Corruption, Political Instability and Firm-Level Export Decisions Kul Kapri 1 Rowan University August 2018 Abstract In this paper I use South Asian firm-level data to examine whether the impact of corruption (political instability) on firm level export decisions depends on political instability (corruption). This study uses the IV probit model to analyze the data collected by the World Bank. The results confirm that Political instability (corruption) reduces the impact of corruption (political instability) on firm-level export decisions. In addition, this analysis yields meaningful policy implication regarding political instability, corruption and a firm-level decision entering the foreign market in developing countries. JEL Codes: F10, F14, O1, K42 Keywords: Exports, Corruption, Political Instability, Firm Heterogeneity, World Bank 1 I thank Lourenco Paz and conference participants at the Southern Economic Association Meeting for helpful comments and suggestions. 1
1 Introduction Corruption is one of the most significant impediments to economic development. Many studies have suggested that corruption reduces human capital, increases crime, discourages investment, lowers the quality of public services and hinders the economic growth. Similarly, political instability is another major obstacle to economic development. However, relatively little is known about how corruption and political instability affect the firm s performance, in particular exporting. Firms might perceive corruption and political instability as an obstacle differently with their sizes, trade orientations and sectors. Empirical evidence on the effects of corruption and political instability on firms export decisions suggest that both of them decrease the probability that a firm only sells in the domestic market and increase the probability that a firm exports (Onley, 2016 & Kapri, 2018). Up to now, the two literatures have developed in a parallel fashion. But, the interaction effect of corruption and political instability has not been addressed. In this paper, I empirically examine whether the effect of corruption (political instability) on firm level export decisions depends on the extent of political instability (degree of corruptibility) to reduce the gap in the literature. This study is important because corruption and political instability are related phenomenon in developing countries and play a vital role in the business environment.. Corruption and political instability impose an additional fixed cost on firms and make them more costly to operate. As a result, fewer firms find it profitable to enter the market. Firms who can afford the extra fixed cost enter the foreign market. Firms who can t afford the extra fixed cost either sell only in the domestic market or exit the market. Hence, both corruption and political instability increase the likelihood that a firm enters the foreign market or exits the market. When the political instability is high, firms can influence the politician to get things done, 2
and hence the impact of corruption on firm decreases. On the other hand, when the degree of corruptibility is high, firms have to pay bribe to get things done, and hence the impact of political instability on firm decreases. To examine the joint effects of corruption and political instability on firm level export decisions, I use firm level data from World Bank s Enterprise Surveys unit. The surveys are conducted in many countries from different regions of the developing world. A random sample of firms is drawn from the population of manufacturing sectors in each country by size, region, and 2-digit industry. The survey years vary by country and region. The sub-sample I use includes a rich set of information on firms from South Asian countries. The estimation strategy used in this paper takes 2SLS approach. In the first stage, corruption and political instability are estimated using instrumental variables with other control variables. In the second stage, I use probit model to explore the interaction effect of corruption and political instability. My results suggest that both corruption and political instability obstacle increases the probability that a firm enters the foreign market. In particular, my estimates suggest that for a one-unit increase in corruption, we expect 7 percentage points increase in the probability of exporting. Also, for a one-unit increase in the political instability, we expect 32.2 percentage points increase in the probability of exporting. Further, I find that the impact of political instability (corruption) on firm s export decision decreases if the level of corruption (political instability) is high. Thus, the estimation results conclude that corruption and political instability increase the firm s probability of entering the foreign market and the impact of corruption (political instability) decreases when the level of political instability (corruption) is high. 3
