Eq. (1) as follows: Corrupt = γ + γ Uncertainty + γ Controls + ω (3b) i,j,t 0 1 i,j,t 2 i,j,t i,j,t Corrupt = α+ β Uncertainty + β Size + β Age + β Exporter i,j,t i,j,t 1 j,t 2 3 i,j,t 4 i,j,t where the Instrument for Corrupt vector consists of Election closeness and + β Government + β Foreign + β GDP + β ∆GDP + β Inflation i,j,t j,t j,t 5 6 i,j,t 7 8 9 j,t ∑ ∑ ∑ Opposition seats, and the Controls vector consists of all the control vari- + β Governance + Industry+ Year+ Country+ ε (2) 10 j,t i,j,t ables from Eq. (2). All the other specifications in Eq. (3b) are as in Eq. k t j (2). Corrupt is either Cheating on taxes or Bribes and other specifications are as in Eq. (1). 3.5. Summary statistics Some studies argue that endogeneity problem (in the form of simultaneity bias) may arise when an economic outcome is regressed on Panel A of Table 2 provides the summary statistics for our pooled a news-based measure of policy uncertainty (see the discussion in Gulen public and private sample. The first heading of the panel displays the and Ion, 2016). In light of this concern, we implement an instrumental opinion of business managers on the severity of the obstacles they face. variable (IV) regression in which we use the closeness of national elec- The responses range from zero (not an obstacle) to four (very severe tions (Election closeness) and the proportion of seats in the national obstacle). On average, managers identify obstacles as less severe when parliament held by the largest opposition party (Opposition seats) as in- the average response for all obstacles is below two (moderate obstacle). struments for Uncertainty in the first stage. The average and median for tax rate indicates that it is a more severe For our first instrument, Election closeness, we rely on the findings concern for firms than other obstacles. The summary statistics for size in Julio and Yook (2012), Boutchkova et al. (2012), and Gungor- indicate that the average private firm in the sample has sales of $10.67 aydinoglu et al. (2017) who claim that when a country’s national elec- million; however, the median value ($149.03 thousand) indicates pos- tions are very close, the policy uncertainty spikes. Thus, this instrument itive sample skewness. The sample also has some micro firms with sales should be a reliable as our Uncertainty variable. We obtain data on all in the 25th percentile at approximately $558,000. The mean (median) national elections from the Database of Political Institutions (Scartascini age is approximately 19 (14) years. Around 16.1% of the firms are ex- et al., 2018), and then we supplement it with a manual data verification porters while only 2.2% have over 10% government ownership. Simi- procedure that closely follows the guidelines in Gungoraydinoglu et al. larly, around 10.5% of the firms in the sample have some foreign (2017). Our Election closeness variable uses the distribution of party ownership. Focusing on country characteristics; the mean (median) votes in the national elections to capture the uncertainty created due to uncertainty index is 0.067 (0.057) that roughly translates to about 6.7% competitive and contested elections. Jens (2017) argues that election (5.7%) word counts for “uncertainty” (or its variants) for each 1000 closeness can be attributed to poor economic performance by the pre- words in the quarterly EIU reports. The mean (median) GDP is around vious administration and that it is indicative of higher economic un- $5023.1 ($2979.0) that indicates considerable variation in the economy certainty in the pre-election period. As such, this variable captures the of the sample countries. The mean and median GDP growth are around uncertainty leading up to the election date. 5%. The mean (median) inflation in the sample countries is 7.1% (5.3%). Our second instrument (Opposition seats), on the other hand, captures Average governance in the countries is on the lower side which is ex- the post-election uncertainty prevalent in a country. Our motivation for pected given that our sample is mostly comprised of underdeveloped or its use comes from the divided government hypothesis (Cutler, 1988; developing countries with weaker governance mechanisms. Kelly, 1993; Sundquist, 1988) which argues that the legislation is less Further, the norm-deviant behavior of the firms is presented sepa- likely to be enacted if the executive branch’s party does not also hold the rately for the pooled sample (bottom of Panel A), for the private firms ’ majority in the legislative branches of the government. Since divided sample (Panel B), and for the public firms ’ sample (Panel C). At least governments are likely to give rise to disagreements between the legis- 25% of firms who respond to the tax question engage in cheating (75th lature and the executive branches, we predict that a larger proportion of percentile value is non-zero) and, on average, they underreport 17.9% of opposition seats in the legislative branches will be positively related their sales to the authorities. At the mean annual sales level of $10.67 with policy uncertainty. Arguably, such political gridlock can affect our million, this number corresponds to an average of $2.34 million unre- dependent variable, Corrupt, only by increasing the uncertainty (Un- ported sales per firm. On average, the firms pay 1.4% of their sales as certainty) which means that the exclusivity criterion for this instrument bribes to government officials that corresponds to an average illegal should also be satisfied. To calculate Opposition seats, we obtain data on payment of about $149,352 per firm per year. 4. Empirical results Note that using several instruments for one endogenous variable is an 4.1. Univariate results acceptable practice in instrumental variable regressions, because it creates over-identified conditions. Even if one of the instruments is deemed irrelevant, Table 3 displays the univariate tests for private and public firms (in the remaining instruments are sufficient for proper identification (please refer to the discussions in section 5.1.2 of Wooldridge, 2002). Panels A and B, respectively). We define high (low) economic 7 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 2 Descriptive statistics. Countries Obs. Mean Std. dev. P25 Median P75 Panel A: Pooled summary statistics Business obstacles Dysfunctional court system 93 103,738 0.921 1.222 0.000 0.000 2.000 Crime, theft, disorder 93 103,738 1.074 1.276 0.000 1.000 2.000 Prevalent corruption 93 103,738 1.566 1.486 0.000 1.000 3.000 High tax rates 93 103,738 1.655 1.369 0.000 2.000 3.000 Ineffective tax administration 93 103,738 1.320 1.292 0.000 1.000 2.000 Anti-competitive practices 93 103,738 1.372 1.368 0.000 1.000 2.000 Reduced access to finance 93 103,738 1.376 1.354 0.000 1.000 2.000 Political instability 93 103,738 1.462 1.449 0.000 1.000 3.000 Enterprise characteristics Size (thousands of USD) 93 103,738 10,668.250 58,333.986 5.579 149.031 1319.697 Age 93 103,738 19.461 17.429 8.000 14.000 24.000 Exporter 93 103,738 0.161 0.368 0.000 0.000 0.000 Government 93 103,738 0.022 0.146 0.000 0.000 0.000 Foreign 93 103,738 0.105 0.307 0.000 0.000 0.000 Country characteristics Uncertainty 93 103,738 0.067 0.055 0.030 0.057 0.088 GDP 93 103,738 5023.057 4785.750 1369.184 2979.005 8573.708 ΔGDP 93 103,738 4.969 3.613 3.050 5.118 7.410 Inflation 93 103,738 7.128 8.060 3.332 5.349 8.778 Governance 93 103,738 -0.754 1.362 -1.682 -0.711 -0.253 Civic norms 59 82,498 0.243 0.112 0.180 0.220 0.282 Religion 90 102,893 0.774 0.419 1.000 1.000 1.000 Norm-deviant behavior Cheating on taxes 68 30,032 0.219 0.311 0.000 0.000 0.400 Bribes 93 72,704 0.014 0.057 0.000 0.000 0.000 Panel B: Summary statistics for private firms Size (thousands of USD) 93 98,865 9970.044 5,6691.797 5.020 140.234 1197.605 Cheating on taxes 68 29,024 0.221 0.311 0.000 0.000 0.400 Bribes 93 69,525 0.014 0.056 0.000 0.000 0.000 Panel C: Summary statistics for public firms Size (thousands of USD) 93 4873 24,833.675 83,821.550 38.687 797.575 7619.794 Cheating on taxes 45 1008 0.163 0.288 0.000 0.000 0.200 Bribes 88 3179 0.018 0.078 0.000 0.000 0.000 This table presents the descriptive statistics for the variables used in this study. Panel A presents the results for our main sample (pooled private and public firms). Panel B (Panel C) presents the summary statistics from the private (public) firms ’ samples; included are a few statistics for only a few key variables that are referred in the text. Uncertainty is the World Uncertainty Index (WUI) developed by Ahir et al. (2019) using the quarterly reports prepared by Economist Intelligence Unit (EIU) about each country. All variables pertaining to business obstacles indicate the severity of an obstacle and range between zero and four, where zero indicates no obstacle and four indicates very severe obstacle. All other variables are defined in Appendix in Table 13. uncertainty as countries belonging to the top (bottom) quintiles of cheating through taxes during high uncertainty periods (7% versus economic uncertainty in a given year. Panel A indicates that private 17.3% of the sales go underreported). However, these public firms firms ’ managers identify all the obstacles as more severe when economic substantially increase the amount of bribes they pay (3.9% versus 1% of uncertainty is higher. Public firms (Panel B) also report a worsening the sales). These results are broadly consistent with our first three hy- business environment due to uncertainty except for tax rates. Thus, both potheses: while both private and public firms accelerate their corrupt private and public firms perceive uncertainty as the cause of worsening behavior when faced with uncertainty, public firms focus more on business obstacles. bribery than cheating on taxes. However, in both panels there are considerable firm- and country- level differences between firms in the low versus high uncertainty sub- 4.2. Policy uncertainty and obstacles facing private firms samples. Therefore, we cannot reach a conclusion based on these results that economic uncertainty really increases business obstacles. It is To corroborate the findings from our univariate analysis, we conduct possible that the firms just perceive it as such. Nevertheless, as the various multivariate tests. Using our pooled sample of 103,738 firm- variable under the norm-deviant headings show, firms cheat more on years, we estimate Eq. (1) to examine the effects of economic uncer- taxes and pay more bribes during high economic uncertainty. Regardless tainty on the obstacles facing all firms. After controlling for both firm of whether the increase in obstacles is real or perceived, both the private and country-level variables, our univariate results persist. The results and the public firms react with increased corruption. Specifically, during reported in Panels A and B of Table 4 show that the coefficient for the the high uncertainty periods, the incremental underreporting of sales by main uncertainty measure is positive and statistically significant in all private firms is about 2.6% of total sales that if evaluated at the mean specifications. All the coefficients from the Poisson regression are sta- sales of $9.97 million (from Table 2), indicates that roughly $259,220 tistically significant at less than the 1% level. The estimates are also additional sales went underreported to the government authorities on economically significant. In Panel B, a one standard deviation increase top of the mean underreported sales of around $2.20 million (=22.1% * in economic uncertainty, ceteris paribus, results in a 9.6% $9.97; from Table 2). Similarly, during high uncertainty periods, the (1.626 × 0.055 ÷ 1.074) increase in the severity of crime, theft, and private firms pay 0.4% more in bribes than they do during low uncer- disorder and a 4.1% (0.545 × 0.055 ÷ 1.372) increase in the severity of tainty periods (1.8% vs. 1.4%). For most small private firms from around anti-competitive practices. These results are in line with our prediction the world, these numbers are not economically negligible. For public in Hypothesis 1 – economic uncertainty worsens managers’ perception firms, on the other hand, these univariate results are quite different. As about the severity of the business obstacles prevalent in a country. Panel B of Table 3 shows, public firms actually reduce the amount of Further, the reported severities of the obstacles increase as the size of 8 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 3 Univariate results. Countries High Low Difference t-statistic Uncertainty Uncertainty Panel A: Univariate results for private firms Business obstacles Dysfunctional court system 93 0.954 0.786 0.168*** 13.73 Crime, theft, disorder 93 1.269 0.980 0.289*** 22.13 Prevalent corruption 93 1.651 1.412 0.239*** 15.93 High tax rates 93 1.540 1.599 -0.059*** -4.25 Ineffective tax administration 93 1.278 1.222 0.056*** 4.33 Anti-competitive practices 93 1.472 1.293 0.179*** 12.94 Reduced access to finance 93 1.328 1.376 -0.048*** -3.48 Political instability 93 1.649 1.198 0.451*** 31.07 Enterprise characteristics Size (thousands of USD) 93 7363.621 18,865.655 -11502.034*** -15.89 Age 93 20.344 17.805 2.539*** 14.67 Exporter 93 0.166 0.129 0.037*** 10.31 Government 93 0.016 0.011 0.005*** 3.99 Foreign 93 0.117 0.099 0.018*** 5.90 Country characteristics Uncertainty 93 0.153 0.019 0.135*** 323.19 GDP 93 5018.214 4710.491 307.723*** 6.35 ΔGDP 93 4.280 5.668 -1.389*** -41.56 Inflation 93 8.052 7.469 0.583*** 6.14 Governance 93 -1.001 -0.801 -0.200*** -13.82 Civic norms 59 0.199 0.254 -0.055*** -46.12 Religion 90 0.881 0.770 0.111*** 28.48 Norm-deviant behavior Cheating on taxes 68 0.251 0.225 0.026*** 4.55 Bribes 93 0.018 0.014 0.004*** 5.80 Panel B: Univariate results for public firms Business obstacles Dysfunctional court system 91 1.256 1.030 0.226*** 3.88 Crime, theft, disorder 91 1.362 1.128 0.234*** 4.01 Prevalent corruption 91 1.814 1.558 0.256*** 3.84 High tax rates 91 1.715 1.742 -0.026 -0.44 Ineffective tax administration 91 1.505 1.256 0.250*** 4.46 Anti-competitive practices 91 1.597 1.328 0.269*** 4.34 Reduced access to finance 91 1.444 1.312 0.132** 2.22 Political instability 91 1.820 1.411 0.409*** 6.49 Enterprise characteristics Size (thousands of USD) 91 19,744.426 35,164.440 -15420.015*** -3.51 Age 91 29.452 27.095 2.357** 2.18 Exporter 91 0.272 0.186 0.085*** 4.51 Government 91 0.238 0.190 0.048*** 2.58 Foreign 91 0.256 0.176 0.079*** 4.27 Country characteristics Uncertainty 91 0.142 0.018 0.124*** 68.91 GDP 91 4377.387 5494.168 -1116.782*** -5.72 ΔGDP 91 3.755 5.393 -1.638*** -10.50 Inflation 91 6.286 6.442 -0.156 -0.42 Governance 91 -1.171 -0.918 -0.253*** -3.95 Civic norms 59 0.191 0.240 -0.049*** -15.73 Religion 91 0.839 0.615 0.223*** 11.06 Norm-deviant behavior Cheating on taxes 45 0.070 0.173 -0.103*** -3.91 Bribes 88 0.039 0.010 0.029*** 5.73 This table presents the univariate analysis based on high and low economic uncertainty in the country. Results for private and public firms are presented separately in Panels A and B, respectively. Uncertainty is the World Uncertainty Index (WUI) developed by Ahir et al. (2019), where high (low) indicates top (bottom) quintile of uncertainty in a given year. All variables pertaining to business obstacles indicate the severity of an obstacle and range between zero and four, where zero indicates no obstacle and four indicates very severe obstacle. All other variables are defined in Appendix in Table 13. the firm increases, except for access to financing. These two results are 4.3. Policy uncertainty and corruption in firms understandable since firms are more likely to encounter challenges as the size of their business grows and are more likely to face financing We next estimate Eq. (2) with our two main dependent variables to issues if they are young and small. Beck et al. (2005) also find this size capture corruption by using a pooled sample of private and public firms. effect in relationship to financing. Among other variables, foreign Panels A and B in Table 5 provide these results. The regressions in panel ownership reduces many of the constraints. This reduction can be A indicate that during high economic uncertainty, all firms cheat more attributed to the potentially better management and access to resources on taxes and pay more bribes. Both of these results are statistically that foreign ownership brings. The coefficient for GDP is consistently significant at the 1% level. The results are also economically significant. negative and highly significant in all eight regression results. This co- For instance, a one standard deviation increase in economic uncertainty efficient means that firms located in low-income countries face more results in about 3.01% (0.548 × 0.055) less sales reported to the gov- severe obstacles than those in middle or high-income countries. ernment authorities for tax purposes. Similarly, a one standard deviation 9 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 4 Economic uncertainty and managerial perception. Panel A: Poisson regression results Dep.variable Managerial perception of severity of business obstacles Obstacle Dysfunctional Crime, theft, Prevalent High tax rates Ineffective tax Anti-competitive Reduced access to Political court system disorder corruption admin. practices finance instability (1) (2) (3) (4) (5) (6) (7) (8) Uncertainty 0.638*** 1.704*** 0.826*** 0.579*** 0.535*** 0.566*** 0.589*** 0.727*** (4.26) (12.99) (9.60) (4.34) (3.65) (5.52) (6.28) (5.92) Size 0.039*** 0.027*** 0.026*** 0.028*** 0.028*** 0.004** -0.002 0.024*** (11.94) (6.35) (11.55) (12.26) (14.60) (2.02) ( 0.75) (11.23) Age 0.035*** -0.007 0.001 0.003 0.004 0.023*** -0.052*** 0.013*** (5.58) ( 1.20) (0.18) (0.67) (0.70) (3.99) ( 8.79) (2.82) Exporter 0.064*** -0.066*** -0.013 -0.032*** 0.034*** -0.142*** -0.049*** 0.007 (4.70) ( 3.60) ( 0.98) ( 3.25) (3.32) ( 7.10) ( 2.66) (0.61) Government -0.086* -0.094*** -0.140*** -0.158*** -0.180*** -0.145*** 0.015 -0.057** ( 1.82) ( 3.68) ( 3.06) ( 5.72) ( 5.08) ( 3.53) (0.61) ( 2.35) Foreign 0.003 -0.056*** -0.038*** -0.078*** -0.039*** -0.139*** -0.238*** -0.009 (0.35) ( 3.76) ( 4.61) ( 11.51) ( 3.49) ( 13.38) ( 17.70) ( 1.08) GDP -1.477*** -1.074*** -1.176*** -1.361*** -1.551*** -0.238* -0.760*** -0.997*** ( 9.74) ( 7.33) ( 12.70) ( 10.48) ( 12.41) ( 1.95) ( 3.98) ( 8.03) ΔGDP 0.008* 0.004 0.007** 0.009* 0.011*** 0.006** 0.003 -0.005** (1.95) (0.79) (2.33) (1.84) (2.76) (2.05) (0.92) ( 2.14) Inflation 0.002 -0.000 0.004* 0.004** 0.007*** -0.000 0.001 0.005*** (0.91) ( 0.20) (1.80) (2.26) (5.08) ( 0.29) (0.53) (3.09) Country -0.057 0.079* -0.002 0.151*** 0.117*** 0.151*** 0.182*** -0.075 Governance ( 1.12) (1.80) ( 0.05) (5.65) (4.83) (4.02) (4.01) ( 1.48) Country fixed Yes Yes Yes Yes Yes Yes Yes Yes effects Industry fixed Yes Yes Yes Yes Yes Yes Yes Yes effects Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Pseudo R 0.128 0.119 0.123 0.105 0.104 0.090 0.084 0.160 Observations 103,738 103,738 103,738 103,738 103,738 103,738 103,738 103,738 Panel B: Pooled OLS regression results Uncertainty 0.646*** 1.626*** 0.627** 0.451** 0.328** 0.545*** 0.373* 0.836*** (3.55) (12.83) (2.67) (2.42) (2.25) (3.36) (1.87) (3.83) Country fixed Yes Yes Yes Yes Yes Yes Yes Yes effects Industry fixed Yes Yes Yes Yes Yes Yes Yes Yes effects Year fixed Yes Yes Yes Yes Yes Yes Yes Yes effects Adjusted R 0.199 0.228 0.259 0.267 0.224 0.195 0.189 0.348 Observations 103,738 103,738 103,738 103,738 103,738 103,738 103,738 103,738 This table presents the results from regressing the severity of obstacles faced by all enterprises (private and public) on the level of economic uncertainty in the country and other explanatory variables. Panel A presents Poisson regression results while Panel B displays OLS regression results. Uncertainty is the World Uncertainty Index (WUI) developed by Ahir et al. (2019). All variables pertaining to business obstacles indicate the interviewee’s opinion about the severity of an obstacle and range between zero and four, where zero indicates no obstacle and four indicates very severe obstacle. Country, industry, and year fixed effects are included. All other variables are defined in Appendix in Table 13. The t-statistics based on industry clustered robust standard errors are shown in parentheses. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively. increase in economic uncertainty translates to about 0.17% with the idea that younger and smaller firms are riskier. Country-level (0.031 × 0.055) increase in bribe payments as a percentage of sales. inflation also induces cheating on taxes and paying more bribes. While Overall, these findings are consistent with Hypothesis 2 in that the small governance can mitigate bribery problem, it contributes to cheating on firms behave more corruptly when there is higher economic taxes. One explanation for this cheating could be the higher tax rate in uncertainty. these countries that could induce an environment where firms are more To enhance the robustness of our results with regards to standard likely to avoid taxes (Kanagaretnam et al., 2018). error clustering, in panel B of Table 5 we apply both country-industry Columns (3) and (4) of Table 5 provide the results from Tobit re- (interacted) fixed effects and cluster the standard errors at the same gressions. Since a majority of the firms do not cheat on taxes or pay level as suggested by Petersen (2009). We continue to find a positive bribes (median values are zero; see Table 2), a Tobit regression with left relationship between firms ’ tendency to engage in corrupt behavior and censoring at zero is an appropriate estimation method in our context. economic uncertainty. The Tobit results are similar to those reported using pooled OLSs. Among the control variables, size, age, GDP, and GDP growth Columns (5) and (6) display the results from propensity score negatively affect corrupt behavior in private firms. This is consistent matching. The treated firms are the ones exposed to high uncertainty. We define high (low) economic uncertainty as countries in the top (bottom) quintile of economic uncertainty in a given year. We imple- ment the matching by first matching treated and control firms based on We report heteroscedasticity-robust standard errors clustered at the in- all characteristics in column (1) and by applying a caliper of 0.05. This dustry level in all our specifications. Following Lederman (2010) and boot- process results in an equal sample of firms belonging to high (treated strapping the standard errors does not alter our results. 10 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 5 Economic uncertainty and corruption. Panel A: Industry, country, and year fixed effects with industry clustered standard errors. Estimation model Pooled OLS Tobit with left censoring Propensity score-matching Dependent variable Cheating on taxes Bribes Cheating on taxes Bribes Cheating on taxes Bribes (1) (2) (3) (4) (5) (6) Uncertainty 0.548*** 0.031*** 0.747*** 0.121*** 7.992** 0.127*** (4.84) (5.62) (4.68) (3.93) (2.33) (2.96) Size -0.012*** -0.001*** -0.026*** -0.003*** -0.037*** -0.002*** ( 7.98) ( 8.41) ( 8.64) ( 4.49) ( 11.87) ( 2.88) Age -0.011*** -0.001 -0.020*** -0.001 -0.001 -0.000 ( 4.65) ( 1.51) ( 4.25) ( 1.27) ( 0.09) ( 0.13) Exporter -0.005 0.001 -0.020* 0.009*** -0.012 0.012*** ( 0.83) (1.29) ( 1.87) (2.69) ( 0.72) (3.13) Government -0.020** 0.005** -0.146*** -0.000 -0.150*** -0.005 ( 2.17) (2.24) ( 5.57) ( 0.03) ( 3.15) ( 0.51) Foreign -0.022*** 0.001 -0.069*** -0.002 -0.072*** 0.001 ( 4.68) (1.04) ( 6.22) ( 0.57) ( 3.26) (0.15) GDP -0.354 -0.009 -0.401 -0.005 -19.148** -0.017 ( 0.64) ( 1.49) ( 0.45) ( 0.14) ( 2.42) ( 0.69) ΔGDP -0.029** -0.001*** -0.055** -0.006*** 0.047** -0.002 ( 2.13) ( 6.56) ( 2.31) ( 7.96) (2.31) ( 1.42) Inflation 0.007 0.000* 0.018*** 0.001* -0.174* 0.001* (1.41) (1.92) (3.18) (1.68) ( 1.69) (1.88) Country Governance 0.265*** -0.001 0.317*** -0.011 4.575** 0.032*** (5.46) ( 0.56) (4.52) ( 1.52) (2.04) (2.72) Country fixed effects Yes Yes Yes Yes Yes Yes Industry fixed effects Yes Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Yes Adjusted/Pseudo R 0.209 0.070 0.150 0.393 0.171 0.440 Observations 30,032 72,704 30,032 72,704 10,704 25,922 Panel B: Industry-country and year fixed effects with industry-country clustered standard errors. Uncertainty 0.685*** 0.037** 0.981*** 0.126** 4.869*** 0.147** (3.75) (2.50) (3.30) (2.07) (3.66) (2.22) Size -0.013*** -0.001*** -0.029*** -0.003*** -0.017*** -0.002*** ( 10.10) ( 6.38) ( 11.19) ( 4.26) ( 7.07) ( 2.69) Age -0.009*** -0.000 -0.016*** -0.001 -0.003 0.000 ( 2.65) ( 1.43) ( 2.66) ( 0.95) ( 0.49) (0.09) Exporter -0.003 0.002** -0.019 0.011*** -0.001 0.014*** ( 0.60) (2.18) ( 1.56) (3.51) ( 0.10) (3.04) Government -0.011 0.005** -0.130*** -0.001 -0.020 0.000 ( 1.24) (2.09) ( 4.90) ( 0.16) ( 1.31) (0.04) Foreign -0.015** 0.001 -0.056*** -0.001 -0.023** 0.001 ( 2.40) (1.43) ( 4.02) ( 0.15) ( 2.05) (0.27) GDP 1.526* -0.014 2.092* -0.016 -15.210*** -0.044 (1.70) ( 1.15) (1.69) ( 0.24) ( 4.18) ( 0.57) ΔGDP -0.034* -0.002*** -0.058* -0.009*** 0.038*** -0.002 ( 1.83) ( 4.00) ( 1.84) ( 5.65) (2.60) ( 0.83) Inflation 0.011 0.000* 0.025 0.001** -0.132*** -0.002 (1.08) (1.72) (1.54) (2.03) ( 2.92) ( 1.34) Country Governance 0.486*** -0.000 0.658*** -0.008 2.575*** -0.016 (3.68) ( 0.06) (3.40) ( 0.55) (3.10) ( 0.62) Industry-country Yes Yes Yes Yes Yes Yes fixed effects Year fixed effects Yes Yes Yes Yes Yes Yes Adjusted/Pseudo R 0.224 0.086 0.180 0.476 0.228 0.527 Observations 30,032 72,704 30,032 72,704 10,704 25,922 This table presents the results from regressing the enterprise-level norm-deviant behavior on the level of economic uncertainty in the country and on other explanatory variables. Private and public firms are pooled together. Columns (1) and (2) present pooled OLS regression results. Columns (3) and (4) present Tobit regression results with left censoring at 0 (many of our private firms do not cheat on taxes or pay bribes). Columns (5) and (6) present propensity score-matching regression results using a symmetrical sample constituted of only the treated firms and their matches; the matching is one-to-one and it is based on all control variables and a caliper of 0.05. Uncertainty is the World Uncertainty Index (WUI) developed by Ahir et al. (2019). Cheating on taxes is the fraction of sales not reported to the tax authorities. Bribes is the fraction of sales paid in informal payments to the government authorities. Country, industry, and year fixed effects are included. All other variables are defined in Appendix in Table 13. In Panel A the t-statistics (shown in parentheses) are based on industry clustered robust standard errors and in Panel B they are based on industry-country clustered robust standard errors. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively. sample) and low (control sample) uncertainty countries. Using a pooled consistent with our baseline results and show that even after matching sample of treated and control observations, we run our OLS tests. The firms on several firm and country characteristics, the uncertainty- coefficient for cheating on taxes and bribes is again positive and statis- corruption relationship prevails. tically significant at the 5% level. Taken together, these findings are 11 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 6 Economic uncertainty and managerial perception: private versus public firms. Dependent variable Managerial perception of severity of business obstacles Obstacle Dysfunctional court Crime, theft, Prevalent High tax Ineffective tax Anti-competitive Reduced access to Political system disorder corruption rates admin. practices finance instability (1) (2) (3) (4) (5) (6) (7) (8) Uncertainty 0.598* 1.718*** 0.493 0.518 0.468 0.511 1.218*** 0.441 (1.88) (5.72) (1.62) (1.57) (1.51) (1.64) (4.85) (1.50) Private -0.030 -0.020 0.027 0.050* 0.037 0.016 0.076*** 0.026 ( 1.15) ( 0.84) (0.93) (1.84) (1.61) (0.66) (2.58) (1.05) Uncertainty × 0.038 -0.018 0.363 0.074 0.079 0.061 -0.660*** 0.312 Private (0.13) (¡0.07) (1.22) (0.25) (0.31) (0.21) (¡2.58) (1.26) Size 0.039*** 0.027*** 0.027*** 0.029*** 0.028*** 0.005** -0.002 0.025*** (11.77) (6.19) (11.63) (12.56) (14.67) (2.07) ( 0.69) (11.24) Age 0.035*** -0.007 0.002 0.005 0.005 0.024*** -0.051*** 0.014*** (5.57) ( 1.31) (0.40) (0.97) (0.87) (4.06) ( 8.73) (3.24) Exporter 0.064*** -0.066*** -0.013 -0.032*** 0.035*** -0.142*** -0.049*** 0.007 (4.66) ( 3.61) ( 0.94) ( 3.18) (3.35) ( 7.11) ( 2.66) (0.66) Government -0.096* -0.101*** -0.124*** -0.139*** -0.165*** -0.139*** 0.026 -0.040* ( 1.95) ( 3.93) ( 2.76) ( 5.19) ( 4.77) ( 3.38) (1.01) ( 1.79) Foreign 0.002 -0.056*** -0.036*** -0.076*** -0.037*** -0.138*** -0.238*** -0.007 (0.25) ( 3.84) ( 4.41) ( 11.10) ( 3.36) ( 13.16) ( 17.42) ( 0.83) GDP -1.473*** -1.071*** -1.184*** -1.366*** -1.556*** -0.240** -0.759*** -1.004*** ( 9.74) ( 7.35) ( 12.85) ( 10.57) ( 12.52) ( 1.96) ( 3.97) ( 8.17) ΔGDP 0.008* 0.004 0.007** 0.009* 0.011*** 0.006** 0.003 -0.005** (1.94) (0.78) (2.34) (1.85) (2.77) (2.06) (0.94) ( 2.11) Inflation 0.002 -0.000 0.004* 0.004** 0.007*** -0.000 0.001 0.005*** (0.91) ( 0.18) (1.78) (2.26) (5.07) ( 0.31) (0.54) (3.03) Country Governance -0.059 0.077* 0.001 0.154*** 0.119*** 0.152*** 0.182*** -0.072 ( 1.15) (1.78) (0.02) (5.80) (4.95) (4.04) (4.06) ( 1.40) Country fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Industry fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Pseudo R 0.128 0.119 0.123 0.105 0.104 0.090 0.084 0.160 Observations 103,738 103,738 103,738 103,738 103,738 103,738 103,738 103,738 This table analyses whether the severity of obstacles faced by the enterprises is perceived differently by private and public firms. Presented are the Poisson regression results. Uncertainty is the World Uncertainty Index (WUI) developed by Ahir et al. (2019). Private indicates that the firm is a small private enterprise. All variables pertaining to business obstacles indicate the interviewee’s opinion about the severity of an obstacle and range between zero and four, where zero indicates no obstacle and four indicates very severe obstacle. Country, industry, and year fixed effects are included. All other variables are defined in Appendix in Table 13. The t-statistics based on industry clustered robust standard errors are shown in parentheses. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively. 4.4. Private versus public firms: preference for different forms of Table 7 shows an important behavioral difference between private and corruption public firms. Because they are more closely monitored, public firms tend to cheat less on taxes than private firms (the coefficient for Uncertain- Next, we make an important distinction between small firms based ty×Private is significantly positive). Put differently, the small private on whether they are subject to monitoring by institutional shareholders firms find it easier to underreport sales for tax purposes than public ones, and other stockholders or whether they are tightly held with limited who are subject to more regulations and more scrutiny from govern- dissemination of information. Small private firms are subject to very mental authorities (Hope et al., 2011). However, the public firms pay a little monitoring. Thus, as expressed in Hypothesis 3, they are likely to higher percentage of their sales in bribes than the private firms. In engage in a different form of corruption than the more closely watched untabulated results, we find that private firms do indeed increase their and larger public firms. bribes when subjected to economic uncertainty; however, the negative First, we check whether private firms feel differently about business sign of the interaction term in Table 7 (see columns (2), (4), and (6)) obstacles than public firms. We run the regression described in Eq. (1) indicates that private firms cannot increase the magnitude of their bribes but instead use an interaction term for private firms (Private dummy). As as much as public firms. Either private firms run out of cash, as indicated displayed in Table 6, the interaction term, Uncertainty×Private, is sta- by the result in column (7) of Table 6, or they find it easier to lie about tistically insignificant which indicates there are no major differences their sales and resort to cheating on taxes as a way to cope with eco- among the managers of public and private firms on whether they nomic hardship. This contrast in preferred forms of corruption among perceive economic uncertainty as an amplifier of business obstacles in private and public firms that are subject to economic uncertainty is a the economy. A possible exception is their perception of financing novel finding for both the “greasing the wheels” (Huntington, 1968; Leff, constraints (column (7)). The managers of private firms feel that the 1964) hypothesis and the policy uncertainty literature (see, among financing constraints are much stronger for them which, in light of the others, Bloom, 2009). findings by Beck et al. (2005) and Hope et al. (2011), is to be expected given that private firms are smaller in size (see Table 2) and the un- certainty likely affects their sources of financing more severely. Second, we analyse whether there are differences between private We basically rerun the results from Table 5 but using only the sample of and public firms with regards to the uncertainty-corruption relationship. private firms. 12 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 7 Economic uncertainty and corruption: private versus public firms. Estimation model Pooled OLS Tobit with left censoring Propensity score-matching Dependent variable Cheating on taxes Bribes Cheating on taxes Bribes Cheating on taxes Bribes (1) (2) (3) (4) (5) (6) Uncertainty -0.155 0.086*** -0.374 0.368*** 6.947* 0.376*** ( 0.44) (2.91) ( 0.47) (5.23) (1.94) (4.54) Private -0.001 -0.001 0.023 0.012 0.038 0.016 ( 0.06) ( 0.38) (0.51) (1.36) (0.54) (1.52) ** Uncertainty × Private 0.735 -0.058* 1.195* -0.262*** 1.097 -0.265*** (2.52) (¡1.92) (1.70) (¡3.14) (1.32) (¡3.20) Size -0.012*** -0.001*** -0.026*** -0.003*** -0.037*** -0.002*** ( 7.62) ( 8.48) ( 8.34) ( 4.58) ( 11.49) ( 2.93) Age -0.010*** -0.001* -0.019*** -0.001 -0.000 -0.000 ( 4.55) ( 1.79) ( 4.08) ( 1.39) ( 0.01) ( 0.14) Exporter -0.005 0.001 -0.020* 0.009*** -0.013 0.012*** ( 0.84) (1.24) ( 1.89) (2.66) ( 0.74) (3.12) Government -0.002 0.003 -0.106*** -0.002 -0.083* -0.006 ( 0.25) (1.39) ( 4.87) ( 0.27) ( 1.93) ( 0.60) Foreign -0.022*** 0.001 -0.069*** -0.002 -0.073*** 0.000 ( 4.65) (0.89) ( 6.28) ( 0.68) ( 3.37) (0.06) GDP -0.468 -0.009 -0.598 -0.003 -19.222** -0.015 ( 0.87) ( 1.41) ( 0.68) ( 0.10) ( 2.40) ( 0.62) ΔGDP -0.027* -0.001*** -0.051** -0.006*** 0.045** -0.002 ( 1.98) ( 6.63) ( 2.10) ( 8.07) (2.21) ( 1.39) Inflation 0.009* 0.000* 0.022*** 0.001* -0.175* 0.001* (1.84) (1.99) (3.88) (1.70) ( 1.68) (1.82) Country Governance 0.284*** -0.001 0.365*** -0.011 4.618** 0.033*** (5.92) ( 0.64) (5.28) ( 1.57) (2.04) (2.87) Country fixed effects Yes Yes Yes Yes Yes Yes Industry fixed effects Yes Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Yes Adjusted/Pseudo R 0.210 0.070 0.150 0.394 0.172 0.441 Observations 30,032 72,704 30,032 72,704 10,704 25,922 This table analyses whether corrupt behavior by the enterprises is different for private and public firms. Columns (1) and (2) present pooled OLS regression results. Columns (3) and (4) present Tobit regression results with left censoring at 0 (many of our private firms do not cheat on taxes or pay bribes). Columns (5) and (6) present propensity score-matching regression results using a symmetrical sample constituted of only the treated firms and their matches; the matching is one-to-one and it is based on all control variables and a caliper of 0.05. Uncertainty is the World Uncertainty Index (WUI) developed by Ahir et al. (2019). Private indicates that the firm is a small private enterprise. Cheating on taxes is the fraction of sales not reported to the tax authorities. Bribes is the fraction of sales paid in informal payments to the government authorities. Country, industry, and year fixed effects are included. All other variables are defined in Appendix in Table 13. The t-statistics based on in- dustry clustered robust standard errors are shown in parentheses. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively. 4.5. Identification through instrumental variable regression 4.6. Mediation analysis: measuring the direct and indirect effects of policy uncertainty To improve the identification of our tests, we turn to an instrumental variable regression, whereby we represent uncertainty with the two Thus far, we show a positive association between rising economic instruments described in subsection 3.4. We predict a positive rela- uncertainty and the worsening of managerial perception regarding busi- tionship between these instruments and economic uncertainty. The re- ness obstacles. We also show that high economic uncertainty induces sults from the IV regressions in Eqs. (3a) and (3b) are displayed in corrupt behavior. However, a worsened managerial perception may also Table 8. As in Table 5, the sample has both private and public firms. affect managerial behavior towards cheating on taxes and paying bribes. Columns (1) and (3) show that both the proportion of opposition seats in Therefore, to further corroborate the association between economic un- the legislature and election closeness are indeed significantly and posi- certainty, managerial perception, and corruption, we perform a media- tively associated with economic uncertainty (with the possible excep- tion analysis (Baron and Kenny, 1986). This analysis allows us to isolate tion being Opposition seats in column (3)). The Cragg-Donald F-statistic the direct and indirect effects of economic uncertainty on corrupt indicates that the instruments are not weak since the values are higher behavior. Specifically, mediation takes place if economic uncertainty af- than the traditional threshold of 10 used by Stock and Yogo (2005). The fects cheating on taxes and paying bribes through another variable. In our Hansen’s j-statistic shows that the instruments do not overidentify the case, the mediator variables are the managerial perceptions regarding the estimation model. eight business obstacles identified in Table 4. The second stage results in columns (2) and (4) lead to qualitatively Fig. 1 depicts the structural equations in the mediation analysis. similar conclusions to the OLS and Tobit specifications in Table 5. Again, Based on the standardized coefficients in the figure, managerial the IV regression results support our Hypothesis 2 – higher economic perception regarding dysfunctional court system; crime, theft, and dis- uncertainty results in more corrupt behavior by firms in the form of order; prevalent corruption; anti-competitive practices; and political cheating on taxes and paying bribes. instability have a positive and indirect effect on tax cheating through rising economic uncertainty. Similarly, for paying bribes, almost all factors have positive and indirect effects. Table 9 shows that the factors that are related to managerial perception explain about 4.2% of the total effect of economic uncertainty on tax cheating and 18.5% of the effect 13 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 8 first alternative measure, Election closeness, was described in subsection Economic uncertainty and corruption: instrumental variable regressions. 3.4. In this subsection, rather than using Election closeness as an instru- ment for the WUI variable, we instead use it as a direct measure of Estimation model IV first IV second IV first IV second stage stage stage stage economic uncertainty (like in Julio and Yook, 2012). This use facilitates Dependent Uncertainty Cheating on Uncertainty Bribes a robustness check related to any possible measurement error bias that variable taxes originates from the construction of WUI by Ahir et al. (2019). (1) (2) (3) (4) In our second measure, we isolate only the local component of a Election closeness 0.010*** 0.009*** country’s economic uncertainty by orthogonalizing the country-level (8.17) (4.91) uncertainty index to the global aggregate uncertainty index available Opposition seats 0.032* 0.014 in Ahir et al. (2019). We run a simple OLS estimate with no constant (1.74) (0.87) Uncertainty 2.309*** 0.150*** term where a country-level measure is the dependent variable and (3.98) (2.61) global aggregate uncertainty is the only explanatory variable. To isolate Size -0.000 0.009*** -0.001* -0.001*** only the orthogonal component of local measures of policy uncertainty, ( 0.24) (3.91) ( 1.67) ( 4.08) we extract the residual values from this regression and use them as our Age -0.001 -0.026*** 0.003*** -0.002*** second uncertainty measure. We carry out this orthogonalization pro- ( 1.59) ( 5.48) (4.30) ( 4.02) Exporter 0.003** -0.044*** 0.005*** 0.001 cedure to obtain the portion of country-level uncertainty that corre- (2.20) ( 4.88) (3.08) (1.04) sponds only to the uncertainties pertaining to economic, political, and Government 0.006** -0.087*** 0.000 0.006** financial aspects of the country itself instead of the global uncertainty (2.28) ( 6.67) (0.14) (2.56) that affects all countries at that time. Foreign -0.000 -0.049*** 0.003** 0.002 Our third measure uses a country’s stock market volatility. This ( 0.20) ( 5.55) (2.04) (1.15) GDP 0.013*** -0.015 0.014*** -0.004*** measure is widely used as a proxy for economic uncertainty and corre- (5.86) ( 1.16) (4.72) ( 2.94) lates positively with the news-based index (Baker et al., 2016) as well as ΔGDP -0.006*** 0.024*** -0.002** -0.000** with the world uncertainty index (Ahir et al., 2019). To calculate the ( 7.63) (4.02) ( 2.20) ( 1.98) stock market volatility, we measure the standard deviation of the major Inflation 0.001*** -0.000 0.001*** -0.000 (3.30) ( 0.34) (4.62) ( 1.10) stock index of each country. We obtain data on stock index values from Country -0.018*** 0.010 -0.015*** 0.000 Compustat’s Global Index Prices dataset that is available at a monthly Governance frequency for only 32 countries in our sample. We then estimate a ( 9.90) (0.89) ( 7.95) (0.02) moving 12-month standard deviation for each index. Further, we Industry fixed Yes No Yes No effects combine this information with WBES data using the month and year of Year fixed effects Yes No Yes No the interview. Since not all countries have a stock market in our sample, Adjusted R 0.394 -0.028 0.200 -0.008 this results in a large reduction in the sample. Observations 29,376 29,376 68,305 68,305 For the fourth measure, we use the natural logarithm of the news- Cragg-Donald F- 1123.461 1070.313 based index available for a subset of sampled countries. In our sample, statistic Hansen j-statistic 0.831 0.998 we are able to obtain this information for Brazil, Chile, Colombia, p-value 0.362 0.318 Greece, India, Mexico, and the Russian Federation. Additionally, we use the European index for countries in Europe to increase our sample size This table presents the instrumental variable (IV) regression results from (this adds 16 more countries to our sample). We calculate the natural regressing the enterprise-level norm-deviant behavior on the level of economic uncertainty in the country and on other explanatory variables. Private and logarithm of the 3-month moving average news-based economic policy public firms are pooled together. Uncertainty is the World Uncertainty Index uncertainty for these countries. Using this measure also causes a large (WUI) developed by Ahir et al. (2019). Our instruments for Uncertainty are: reduction in the sample. Election closeness, which is the degree of competitiveness of the most recent 19 The results from using these four measures are provided in national election (for executive branch) in the country and Opposition seats, Table 10. The coefficients in all our specifications are positive and sta- which is the proportion of seats held by the largest opposition party in the given tistically significant at the 1% level. Hence, our results are not driven by country and year. Cheating on taxes is the fraction of sales not reported to the tax the worldwide changes in the economic uncertainty or the measurement authorities. Bribes is the fraction of sales paid in informal payments to the government authorities. Country, industry, and year fixed effects are included. All other variables are defined in Appendix in Table 13. The t-statistics based on industry clustered robust standard errors are shown in parentheses. Orthogonalization is implemented by running the following OLS: WUI = j,t Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, β GlobalWUI + є , where WUI is the world uncertainty index of a country j t j,t j,t j,t respectively. during time t, and GlobalWUI during time t is represented by the unweighted average of countries’ WUIs. Ahir et al. (2019) refer to this global index as the on paying bribes. While the indirect effects are both statistically sig- unweighted Global WUI during that period t (see their Fig. 1B). Note that the nificant, the indirect economic effect of a worse managerial perception above regression has no constant term. The local component of WUI for a is higher on paying bribes. country j in time t is the residual term, є . j,t Since we choose only one index per country (the largest stock index in terms of market capitalization), our final sample has one entry per country, month, 5. Further analyses and year. The 32 countries with both bribery data and stock index data at the time of the survey are Argentina, Bangladesh, Brazil, Bulgaria, Chile, Croatia, 5.1. Alternative measures of policy uncertainty Czech Republic, Ecuador, Egypt, Greece, Hungary, India, Indonesia, Jamaica, Jordan, Kenya, Lithuania, Malaysia, Mexico, Morocco, Nigeria, Pakistan, Peru, To reduce the measurement error issues with our uncertainty vari- Philippines, Poland, Russian Federation, Slovakia, Slovenia, Sri Lanka, ables, we use four additional measures of economic uncertainty. Our Thailand, Turkey, and Venezuela. The three alternative uncertainty measures and our original economic un- certainty measure (WUI) are strongly and positively correlated with each other. This is consistent with the findings of Ahir et al. (2019), who test their measure The indirect effects are calculated using the following formula: (indirect of uncertainty index against news-based economic policy uncertainty of Baker effect of managerial perception/total effect of uncertainty)*100. For Cheating on et al. (2016) and the stock market volatility and they report significantly pos- taxes: (0.004/0.096)*100 = 4.2% and for Bribes: (0.005/0.027)*100 = 18.5% itive correlations. 14 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Fig. 1. Mediation analysis: Isolating the direct and indirect effects of policy uncertainty. This figure presents the structural equation modeling of the mediation analysis to assess the direct and indirect effects of economic policy uncertainty on cheating on taxes and paying bribes. Uncertainty is the country-level World Uncertainty Index (WUI) developed by Ahir et al. (2019) using the country reports prepared by Economist Intelligence Unit (EIU). Standardized coefficients are reported next to each line. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively. 15 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 9 Economic uncertainty and corruption: mediation analysis. Dependent variable Cheating on taxes Cheating on taxes Bribes Bribes (1) (2) (3) (4) Total effect of Uncertainty 0.096*** 0.031*** (4.84) (5.62) Direct effect of Uncertainty 0.101*** 0.027*** (5.13) (4.73) 0.004** 0.005*** Mediating effect of Managerial (2.28) (3.89) perception Size -0.143*** -0.150*** -0.070*** -0.078*** ( 7.98) ( 8.00) ( 8.41) ( 9.48) Age -0.027*** -0.027*** -0.007 -0.008* ( 4.65) ( 4.64) ( 1.51) ( 1.71) Exporter -0.006 -0.006 0.008 0.007 ( 0.83) ( 0.81) (1.29) (1.19) Government -0.012** -0.009* 0.013** 0.015*** ( 2.17) ( 1.78) (2.24) (2.75) Foreign -0.023*** -0.023*** 0.006 0.006 ( 4.68) ( 4.89) (1.04) (1.13) GDP -1.223 -1.608 -0.169 -0.064 ( 0.64) ( 0.86) ( 1.49) ( 0.55) ΔGDP -0.267** -0.240** -0.064*** -0.062*** ( 2.13) ( 1.99) ( 6.56) ( 6.23) Inflation 0.226 0.243 0.029* 0.025 (1.41) (1.53) (1.92) (1.61) Country Governance 1.231*** 1.245*** -0.026 -0.021 (5.46) (5.43) ( 0.56) ( 0.45) Percent of total effect mediated 4.2% 18.5% Country fixed effects Yes Yes Yes Yes Industry fixed effects Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Adjusted/Overall R 0.209 0.221 0.070 0.110 Observations 30,032 30,032 30,032 30,032 This table presents the mediation analysis results from regressing the enterprise-level norm-deviant behavior on the level of economic uncertainty in the country and on other explanatory variables. Private and public firms are pooled together. Uncertainty is the World Uncertainty Index (WUI) developed by Ahir et al. (2019). Cheating on taxes is the fraction of sales not reported to the tax authorities. Bribes is the fraction of sales paid in informal payments to the government authorities. Managerial perception is the total effect of all variables pertaining to business obstacles indicating the interviewee’s opinion about the severity of an obstacle. Country, industry, and year fixed effects are included. All other variables are defined in Appendix in Table 13. The t-statistics based on industry clustered robust standard errors are shown in parentheses. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively. error in our main uncertainty variable. These robustness tests further 5.2. Political and societal factors and the uncertainty-corruption corroborate our claim that rising economic uncertainty induces many relationship private firms around the Globe to cheat on taxes and pay more bribes. These findings also confirm that the link between uncertainty and cor- Next, we explore the heterogeneity of our sample and examine how ruption exists even for the larger and more visible countries that are different political, economic, and societal factors enhance or reduce the calculated by Baker et al. (2016) economic uncertainty index (the uncertainty-corruption relationship. above-mentioned 23 countries that are either located in Europe or are among the largest emerging markets). Thus, the negative externality of 5.2.1. Political and economic dimensions rising economic uncertainty affects all countries regardless of their size First, we focus on the political and economic differences across and geographic location. countries. Previous evidence indicates that parliamentary systems are more likely to have large policy swings and to create a more uncertain environment for the economic agents (Julio and Yook, 2012). We therefore predict that the uncertainty-corruption relationship will be stronger in countries with a parliamentary system of government. We test this prediction by interacting a dummy variable that equals one if the country in which the firm is located in has a parliamentary system of Another widely used measure of policy uncertainty is national elections (see government, and zero otherwise. The results for this specification are for instance, Julio and Yook, 2012). Election-based policy uncertainty in- provided in columns (1) and (2) of Table 11. The interaction term is dicators have the advantage of capturing the exogenous variation in economic positive and statistically significant only for cheating on taxes. During uncertainty, however they do not tell us how much election uncertainty (relative to other countries’ elections) exists at the time of a given country’s high economic uncertainty, the firms located in countries with parlia- election. Also, these measures assume absence of uncertainty during mentary systems cheat more on taxes but do not pay more bribes relative non-election years; yet our sample is comprised of primarily low-income and to firms located in countries with presidential systems. This is consistent underdeveloped countries, where policy uncertainty is presumably very high with our prediction. even in the absence of elections. Furthermore, using election-based indicators We also analyze how the level of economic development (i.e., per- (which are in the form of dummies that take a value of one every 4–5 years capita income) of each country influences the uncertainty-corruption during the election year) leads to large attrition of our sample, which is based relationship. We obtain country-level income groupings from the on surveys that are not conducted every year (see Table 1). Therefore, in our World Bank. We predict that the private firms located in low-income context, using non-election-based proxies for economic uncertainty seems more countries will engage more in cheating on taxes and paying bribes due appropriate. 16 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 10 Economic uncertainty and corruption: alternative measure of economic uncertainty. Uncertainty measure Election closeness Orthogonalized uncertainty Stock market volatility News-based uncertainty (EPUI) Dependent variable Cheating on taxes Bribes Cheating on taxes Bribes Cheating on taxes Bribes Cheating on taxes Bribes (1) (2) (3) (4) (5) (6) (7) (8) Uncertainty 0.020*** 0.001*** 0.678*** 0.128*** 0.254*** 0.030*** 0.307*** 0.036*** (4.53) (3.74) (4.12) (4.29) (4.20) (4.55) (2.97) (3.10) Size 0.009*** -0.001*** -0.026*** -0.003*** -0.028*** -0.003*** -0.024*** -0.004*** (5.43) ( 5.32) ( 8.58) ( 4.48) ( 10.53) ( 3.29) ( 7.39) ( 3.88) Age -0.026*** -0.001*** -0.020*** -0.001 -0.022** -0.002 -0.022*** 0.001 ( 7.58) ( 3.19) ( 4.26) ( 1.28) ( 2.41) ( 1.15) ( 3.05) (0.51) Exporter -0.036*** 0.002 -0.020* 0.009*** -0.018 0.012*** 0.006 -0.001 ( 4.16) (1.67) ( 1.89) (2.68) ( 0.81) (3.33) (0.45) ( 0.30) Government -0.082*** 0.006** -0.147*** -0.000 -0.173*** 0.019 -0.187*** -0.021** ( 9.58) (2.38) ( 5.58) ( 0.04) ( 2.86) (1.37) ( 5.45) ( 1.97) Foreign -0.051*** 0.002 -0.070*** -0.002 -0.054*** 0.011 -0.110*** 0.006 ( 7.00) (1.51) ( 6.26) ( 0.57) ( 3.18) (1.60) ( 5.17) (1.47) GDP 0.014* -0.002*** -0.372 -0.005 -0.635*** -0.104** 0.096 -0.314*** (1.85) ( 3.76) ( 0.42) ( 0.15) ( 3.69) ( 2.20) (0.75) ( 9.16) ΔGDP 0.009*** -0.001*** -0.056** -0.006*** 0.140** -0.004* -0.532* 0.002* (4.02) ( 8.64) ( 2.36) ( 7.95) (2.35) ( 1.94) ( 1.76) (1.65) Inflation 0.002*** 0.000 0.018*** 0.001* -0.187*** -0.001 -0.086** -0.002*** (3.35) (1.59) (3.16) (1.65) ( 3.29) ( 0.68) ( 2.15) ( 3.63) Country Governance -0.028*** -0.002*** 0.302*** -0.011 -0.853*** -0.025 -1.392* -0.003 ( 5.58) ( 5.49) (4.31) ( 1.51) ( 3.44) ( 1.04) ( 1.76) ( 0.34) Country fixed effects No No Yes Yes Yes Yes Yes Yes Industry fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Pseudo R 0.089 0.032 0.150 0.393 0.116 0.406 0.148 0.735 Observations 29,598 69,549 30,032 72,704 12,656 35,711 11,481 24,717 This table presents the relationship between the enterprise-level norm-deviant behavior and the level of economic uncertainty when alternative measures of uncer- tainty are used. In columns (1) and (2), the economic uncertainty measure is Election closeness, which is the degree of competitiveness of the most recent national election (for executive branch) in the country. In columns (3) and (4), the economic uncertainty measure is the orthogonalized measure of uncertainty (i.e., country- specific WUI orthogonalized by the aggregated uncertainty index for the whole World provided in Ahir et al., 2019). In columns (5) and (6), the measure is stock market volatility, calculated as the standard deviation of the major stock index in the given country over the past 12 months. This index can be calculated for only 20 countries in our sample. In columns (7) and (8) the measure is the weighted three-month moving average of the news-based economic policy uncertainty index (EPUI) created by Baker et al. (2016), which is available only for 23 out of 93 countries in our sample. Private and public firms are pooled together. Cheating on taxes is the fraction of sales not reported to the tax authorities. Bribes is the fraction of sales paid in informal payments to the government authorities. Country, industry, and year fixed effects are included. All other variables are defined in Appendix in Table 13. The t-statistics based on industry clustered robust standard errors are shown in parentheses. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively. to the lack of oversight and transparency by well-funded and sophisti- calculate country-level trust and norms of civic cooperation measures. cated regulatory authorities. We, therefore, interact an indicator vari- We then rerun our baseline model with an interaction term. The results able which equals one if the firm is located in a low-income country, and reported in columns (1) and (2) of Table 12, show that firms located in zero otherwise. The results provided in columns (3) and (4) of Table 11 countries with higher levels of civic norms engage in less cheating on are consistent with our predictions. The coefficient for the interaction taxes, but civic norms do not mitigate bribery. This mixed result in- term is positive and statistically significant in both specifications that dicates that civic norms, as a replacement for a country’s governance indicates the marginal effect of economic uncertainty on corrupt institutions, reduces corrupt behavior to some extent when faced with behavior in private firms is higher in low-income and underdeveloped higher economic uncertainty. These results are largely consistent with countries. Thus, economic underdevelopment of a country amplifies the the literature on trust (Aghion et al., 2010; Kanagaretnam et al., 2018) corruptive effects that higher policy uncertainty has on small private and Hypothesis 4. firms. Next, we look at how religiosity can affect the uncertainty-corruption relationship. We obtain data on country-level religiosity from a Gallup 5.2.2. The role of societal norms poll conducted in 2009 that asked individuals in each country about the Previous evidence indicates that societal norms reduce norm-deviant importance of religion in their lives. Using this information, we create a behavior. Based on the findings of Putnam et al. (1994), La Porta et al. dummy variable that equals one if more than 70% of the residents in a (1997), and Fisman and Miguel (2007), we argue that societal norms are country replied “yes” when asked, “Is religion important in your daily a strong determinant of whether corruption exists. Similarly, Kanagar- life?”, and zero otherwise. Religion is shown to reduce levels of etnam et al. (2018) and Hasan et al. (2017b) show that societal trust and perceived corruption (Mensah, 2014) as well as mitigate corporate social capital are negatively associated with corporate tax avoidance. Jin misconduct such as financial reporting fraud (McGuire et al., 2012). We, et al. (2017) show that banks in high social capital regions in the US therefore, predict that the religiosity of a country’s residents can break, were less prone to financial failures in the recent global financial crisis or at least mitigate, the uncertainty-corruption relationship. The results due to their higher transparency. Furthermore, following the arguments for this specification are provided in columns (3) and (4) of Table 12. in Aghion et al. (2010), we predict that trust becomes vital when The interaction terms are once again negative and statistically signifi - country-level governance institutions are weak. Consistent with this cant. These results are also consistent with Hypothesis 4 that when prediction, we argue that civic norms in the form of trust and civic country-level regulatory agencies start becoming ineffective (i.e., cor- cooperation negatively influence the uncertainty-corruption ruption in private firms starts rising together with the economic un- relationship. certainty), religiosity as a social norm can have a certain ameliorating To test this prediction, we collect data from the World Values Survey effect in reducing cheating and bribery. (WVS) and follow the methodology in Knack and Keefer (1997) to 17 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 11 Economic uncertainty and corruption: do country’s political system and income level matter?. Dependent variable Cheating on taxes Bribes Cheating on taxes Bribes Country Characteristic Parliamentary system of Parliamentary system of Low-income Low-income government government countries countries (1) (2) (3) (4) Uncertainty 0.834*** 0.070*** 0.964*** 0.040* (3.86) (3.44) (3.63) (1.78) Characteristic -0.161*** 0.007 0.161*** 0.088*** ( 3.48) (0.68) (3.48) (8.33) ** Uncertainty × Country Characteristic 2.055*** 0.036 1.086 0.103* (5.21) (0.31) (2.07) (1.69) Size 0.008*** -0.002*** 0.009*** -0.002*** (2.95) ( 2.85) (3.28) ( 3.19) Age -0.047*** -0.006*** -0.042*** -0.004*** ( 6.35) ( 4.26) ( 4.94) ( 3.16) Exporter -0.070*** 0.013*** -0.081*** 0.011*** ( 4.83) (3.84) ( 5.23) (3.57) Government -0.234*** 0.004 -0.284*** -0.008 ( 5.54) (0.44) ( 11.77) ( 1.04) Foreign -0.136*** -0.002 -0.123*** -0.002 ( 7.16) ( 0.43) ( 7.15) ( 0.40) GDP -0.007 -0.014*** 0.094*** 0.021*** ( 0.62) ( 5.11) (5.18) (4.93) ΔGDP 0.021*** -0.002*** 0.017*** -0.004*** (4.50) ( 3.93) (4.02) ( 6.24) Inflation 0.002** 0.000 -0.002 -0.000 (2.12) (1.18) ( 1.20) ( 1.29) Country Governance -0.006 -0.018*** -0.044*** -0.024*** ( 1.04) ( 11.82) ( 5.02) ( 10.48) Country fixed effects Yes Yes Yes Yes Industry fixed effects Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Pseudo R 0.062 0.233 0.065 0.258 Observations 28,737 68,933 30,032 72,118 P(Uncertainty + Characteristic + Uncertainty × 0.000 0.000 0.000 0.000 Characteristic = 0) The table conducts cross-country analyses to assess the role of political system and income level. The results are obtained from regressing the enterprise-level norm- deviant behavior on the level of economic uncertainty, country characteristic, and other explanatory variables. In columns (1) and (2), Characteristic equals one if the country has a parliamentary system of government, and zero otherwise. In columns (3) and (4), Characteristic equals one if the country belongs to the low-income group based on World Bank’s classification, and zero otherwise. Private and public firms are pooled together. Uncertainty is the World Uncertainty Index (WUI) developed by Ahir et al. (2019). Cheating on taxes is the fraction of sales not reported to the tax authorities. Bribes is the fraction of sales paid in informal payments to the government authorities. Country, industry, and year fixed effects are included. All other variables are defined in Appendix in Table 13. The t-statistics based on industry clustered robust standard errors are shown in parentheses. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively. 5.2.3. Other socioeconomic dimensions in breaking the link between economic uncertainty and firm-level In addition to the above dimensions, we consider several other so- corruption. cioeconomic, political, and geographic aspects of a society (or a coun- try). We conduct a series of cross-country tests (untabulated) to 5.3. An alternative measure of corruption: time spent with government understand what influences the relationship between policy uncertainty 21 authorities and corruption. First, we test how cultural aspects of a society can influence the uncertainty-corruption relationship. We obtain data on In an untabulated test, we also analyze whether the management cultural dimensions from Hofstede (1980). Consistent with Husted spends more time with government authorities. This measure may not (1999), we find that societies at higher levels in the uncertainty avoid- directly relate to corruption, but it still reflects the overall inefficiency of ance index (UAI) pay more bribes during high economic uncertainty. For the firm in its involvement with the government. While the WBES survey cheating on taxes, the cross-country results split according to the UAI does not provide the reason for the time spent with government au- and are not statistically significant at conventional levels. thorities, we can safely assume that any excessive or unnecessary time Second, we find that there are some differences between the coun- spent with government authorities is a sign of norm-deviant behavior in tries located in different continents (Africa, Asia, Europe, and North and the form of lobbying or another such activity. South Americas) in terms of cheating on taxes and paying bribes. For To explore this association, we use the amount of time that man- instance, cheating is more prevalent in Africa and Europe, while bribery agement spends with the government authorities in a given year as a is more common in Asia. However, regardless of the continent they are percentage of their total working hours. The mean (median) percentage located on, the private firms react to policy uncertainty by engaging in time spent with the government is 10.44% (4.00%) of the management’s one form of norm-deviant behavior or the other. Furthermore, as we total working hours during a year. Our regression analyses estimate that demonstrated earlier, regardless of the continent, a country’s socio- during high economic uncertainty, firms ’ management spends more economic features and various political institutions do make a difference time with the government. The coefficient for our uncertainty measure is statistically significant at the 5% level and its economic impact is as follows: A one standard deviation increase in our WUI measure trans- lates into about 0.49% more time spent with the government, which While these tests are untabulated, they are available from the authors on request. roughly corresponds to two additional days. 18 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 12 Economic uncertainty and corruption: the mitigating role of civic norms and religiosity. Dependent variable Cheating on taxes Bribes Cheating on taxes Bribes Social Norm Civic norms Civic norms Religion Religion (1) (2) (3) (4) Uncertainty 12.002*** 0.146 3.432*** 0.139** (5.10) (1.52) (5.65) (2.39) Social Norm 2.485 -0.018 0.326*** 0.006 (1.35) ( 0.49) (6.89) (1.14) Uncertainty × Social Norm -13.364*** -0.407 -2.533*** -0.117* (¡4.34) (¡1.43) (¡3.65) (¡1.85) Size -0.027*** -0.001** 0.008*** -0.002*** ( 8.16) ( 2.05) (3.72) ( 3.02) Age -0.016** -0.002 -0.040*** -0.004*** ( 2.40) ( 1.45) ( 5.59) ( 3.42) Exporter -0.015 0.008** -0.071*** 0.012*** ( 1.13) (1.98) ( 5.11) (3.71) Government -0.157*** 0.005 -0.227*** -0.000 ( 5.84) (0.46) ( 9.07) ( 0.01) Foreign -0.051*** 0.003 -0.125*** -0.002 ( 3.58) (0.76) ( 9.31) ( 0.47) GDP 3.041*** -0.043 0.020* -0.014*** (5.30) ( 1.23) (1.71) ( 4.98) ΔGDP 0.204*** -0.002*** 0.022*** -0.003*** (4.67) ( 2.99) (5.13) ( 4.68) Inflation 0.068*** -0.000 0.004*** 0.000 (4.37) ( 0.71) (3.53) (1.36) Country Governance 0.773*** -0.032*** -0.001 -0.020*** (4.94) ( 2.91) ( 0.16) ( 10.19) Country fixed effects Yes Yes No No Industry fixed effects Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Pseudo R 0.140 0.397 0.072 0.230 Observations 22,990 56,793 29,902 72,352 P(Uncertainty + Norm + 0.000 0.000 0.000 0.000 Uncertainty × Norm = 0) The table analyses the role of social norms, such as civic norms and religiosity. The results are obtained by regressing the enterprise-level norm-deviant behavior on the level of economic uncertainty, social norm, and other explanatory variables. Civic norms is the average level of trust in others and norms of civic cooperation in a country based on World Values Survey (WVS) data, normalized to range between zero and one. Religion equals one if more than 70% of the country’s residents replied ‘yes’ when asked, ‘Is religion important in your daily life?’ in a Gallop Poll in 2009. Private and public firms are pooled together. Cheating on taxes is the fraction of sales not reported to the tax authorities. Bribes is the fraction of sales paid in informal payments to the government authorities. Uncertainty is the World Uncertainty Index (WUI) developed by Ahir et al. (2019). Country, industry, and year fixed effects are included. All other variables are defined in Appendix in Table 13. The t-statistics based on industry clustered robust standard errors are shown in parentheses. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively. We also test whether firms get more frequent visits from tax au- predicted by “the greasing the wheels” hypothesis (Acemoglu and Ver- thorities during higher uncertainty periods. We find that tax authorities dier, 2000; Leff, 1964; Mendoza et al., 2015), firms engage in corrupt visit firms less frequently during such times. This infrequency further behavior to reduce the inefficiencies associated with the elevated policy indicates that the time spent with government authorities is not due to uncertainty that originates from political and governmental activities. taxes, but perhaps due to another activity such as lobbying. Less over- We report empirical evidence consistent with this argument. Specif- sight from authorities during uncertain times paves the way for norm- ically, we show that firms pay more bribes and cheat more often on taxes deviant behavior like cheating on taxes. in the presence of higher economic policy uncertainty. These results are robust to endogeneity bias; industry, year, and country fixed effects; and 6. Conclusion alternative measures of policy uncertainty. We use the heterogeneity of our sample to find that when facing We study how policy uncertainty contributes to the severity of the policy uncertainty, private firms ’ corruption is different from that of obstacles that face private firms in 93 countries. The firms report that public firms; the former cheats by underreporting taxes and the latter they face more severe constraints and higher socioeconomic in- pays more bribes. Different political, legal, and societal factors also efficiencies when policy uncertainty is higher. We argue that the higher affect the uncertainty-corruption relationship. Since a parliamentary unpredictability of future economic and political outcomes and more system of government is more prone to large economic policy swings, we severe business obstacles induce an environment of uncertainty and find that firms located in such countries are more likely to cheat on inefficiency. When facing an uncertain environment, firm managers taxes. The uncertainty-corruption relationship is stronger in firms become more worried about their own future (Hogg, 2007) and pay located in low-income countries. On the other hand, societal factors such more attention to negative events and obstacles (Arenas et al., 2006). as civic norms and religion mitigate the uncertainty-corruption rela- Some of them ultimately resort to corruption to cope with the high tionship. These results provide new insights on how policy uncertainty uncertainty and the perceived or real inefficiencies it brings. As influences corporate practices of small firms around the world. 19 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Appendix Table 13 Variable definitions. Variable Description Source Uncertainty The World Uncertainty Index (WUI) developed by Ahir et al. (2019) Ahir et al. (2019) Election closeness The degree of competitiveness of the most recent national election (for executive branch) in the country Database of Political Institutions Opposition seats The proportion of seats held by the largest opposition party in the given country and year. Database of Political Institutions Cheating on taxes The fraction of sales not reported to the tax authorities WBES Bribes The fraction of sales paid in informal payments to the government authorities WBES Dysfunctional court How much of an obstacle is the court system to your enterprise? Ranges between 0 (no obstacle) and 5 (very severe WBES system obstacle) Crime, theft, disorder How much of an obstacle is crime, theft, and disorder to your enterprise? Ranges between 0 (no obstacle) and 5 (very WBES severe obstacle) Prevalent corruption How much of an obstacle is corruption to your enterprise? Ranges between 0 (no obstacle) and 5 (very severe obstacle) WBES High tax rates How much of an obstacle is tax rates to your enterprise? Ranges between 0 (no obstacle) and 5 (very severe obstacle) WBES Ineffective tax How much of an obstacle is tax administration to your enterprise? Ranges between 0 (no obstacle) and 5 (very severe WBES administration obstacle) Anti-competitive practices How much of an obstacle is practices of competitors in informal sector to your enterprise? Ranges between 0 (no obstacle) WBES and 5 (very severe obstacle) Reduced access to finance How much of an obstacle is access to finance to your enterprise? Ranges between 0 (no obstacle) and 5 (very severe WBES obstacle) Political instability How much of an obstacle is political instability to your enterprise? Ranges between 0 (no obstacle) and 5 (very severe WBES obstacle) Size The firm ’s total sales in the previous fiscal year (thousands of USD) WBES Age The firm age in years WBES Exporter Equals one if at least ten percent of firm ’s sales are exported, and zero otherwise WBES Government Equals one if at least ten percent of the firm is owned by the government, and zero otherwise WBES Foreign Equals one if at least ten percent of the firm is owned by a foreign entity, and zero otherwise WBES GDP The gross domestic product per capita for the country World Bank ΔGDP The growth in GDP for the country World Bank Inflation The inflation rate for the country World Bank Governance The first principal component of six governance indicators for each country (see Kaufmann et al., 2009) World Bank Civic norms The average level of trust in others and norms of civic cooperation in a country. WVS Religion Equals one if more than 70% of the country’s residents replied ‘yes’ when asked, ‘Is religion important in your daily life?’ Gallop in a Gallop Poll in 2009 Stock market volatility Calculated as the standard deviation of the major stock index in the given country over the past 12 months Compustat Global News-based uncertainty The weighted three-month moving average of the news-based economic policy uncertainty index Baker et al. (2016) Beck, T., Demirgüç-Kunt, A., Maksimovic, V., 2005. Financial and legal constraints to References growth: does firm size matter? J. Financ. 60, 137–177. Beltratti, A., Stulz, R.M., 2012. The credit crisis around the globe: why did some banks Acemoglu, D., Verdier, T., 2000. The choice between market failures and corruption. Am. perform better? J. Financ. Econ. 105, 1–17. Econ. Rev. 90, 194–211. Bentolila, S., Bertola, G., 1990. Firing costs and labour demand: how bad is Adelopo, I., Rufai, I., 2020. Trust deficit and anti-corruption initiatives. J. Bus. Ethics eurosclerosis? Rev. Econ. Stud. 57, 381. 163, 429–449. Bernanke, B.S., 1983. Irreversibility, uncertainty, and cyclical investment. Q. J. Econ. 98, Aghion, P., Algan, Y., Cahuc, P., Shleifer, A., 2010. Regulation and distrust. Q. J. Econ. 85–106. 125, 1015–1049. Bhattacharya, U., Hsu, P.H., Tian, X., Xu, Y., 2017. What affects innovation more: policy Aguilera, R.V., Vadera, A.K., 2008. The dark side of authority: antecedents, mechanisms, or policy uncertainty? J. Financ. Quant. Anal. 52, 1869–1901. and outcomes of organizational corruption. J. Bus. Ethics 77, 431–449. Bloom, N., 2009. The impact of uncertainty shocks. Econometrica 77, 623–685. Ahir, H., Bloom, N., Furceri, D., 2019. The world uncertainty index. Stanf. Inst. Econ. Bloom, N., Bond, S., Van Reenen, J., 2007. Uncertainty and investment dynamics. Rev. Policy Res. Econ. Stud. 74, 391–415. ¨¨ ¨ Airaksinen, A., Luomaranta, H., Roodhuijzen, A., Alajaasko, P., 2016. Statistics on Small Bonaime, A., Gulen, H., Ion, M., 2018. Does policy uncertainty affect mergers and and Medium-Sized Enterprises - Statistics Explained, Eurostat Statistics Explained. acquisitions? J. Financ. Econ. 129, 531–558. Amore, M.D., Minichilli, A., 2018. Local political uncertainty, family control, and Boutchkova, M., Doshi, H., Durnev, A., Molchanov, A., 2012. Precarious politics and investment behavior. J. Financ. Quant. Anal. 53, 1781–1804. return volatility. Rev. Financ. Stud. 25, 1111–1154. An, H., Chen, Y., Luo, D., Zhang, T., 2016. Political uncertainty and corporate Braun, M., Di Tella, R., 2004. Inflation, inflation variability, and corruption. Econ. Polit. investment: evidence from China. J. Corp. Financ. 36, 174–189. 16, 77–100. Anand, V., Ashforth, B.E., Joshi, M., 2004. Business as usual: the acceptance and Brav, O., 2009. Access to capital, capital structure, and the funding of the firm. J. Financ. perpetuation of corruption in organizations. Acad. Manag. Exec. 18, 39–53. 64, 263–308. Arenas, A., Tabernero, C., Briones, E., 2006. Effects of goal orientation, error orientation Brogaard, J., Detzel, A., 2015. The asset-pricing implications of government economic and self-efficacy on performance in an uncertain situation. Soc. Behav. Pers. 34, policy uncertainty. Manag. Sci. 61, 3–18. 569–586. Caballero, R.J., Engel, E.M.R.A., Haltiwanger, J.C., 1995. Plant-level adjustment and Ashforth, B.E., Anand, V., 2003. The normalization of corruption in organizations. Res. aggregate investment dynamics. Brook. Pap. Econ. Act. 1995, 1–54. Organ. Behav. 25, 1–52. Callen, J.L., Fang, X., 2015. Religion and stock price crash risk. J. Financ. Quant. Anal. Ashraf, B.N., Shen, Y., 2019. Economic policy uncertainty and banks’ loan pricing. 50, 169–195. J. Financ. Stab. 44, 100695. Chan, Y.C., Saffar, W., Wei, K.C.J., 2021. How economic policy uncertainty affects the Asker, J., Farre-Mensa, J., Ljungqvist, A., 2011. Comparing the investment behavior of cost of raising equity capital: Evidence from seasoned equity offerings. J. Financ. public and private firms. Natl. Bur. Econ. Res. Stab. 53, 100841. Baker, S.R., Bloom, N., Davis, S.J., 2016. Measuring economic policy uncertainty. Q. J. Chen, F., Hope, O.K., Li, Q., Wang, X., 2011. Financial reporting quality and investment Econ. 131, 1593–1636. efficiency of private firms in emerging markets. Account. Rev. 86, 1255–1288. Bardhan, P., 1997. Corruption and development: a review of issues. J. Econ. Lit. 35, Clarke, G., Xu, L., 2002. Ownership, competition, and corruption: bribe takers versus 1320–1346. bribe payers. World Bank Policy Res. Work. Pap. 2783. Baron, R.M., Kenny, D.A., 1986. The moderator-mediator variable distinction in social Cloyd, C.B., Pratt, J., Stock, T., 1996. The use of financial accounting choice to support psychological research. Conceptual, strategic, and statistical considerations. J. Pers. aggressive tax positions: public and private firms. J. Account. Res. 34, 23. Soc. Psychol. 51, 1173–1182. 20 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Cohen, L., Coval, J., Malloy, C., 2011. Do powerful politicians cause corporate La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R.W., 1997. Trust in large downsizing? J. Polit. Econ. 119, 1015–1060. organizations. Am. Econ. Rev. 87, 333–338. Cohen, L., Malloy, C.J., 2014. Friends in high places. Am. Econ. J. Econ. Policy 6, 63–91. Lawless, M., O’Connell, B., O’Toole, C., 2015. SME recovery following a financial crisis: Colak, G., Durnev, A., Qian, Y., 2017. Political uncertainty and IPO activity: evidence does debt overhang matter? J. Financ. Stab. 19, 45–59. from U.S. gubernatorial elections. J. Financ. Quant. Anal. 52, 2523–2564. Leahy, J.V., Whited, T.M., 1996. The effect of uncertainty on investment: some stylized ¨ ¨ Colak, G., Gungoraydinoglu, A., Oztekin, O., 2018. Global leverage adjustments, facts. J. Money Credit Bank. 28, 64–83. uncertainty, and country institutional strength. J. Financ. Intermed. 35, 41–56. Lederman, D., 2010. An international multilevel analysis of product innovation. J. Int. Collins, J.D., Uhlenbruck, K., Rodriguez, P., 2009. Why firms engage in corruption: a top Bus. Stud. 41, 606–619. management perspective. J. Bus. Ethics 87, 89–108. Leff, N.H., 1964. Economic development through bureaucratic corruption. Am. Behav. Cooper, R.W., Haltiwanger, J.C., 2006. On the nature of capital adjustment costs. Rev. Sci. 8, 8–14. Econ. Stud. 73, 611–633. Mauro, P., 1995. Corruption and growth. Q. J. Econ. 110, 681–712. Cutler, L.N., 1988. Some reflections about divided government. Pres. Stud. Q 18, McGuire, S.T., Omer, T.C., Sharp, N.Y., 2012. The impact of religion on financial 485–492. reporting irregularities. Account. Rev. 87, 645–673. D’Souza, J., Megginson, W.L., Ullah, B., Wei, Z., 2017. Growth and growth obstacles in Mendoza, R.U., Lim, R.A., Lopez, A.O., 2015. Grease or sand in the wheels of commerce? transition economies: privatized versus de novo private firms. J. Corp. Financ. 42, Firm level evidence on corruption and SMES. J. Int. Dev. 27, 415–439. 422–438. Mensah, Y.M., 2014. An analysis of the effect of culture and religion on perceived Datta, S., Doan, T., Iskandar-Datta, M., 2019. Policy uncertainty and the maturity corruption in a global context. J. Bus. Ethics 121, 255–282. structure of corporate debt. J. Financ. Stab. 44, 100694. M´ eon, P.G., Sekkat, K., 2005. Does corruption grease or sand the wheels of growth? Dixit, A.K., Pindyck, R.S., 2012. Investment Under Uncertainty. Princeton University Public Choice 122, 69–97. Press. M´ eon, P.G., Weill, L., 2010. Is corruption an efficient grease? World Dev. 38, 244–259. Drobetz, W., Ghoul, El, Guedhami, S., Janzen, M, O., 2018. Policy uncertainty, Mertzanis, C., 2019. Family ties, institutions and financing constraints in developing investment, and the cost of capital. J. Financ. Stab. 39, 28–45. countries. J. Bank. Financ. 108, 105650. Edgerton, J., 2012. Agency problems in public firms: evidence from corporate jets in Michaely, R., Roberts, M.R., 2012. Corporate dividend policies: lessons from private leveraged buyouts. J. Financ. 67, 2187–2213. firms. Rev. Financ. Stud. 25, 711–746. Fisman, R., Miguel, E., 2007. Corruption, norms, and legal enforcement: evidence from Nagar, V., Petroni, K., Wolfenzon, D., 2011. Governance problems in closely held diplomatic parking tickets. J. Polit. Econ. 115, 1020–1048. corporations. J. Financ. Quant. Anal. 46, 943–966. Francis, B.B., Hasan, I., Zhu, Y., 2014. Political uncertainty and bank loan contracting. Nagar, V., Schoenfeld, J., Wellman, L., 2019. The effect of economic policy uncertainty J. Empir. Financ. 29, 281–286. on investor information asymmetry and management disclosures. J. Account. Econ. Gao, H., Li, K., 2015. A comparison of CEO pay-performance sensitivity in privately-held 67, 36–57. and public firms. J. Corp. Financ. 35, 370–388. Nguyen, N.H., Phan, H.V., 2017. Policy uncertainty and mergers and acquisitions. Goel, R.K., Ram, R., 2013. Economic uncertainty and corruption: evidence from a large J. Financ. Quant. Anal. 52, 613–644. cross-country data set. Appl. Econ. 45, 3462–3468. Pastor, L., Veronesi, P., 2012. Uncertainty about government policy and stock prices. Gulen, H., Ion, M., 2016. Policy uncertainty and corporate investment. Rev. Financ. Stud. J. Financ. 67, 1219–1264. 29, 523–564. Pastor, L., Veronesi, P., 2013. Political uncertainty and risk premia. J. Financ. Econ. 110, ¨ ¨ Gungoraydinoglu, A., Colak, G., Oztekin, O., 2017. Political environment, financial 520–545. ´ ´ intermediation costs, and financing patterns. J. Corp. Financ. 44, 167–192. Pena Lopez, J.A., Sanchez Santos, J.M., 2014. Does corruption have social roots? The role Hasan, I., Hoi, C.K., Wu, Q., Zhang, H., 2017a. Social capital and debt contracting: of culture and social capital. J. Bus. Ethics 122, 697–708. evidence from bank loans and public bonds. J. Financ. Quant. Anal. 52, 1017–1047. Petersen, M.A., 2009. Estimating standard errors in finance panel data sets: Comparing Hasan, I., Hoi, C.K.S., Wu, Q., Zhang, H., 2017b. Does social capital matter in corporate approaches. Rev. Financ. Stud. 22 (1), 435–480. decisions? Evidence from corporate tax avoidance. J. Account. Res. 55, 629–668. Phan, H.V., Nguyen, N.H., Nguyen, H.T., Hegde, S., 2019. Policy uncertainty and firm Hofstede, G., 1980. Culture’s Consequences: International Differences in Work-related cash holdings. J. Bus. Res. 95, 71–82. Values. Sage Publications,, Beverly Hills. Putnam, R.D., Leonardi, R., Nanetti, R., 1994. Making Democracy Work: Civic Traditions Hogg, M.A., 2007. Uncertainty-identity theory. Adv. Exp. Soc. Psychol. 39, 69–126. in Modern Italy. Hope, O.K., Thomas, W., Vyas, D., 2011. Financial credibility, ownership, and financing Rodriguez, P., Uhlenbruck, K., Eden, L., 2005. Government corruption and the entry constraints in private firms. J. Int. Bus. Stud. 42, 935–957. strategies of multinationals. Acad. Manag. Rev. 30, 383–396. Huntington, S.P., 1968. Political Order in Changing Societies. Yale University Press. Saunders, A., Steffen, S., 2011. The costs of being private: evidence from the loan market. Husted, B.W., 1999. Wealth, culture, and corruption. J. Int. Bus. Stud. 30, 339–360. Rev. Financ. Stud. 24, 4091–4122. International Monetary Fund, 2012. Coping with high debt and sluggish growth. World Scartascini, C., Cruz, C., Keefer, P., 2018. The Database of Political Institutions 2017. Econ. Financ. Surv., 2012, 1–250. Inter-American Dev. Bank. Jens, C.E., 2017. Political uncertainty and investment: causal evidence from U.S. Shang, L., Lin, J.-C., Saffar, W., 2021. Does economic policy uncertainty drive the gubernatorial elections. J. Financ. Econ. 124, 563–579. initiation of corporate lobbying? J. Corp. Financ. 70, 102053. Jiang, F., John, K., Li, C.W., Qian, Y., 2018. Earthly reward to the religious: religiosity Shleifer, A., Vishny, R.W., 1993. Corruption. Q. J. Econ. 108, 599–617. and the costs of public and private debt. J. Financ. Quant. Anal. 53, 2131–2160. Slemrod, J., 2007. Cheating ourselves: the economics of tax evasion. J. Econ. Perspect. Jin, J.Y., Kanagaretnam, K., Lobo, G.J., Mathieu, R., 2017. Social capital and bank 21, 25–48. stability. J. Financ. Stab. 32, 99–114. Staiger, D., Stock, J.H., 1997. Instrumental variables regression with weak instruments. Julio, B., Yook, Y., 2012. Political uncertainty and corporate investment cycles. Econometrica 65, 557. J. Financ. 67, 45–84. Stepan, A., Skach, C., 1993. Constitutional frameworks and democratic consolidation: Kanagaretnam, K., Lee, J., Lim, C.Y., Lobo, G., 2018. Societal trust and corporate tax parliamentarianism versus presidentialism. World Polit. 46, 1–22. avoidance. Rev. Account. Stud. 23, 1588–1628. Stock, J.H., Yogo, M., 2005. Testing for weak instruments in Linear IV regression. In: Kaufmann, D., Kraay, A., Mastruzzi, M., 2009. Governance Matters VIII: Aggregate and Identification and Inference for Econometric Models: Essays in Honor of Thomas Individual Governance Indicators. The World Bank, pp. 1996–2008. Rothenberg. Cambridge University Press, pp. 80–108. Kaufmann, D., Wei, S.-J., 2000. Does “Grease Money” speed up the wheels of commerce? Sundquist, J.L., 1988. Needed: a political theory for the new era of coalition government IMF Work. Pap. 00, 1–21. in the United States. Political Sci. Q. 103, 613. Kelly, B., Pastor, L., Veronesi, P., 2016. The price of political uncertainty: theory and Svensson, J., 2003. Who must pay bribes and how much? Evidence from a cross section evidence from the option market. J. Financ. 71, 2417–2480. of firms. Q. J. Econ. 118, 207–230. Kelly, S.Q., 1993. Divided we govern? A reassessment. Polity 25, 475–484. Treisman, D., 2007. What have we learned about the causes of corruption from ten years Khalil, S., Saffar, W., Trabelsi, S., 2015. Disclosure standards, auditing infrastructure, and of cross-national empirical research? Annu. Rev. Polit. Sci. 10, 211–244. bribery mitigation. J. Bus. Ethics 132, 379–399. Van Den Bos, K., Euwema, M.C., Poortvliet, P.M., Maas, M., 2007. Uncertainty Kim, C., Pantzalis, C., Chul Park, J., 2012. Political geography and stock returns: the management and social issues: uncertainty as an important determinant of reactions value and risk implications of proximity to political power. J. Financ. Econ. 106, to socially deviating people. J. Appl. Soc. Psychol. 37, 1726–1756. 196–228. Wellalage, N.H., Locke, S., Samujh, H., 2019. Corruption, gender and credit constraints: Klassen, K.J., 1997. The impact of inside ownership concentration on the trade-off evidence from South Asian SMEs. J. Bus. Ethics 159, 267–280. between financial and tax reporting. Account. Rev. 72, 455–474. Wooldridge, J.M., 2002. Econometric analysis of Cross Section and Panel Data. MIT Knack, S., Keefer, P., 1997. Does social capital have an economic payoff? A cross-country Press. investigation. Q. J. Econ. 112, 1251–1288. Xie, X., Qi, G., Zhu, K.X., 2019. Corruption and new product innovation: examining firms ’ ethical dilemmas in transition economies. J. Bus. Ethics 160, 107–125. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png SSRN Electronic Journal Unpaywall

Economic Uncertainty and Corruption: Evidence From Public and Private Firms

SSRN Electronic JournalJan 1, 2021

Loading next page...
