Corruption Amongst India s Politicians: Insights from Unusual Data Rikhil R. Bhavnani November 2010 1 / 17
Questions How corrupt are India s politicians? What are (excess) gains to office in India? What proportion of politicians have excess gains? Who is corrupt? What are the consequences? 2 / 17
Motivation Corruption affects political and economic outcomes it undermines trust in the political process impacts economic growth linked to organized crime and violence Political corruption is said to be the fount of all corruption Yet we have few measures of it 3 / 17
Background In India, corruption amongst politicians is thought to have worsened over time be more prevalent in the North be caused due to low salaries, government intervention, poor campaign-finance laws, ethnic politics, anti-incumbency Why study India? 85th-most corrupt country (of 180); intra-country variation great data lots of governance reforms 4 / 17
Data Since 2003, candidates have had to declare their own, spouse, dependent assets liabilities education age criminal charges, convictions Data for 2 consecutive elections, 2003/04 and 2008/09 Recently coded 70,000+ data sheets in 8 languages to yield records for 11 states. Results are preliminary 5 / 17
An example Mayawati s assets in 2004 $0.43m in 2008 $13m (72 properties, 54 bank accounts, among India s highest tax payers) But she had a salary of $50,000/year 6 / 17
Empirical strategy Gains to office E [ A W X = 0] E [ A L X = 0] Prop. suspect politicians Pr( A W > t X = 0) Pr( A L > t X = 0) Correlates, consequences of corruption 7 / 17
Bias? I probably underestimate the gains to office since politicians probably underreport their assets politicians might have spent their gains might have not monetized their gains I will overestimate the gains to office and corruption though if inefficient politicians are denied tickets reporting improves over time 8 / 17
The gains to office Winning increases candidate assets by Rs. 6.6 million or 135% 9 / 17
The gains to office Winning increases candidate assets by Rs. 6.6 million or 135% 9 / 17
Robustness tests The results are unchanged using different winning margins/bandwidths using placebos ( winning at different thresholds) controlling for selection (Heckman), covariates (OLS, matching) 10 / 17
What proportion of politicians are suspect? Rs. 0.4-2.7 million of the gain in assets could be legitimate 11 / 17
What proportion of politicians are suspect? Rs. 0.4-2.7 million of the gain in assets could be legitimate 11 / 17
Is this really corruption? We might expect suspect-politicians to shift wealth to family members shift wealth to jewelry shift wealth to (non-agricultural) land have a previous criminal record be more likely to be charged with crimes be less likely to have tax IDs 12 / 17
Is this really corruption? We might expect suspect-politicians to shift wealth to family members shift wealth to jewelry shift wealth to (non-agricultural) land have a previous criminal record be more likely to be charged with crimes be less likely to have tax IDs 12 / 17
Correlates of being suspect Figure: Pr(suspect-politician) 13 / 17
Effects of suspect politicians? Figure: Pr(winning in 2008/09) 14 / 17
Tentative conclusions, implications Preliminary estimates unexpectedly varied gains to office less-than-expected corruption perceived and actual corruption are different fewer people might oppose reform than expected Correlates positive: governing coalition, cabinet members, from Delhi negative: ST politicians, from Rajasthan and Orissa no effect: initial wealth Outcomes corruption improves the chances of winning the next election 15 / 17
Thank you 16 / 17
Remaining questions Differences between national and state-level patterns? Causal estimates of effects of attributes Which cabinet members do well? 17 / 17