This article contributes to the existing literature in three important ways. First, to the best of my knowledge, this is the first article to examine the interaction effect of the corruption and political instability on the firm s decision to enter the foreign market (i.e., becoming an exporter or a global firm). Second, this paper proposes various instruments for corruption and political instability to address endogeneity issues. Third, these results have important policy implications for developing countries, in particular, South Asian region where corruption prevalence is relatively high. Corruption and PI do impose a cost in the economy so that fewer firms (who are more productive, i.e., larger) will be able enter the market. In other words, many firms (who cannot afford such kind of additional cost) have to leave the market. The remainder of the paper is organized as follows. Section 2 explores the link between political instability, corruption and export decisions, while Section 3 describes the data. Section 4 discusses the empirical strategy. Section 5 presents the results. Finally, section 6 concludes. 2 Political Instability, Corruption, and Export Decisions In this section I describe a simple conceptual framework with which I examine the interaction effect of corruption and political instability (PI) on firm s export decisions. The insight of this framework comes from Melitz s Model (2003) based on heterogeneous firms. According to this model, heterogeneous firms differ in terms of their productivities and there are fixed costs associated with exporting. Olney (2016) extends the Melitz s model (2003) and finds that corruption decreases the probability that a firm will only sell domestically. Moreover, conditional on exporting, corruption increases the likelihood that a firm indirectly exports and decreases the likelihood that a firm directly exports. This leads me to test the following hypothesis: 4
Hypothesis 1: An increase in corruption increases the likelihood that a firm enters the foreign market. Relatively little is known about how PI affects the firm s performance. Amr (2017) finds that the perception of PI has a negative causal effect on firms sales and employment growth. Based on the author s claim, that is the first paper that tests the causal effect of PI on firm s performance. To the best of my knowledge, very little is known about the relationship between PI and firm s export decisions (Kapri, 2018). Since PI reduces firm s performance, under this situation, either firm has to increase performance or leave the market. When the existing firms increase performance, I believe, they enter the foreign market. Also, when firms perceive PI as a significant obstacle, it contributes an additional cost to the production process (Kapri, 2018). Towards this end, even though this relationship is not the main contribution of this article but supporting, I test the following hypothesis: Hypothesis 2: An increase in political instability increases the likelihood that a firm enters the foreign market. When the level of PI is high, firms can influence the politician to get things done, and hence the impact of corruption on firms exporting decisions would be lower compared to the situation when the level of PI is low. If the level of PI is low, i.e. more stable system, firms have to go through the bureaucratic channels and have to pay bribe to government officials to get things done fast. On the other hand, when the degree of corruptibility is high, firms have to pay bribe to get things done (political channel may not work) and hence, the impact of PI on firms exporting decision would go down. Similarly, when the degree of corruption is low, firms use the politician to get things done, and hence, the impact of PI on firms exporting decision would go 5
