Abstract
We present the findings of an experimental game of bribery based on Mookherjee and Png's model where inspectors are hired to find evidence against firm owners who have violated some regulation. Inspectors choose costly effort that determines the probability of finding evidence and allows them to fine the owner. Bribes may occur before or after the inspector has exerted effort and found evidence. Inspectors consistently demanded bribes below the Nash equilibrium prediction and exerted effort below the payoff-maximizing level. These results raise questions about the robustness of theoretical results regarding the efficiency of using bribes to motivate inspections.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
In their summary of the literature, Dusek et al. [2004] identify other social ills associated with high rates of corruption and bribery such as: increased poverty rates and decreased foreign private investment, economic growth rates, government revenues and infrastructure, and social equality.
See Samuel [2009] for a theoretical model of preemptive and ex-post bribery.
The most similar experiments that use endogenous effort are Dittrich and Kocher [2006] and Fehr et al. [1997], but they examine the role of different wage contracts on employee effort. Specifically, they find that, despite effort being non-contractible and unobservable in their experiment, workers frequently exert effort that is above the payoff-maximizing level when they perceive that they will be rewarded with higher wages (even when these higher wages are not guaranteed).
In real situations, agents may have to manage reputations or repeated play, but the basic structure is still one of bargaining after sunk costs. We focus on the one-shot game instead of one of reputation or repeated play in order to reflect Mookherjee and Png [1995] most closely.
As discussed in Lambsdorff and Frank [2007], the relevant “sharing norm” in a game of corruption is reciprocity, not fairness. Whereas fairness indicates a sharing norm that would proscribe society-harming corruption, reciprocity permits the provision of surplus gained from illegal behavior.
To illustrate, Abbink et al. [2002] also analyze bribery as a reciprocity game, but implement it differently. In their game the first player offers a bribe to the second, who may reject the offer. The second then chooses whether to create a large surplus for the first at a small personal cost. The game features right of refusal, reneging, transfer fees, multiplication of transferred values, an exogenous chance of being caught and punished, and externalities levied against other players. Given their interest in long-term relationships, they implement 30-round sessions in which players have the same partner in every round. Their results suggest that repeated interaction strengthens reciprocal behavior in that first movers are more likely to transfer money (pay bribes) to second movers.
These experiments also find that when positive and negative reciprocity are observed in the lab, negative reciprocity appears to occur more frequently.
Recognizing the challenges of using field data to understand bureaucrat behavior is not the same as establishing the external validity of experiments on corrupt behaviors. As Abbink [2006] notes, the standard use of students and the inherently artificial nature of experiments weakens their external validity but are still an excellent way to test theoretical models (such as we do with Mookherjee and Png [1995]). In response, Abbink notes that “…[field research] suffer[s] from the noise, identification problems, and lack of control” (p. 1). Although not a direct test of external validity, Barr et al. [2004] implement an embezzlement experiment with health workers, and their results generally support prior experimental research on students, giving some empirical support for the external validity of corruption experiments. Using potentially corrupt bureaucrats as participants has problems as well, as they would be aware of being observed making choices about behaviors similar to their own illegal activities and likely behave in contrived ways. Abbink [2006, p. 436] notes that “(a)lthough it is naturally impossible to prove the external validity of experimental results, such parallel investigations could dramatically add to the robustness of the stylized facts we can identify in laboratory experiments.” Dusek et al. [2004] claim experiments are useful for designing incentive-compatible and effective anti-corruption measures. The fact that real situations often feature indefinitely repeated interactions indicates that more trust and reciprocity likely occur in real settings, and so experiments would underestimate the extent of corruption. Knowing the direction of bias informs us how to interpret experimental results, and thus they are still useful for analyzing public policy. For example, they note that since college students seem to ignore spillovers onto other participants, the implication is clear that corrupt bureaucrats would likely not either, and that moral suasion and ad campaigns will be of little help. Further, if college students have trouble correctly using percentages of being caught, respond to wage increases by decreasing corrupt behaviors, exhibit inequality aversion, and respond to beliefs about the prevalence of corruption, we can safely assume bureaucrats and bribe-payers likely exhibit these stylized behaviors as well. Thus, while experiments may not exhibit external validity in the overall magnitude of bribery, they would be informative of underlying behaviors.