 
/lp/unpaywall/economic-uncertainty-and-corruption-evidence-from-public-and-private-2elH94EvC8

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher
Unpaywall
ISSN
1556-5068
DOI
10.2139/ssrn.3894468
Publisher site
See Article on Publisher Site

Abstract

E Ec co on no om miic c u un nc ce er rtta aiin ntty y a an nd d c co or rr ru up pttiio on n:: e ev viid de en nc ce e ffr ro om m p pu ub blliic c a an nd d p pr riiv va atte e ffiir rm ms s Mansoor Afzali, Gonal Colak, Mengchuan Fu P Pu ub blliic ca attiio on n d da atte e 01-12-2021 L Liic ce en nc ce e This work is made available under the CC BY 4.0 licence and should only be used in accordance with that licence. For more information on the specific terms, consult the repository record for this item. D Do oc cu um me en ntt V Ve er rs siio on n Published version C Ciitta attiio on n ffo or r tth hiis s w wo or rk k ( (A Am me er riic ca an n P Ps sy yc ch ho ollo og giic ca all A As ss so oc ciia attiio on n 7 7tth h e ed diittiio on n) ) Afzali, M., Colak, G., & Fu, M. (2021). Economic uncertainty and corruption: evidence from public and private firms (Version 2). University of Sussex. https://hdl.handle.net/10779/uos.23487545.v2 P Pu ub blliis sh he ed d iin n Journal of Financial Stability L Liin nk k tto o e ex xtte er rn na all p pu ub blliis sh he er r v ve er rs siio on n https://doi.org/10.1016/j.jfs.2021.100936 C Co op py yr riig gh htt a an nd d r re eu us se e:: This work was downloaded from Sussex Research Open (SRO). This document is made available in line with publisher policy and may differ from the published version. Please cite the published version where possible. Copyright and all moral rights to the version of the paper presented here belong to the individual author(s) and/or other copyright owners unless otherwise stated. For more information on this work, SRO or to report an issue, you can contact the repository administrators at sro@sussex.ac.uk. Discover more of the University’s research at https://sussex.figshare.com/ Journal of Financial Stability 57 (2021) 100936 Contents lists available at ScienceDirect Journal of Financial Stability journal homepage: www.elsevier.com/locate/jfstabil Economic uncertainty and corruption: Evidence from public and private firms a b, * c Mansoor Afzali , Gonül Ҫolak , Mengchuan Fu Hanken School of Economics, Department of Accounting, P.O. Box 479, FI-00101 Helsinki, Finland Hanken School of Economics, Department of Finance and Economics, P.O. Box 479, FI-00101 Helsinki, Finland Fordham University, Gabelli School of Business, 45 Columbus Avenue, Fifth Floor, New York, NY 10023, United States ARTICLE INFO ABSTRACT JEL classification: We study the influence of policy uncertainty on the moral behavior of firms. When facing uncertainty, managers D80 perceive various socioeconomic obstacles as more severe and disruptive to their business. Using data from policy F30 uncertainty spouts in 93 countries, we document that some firms engage in norm-deviant behavior by cheating G38 on taxes and paying more bribes. While private firms prefer to cheat on taxes, public firms choose bribery as a O43 favorite tool to “grease the wheels” during periods of uncertainty. Strong social capital (local trust and religi- P48 osity) breaks this link between uncertainty and corruption. Keywords: Economic policy uncertainty Private firms Corruption Bribery Cheating on taxes Trust Religiosity 1. Introduction (Bonaime et al., 2018; Nguyen and Phan, 2017), and cash holdings (Phan et al., 2019). Political and regulatory authorities generate uncertainty through Most of this research, however, focuses on publicly traded large their decision-making process, and this uncertainty can influence firms ’ corporations. We add to this literature by studying the influence of business operations (Bloom, 2009). Several recent studies have linked policy uncertainty on small firms throughout the world, most of which policy-related economic uncertainty (policy uncertainty) to various are private firms. While we analyze both private and public firms, we managerial decisions by public firms. Julio and Yook (2012), Gulen and try to shed more light on private enterprises since they make up more Ion (2016), and Jens (2017) show that policy uncertainty has a sub- than 99% of the business entities in most countries, and they have in stantial impact on firms ’ investment decisions, and Bhattacharya et al. aggregate four times as many employees, three times the revenues, and (2017) report that it affects firms ’ innovativeness. This uncertainty also twice as many assets as listed firms (Chen et al., 2011). For instance, the affects public corporations’ decisions on external financing (Ashraf and privately held firms in the United States produce 51% of the gross na- Shen, 2019; Chan et al., 2021; Colak et al., 2017; Datta et al., 2019; tional output and employ more than half of the labor force (Nagar et al., Francis et al., 2014; Gungoraydinoglu et al., 2017), acquisition activity 2011). In the European Union, small firms with less than 250 employees, This paper has been circulated also under the name “Resorting to Corruption When Facing Policy Uncertainty: Global Evidence.” The authors would like to thank the editor (Iftekhar Hasan), the two anonymous referees, and the seminar participants at Hanken School of Economics and University of Southern Denmark. * Corresponding author. E-mail addresses: mansoor.afzali@hanken.fi (M. Afzali), gonul.colak@hanken.fi (G. Ҫolak), mfu10@fordham.edu (M. Fu). The link between privately held (non-publicly trading) firms and political uncertainty has been analyzed within specific countries. Amore and Minichilli (2018), for instance, analyze the investment behavior of family firms in Italy that are exposed to economic uncertainty. An et al. (2016) analyze the investments of private vs. state-owned enterprises in China, although the term “private firms ” in the context of China encompasses those listed on the exchange that are not owned by the local or federal government. https://doi.org/10.1016/j.jfs.2021.100936 Received 22 March 2021; Received in revised form 25 August 2021; Accepted 27 August 2021 Available online 8 September 2021 1572-3089/© 2021 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 employ more than 66% of the labor force (Airaksinen et al., 2016). to engage in corruption in order to “grease the wheels” of their busi- While researchers have examined financing and growth constraints in nesses. Similarly, Husted (1999) claim that when uncertainty avoidance private firms (for instance, Beck et al., 2005; D’Souza et al., 2017; Hope (a la Hofstede, 1980) in a country is high, firms engage more in corrupt et al., 2011; Mertzanis, 2019), we do not know how these small and behavior. Rodriguez et al. (2005) note that corrupt economies are often relatively more fragile firms cope with the pressure generated by the characterized by widespread uncertainties. When faced with uncer- rising uncertainty. tainty, managers will rationalize their norm-deviant behavior and will The main obstacle in studying private firms ’ reaction to policy un- even make it part of the corporate culture (Anand et al., 2004; Ashforth certainty is the lack of data on them in many smaller countries and the and Anand, 2003). In light of these findings, we hypothesize that most unavailability of a proper policy uncertainty measure in those countries. firms from around the world will reduce their exposure to economic Using the enterprise-level data from the World Bank Enterprise Survey uncertainty shocks by engaging in corruption. (henceforth WBES), we study how policy uncertainty affects private To test this prediction, we examine two aspects of business activity firms in various countries. This data not only provides detailed infor- that pertain to corruption. These include, cheating on taxes and paying mation on the structure and financials of both private and public firms, bribes. We show that during times of high policy uncertainty, small but the survey also seeks the managers’ opinions on the main obstacles businesses cheat more on taxes and pay more bribes (both measured as (inefficiencies) in the business environment. This survey data is avail- percent of sales). All of these results are statistically significant at the 1% able for around 146,000 firms in 143 countries. level and are robust to the inclusion of several firm-level and country- The unavailability of a local measure of policy uncertainty for many specific characteristics as well as industry, country, and year fixed ef- countries around the world has been a major challenge for researchers. fects. The results are economically significant too; for instance, a one While the news-based index of Baker et al. (2016) is widely used to standard deviation increase in our economic uncertainty measure capture economic policy uncertainty, it is predominantly available for translates to 3.5% less sales reported to the government for tax purposes. the developed or the large developing nations. To address this chal- We also document an interesting divergence between private versus lenge, Ahir et al. (2019) develop a world uncertainty index for 143 public firms ’ reaction to economic uncertainty. Private firms ’ preferred countries for each quarter since 1996Q1. Their method relies on textual form of corruption is cheating on taxes, probably because they find it analysis of the Economist Intelligence Unit’s (EIU) quarterly country easier to fudge their actual sales numbers subject to taxes. Public firms, reports, and it is designed in a similar fashion to Baker et al.’s (2016) on the other hand, are relatively less likely to cheat on taxes and instead economic policy uncertainty index. prefer to cope with uncertainty by paying more bribes. This is a novel We combine this world uncertainty index (WUI) with the WBES data finding for the literatures on corruption and policy uncertainty. for the period of 2002–2018 to get a large sample of 96,769 firm-year To improve identification and to increase the reliability of our observations for small business firms from 93 countries that cover 16 findings, we use instrumental variable (IV) regressions, where election regions within Africa, Americas, Asia, and Europe. Using this sample, we closeness and the proportion of seats held by the largest opposition party assess how economic uncertainty influences the business environment in the legislature serve as instruments for uncertainty. We continue to that firms operate in. Since our data is survey based, our goal is not to find higher levels of corruption after using these proxies and IV re- establish a causal link between uncertainty and business constraints but gressions. Using a carefully balanced sample obtained by applying rather to provide evidence that uncertainty alters managerial perception propensity score matching, where we match on all firm and country- of the severity of the business constraints that are prevalent in a country. level characteristics, produces similar results. To further improve the We find that as policy uncertainty rises, the managerial opinion robustness of our results and to reduce the effects of measurement error (measured by the responses to survey questions) about the severity of in our estimations, we employ four alternative measures of economic business obstacles increases. These obstacles include, a dysfunctional policy uncertainty. First, using the voting outcomes from the national court system, high crime levels (theft and disorder), various forms of elections in each country (as in Julio and Yook, 2012), we construct a public sector corruption, high tax rates, ineffective tax administration, measure (Election closeness) to capture the uncertainty generated around anti-competitive behavior of informal businesses, reduced access to contested elections. Second, we construct the orthogonalized measure of finance, and political instability. policy uncertainty by running a regression of country-level index This relationship between uncertainty and managerial opinion is to (country-specific WUI) on the aggregate global uncertainty (available in be expected because individuals (firm managers) tend to perceive Ahir et al., 2019) and use the residuals from this regression as our un- increased uncertainty as a threat to their livelihood (Hogg, 2007). As certainty measure in the second stage. Third, we calculate stock market studies have shown that uncertainty amplifies people’s reactions to volatility by estimating the standard deviation in returns on the main negative events (Arenas et al., 2006), managers are likely to react more stock index in each country. Fourth, we use the news-based index of negatively to existing market frictions and business obstacles. That is, Baker et al. (2016) for countries with available data on electronic news the unpredictability of future economic and political outcomes induces articles. These four alternative measures do not alter our qualitative an environment of uncertainty and inefficiency. Leff (1964), Acemoglu conclusions. and Verdier (2000), and Mendoza et al. (2015) argue that when eco- Policy uncertainty can directly affect corruption by inducing external nomic agents think that a country’s institutions are ineffective, they tend conditions that incentivize norm-deviant behavior in firms. It can also affect corruption indirectly by changing managerial perception regarding obstacles facing their businesses that in turn can trigger corrupt managerial actions. To test this indirect affect, we perform a See https://www.enterprisesurveys.org/en/data. formal mediation analysis (Baron and Kenny, 1986). Our mediation tests As of January 2020, the measure was available for only 23 countries: show that worsened perception of managers regarding business obsta- Australia, Brazil, Canada, Chile, China, Colombia, France, Germany, Greece, cles explains 4.2% of the relationship between economic uncertainty Hong Kong, India, Ireland, Italy, Japan, Mexico, Netherlands, Russia, and cheating on taxes. The indirect effect of paying bribes explains about Singapore, South Korea, Spain, Sweden, the United Kingdom, and the United 18.5% of the total effect. Thus, economic uncertainty affects corrupt States. behavior both directly and indirectly through the mediating role of In this study, we use the concepts of economic uncertainty and policy un- managerial perception. certainty as close substitute for each other. While conceptually there can be We next explore how cross-country heterogeneity affects the rela- certain differences between these uncertainties, due to lack of separate country- tionship between policy uncertainty and corruption. We first look at the level measures for each type of uncertainty, we work with the WUI, which is constitutional framework of the country. A parliamentary system is referred to by its creators (Ahir et al., 2019) with the generic name of “world uncertainty index.” characterized by simultaneous changes in the control of the 2 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 government’s executive branch and the legislative branch that thereby trust in others and religiosity reduce such corruption. Overall, these increases the freedom of politicians when trying to consolidate de- results are consistent with the idea that uncertainty induces an envi- mocracy (Stepan and Skach, 1993). A presidential system, on the other ronment where norm-deviant behavior prevails, and a country’s social hand, is characterized by a high degree of checks and balances that often capital ameliorates this negative effect. minimize policy fluctuations and imposes constraints on the adoption of new laws and regulations. Julio and Yook (2012) and Gungoraydinoglu 2. Theoretical motivation and hypotheses development et al. (2017) argue that in general, presidential systems have less po- litical uncertainty; that is, parliamentary systems are more likely to Studies like Bernanke (1983), Bloom (2009), and (Pastor and Ver- experience greater policy-related economic uncertainty. We therefore onesi, 2012) have set the theoretical foundations of the literature on the predict the relationship between policy uncertainty and corruption to be consequences of economic policy uncertainty (Pastor and Veronesi, more significant in countries with a parliamentary system of govern- 2012). As predicted by these studies, empirical tests confirm that eco- ment. To test this prediction, we interact our main uncertainty measure nomic policy uncertainty has broad macroeconomic and financial im- (WUI) with an indicator variable for a country with a parliamentary plications (Petersen, 2009; Bernanke, 1983; Bloom et al., 2007; Leahy system of government. We find that firms located in these countries and Whited, 1996). Similarly, reports prepared after the 2008 financial cheat more on taxes. However, we do not find any statistically signifi - crisis show that the period of high economic uncertainty that followed cant differences between the levels of bribes under the two systems. We the recession created business obstacles and impeded growth and re- also explore whether this uncertainty-corruption relationship is influ - covery (International Monetary Fund, 2012). In particular, policy un- enced by the economic conditions in a country. We predict that this certainty lowers investment and delays the implementation of projects relationship will be stronger in low-income regions due to the general (Dixit and Pindyck, 2012; Drobetz et al., 2018), reduces employment lack of oversight and transparency. Consistent with our prediction, growth (Bentolila and Bertola, 1990), boosts interest rates (Ashraf and private firms located in low-income countries have a greater tendency to Shen, 2019), and suppresses asset valuations (Brogaard and Detzel, engage in cheating on taxes and paying bribes when economic uncer- 2015; Kelly et al., 2016; Kim et al., 2012; Pastor and Veronesi, 2012, tainty increases. 2013). Studies have indicated that cultural and sociological norms in the The above findings imply that policy uncertainty can disrupt firms’ country can mitigate norm-deviant behavior. For instance, Kanagar- regular business activities and in turn can create new impediments or etnam et al. (2018) use a sample of firms from 25 countries to show that can exacerbate existing frictions in the economy. Thus, we first analyze societal trust is negatively correlated with corporate tax avoidance. how policy uncertainty affects managers’ perceptions about the efficient Aghion et al. (2010) claim that trust is vital when country-level gover- functioning of the regulatory authorities, the governmental agencies, nance institutions are weak. Similarly, studies have shown that religi- and various other socioeconomic institutions within a country. The osity affects corporate outcomes (Callen and Fang, 2015; Jiang et al., research in different studies indicates that individuals perceive various 2018; McGuire et al., 2012). Therefore, we examine how societal trust types of uncertainties as a threat to themselves (Hogg, 2007). Uncer- and religious norms affect the link between uncertainty and corruption. tainty further intensifies people’s reactions to negative events (Arenas To measure trust and civic norms (social capital), we follow Knack and et al., 2006; Van Den Bos et al., 2007). Thus, managers should react Keefer (1997) and use the World Values Survey (WVS). Specifically, we more negatively to existing economic frictions and business obstacles define social capital as the average of trust in others and norms of civic when the uncertainty about their firms’ future increases. On the basis of cooperation. Our religiosity measure is an indicator variable equal to the above arguments, we formulate our first hypothesis: one if more than 70% of the country’s residents replied “yes” when Hypothesis 1. Policy uncertainty worsens managers’ perception about the asked, “Is religion important in your daily life?” in a Gallop Poll in 2009, severity of the business obstacles prevalent in a country. and zero otherwise. Using these measures, we show that both societal norms and religiosity mitigate the uncertainty-corruption relationship. If business obstacles are perceived to be more severe by the firm We contribute to the literature in several ways. First, we extend the managers during periods of high uncertainty, then these managers literature on policy uncertainty to private firms. Due to lack of data on should take precautionary action to ameliorate the effects of such ob- private firms, most of the literature about policy uncertainty focuses on stacles on their businesses. Put differently, the unpredictability of eco- publicly traded firms. For instance, Nagar et al. (2019) assert that policy nomic and political outcomes can create an uncertain environment in uncertainty contributes to increasing information asymmetry between which firms constantly look for ways to mitigate their exposure to these public firms and their investors. Boutchkova et al. (2012) claim that uncertainties. One such remedy is engaging in corruption and other local policy uncertainty translates into higher systematic (stock market) illegal practices. Husted (1999) and Rodriguez et al. (2005) argue that risk, while global shocks result in higher firm-specific return volatility. firms engage in corruption to reduce the uncertainty in their business Politically motivated local (state level) fiscal shocks also tend to reduce dealings with government officials. Using data from India, Collins et al. the productivity and the output of local public firms (Cohen et al., 2011). (2009) show that corruption can be costly to the society but vital in Furthermore, heightened policy debates slow down the public firm’s reducing regulatory uncertainty for firms, especially when the managers ability to adjust to the optimal capital structure by using stock issuances perceive corruption as increasing their firms’ survival chances. Man- (Colak et al., 2018). We contribute to this line of research by showing agers then rationalize their norm-deviant behaviors, and corruption that policy uncertainty can induce corrupt business practices such as becomes part of the corporate culture (Anand et al., 2004; Ashforth and bribery and tax cheating in both public and private firms. Anand, 2003). We are the first to link policy uncertainty to corruption in private Consistent with this view, we argue that uncertainty increases the firms. While some studies have linked uncertainty to corruption (e.g., unpredictability of future economic and political outcomes that explains Braun and Di Tella, 2004; Goel and Ram, 2013; Treisman, 2007), they why firms engage in shady behavior to reduce their exposure to the largely confine their empirical evidence to cross-country analyses using potentially detrimental consequences of uncertainty. For instance, high macro-level corruption indices. We fill this gap by studying how policy policy uncertainty can induce firms to engage in more aggressive po- uncertainty effects the behavior of small firms. During times of high litical lobbying activities (Shang et al., 2021). Kaufmann and Wei (2000) economic uncertainty, firms face more severe obstacles that impede their business, and they respond by engaging in norm-deviant behavior. While private firms cheat through underreporting sales and paying lower taxes, the public firms prefer to “grease the wheels” through bribery. Further, our results also indicate that societal norms such as 3 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 claim that firms engage in corruption to reduce various inefficiencies. delay their decision to raise new equity capital. However, the rest of the At the country level, there is some evidence that macroeconomic un- private firms proceed with their IPO events and are willing to accept certainty has a positive association with corruption; Braun and Di Tella lower price-to-value ratios for their equity. (2004) and Treisman (2007) find that countries with high variability in In view of the differences in the nature of public and private firms, inflation have higher levels of corruption. Cohen and Malloy (2014) the assumption that they may adopt different norm-deviant behavior show that political favoritism (a form of corruption) is prevalent within when faced with high economic uncertainty is a reasonable one. Typi- US Senate as alumni’s network connections to powerful Senate com- cally, compared with public firms, private firms have less pressure from mittee chairmen help those ex-senators secure additional discretionary the capital market and are therefore more likely to actively reduce taxes earmark funding for local constituents. by deliberately reporting lower financial earnings (Slemrod, 2007). Many studies have pointed to the costs and benefits of corruption Conversely, because public firms need to disclose their activities through when widespread inefficiency prevails and creates serious economic financial reports or discussions with analysts, the perceived benefits of frictions (Huntington, 1968; Meon and Sekkat, 2005; Shleifer and tax evasion may be less than the perceived costs of putting themselves in Vishny, 1993). For instance, corruption can facilitate trade and improve a disadvantaged position (Cloyd et al., 1996). Additionally, due to the efficiency by allowing private sector agents to circumvent cumbersome high concentration of ownership, private firms generally have lower regulations (Bardhan, 1997; Leff, 1964; Mauro, 1995). Furthermore, costs for financial reporting that are associated with reducing taxes and Leff (1964), Acemoglu and Verdier (2000), Meon and Weill (2010), and are more likely to conduct transactions that generate tax savings Mendoza et al. (2015) theorize that when a country’s institutions are (Klassen, 1997). Therefore, we predict that private firms will be more ineffective, corruption is a useful tool to “grease the wheels”, especially willing to cheat through taxation to deal with economic uncertainties. at the small firm level. Although both tax evasion and bribery are corrupt practices, their We build on this literature and argue that a country’s economic penalty functions are different. In the process of tax evasion, the firm uncertainty increases the severity of business constraints for all firms may be held liable for the fraudulent acts of its internal accountant. But (Hypothesis 1). We further predict that these conditions induce an in the process of bribery, individuals may be liable (e.g., imprisonment) environment of inefficiency. To reduce this inefficiency, some managers if involved in any corrupt payment scheme. As the ownership of private can engage in paying bribes and cheating on taxes. This idea is largely firms is more concentrated, their managers will increase their efforts to consistent with the theoretical models of Aguilera and Vadera (2008) in control bribery to avoid responsibility (Clarke and Xu, 2002). The scale which they argue that corruption occurs when there is a context or of bribery can also vary with firm ’s bargaining power (Svensson, 2003). environment that makes such actions