Bureaucratic Competition and Corruption

The experimental treatments were systematically designed to examine critical aspects of the theoretical model: the interplay between bureaucratic competition and corruption, and the influence of GDP significance in the cadre evaluation process. Specifically, the first treatment (T1) featured high competition coupled with moderate GDP importance, serving as our baseline scenario.

In the second treatment (T2), GDP significance remained constant, but competition was reduced. This adjustment positioned candidate A more favorably, enabling us to directly assess the impact of diminished bureaucratic competition on corruption behaviors.

The third treatment (T3) retained high competition levels but significantly lowered the GDP importance in evaluations. This allowed us to investigate how corruption dynamics shift when GDP performance becomes less critical in the promotion criteria.

Finally, the fourth treatment (T4) substantially increased the significance of GDP while maintaining other parameters identical to the first treatment. This scenario provided a clear contrast to evaluate how corruption evolves when economic performance dominates the evaluation process.

In summary, the theoretical model predicts the following relationship predictions:

As the competition level increased from T2 to T1, the bribery level would increase, and the graft level would decrease: Bribery (T1) > Bribery (T2), Graft (T1) < Graft (T2). When GDP importance is less significant, bribery would be no different but graft would increase significantly: Bribery (T1) = Bribery (T3), Graft (T1) < Graft (T3). When GDP becomes very important in the evaluation process, both bribery and graft would be smaller than the baseline: Bribery (T1) > Bribery (T4), Graft (T1) > Graft (T4).

These are the hypotheses we are testing in the experiment.

2025-12-15

2026-01-30

Bribery (office-buying corruption) and Graft (rent-seeking corruption)

Each experimental treatment consisted of one session comprising 60 periods. In each period, two candidates—labeled A and B—competed against each other. At the beginning of every session, 20 participants were randomly and anonymously assigned to roles (candidate A or candidate B). These roles remained fixed for the duration of the 60-period session. Participants knew their own roles but did not know the identities of their counterparts.

Within each period, participants were randomly assigned to one of ten groups, each containing exactly two individuals, one candidate A and one candidate B. Although group composition changed every period, each participant retained their original role. For example, a participant designated as candidate A could be paired with candidate B in group 3 for the first period and reassigned to candidate B in group 8 in the subsequent period. This continuous reshuffling of groups aimed to minimize potential collusion among candidates.

Neutral language was deliberately employed throughout the experiment to prevent unintended positive or negative connotations beyond the intended incentive structures. Participants made two decisions in each period: (a) Investment in good 1, ranging from 0 to 20; (b) Investment in good 2, which could not exceed the investment made in good 1.

Participants were explicitly informed that investment in good 1 decreased their probability of winning, while investment in good 2 increased their chances.

At the outset of the session, participants were clearly informed about their payoff structure. Their payoffs depended on their probability of winning, influenced by their investments in both goods. Each investment incurred a variable cost, while the remaining amount after subtracting the investment in good 2 from good 1 provided an additional payoff.

To facilitate informed decision-making, an "expected payoff calculator" was available on participants' decision screens. This tool allowed them to explore various investment combinations and their corresponding payoffs before finalizing their decisions. Participants could utilize the calculator without restriction throughout each period. Additionally, a log sheet was provided to participants, enabling them to systematically track earnings across periods, thereby gaining insights into the implications of their previous investment choices.

Finally, control questions were administered at the beginning of each session to ensure that all participants thoroughly understood the instructions and mechanics of the experiment.

Randomization is done by a computer algorithm.

Individual

No

(# clusters equal to number of observations below.)

4800 observations consisting of four treatments of 20 students and 60 rounds in each treatment.

Each treatment consists of 1200 observations.