In order to prevent the inspector from falsely accusing the owner, inspectors must present verifiable evidence before they can collect a fine from the owner.
The choice with the lowest variance is 0.70, whereas the choice with the smallest potential loss is 0 cents. Thus, strongly risk-averse agents will always choose 0.70, whereas strongly loss-averse agents will always choose 0 cents. In our experiments, however, we rarely see either.
Note bribe demands were in 0.05 increments. Although profit-maximizing owners would be indifferent to a bribe request of 1.50, we assume (here) that inspectors would not offer a weakly dominated bribe request.
We are grateful to an anonymous referee for raising these issues.
Although previous experimental research on bribery suggests an externality does not affect the likelihood of bribes [Abbink et al. 2002], the presence of a negative externality (due to the firm's non-compliance) may encourage more intensive inspections if inspectors are empathetic or hold norms of fairness. Such findings would also have implications for Harrington's [1988] model of inspections and firm compliance, and its experimental implementation in Cason and Gangadharan [2005].
We note there are limited design options available. Our approach introduces contagion, potentially leaving potential for both training effects (an attempt to influence future partners indirectly through choices against current partners) and reputation-building. Although we did not attempt to measure the effects of this contamination, we believe the effects to be smaller than that from repeated pairings, which maximize the incentives for training and reputation effects. Further, such a super-game would neutralize the timing effect of bribery, as one period's ex-post bribe would be directly followed by the next period's preemptive bribe. A third alternative would be to bring in participants for a single round of play. Given the high cost of attracting participants and substantial time to give instructions, such an approach was infeasible. In our econometric analysis below we analyze all rounds of data (using HLM techniques), but as a robustness check also analyze only the first round of play. (See Harrison [2007] for further caveats on examining only the first round of data).
Note that two practice rounds of each treatment were done without financial incentives to allow for learning about the flow and structure of the treatment and questions. We recorded and report only those rounds played for money. Participants were informed of the number of rounds of each treatment verbally and through the format of their information-recording forms.
Instructions for the Full treatment are included in the Appendix.
Hypothesis 1 is one-tailed, with evidence against the hypothesis being bribes below 1.45; since permissible bribes were in 0.05 increments, 1.45 is the largest bribe demand that gives the owner a non-zero payoff.
Since we expect to observe some bribe demands being rejected, which is off-equilibrium path behavior, we test for this hypothesis.
Cameron and Trivedi [2005] state that when the structure of the data consists of a short panel, this is the correct way to model the hierarchical structure. We include session as a level in order to account for possible session-level effects. However, our qualitative findings remain unchanged whether or not session is included as a level in our HLM estimates.
The maximum likelihood estimator is preferred because it is more efficient and its standard error is weakly greater than the Ordinary Least Squares (OLS) estimator. If one is not interested in the within-cluster dependence, then the OLS “sandwich” estimator may also be used. This sandwich estimator takes this clustering into account to produce “robust” standard errors [see Rabe-Hesketh and Skrondal 2008, pp. 68 ff.]. For robustness we estimate our model using both estimators but only present the HLM estimates. OLS results are included for comparison, and not discussed here. Similarly, we pool the sessions at various levels (both identical sessions and all four sessions, when applicable) and find no qualitative difference in our results.
This HLM specification allows for both random coefficients and slopes by owner and is effectively an extension of a random effects logit model.
Since we are only interested in estimating and conducting hypothesis tests on the coefficients, we do not discuss these variance estimates.