feasible and there are factors that Given that public firms and public officials may have a closer relation- encourage agents to act in such ways. ship, bribery by public firms may be more cost effective than that by private firms. Our next hypothesis focuses on such differences between Hypothesis 2. Higher policy uncertainty is associated with more severe public and private firms: corruption in the form of cheating on taxes or paying more bribes by firms. Hypothesis 3. High policy uncertainty induces a different norm-deviant Studies have shown well that policy uncertainty can influence a behavior in private firms than in public firms; private firms prefer cheating manager’s decisions at the firm level. For instance, Julio and Yook through taxes and public firms prefer paying more bribes. (2012), Gulen and Ion (2016), and Jens (2017) show that policy un- certainty influences a firm ’s investment decisions. Political uncertainty The social norm theory argues that social constructs such as trust and can also affect the financing of projects by delaying the decision to go civic norms reduce norm-deviant behavior among all economic agents. public (Colak et al., 2017) or by impeding equity or bond issuance Based on the findings of Putnam et al. (1994), La Porta et al. (1997), and (Colak et al., 2018; Francis et al., 2014; Gungoraydinoglu et al., 2017; Fisman and Miguel (2007), we argue that societal trust is a reliable Chan et al., 2021; Datta et al., 2019). Policy uncertainty is also shown to deterrent of corruption. For instance, Kanagaretnam et al. (2018) show affect technological innovation (Bhattacharya et al., 2017) and cash that societal trust is negatively associated with corporate tax avoidance. holdings (Phan et al., 2019). However, almost everything we know Consistent with Adelopo and Rufai (2020), we argue that trust inhibits about the relationship between policy uncertainty and managerial de- corruption. Furthermore, following the arguments in Aghion et al. cisions is based on the behavior of publicly listed firms that are in the US. (2010), we predict that trust becomes vital when country-level gover- In this study, we focus on private firms across the world and their nance institutions are weak. Similarly, society’s religiosity can reduce managerial decisions under uncertainty. We contrast their behavior to corporate misconduct such as financial reporting fraud (McGuire et al., the behavior of public firms. How exactly small private firms cope with 2012). Mensah (2014) also argues that religion reduces the levels of uncertainty is still unclear. There is some scant evidence that indicates perceived corruption in the country. Thus, we hypothesize that societal private firms react somewhat differently to policy uncertainty than trust and religiosity will break the link between policy uncertainty and public firms. A study by Amore and Minichilli (2018), for instance, fo- corruption for all firms: cuses on a single country (Italy) and shows that family firms are more Hypothesis 4. Societal norms such as trust and religiosity will ameliorate likely to invest during periods of high uncertainty. Colak et al. (2017) the corruptive effects of higher policy uncertainty on managerial behavior in shows that when facing policy uncertainty, about 15% of private firms all firms. Kaufmann and Wei (2000) also provide evidence that when firms pay more bribes, they spend more, not less time, with the governmental authorities. In untabulated results, we find results consistent with this argument. The literature on private firms ’ decision-making is growing. Caballero et al. (1995), Cooper and Haltiwanger (2006), and Asker et al. (2011) claim that Different dimensions of social capital can have different effects on the level private firms ’ investment policies are different than public firms. Michaely and of corruption (Pena Lopez ´ and S´ anchez Santos, 2014). In particular, strong Roberts (2012) show that private firms smooth dividends less than public firms. bonds between individuals can positively relate to corruption. On the other Gao and Li (2015) report that private firms ’ CEO compensation is less sensitive hand, linking and bridging social capital can inhibit individuals or groups from to performance; however, Edgerton (2012) shows that public firms overuse engaging in corruption. Recent evidence, however, shows that both civic norms corporate jets compared to private firms. Financing options, growth constraints, and dense social networks can mitigate norm deviance and reduce transaction and capital structure policies are also different in private firms (Beck et al., costs (Hasan et al., 2017a, 2017b). In our analysis, we measure trust using the 2005; Brav, 2009; D’Souza et al., 2017; Hope et al., 2011; Mertzanis, 2019; civic norm (linking and bridging) component of social capital and expect a Saunders and Steffen, 2011). negative link between trust and corruption. 4 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 3. Research design and sample selection perceptions about the seriousness of the obstacle. In this section, we describe the measures for economic policy un- 3.3. Sample selection certainty, the details of our empirical design, and the sample selection procedure. We also present the summary statistics. We start with the raw firm-level data from the WBES. This initial sample has 104,638 firm-years of which 4897 (or 4.7% of the sample) belong to publicly listed firms. Since in WBES all values of sales are in 3.1. World economic uncertainty index local currency units, we convert these values to US dollars using the exchange rate at the end of the month in which the survey was con- We define economic policy uncertainty using the newly developed ducted. Consistent with the “grease the wheels” hypothesis (Acemoglu world uncertainty index (WUI) of Ahir et al. (2019). The WUI utilizes the and Verdier, 2000; Mendoza et al., 2015), we focus on small and Economist Intelligence Unit’s (EIU) quarterly country reports to capture medium-level firms. We, thus, screen out the very large firms, as well as the economic, political, and financial trends in a country by using a the firms whose sales are missing or negative. These screening only five-step comprehensive process that is developed by experienced ana- affects about 1.9% of our original sample of firms. lysts within each country. Ahir et al. (2019) construct the index for each We then match the data from this survey with the uncertainty index. country by counting the frequency of the word “uncertainty” and its This index is constructed for each quarter, and we therefore take the variants in the EIU reports. To make the index comparable, they divide timing into consideration when matching. Using the date of the inter- the word count by the total number of words used in the report. The view we are able to precisely match the firm data (WBES) to the WUI uncertainty index does well in capturing important global events and is measure using country, year, and quarter. We remove any country positively correlated with the index for economic uncertainty and the with fewer than 50 observations per year (affects only 184 firm-years). volatility in the stock market. To further smooth the idiosyncratic usage Our final sample contains 103,738 firm-year observations from 93 of the word “uncertainty” in a particular quarterly report and to remove countries that cover the period from 2002 to 2018. idiosyncratic spikes in economic uncertainty that may not last long Table 1 provides the sample composition with the number of ob- enough to change managers’ behavior, we use the three-quarter servations for each country, the number of surveys of each country, and weighted moving average of this index as our main uncertainty mea- the mean uncertainty index. The sample is highly inclusive with gross sure. To reduce potential measurement problems related to the WUI, we domestic product (GDP) per capita ranging from $225 in Burundi to use three alternative measures of uncertainty in a series of robustness $27,698 in Greece. The largest number of observations comes from India checks (see Section 5). These alternative measures are available for a (10,322), followed by the Russian Federation (3004), and Chile (2721). smaller set of countries. The mean policy uncertainty index is higher in the Central Asian countries, the Middle East, and Africa. For majority of our sample 3.2. World Bank’s Enterprise Surveys (WBES) countries, the survey is conducted two or more times. We can therefore use country and year fixed effects that should increase the identification The World Bank’s Enterprise Survey (WBES) covers a wide range of strength of our tests. topics related to a country’s business environment, such as access to The survey data from WBES are well populated for the question(s) financing, gender, corruption, infrastructure, innovation, competition, related to bribery and are used to create our left-hand side variable that informality, and performance indicators. The survey asks a series of measures the percentage of sales paid as bribes during a given year. questions and allows business owners and top managers to express their However, the survey data related to firms’ cheating on taxes is available views on the business environment, growth opportunities, financial and for 68 out of 93 countries in our sample. Thus, when analyzing the tax legal barriers, and corruption issues that help to identify the obstacles cheating behavior of firms, our sample size is reduced to 30,032 firm- that hinder the performance and growth of firms. It is considered to have year observations. the most comprehensive firm-level data on emerging markets and The data on macro-level variables for each country are obtained from developing economies from all over the world. While this survey is conducted in several waves, the World Bank conveniently provides combined datasets for several countries for the periods of 2002–2005 According to WBES guidelines, negative sales values indicate that the firm and 2006–2019. Despite some differences, the two datasets are largely 8 did not report their sales during the survey. We also exclude all firms with sales the same and compatible with each other. equal to zero since they are likely to be micro firms with no activity during the The financial accounting studies have used the WBES data before. year. To focus on small firms, we place an upper limit of $100 million in sales. They have used it to study growth (Beck et al., 2005; D’Souza et al., WBES does not provide the exact date of the interview for the 2002–2005 2017), financing constraints (Hope et al., 2011; Mertzanis, 2019; sample. We therefore assume that the interview was conducted in the fourth Wellalage et al., 2019), and product innovation (Lederman, 2010; Xie quarter of the year. In robustness checks, we change this to the first, second, et al., 2019) for firms as well as the association between auditing and third quarters. Our baseline results remain qualitatively unchanged. infrastructure and bribery mitigation (Khalil et al., 2015), recovery of If countries appeared only once in the sample, our results could simply SMEs from a financial crisis (Lawless et al., 2015), and the link between reflect the cross-country differences. For instance, firms in countries with higher uncertainty are also likely to face more obstacles. In our analyses, this investment efficiency and financial reporting quality (Chen et al., 2011). issue is of little concern since we apply country and year fixed effects. We also From the WBES data we identify eight major obstacles (in- control for several other macro-level variables which further reduce the prob- efficiencies) for which all private firms have sufficient data. These ob- ability that we capture mere cross-country differences. stacles are the court system, crime, corruption, tax rate, tax Information on cheating on taxes is available for Albania, Angola, administration, anticompetitive behavior of the informal sector, Argentina, Armenia, Belarus, Benin, Bolivia, Bosnia-Herzegovina, Botswana, financing, and political instability. The severity of an obstacle is Brazil, Bulgaria, Burundi, Cambodia, Chile, Colombia, Croatia, Czech Republic, measured by the survey, and thus it reflects managers’ opinions and Dominican Republic, Ecuador, El Salvador, Georgia, Ghana, Greece, Guatemala, Guinea, Honduras, Hungary, India, Indonesia, Jamaica, Jordan, Kazakhstan, Kenya, Kyrgyzstan, Latvia, Lebanon, Liberia, Lithuania, Madagascar, Malawi, We merge the two datasets on all required identification variables. The Mali, Mauritania, Mexico, Moldova, Mongolia, Namibia, Nicaragua, Panama, industries are defined slightly differently in the two datasets, and so we Paraguay, Peru, Poland, Republic of North Macedonia, Romania, Russian manually adjust the industry names in the 2002–2005 dataset to correspond to Federation, Rwanda, Senegal, Slovakia, Slovenia, South Africa, Tajikistan, the 2006–2019 dataset. After combining the datasets, we remove the duplicates Tanzania, Turkey, Uganda, Ukraine, Uruguay, Uzbekistan, Vietnam, and based on the unique identification number and year of the firm. Zambia. 5 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 1 Sample composition. Country Obs. No. of surveys Uncertainty index Country Obs. No. of surveys Uncertainty index Afghanistan 425 2 0.15 Liberia 258 2 0.06 Albania 609 3 0.07 Lithuania 693 4 0.04 Angola 647 2 0.00 Madagascar 525 2 0.05 Argentina 2654 3 0.15 Malawi 584 3 0.08 Armenia 711 3 0.02 Malaysia 708 1 0.08 Azerbaijan 402 2 0.00 Mali 930 4 0.08 Bangladesh 2537 2 0.07 Mauritania 327 2 0.00 Belarus 986 4 0.02 Mexico 2646 2 0.09 Benin 380 3 0.07 Moldova 705 4 0.09 Bolivia 980 3 0.07 Mongolia 602 3 0.08 Bosnia and Herzegovina 642 3 0.07 Morocco 239 1 0.06 Botswana 541 2 0.03 Mozambique 1031 2 0.12 Brazil 2575 2 0.09 Namibia 604 2 0.02 Bulgaria 1312 4 0.07 Nepal 682 2 0.05 Burundi 403 2 0.04 Nicaragua 1433 4 0.11 Cambodia 506 2 0.08 Niger 175 2 0.08 Cameroon 576 2 0.10 Nigeria 1953 1 0.08 Chad 266 2 0.06 Pakistan 532 1 0.02 Chile 2721 3 0.04 Panama 638 2 0.01 Colombia 2619 3 0.06 Paraguay 1077 3 0.06 Croatia 889 3 0.05 Peru 2309 3 0.06 Czech Republic 620 3 0.04 Philippines 1763 2 0.07 Cote ˆ d’Ivoire 580 2 0.08 Poland 1230 4 0.08 Dominican Republic 626 3 0.03 Macedonia 660 3 0.03 Ecuador 1558 4 0.10 Romania 1031 3 0.08 Egypt 2441 2 0.08 Russian Federation 3004 3 0.05 El Salvador 1945 4 0.07 Rwanda 361 2 0.05 Eritrea 117 1 0.00 Senegal 1047 3 0.04 Ethiopia 857 2 0.00 Slovakia 454 3 0.04 Georgia 534 3 0.04 Slovenia 661 3 0.06 Ghana 942 2 0.07 South Africa 1502 2 0.10 Greece 959 2 0.05 Sri Lanka 460 1 0.04 Guatemala 1602 4 0.09 Tajikistan 521 3 0.03 Guinea 275 2 0.22 Tanzania 990 3 0.03 Honduras 1277 4 0.04 Thailand 768 1 0.15 Hungary 830 3 0.09 Togo 224 2 0.09 India 10,322 2 0.04 Turkey 2606 4 0.10 Indonesia 2696 3 0.04 Uganda 1023 2 0.02 Jamaica 305 2 0.05 Ukraine 1368 3 0.07 Jordan 901 2 0.01 Uruguay 1179 3 0.05 Kazakhstan 1054 3 0.01 Uzbekistan 846 4 0.07 Kenya 2047 3 0.13 Venezuela 173 1 0.03 Kyrgyzstan 580 4 0.08 Viet Nam 2138 3 0.04 Lao PDR 642 3 0.01 Yemen 456 2 0.02 Latvia 570 3 0.05 Zambia 1219 3 0.10 Lebanon 493 2 0.19 Zimbabwe 1123 2 0.22 Lesotho 126 1 0.08 This table presents the sample composition with number of observations, years of survey, and the average uncertainty index for each country. Uncertainty is the country-level World Uncertainty Index (WUI) developed by Ahir et al. (2019) using the country reports prepared by Economist Intelligence Unit (EIU). the World Bank. The country-level governance data is from Kaufmann where Obstacle indicates the perceived severity of one of the eight ob- et al. (2009). These data consist of several hundred individual variables stacles facing private firms. Management’s responses range from zero that measure six different governance categories: political stability, (not an obstacle) to four (a very severe obstacle). Uncertainty is the government effectiveness, regulatory quality, law enforcement, and country-level measure of policy uncertainty (WUI). Subscripts i, j, and t corruption as well as the extent to which a country’s citizens are able to indicate firm, country, and time. Since Obstacle is a count variable be- participate in selecting their government. We follow Kaufmann et al. tween zero and four, we run count regressions that assume a Poisson (2009) and Beltratti and Stulz (2012) and create a composite index (by distribution. Assuming a negative binomial distribution or implement- extracting the principal component) of these six variables for each ing a pooled OLS does not alter our qualitative conclusions. country. Following other studies (e.g., Beck et al., 2005), we add the firm ’s size (Size) that is defined as the natural logarithm of total sales and its 3.4. Baseline model age (Age) that is defined as the natural logarithm of one plus the dif- ference between the survey year and the year of incorporation; and our To assess the relationship between the severity of business obstacles control variables are an indicator for firms that export more than 10% of and economic uncertainty, we estimate the following model: their sales (Exporter), indicators for when the government (Government) or a foreign entity (Foreign) own more than 10% of the firm, gross do- Obstacle = α+ β Uncertainty + β Size + β Age + β Exporter i,j,t i,j,t 1 j,t 2 3 i,j,t 4 i,j,t mestic product per capita (GDP) and its growth rate (ΔGDP), and + β Government + β Foreign + β GDP + β ∆GDP + β Inflation i,j,t j,t j,t 5 6 i,j,t 7 8 9 j,t country-level inflation (Inflation ). Additionally, we control for ∑ ∑ ∑ + β Country Governance + Industry+ Year+ Country+ ε j,t i,j,t country-level governance as in Beltratti and Stulz (2012) by using the k t j first principal component of six indicators of governance in the country (1) (Country Governance). Following Hope et al. (2011), we also add 6 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 industry, year, and country fixed effects. The standard errors are the composition of the legislative branches in each country in a given heteroscedasticity-robust and clustered at the industry level. All vari- year from the Database of Political Institutions. We then divide the total ables are defined in Table 13 of the Appendix. seats of the largest opposition party by the total seats in the house to To examine how economic uncertainty affects managerial behavior obtain the proportion of seats held by the largest opposition party. The and corruption, we look at two measures. First, we calculate the fraction higher the proportion of opposition seats, the more likely is the gridlock. of sales not reported to the tax authorities (Cheating on taxes). The survey To formally test the quality of our instruments, we apply Stock and specifically asks the interviewee what percentage of sales were reported Yogo’s (2005) weak instrument test. We also report the F-statistics from for taxation. Although the variable is not available for all countries, it is the test of under-identification (instruments are irrelevant) and follow well populated for vast majority of the countries (68 out of 93). This is a the instrument assessment guidelines proposed in Staiger and Stock direct measure of corruption as firms are required to report all sales for (1997). Formally, our instrumental variable (IV) regression is summa- tax purposes. Our second measure, Bribes, is the fraction of sales paid as rized as: “unofficial payments” to the government authorities. This measure Uncertainty = δ + δ Instruments for Corrupt + δ Controls + μ 0 1 2 i,j,t j,t i,j,t i,j,t captures corrupt behavior on the part of private firms as well as the (3a) government authorities. Following these definitions, we modify Eq. (1) as follows: Corrupt = γ + γ Uncertainty + γ Controls + ω (3b) i,j,t 0 1 i,j,t 2 i,j,t i,j,t Corrupt = α+ β Uncertainty + β Size + β Age + β Exporter i,j,t i,j,t 1 j,t 2 3 i,j,t 4 i,j,t where the Instrument for Corrupt vector consists of Election closeness and + β Government + β Foreign + β GDP + β ∆GDP + β Inflation i,j,t j,t j,t 5 6 i,j,t 7 8 9 j,t ∑ ∑ ∑ Opposition seats, and the Controls vector consists of all the control vari- + β Governance + Industry+ Year+ Country+ ε (2) 10 j,t i,j,t ables from Eq. (2). All the other specifications in Eq. (3b) are as in Eq. k t j (2). Corrupt is either Cheating on taxes or Bribes and other specifications are as in Eq. (1). 3.5. Summary statistics Some studies argue that endogeneity problem (in the form of simultaneity bias) may arise when an economic outcome is regressed on Panel A of Table 2 provides the summary statistics for our pooled a news-based measure of policy uncertainty (see the discussion in Gulen public and private sample. The first heading of the panel displays the and Ion, 2016). In light of this concern, we implement an instrumental opinion of business managers on the severity of the obstacles they face. variable (IV) regression in which we use the closeness of national elec- The responses range from zero (not an obstacle) to four (very severe tions (Election closeness) and the proportion of seats in the national obstacle). On average, managers identify obstacles as less severe when parliament held by the largest opposition party (Opposition seats) as in- the average response for all obstacles is below two (moderate obstacle). struments for Uncertainty in the first stage. The average and median for tax rate indicates that it is a more severe For our first instrument, Election closeness, we rely on the findings concern for firms than other obstacles. The summary statistics for size in Julio and Yook (2012), Boutchkova et al. (2012), and Gungor- indicate that the average private firm in the sample has sales of $10.67 aydinoglu et al. (2017) who claim that when a country’s national elec- million; however, the median value ($149.03 thousand) indicates pos- tions are very close, the policy uncertainty spikes. Thus, this instrument itive sample skewness. The sample also has some micro firms with sales should be a reliable as our Uncertainty variable. We obtain data on all in the 25th percentile at approximately $558,000. The mean (median) national elections from the Database of Political Institutions (Scartascini age is approximately 19 (14) years. Around 16.1% of the firms are ex- et al., 2018), and then we supplement it with a manual data verification porters while only 2.2% have over 10% government ownership. Simi- procedure that closely follows the guidelines in Gungoraydinoglu et al. larly, around 10.5% of the firms in the sample have some foreign (2017). Our Election closeness variable uses the distribution of party ownership. Focusing on country characteristics; the mean (median) votes in the national elections to capture the uncertainty created due to uncertainty index is 0.067 (0.057) that roughly translates to about 6.7% competitive and contested elections. Jens (2017) argues that election (5.7%) word counts for “uncertainty” (or its variants) for each 1000 closeness can be attributed to poor economic performance by the pre- words in the quarterly EIU reports. The mean (median) GDP is around vious administration and that it is indicative of higher economic un- $5023.1 ($2979.0) that indicates considerable variation in the economy certainty in the pre-election period. As such, this variable captures the of the sample countries. The mean and median GDP growth are around uncertainty leading up to the election date. 5%. The mean (median) inflation in the sample countries is 7.1% (5.3%). Our second instrument (Opposition seats), on the other hand, captures Average governance in the countries is on the lower side which is ex- the post-election uncertainty prevalent in a country. Our motivation for pected given that our sample is mostly comprised of underdeveloped or its use comes from the divided government hypothesis (Cutler, 1988; developing countries with weaker governance mechanisms. Kelly, 1993; Sundquist, 1988) which argues that the legislation is less Further, the norm-deviant behavior of the firms is presented sepa- likely to be enacted if the executive branch’s party does not also hold the rately for the pooled sample (bottom of Panel A), for the private firms ’ majority in the legislative branches of the government. Since divided sample (Panel B), and for the public firms ’ sample (Panel C). At least governments are likely to give rise to disagreements between the legis- 25% of firms who respond to the tax question engage in cheating (75th lature and the executive branches, we predict that a larger proportion of percentile value is non-zero) and, on average, they underreport 17.9% of opposition seats in the legislative branches will be positively related their sales to the authorities. At the mean annual sales level of $10.67 with policy uncertainty. Arguably, such political gridlock can affect our million, this number corresponds to an average of $2.34 million unre- dependent variable, Corrupt, only by increasing the uncertainty (Un- ported sales per firm. On average, the firms pay 1.4% of their sales as certainty) which means that the exclusivity criterion for this instrument bribes to government officials that corresponds to an average illegal should also be satisfied. To calculate Opposition seats, we obtain data on payment of about $149,352 per firm per year. 4. Empirical results Note that using several instruments for one endogenous variable is an 4.1. Univariate results acceptable practice in instrumental variable regressions, because it creates over-identified conditions. Even if one of the instruments is deemed irrelevant, Table 3 displays the univariate tests for private and public firms (in the remaining instruments are sufficient for proper identification (please refer to the discussions in section 5.1.2 of Wooldridge, 2002). Panels A and B, respectively). We define high (low) economic 7 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 2 Descriptive statistics. Countries Obs. Mean Std. dev. P25 Median P75 Panel A: Pooled summary statistics Business obstacles Dysfunctional court system 93 103,738 0.921 1.222 0.000 0.000 2.000 Crime, theft, disorder 93 103,738 1.074 1.276 0.000 1.000 2.000 Prevalent corruption 93 103,738 1.566 1.486 0.000 1.000 3.000 High tax rates 93 103,738 1.655 1.369 0.000 2.000 3.000 Ineffective tax administration 93 103,738 1.320 1.292 0.000 1.000 2.000 Anti-competitive practices 93 103,738 1.372 1.368 0.000 1.000 2.000 Reduced access to finance 93 103,738 1.376 1.354 0.000 1.000 2.000 Political instability 93 103,738 1.462 1.449 0.000 1.000 3.000 Enterprise characteristics Size (thousands of USD) 93 103,738 10,668.250 58,333.986 5.579 149.031 1319.697 Age 93 103,738 19.461 17.429 8.000 14.000 24.000 Exporter 93 103,738 0.161 0.368 0.000 0.000 0.000 Government 93 103,738 0.022 0.146 0.000 0.000 0.000 Foreign 93 103,738 0.105 0.307 0.000 0.000 0.000 Country characteristics Uncertainty 93 103,738 0.067 0.055 0.030 0.057 0.088 GDP 93 103,738 5023.057 4785.750 1369.184 2979.005 8573.708 ΔGDP 93 103,738 4.969 3.613 3.050 5.118 7.410 Inflation 93 103,738 7.128 8.060 3.332 5.349 8.778 Governance 93 103,738 -0.754 1.362 -1.682 -0.711 -0.253 Civic norms 59 82,498 0.243 0.112 0.180 0.220 0.282 Religion 90 102,893 0.774 0.419 1.000 1.000 1.000 Norm-deviant behavior Cheating on taxes 68 30,032 0.219 0.311 0.000 0.000 0.400 Bribes 93 72,704 0.014 0.057 0.000 0.000 0.000 Panel B: Summary statistics for private firms Size (thousands of USD) 93 98,865 9970.044 5,6691.797 5.020 140.234 1197.605 Cheating on taxes 68 29,024 0.221 0.311 0.000 0.000 0.400 Bribes 93 69,525 0.014 0.056 0.000 0.000 0.000 Panel C: Summary statistics for public firms Size (thousands of USD) 93 4873 24,833.675 83,821.550 38.687 797.575 7619.794 Cheating on taxes 45 1008 0.163 0.288 0.000 0.000 0.200 Bribes 88 3179 0.018 0.078 0.000 0.000 0.000 This table presents the descriptive statistics for the variables used in this study. Panel A presents the results for our main sample (pooled private and public firms). Panel B (Panel C) presents the summary statistics from the private (public) firms ’ samples; included are a few statistics for only a few key variables that are referred in the text. Uncertainty is the World Uncertainty Index (WUI) developed by Ahir et al. (2019) using the quarterly reports prepared by Economist Intelligence Unit (EIU) about each country. All variables pertaining to business obstacles indicate the severity of an obstacle and range between zero and four, where zero indicates no obstacle and four indicates very severe obstacle. All other variables are defined in Appendix in Table 13. uncertainty as countries belonging to the top (bottom) quintiles of cheating through taxes during high uncertainty periods (7% versus economic uncertainty in a given year. Panel A indicates that private 17.3% of the sales go underreported). However, these public firms firms ’ managers identify all the obstacles as more severe when economic substantially increase the amount of bribes they pay (3.9% versus 1% of uncertainty is higher. Public firms (Panel B) also report a worsening the sales). These results are broadly consistent with our first three hy- business environment due to uncertainty except for tax rates. Thus, both potheses: while both private and public firms accelerate their corrupt private and public firms perceive uncertainty as the cause of worsening behavior when faced with uncertainty, public firms focus more on business obstacles. bribery than cheating on taxes. However, in both panels there are considerable firm- and country- level differences between firms in the low versus high uncertainty sub- 4.2. Policy uncertainty and obstacles facing private firms samples. Therefore, we cannot reach a conclusion based on these results that economic uncertainty really increases business obstacles. It is To corroborate the findings from our univariate analysis, we conduct possible that the firms just perceive it as such. Nevertheless, as the various multivariate tests. Using our pooled sample of 103,738 firm- variable under the norm-deviant headings show, firms cheat more on years, we estimate Eq. (1) to examine the effects of economic uncer- taxes and pay more bribes during high economic uncertainty. Regardless tainty on the obstacles facing all firms. After controlling for both firm of whether the increase in obstacles is real or perceived, both the private and country-level variables, our univariate results persist. The results and the public firms react with increased corruption. Specifically, during reported in Panels A and B of Table 4 show that the coefficient for the the high uncertainty periods, the incremental underreporting of sales by main uncertainty measure is positive and statistically significant in all private firms is about 2.6% of total sales that if evaluated at the mean specifications. All the coefficients from the Poisson regression are sta- sales of $9.97 million (from Table 2), indicates that roughly $259,220 tistically significant at less than the 1% level. The estimates are also additional sales went underreported to the government authorities on economically significant. In Panel B, a one standard deviation increase top of the mean underreported sales of around $2.20 million (=22.1% * in economic uncertainty, ceteris paribus, results in a 9.6% $9.97; from Table 2). Similarly, during high uncertainty periods, the (1.626 × 0.055 ÷ 1.074) increase in the severity of crime, theft, and private firms pay 0.4% more in bribes than they do during low uncer- disorder and a 4.1% (0.545 × 0.055 ÷ 1.372) increase in the severity of tainty periods (1.8% vs. 1.4%). For most small private firms from around anti-competitive practices. These results are in line with our prediction the world, these numbers are not economically negligible. For public in Hypothesis 1 – economic uncertainty worsens managers’ perception firms, on the other hand, these univariate results are quite different. As about the severity of the business obstacles prevalent in a country. Panel B of Table 3 shows, public firms actually reduce the amount of Further, the reported severities of the obstacles increase as the size of 8 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 3 Univariate results. Countries High Low Difference t-statistic Uncertainty Uncertainty Panel A: Univariate results for private firms Business obstacles Dysfunctional court system 93 0.954 0.786 0.168*** 13.73 Crime, theft, disorder 93 1.269 0.980 0.289*** 22.13 Prevalent corruption 93 1.651 1.412 0.239*** 15.93 High tax rates 93 1.540 1.599 -0.059*** -4.25 Ineffective tax administration 93 1.278 1.222 0.056*** 4.33 Anti-competitive practices 93 1.472 1.293 0.179*** 12.94 Reduced access to finance 93 1.328 1.376 -0.048*** -3.48 Political instability 93 1.649 1.198 0.451*** 31.07 Enterprise characteristics Size (thousands of USD) 93 7363.621 18,865.655 -11502.034*** -15.89 Age 93 20.344 17.805 2.539*** 14.67 Exporter 93 0.166 0.129 0.037*** 10.31 Government 93 0.016 0.011 0.005*** 3.99 Foreign 93 0.117 0.099 0.018*** 5.90 Country characteristics Uncertainty 93 0.153 0.019 0.135*** 323.19 GDP 93 5018.214 4710.491 307.723*** 6.35 ΔGDP 93 4.280 5.668 -1.389*** -41.56 Inflation 93 8.052 7.469 0.583*** 6.14 Governance 93 -1.001 -0.801 -0.200*** -13.82 Civic norms 59 0.199 0.254 -0.055*** -46.12 Religion 90 0.881 0.770 0.111*** 28.48 Norm-deviant behavior Cheating on taxes 68 0.251 0.225 0.026*** 4.55 Bribes 93 0.018 0.014 0.004*** 5.80 Panel B: Univariate results for public firms Business obstacles Dysfunctional court system 91 1.256 1.030 0.226*** 3.88 Crime, theft, disorder 91 1.362 1.128 0.234*** 4.01 Prevalent corruption 91 1.814 1.558 0.256*** 3.84 High tax rates 91 1.715 1.742 -0.026 -0.44 Ineffective tax administration 91 1.505 1.256 0.250*** 4.46 Anti-competitive practices 91 1.597 1.328 0.269*** 4.34 Reduced access to finance 91 1.444 1.312 0.132** 2.22 Political instability 91 1.820 1.411 0.409*** 6.49 Enterprise characteristics Size (thousands of USD) 91 19,744.426 35,164.440 -15420.015*** -3.51 Age 91 29.452 27.095 2.357** 2.18 Exporter 91 0.272 0.186 0.085*** 4.51 Government 91 0.238 0.190 0.048*** 2.58 Foreign 91 0.256 0.176 0.079*** 4.27 Country characteristics Uncertainty 91 0.142 0.018 0.124*** 68.91 GDP 91 4377.387 5494.168 -1116.782*** -5.72 ΔGDP 91 3.755 5.393 -1.638*** -10.50 Inflation 91 6.286 6.442 -0.156 -0.42 Governance 91 -1.171 -0.918 -0.253*** -3.95 Civic norms 59 0.191 0.240 -0.049*** -15.73 Religion 91 0.839 0.615 0.223*** 11.06 Norm-deviant behavior Cheating on taxes 45 0.070 0.173 -0.103*** -3.91 Bribes 88 0.039 0.010 0.029*** 5.73 This table presents the univariate analysis based on high and low economic uncertainty in the country. Results for private and public firms are presented separately in Panels A and B, respectively. Uncertainty is the World Uncertainty Index (WUI) developed by Ahir et al. (2019), where high (low) indicates top (bottom) quintile of uncertainty in a given year. All variables pertaining to business obstacles indicate the severity of an obstacle and range between zero and four, where zero indicates no obstacle and four indicates very severe obstacle. All other variables are defined in Appendix in Table 13. the firm increases, except for access to financing. These two results are 4.3. Policy uncertainty and corruption in firms understandable since firms are more likely to encounter challenges as the size of their business grows and are more likely to face financing We next estimate Eq. (2) with our two main dependent variables to issues if they are young and small. Beck et al. (2005) also find this size capture corruption by using a pooled sample of private and public firms. effect in relationship to financing. Among other variables, foreign Panels A and B in Table 5 provide these results. The regressions in panel ownership reduces many of the constraints. This reduction can be A indicate that during high economic uncertainty, all firms cheat more attributed to the potentially better management and access to resources on taxes and pay more bribes. Both of these results are statistically that foreign ownership brings. The coefficient for GDP is consistently significant at the 1% level. The results are also economically significant. negative and highly significant in all eight regression results. This co- For instance, a one standard deviation increase in economic uncertainty efficient means that firms located in low-income countries face more results in about 3.01% (0.548 × 0.055) less sales reported to the gov- severe obstacles than those in middle or high-income countries. ernment authorities for tax purposes. Similarly, a one standard deviation 9 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 4 Economic uncertainty and managerial perception. Panel A: Poisson regression results Dep.variable Managerial perception of severity of business obstacles Obstacle Dysfunctional Crime, theft, Prevalent High tax rates Ineffective tax Anti-competitive Reduced access to Political court system disorder corruption admin. practices finance instability (1) (2) (3) (4) (5) (6) (7) (8) Uncertainty 0.638*** 1.704*** 0.826*** 0.579*** 0.535*** 0.566*** 0.589*** 0.727*** (4.26) (12.99) (9.60) (4.34) (3.65) (5.52) (6.28) (5.92) Size 0.039*** 0.027*** 0.026*** 0.028*** 0.028*** 0.004** -0.002 0.024*** (11.94) (6.35) (11.55) (12.26) (14.60) (2.02) ( 0.75) (11.23) Age 0.035*** -0.007 0.001 0.003 0.004 0.023*** -0.052*** 0.013*** (5.58) ( 1.20) (0.18) (0.67) (0.70) (3.99) ( 8.79) (2.82) Exporter 0.064*** -0.066*** -0.013 -0.032*** 0.034*** -0.142*** -0.049*** 0.007 (4.70) ( 3.60) ( 0.98) ( 3.25) (3.32) ( 7.10) ( 2.66) (0.61) Government -0.086* -0.094*** -0.140*** -0.158*** -0.180*** -0.145*** 0.015 -0.057** ( 1.82) ( 3.68) ( 3.06) ( 5.72) ( 5.08) ( 3.53) (0.61) ( 2.35) Foreign 0.003 -0.056*** -0.038*** -0.078*** -0.039*** -0.139*** -0.238*** -0.009 (0.35) ( 3.76) ( 4.61) ( 11.51) ( 3.49) ( 13.38) ( 17.70) ( 1.08) GDP -1.477*** -1.074*** -1.176*** -1.361*** -1.551*** -0.238* -0.760*** -0.997*** ( 9.74) ( 7.33) ( 12.70) ( 10.48) ( 12.41) ( 1.95) ( 3.98) ( 8.03) ΔGDP 0.008* 0.004 0.007** 0.009* 0.011*** 0.006** 0.003 -0.005** (1.95) (0.79) (2.33) (1.84) (2.76) (2.05) (0.92) ( 2.14) Inflation 0.002 -0.000 0.004* 0.004** 0.007*** -0.000 0.001 0.005*** (0.91) ( 0.20) (1.80) (2.26) (5.08) ( 0.29) (0.53) (3.09) Country -0.057 0.079* -0.002 0.151*** 0.117*** 0.151*** 0.182*** -0.075 Governance ( 1.12) (1.80) ( 0.05) (5.65) (4.83) (4.02) (4.01) ( 1.48) Country fixed Yes Yes Yes Yes Yes Yes Yes Yes effects Industry fixed Yes Yes Yes Yes Yes Yes Yes Yes effects Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Pseudo R 0.128 0.119 0.123 0.105 0.104 0.090 0.084 0.160 Observations 103,738 103,738 103,738 103,738 103,738 103,738 103,738 103,738 Panel B: Pooled OLS regression results Uncertainty 0.646*** 1.626*** 0.627** 0.451** 0.328** 0.545*** 0.373* 0.836*** (3.55) (12.83) (2.67) (2.42) (2.25) (3.36) (1.87) (3.83) Country fixed Yes Yes Yes Yes Yes Yes Yes Yes effects Industry fixed Yes Yes Yes Yes Yes Yes Yes Yes effects Year fixed Yes Yes Yes Yes Yes Yes Yes Yes effects Adjusted R 0.199 0.228 0.259 0.267 0.224 0.195 0.189 0.348 Observations 103,738 103,738 103,738 103,738 103,738 103,738 103,738 103,738 This table presents the results from regressing the severity of obstacles faced by all enterprises (private and public) on the level of economic uncertainty in the country and other explanatory variables. Panel A presents Poisson regression results while Panel B displays OLS regression results. Uncertainty is the World Uncertainty Index (WUI) developed by Ahir et al. (2019). All variables pertaining to business obstacles indicate the interviewee’s opinion about the severity of an obstacle and range between zero and four, where zero indicates no obstacle and four indicates very severe obstacle. Country, industry, and year fixed effects are included. All other variables are defined in Appendix in Table 13. The t-statistics based on industry clustered robust standard errors are shown in parentheses. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively. increase in economic uncertainty translates to about 0.17% with the idea that younger and smaller firms are riskier. Country-level (0.031 × 0.055) increase in bribe payments as a percentage of sales. inflation also induces cheating on taxes and paying more bribes. While Overall, these findings are consistent with Hypothesis 2 in that the small governance can mitigate bribery problem, it contributes to cheating on firms behave more corruptly when there is higher economic taxes. One explanation for this cheating could be the higher tax rate in uncertainty. these countries that could induce an environment where firms are more To enhance the robustness of our results with regards to standard likely to avoid taxes (Kanagaretnam et al., 2018). error clustering, in panel B of Table 5 we apply both country-industry Columns (3) and (4) of Table 5 provide the results from Tobit re- (interacted) fixed effects and cluster the standard errors at the same gressions. Since a majority of the firms do not cheat on taxes or pay level as suggested by Petersen (2009). We continue to find a positive bribes (median values are zero; see Table 2), a Tobit regression with left relationship between firms ’ tendency to engage in corrupt behavior and censoring at zero is an appropriate estimation method in our context. economic uncertainty. The Tobit results are similar to those reported using pooled OLSs. Among the control variables, size, age, GDP, and GDP growth Columns (5) and (6) display the results from propensity score negatively affect corrupt behavior in private firms. This is consistent matching. The treated firms are the ones exposed to high uncertainty. We define high (low) economic uncertainty as countries in the top (bottom) quintile of economic uncertainty in a given year. We imple- ment the matching by first matching treated and control firms based on We report heteroscedasticity-robust standard errors clustered at the in- all characteristics in column (1) and by applying a caliper of 0.05. This dustry level in all our specifications. Following Lederman (2010) and boot- process results in an equal sample of firms belonging to high (treated strapping the standard errors does not alter our results. 10 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 5 Economic uncertainty and corruption. Panel A: Industry, country, and year fixed effects with industry clustered standard errors. Estimation model Pooled OLS Tobit with left censoring Propensity score-matching Dependent variable Cheating on taxes Bribes Cheating on taxes Bribes Cheating on taxes Bribes (1) (2) (3) (4) (5) (6) Uncertainty 0.548*** 0.031*** 0.747*** 0.121*** 7.992** 0.127*** (4.84) (5.62) (4.68) (3.93) (2.33) (2.96) Size -0.012*** -0.001*** -0.026*** -0.003*** -0.037*** -0.002*** ( 7.98) ( 8.41) ( 8.64) ( 4.49) ( 11.87) ( 2.88) Age -0.011*** -0.001 -0.020*** -0.001 -0.001 -0.000 ( 4.65) ( 1.51) ( 4.25) ( 1.27) ( 0.09) ( 0.13) Exporter -0.005 0.001 -0.020* 0.009*** -0.012 0.012*** ( 0.83) (1.29) ( 1.87) (2.69) ( 0.72) (3.13) Government -0.020** 0.005** -0.146*** -0.000 -0.150*** -0.005 ( 2.17) (2.24) ( 5.57) ( 0.03) ( 3.15) ( 0.51) Foreign -0.022*** 0.001 -0.069*** -0.002 -0.072*** 0.001 ( 4.68) (1.04) ( 6.22) ( 0.57) ( 3.26) (0.15) GDP -0.354 -0.009 -0.401 -0.005 -19.148** -0.017 ( 0.64) ( 1.49) ( 0.45) ( 0.14) ( 2.42) ( 0.69) ΔGDP -0.029** -0.001*** -0.055** -0.006*** 0.047** -0.002 ( 2.13) ( 6.56) ( 2.31) ( 7.96) (2.31) ( 1.42) Inflation 0.007 0.000* 0.018*** 0.001* -0.174* 0.001* (1.41) (1.92) (3.18) (1.68) ( 1.69) (1.88) Country Governance 0.265*** -0.001 0.317*** -0.011 4.575** 0.032*** (5.46) ( 0.56) (4.52) ( 1.52) (2.04) (2.72) Country fixed effects Yes Yes Yes Yes Yes Yes Industry fixed effects Yes Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Yes Adjusted/Pseudo R 0.209 0.070 0.150 0.393 0.171 0.440 Observations 30,032 72,704 30,032 72,704 10,704 25,922 Panel B: Industry-country and year fixed effects with industry-country clustered standard errors. Uncertainty 0.685*** 0.037** 0.981*** 0.126** 4.869*** 0.147** (3.75) (2.50) (3.30) (2.07) (3.66) (2.22) Size -0.013*** -0.001*** -0.029*** -0.003*** -0.017*** -0.002*** ( 10.10) ( 6.38) ( 11.19) ( 4.26) ( 7.07) ( 2.69) Age -0.009*** -0.000 -0.016*** -0.001 -0.003 0.000 ( 2.65) ( 1.43) ( 2.66) ( 0.95) ( 0.49) (0.09) Exporter -0.003 0.002** -0.019 0.011*** -0.001 0.014*** ( 0.60) (2.18) ( 1.56) (3.51) ( 0.10) (3.04) Government -0.011 0.005** -0.130*** -0.001 -0.020 0.000 ( 1.24) (2.09) ( 4.90) ( 0.16) ( 1.31) (0.04) Foreign -0.015** 0.001 -0.056*** -0.001 -0.023** 0.001 ( 2.40) (1.43) ( 4.02) ( 0.15) ( 2.05) (0.27) GDP 1.526* -0.014 2.092* -0.016 -15.210*** -0.044 (1.70) ( 1.15) (1.69) ( 0.24) ( 4.18) ( 0.57) ΔGDP -0.034* -0.002*** -0.058* -0.009*** 0.038*** -0.002 ( 1.83) ( 4.00) ( 1.84) ( 5.65) (2.60) ( 0.83) Inflation 0.011 0.000* 0.025 0.001** -0.132*** -0.002 (1.08) (1.72) (1.54) (2.03) ( 2.92) ( 1.34) Country Governance 0.486*** -0.000 0.658*** -0.008 2.575*** -0.016 (3.68) ( 0.06) (3.40) ( 0.55) (3.10) ( 0.62) Industry-country Yes Yes Yes Yes Yes Yes fixed effects Year fixed effects Yes Yes Yes Yes Yes Yes Adjusted/Pseudo R 0.224 0.086 0.180 0.476 0.228 0.527 Observations 30,032 72,704 30,032 72,704 10,704 25,922 This table presents the results from regressing the enterprise-level norm-deviant behavior on the level of economic uncertainty in the country and on other explanatory variables. Private and public firms are pooled together. Columns (1) and (2) present pooled OLS regression results. Columns (3) and (4) present Tobit regression results with left censoring at 0 (many of our private firms do not cheat on taxes or pay bribes). Columns (5) and (6) present propensity score-matching regression results using a symmetrical sample constituted of only the treated firms and their matches; the matching is one-to-one and it is based on all control variables and a caliper of 0.05. Uncertainty is the World Uncertainty Index (WUI) developed by Ahir et al. (2019). Cheating on taxes is the fraction of sales not reported to the tax authorities. Bribes is the fraction of sales paid in informal payments to the government authorities. Country, industry, and year fixed effects are included. All other variables are defined in Appendix in Table 13. In Panel A the t-statistics (shown in parentheses) are based on industry clustered robust standard errors and in Panel B they are based on industry-country clustered robust standard errors. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively. sample) and low (control sample) uncertainty countries. Using a pooled consistent with our baseline results and show that even after matching sample of treated and control observations, we run our OLS tests. The firms on several firm and country characteristics, the uncertainty- coefficient for cheating on taxes and bribes is again positive and statis- corruption relationship prevails. tically significant at the 5% level. Taken together, these findings are 11 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 6 Economic uncertainty and managerial perception: private versus public firms. Dependent variable Managerial perception of severity of business obstacles Obstacle Dysfunctional court Crime, theft, Prevalent High tax Ineffective tax Anti-competitive Reduced access to Political system disorder corruption rates admin. practices finance instability (1) (2) (3) (4) (5) (6) (7) (8) Uncertainty 0.598* 1.718*** 0.493 0.518 0.468 0.511 1.218*** 0.441 (1.88) (5.72) (1.62) (1.57) (1.51) (1.64) (4.85) (1.50) Private -0.030 -0.020 0.027 0.050* 0.037 0.016 0.076*** 0.026 ( 1.15) ( 0.84) (0.93) (1.84) (1.61) (0.66) (2.58) (1.05) Uncertainty × 0.038 -0.018 0.363 0.074 0.079 0.061 -0.660*** 0.312 Private (0.13) (¡0.07) (1.22) (0.25) (0.31) (0.21) (¡2.58) (1.26) Size 0.039*** 0.027*** 0.027*** 0.029*** 0.028*** 0.005** -0.002 0.025*** (11.77) (6.19) (11.63) (12.56) (14.67) (2.07) ( 0.69) (11.24) Age 0.035*** -0.007 0.002 0.005 0.005 0.024*** -0.051*** 0.014*** (5.57) ( 1.31) (0.40) (0.97) (0.87) (4.06) ( 8.73) (3.24) Exporter 0.064*** -0.066*** -0.013 -0.032*** 0.035*** -0.142*** -0.049*** 0.007 (4.66) ( 3.61) ( 0.94) ( 3.18) (3.35) ( 7.11) ( 2.66) (0.66) Government -0.096* -0.101*** -0.124*** -0.139*** -0.165*** -0.139*** 0.026 -0.040* ( 1.95) ( 3.93) ( 2.76) ( 5.19) ( 4.77) ( 3.38) (1.01) ( 1.79) Foreign 0.002 -0.056*** -0.036*** -0.076*** -0.037*** -0.138*** -0.238*** -0.007 (0.25) ( 3.84) ( 4.41) ( 11.10) ( 3.36) ( 13.16) ( 17.42) ( 0.83) GDP -1.473*** -1.071*** -1.184*** -1.366*** -1.556*** -0.240** -0.759*** -1.004*** ( 9.74) ( 7.35) ( 12.85) ( 10.57) ( 12.52) ( 1.96) ( 3.97) ( 8.17) ΔGDP 0.008* 0.004 0.007** 0.009* 0.011*** 0.006** 0.003 -0.005** (1.94) (0.78) (2.34) (1.85) (2.77) (2.06) (0.94) ( 2.11) Inflation 0.002 -0.000 0.004* 0.004** 0.007*** -0.000 0.001 0.005*** (0.91) ( 0.18) (1.78) (2.26) (5.07) ( 0.31) (0.54) (3.03) Country Governance -0.059 0.077* 0.001 0.154*** 0.119*** 0.152*** 0.182*** -0.072 ( 1.15) (1.78) (0.02) (5.80) (4.95) (4.04) (4.06) ( 1.40) Country fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Industry fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Pseudo R 0.128 0.119 0.123 0.105 0.104 0.090 0.084 0.160 Observations 103,738 103,738 103,738 103,738 103,738 103,738 103,738 103,738 This table analyses whether the severity of obstacles faced by the enterprises is perceived differently by private and public firms. Presented are the Poisson regression results. Uncertainty is the World Uncertainty Index (WUI) developed by Ahir et al. (2019). Private indicates that the firm is a small private enterprise. All variables pertaining to business obstacles indicate the interviewee’s opinion about the severity of an obstacle and range between zero and four, where zero indicates no obstacle and four indicates very severe obstacle. Country, industry, and year fixed effects are included. All other variables are defined in Appendix in Table 13. The t-statistics based on industry clustered robust standard errors are shown in parentheses. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively. 4.4. Private versus public firms: preference for different forms of Table 7 shows an important behavioral difference between private and corruption public firms. Because they are more closely monitored, public firms tend to cheat less on taxes than private firms (the coefficient for Uncertain- Next, we make an important distinction between small firms based ty×Private is significantly positive). Put differently, the small private on whether they are subject to monitoring by institutional shareholders firms find it easier to underreport sales for tax purposes than public ones, and other stockholders or whether they are tightly held with limited who are subject to more regulations and more scrutiny from govern- dissemination of information. Small private firms are subject to very mental authorities (Hope et al., 2011). However, the public firms pay a little monitoring. Thus, as expressed in Hypothesis 3, they are likely to higher percentage of their sales in bribes than the private firms. In engage in a different form of corruption than the more closely watched untabulated results, we find that private firms do indeed increase their and larger public firms. bribes when subjected to economic uncertainty; however, the negative First, we check whether private firms feel differently about business sign of the interaction term in Table 7 (see columns (2), (4), and (6)) obstacles than public firms. We run the regression described in Eq. (1) indicates that private firms cannot increase the magnitude of their bribes but instead use an interaction term for private firms (Private dummy). As as much as public firms. Either private firms run out of cash, as indicated displayed in Table 6, the interaction term, Uncertainty×Private, is sta- by the result in column (7) of Table 6, or they find it easier to lie about tistically insignificant which indicates there are no major differences their sales and resort to cheating on taxes as a way to cope with eco- among the managers of public and private firms on whether they nomic hardship. This contrast in preferred forms of corruption among perceive economic uncertainty as an amplifier of business obstacles in private and public firms that are subject to economic uncertainty is a the economy. A possible exception is their perception of financing novel finding for both the “greasing the wheels” (Huntington, 1968; Leff, constraints (column (7)). The managers of private firms feel that the 1964) hypothesis and the policy uncertainty literature (see, among financing constraints are much stronger for them which, in light of the others, Bloom, 2009). findings by Beck et al. (2005) and Hope et al. (2011), is to be expected given that private firms are smaller in size (see Table 2) and the un- certainty likely affects their sources of financing more severely. Second, we analyse whether there are differences between private We basically rerun the results from Table 5 but using only the sample of and public firms with regards to the uncertainty-corruption relationship. private firms. 