Combining topdown and bottomup accountability Evidence from a
Combining top-down and bottom-up accountability: Evidence from a bribery experiment Danila Serra University of Oxford and CSAE Fourth Ph. D. Presentation Meeting January 18 th, 2009
Outline of the presentation n Collusive bribery n Top-down vs. Bottom-up n Methodology: the bribery experiment n Results n Conclusion Serra, D. "Combining top-down and bottom-up accountability" 2
The Top-down Approach n Evidence from randomized interventions: q q q Olken (2007 JPE): effective in reducing missing expenditure in a public road construction – 100% probability of detection Kremer and Chen (2001): ineffective, due to lack of committment of the monitors (monitors left unaccountable) Duflo and Hanna (2005): effective in reducing teachers’ absence – 100% probability of detection Serra, D. "Combining top-down and bottom-up accountability" 3
The Bottom-up Approach n Evidence from randomized interventions: q q q Olken (2007 JPE): ineffective in reducing corruption in the form of total missing expenses for a road construction; Banerjee, Deaton and Duflo (2005): ineffective in reducing health workers’ absence Bjorkman and Svensson (2008 QJE): effective in improving health outcomes Serra, D. "Combining top-down and bottom-up accountability" 4
Research question Could a system that combines top-down auditing and bottom-up monitoring effectively curb corruption, even in a weak institutional environment (i. e. low probability of detection)? Serra, D. "Combining top-down and bottom-up accountability" 5
Research question Put another way Could top-down auditing be effective in a weak institutional environment if it was triggered by bottom-up monitoring? Could bottom-up monitoring be effective if it was able to activate top-down auditing with some, even low, probability? Serra, D. "Combining top-down and bottom-up accountability" 6
The bribery game in detail n n n n In each session: 5 “private citizens”, 5 “public officials” and 5 “other members of society” Citizens and officials play in pairs The public official can demand a bribe b from the citizen in exchange for a better, quicker or illegal service The “corrupt” service gives the citizen a monetary gain G In order to provide the corrupt service the official has to sustain an administrative cost K The provision of the corrupt service generates a cost h for “other members of society” If citizen and public official do not engage in corruption they get Yc and Yp Play is one-shot and anonymous – No trust issues Serra, D. "Combining top-down and bottom-up accountability" 7
Experimental Treatments n Control: no monitoring; n Top-down auditing: If corruption takes place, there is an exogenous probability pe that the official (but not the citizen) is punished. n “Combined” monitoring: q The citizen can “report” the corrupt official. q If corruption takes place and the citizen reports the official, there is an exogenous probability pe (the same as before) that the official is punished. Serra, D. "Combining top-down and bottom-up accountability" 8
The Experimental Design The “combined” monitoring treatment 1) The decision to “report” is not costly to the citizen: perfect information on how to report and full leniency § Why? This is collusive corruption… 2) Citizens’ reports trigger top-down auditing, i. e. the probability of detection under the “combined system” is lower than under top-down auditing. n Why? In order to have “conservative” results… Serra, D. "Combining top-down and bottom-up accountability" 9
The Experimental Design The “combined” monitoring treatment 3) The decision to report the official is independent of the size of the bribe demanded, i. e. the citizen decides whether or not to report before knowing the amount of the bribe demanded. § Why? Two reasons… 4) Top-down auditing is triggered only if the citizen reports the official and pays the bribe, i. e. if the corrupt service is provided. n Why? Need for evidence… Serra, D. "Combining top-down and bottom-up accountability" 10
Public official’s decision-making under no monitoring n n If the official and the citizen are self-interested, the official will demand for a bribe if and only if b > K The citizen will pay if and only if: b < G SPNE: If G > K, then the official will demand a bribe equal to (G – ε ); the citizen will pay the bribe and the other members of society will suffer a monetary loss. Serra, D. "Combining top-down and bottom-up accountability" 11
Under top-down auditing n If corruption takes place the official will receive a sanction S and will have to give back the bribe with probability pe. A risk-neutral official will now demand a bribe if and only if: (1 - pe)(Yp + b – K) + pe (Yp – K – S ) > Yp b > (K + pe. S)/(1 – pe) n Nothing changes for the citizen. SPNE: If G > (K + pe. S)/(1 – pe), then the official will demand a bribe equal to (G –ε ). The citizen will pay the bribe…. Serra, D. "Combining top-down and bottom-up accountability" 12
