Electoral competition and corruption: Theory and evidence from India

Transcription

1 Electoral competition and corruption: Theory and evidence from India Farzana Afridi (ISI, Delhi) Amrita Dhillon (King s College London) Eilon Solan (Tel Aviv University) June 25-26, 2018 ABCDE Conference, Washington DC

2 Motivation The persistence of corruption in low income democracies like India is puzzling: If electoral competition is fair, and if voters care about honesty, forward looking incumbents should respond to re-election incentives by lowering corruption today.

3 This paper When legal sanctions are not binding, can electoral competition, in the presence of mandated exposure of corruption, discipline politician behavior? Does electoral competition affect the nature of corruption? All corruption may not be viewed as equally bad from the voters point of view - pilferage from public vs. private goods

4 Preview of findings Our main theoretical results predict a non-linear, U- shaped, relationship between electoral competition and corruption. Lower responsiveness to competition when the electoral bias is against the incumbent. Electoral competition can, at best, control only those types of corruption which affect voters directly. Higher responsiveness of corruption when it concerns citizens directly.

5 Theoretical Model Assumptions: Two parties, L and R and a continuum of potential candidates. Incumbent from party L and challenger from party R situated at [-1,1], respectively. Political leaders care about rents from office or corruption (x), and getting elected: δδ tt 1 xx tt PP tt tt=1 Where P is the probability that the incumbent is in power at stage t and δδ is the incumbent's discount factor, which is common to all potential candidates

6 Theoretical Model Assumptions: There is a continuum of voters with ideology distributed on [-1,1] Position of median voter, β, denotes electoral competition. β< 0 indicates advantage with L party to begin with. Voters care about ideology and corruption: Uj(x; i) = - x - (zj - i)² for a voter j with ideal point zj 2 [-1; 1]

7 Theoretical Model Timeline: In every period the incumbent chooses the proportion, x, to steal from a public fund of given size. An audit is conducted of the public expenditures whose findings are revealed publicly. Voters observe whether revealed corruption is higher than challenger s reputation for corruption uniformly distributed on [0,1]. Election takes place. If the incumbent is re-elected, the game continues to next stage otherwise terminates.

8 Theoretical Model Corruption (x) as a function of electoral competition (β)

9 Theoretical Model Nature of corruption: When theft is private (citizens are affected directly) then the relationship between electoral competition and corruption is as given in Figure. When theft is public (citizens only affected indirectly) then the relationship is insignificant - theft is maximum possible.

10 Predictions When electoral bias is towards the incumbent, average corruption follows a U-shape: corruption level is high when competition is either very low or very high, while in the intermediate range corruption is decreasing with competition. When electoral bias is against the incumbent, corruption remains high regardless of competition. For a wide range of parameters, the non-linearity in corruption relative to competition is driven by the private component of corruption.

11 Our setting Electoral competition for the position of village council (GP) head or sarpanch in erstwhile Andhra Pradesh (AP). The implementation of one of the largest public programs, the National Rural Employment Guarantee Act (NREGA). Independent audits of NREGA projects in all GPs.

12 NREGA Private component : Expenditures on the wage component of the program; irregularities registered during the door-to-door verification process by the audit team. Public component : Expenditures on the material component of the NREGA projects; audit at project sites. Data extracted and codified from official audit reports of all NREGA audits conducted during in 300 randomly sampled GPs along with data on election to the position of GP head (sarpanch) in 2006.

