Gender and Bureaucratic Corruption (Tables)

Table 1: Summary statistics for the Italian data Female Investigated Age RUP (in 2018) RUP born in same Region Observations All Male Female Difference (1) (2) (3) (3)-(2) 0.17 (0.38) 0.09 (0.28) 57.83 (9.43) 0.87 (0.33) 0.09 (0.29) 58.97 (9.16) 0.88 (0.32) 0.05 (0.22) 52.45 (8.83) 0.84 (0.37) -0.04 *** [0.000] -6.52 *** [0.000] -0.04 *** [0.000] 21,277 17,574 3,703 21,277 Notes: Investigated is an indicator variable equal to 1 if the contracting officer overseeing each contract (the “Responsabile Unico del Procedimento”, or RUP) has been investigated for corruption. PA stands for Procurement Authority. The sample includes the universe of RUP-PA unique pairs. See text for further details. Standard deviations in parentheses. Pvalues in squared brackets. Significance: ∗ p < 0.1,∗∗ p < 0.05,∗∗∗ p < 0.01. Table 2: Summary statistics for the Chinese data All Male Female Difference (1) (2) (3) (3)-(2) Panel A: Full Sample Female 0.04 (0.19) Investigated for Corruption 0.08 (0.26) Masters 0.32 (0.46) Doctor 0.10 (0.30) Year of First Appointment 1997.56 (9.83) Age 56.98 (27.40) Observations 3133 0.08 (0.27) 0.30 (0.46) 0.10 (0.30) 1997.34 (9.80) 56.85 (27.77) 0.03 (0.18) 0.59 (0.49) 0.12 (0.33) 2003.13 (8.83) 60.09 (15.28) -0.04 ** 0.011 0.29 *** 0.000 0.03 0.360 5.80 *** 0.000 3.24 ** 0.029 3013 120 3133 Panel B: Started post > Female 0.05 (0.22) Investigated for Corruption 0.12 0.12 0.04 (0.32) (0.33) (0.20) Masters 0.48 0.47 0.64 (0.50) (0.50) (0.48) Doctor 0.16 0.16 0.15 (0.37) (0.37) (0.36) Year of First Appointment 2004.11 2004.00 2006.18 (5.90) (5.90) (5.59) Age 59.59 59.69 57.79 (13.19) (13.32) (10.39) Observations 1878 1778 100 1998 -0.08 *** [0.000] 0.17 *** [0.001] -0.01 [0.756] 2.18 *** [0.000] -1.90 * [0.079] 1878 Notes: The sample in Panel A is the set of Chinese officials who held the position of prefecture mayor or party secretary between 1979 and 2014. In Panel B the sample is limited to individuals who started such a position in 1998 or later. Investigated is an indicator variable denoting that the official was investigated for corruption. Year of First Appointment is the year the official first held a position of prefecture mayor or party secretary. See text for further details. Standard deviations in parentheses. P-values in squared brackets. Significance: * significant at 10%; ** significant at 5%; *** significant at 1%. Table 3: Probability of Italian RUP investigation as a function of gender Dependent variable: Investigated (2) (1) (3) (4) (5) -0.0295 [0.00569] ∗∗∗ -0.0214 [0.00565] -0.0188 [0.00644] Age (log) 0.0598∗∗∗ [0.0145] 0.0906∗∗∗ [0.0145] 0.0875∗∗∗ [0.0168] RUP born in same Municipality -0.0228∗∗∗ [0.00507] -0.0158∗∗∗ [0.00507] -0.0151∗∗ [0.00605] Tot. Auctions managed by RUP (log) 0.0324∗∗∗ [0.00217] 0.0336∗∗∗ [0.00242] 0.0349∗∗∗ [0.00253] Average Value of Auctions (log) 0.00821∗∗∗ [0.00270] 0.00207 [0.00325] 0.00190 [0.00347] Average Number of Bidders (log) -0.0119∗∗∗ [0.00196] -0.00820∗∗∗ [0.00245] -0.00890∗∗∗ [0.00254] Yes Yes Yes No No No 0.0867 21277 0.0937 Yes No Yes Yes No No 0.0802 17831 0.370 RegionXYear FE PA Controls RUP Controls PA FE Age & N.Contracts FE Municipalities only Dep. Var. Mean Observations Adjusted R-sq -0.0338 [0.00560] Yes No No No No No 0.0867 21277 0.0704 Yes No Yes No No No 0.0867 21277 0.0738 ∗∗∗ (6) ∗∗∗ Female ∗∗∗ ∗∗∗ -0.0176 [0.00653] -0.0178∗∗∗ [0.00680] 0.0909∗∗∗ [0.0177] -0.0163∗∗∗ [0.00611] Yes No Yes Yes Yes No 0.0802 17823 0.381 -0.0149∗∗ [0.00613] Yes No Yes Yes No Yes 0.0823 16774 0.372 Notes: The