up. Thus, corruption attribute and political instability are related phenomena in developing countries. Campante et al. (2009) develop the model and investigate the relationship between the level of corruption and the degree of political instability. Using a country level data the paper finds U-shaped relationship between country indices of corruption perception and various measures of political stability including average tenures of chief executives and government parties. Fredrikson and Svensson (2003) develop the model on PI, corruption and policy formation; and find that PI has a negative impact on the rigorousness of environmental regulations if the level of corruption is low, but a positive effect when the degree of corruption is high. The paper also finds that Corruption reduces the rigidity of environmental regulations, but the effect disappears with higher level of PI. This leads me to test the following hypothesis: Hypothesis 3: Political instability reduces the impact of corruption on firm-level export decision. And, Corruption reduces the impact of Political Instability on firm-level export decision. 3 Data The firm-level data used in this paper are collected and maintained by World Bank s Enterprise Surveys unit. The surveys are conducted in many countries from different regions of the developing world. A random sample of firms is drawn from the population of manufacturing sectors in each country by size, region, and two-digit industry. The survey years vary by country and region. The pooled cross-sectional data used in the study includes years 2007, 2008, 2009, 6
2011, 2013 and 2014. 2 A total of 12,340 firms from 6 countries (Afghanistan, Bangladesh, India, Nepal, Pakistan and Srilanka) of South Asian region are used in the analysis. 3 Table 1 shows numbers of firms by country and year. The World Bank s Enterprise Surveys are unique because, in addition to standard information on production, it provides information on firm s business environment. In particular, the survey collects information on corruption and political instability. The corruption variable used in this paper is an answer to the question- To what degree is corruption an obstacle to the current operations? Firms had the choice of selecting one of the five options: (a) no obstacle, (b) minor obstacle, (c) moderate obstacle, (d) major obstacle and (e) very severe obstacle. Whereas no obstacle takes value 0 and very severe obstacle takes value 4. Similarly, the political instability variable used in this paper is an answer to the question- To what degree is political instability an obstacle to the current operations? Firms had the choice of selecting one of the five options: ranges from No obstacle (taking a value of 0) to Very severe obstacle (taking a value of 4). Other business environment related variables include informal sector obstacle, informal sector competition, bribes requested by officials, losses due to the TRVA (theft, robbery, vandalism or arson), and many others. In addition to collecting information on the business environment faced by firms, the survey collects the information on standard variables such as sales, employment, raw materials, labor cost, capital, imports, exports, age, use of international quality certificate, and foreign licenses and the gender of manager/owner. 4 The descriptive statistics are shown in the Table 2. 2 The surveys are conducted in different years in different countries. The countries and survey conducted years are given as follows. Afghanistan: 2008 and 2014; Bangladesh 2007 and 2013; India: 2014; Nepal 2009 and 2013; Pakistan 2007 and 2013 and Srilanka: 2011. 3 The number of firms used in the empirical analysis varies across specifications due to the missing observations for variables of interest. 4 The data in the surveys refers to the previous fiscal year of surveys conducted years. Also, surveys include lag 3 sales and labor variable. 7
From the table we can see that numbers of firms used in the empirical analysis are different. This is due to the missing observation of variables of interests. Data in the surveys are expressed in local currencies. Those local currencies are converted into US dollars using the exchange rate presented in the International Financial Statistics database. Nominal values are deflated using the GDP deflator (base year 2010) from the World Bank Development Indicators database. 4 Econometric Strategy 4.1 Effect of corruption on exporting To examine the effect of corruption on firm-level export decision I estimate the following econometric specification: Export_dum ijtc = β 0 + β 1 Corrupt_Obstacle ijtc + γ X ijtc + δi f + λi j + θi c +µi t +ε ijtc (1) where Export_dum ijtc is the export dummy variable for firm i in two-digit industry j, year t, and country c. It takes value 1 if the firm exports and 0 otherwise. Corrupt_Obstacle ijtc is the corruption obstacle variable takes values from 0 (no obstacle) to 4 (very severe obstacle). X ijtc is a set of firm-level control variables that includes the natural logarithm of sales per worker, age, age squared, and indicators whether a firm has foreign technology, international quality certificate, female owner and female manager. Finally, I f, I j, I c, and I t are firm-size 5 fixed effects, industry fixed effects, country fixed effects and survey year fixed effects while ε ijtc is a classical error term. Equation (1) is estimated using a probit regression since the dependent variable is binary. I also estimate this specification using global firm dummy as a dependent variable. If a firm enters the foreign market as an exporter, importer, or both, that firm is defined as a global firm in this paper. 4.2 Instrument for Corruption Obstacle 5 Firm size is defined by the number of employees, where 5-19 employees indicates the firm is small, 20-99 is medium, and 100+ is large. 8