References
Abbink, K. 2006. Laboratory Experiments on Corruption, in International Handbook on the Economics of Corruption, edited by S. Rose-Ackerman. Northampton, USA: Edward Elgar Publishing Limited, 418–437.
Abbink, K., B. Irlenbusch, and E. Renner . 2000. The Moonlighting Game: An Experimental Study on Reciprocity and Retribution. Journal of Economic Behavior and Organization, 42: 265–277.
Abbink, K., B. Irlenbusch, and E. Renner . 2002. An Experimental Bribery Game. Journal of Law, Economics, and Organization, 18: 428–454.
Abbink, K., and H. Hennig-Schmidt . 2006. Neutral Versus Loaded Instructions in a Bribery Experiment. Experimental Economics, 9: 103–121.
Bardhan, P. 2006. The Economist's Approach to the Problem of Corruption. World Development, 34: 341–348.
Barr, A., M. Lindelow, and P. Serneels . 2004. To Serve the Community or Oneself: The Public Servant's Dilemma, World Bank Policy Research Working Paper 3187.
Becker, G., and G. Stigler . 1974. Law Enforcement, Malfeasance and Compensation of Enforcers. Journal of Legal Studies, 3: 1–18.
Berg, J., J. Dickhaut, and K. McCabe . 1995. Trust, Reciprocity and Social History. Games and Economic Behavior, 10: 122–142.
Binmore, K., A. Shaked, and J. Sutton . 1989. An Outside Option Experiment. The Quarterly Journal of Economics, 104: 753–770.
Boycko, M., A. Shleifer, and R. Vishny . 1995. Privatizing Russia. Cambridge: MIT Press.
Bryk, A.S., and S.W. Raudenbush . 1992. Hierarchical Linear Models. California USA: Sage Publications.
Camerer, C., and R. Thaler . 1995. Anomalies: Ultimatums, Dictators and Manners. Journal of Economic Perspectives, 9: 209–219.
Cameron, A., and P. Trivedi . 2005. Microeconomics: Methods and Applications. Cambridge: Cambridge University Press.
Cason, T.N., and L. Gangadharan . 2005. An Experimental Study of Compliance and Leverage in Auditing and Regulatory Enforcement. Economic Inquiry, 44: 352–366.
Castro, F.M. 2011. To Bribe or not to Bribe: An Experimental Analysis of Corruption, DSSGES Working Paper, Università Mediterranea di Reggio, Calabria.
Cherry, T., P. Frykblom, and J. Shogren . 2002. Hardnose the Dictator. American Economic Review, 92: 1218–1221.
Dittrich, D., and M. Kocher . 2006. Monitoring and Pay: An Experiment on Employee Performance under Endogenous Supervision, CEPR Discussion Paper No. 5962.
Duffy, J., and N.J. Feltovich . 2006. Words, Deeds and Lies: Strategic Behavior in Games with Multiple Signals. Review of Economic Studies, 73: 669–688.
Dusek, L., A. Ortmann, and L. Lizal . 2004. Understanding Corruption and Corruptibility Through Experiments: A Primer, CERGE-EI Discussion Paper No. 2004 -136.
Fehr, E., S. Gachter, and G. Kirchsteiger . 1997. Reciprocity as a Contract Enforcement Device: Experimental Evidence. Econometrica, 65: 833–860.
Greenhouse, S. 2008. School Bus Safety Officers are Accused of Soliciting Bribes, New York Times (May 14).
Guth, W., and E. van Damme . 1998. Information, Strategic Behavior, and Fairness in Ultimatum Bargaining: An Experimental Study. Journal of Mathematical Psychology, 42: 227–247.
Harrington, W. 1988. Enforcement Leverage when Penalties are Restricted. Journal of Public Economics, 37: 29–53.
Harrison, G. 2007. House Money Effects in Public Good Experiments: Comment. Experimental Economics, 10: 429–437.
