Corruption and Firm Growth - Evidence From Around the World (Figures)

Figure 1: Relationship between BribeShare and growth – flexible functional form Source: Enterprise Surveys and authors’ calculations. Note: This figure graphs the coefficients from a regression of sales growth on a finely partitioned set of indicator variables for BribeShare (the first column in Table A15). The dependent variable is annual growth of sales (half of the difference between the logarithm of sales in the last fiscal year and the logarithm of sales three fiscal years ago, both expressed in US dollars at market exchange rates). Control variables include region-sector-year fixed effects, ownership (state/foreign/domestic), the logarithm of firm age, exporter status, propensity to complain (the kvetch index), the logarithm of employment three years ago, the logarithm of sales in US dollars three years ago. We include bars to indicate the 95% confidence intervals of the coefficient estimates, with standard errors clustered at the level of industry*region*survey wave. Table 1: Selection into paying bribes Dep. var: Non-zero BribeShare Sales 3 years ago, USD, log Profit/sales (1) (2) (3) 0.000632 (0.000918) 0.000427 (0.000955) -0.000593 (0.00197) 0.0263** (0.0125) 0.00724* (0.00423) -0.0281** (0.0111) 0.0162** (0.00685) 0.0352 (0.0383) -0.00342 (0.00358) Propensity to complain (kvetch index) 0.0131*** (0.00200) -0.0237*** (0.00568) 0.0131*** (0.00338) -0.000810 (0.0117) -0.00235 (0.00196) Foreign ownership Exporter State ownership Age, years, log R-squared Observations Region-Sector-Year FE Firm-level controls 0.296 87829 Yes No 0.297 87829 Yes All 0.343 14844 Yes All Source: Enterprise Surveys and authors’ calculations. Note: Standard errors in parentheses are clustered at the level of industry*region*survey wave. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively. Estimated using ordinary least squares. The dependent variable is a dummy variable indicating whether the firm paid non-zero bribes. All regressions control for region-sector-year fixed effects. Table 2: BribeShare and growth of sales Dep. var.: Growth of sales BribeShare Non-zero BribeShare 1 -0.053 (0.052) Informal payment categories 0 < b ≤ 0.5% 2 3 4 0.010** (0.004) -0.139** (0.060) 0.018*** (0.005) -0.232*** (0.058) 0.021*** (0.005) 0.047*** (0.008) 0.015** (0.007) -0.001 (0.007) 0.005 (0.009) -0.015* (0.009) -0.025 (0.017) 0.5 < b ≤ 1% 1 < b ≤ 5% 5 < b ≤ 10% 10 < b ≤ 20% > 20% R-squared Observations Region-Sector-Year FE Firm-level controls 5 0.237 87829 Yes 0.237 87829 Yes 0.237 87829 Yes 0.293 87829 Yes All 0.294 87829 Yes All Source: Enterprise Surveys and authors’ calculations. Note: Standard errors in parentheses are clustered at the level of industry*region*survey wave. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively. Estimated using ordinary least squares. The dependent variable is annual growth of sales (half of the difference between the logarithm of sales in the last fiscal year and the logarithm of sales three fiscal years ago, both expressed in US dollars at market exchange rates). All regressions control for region-sector-year fixed effects. Specifications in columns 4-5 also control for ownership (state/foreign/domestic), the logarithm of firm age, exporter status, propensity to complain (the kvetch index), the logarithm of employment three years ago, the logarithm of sales in US dollars three years ago. In column (5), the omitted category is BribeShare = 0. Table 3: High-corruption vs low-corruption economies Dep. var.: Growth of sales BribeShare Non-zero BribeShare R-squared Observations Region-Sector-Year FE Firm-level controls High corruption (1) Low corruption (2) -0.194** (0.076) 0.030*** (0.007) -0.301*** (0.084) 0.007 (0.007) 0.306 42341 Yes All 0.275 45488 Yes All Source: Enterprise Surveys and authors’ calculations. Note: Standard errors in parentheses are clustered at the level of industry*region*survey wave. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively. Estimated using ordinary least squares. The dependent variable is annual growth of sales (half of the difference between the logarithm of sales in the last fiscal year and the logarithm of sales three fiscal years ago, both expressed in US dollars at market exchange rates). All regressions control for ownership (state/foreign/domestic), the logarithm of firm age, exporter status, propensity to complain (the kvetch index), the logarithm of employment three years ago, the logarithm of sales in US dollars three years ago, region-sector-year fixed effects. Lowcorruption economies are those with the Worldwide Governance Indicator of control of corruption of -0.43 or above (sample median); the rest are high-corruption economies. Table 4: Firms that do and do not complain about corruption Dep. var.: Growth of sales BribeShare Non-zero BribeShare Propensity to complain (kvetch index) R-squared Observations Region-Sector-Year FE Firm-level controls Corruption at least a moderate obstacle (1) Corruption a minor or no obstacle (2) -0.284*** (0.069) 0.025*** (0.006) 0.002 (0.002) -0.037 (0.111) 0.015* (0.008) 0.007*** (0.003) 0.337 44572 Yes All 0.303 41591 Yes All Source: Enterprise Surveys and authors’ calculations. Note: Standard errors in parentheses are clustered at the level of industry*region*survey wave. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively. Estimated using ordinary least squares. The dependent variable is annual growth of sales (half of the difference between the logarithm of sales in the last fiscal year and the logarithm of sales three fiscal years ago, both expressed in US dollars at market exchange rates). All regressions control for ownership (state/foreign/domestic), the logarithm of firm age, exporter status, propensity to complain (the kvetch index), the logarithm of employment three years ago, the logarithm of sales in US dollars three years ago, region-sector-year fixed effects. Column 1 includes only businesses that state that corruption is at least a moderate obstacle, column 2 includes businesses that state that corruption is no obstacle or a minor obstacle to their operations. Appendix Figures and Tables Figure A1: Correlations between BribeShare and corruption as an obstacle Source: Enterprise Surveys and authors’ calculations. Note: Includes observations in the estimation sample only. For countries with multiple surveys, we use the average Control of Corruption score across the survey years. Marker size indicates the number of underlying observations. The “Corruption is an obstacle” question is answered on a 5-point scale from 0 (“not an obstacle”) to 4 (“very severe obstacle”). Figure A2: Distribution of BribeShare between 0 and 10% Source: Enterprise Surveys and authors’ calculations. Note: The graph provides a histogram of BribeShare for all firms reporting bribes as a fraction of sales between 0 and 10%. Figure A3: Distribution of BribeShare between 0 and 0.1% Source: Enterprise Surveys and authors’ calculations. Note: The graph provides a histogram of BribeShare for all firms reporting bribes as a fraction of sales between 0 and 0.1%. Table A1: Correlates of missing data Dep. var.: Sales growth, per annum Exporter State ownership Age, years, log Log employees Missing informal payment (1) (2) 0.003 (0.002) 0.010*** (0.003) -0.008 (0.010) -0.000 (0.000) Control of corruption index Foreign ownership -0.001 (0.005) 0.016** (0.007) -0.022 (0.021) -0.000 (0.000) 0.006** (0.003) -0.011 (0.031) Missing sales growth (3) (4) -0.010*** (0.003) -0.002 (0.010) -0.002*** (0.000) -0.008*** (0.001) -0.011*** (0.003) -0.017* (0.009) 0.020*** (0.005) 0.009* (0.005) Firm age, years 1 0.719*** (0.013) 0.706*** (0.009) 0.442*** (0.013) 0.111*** (0.009) 0.021*** (0.006) 0.001 (0.005) 0.013** (0.005) -0.006 (0.005) -0.005 (0.005) 0.005 (0.005) 2 3 4 5 6 7 8 9 10 R-squared Observations Region-Sector-Year FE Sector-Year FE -0.003** (0.001) 0.277 112307 Yes No 0.023 114891 No Yes 0.252 149418 Yes No 0.334 149418 Yes No Source: Enterprise Surveys and authors’ calculations. Note: Standard errors in parentheses are clustered at the level of industry*region*survey wave. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively. Estimated using ordinary least squares. The dependent variable in columns 1 and 2 is a dummy variable equal to 1 for observations with missing BribeShare. The dependent variable in columns 3 and 4 is a dummy variable equal to 1 for observations with missing sales growth data. Table A2: Descriptive statistics Variables Mean St. dev. Median Min Max BribeShare Non-zero BribeShare Sales growth, per annum Labor productivity growth, per annum Sales growth, real terms, per annum Productivity growth, real terms, p.a. Employment growth, per annum Sales 3 years ago, USD, log Employment 3 years ago, USD, log Age, years, log Foreign ownership State ownership Exporter Propensity to complain (kvetch index) 0.008 0.148 0.066 0.019 0.035 -0.012 0.047 13.258 3.248 2.760 0.073 0.015 0.199 0.001 0.034 0.355 0.340 0.347 0.331 0.341 0.193 2.444 1.413 0.707 0.242 0.121 0.399 0.684 0.000 0.000 0.050 0.008 0.019 -0.023 0.000 13.215 2.996 2.773 0.000 0.000 0.000 -0.091 0.000 0.000 -1.969 -3.708 -2.142 -3.795 -4.736 0.411 0.000 0.000 0.000 0.000 0.000 -2.588 0.500 1.000 1.998 5.300 2.124 5.230 3.719 28.478 13.122 5.421 1.000 1.000 1.000 2.737 Source: Enterprise Surveys and authors’ calculations. Note: Based on 87,829 firms across 141 economies surveyed as part of Enterprise Surveys in 2006-20. Table A3: Informal payments over time Variable 2006-2010 Time period 2011-2015 2016-2020 Average bribe share 0.00992 0.00697 0.00610 Average bribe share for firms reporting non-zero bribes 0.0471 0.0493 0.0605 Share of firms reporting non-zero bribe share 0.211 0.141 0.101 Number of observations 25481 33746 28602 Source: Enterprise Surveys and authors’ calculations. Note: Averages across all firms surveyed during the time periods shown. Table A4: BribeShare and growth of sales, with weights Dep. var.: Growth of sales BribeShare Non-zero BribeShare (1) -0.152* (0.087) Informal payment categories 0 < b ≤ 0.1% (2) 0.014* (0.008) (3) -0.302*** (0.103) 0.031*** (0.010) (4) -0.398*** (0.105) 0.037*** (0.009) 0.111*** (0.022) 0.041** (0.016) 0.018 (0.015) 0.011 (0.013) 0.009 (0.019) -0.021 (0.014) 0.1 < b ≤ 0.5% 0.5 < b ≤ 1% 1 < b ≤ 5% 5 < b ≤ 10% > 10% R-squared Observations Region-Sector-Year FE Firm-level controls (5) 0.302 87286 Yes 0.302 87286 Yes 0.303 87286 Yes 0.361 87286 Yes All 0.361 87286 Yes All Source: Enterprise Surveys and authors’ calculations. Note: Standard errors in parentheses are clustered at the level of industry*region*survey wave. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively. Estimated using ordinary least squares. The dependent variable is annual growth of sales (half of the difference between the logarithm of sales in the last fiscal year and the logarithm of sales three fiscal years ago, both expressed in US dollars at market exchange rates). All regressions control for region-sector-year fixed effects. Specifications in columns 4-5 also control for ownership (state/foreign/domestic), the logarithm of firm age, exporter status, propensity to complain (the kvetch index), the logarithm of employment three years ago, the logarithm of sales in US dollars three years ago. In column (5), the omitted category is BribeShare = 0. Observations are weighted using survey weights. Table A5: Baseline regressions: coefficients on control variables Dep. var.: Growth of sales BribeShare Non-zero BribeShare Informal payment categories 0 < b ≤ 0.5% 1 -0.232*** (0.058) 0.021*** (0.005) -0.061*** (0.002) 0.056*** (0.002) 0.004** (0.002) 0.049*** (0.005) 0.045*** (0.003) 0.014 (0.012) -0.034*** (0.002) 0.776*** (0.019) 0.047*** (0.008) 0.015** (0.007) -0.001 (0.007) 0.005 (0.009) -0.015* (0.009) -0.025 (0.017) -0.061*** (0.002) 0.056*** (0.002) 0.004** (0.002) 0.048*** (0.005) 0.044*** (0.003) 0.013 (0.012) -0.034*** (0.002) 0.778*** (0.019) 0.293 87829 Yes All 0.294 87829 Yes All 0.5 < b ≤ 1% 1 < b ≤ 5% 5 < b ≤ 10% 10 < b ≤ 20% > 20% Sales 3 years ago, USD, log Employment 3 years ago, USD, log Propensity to complain (kvetch index) Foreign ownership Exporter State ownership Age, years, log Constant R-squared Observations Region-Sector-Year FE Firm-level controls 2 Source: Authors’ calculations. Note: Standard errors in parentheses are clustered at the level of industry*region*survey wave. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively. Estimated using ordinary least squares. The specifications are the same as those in Table 2’s columns (4) and (5). The dependent variable is annual growth of sales (half of the difference between the logarithm of sales in the last fiscal year and the logarithm of sales three fiscal years ago, both expressed in US dollars at market exchange rates). All regressions control for region-sector-year fixed effects, ownership (state/foreign/domestic), the logarithm of firm age, exporter status, propensity to complain (the kvetch index), the logarithm of employment three years ago, the logarithm of sales in US dollars three years ago. Table A6: High-corruption vs low-corruption economies by quartile Dep. var.: Growth of sales Very high corruption (Q1) (1) (2) High corruption (Q2) (3) (4) Low corruption (Q3) (5) (6) BribeShare -0.239** (0.106) 0.034*** (0.009) -0.146 (0.104) 0.026*** (0.009) -0.259*** (0.084) 0.016** (0.008) Non-zero BribeShare Informal payment categories 0 < b ≤ 0.1% 0.113*** (0.021) 0.030** (0.015) 0.024** (0.012) 0.003 (0.012) 0.012 (0.016) -0.001 (0.014) 0.1 < b ≤ 0.5% 0.5 < b ≤ 1% 1 < b ≤ 5% 5 < b ≤ 10% > 10% R-squared Observations Region-Sector-Year FE Firm-level controls 0.333 18750 Yes All 0.334 18750 Yes All 0.093*** (0.021) 0.020 (0.020) 0.018 (0.015) 0.008 (0.015) 0.024 (0.017) -0.013 (0.016) 0.280 21250 Yes All 0.281 21250 Yes All Very low corruption (Q4) (7) (8) -0.424** (0.187) -0.010 (0.013) 0.061*** (0.015) -0.011 (0.015) 0.006 (0.012) -0.000 (0.012) 0.007 (0.014) -0.023* (0.013) 0.284 30518 Yes All 0.284 30518 Yes All 0.036 (0.026) -0.060* (0.034) 0.021 (0.023) -0.045** (0.019) -0.077*** (0.026) -0.076** (0.031) 0.244 17311 Yes All 0.245 17311 Yes All Source: Enterprise Surveys, Polity IV and authors’ calculations. Note: Standard errors in parentheses are clustered at the level of industry*region*survey wave. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively. Estimated using ordinary least squares. The dependent variable is annual growth of sales (half of the difference between the logarithm of sales in the last fiscal year and the logarithm of sales three fiscal years ago, both expressed in US dollars at market exchange rates). All regressions control for ownership (state/foreign/domestic), the logarithm of firm age, exporter status, propensity to complain (the kvetch index), the logarithm of employment three years ago, the logarithm of sales in US dollars three years ago, region-sector-year fixed effects. Countries that are defined as having very high corruption have a control of corruption score of less than -0.83, countries in the second quartile of less than -0.45, countries in the third quartile of less than -0.02. All other countries are considered to have very low corruption. Table A7: Informal payments and productivity growth Dep. var.: Labor productivity growth BribeShare Non-zero BribeShare 1 -0.046 (0.053) Informal payment categories 0 < b ≤ 0.5% 2 3 4 0.008* (0.004) -0.112* (0.061) 0.014*** (0.005) -0.211*** (0.060) 0.017*** (0.005) -0.088*** (0.002) 0.025*** (0.001) 0.002 (0.002) 0.037*** (0.005) 0.015*** (0.003) -0.001 (0.012) -0.006*** (0.002) 0.831*** (0.021) 0.037*** (0.008) 0.007 (0.007) 0.000 (0.007) 0.003 (0.009) -0.010 (0.009) -0.032* (0.018) -0.088*** (0.002) 0.025*** (0.001) 0.002 (0.002) 0.037*** (0.005) 0.015*** (0.003) -0.001 (0.012) -0.006*** (0.002) 0.832*** (0.021) 0.305 87780 Yes All 0.305 87780 Yes All 0.5 < b ≤ 1% 1 < b ≤ 5% 5 < b ≤ 10% 10 < b ≤ 20% > 20% Log labor productivity t-2 (USD) Employment 3 years ago, USD, log Propensity to complain (kvetch index) Foreign ownership Exporter State ownership Age, years, log Constant R-squared Observations Region-Sector-Year FE Firm-level controls 0.020*** (0.000) 0.018*** (0.001) 0.018*** (0.001) 0.201 87780 Yes 0.201 87780 Yes 0.201 87780 Yes 5 Source: Authors’ calculations. Note: Standard errors in parentheses at clustered at the level of industry*region*survey wave. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively. Estimated using ordinary least squares. The dependent variable is annual growth of sales per worker (half of the difference between the logarithm of sales per worker in the last fiscal year and the logarithm of sales three fiscal years ago, both expressed in US dollars at market exchange rates). All regressions control for region-sector-year fixed effects. Specifications in columns 4-5 also control for ownership (state/foreign/domestic), the logarithm of firm age, exporter status, propensity to complain (the kvetch index), the logarithm of employment three years ago, the logarithm of sales in US dollars three years ago. Table A8: Informal payments and employment growth Dep. var.: Employment growth BribeShare Non-zero BribeShare 1 -0.010 (0.028) Informal payment categories 0 < b ≤ 0.5% 2 3 4 0.003 (0.003) -0.031 (0.032) 0.004 (0.003) -0.061** (0.031) 0.006** (0.003) -0.029*** (0.001) 0.003*** (0.001) 0.027*** (0.003) 0.039*** (0.002) 0.012** (0.006) -0.026*** (0.001) 0.203*** (0.005) 0.017*** (0.004) 0.009* (0.004) -0.005 (0.005) 0.001 (0.005) -0.008 (0.006) -0.001 (0.009) -0.029*** (0.001) 0.003*** (0.001) 0.027*** (0.003) 0.039*** (0.002) 0.012** (0.006) -0.026*** (0.001) 0.203*** (0.005) 0.180 87780 Yes All 0.180 87780 Yes All 0.5 < b ≤ 1% 1 < b ≤ 5% 5 < b ≤ 10% 10 < b ≤ 20% > 20% Employment 3 years ago, USD, log Propensity to complain (kvetch index) Foreign ownership Exporter State ownership Age, years, log Constant R-squared Observations Region-Sector-Year FE Firm-level controls 0.047*** (0.000) 0.046*** (0.000) 0.046*** (0.000) 0.135 87780 Yes 0.135 87780 Yes 0.135 87780 Yes 5 Source: Authors’ calculations. Note: Standard errors in parentheses at clustered at the level of industry*region*survey wave. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively. Estimated using ordinary least squares. The dependent variable is annual growth of employment (half of the difference between the logarithm of employment in the last fiscal year and the logarithm of employment three fiscal years ago. All regressions control for region-sector-year fixed effects. Specifications in columns 4-5 also control for ownership (state/foreign/domestic), the logarithm of firm age, exporter status, propensity to complain (the kvetch index), the logarithm of employment three years ago. Table A9: Firms that do and do not complain about corruption relative to the country average Dep. var.: Growth of sales BribeShare Non-zero BribeShare Propensity to complain (kvetch index) R-squared Observations Region-Sector-Year FE Firm-level controls Complain about corruption more than country average (1) Complain about corruption less than country average (2) -0.304*** (0.075) 0.026*** (0.007) 0.004 (0.003) -0.039 (0.101) 0.010 (0.007) 0.004* (0.002) 0.345 37437 Yes All 0.306 45915 Yes All Source: Enterprise Surveys and authors’ calculations. Note: Standard errors in parentheses are clustered at the level of industry*region*survey wave. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively. Estimated using ordinary least squares. The dependent variable is annual growth of sales (half of the difference between the logarithm of sales in the last fiscal year and the logarithm of sales three fiscal years ago, both expressed in US dollars at market exchange rates). All regressions control for ownership (state/foreign/domestic), the logarithm of firm age, exporter status, propensity to complain (the kvetch index), the logarithm of employment three years ago, the logarithm of sales in US dollars three years ago, region-sector-year fixed effects. Column 1 includes only businesses whose response to the question “How much of an obstacle is corruption" is higher than the country-year average, while column 2 only includes businesses where the response is lower. Table A10: Firms that do and do not complain about corruption, and do not pay bribes Dep. var.: Growth of sales Corruption at least a moderate obstacle Complain about corruption more than country mean (1) -0.012*** (0.003) (2) Propensity to complain (kvetch index) 0.006*** (0.002) -0.009*** (0.003) 0.006*** (0.002) R-squared Observations Region-Sector-Year FE Firm-level controls 0.294 71979 Yes All 0.294 71979 Yes All Source: Enterprise Surveys and authors’ calculations. Note: Standard errors in parentheses are clustered at the level of industry*region*survey wave. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively. Estimated using ordinary least squares. The dependent variable is annual growth of sales (half of the difference between the logarithm of sales in the last fiscal year and the logarithm of sales three fiscal years ago, both expressed in US dollars at market exchange rates). All regressions control for ownership (state/foreign/domestic), the logarithm of firm age, exporter status, propensity to complain (the kvetch index), the logarithm of employment three years ago, the logarithm of sales in US dollars three years ago, region-sector-year fixed effects. Includes only businesses that do not pay bribes. Table A11: Number of competitors Dep. var.: BribeShare BribeShare x Competition Non-zero BribeShare Non-zero BribeShare x Competition Number of competitors Growth of sales (1) BribeShare (3) 0.002 (0.318) -0.071 (0.121) 0.022 (0.024) -0.003 (0.009) -0.007** (0.003) Number of competitors 2-5 More than 5 R-squared Observations Sector FE Firm-level controls Non-zero bribe dummy (2) 0.267 55385 Yes All 0.024*** (0.006) 0.022*** (0.006) 0.001 (0.001) 0.001 (0.001) 0.268 55385 Yes All 0.173 55385 Yes All Source: Enterprise Surveys and authors’ calculations. Note: Standard errors in parentheses are clustered at the level of industry*region*survey wave. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively. Estimated using ordinary least squares. The dependent variable in column 1 is annual growth of sales (half of the difference between the logarithm of sales in the last fiscal year and the logarithm of sales three fiscal years ago, both expressed in US dollars at market exchange rates). In column 2, the dependent variable is the share of informal payments in sales and in column 3, it is a dummy variable indicating whether the business made non-zero informal payments. Competition is a categorical variable indicating whether the business faces zero or one, two to five, or more than five competitors in its main market. In column 2, zero or one competitors is the baseline category. All regressions control for ownership (state/foreign/domestic), the logarithm of firm age, exporter status, propensity to complain (the kvetch index), the logarithm of employment three years ago, the logarithm of sales in US dollars three years ago, region-sector-year fixed effects. Table A12: Firms vulnerable to government decision making Dep. var.: Growth of sales Government contract (1) (2) Yes No BribeShare -0.283* -0.286*** (0.145) (0.076) 0.016 0.020*** (0.012) (0.006) -0.285** (0.112) 0.007 (0.009) -0.098 (0.097) 0.018* (0.009) -0.391***-0.226*** (0.136) (0.066) 0.017 0.023*** (0.011) (0.005) 0.388 13079 Yes All 0.292 18712 Yes All 0.343 17904 Yes All 0.319 16548 Yes All Non-zero BribeShare R-squared Observations Sector FE Firm-level controls 0.291 65896 Yes All Import (3) (4) Yes No Export (5) (6) Yes No 0.307 70077 Yes All Source: Enterprise Surveys and authors’ calculations. Note: Standard errors in parentheses are clustered at the level of industry*region*survey wave. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively. Estimated using ordinary least squares. The dependent variable is annual growth of sales (half of the difference between the logarithm of sales in the last fiscal year and the logarithm of sales three fiscal years ago, both expressed in US dollars at market exchange rates). All regressions control for ownership (state/foreign/domestic), the logarithm of firm age, exporter status, propensity to complain (the kvetch index), the logarithm of employment three years ago, the logarithm of sales in US dollars three years ago, region-sector-year fixed effects. Column 1 includes only businesses that have secured or attempted to secure a government contract in the last 12 months, column 3 only includes businesses that import any of their material inputs and supplies, and column 5 only businesses that export. Table A13: Baseline results by sector Dep. var.: Growth of sales BribeShare Non-zero BribeShare R-squared Observations Region, Year FE Firm-level controls Light manufacturing (1) Heavy manufacturing (2) -0.236* (0.121) 0.011 (0.009) -0.274*** (0.098) 0.013* (0.008) 0.228 25872 Yes All 0.242 26480 Yes All Utilities & construction (3) -0.554** (0.275) 0.037 (0.028) 0.328 4444 Yes All Wholesale & retail (4) Others (5) -0.226* (0.126) 0.021** (0.010) -0.284* (0.161) 0.032** (0.015) 0.248 22297 Yes All 0.343 8678 Yes All Source: Enterprise Surveys and authors’ calculations. Note: Standard errors in parentheses are clustered at the level of industry*region*survey wave. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively. Estimated using ordinary least squares. The dependent variable is annual growth of sales (half of the difference between the logarithm of sales in the last fiscal year and the logarithm of sales three fiscal years ago, both expressed in US dollars at market exchange rates). All regressions control for ownership (state/foreign/domestic), the logarithm of firm age, exporter status, propensity to complain (the kvetch index), the logarithm of employment three years ago, the logarithm of sales in US dollars three years ago, region-sector-year fixed effects. Sectors are classified by ISIC Rev. 3.1. Light manufacturing includes manufacturing of food and beverages, textile and paper products. Heavy manufacturing includes production of chemicals, plastic, metal, wood and electrical products, furniture and machinery. Table A14: Democracies vs non-democracies, low vs middle income countries Dep. var.: Growth of sales BribeShare Non-zero BribeShare R-squared Observations Region-Sector-Year FE Firm-level controls Democracies (1) Non-democracies (2) Low income (3) Middle income (4) -0.187** (0.075) 0.013** (0.006) -0.273*** (0.087) 0.034*** (0.008) -0.179** (0.072) 0.026*** (0.006) -0.347*** (0.101) 0.015* (0.009) 0.261 52506 Yes All 0.334 32234 Yes All 0.280 46675 Yes All 0.318 31088 Yes All Source: Enterprise Surveys and authors’ calculations. Note: Standard errors in parentheses are clustered at the level of industry*region*survey wave. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively. Estimated using ordinary least squares. The dependent variable is annual growth of sales (half of the difference between the logarithm of sales in the last fiscal year and the logarithm of sales three fiscal years ago, both expressed in US dollars at market exchange rates). All regressions control for ownership (state/foreign/domestic), the logarithm of firm age, exporter status, propensity to complain (the kvetch index), the logarithm of employment three years ago, the logarithm of sales in US dollars three years ago, region-sector-year fixed effects. Column 1: countries with Polity IV score of 6 or above; column 2: countries with Polity IV score below 6; column 3: country-years when the World Bank classifies the country as a low-income country (GDP per capita in constant USD of 4035 or less); column 4: country-years when the World Bank classifies the country as a middle-income country (GDP per capita in constant USD of more than 4035 and less or equal 12,475). Table A15: Fine-grained bribe share dummies Dep. var.: Growth of sales 0 < b <= 0.01% 0.01% < b <= 0.03% 0.03% < b <= 0.06% 0.06% < b <= 0.1% 0.1% < b <= 0.5% 0.5% < b <= 1% 1% < b <= 5% 5% < b <= 10% > 10% (1) 0.083*** (0.019) 0.100*** (0.015) 0.079*** (0.019) 0.062*** (0.017) 0.009 (0.009) 0.014** (0.007) -0.002 (0.007) 0.004 (0.009) -0.018** (0.008) BribeShare BribeShare squared BribeShare cubed Non-zero BribeShare R-squared Observations Sector FE Firm-level controls 0.294 87829 Yes All (2) -0.341*** (0.122) 0.332 (0.380) (3) 0.024*** (0.005) -0.636*** (0.205) 2.644* (1.545) -3.742 (2.568) 0.028*** (0.006) 0.293 87829 Yes All 0.293 87829 Yes All Source: Enterprise Surveys, Polity IV and authors’ calculations. Note: Standard errors in parentheses are clustered at the level of industry*region*survey wave. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively. Estimated using ordinary least squares. The dependent variable is annual growth of sales (half of the difference between the logarithm of sales in the last fiscal year and the logarithm of sales three fiscal years ago, both expressed in US dollars at market exchange rates). All regressions control for ownership (state/foreign/domestic), the logarithm of firm age, exporter status, propensity to complain (the kvetch index), the logarithm of employment three years ago, the logarithm of sales in US dollars three years ago, region-sector-year fixed effects. Table A16: Absolute bribes and growth Dep. var.: Growth of sales Absolute BribeShare, USD, log Non-zero BribeShare (1) -0.002*** (0.000) -0.013*** (0.002) 0.126*** (0.015) -0.059*** (0.002) 0.056*** (0.002) 0.004** (0.002) 0.048*** (0.005) 0.044*** (0.003) 0.013 (0.012) -0.034*** (0.002) 0.237 87829 Yes No 0.294 87829 Yes All Sales 3 years ago, USD, log Employment 3 years ago, USD, log Propensity to complain (kvetch index) Foreign ownership Exporter State ownership Age, years, log R-squared Observations Sector FE Firm-level controls (2) Source: Enterprise Surveys, Polity IV and authors’ calculations. Note: Standard errors in parentheses are clustered at the level of industry*region*survey wave. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively. Estimated using ordinary least squares. The dependent variable is annual growth of sales (half of the difference between the logarithm of sales in the last fiscal year and the logarithm of sales three fiscal years ago, both expressed in US dollars at market exchange rates). All regressions control for ownership (state/foreign/domestic), the logarithm of firm age, exporter status, propensity to complain (the kvetch index), the logarithm of employment three years ago, the logarithm of sales in US dollars three years ago, region-sector-year fixed effects. Absolute BribeShare is calculated as the informal payment share plus 1 times total sales in USD three years ago. Table A17: Government kvetch index Dep. var.: Growth of sales Sales 3 years ago, USD, log Propensity to complain (kvetch index) Government kvetch index Foreign ownership Exporter State ownership Age, years, log (1) Non-zero BribeShare -0.000 (0.001) 0.004* (0.002) 0.051*** (0.003) -0.019*** (0.006) 0.012*** (0.003) 0.004 (0.012) -0.002 (0.002) -0.061*** (0.002) 0.004** (0.002) -0.002 (0.002) 0.048*** (0.005) 0.044*** (0.003) 0.013 (0.012) -0.034*** (0.002) -0.233*** (0.058) 0.022*** (0.005) 0.056*** (0.002) 0.309 87625 Yes All 0.294 87625 Yes All BribeShare Non-zero BribeShare Employment 3 years ago, USD, log R-squared Observations Sector FE Firm-level controls (2) Growth of sales Source: Enterprise Surveys, Polity IV and authors’ calculations. Note: Standard errors in parentheses are clustered at the level of industry*region*survey wave. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively. Estimated using ordinary least squares. The dependent variable is annual growth of sales (half of the difference between the logarithm of sales in the last fiscal year and the logarithm of sales three fiscal years ago, both expressed in US dollars at market exchange rates). All regressions control for ownership (state/foreign/domestic), the logarithm of firm age, exporter status, propensity to complain (the kvetch index), the logarithm of employment three years ago, the logarithm of sales in US dollars three years ago, region-sector-year fixed effects. The government kvetch index combines complaints about tax rates, tax administration, business licensing and permits, and political instability. Table A18: Re-surveyed firms Dep. var.: Growth of sales BribeShare (lagged) Non-zero BribeShare (lagged) (1) -0.102 (0.105) 0.020* (0.012) BribeShare Non-zero BribeShare (2) (3) -0.193 (0.126) 0.019 (0.013) -0.282*** (0.059) 0.018*** (0.005) Years between surveys 3 4 5 6 7 8 9 R-squared Observations Fixed effects Firm-level controls -0.016 (0.024) 0.001 (0.040) -0.037 (0.052) -0.038 (0.056) -0.124** (0.060) -0.104* (0.061) -0.148* (0.089) 0.218 10691 Region, sector, year All 0.216 10691 Region, sector, year All 0.213 87829 Region, sector, year All Source: Enterprise Surveys, Polity IV and authors’ calculations. Note: Standard errors in parentheses are clustered at the level of industry*region*survey wave. ***, **, * denote statistical significance at the 1%, 5% and 10% levels, respectively. Estimated using ordinary least squares. The dependent variable is annual growth of sales (half of the difference between the logarithm of sales in the last fiscal year and the logarithm of sales three fiscal years ago, both expressed in US dollars at market exchange rates). All regressions control for ownership (state/foreign/domestic), the logarithm of firm age, exporter status, propensity to complain (the kvetch index), the logarithm of employment three years ago, the logarithm of sales in US dollars three years ago, region-sector-year fixed effects. The specification in column 1 controls for lagged BribeShare and non-zero BribeShare from the previous survey wave. For comparison purposes, the specification in column 2 is estimated for the same sample as in column 1, but includes contemporaneous informal payment variables. Column 3 repeats the baseline results from Table 2, column 4. Appendix B: Model with contract menus In our main model in Section 6 we assume that the bureaucrat does not have information about firms’ cost functions and offers the same bribe rate b to all firms. In this Appendix we consider a model in which the bureaucrat offers firms a menu of contracts, and firms rationally self-select into different contracts. Offering a menu should weakly increase the bureaucrat’s payoff (as he can always choose a degenerate menu of a single bribe rate). However, this setting relies on the assumption of enforceability of rather complex contracts which may or may not hold given that bribe transactions take place outside of the legal system. For the sake of simplicity, we consider