dependent variable, Investigated, is an indicator equal to 1 if the public official in charge of the auction (the RUP) has been investigated. The analysis is conducted on a panel of RUP-PA observations. PA Controls include a set of dummies for the type of PA (Central, Region and other local authority, Hospitals and Universities, Transportation), the log of the total number of auctions done by the PA during the sample period, the log of the total number of RUPs observed in the PA during the sample period, the log of the total number of auctions managed by the RUP, the log of the average value of auctions managed by the RUP and the log of the average number of bidders participating in auctions managed by the RUP. Individual controls include the log of the age of the RUP, an indicator for whether the RUP was born in the same city where she operates as a RUP and a set of dummies for the region of birth of the RUP. In Column 6 the sample is restricted to contracting authorities that are municipalities, which represent 54% of the sample. This specification excludes administrations that do not map to a specific local geography, such as educational institutions, hospitals, and public companies dealing with the management of motorway sections under concession. Robust standard errors clustered at the RUP level are in parentheses. Significance: ∗ p < 0.1,∗∗ p < 0.05,∗∗∗ p < 0.01. Table 4: Probability of Italian RUP investigation as a function of gender, heterogeneity analyses Female=1 Female=1 × Born in Municipality=1 (1) (2) (3) (4) (5) (6) (7) (8) (9) -0.0247∗∗∗ [0.00757] -0.0125∗ [0.00707] -0.0285∗∗∗ [0.00693] -0.0236∗∗∗ [0.00706] -0.0115 [0.00821] -0.0123 [0.00891] -0.0110 [0.00953] -0.00241 [0.0103] 0.00655 [0.0223] 0.0272∗∗ [0.0139] Female=1 × Local Politician=1 0.0178 [0.0194] -0.0353 [0.0216] -0.00385 [0.0297] 0.0340∗∗ [0.0172] Female=1 × South=1 Female=1 × Mafia Municipality=1 -0.00883 [0.0322] 0.0295 [0.0194] 0.0215 [0.0242] -0.0167∗∗ [0.00814] Female=1 × Female LFP (SD) Female=1 × Share Female RUPs (SD) 0.0123 [0.0161] -0.0112 [0.0110] -0.0102 [0.0188] -0.0138∗ [0.00764] Female=1 × Share Female PA (SD) Female=1 × Avg. Income (SD) RegionXYear FE PA Controls Indiv. Controls PA FE Municipalities only Dep. Var. Mean Observations R-sq Yes No Yes Yes No 0.0823 16771 0.372 Yes No Yes Yes No 0.0823 16771 0.374 Yes No Yes Yes No 0.0828 16506 0.372 Yes No Yes Yes No 0.0827 16292 0.370 Yes No Yes Yes No 0.0829 16136 0.373 Yes No Yes Yes No 0.0823 16771 0.372 Yes No Yes Yes No 0.0884 10126 0.378 -0.00962 [0.00880] -0.0409∗∗ [0.0163] -0.0652∗∗ [0.0265] Yes No Yes Yes No 0.0824 16763 0.372 Yes No Yes Yes No 0.0891 9518 0.382 Notes:The dependent variable, Investigated is an indicator equal to 1 if the public official in charge of the auction (the RUP) has been investigated. BorninM unicipality is a dummy equal to 1 if the RUP was born in the same municipality in which she works as a RUP. LocalP olitician is a dummy equal to 1 if the RUP was ever elected as a local politician. South is a dummy equal to 1 if the PA is located in one of the Southern Regions of Italy (Abruzzo, Basilicata, Campania, Calabria, Molise, Puglia, Sardinia, Sicily). M af iaM unicipality is a dummy equal to 1 if the PA is located in a municipality ever subject to mafia investigations. F emaleLF P (SD) is the standardized value of Female Labor Force Participation in the municipality where the PA is located. ShareF emaleRU P s(SD) is the standardized value of the share of female RUPs in the PA. ShareF emaleP A(SD) is the standardized value of the share of