12 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 7 Economic uncertainty and corruption: private versus public firms. Estimation model Pooled OLS Tobit with left censoring Propensity score-matching Dependent variable Cheating on taxes Bribes Cheating on taxes Bribes Cheating on taxes Bribes (1) (2) (3) (4) (5) (6) Uncertainty -0.155 0.086*** -0.374 0.368*** 6.947* 0.376*** ( 0.44) (2.91) ( 0.47) (5.23) (1.94) (4.54) Private -0.001 -0.001 0.023 0.012 0.038 0.016 ( 0.06) ( 0.38) (0.51) (1.36) (0.54) (1.52) ** Uncertainty × Private 0.735 -0.058* 1.195* -0.262*** 1.097 -0.265*** (2.52) (¡1.92) (1.70) (¡3.14) (1.32) (¡3.20) Size -0.012*** -0.001*** -0.026*** -0.003*** -0.037*** -0.002*** ( 7.62) ( 8.48) ( 8.34) ( 4.58) ( 11.49) ( 2.93) Age -0.010*** -0.001* -0.019*** -0.001 -0.000 -0.000 ( 4.55) ( 1.79) ( 4.08) ( 1.39) ( 0.01) ( 0.14) Exporter -0.005 0.001 -0.020* 0.009*** -0.013 0.012*** ( 0.84) (1.24) ( 1.89) (2.66) ( 0.74) (3.12) Government -0.002 0.003 -0.106*** -0.002 -0.083* -0.006 ( 0.25) (1.39) ( 4.87) ( 0.27) ( 1.93) ( 0.60) Foreign -0.022*** 0.001 -0.069*** -0.002 -0.073*** 0.000 ( 4.65) (0.89) ( 6.28) ( 0.68) ( 3.37) (0.06) GDP -0.468 -0.009 -0.598 -0.003 -19.222** -0.015 ( 0.87) ( 1.41) ( 0.68) ( 0.10) ( 2.40) ( 0.62) ΔGDP -0.027* -0.001*** -0.051** -0.006*** 0.045** -0.002 ( 1.98) ( 6.63) ( 2.10) ( 8.07) (2.21) ( 1.39) Inflation 0.009* 0.000* 0.022*** 0.001* -0.175* 0.001* (1.84) (1.99) (3.88) (1.70) ( 1.68) (1.82) Country Governance 0.284*** -0.001 0.365*** -0.011 4.618** 0.033*** (5.92) ( 0.64) (5.28) ( 1.57) (2.04) (2.87) Country fixed effects Yes Yes Yes Yes Yes Yes Industry fixed effects Yes Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Yes Adjusted/Pseudo R 0.210 0.070 0.150 0.394 0.172 0.441 Observations 30,032 72,704 30,032 72,704 10,704 25,922 This table analyses whether corrupt behavior by the enterprises is different for private and public firms. Columns (1) and (2) present pooled OLS regression results. Columns (3) and (4) present Tobit regression results with left censoring at 0 (many of our private firms do not cheat on taxes or pay bribes). Columns (5) and (6) present propensity score-matching regression results using a symmetrical sample constituted of only the treated firms and their matches; the matching is one-to-one and it is based on all control variables and a caliper of 0.05. Uncertainty is the World Uncertainty Index (WUI) developed by Ahir et al. (2019). Private indicates that the firm is a small private enterprise. Cheating on taxes is the fraction of sales not reported to the tax authorities. Bribes is the fraction of sales paid in informal payments to the government authorities. Country, industry, and year fixed effects are included. All other variables are defined in Appendix in Table 13. The t-statistics based on in- dustry clustered robust standard errors are shown in parentheses. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively. 4.5. Identification through instrumental variable regression 4.6. Mediation analysis: measuring the direct and indirect effects of policy uncertainty To improve the identification of our tests, we turn to an instrumental variable regression, whereby we represent uncertainty with the two Thus far, we show a positive association between rising economic instruments described in subsection 3.4. We predict a positive rela- uncertainty and the worsening of managerial perception regarding busi- tionship between these instruments and economic uncertainty. The re- ness obstacles. We also show that high economic uncertainty induces sults from the IV regressions in Eqs. (3a) and (3b) are displayed in corrupt behavior. However, a worsened managerial perception may also Table 8. As in Table 5, the sample has both private and public firms. affect managerial behavior towards cheating on taxes and paying bribes. Columns (1) and (3) show that both the proportion of opposition seats in Therefore, to further corroborate the association between economic un- the legislature and election closeness are indeed significantly and posi- certainty, managerial perception, and corruption, we perform a media- tively associated with economic uncertainty (with the possible excep- tion analysis (Baron and Kenny, 1986). This analysis allows us to isolate tion being Opposition seats in column (3)). The Cragg-Donald F-statistic the direct and indirect effects of economic uncertainty on corrupt indicates that the instruments are not weak since the values are higher behavior. Specifically, mediation takes place if economic uncertainty af- than the traditional threshold of 10 used by Stock and Yogo (2005). The fects cheating on taxes and paying bribes through another variable. In our Hansen’s j-statistic shows that the instruments do not overidentify the case, the mediator variables are the managerial perceptions regarding the estimation model. eight business obstacles identified in Table 4. The second stage results in columns (2) and (4) lead to qualitatively Fig. 1 depicts the structural equations in the mediation analysis. similar conclusions to the OLS and Tobit specifications in Table 5. Again, Based on the standardized coefficients in the figure, managerial the IV regression results support our Hypothesis 2 – higher economic perception regarding dysfunctional court system; crime, theft, and dis- uncertainty results in more corrupt behavior by firms in the form of order; prevalent corruption; anti-competitive practices; and political cheating on taxes and paying bribes. instability have a positive and indirect effect on tax cheating through rising economic uncertainty. Similarly, for paying bribes, almost all factors have positive and indirect effects. Table 9 shows that the factors that are related to managerial perception explain about 4.2% of the total effect of economic uncertainty on tax cheating and 18.5% of the effect 13 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 8 first alternative measure, Election closeness, was described in subsection Economic uncertainty and corruption: instrumental variable regressions. 3.4. In this subsection, rather than using Election closeness as an instru- ment for the WUI variable, we instead use it as a direct measure of Estimation model IV first IV second IV first IV second stage stage stage stage economic uncertainty (like in Julio and Yook, 2012). This use facilitates Dependent Uncertainty Cheating on Uncertainty Bribes a robustness check related to any possible measurement error bias that variable taxes originates from the construction of WUI by Ahir et al. (2019). (1) (2) (3) (4) In our second measure, we isolate only the local component of a Election closeness 0.010*** 0.009*** country’s economic uncertainty by orthogonalizing the country-level (8.17) (4.91) uncertainty index to the global aggregate uncertainty index available Opposition seats 0.032* 0.014 in Ahir et al. (2019). We run a simple OLS estimate with no constant (1.74) (0.87) Uncertainty 2.309*** 0.150*** term where a country-level measure is the dependent variable and (3.98) (2.61) global aggregate uncertainty is the only explanatory variable. To isolate Size -0.000 0.009*** -0.001* -0.001*** only the orthogonal component of local measures of policy uncertainty, ( 0.24) (3.91) ( 1.67) ( 4.08) we extract the residual values from this regression and use them as our Age -0.001 -0.026*** 0.003*** -0.002*** second uncertainty measure. We carry out this orthogonalization pro- ( 1.59) ( 5.48) (4.30) ( 4.02) Exporter 0.003** -0.044*** 0.005*** 0.001 cedure to obtain the portion of country-level uncertainty that corre- (2.20) ( 4.88) (3.08) (1.04) sponds only to the uncertainties pertaining to economic, political, and Government 0.006** -0.087*** 0.000 0.006** financial aspects of the country itself instead of the global uncertainty (2.28) ( 6.67) (0.14) (2.56) that affects all countries at that time. Foreign -0.000 -0.049*** 0.003** 0.002 Our third measure uses a country’s stock market volatility. This ( 0.20) ( 5.55) (2.04) (1.15) GDP 0.013*** -0.015 0.014*** -0.004*** measure is widely used as a proxy for economic uncertainty and corre- (5.86) ( 1.16) (4.72) ( 2.94) lates positively with the news-based index (Baker et al., 2016) as well as ΔGDP -0.006*** 0.024*** -0.002** -0.000** with the world uncertainty index (Ahir et al., 2019). To calculate the ( 7.63) (4.02) ( 2.20) ( 1.98) stock market volatility, we measure the standard deviation of the major Inflation 0.001*** -0.000 0.001*** -0.000 (3.30) ( 0.34) (4.62) ( 1.10) stock index of each country. We obtain data on stock index values from Country -0.018*** 0.010 -0.015*** 0.000 Compustat’s Global Index Prices dataset that is available at a monthly Governance frequency for only 32 countries in our sample. We then estimate a ( 9.90) (0.89) ( 7.95) (0.02) moving 12-month standard deviation for each index. Further, we Industry fixed Yes No Yes No effects combine this information with WBES data using the month and year of Year fixed effects Yes No Yes No the interview. Since not all countries have a stock market in our sample, Adjusted R 0.394 -0.028 0.200 -0.008 this results in a large reduction in the sample. Observations 29,376 29,376 68,305 68,305 For the fourth measure, we use the natural logarithm of the news- Cragg-Donald F- 1123.461 1070.313 based index available for a subset of sampled countries. In our sample, statistic Hansen j-statistic 0.831 0.998 we are able to obtain this information for Brazil, Chile, Colombia, p-value 0.362 0.318 Greece, India, Mexico, and the Russian Federation. Additionally, we use the European index for countries in Europe to increase our sample size This table presents the instrumental variable (IV) regression results from (this adds 16 more countries to our sample). We calculate the natural regressing the enterprise-level norm-deviant behavior on the level of economic uncertainty in the country and on other explanatory variables. Private and logarithm of the 3-month moving average news-based economic policy public firms are pooled together. Uncertainty is the World Uncertainty Index uncertainty for these countries. Using this measure also causes a large (WUI) developed by Ahir et al. (2019). Our instruments for Uncertainty are: reduction in the sample. Election closeness, which is the degree of competitiveness of the most recent 19 The results from using these four measures are provided in national election (for executive branch) in the country and Opposition seats, Table 10. The coefficients in all our specifications are positive and sta- which is the proportion of seats held by the largest opposition party in the given tistically significant at the 1% level. Hence, our results are not driven by country and year. Cheating on taxes is the fraction of sales not reported to the tax the worldwide changes in the economic uncertainty or the measurement authorities. Bribes is the fraction of sales paid in informal payments to the government authorities. Country, industry, and year fixed effects are included. All other variables are defined in Appendix in Table 13. The t-statistics based on industry clustered robust standard errors are shown in parentheses. Orthogonalization is implemented by running the following OLS: WUI = j,t Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, β GlobalWUI + є , where WUI is the world uncertainty index of a country j t j,t j,t j,t respectively. during time t, and GlobalWUI during time t is represented by the unweighted average of countries’ WUIs. Ahir et al. (2019) refer to this global index as the on paying bribes. While the indirect effects are both statistically sig- unweighted Global WUI during that period t (see their Fig. 1B). Note that the nificant, the indirect economic effect of a worse managerial perception above regression has no constant term. The local component of WUI for a is higher on paying bribes. country j in time t is the residual term, є . j,t Since we choose only one index per country (the largest stock index in terms of market capitalization), our final sample has one entry per country, month, 5. Further analyses and year. The 32 countries with both bribery data and stock index data at the time of the survey are Argentina, Bangladesh, Brazil, Bulgaria, Chile, Croatia, 5.1. Alternative measures of policy uncertainty Czech Republic, Ecuador, Egypt, Greece, Hungary, India, Indonesia, Jamaica, Jordan, Kenya, Lithuania, Malaysia, Mexico, Morocco, Nigeria, Pakistan, Peru, To reduce the measurement error issues with our uncertainty vari- Philippines, Poland, Russian Federation, Slovakia, Slovenia, Sri Lanka, ables, we use four additional measures of economic uncertainty. Our Thailand, Turkey, and Venezuela. The three alternative uncertainty measures and our original economic un- certainty measure (WUI) are strongly and positively correlated with each other. This is consistent with the findings of Ahir et al. (2019), who test their measure The indirect effects are calculated using the following formula: (indirect of uncertainty index against news-based economic policy uncertainty of Baker effect of managerial perception/total effect of uncertainty)*100. For Cheating on et al. (2016) and the stock market volatility and they report significantly pos- taxes: (0.004/0.096)*100 = 4.2% and for Bribes: (0.005/0.027)*100 = 18.5% itive correlations. 14 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Fig. 1. Mediation analysis: Isolating the direct and indirect effects of policy uncertainty. This figure presents the structural equation modeling of the mediation analysis to assess the direct and indirect effects of economic policy uncertainty on cheating on taxes and paying bribes. Uncertainty is the country-level World Uncertainty Index (WUI) developed by Ahir et al. (2019) using the country reports prepared by Economist Intelligence Unit (EIU). Standardized coefficients are reported next to each line. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively. 15 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 9 Economic uncertainty and corruption: mediation analysis. Dependent variable Cheating on taxes Cheating on taxes Bribes Bribes (1) (2) (3) (4) Total effect of Uncertainty 0.096*** 0.031*** (4.84) (5.62) Direct effect of Uncertainty 0.101*** 0.027*** (5.13) (4.73) 0.004** 0.005*** Mediating effect of Managerial (2.28) (3.89) perception Size -0.143*** -0.150*** -0.070*** -0.078*** ( 7.98) ( 8.00) ( 8.41) ( 9.48) Age -0.027*** -0.027*** -0.007 -0.008* ( 4.65) ( 4.64) ( 1.51) ( 1.71) Exporter -0.006 -0.006 0.008 0.007 ( 0.83) ( 0.81) (1.29) (1.19) Government -0.012** -0.009* 0.013** 0.015*** ( 2.17) ( 1.78) (2.24) (2.75) Foreign -0.023*** -0.023*** 0.006 0.006 ( 4.68) ( 4.89) (1.04) (1.13) GDP -1.223 -1.608 -0.169 -0.064 ( 0.64) ( 0.86) ( 1.49) ( 0.55) ΔGDP -0.267** -0.240** -0.064*** -0.062*** ( 2.13) ( 1.99) ( 6.56) ( 6.23) Inflation 0.226 0.243 0.029* 0.025 (1.41) (1.53) (1.92) (1.61) Country Governance 1.231*** 1.245*** -0.026 -0.021 (5.46) (5.43) ( 0.56) ( 0.45) Percent of total effect mediated 4.2% 18.5% Country fixed effects Yes Yes Yes Yes Industry fixed effects Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Adjusted/Overall R 0.209 0.221 0.070 0.110 Observations 30,032 30,032 30,032 30,032 This table presents the mediation analysis results from regressing the enterprise-level norm-deviant behavior on the level of economic uncertainty in the country and on other explanatory variables. Private and public firms are pooled together. Uncertainty is the World Uncertainty Index (WUI) developed by Ahir et al. (2019). Cheating on taxes is the fraction of sales not reported to the tax authorities. Bribes is the fraction of sales paid in informal payments to the government authorities. Managerial perception is the total effect of all variables pertaining to business obstacles indicating the interviewee’s opinion about the severity of an obstacle. Country, industry, and year fixed effects are included. All other variables are defined in Appendix in Table 13. The t-statistics based on industry clustered robust standard errors are shown in parentheses. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively. error in our main uncertainty variable. These robustness tests further 5.2. Political and societal factors and the uncertainty-corruption corroborate our claim that rising economic uncertainty induces many relationship private firms around the Globe to cheat on taxes and pay more bribes. These findings also confirm that the link between uncertainty and cor- Next, we explore the heterogeneity of our sample and examine how ruption exists even for the larger and more visible countries that are different political, economic, and societal factors enhance or reduce the calculated by Baker et al. (2016) economic uncertainty index (the uncertainty-corruption relationship. above-mentioned 23 countries that are either located in Europe or are among the largest emerging markets). Thus, the negative externality of 5.2.1. Political and economic dimensions rising economic uncertainty affects all countries regardless of their size First, we focus on the political and economic differences across and geographic location. countries. Previous evidence indicates that parliamentary systems are more likely to have large policy swings and to create a more uncertain environment for the economic agents (Julio and Yook, 2012). We therefore predict that the uncertainty-corruption relationship will be stronger in countries with a parliamentary system of government. We test this prediction by interacting a dummy variable that equals one if the country in which the firm is located in has a parliamentary system of Another widely used measure of policy uncertainty is national elections (see government, and zero otherwise. The results for this specification are for instance, Julio and Yook, 2012). Election-based policy uncertainty in- provided in columns (1) and (2) of Table 11. The interaction term is dicators have the advantage of capturing the exogenous variation in economic positive and statistically significant only for cheating on taxes. During uncertainty, however they do not tell us how much election uncertainty (relative to other countries’ elections) exists at the time of a given country’s high economic uncertainty, the firms located in countries with parlia- election. Also, these measures assume absence of uncertainty during mentary systems cheat more on taxes but do not pay more bribes relative non-election years; yet our sample is comprised of primarily low-income and to firms located in countries with presidential systems. This is consistent underdeveloped countries, where policy uncertainty is presumably very high with our prediction. even in the absence of elections. Furthermore, using election-based indicators We also analyze how the level of economic development (i.e., per- (which are in the form of dummies that take a value of one every 4–5 years capita income) of each country influences the uncertainty-corruption during the election year) leads to large attrition of our sample, which is based relationship. We obtain country-level income groupings from the on surveys that are not conducted every year (see Table 1). Therefore, in our World Bank. We predict that the private firms located in low-income context, using non-election-based proxies for economic uncertainty seems more countries will engage more in cheating on taxes and paying bribes due appropriate. 16 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 10 Economic uncertainty and corruption: alternative measure of economic uncertainty. Uncertainty measure Election closeness Orthogonalized uncertainty Stock market volatility News-based uncertainty (EPUI) Dependent variable Cheating on taxes Bribes Cheating on taxes Bribes Cheating on taxes Bribes Cheating on taxes Bribes (1) (2) (3) (4) (5) (6) (7) (8) Uncertainty 0.020*** 0.001*** 0.678*** 0.128*** 0.254*** 0.030*** 0.307*** 0.036*** (4.53) (3.74) (4.12) (4.29) (4.20) (4.55) (2.97) (3.10) Size 0.009*** -0.001*** -0.026*** -0.003*** -0.028*** -0.003*** -0.024*** -0.004*** (5.43) ( 5.32) ( 8.58) ( 4.48) ( 10.53) ( 3.29) ( 7.39) ( 3.88) Age -0.026*** -0.001*** -0.020*** -0.001 -0.022** -0.002 -0.022*** 0.001 ( 7.58) ( 3.19) ( 4.26) ( 1.28) ( 2.41) ( 1.15) ( 3.05) (0.51) Exporter -0.036*** 0.002 -0.020* 0.009*** -0.018 0.012*** 0.006 -0.001 ( 4.16) (1.67) ( 1.89) (2.68) ( 0.81) (3.33) (0.45) ( 0.30) Government -0.082*** 0.006** -0.147*** -0.000 -0.173*** 0.019 -0.187*** -0.021** ( 9.58) (2.38) ( 5.58) ( 0.04) ( 2.86) (1.37) ( 5.45) ( 1.97) Foreign -0.051*** 0.002 -0.070*** -0.002 -0.054*** 0.011 -0.110*** 0.006 ( 7.00) (1.51) ( 6.26) ( 0.57) ( 3.18) (1.60) ( 5.17) (1.47) GDP 0.014* -0.002*** -0.372 -0.005 -0.635*** -0.104** 0.096 -0.314*** (1.85) ( 3.76) ( 0.42) ( 0.15) ( 3.69) ( 2.20) (0.75) ( 9.16) ΔGDP 0.009*** -0.001*** -0.056** -0.006*** 0.140** -0.004* -0.532* 0.002* (4.02) ( 8.64) ( 2.36) ( 7.95) (2.35) ( 1.94) ( 1.76) (1.65) Inflation 0.002*** 0.000 0.018*** 0.001* -0.187*** -0.001 -0.086** -0.002*** (3.35) (1.59) (3.16) (1.65) ( 3.29) ( 0.68) ( 2.15) ( 3.63) Country Governance -0.028*** -0.002*** 0.302*** -0.011 -0.853*** -0.025 -1.392* -0.003 ( 5.58) ( 5.49) (4.31) ( 1.51) ( 3.44) ( 1.04) ( 1.76) ( 0.34) Country fixed effects No No Yes Yes Yes Yes Yes Yes Industry fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Pseudo R 0.089 0.032 0.150 0.393 0.116 0.406 0.148 0.735 Observations 29,598 69,549 30,032 72,704 12,656 35,711 11,481 24,717 This table presents the relationship between the enterprise-level norm-deviant behavior and the level of economic uncertainty when alternative measures of uncer- tainty are used. In columns (1) and (2), the economic uncertainty measure is Election closeness, which is the degree of competitiveness of the most recent national election (for executive branch) in the country. In columns (3) and (4), the economic uncertainty measure is the orthogonalized measure of uncertainty (i.e., country- specific WUI orthogonalized by the aggregated uncertainty index for the whole World provided in Ahir et al., 2019). In columns (5) and (6), the measure is stock market volatility, calculated as the standard deviation of the major stock index in the given country over the past 12 months. This index can be calculated for only 20 countries in our sample. In columns (7) and (8) the measure is the weighted three-month moving average of the news-based economic policy uncertainty index (EPUI) created by Baker et al. (2016), which is available only for 23 out of 93 countries in our sample. Private and public firms are pooled together. Cheating on taxes is the fraction of sales not reported to the tax authorities. Bribes is the fraction of sales paid in informal payments to the government authorities. Country, industry, and year fixed effects are included. All other variables are defined in Appendix in Table 13. The t-statistics based on industry clustered robust standard errors are shown in parentheses. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively. to the lack of oversight and transparency by well-funded and sophisti- calculate country-level trust and norms of civic cooperation measures. cated regulatory authorities. We, therefore, interact an indicator vari- We then rerun our baseline model with an interaction term. The results able which equals one if the firm is located in a low-income country, and reported in columns (1) and (2) of Table 12, show that firms located in zero otherwise. The results provided in columns (3) and (4) of Table 11 countries with higher levels of civic norms engage in less cheating on are consistent with our predictions. The coefficient for the interaction taxes, but civic norms do not mitigate bribery. This mixed result in- term is positive and statistically significant in both specifications that dicates that civic norms, as a replacement for a country’s governance indicates the marginal effect of economic uncertainty on corrupt institutions, reduces corrupt behavior to some extent when faced with behavior in private firms is higher in low-income and underdeveloped higher economic uncertainty. These results are largely consistent with countries. Thus, economic underdevelopment of a country amplifies the the literature on trust (Aghion et al., 2010; Kanagaretnam et al., 2018) corruptive effects that higher policy uncertainty has on small private and Hypothesis 4. firms. Next, we look at how religiosity can affect the uncertainty-corruption relationship. We obtain data on country-level religiosity from a Gallup 5.2.2. The role of societal norms poll conducted in 2009 that asked individuals in each country about the Previous evidence indicates that societal norms reduce norm-deviant importance of religion in their lives. Using this information, we create a behavior. Based on the findings of Putnam et al. (1994), La Porta et al. dummy variable that equals one if more than 70% of the residents in a (1997), and Fisman and Miguel (2007), we argue that societal norms are country replied “yes” when asked, “Is religion important in your daily a strong determinant of whether corruption exists. Similarly, Kanagar- life?”, and zero otherwise. Religion is shown to reduce levels of etnam et al. (2018) and Hasan et al. (2017b) show that societal trust and perceived corruption (Mensah, 2014) as well as mitigate corporate social capital are negatively associated with corporate tax avoidance. Jin misconduct such as financial reporting fraud (McGuire et al., 2012). We, et al. (2017) show that banks in high social capital regions in the US therefore, predict that the religiosity of a country’s residents can break, were less prone to financial failures in the recent global financial crisis or at least mitigate, the uncertainty-corruption relationship. The results due to their higher transparency. Furthermore, following the arguments for this specification are provided in columns (3) and (4) of Table 12. in Aghion et al. (2010), we predict that trust becomes vital when The interaction terms are once again negative and statistically signifi - country-level governance institutions are weak. Consistent with this cant. These results are also consistent with Hypothesis 4 that when prediction, we argue that civic norms in the form of trust and civic country-level regulatory agencies start becoming ineffective (i.e., cor- cooperation negatively influence the uncertainty-corruption ruption in private firms starts rising together with the economic un- relationship. certainty), religiosity as a social norm can have a certain ameliorating To test this prediction, we collect data from the World Values Survey effect in reducing cheating and bribery. (WVS) and follow the methodology in Knack and Keefer (1997) to 17 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 11 Economic uncertainty and corruption: do country’s political system and income level matter?. Dependent variable Cheating on taxes Bribes Cheating on taxes Bribes Country Characteristic Parliamentary system of Parliamentary system of Low-income Low-income government government countries countries (1) (2) (3) (4) Uncertainty 0.834*** 0.070*** 0.964*** 0.040* (3.86) (3.44) (3.63) (1.78) Characteristic -0.161*** 0.007 0.161*** 0.088*** ( 3.48) (0.68) (3.48) (8.33) ** Uncertainty × Country Characteristic 2.055*** 0.036 1.086 0.103* (5.21) (0.31) (2.07) (1.69) Size 0.008*** -0.002*** 0.009*** -0.002*** (2.95) ( 2.85) (3.28) ( 3.19) Age -0.047*** -0.006*** -0.042*** -0.004*** ( 6.35) ( 4.26) ( 4.94) ( 3.16) Exporter -0.070*** 0.013*** -0.081*** 0.011*** ( 4.83) (3.84) ( 5.23) (3.57) Government -0.234*** 0.004 -0.284*** -0.008 ( 5.54) (0.44) ( 11.77) ( 1.04) Foreign -0.136*** -0.002 -0.123*** -0.002 ( 7.16) ( 0.43) ( 7.15) ( 0.40) GDP -0.007 -0.014*** 0.094*** 0.021*** ( 0.62) ( 5.11) (5.18) (4.93) ΔGDP 0.021*** -0.002*** 0.017*** -0.004*** (4.50) ( 3.93) (4.02) ( 6.24) Inflation 0.002** 0.000 -0.002 -0.000 (2.12) (1.18) ( 1.20) ( 1.29) Country Governance -0.006 -0.018*** -0.044*** -0.024*** ( 1.04) ( 11.82) ( 5.02) ( 10.48) Country fixed effects Yes Yes Yes Yes Industry fixed effects Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Pseudo R 0.062 0.233 0.065 0.258 Observations 28,737 68,933 30,032 72,118 P(Uncertainty + Characteristic + Uncertainty × 0.000 0.000 0.000 0.000 Characteristic = 0) The table conducts cross-country analyses to assess the role of political system and income level. The results are obtained from regressing the enterprise-level norm- deviant behavior on the level of economic uncertainty, country characteristic, and other explanatory variables. In columns (1) and (2), Characteristic equals one if the country has a parliamentary system of government, and zero otherwise. In columns (3) and (4), Characteristic equals one if the country belongs to the low-income group based on World Bank’s classification, and zero otherwise. Private and public firms are pooled together. Uncertainty is the World Uncertainty Index (WUI) developed by Ahir et al. (2019). Cheating on taxes is the fraction of sales not reported to the tax authorities. Bribes is the fraction of sales paid in informal payments to the government authorities. Country, industry, and year fixed effects are included. All other variables are defined in Appendix in Table 13. The t-statistics based on industry clustered robust standard errors are shown in parentheses. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively. 5.2.3. Other socioeconomic dimensions in breaking the link between economic uncertainty and firm-level In addition to the above dimensions, we consider several other so- corruption. cioeconomic, political, and geographic aspects of a society (or a coun- try). We conduct a series of cross-country tests (untabulated) to 5.3. An alternative measure of corruption: time spent with government understand what influences the relationship between policy uncertainty 21 authorities and corruption. First, we test how cultural aspects of a society can influence the uncertainty-corruption relationship. We obtain data on In an untabulated test, we also analyze whether the management cultural dimensions from Hofstede (1980). Consistent with Husted spends more time with government authorities. This measure may not (1999), we find that societies at higher levels in the uncertainty avoid- directly relate to corruption, but it still reflects the overall inefficiency of ance index (UAI) pay more bribes during high economic uncertainty. For the firm in its involvement with the government. While the WBES survey cheating on taxes, the cross-country results split according to the UAI does not provide the reason for the time spent with government au- and are not statistically significant at conventional levels. thorities, we can safely assume that any excessive or unnecessary time Second, we find that there are some differences between the coun- spent with government authorities is a sign of norm-deviant behavior in tries located in different continents (Africa, Asia, Europe, and North and the form of lobbying or another such activity. South Americas) in terms of cheating on taxes and paying bribes. For To explore this association, we use the amount of time that man- instance, cheating is more prevalent in Africa and Europe, while bribery agement spends with the government authorities in a given year as a is more common in Asia. However, regardless of the continent they are percentage of their total working hours. The mean (median) percentage located on, the private firms react to policy uncertainty by engaging in time spent with the government is 10.44% (4.00%) of the management’s one form of norm-deviant behavior or the other. Furthermore, as we total working hours during a year. Our regression analyses estimate that demonstrated earlier, regardless of the continent, a country’s socio- during high economic uncertainty, firms ’ management spends more economic features and various political institutions do make a difference time with the government. The coefficient for our uncertainty measure is statistically significant at the 5% level and its economic impact is as follows: A one standard deviation increase in our WUI measure trans- lates into about 0.49% more time spent with the government, which While these tests are untabulated, they are available from the authors on request. roughly corresponds to two additional days. 18 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Table 12 Economic uncertainty and corruption: the mitigating role of civic norms and religiosity. Dependent variable Cheating on taxes Bribes Cheating on taxes Bribes Social Norm Civic norms Civic norms Religion Religion (1) (2) (3) (4) Uncertainty 12.002*** 0.146 3.432*** 0.139** (5.10) (1.52) (5.65) (2.39) Social Norm 2.485 -0.018 0.326*** 0.006 (1.35) ( 0.49) (6.89) (1.14) Uncertainty × Social Norm -13.364*** -0.407 -2.533*** -0.117* (¡4.34) (¡1.43) (¡3.65) (¡1.85) Size -0.027*** -0.001** 0.008*** -0.002*** ( 8.16) ( 2.05) (3.72) ( 3.02) Age -0.016** -0.002 -0.040*** -0.004*** ( 2.40) ( 1.45) ( 5.59) ( 3.42) Exporter -0.015 0.008** -0.071*** 0.012*** ( 1.13) (1.98) ( 5.11) (3.71) Government -0.157*** 0.005 -0.227*** -0.000 ( 5.84) (0.46) ( 9.07) ( 0.01) Foreign -0.051*** 0.003 -0.125*** -0.002 ( 3.58) (0.76) ( 9.31) ( 0.47) GDP 3.041*** -0.043 0.020* -0.014*** (5.30) ( 1.23) (1.71) ( 4.98) ΔGDP 0.204*** -0.002*** 0.022*** -0.003*** (4.67) ( 2.99) (5.13) ( 4.68) Inflation 0.068*** -0.000 0.004*** 0.000 (4.37) ( 0.71) (3.53) (1.36) Country Governance 0.773*** -0.032*** -0.001 -0.020*** (4.94) ( 2.91) ( 0.16) ( 10.19) Country fixed effects Yes Yes No No Industry fixed effects Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Pseudo R 0.140 0.397 0.072 0.230 Observations 22,990 56,793 29,902 72,352 P(Uncertainty + Norm + 0.000 0.000 0.000 0.000 Uncertainty × Norm = 0) The table analyses the role of social norms, such as civic norms and religiosity. The results are obtained by regressing the enterprise-level norm-deviant behavior on the level of economic uncertainty, social norm, and other explanatory variables. Civic norms is the average level of trust in others and norms of civic cooperation in a country based on World Values Survey (WVS) data, normalized to range between zero and one. Religion equals one if more than 70% of the country’s residents replied ‘yes’ when asked, ‘Is religion important in your daily life?’ in a Gallop Poll in 2009. Private and public firms are pooled together. Cheating on taxes is the fraction of sales not reported to the tax authorities. Bribes is the fraction of sales paid in informal payments to the government authorities. Uncertainty is the World Uncertainty Index (WUI) developed by Ahir et al. (2019). Country, industry, and year fixed effects are included. All other variables are defined in Appendix in Table 13. The t-statistics based on industry clustered robust standard errors are shown in parentheses. Coefficients marked with *, **, and *** are significant at 10%, 5%, and 1%, respectively. We also test whether firms get more frequent visits from tax au- predicted by “the greasing the wheels” hypothesis (Acemoglu and Ver- thorities during higher uncertainty periods. We find that tax authorities dier, 2000; Leff, 1964; Mendoza et al., 2015), firms engage in corrupt visit firms less frequently during such times. This infrequency further behavior to reduce the inefficiencies associated with the elevated policy indicates that the time spent with government authorities is not due to uncertainty that originates from political and governmental activities. taxes, but perhaps due to another activity such as lobbying. Less over- We report empirical evidence consistent with this argument. Specif- sight from authorities during uncertain times paves the way for norm- ically, we show that firms pay more bribes and cheat more often on taxes deviant behavior like cheating on taxes. in the presence of higher economic policy uncertainty. These results are robust to endogeneity bias; industry, year, and country fixed effects; and 6. Conclusion alternative measures of policy uncertainty. We use the heterogeneity of our sample to find that when facing We study how policy uncertainty contributes to the severity of the policy uncertainty, private firms ’ corruption is different from that of obstacles that face private firms in 93 countries. The firms report that public firms; the former cheats by underreporting taxes and the latter they face more severe constraints and higher socioeconomic in- pays more bribes. Different political, legal, and societal factors also efficiencies when policy uncertainty is higher. We argue that the higher affect the uncertainty-corruption relationship. Since a parliamentary unpredictability of future economic and political outcomes and more system of government is more prone to large economic policy swings, we severe business obstacles induce an environment of uncertainty and find that firms located in such countries are more likely to cheat on inefficiency. When facing an uncertain environment, firm managers taxes. The uncertainty-corruption relationship is stronger in firms become more worried about their own future (Hogg, 2007) and pay located in low-income countries. On the other hand, societal factors such more attention to negative events and obstacles (Arenas et al., 2006). as civic norms and religion mitigate the uncertainty-corruption rela- Some of them ultimately resort to corruption to cope with the high tionship. These results provide new insights on how policy uncertainty uncertainty and the perceived or real inefficiencies it brings. As influences corporate practices of small firms around the world. 19 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Appendix Table 13 Variable definitions. Variable Description Source Uncertainty The World Uncertainty Index (WUI) developed by Ahir et al. (2019) Ahir et al. (2019) Election closeness The degree of competitiveness of the most recent national election (for executive branch) in the country Database of Political Institutions Opposition seats The proportion of seats held by the largest opposition party in the given country and year. Database of Political Institutions Cheating on taxes The fraction of sales not reported to the tax authorities WBES Bribes The fraction of sales paid in informal payments to the government authorities WBES Dysfunctional court How much of an obstacle is the court system to your enterprise? Ranges between 0 (no obstacle) and 5 (very severe WBES system obstacle) Crime, theft, disorder How much of an obstacle is crime, theft, and disorder to your enterprise? Ranges between 0 (no obstacle) and 5 (very WBES severe obstacle) Prevalent corruption How much of an obstacle is corruption to your enterprise? Ranges between 0 (no obstacle) and 5 (very severe obstacle) WBES High tax rates How much of an obstacle is tax rates to your enterprise? Ranges between 0 (no obstacle) and 5 (very severe obstacle) WBES Ineffective tax How much of an obstacle is tax administration to your enterprise? Ranges between 0 (no obstacle) and 5 (very severe WBES administration obstacle) Anti-competitive practices How much of an obstacle is practices of competitors in informal sector to your enterprise? Ranges between 0 (no obstacle) WBES and 5 (very severe obstacle) Reduced access to finance How much of an obstacle is access to finance to your enterprise? Ranges between 0 (no obstacle) and 5 (very severe WBES obstacle) Political instability How much of an obstacle is political instability to your enterprise? Ranges between 0 (no obstacle) and 5 (very severe WBES obstacle) Size The firm ’s total sales in the previous fiscal year (thousands of USD) WBES Age The firm age in years WBES Exporter Equals one if at least ten percent of firm ’s sales are exported, and zero otherwise WBES Government Equals one if at least ten percent of the firm is owned by the government, and zero otherwise WBES Foreign Equals one if at least ten percent of the firm is owned by a foreign entity, and zero otherwise WBES GDP The gross domestic product per capita for the country World Bank ΔGDP The growth in GDP for the country World Bank Inflation The inflation rate for the country World Bank Governance The first principal component of six governance indicators for each country (see Kaufmann et al., 2009) World Bank Civic norms The average level of trust in others and norms of civic cooperation in a country. WVS Religion Equals one if more than 70% of the country’s residents replied ‘yes’ when asked, ‘Is religion important in your daily life?’ Gallop in a Gallop Poll in 2009 Stock market volatility Calculated as the standard deviation of the major stock index in the given country over the past 12 months Compustat Global News-based uncertainty The weighted three-month moving average of the news-based economic policy uncertainty index Baker et al. (2016) Beck, T., Demirgüç-Kunt, A., Maksimovic, V., 2005. Financial and legal constraints to References growth: does firm size matter? J. Financ. 60, 137–177. Beltratti, A., Stulz, R.M., 2012. The credit crisis around the globe: why did some banks Acemoglu, D., Verdier, T., 2000. The choice between market failures and corruption. Am. perform better? J. Financ. Econ. 105, 1–17. Econ. Rev. 90, 194–211. Bentolila, S., Bertola, G., 1990. Firing costs and labour demand: how bad is Adelopo, I., Rufai, I., 2020. Trust deficit and anti-corruption initiatives. J. Bus. Ethics eurosclerosis? Rev. Econ. Stud. 57, 381. 163, 429–449. Bernanke, B.S., 1983. Irreversibility, uncertainty, and cyclical investment. Q. J. Econ. 98, Aghion, P., Algan, Y., Cahuc, P., Shleifer, A., 2010. Regulation and distrust. Q. J. Econ. 85–106. 125, 1015–1049. Bhattacharya, U., Hsu, P.H., Tian, X., Xu, Y., 2017. What affects innovation more: policy Aguilera, R.V., Vadera, A.K., 2008. The dark side of authority: antecedents, mechanisms, or policy uncertainty? J. Financ. Quant. Anal. 52, 1869–1901. and outcomes of organizational corruption. J. Bus. Ethics 77, 431–449. Bloom, N., 2009. The impact of uncertainty shocks. Econometrica 77, 623–685. Ahir, H., Bloom, N., Furceri, D., 2019. The world uncertainty index. Stanf. Inst. Econ. Bloom, N., Bond, S., Van Reenen, J., 2007. Uncertainty and investment dynamics. Rev. Policy Res. Econ. Stud. 74, 391–415. ¨¨ ¨ Airaksinen, A., Luomaranta, H., Roodhuijzen, A., Alajaasko, P., 2016. Statistics on Small Bonaime, A., Gulen, H., Ion, M., 2018. Does policy uncertainty affect mergers and and Medium-Sized Enterprises - Statistics Explained, Eurostat Statistics Explained. acquisitions? J. Financ. Econ. 129, 531–558. Amore, M.D., Minichilli, A., 2018. Local political uncertainty, family control, and Boutchkova, M., Doshi, H., Durnev, A., Molchanov, A., 2012. Precarious politics and investment behavior. J. Financ. Quant. Anal. 53, 1781–1804. return volatility. Rev. Financ. Stud. 25, 1111–1154. An, H., Chen, Y., Luo, D., Zhang, T., 2016. Political uncertainty and corporate Braun, M., Di Tella, R., 2004. Inflation, inflation variability, and corruption. Econ. Polit. investment: evidence from China. J. Corp. Financ. 36, 174–189. 16, 77–100. Anand, V., Ashforth, B.E., Joshi, M., 2004. Business as usual: the acceptance and Brav, O., 2009. Access to capital, capital structure, and the funding of the firm. J. Financ. perpetuation of corruption in organizations. Acad. Manag. Exec. 18, 39–53. 64, 263–308. Arenas, A., Tabernero, C., Briones, E., 2006. Effects of goal orientation, error orientation Brogaard, J., Detzel, A., 2015. The asset-pricing implications of government economic and self-efficacy on performance in an uncertain situation. Soc. Behav. Pers. 34, policy uncertainty. Manag. Sci. 61, 3–18. 569–586. Caballero, R.J., Engel, E.M.R.A., Haltiwanger, J.C., 1995. Plant-level adjustment and Ashforth, B.E., Anand, V., 2003. The normalization of corruption in organizations. Res. aggregate investment dynamics. Brook. Pap. Econ. Act. 1995, 1–54. Organ. Behav. 25, 1–52. Callen, J.L., Fang, X., 2015. Religion and stock price crash risk. J. Financ. Quant. Anal. Ashraf, B.N., Shen, Y., 2019. Economic policy uncertainty and banks’ loan pricing. 50, 169–195. J. Financ. Stab. 44, 100695. Chan, Y.C., Saffar, W., Wei, K.C.J., 2021. How economic policy uncertainty affects the Asker, J., Farre-Mensa, J., Ljungqvist, A., 2011. Comparing the investment behavior of cost of raising equity capital: Evidence from seasoned equity offerings. J. Financ. public and private firms. Natl. Bur. Econ. Res. Stab. 53, 100841. Baker, S.R., Bloom, N., Davis, S.J., 2016. Measuring economic policy uncertainty. Q. J. Chen, F., Hope, O.K., Li, Q., Wang, X., 2011. Financial reporting quality and investment Econ. 131, 1593–1636. efficiency of private firms in emerging markets. Account. Rev. 86, 1255–1288. Bardhan, P., 1997. Corruption and development: a review of issues. J. Econ. Lit. 35, Clarke, G., Xu, L., 2002. Ownership, competition, and corruption: bribe takers versus 1320–1346. bribe payers. World Bank Policy Res. Work. Pap. 2783. Baron, R.M., Kenny, D.A., 1986. The moderator-mediator variable distinction in social Cloyd, C.B., Pratt, J., Stock, T., 1996. The use of financial accounting choice to support psychological research. Conceptual, strategic, and statistical considerations. J. Pers. aggressive tax positions: public and private firms. J. Account. Res. 34, 23. Soc. Psychol. 51, 1173–1182. 20 M. Afzali et al. Journal of Financial Stability 57 (2021) 100936 Cohen, L., Coval, J., Malloy, C., 2011. Do powerful politicians cause corporate La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R.W., 1997. Trust in large downsizing? J. Polit. Econ. 119, 1015–1060. organizations. Am. Econ. Rev. 87, 333–338. Cohen, L., Malloy, C.J., 2014. Friends in high places. Am. Econ. J. Econ. Policy 6, 63–91. Lawless, M., O’Connell, B., O’Toole, C., 2015. SME recovery following a financial crisis: Colak, G., Durnev, A., Qian, Y., 2017. Political uncertainty and IPO activity: evidence does debt overhang matter? J. Financ. Stab. 19, 45–59. from U.S. gubernatorial elections. J. Financ. Quant. Anal. 52, 2523–2564. Leahy, J.V., Whited, T.M., 1996. The effect of uncertainty on investment: some stylized ¨ ¨ Colak, G., Gungoraydinoglu, A., Oztekin, O., 2018. Global leverage adjustments, facts. J. Money Credit Bank. 28, 64–83. uncertainty, and country institutional strength. J. Financ. Intermed. 35, 41–56. Lederman, D., 2010. An international multilevel analysis of product innovation. J. Int. Collins, J.D., Uhlenbruck, K., Rodriguez, P., 2009. Why firms engage in corruption: a top Bus. Stud. 41, 606–619. management perspective. J. Bus. Ethics 87, 89–108. Leff, N.H., 1964. Economic development through bureaucratic corruption. Am. Behav. Cooper, R.W., Haltiwanger, J.C., 2006. On the nature of capital adjustment costs. Rev. Sci. 8, 8–14. Econ. Stud. 73, 611–633. Mauro, P., 1995. Corruption and growth. Q. J. Econ. 110, 681–712. Cutler, L.N., 1988. Some reflections about divided government. Pres. Stud. Q 18, McGuire, S.T., Omer, T.C., Sharp, N.Y., 2012. The impact of religion on financial 485–492. reporting irregularities. Account. Rev. 87, 645–673. D’Souza, J., Megginson, W.L., Ullah, B., Wei, Z., 2017. Growth and growth obstacles in Mendoza, R.U., Lim, R.A., Lopez, A.O., 2015. Grease or sand in the wheels of commerce? transition economies: privatized versus de novo private firms. J. Corp. Financ. 42, Firm level evidence on corruption and SMES. J. Int. Dev. 27, 415–439. 422–438. Mensah, Y.M., 2014. An analysis of the effect of culture and religion on perceived Datta, S., Doan, T., Iskandar-Datta, M., 2019. Policy uncertainty and the maturity corruption in a global context. J. Bus. Ethics 121, 255–282. structure of corporate debt. J. Financ. Stab. 44, 100694. M´ eon, P.G., Sekkat, K., 2005. Does corruption grease or sand the wheels of growth? Dixit, A.K., Pindyck, R.S., 2012. Investment Under Uncertainty. Princeton University Public Choice 122, 69–97. Press. M´ eon, P.G., Weill, L., 2010. Is corruption an efficient grease? World Dev. 38, 244–259. Drobetz, W., Ghoul, El, Guedhami, S., Janzen, M, O., 2018. Policy uncertainty, Mertzanis, C., 2019. Family ties, institutions and financing constraints in developing investment, and the cost of capital. J. Financ. Stab. 39, 28–45. countries. J. Bank. Financ. 108, 105650. Edgerton, J., 2012. Agency problems in public firms: evidence from corporate jets in Michaely, R., Roberts, M.R., 2012. Corporate dividend policies: lessons from private leveraged buyouts. J. Financ. 67, 2187–2213. firms. Rev. Financ. Stud. 25, 711–746. Fisman, R., Miguel, E., 2007. Corruption, norms, and legal enforcement: evidence from Nagar, V., Petroni, K., Wolfenzon, D., 2011. Governance problems in closely held diplomatic parking tickets. J. Polit. Econ. 115, 1020–1048. corporations. J. Financ. Quant. Anal. 46, 943–966. Francis, B.B., Hasan, I., Zhu, Y., 2014. Political uncertainty and bank loan contracting. Nagar, V., Schoenfeld, J., Wellman, L., 2019. The effect of economic policy uncertainty J. Empir. Financ. 29, 281–286. on investor information asymmetry and management disclosures. J. Account. Econ. Gao, H., Li, K., 2015. A comparison of CEO pay-performance sensitivity in privately-held 67, 36–57. and public firms. J. Corp. Financ. 35, 370–388. Nguyen, N.H., Phan, H.V., 2017. Policy uncertainty and mergers and acquisitions. Goel, R.K., Ram, R., 2013. Economic uncertainty and corruption: evidence from a large J. Financ. Quant. Anal. 52, 613–644. cross-country data set. Appl. Econ. 45, 3462–3468. Pastor, L., Veronesi, P., 2012. Uncertainty about government policy and stock prices. Gulen, H., Ion, M., 2016. Policy uncertainty and corporate investment. Rev. Financ. Stud. J. Financ. 67, 1219–1264. 29, 523–564. Pastor, L., Veronesi, P., 2013. Political uncertainty and risk premia. J. Financ. Econ. 110, ¨ ¨ Gungoraydinoglu, A., Colak, G., Oztekin, O., 2017. Political environment, financial 520–545. ´ ´ intermediation costs, and financing patterns. J. Corp. Financ. 44, 167–192. Pena Lopez, J.A., Sanchez Santos, J.M., 2014. Does corruption have social roots? The role Hasan, I., Hoi, C.K., Wu, Q., Zhang, H., 2017a. Social capital and debt contracting: of culture and social capital. J. Bus. Ethics 122, 697–708. evidence from bank loans and public bonds. J. Financ. Quant. Anal. 52, 1017–1047. Petersen, M.A., 2009. Estimating standard errors in finance panel data sets: Comparing Hasan, I., Hoi, C.K.S., Wu, Q., Zhang, H., 2017b. Does social capital matter in corporate approaches. Rev. Financ. Stud. 22 (1), 435–480. decisions? Evidence from corporate tax avoidance. J. Account. Res. 55, 629–668. Phan, H.V., Nguyen, N.H., Nguyen, H.T., Hegde, S., 2019. Policy uncertainty and firm Hofstede, G., 1980. Culture’s Consequences: International Differences in Work-related cash holdings. J. Bus. Res. 95, 71–82. Values. Sage Publications,, Beverly Hills. Putnam, R.D., Leonardi, R., Nanetti, R., 1994. Making Democracy Work: Civic Traditions Hogg, M.A., 2007. Uncertainty-identity theory. Adv. Exp. Soc. Psychol. 39, 69–126. in Modern Italy. Hope, O.K., Thomas, W., Vyas, D., 2011. Financial credibility, ownership, and financing Rodriguez, P., Uhlenbruck, K., Eden, L., 2005. Government corruption and the entry constraints in private firms. J. Int. Bus. Stud. 42, 935–957. strategies of multinationals. Acad. Manag. Rev. 30, 383–396. Huntington, S.P., 1968. Political Order in Changing Societies. Yale University Press. Saunders, A., Steffen, S., 2011. The costs of being private: evidence from the loan market. Husted, B.W., 1999. Wealth, culture, and corruption. J. Int. Bus. Stud. 30, 339–360. Rev. Financ. Stud. 24, 4091–4122. International Monetary Fund, 2012. Coping with high debt and sluggish growth. World Scartascini, C., Cruz, C., Keefer, P., 2018. The Database of Political Institutions 2017. Econ. Financ. Surv., 2012, 1–250. Inter-American Dev. Bank. Jens, C.E., 2017. Political uncertainty and investment: causal evidence from U.S. Shang, L., Lin, J.-C., Saffar, W., 2021. Does economic policy uncertainty drive the gubernatorial elections. J. Financ. Econ. 124, 563–579. initiation of corporate lobbying? J. Corp. Financ. 70, 102053. Jiang, F., John, K., Li, C.W., Qian, Y., 2018. Earthly reward to the religious: religiosity Shleifer, A., Vishny, R.W., 1993. Corruption. Q. J. Econ. 108, 599–617. and the costs of public and private debt. J. Financ. Quant. Anal. 53, 2131–2160. Slemrod, J., 2007. Cheating ourselves: the economics of tax evasion. J. Econ. Perspect. Jin, J.Y., Kanagaretnam, K., Lobo, G.J., Mathieu, R., 2017. Social capital and bank 21, 25–48. stability. J. Financ. Stab. 32, 99–114. Staiger, D., Stock, J.H., 1997. Instrumental variables regression with weak instruments. Julio, B., Yook, Y., 2012. Political uncertainty and corporate investment cycles. Econometrica 65, 557. J. Financ. 67, 45–84. Stepan, A., Skach, C., 1993. Constitutional frameworks and democratic consolidation: Kanagaretnam, K., Lee, J., Lim, C.Y., Lobo, G., 2018. Societal trust and corporate tax parliamentarianism versus presidentialism. World Polit. 46, 1–22. avoidance. Rev. Account. Stud. 23, 1588–1628. Stock, J.H., Yogo, M., 2005. Testing for weak instruments in Linear IV regression. In: Kaufmann, D., Kraay, A., Mastruzzi, M., 2009. Governance Matters VIII: Aggregate and Identification and Inference for Econometric Models: Essays in Honor of Thomas Individual Governance Indicators. The World Bank, pp. 1996–2008. Rothenberg. Cambridge University Press, pp. 80–108. Kaufmann, D., Wei, S.-J., 2000. Does “Grease Money” speed up the wheels of commerce? Sundquist, J.L., 1988. Needed: a political theory for the new era of coalition government IMF Work. Pap. 00, 1–21. in the United States. Political Sci. Q. 103, 613. Kelly, B., Pastor, L., Veronesi, P., 2016. The price of political uncertainty: theory and Svensson, J., 2003. Who must pay bribes and how much? Evidence from a cross section evidence from the option market. J. Financ. 71, 2417–2480. of firms. Q. J. Econ. 118, 207–230. Kelly, S.Q., 1993. Divided we govern? A reassessment. Polity 25, 475–484. Treisman, D., 2007. What have we learned about the causes of corruption from ten years Khalil, S., Saffar, W., Trabelsi, S., 2015. Disclosure standards, auditing infrastructure, and of cross-national empirical research? Annu. Rev. Polit. Sci. 10, 211–244. bribery mitigation. J. Bus. Ethics 132, 379–399. Van Den Bos, K., Euwema, M.C., Poortvliet, P.M., Maas, M., 2007. Uncertainty Kim, C., Pantzalis, C., Chul Park, J., 2012. Political geography and stock returns: the management and social issues: uncertainty as an important determinant of reactions value and risk implications of proximity to political power. J. Financ. Econ. 106, to socially deviating people. J. Appl. Soc. Psychol. 37, 1726–1756. 196–228. Wellalage, N.H., Locke, S., Samujh, H., 2019. Corruption, gender and credit constraints: Klassen, K.J., 1997. The impact of inside ownership concentration on the trade-off evidence from South Asian SMEs. J. Bus. Ethics 159, 267–280. between financial and tax reporting. Account. Rev. 72, 455–474. Wooldridge, J.M., 2002. Econometric analysis of Cross Section and Panel Data. MIT Knack, S., Keefer, P., 1997. Does social capital have an economic payoff? A cross-country Press. investigation. Q. J. Econ. 112, 1251–1288. Xie, X., Qi, G., Zhu, K.X., 2019. Corruption and new product innovation: examining firms ’ ethical dilemmas in transition economies. J. Bus. Ethics 160, 107–125.

Journal

SSRN Electronic JournalUnpaywall

Published: Jan 1, 2021

There are no references for this article.