Hunt, J., and S. Laszlo . 2005. Bribery: Who Pays, Who Refuses, What are the Payoffs? NBER Working Paper 11635.
Klitgaard, R. 1988. Controlling Corruption. Berkeley, CA: University of California Press.
Lambsdorff, J., and B. Frank . 2007. Corrupt Reciprocity — An Experiment, University of Passau Discussion Paper V-51-07.
Lubin, N. 2003. Who's Watching the Watchdogs? Journal of International Affairs, 56: 43–57.
Mookherjee, D., and I. Png . 1995. Corruptible Law Enforcers: How Should They Be Compensated? The Economic Journal, 105: 145–159.
Olken, B., and P. Barron . 2007. The Simple Economics of Extortion: Evidence from Trucking in Aceh, NBER Working Paper No. 13145, Cambridge: National Bureau of Economic Research.
Oosterbeek, H., J. Sonnemans, and S. vanVelzen . 1998. Bargaining with Endogenous Pie Size and Disagreement Points; A Holdup Experiment, Working Paper.
Phillips, O., R. Battalio, and C. Kogut . 1991. Sunk and Opportunity Costs in Valuation and Bidding. Southern Economic Journal, 58: 112–128.
Polinsky, A., and S. Shavell . 2001. Corruption and Optimal Law Enforcement. Journal of Public Economics, 81: 1–24.
Rabe-Hesketh, S., and A. Skrondal . 2008. Multilevel and Longitudinal Modeling Using Stata, 2nd ed., College Station, TX: Stata Press.
Rubin, E. 2006. In the Land of the Taliban, New York Times Magazine, October 22: 91.
Samuel, A. 2009. Preemptive Collusion among Corruptible Law Enforcers. Journal of Economic Behavior and Organization, 71: 441–450.
Schmitt, P. 2004. On perceptions of Fairness: The Role of Valuations, Outside Options, and Information in Ultimatum Bargaining Games. Experimental Economics, 7: 49–73.
Schulze, G., and B. Frank . 2003. Deterrence Versus Intrinsic Motivation: Experimental Evidence on the Determinants of Corruptibility. Economics of Governance, 4: 143–160.
Simpson, J., and Z. Bojovic . 2005. Investigation: Serbia Losing Customs Corruption Battle, Institute for War and Peace Reporting, BCR No 541, 11-Feb-05.
Svensson, J. 2003. Who Must Pay Bribes and How Much? Evidence from a Cross Section of Firms. The Quarterly Journal of Economics, 118: 207–230.
Acknowledgements
We wish to thank three anonymous referees, Emmanuel Dechenaux, and Christopher Morrell, as well as seminar participants at the 2007 International Industrial Organization Conference and at the workshop on Law, Economics, and Institutions — University of Birmingham (UK). Alden Van Slokema, Andrew Vlietstra, and Laura Samuel provided excellent assistance in running our experiments. Financial support from the Calvin College Center for Social Research, Loyola University — Maryland, and Grand Valley State University is also gratefully acknowledged.
Author information
Authors and Affiliations
APPENDIX
APPENDIX
Instructions for the full bribe treatment
General information
You are going to participate in an experiment that simulates an inspector–inspectee relationship. These instructions explain how the decisions you and the other participants make determine the total cash payment you will receive at the end of the session. During the session you must keep track of your payments on the forms provided to you. At the end of the session, your total profits will be paid to you in cash.
Background
The Department of Health has just instituted a new set of health code requirements for restaurants. All restaurant owners are currently in violation of these new health code requirements. The Department of Health has hired health inspectors to investigate these restaurants. Inspectors must exert costly effort in order to collect evidence against a restaurant owner. By exerting a higher degree of effort inspectors are able to find evidence of the owner's guilt with greater probability. However, a higher degree of effort costs more than a lower degree of effort. This relationship between the inspector's cost of effort is given by the following table:
For example, if an inspector incurs $0.20 of effort, that inspector will have a 40 percent chance of finding evidence (of health code violations) against the owner.