The goal of the empirical specification is to identify the causal effect of corruption on firm-level export decision. Since the corruption is related to business environment characteristics and exporting is a firm-specific decision, it is likely that equation (1) is identifying this causal effect. Also, the corruption variable here I use does not include the actual bribes paid by firms. Moreover, the specification includes firm-level controls, firm size fixed effects, industry fixed effects, country fixed effects, and year fixed effects that further reduce the problem of endogeneity. However, there may still some concerns about the identification of the empirical specification that could lead biased estimates. The first concern to identification is the possibility that both firm-level decisions and corruption are correlated with unobservable firm characteristics. Given that the data are a pooled cross-section, the specification does not include firm fixed effects to control for time-invariant firm unobservable characteristics. The second threat to identification is reverse causality. A reverse causality bias may exist as low performing firms (less productive firms) tend to report severe corruption obstacle whereas high performing firms (more productive firms, i.e. exporters/global firms) have the enough resources to overcome obstacles, as a result they may report no obstacle or very low obstacle. The third threat to identification is measurement error. Measurement error is likely a problem because corruption is a kind of secretive issue and thus becomes a noisy variable. To overcome potential endogeneity, I employ an Instrumental Variable (IV) approach. In particular, this analysis uses four different instruments. First two firm-level instruments for corruption obstacle, following Pless and Fell (2017), are informal sector obstacle and informal sector competition. The informal sector obstacle is a measure of how severe the informal sector obstacle to operate firms. Note that the informal sector is a major engine for employment 9
and growth in many developing countries (Scheider, 2002). In the survey, firms were asked to rank the severity of this obstacle on a scale from 1 (not severe) to 4 (very severe). I believe that informal sector obstacle is a valid instrument because it is not directly linked to the firm s own management skills, i.e. exporting. And, it is possible that this obstacle is directly related to good governance, and as a result, corruption variable that I use in this study. The survey also has another question related to informal sector. Firms were asked whether they have to compete against unregistered or informal firms. If the firm said yes, the informal sector competition is equal to one and a zero otherwise. This is likely related to corruption obstacle since there might be an incentive to supply bribes to secure services if the firm realizes that it is competing with the informal sector (Pless and Fell, 2017). But again this is not directly linked to the firm s own management skills. The third instrument for corruption obstacle that I use is the average corruption level of other firms within the same country-industry-year as in Fisman and Svensson (2007). In particular, for each country-industry-year, the average level of corruption of firms not including firm i itself is calculated. This average level of corruption is used as an instrument for firm i's corruption obstacle. The fourth instrument used for this analysis is number of bribes requested as in Olney (2016). To construct this instrument, I use firm-level information on whether gifts or any kind of informal payments were expected by government officials. The survey provides seven different kinds of binary variables that indicate whether a bribe was requested in order to obtain an electricity connection, a water connection, a construction permit, an import license, an operating license, a government inspection, and a get things done. Due to the nature of firms, they were requested bribes for few or all kinds of services. So, total number of bribes ( number of bribes 10