female employees in the PA. Avg.Income(SD) is the standardized value of the average income (from tax revenues) in the municipality where the PA is located. The analysis is conducted on a panel of RUP-PA observations. PA Controls include a set of dummies for the type of PA (Central, Region and other local authority, Hospitals and Universities, Transportation), the log of the total number of auctions done by the PA during the sample period, the log of the total number of RUPs observed in the PA during the sample period, the log of the total number of auctions managed by the RUP, the log of the average value of auctions managed by the RUP and the log of the average number of bidders participating in auctions managed by the RUP. Individual controls include the log of the age of the RUP, an indicator for whether the RUP was born in the same city where she operates as a RUP and a set of dummies for the region of birth of the RUP. Robust standard errors clustered at the RUP level are in parentheses. Significance: ∗ p < 0.1,∗∗ p < 0.05,∗∗∗ p < 0.01. Table 5: Selection on Unobservables in Italian Data Baseline Effect [R2] Controlled Effect [R2] Identified Set δ for β = 0 given Rmax (1) (2) (3) (4) Panel A: Without Municipality FE Female -0.033 [0.070] -0.020 [0.092] [-0.003,-0.020] 1.11 Panel B: Including Municipality FE Female -0.033 [0.070] -0.018 [0.504] [-0.006,-0.018] 1.39 Notes: We perform the test suggested by Oster (2019). In Panel A, Column 1, we report the coefficient from our baseline regression, including only region-times-year fixed effects. In Column 2, we report the coefficient from our regression including all available controls (except for PA fixed effects). In Column 3, we calculate the identified set for the biasadjusted treatment effect, using the suggested values of Rmax = 1.3R and δ = 1. In Column 4, we calculate the value of δ s.t. β = 0, assuming again that Rmax = 1.3R , where R is the observed R-squared from the regression in Column 2. Panel B repeats the same exercise but includes PA fixed effects for the regression in Column 2. Rmax is the R-squared of a hypothetical regression including all observables and unobservables, while δ represents the degree of selection on observed versus unobserved variables. For further details, see Section B. 1. Table 6: Defensive Actions and Insurance in Italian Survey Data Defensive Action (1) Female Dep. var. Mean Dep. var. SD Demographics Risk Observations R-sq ∗∗∗ (2) Own Insurance (3) (4) (5) ∗∗ (6) 5.484 [1.711] ∗∗ 4.173 [1.852] ∗∗ ∗∗∗ 4.237 [1.898] -11.34 [4.962] -14.33 [5.307] -13.12∗∗ [6.015] 31.05 18.95 No No 508 0.0226 30.72 18.39 Yes No 402 0.0302 29.49 17.04 Yes Yes 323 0.0721 36.86 48.30 No No 388 0.0531 37.05 48.36 Yes No 359 0.0949 39.18 48.90 Yes Yes 291 0.0912 Notes: In Columns 1-3, the dependent variable, Def ensiveActions, is defined as the average probability of choosing the ”defensive option” across the four scenario questions (see Appendix B). Demographics controls include tenure in the public sector, age, age squared, as well as fixed effects for the type of PA the RUP works at; Risk controls include RUP risk attitudes as captured by survey responses (see Appendix B). The dependent variable in Columns 4-6, Insurance, reflects whether the RUP has insurance against the risk of administrative damages. Robust standard errors in parentheses. Significance: ∗ p < 0.1,∗∗ p < 0.05,∗∗∗ p < 0.01. Table 7: Probability of Chinese officials’ investigation as a function of gender Dependent variable: Investigated Full Sample Started post > 1998 (1) (2) (3) (4) (5) -0.069∗∗∗ [0.0177] -0.067∗∗∗ [0.0176] -0.079∗∗∗ [0.0178] 0.057∗ [0.0252] 0.038∗ [0.0154] 0.007 [0.0127] -0.232∗∗∗ [0.0688] -1.047∗∗∗ [0.305] -0.092∗∗∗ [0.0215] 0.053 [0.0278] 0.041∗ [0.0190] 0.010 [0.0287] -0.303∗ [0.134] -1.344∗ [0.570] -0.086∗∗∗ [0.0217] 0.038 [0.0269] 0.030 [0.0193] 0.026 [0.0298] Province FE Start year FE Age Cohort FE No Yes No Yes Yes No Yes Yes No Yes Yes No Yes Yes Yes Dep. Var. Mean Observations Adjusted R-sq 0.075 3133 0.044 0.075 3133 0.044 0.075 3133 0.056 0.117 1874 0.027 0.121 1807 0.028 Female Doctor Masters No Degree Age (log) Age Missing Notes: The sample in columns (1)-(3) is the set of Chinese officials who held the position of prefecture mayor or party secretary between 1979 and 2014; in columns (4) and (5) the sample is limited to individuals who started such a position in 1998 or later. The outcome in all columns is an indicator variable denoting that the official was investigated for corruption. Please see text for further details. Robust standard errors in parentheses. Significance: ∗ p < 0.1,∗∗ p < 0.05,∗∗∗ p < 0.01. Table 8: Selection on Unobservables in Chinese Data Baseline Effect [R2] Controlled Effect [R2] Identified Set δ for β = 0 given Rmax (1) (2) (3) (4) Panel A: Without Province FE Female -0.069 [0.041] -0.081 [0.053] [-0.097,-0.081] -1.16 Panel B: Including Province FE Female -0.069 [0.041] -0.080 [0.065] [-0.089,-0.080] -7.95 Notes: We perform the test suggested by Oster (2019). In Panel A, Column 1, we report the coefficient from our baseline regression, including cohort fixed effects. In Column 2, we report the coefficient from our regression, including all available controls (except for Province fixed effects). In Column 3, we calculate the identified set for the bias-adjusted treatment effect, using the suggested values of Rmax = 1.3R and δ = 1. In Column 4, we calculate the value of δ s.t. β = 0, assuming again that Rmax = 1.3R , where R is the observed R-squared from the regression in Column 2. Panel B repeats the same exercise but including Province fixed effects for the regression in Column 2. Rmax is the R-squared of a hypothetical regression including all observables and unobservables, while δ represents the degree of selection on observed versus unobserved variables. For further details, see Section B. 1. Table A.1: Probability of Italian RUP investigation as a function of gender, including a control for whether RUP had been a local politician Dependent variable: Investigated (1) (2) (3) (4) (5) (6) -0.0338∗∗∗ [0.00515] -0.0261∗∗∗ [0.00527] -0.0187∗∗∗ [0.00523] -0.0165∗∗∗ [0.00620] -0.0176∗∗∗ [0.00624] -0.0155∗∗ [0.00648] Age (log) 0.0539∗∗∗ [0.0132] 0.0846∗∗∗ [0.0132] 0.0828∗∗∗ [0.0162] RUP born in same Municipality -0.0187∗∗∗ [0.00463] -0.0129∗∗∗ [0.00467] -0.0130∗∗ [0.00577] Tot. Auctions managed by RUP (log) 0.0320∗∗∗ [0.00222] 0.0334∗∗∗ [0.00223] 0.0346∗∗∗ [0.00233] Average Value of Auctions (log) 0.00825∗∗∗ [0.00271] 0.00200 [0.00319] 0.00180 [0.00337] Average Number of Bidders (log) -0.0114∗∗∗ [0.00194] -0.00796∗∗∗ [0.00245] -0.00866∗∗∗ [0.00255] Yes Yes Yes No No No 0.0867 21277 0.0964 Yes No Yes Yes No No 0.0802 17831 0.371 Female RegionXYear FE PA Controls RUP Controls PA FE Age & N.Contracts FE Municipalities only Dep. Var. Mean Observations Adjusted R-sq Yes No No No No No 0.0867 21277 0.0704 Yes No Yes No No No 0.0867 21277 0.0773 0.0856∗∗∗ [0.0171] -0.0163∗∗∗ [0.00578] Yes No Yes Yes Yes No 0.0802 17823 0.381 -0.0128∗∗ [0.00591] Yes No Yes Yes No Yes 0.0823 16774 0.374 Notes: The dependent variable, Investigated is an