a setting with only two types of firms: “high” and “low” with cost functions ci (y), i ∈ {H, L}, respectively. The high type has a lower total cost cH (y) ≤ cL (y) and a lower marginal cost: c0H (y) ≤ c0L (y). The latter inequality is equivalent to the Spence-Mirrlees single-crossing condition.18 We assume that the share of high type is λ ∈ (0, 1). In an equilibrium in which both types choose to pay bribes, B offers (at most) two contracts, each including a bribe rate bi and an output level yi so that the contract with bi and yi is chosen by type i. Without loss of generality, we will use the total bribe amount Ti = bi yi rather than the bribe rate. The menu of two contracts {(yH , TH ) , (yL , TL )} should satisfy the following four constraints. First, for each i = H, L there is an incentive compatibility constraint: yi − ci (yi ) − Ti ≥ yj − ci (yj ) − Tj , (5) where j = H, L and j 6= i. Second, for each type i = H, L there should be a participation constraint: yi − ci (yi ) − Ti ≥ (1 − k)πi (1). (6) The right-hand side of this inequality is the firm’s payoff in case the firm refuses to pay bribes and complains to the public (see the analysis in section 6.4). Under some sets of parameters there may be equilibria in which the bureaucrat prefers to receive bribes only from one type. Since these equilibria are similar to those in the main model, in this Appendix we will focus on the equilibria in which the participation constraints (6) hold for both types. The bureaucrat thus chooses yH , yL , TH , TL subject to (5)-(6) in order to maximize his total bribe: λTH + (1 − λ)TL . (7) 18 This condition is consistent with Assumption 3. For example, c0H (y ∗ ) = c0L (y ∗ ) = 1 for a certain y ∗ but c0H (y) ≤ c0L (y) for all other y. Rearranging the terms in the two constraints (5) for i = L, H, we obtain (yH − yL ) − [cL (yH ) − cL (yL )] ≤ TH − TL ≤ (yH − yL ) − [cH (yH ) − cH (yL )] . The Spence-Mirrlees condition c0H (y) ≤ c0L (y) implies that the expression in the brackets is weakly smaller for the high type. Thus, if one of these two inequalities is binding, the other one follows from the Spence-Mirrlees condition. For each type i at least one constraint, (5) and/or (6), should bind — otherwise the bureaucrat can increase Ti , strictly improving his payoff. Let us first explore the conventional case in which the participation constraint (6) is binding for the low type and the incentive compatibility constraint (5) is binding for the high type. Solving for TH and TL and substituting into (7), we find the expression for the bureaucrat’s objective function: −(1 − k)πL (1) + (1 − λ)yL − cL (yL ) + λcH (yL ) + λ (yH − cH (yH )) . The first order conditions c0H (yH ) = 1 and c0L (yL ) − λc0H (yL ) = 1 − λ imply that yH = y ∗ and yL ≤ y ∗ . This in turn implies that TH ≥ TL . Therefore, if we assume that bureaucrats can offer menus of contracts, we should observe a positive correlation between total bribe amount and firm growth, a prediction that is at odds with our empirical findings. Ruling out other equilibria. While the equilibrium with binding incentive compatibility constraint for the high type is standard in conventional adverse selection settings (where the reservation utility is the same for both types), in our case there may also be equilibria in which the high type’s participation constraint is binding. Indeed, in our case, the reservation utility (1 − k)πi (1) is higher for the high type. Let us first rule out equilibria in which (6) is binding for the high type and (5) is binding for the low type. If this were the case, the first order conditions c0L (yL ) = 1 and c0H (yH ) − (1 − λ)c0L (yH ) = λ would imply that yH ≥ yL = y ∗ . Let us now check when the remaining constraints would be satisfied in this case. The high type’s incentive compatibility constraint is yH − cH (yH ) − TH ≥ yL − cH (yL ) − TL . As the left-hand side equals (1 − k)πH (1) and yL = y ∗ , this constraint is equivalent to TL ≥ kπH (1). On the other hand, the low type’s participation constraint yL − ci (yL ) − Ti ≥ (1 − k)πL (1) is equivalent to TL ≤ kπL (1). These are only compatible in the case where πH (1) = πL (1); however, in this case all constraints are binding, so this is a special case of the equilibrium we considered above. Let us now rule out equilibria where both participation constraints (6) bind and neither incentive compatibility constraint does. In this case, the bureaucrat’s first order conditions imply that both types produce efficient amounts yi = y ∗ and Ti = kπi (1). We now need to check if these satisfy incentive compatibility constraints (5). The low type’s incentive compatibility constraint is equivalent to πH (1) ≥ πL (1) while the high type’s incentive compatibility constraint is equivalent to πH (1) ≤ πL (1). Again, these are only compatible in the degenerate case where πH (1) = πL (1). This is the case where all four constraints are binding, again a special case of the equilibrium we considered above.