requested variable) is constructed based on how many bribes were requested by government officials. Of course, this is directly related to the corruption obstacle but not the firm s management skills. Hence, utilizing these four IVs, corruption obstacle is estimated in the first stage. And then the econometric specification (1) is re-estimated using estimated corruption obstacle in the second stage. The subsequent results indicate that all of these instruments are strong enough. They only affect the firm-level export decisions through their impact on corruption obstacle. 4.3 Effects of political instability on exporting To examine the effect of PI on firm-level export decision I estimate the following econometric specification: Export_dum ijtc = β 0 + β 1 PI_Obstacle ijtc + γ X ijtc + δi f + λi j + θi c +µi t +ε ijtc (2) where PI ijtc is the political instability obstacle for firm i in two-digit industry j, year t, and country c. It takes values from 0 (no obstacle) to 4 (very severe obstacle). As I have discussed in the previous sub-section 4.2, there might be few endogeneity issues. To address these issues, I use two instruments: Losses due to the theft, robbery, vandalism or arson, and the average PI of other firms within the same country-industry-year. Equation (2) is also estimated using probit regression. 4.4 Interaction effect of corruption and political instability on exporting After finding the causal effect of corruption and political instability on a firm s export decisions, I examine the interaction effect of corruption and PI to investigate whether the PI (corruption) reduces the impact of corruption (PI) on a firm s export decisions using following econometric specification: 11
Export_dum ijtc = β 0 + β 1 Corrupt_Obstacle ijtc + β 2 PI_dum ijtc + β 3 Corrupt_Obstacle ijtc * PI_dum ijtc +γ X ijtc + δi f + λi j + θi c +µi t +ε ijtc (3) where PI_dum ijtc is the political instability obstacle dummy for firm i in two-digit industry j, year t, and country c. It takes value 1 if the PI ranges from 2 (major obstacle) to 4 (very severe obstacle), and 0 otherwise. I also estimate the equation with the interaction between PI and corruption obstacle dummy to find whether the corruption reduces the impact of PI on a firm s export decisions. Corruption dummy takes value 1 if corruption ranges from 2 (major obstacle) to 4 (very severe obstacle), and 0 otherwise. This equation (3) is also estimated using probit regression. 4.5 Pseudo-Panel Regression Analysis (Need to work on it?) As I have mentioned in Sub-section 4.2, export decisions could be correlated with shocks at the firm level over the time period, and the pooled cross-sectional data do not allow us to include firm fixed effects as we do in a traditional panel data. Under this circumstance, I create means-based pseudo panel in which cohorts of firms with similar characteristics are tracked over time. Deaton (1985) showed that the pseudo panel has the advantages of less requirements of construction of the data set. 6 Basically, this pseudo panel is very useful to analyze the micro data from developing world since it is hard to find true panel over there. I use three firm-level characteristics namely; firm size, industry and country to develop cohorts. I choose these three firm-level attributes because they capture the bigger share of corruption climate as a proxy for the propensity to engage in corrupt activities (Pless and Fell, 6 Deaton (1985) identified four advantages of pseudo panel. First, data from different sources can be combined into a single data set of pseudo panel. Second, attrition problem is minimized compared to the true panel. Third, individual s response error is controlled by using cohort means. Fourth, it is easy to handle from individual data to large macro data. 12
2017). Also, these are time-invariant firm-level characteristics. The econometric specification for pseudo-panel model is written as: Export _ dum = β + β Corrupt _ obstacle + γx + ηi + κi + ε (4) ct 0 2 ct ct c t ct where Export _ dum is the average value of all observed export dummy within cohort c in time ct t. Similarly, other variables are also averages of observed values with cohort c in time t. I c is cohort-level fixed effect to control for time-invariant unobservable characteristics at the cohort level. I t is time fixed effect. 