Under combined monitoring n Assuming that the official expects that the citizen will report him with probability pc he will now demands a bribe if: (1 - pc pe)(Yp + b – K) + pc pe (Yp – K – S ) > Yp b > (K + pc pe. S)/ (1 – pc pe) n Nothing changes for the citizen. SPNE: If G > (K + pc pe. S)/ (1 – pc pe) the official will demand a bribe equal to (G – ε); the citizen will pay the bribe, and either report or not report. Serra, D. "Combining top-down and bottom-up accountability" 13
Parameterization n n Yp = Yc = 35 G (1 G = € 0. 30) Yo = 25 G G = 15 G K = 5 G h = 4 G Under external monitoring and combined accountability: n Pe=0. 04 n S=10 G + b Under these parameters the monetary incentives for a (risk-neutral) public official to engage in corruption under no monitoring, external monitoring and bottom-up monitoring are identical Serra, D. "Combining top-down and bottom-up accountability" 14
Predictions 1. If the public officials are self-interested and risk-neutral, then: All officials will act corruptly and demand a bribe equal to 14. We should observe no differences in behaviour across treatments 2. If the public officials are self-interested and risk-averse then: We should observe less corruption under top-down monitoring than under “combined” monitoring and no monitoring 3. If the public officials are intrinsically motivated and suffer a moral cost for harming others and/or acting illegally, they migt abstain from corruption or demand a bribe lower than 14. Still: We should observe no differences in behaviour across treatments Serra, D. "Combining top-down and bottom-up accountability" 15
Experimental sessions 12 sessions involving a total of 180 Oxford University students, 60 “citizens, 60 “officials” and 60 “other members of society” Recruitment via emails and posters in November 2007 Experimental sessions No monitoring 4 sessions (60 participants) Top-down monitoring 4 sessions (60 participants) “Combined” monitoring 4 sessions (60 participants) Serra, D. "Combining top-down and bottom-up accountability" 16
Results: Public officials’ decision to demand a bribe Serra, D. "Combining top-down and bottom-up accountability" 17
Why did we get this surprising result? Three possible reasons: 1) Betrayal aversion (Bohnet and Zeckhauser, 2004) 2) Fear of social disapproval 3) Conjenction fallacy (Tversky and Kahneman, 1983) Serra, D. "Combining top-down and bottom-up accountability" 18
Conclusions n Top-down auditing in a weak institutional environment may prove ineffective n Anti-corruption mechanisms based on top-down auditing triggered by bottom-up monitoring can be highly effective in curbing corruption, even in a weak institutional environment. n WHY? Future work Serra, D. "Combining top-down and bottom-up accountability" 19
Scope for future work n What if citizens’ reports are conditional on the size of the bribe? n Does the result hold under well-functioning institutions? n Could we bring the experiment to the field? Serra, D. "Combining top-down and bottom-up accountability" 20
THANK YOU danila. serra@economics. ox. ac. uk Serra, D. "Combining top-down and bottom-up accountability" 21
The “combined” monitoring treatment Public Official Demand b = 20 b=1 Citizen Report ND (35, 35) NR Citizen Pay NP P NP Nature (35, 35) p = 0. 04 (20, 50 -b) p=0. 96 (30+b, 50 -b) (35, 35) (30+b, 50 -b) Serra, D. "Combining top-down and bottom-up accountability" 22
Results: The decision to demand a Dependent variable: bribe Dummy variable equal to 1 if the official demanded a bribe Top-down “Combined” (1) -0. 08 [0. 493] -0. 28 [0. 028]** Female Age “Religious person” Economist Observations Pseudo R-squared 60 0. 105 (2) -0. 13 [0. 279] -0. 32 [0. 032]** 0. 05 [0. 535] -0. 01 [0. 192] 0. 11 [0. 064]* 0. 12 [0. 080]* 60 0. 179 Note: the p values (in brackets) correspond to robust standard errors, which have been adjusted to account for clustering within sessions. We report marginal effects of continuous variables and the effect of a change from 0 to 1 for dichotomous variables. * significant at 10%; ** significant at 5%; *** significant at 1% Serra, D. "Combining top-down and bottom-up accountability" 23
Results: The size of the bribe Equal payoff Nash players No bribe Serra, D. "Combining top-down and bottom-up accountability" 24
Results: The size of the bribe Dependent variable: Amount of the bribe demanded OLS on selected sample OLS on full sample (1) (2) (3) (4) Top-down “Combined” 1. 92 [0. 003]*** 1. 45 [0. 227] Female Age “Religious person” Economist Constant Observations R-squared 10. 47 [0. 000]*** 51 0. 152 1. 74 [0. 018]** 1. 23 [0. 324] -1. 25 [0. 047]** -0. 08 [0. 210] 0. 14 [0. 843] -0. 44 [0. 587] 12. 98 [0. 000]*** 51 0. 252 Serra, D. "Combining top-down and bottom-up accountability" 1. 20 [0. 195] -1. 60 [0. 086]* 9. 95 [0. 000]*** 60 0. 063 0. 64 [0. 631] -2. 01 [0. 093]* -0. 89 [0. 519] -0. 23 [0. 235] 1. 35 [0. 317] 0. 77 [0. 542] 14. 94 [0. 005]*** 60 0. 124 25
Results: Citizen’s willingness to pay a bribe Serra, D. "Combining top-down and bottom-up accountability" 26
Results: Citizens’ maximum payable bribe Serra, D. "Combining top-down and bottom-up accountability" 27
Reports and maximum payable bribes Serra, D. "Combining top-down and bottom-up accountability" 28
- Slides: 28