Each time an inspector successfully collects evidence against a restaurant, the inspector is paid $1 minus the cost of effort. If the inspector does not successfully collect evidence, the inspector still pays the cost of effort whereas the owner is receives $1.50.
Your roles
You have the same role assigned in the previous round, either an Owner or an Inspector. As earlier, you will never know the identity of your partner for any round, and are matched in such a way that you will never play the same person twice in this session.
Recording your payoffs
All inspectors must have a short form and an inspector's long form and all owners an owners’ long form. Please do NOT write down anything other than the requested information (additional messages or comments will result in a penalty for the round).
The game
The stages of the game take place in the following order:
Stage 1: Keeping in mind that no owners are in compliance with the health code, the inspector specifies a bribe amount (in 5 cent increments) that he/she is willing to accept in exchange for not inspecting the restaurant. Inspectors may choose NOT to request a bribe at this stage.
Records: The inspector records this initial decision on the short and long forms. The inspector is then matched with an owner who receives the inspector's offer on the short form. Please make sure to fill out all the information pertaining to the inspector on both the short and the long forms.
Stage 2: The restaurant owner responds by either accepting or rejecting the bribe.
Records: The owner records their choice on both the short and long forms.
Stage 3: If the owner accepts the bribe, the inspector does not exert any effort and the round ends. In this case the owner gets $1.50 minus the bribe and the inspector receives the amount of the bribe. Inspectors will be informed if their bribe is rejected, and they will proceed to stage 4.
Records: Owners and inspectors record their payoffs (if applicable) and actions on the long form.
Stage 4: If the Inspector chose not to request a bribe in stage 1 or the inspector's bribe was rejected in stage 2, the inspector must choose a level of effort and a probability of finding evidence against a guilty owner, according to the above table. Once all inspectors have chosen their effort level a die is thrown.
If the die roll is equal to or below the inspectors’ chance of finding evidence, the inspector receives a payoff of $1 minus the cost of effort. These inspectors have found evidence against their corresponding restaurant owner and the corresponding owner receives a payoff of $0. Inspectors who find evidence against their assigned owner will be announced.
If the die roll is above the inspectors’ chance of finding evidence, the inspector receives a payoff of $0 but still incurs the cost of effort. These inspectors have not found any evidence against their corresponding restaurant owner and the owner receives a payoff of $1.50. Note that it is possible to have negative profit for a round.
Records: Inspectors record the outcome of the dice role on the “Inspection Outcome” column whether they found Evidence or No-evidence, and their earnings in that round. Owners record the result of the inspection and their earnings.
Stage 5: If the Inspector has found evidence (according to the outcome of the die roll in stage 4), he/she must specify a bribe amount that he/she is willing to accept in exchange for not reporting the restaurant. Please realize that you may choose NOT to request a bribe at this stage.
Records: ALL INSPECTORS, regardless of whether they have found evidence against the Owner, must fill out a bribe form. All inspectors must state their effort level and whether he has found evidence or not found evidence. If no evidence has been found, Inspectors CANNOT request a bribe. However, if the evidence has been found, Inspectors may request a bribe, but do not have to.
Stage 6: Owners choose to accept or reject their Inspector's bribe. If the bribe is accepted the Owner receives $1.50 minus the bribe paid and the Inspector receives the bribe minus the cost of effort. If the bribe is rejected, an owner receives $0 for that round and the Inspector receives a payoff of $1 minus the cost of effort.
Records: Inspector–Owner pairs record their payoff on their long forms.
Are there any questions?
Rights and permissions
About this article
Cite this article
Lowen, A., Samuel, A. Bribery and Endogenous Monitoring Effort: An Experimental Study. Eastern Econ J 38, 356–380 (2012). https://doi.org/10.1057/eej.2011.18
Published:
Issue Date:
DOI: https://doi.org/10.1057/eej.2011.18