5 Results In this section, I look at the effects of corruption and PI on firm-level decision to enter the foreign market. First, I examine the impact of corruption and PI (without instruments) as in equations (1) and (2). The estimated marginal effects (probit model) are presented in Table 3. Columns 1, 3 and 5 have export dummy dependent variable and columns 2, 4 and 6 have global firm dummy dependent variable. All regressions include firm-size effects, industry effects, country effects, year effects, and other firm-level controls (such as sales per worker (lag 3), age, age squared, foreign technology dummy, international quality certificate dummy, female manager dummy and female owner dummy). The result in the first column indicates that if the corruption increases by one unit, the firm s probability of entering the export market increases by 1.1 percentage points. Similarly, if the corruption increases by one unit, the firm s probability of becoming a global firm (exporting, importing, or both) increases by 1.4 percentage points (column 2). Columns (3) and (4) show the impact of PI on firms export decision. These impacts are almost the same as the impact of corruption on exporting. When the regressions include both corruption and PI, columns (5) and (6), the impact of both variables corruption and PI on exporting decreases and the impact of PI is slightly higher than the impact of corruption. In 13
overall, this table shows that both the corruption and PI increase the likelihood that a firm is an exporter (a global firm) relative to being purely domestic. Table 5 presents the results using instrumental variables. 7 In this table we can see that firm s probability of entering the export market (becoming a global firm) increases with an increase in both corruption and PI. For instance, for a one-unit increase in the corruption, we expect 7 percentage point increase in the firm s probability of entering the export market. All regressions include firm-size effects, industry effects, country effects, year effects, and other firm-level controls (such as sales per worker (lag 3), age, age squared, foreign technology dummy, international quality certificate dummy, female manager dummy and female owner dummy) as in Table 3. When endogeneity issues are addressed, the impact of corruption and PI on export dummy (global firm dummy) becomes bigger. Table 6 presents the estimated results using the equation (3). In this table, we can see that the sign of the interaction term is negative. Even though, coefficients of interaction terms are not statistically significant, interaction term and other two variables are jointly significant. Hence, this indicates that in the corrupt climate the impact of PI on firm s export decision decreases. Similarly, if the PI is high, the effect of corruption on firm s export decision decreases. Thus, the estimation results conclude that corruption and PI increase the firm s probability of entering the foreign market and the impact of corruption (PI) decreases when the level of PI (corruption) is high. 6 Conclusion This paper contributes to a growing literature documenting the impact of corruption and PI on the firm-level decisions to enter the foreign market. The first insight of the paper is to know how corruption affects the firm s decision to enter the foreign market. The second insight 7 First stage results are presented in Table 4. 14
of the paper is to know how PI affects the firm s decision to enter the foreign market. The final, and the main, insight of the paper is to know the firm s export decision taking into consideration the degrees of corruptibility and political turbulence, i.e. the interaction effects of corruption and PI on the firm s decision to enter the foreign market. The paper used the World Bank s enterprises survey data to test proposed hypotheses using probit model. Endogeneity issues are addressed using instrumental variables. The results show that corruption and PI positively affect the firm-level decisions to enter the foreign market, and the impact of PI is higher than the impact of corruption. The impact of PI on firm-level decisions entering the global market is lower in the corrupt climate than the non-corrupt climate. And, the impact of corruption on firm-level decisions entering the global market is higher with low level of PI. This article makes two important contributions. First, to the best of my knowledge, this is the only article to examine the interaction effect of the corruption and PI on the firm s decision to enter the foreign market. Second, this analysis yields meaningful policy implication regarding PI, corruption and a firm-level decision entering foreign market in developing countries, in particular, South Asian region where corruption prevalence is relatively high. Corruption and PI do impose a cost in the economy so that fewer firms (who are more productive) will be able enter the market, and furthermore the foreign market. In other words, many firms (who cannot afford such kind of additional cost) have to leave the market. 15