indicator equal to 1 if the public official in charge of the auction (the RUP) has been investigated. The analysis is conducted on a panel of RUP-PA observations. PA Controls include a set of dummies for the type of PA (Central, Region and other local authority, Hospitals and Universities, Transportation), the log of the total number of auctions done by the PA during the sample period, the log of the total number of RUPs observed in the PA during the sample period, the log of the total number of auctions managed by the RUP, the log of the average value of auctions managed by the RUP and the log of the average number of bidders participating in auctions managed by the RUP. Individual controls include the log of the age of the RUP, an indicator for whether the RUP was born in the same city where she operates as a RUP, an indicator for whether RUP had been a local politician and a set of dummies for the region of birth of the RUP. In column 6 the sample is restricted to contracting authorities that are municipalities, which represent 54% of the sample. This specification excludes administrations that do not map to a specific local geography, such as educational institutions, hospitals, and public companies dealing with the management of motorway sections under concession. Robust standard errors clustered at the RUP level are in parentheses. Significance: ∗ p < 0.1,∗∗ p < 0.05,∗∗∗ p < 0.01. Table A.2: Summary statistics by North vs. South North South Tot. Auctions managed by RUP Investigated Share discretionary contracts Male, North Female, North Male, South Female, South F-M, North F-M, South (1) (2) (3) (4) (5) (6) (3)-(4) (5)-(6) 6.77 (9.59) 0.06 (0.24) 0.26 (0.27) 5.48 (6.90) 0.13 (0.34) 0.17 (0.20) 7.05 (9.88) 0.07 (0.26) 0.26 (0.27) 5.71 (8.30) 0.04 (0.19) 0.29 (0.28) 5.52 (6.98) 0.13 (0.34) 0.17 (0.20) 5.11 (6.20) 0.12 (0.32) 0.14 (0.17) -1.34 *** [0.000] -0.04 *** [0.000] 0.03 *** [0.000] -0.41 * [0.097] -0.02 [0.194] -0.03 *** [0.000] Notes: Investigated is an indicator variable equal to 1 if the contracting officer overseeing each contract (the “Responsabile Unico del Procedimento”, or RUP) has been investigated for corruption. PA stands for Procurement Authority. The sample includes the universe of RUP-PA unique pairs. See text for further details. Standard deviations in parentheses. P-values in squared brackets. Significance: ∗ p < 0.1,∗∗ p < 0.05,∗∗∗ p < 0.01. Table A.3: Gender Differences in the type of auctions (1) DiscretProc (2) DiscretP roclowN (3) DiscretProc (4) DiscretP roclowN 0.0168∗ [0.00874] 0.0113∗∗ [0.00490] -0.0146 [0.00960] -0.000689 [0.00548] -0.191 [0.210] -0.140∗∗∗ [0.0471] 0.00599 [0.00798] 0.00570∗∗∗ [0.00175] Constant 1.854 [1.380] 0.906∗∗∗ [0.317] 0.372∗∗∗ [0.00151] 0.0544∗∗∗ [0.000864] Dep. Var. Mean Observations R-sq Geog. FE 0.225 109875 0.442 0.225 109875 0.108 0.225 108717 0.577 PA 0.225 108717 0.249 PA Female Population (log) Population Sq. (log) Notes: The dependent variable is indicated on top of each column.DiscretP roc denotes negotiated procedures. DiscretP roclowN denotes negotiated procedures with fewer than the legally mandated number of bidders. The analysis is conducted at the contract level. All regressions include year fixed effects, a linear control for reserve price (in log) and 5 dummies for different contract size thresholds (up to 100k, 100-150k, 150-300k, 300-500k, 500k-1mil, 1-1.5mil, over 1.5mil) as well as controls for contract characteristics: 4 dummies for category type (Civil Building, Roadworks, Specialized Works or Others), 1 dummy for whether