13 Methodology Our empirical specification, utilizing the panel structure, is given by: Irregularity jklmt = β 0 + β 1 competition jkl + β 2 competition 2 jkl + β 3 X jkl + β t year t + δ lt (D l *year t ) + δ m audit m + δ k D k + ε jklmt Competition in GP j in mandal k in district l= 1-Margin of victory audit m : audit round fixed effect year t : secular trends D k : mandal fixed effect D l *year t : district specific trends Electoral Bias: self-reported probability of winning next election

14 Effect of electoral completion on NREGA irregularities by electoral bias (GP-audit level, ) Professional auditor reported irregularities Total irregularities Irregularities in private good Irregularities in public good (1) (2) (3) Electoral bias in favor of incumbent Electoral Competition ** *** (12.598) (8.584) (6.095) Electoral Competition ** 23.08*** (7.734) (5.338) (3.739) U-shape test [Overall P-value] [0.022] [0.007] - N R Electoral bias against incumbent Electoral Competition (6.593) (4.070) (4.243) Electoral Competition (4.242) (2.618) (2.790) U-shape test [Overall P-value] N R Note: Controls include sarpanch characteristics (age, age square, dummy for secondary education completed, dummy for graduate and above education; dummy for prior political experience, affiliated to INC) GP characteristics (main GP of mandal, medical, communication, banking, paved road, middle school in GP, distance from town, proportion of cultivated area which is irrigated, population density, dummy for SC, ST, OBC, woman reserved sarpanch candidate, sarpanch elected unanimously), mandal, audit round, year fixed effects and district specific trends. Standard errors, clustered at the GP level, reported in parentheses. Electoral competition is in the interval (0, 0.99). Extremum outside interval in lower panel, column (3). Significant at *10%, **5% and ***1%.

15 Effect of electoral competition on irregularities in the private component by electoral bias (GP-audit level, ) Professional auditor reported irregularities Impersonations in wage payments Irregularities in own labor records Private good irregularities Non-payment/ delay of wages Payment of bribes to receive due wages Non-provision of work Poor quality of asset (1) (2) (3) (4) (5) (6) Electoral bias in favour of incumbent Electoral Competition *** *** *** (2.622) (2.374) (1.147) (6.244) (1.422) (2.287) Electoral Competition *** 4.911*** 1.417** *** (1.645) (1.391) (0.670) (3.886) (0.886) (1.304) U-shape test [Overall P-value] [0.023] [0.151] [0.053] [0.155] - [0.000] N R Electoral bias against incumbent Electoral Competition 4.667** *** (2.281) (2.410) (1.387) (1.981) (0.507) (1.781) Electoral Competition ** * 0.817*** (1.474) (1.545) (0.864) (1.286) (0.306) (1.155) U-shape test [Overall P-value] [0.064] [0.028] - N R Note: Controls as elucidated in Table 3. Standard errors, clustered at the GP level, reported in parentheses. Significant at *10%, **5% and ***1%. Extremum outside interval in lower panel, column (6).

16 Effect of electoral completion on NREGA irregularities (GP level, ) Professional auditor reported irregularities Total irregularities Irregularities in private good Irregularities in public good (1) (2) (3) (4) (5) (6) Electoral competition *** *** (1.722) (8.301) (1.380) (6.250) (0.967) (4.599) Electoral competition *** 13.00*** (5.599) (4.143) (3.191) U-shape test [Overall P-value] [0.006] [0.005] [0.157] Mandal FE N R Note: Controls as elucidated in Table 3, excluding, audit round, year and district specific trends. U-shape test for columns (2), (4) and (6). Standard errors in parentheses. Significant at *10%, **5% and ***1%.

17 Electoral competition and total irregularities Professional auditor reported irregularities Bias in favor of incumbent Bias against incumbent electoral competition electoral competition Note: Fitted values and 95% confidence interval (Table 3, column 1). The second column reports fitted values for same specification but excludes extreme value - 27 GPs where the sarpanch was elected unanimously, i.e. electoral competition was 0.

18 Conclusions Too much or too little electoral competition can lead to high levels of corruption in public programs. Mandated exposure of corruption can lead to better governance outcomes, even in the absence of legal sanctions, only when there is moderate electoral competition. The response of malfeasance to electoral competition differs by the type of good: recent measures like biometrics could be of limited effectiveness in curtailing theft from the material (public) component of program expenditures.