References Campante, Filipe R., Davin Chor, and Quoc-Anh Do. 2009. Instability and the Incentives for Corruption. Economics & Politics, 21: 42-92. Deaton, Angus. 1985. Panel data from time series of cross sections. Journal of Econometrics, 30: 109-126. Fisman, Raymond, and Jakob Svensson. 2007. Are corruption and Taxation Really Harmful to Growth? Firm Level Evidence. Journal of Development Economics, 83(1): 63-75. Fredriksson Per, and Jakob Svensson. 2003. Political Instability, Corruption and Policy Formation: the Case of Environmental Policy. Journal of Public Economics, 87: 1383-1405. Hosny, Amr. 2017. Political Stability, Firm Characteristics and Performance: Evidence from 6,083 Private Firms in the Middle East. Review of Middle East Economics and Finance, 13(1): 1-21. Kapri, Kul. 2019. Impact of Political Stability on Firm-Level Export Decisions. International Review of Economics and Finance. Melitz, Marc. 2003. The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity. Econometrica, 71(6): 1695-1725. Olney, William W. 2016. Impact of Corruption on Firm-Level Exprot Decisions. Economic Inquiry, 54(2): 1105-1127. Pless, Jacquenlyn, and Harrison Fell. 2017. Bribes, Bureaucracies, and Blackouts: Towards Understanding How Corruption at the Firm Level Impacts Electricity Reliability. Resource and Energy Economics, 47: 36-55. Schneider, Friedrich. 2002. Size and Measurement of the Informal Economy in 110 countries Around the World. Research Working Paper, World Bank. 16
Table 1: Number of Firms by Country Country Year Number of Firms Afghanistan 2008 120 Afghanistan 2014 139 Bangladesh 2007 1193 Bangladesh 2013 1179 India 2014 7162 Nepal 2009 136 Nepal 2013 241 Pakistan 2007 784 Pakistan 2013 1024 SriLanka 2011 362 17
Table 2: Summary Statistics Variable Obs Mean Std. Dev. Min Max Export Dummy 12340 0.216208 0.411674 0 1 Global Firm Dummy 12109 0.335949 0.472341 0 1 Log of Sales per Worker_Lag3 10174 9.765471 1.39297 3.855806 15.20425 Age 11426 20.28829 14.45816 0 162 Age Squared 11426 620.6349 1049.423 0 26244 Foreign Technology Dummy 12232 0.103417 0.304516 0 1 International Quality Certificate Dummy 12151 0.378158 0.484947 0 1 Female Manager Dummy 10343 0.068065 0.25187 0 1 Female Owner Dummy 12295 0.16462 0.370852 0 1 Corruption Obstacle 12212 2.244022 1.324347 0 4 Political Instability 12235 1.715897 1.436022 0 4 Informal Sector Competition 10177 0.400413 0.490006 0 1 Informal Sector Obstacle 11937 1.074055 1.156798 0 4 Electricity Connection Bribe Dummy 610 0.404918 0.491279 0 1 Water Connection Bribe Dummy 165 0.339394 0.474945 0 1 Construction Permit Bribe Dummy 668 0.371258 0.483503 0 1 Inspection Bribe Dummy 6337 0.279154 0.448619 0 1 Import License Bribe Dummy 947 0.379092 0.485417 0 1 Operating License Bribe Dummy 2826 0.283086 0.450577 0 1 Losses Due to TRVA Dummy 12290 0.06607 0.248415 0 1 Note: All monetary values are in US dollar deflated by GDP deflator, where 2010 is the base year. 18
Table 3: Effect of corruption and PI on firm-level export decisions (without IVs, probit model) Dependent variables: Export Dummy (col. 1, 3 and 5) and Global Firm Dummy (col. 2, 4 and 6) VARIABLES (1) (2) (3) (4) (5) (6) Global Global Global Export Firm Export Firm Export Firm Dummy Dummy Dummy Dummy Dummy Dummy Corruption Obstacle 0.011*** 0.014*** 0.007* 0.008 (0.004) (0.005) (0.004) (0.006) Political Instability Obstacle 0.011*** 0.019*** 0.008** 0.015*** (0.003) (0.003) (0.004) (0.005) Log of Sales per Worker_Lag3 0.023*** 0.038*** 0.022*** 0.038*** 0.023*** 0.038*** (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) Age 0.003*** 0.003*** 0.003*** 0.003*** 0.003*** 0.003*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Age Squared -0.000** -0.000-0.000** -0.000* -0.000** -0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Foreign Technology Dummy 0.016 0.035* 0.013 0.031 0.015 0.033* (0.014) (0.019) (0.014) (0.020) (0.014) (0.019) International Quality Certificate Dummy 0.100*** 0.091*** 0.101*** 0.094*** 0.101*** 0.092*** (0.012) (0.012) (0.012) (0.012) (0.012) (0.012) Female Manager Dummy 0.037*** 0.052*** 0.037*** 0.051*** 0.036*** 0.050*** (0.013) (0.015) (0.014) (0.015) (0.013) (0.015) Female Owner Dummy 0.053*** 0.075*** 0.054*** 0.077*** 0.054*** 0.076*** (0.011) (0.013) (0.011) (0.013) (0.011) (0.013) Firm Size Effects Yes Yes Yes Yes Yes Yes Industry Effects Yes Yes Yes Yes Yes Yes Country Effects Yes Yes Yes Yes Yes Yes Year Effects Yes Yes Yes Yes Yes Yes Observations 8,454 8,437 8,487 8,471 8,439 8,424 Reported values are estimated marginal effects (Probit model) with robust standard errors clustered at the countryindustry-year level in brackets. All regressions include firm-size effects, country effects, industry effects, and year effects. Dependent variables are export dummy and global firm dummy. *** p<0.01, ** p<0.05, * p<0.1. 19