the contract was awarded under urgency and 1 dummy for whether the object of the contract entailed maintenance. Robust standard errors clustered at the PA level are in parentheses. ∗ p < 0.1,∗∗ p < 0.05,∗∗∗ p < 0.01. Robust standard errors clustered at the administration level are in parentheses.Significance: ∗ p < 0.1,∗∗ p < 0.05,∗∗∗ p < 0.01. Table A.4: Summary statistics for the Italian survey data Female Defensive action (avg.) Time devoted to check compliance with regulations Risk of Proceedings Risk of Proceedings, ANAC Risk of Proceedings, Court of Accounts Risk of Proceedings, Judiciary Risk of Proceedings, higher admin Media Risk Complexity of Compliance with regulation Knowledge of regulation Share of own defensive actions Share of defensive actions in the PA Change of share in the last 5 years Risk of civil responsibility sactions Risk of admin responsibility sactions Has Insurance from PA for civil responsibility Is aware of insurance for admin responsibility Has insurance for admin responsibility Observations All Male Female Difference (1) (2) (3) (3)-(2) 0.47 (0.50) 30.56 (18.76) 68.24 (28.13) 51.78 (28.63) 41.91 (29.35) 47.66 (31.43) 44.28 (33.64) 42.93 (31.73) 46.62 (28.39) 74.74 (20.00) 65.52 (20.80) 30.60 (28.33) 51.08 (27.53) 29.10 (39.13) 42.96 (27.25) 53.30 (28.82) 17.13 (37.72) 0.79 (0.41) 36.75 (48.27) 28.51 (20.29) 62.21 (29.09) 50.58 (28.48) 40.60 (28.22) 46.81 (30.96) 45.12 (33.42) 40.71 (31.44) 44.94 (29.29) 75.04 (19.77) 67.20 (19.66) 31.12 (27.88) 51.80 (28.11) 32.53 (38.85) 43.09 (28.27) 51.83 (29.78) 20.61 (40.54) 0.78 (0.41) 44.44 (49.82) 32.84 (16.64) 74.83 (25.52) 53.18 (28.81) 43.43 (30.62) 48.66 (32.03) 43.23 (33.98) 45.58 (31.96) 48.55 (27.26) 74.40 (20.30) 63.66 (21.90) 29.96 (28.94) 50.18 (26.83) 24.23 (39.15) 42.81 (26.08) 55.04 (27.62) 13.24 (33.97) 0.80 (0.40) 28.42 (45.22) 4.33 ** [0.015] 12.62 *** [0.000] 2.61 [0.377] 2.83 [0.359] 1.85 [0.572] -1.90 [0.593] 4.86 [0.139] 3.61 [0.200] -0.64 [0.743] -3.54 * [0.083] -1.16 [0.701] -1.62 [0.578] -8.30 * [0.068] -0.28 [0.919] 3.21 [0.265] -7.38 ** [0.040] 0.02 [0.553] -16.03 *** [0.001] 420 224 196 420 Notes: All variables (except Female) vary from 0 to 100. Standard deviations in parentheses. P-values in squared brackets. Significance: ∗ p < 0.1,∗∗ p < 0.05,∗∗∗ p < 0.01. Table A.5: Probability of Chinese officials’ investigation under the 2013 anticorruption crackdown, as a function of gender Dependent variable: Investigated Full Sample Started post > 1998 (1) (2) (3) (4) (5) -0.069∗∗∗ [0.0177] -0.067∗∗∗ [0.0176] -0.079∗∗∗ [0.0178] 0.057∗ [0.0252] 0.038∗ [0.0154] 0.007 [0.0127] -0.232∗∗∗ [0.0688] -1.047∗∗∗ [0.305] -0.092∗∗∗ [0.0215] 0.053 [0.0278] 0.041∗ [0.0190] 0.010 [0.0287] -0.303∗ [0.134] -1.344∗ [0.570] -0.086∗∗∗ [0.0217] 0.038 [0.0269] 0.030 [0.0193] 0.026 [0.0298] Province FE Start year FE Age Cohort FE No Yes No Yes Yes No Yes Yes No Yes Yes No Yes Yes Yes Dep. Var. Mean Observations Adjusted R-sq 0.075 3133 0.044 0.075 3133 0.044 0.075 3133 0.056 0.117 1874 0.027 0.121 1807 0.028 Female Doctor Masters No Degree Age (log) Age Missing Notes: The sample in columns (1)-(3) is the set of Chinese officials who held the position of prefecture mayor or party secretary during 1979-2014; in columns (4) and (5) the sample is limited to individuals who started such a position 1998 or later. The outcome in all columns is an indicator variable denoting that the official was investigated for corruption. Please see text for further details. Robust standard errors in parentheses. Significance: ∗ p < 0.1,∗∗ p < 0.05,∗∗∗ p < 0.01.