Table 4: Estimation of corruption obstacle and PI obstacle using IVs VARIABLES Corruption Obstacle PI Obstacle Informal Sector Obstacle 0.111*** (0.022) Informal Sector Competition Dummy -0.043 (0.043) Average Corruption Obstacle by CIY 0.097 (0.150) Number of Bribes Requested 0.143*** (0.029) Average PI Obstacle by CIY 0.212* (0.112) Losses TRVA Dummy -0.055 (0.059) Constant 1.386*** 2.208*** (0.401) (0.363) Observations 8,282 8,463 R-squared 0.076 0.306 Number of Clusters 100 100 Reported values are estimated by OLS and include firm-size effects, country effects, industry effects, and year effects. Regressions also include sales per worker, age. age squared, foreign technology dummy, international quality certificate dummy, female manager dummy and female owner dummy. Standard errors in parentheses are clustered at the country- industry-year level. *** p<0.01, ** p<0.05, * p<0.1. 20
Table 5: Effect of corruption and PI on firm-level export decisions ( probit IVs results) Dependent variables: Export Dummy (col. 1, 3 and 5) and Global Firm Dummy (col. 2, 4 and 6) VARIABLES (1) (2) (3) (4) (5) (6) Export Dummy Global Firm Dummy Export Dummy Global Firm Dummy Export Dummy Global Firm Dummy Estimated Corruption Obstacle 0.070** 0.088** 0.068** 0.086** (0.031) (0.036) (0.030) (0.036) Estimated Political Instability Obstacle 0.322** 0.286* 0.311* 0.325** (0.154) (0.156) (0.161) (0.164) Log of Sales per Worker_Lag3 0.021*** 0.035*** 0.025*** 0.041*** 0.024*** 0.039*** (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) Age 0.003*** 0.003*** 0.003*** 0.003*** 0.003*** 0.003*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Age Squared -0.000* -0.000-0.000** -0.000* -0.000** -0.000* (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Foreign Technology Dummy 0.029** 0.052*** 0.020 0.037* 0.035** 0.059*** (0.013) (0.016) (0.016) (0.021) (0.014) (0.017) International Quality Certificate Dummy 0.097*** 0.086*** 0.110*** 0.101*** 0.106*** 0.096*** (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) Female Manager Dummy 0.022 0.035** -0.043-0.017-0.058-0.048 (0.015) (0.017) (0.044) (0.045) (0.046) (0.047) Female Owner Dummy 0.055*** 0.077*** 0.091*** 0.109*** 0.090*** 0.114*** (0.011) (0.013) (0.021) (0.022) (0.022) (0.022) Firm Size Effects Yes Yes Yes Yes Yes Yes Industry Effects Yes Yes Yes Yes Yes Yes Country Effects Yes Yes Yes Yes Yes Yes Year Effects Yes Yes Yes Yes Yes Yes Observations 8,275 8,258 8,456 8,440 8,246 8,231 Reported values are estimated marginal effects (Probit model) with robust standard errors clustered at the countryindustry-year level in brackets. All regressions include firm-size effects, country effects, industry effects, and year effects. Dependent variables are export dummy and global firm dummy. *** p<0.01, ** p<0.05, * p<0.1. 21
Table 6: Effect of interaction of corruption and political instability on firm-level export decisions Dependent variables: Export Dummy (col. 1 and 3) and Global Firm Dummy (col. 2 and 4) VARIABLES (1) (2) (3) (4) Export Dummy Global Firm Dummy Export Dummy Global Firm Dummy Estimated Corruption Obstacle 0.070** 0.095** (0.035) (0.041) PI Obstacle Dummy 0.082 0.147** (0.052) (0.064) Estimated Corruption Obstacle * PI Obstacle Dummy -0.021-0.043 (0.023) (0.028) Estimated PI Obstacle 0.336** 0.303* (0.157) (0.162) Corruption Obstacle Dummy 0.050** 0.054* (0.022) (0.028) Estimated PI Obstacle *Corrupt. Obstacle Dummy -0.014-0.012 (0.011) (0.016) Log of Sales per Worker_Lag3 0.021*** 0.036*** 0.026*** 0.041*** (0.005) (0.005) (0.005) (0.006) Age 0.003*** 0.003*** 0.003*** 0.003*** (0.001) (0.001) (0.001) (0.001) Age Squared -0.000* -0.000-0.000** -0.000* (0.000) (0.000) (0.000) (0.000) Foreign Technology Dummy 0.028** 0.052*** 0.022 0.039* (0.012) (0.016) (0.016) (0.021) International Quality Certificate Dummy 0.097*** 0.087*** 0.109*** 0.099*** (0.013) (0.013) (0.013) (0.012) Female Manager Dummy 0.021 0.034** -0.045-0.022 (0.015) (0.017) (0.044) (0.047) Female Owner Dummy 0.055*** 0.077*** 0.091*** 0.109*** (0.011) (0.013) (0.021) (0.022) Observations 8,260 8,245 8,408 8,393 Reported values are estimated marginal effects (Probit model) with robust standard errors clustered at the countryindustry-year level in brackets. All regressions include firm-size effects, country effects, industry effects, and year effects. Dependent variables are export dummy and global firm dummy. *** p<0.01, ** p<0.05, * p<0.1. 22