The Moral Economy of Supervisor Kickbacks and Its Significance for Corruption

Intervention: The top-up was a cash transfer given to police agents on some days. The cash transfer was unconditional, thus involved no quid-pro-quo, and hence did not amount to a bribe. The average top-up given to each agent amounted to 17.62 in a day. For each agent, the value of the top-up received on a given day was calculated in which they received a top-up was calculated to be equal to their total daily income. To calculate this amount, we estimated the number of bribes that each team of agents received, on average, daily, in preparatory work, and added those to their official wages. Each team was selected an equal number of days to receive the top-up. What days they received it was random.

We embedded two additional sub-treatments:

Sub-treatment 1: We also randomized whether the project staff informed the supervisor of a specific top-up taking place on a given day for a team. In one treatment arm, we did not disclose the specific day and time of the top-up to the supervisor. This was consistent with our agreement with the supervisors, who allowed for occasional transfers to their agents. In another arm, we notified the supervisor in the morning and prior to the top-up to a given team that day, that the team had been selected for a top-up on that day.

Sub-treatment 2: the level of the top-up; we assigned high and low values of the top-up in randomly selected teams x days

2015-06-20

2015-07-20

We focus on the monetary kickbacks from the top-up on a given day. To measure this, an auditor asked the agents at the end of each day to describe how they allocated the top-up. Since each team may have received a different top-up and may have shared a different amount, the auditor conducted private conversations with each team every evening.

A pillar of this study is that we take advantage of the fact that, following six months of preparation, we had created trust relations between the agents and the top-up auditor, and the auditor had been introduced to the agents by project members who were already well-known and respected by the traffic police staff. By the time the study started, the agents already had good relations with the auditor. This trust was strengthened as the auditor's mission was to ensure the agents were paid. The first mandate of the auditor was to ensure that the agents received the cash they were supposed to.

n addition to verifying whether the agents that should receive a top-up did receive it, the auditor also asked the agents to describe how they allocated the top-up. Since each team may have received a different top-up and may have shared a different amount, the auditor conducted private conversations with each team every evening.

It was made clear at the start of this component that the auditor had no interest in observing low or high transfers, and that the amount transferred to the supervisor had no bearing on future benefits. Agents were made aware that the reported sharing would have no effect on future top-ups. It was evident in informal conversations after a few days that the agents felt comfortable talking about the kickbacks with the auditor, and that those conversations were not sensitive.

This end point allows us to measure, for example the "marginal kickback rate" and the "average kickback rate", defined respectively as the USD payment to the supervisor for each additional USD of income, and the average USD payment to the supervisor over the income given in top-up.

We also measure the behavior of the agents. Notably, we collect proxies of effort on the official prerogatives of the traffic police agency (absenteeism, traffic jams, accidents, detentions and escorts to the police station) and on the unofficial prerogatives (bribe taking and driver harassment)

To measure the behavior of the agents, we draw on the data from the method described in Sanchez de la Sierra, Xie, Titeca, Malukisa, Lameke (2022): "The Real State: Inside the Congo's Traffic Police Agency", consisting in observers in the street observing the police officers and recording their behavior in a smart phone.

Main Intervention: Each team of agents was entitled to receive the top-up on half of the study days. We randomly assigned, within each team strata, an equal number of days for all team members to receive the top-up. Thus, the randomization clusters are 337 team x days.

Sub-treatment 1: Supervisor notified. Amongst the team-days that had been randomly selected for a top-up, we randomly selected team-days to the supervisor notified vs supervisor not notified treatments, stratifying by team of agents.

Sub-treatment 2: We randomly assigned high top-up vs. low top-up, among the team-days that were assigned to receive a top-up. Separately for each team, and among the days they were selected to receive the top-up, we randomly assigned days to high top-up vs. low top-up within supervisor notified and not notified treatment status. This generates three randomly assigned levels: zero, .5, and 1.5.

Between 6 am and 7 am, the project coordinator placed calls to each supervisor to provide information about the top-up taking place, and the specific amount of the top-up for that day, only for the teams that had been assigned to notifying the supervisor---but not for the rest. Supervisors knew that they were not notified of all the payments and the number of teams treated in a day was random, hence they were unable to guess whether other teams received payments. Between 7 am and 8 am, the coordinator met with the couriers, and instructed them on which teams had been selected for each treatment arm, withdrew cash from the project's safe, and handed it over to the couriers.

Randomization done in office by a computer. Implementers were not informed of the random assignment until each day at 6am when they had to deliver cash, to prevent them from using that information in ways that could undermine statistical credibility.

Cluster: Team of police agents in the street X day

Yes

375 Team of police agents in the street X day clusters

Approximately 1,575 police agents in the street X day observations.

Control: 186

Top-up: 186, of which:

Sub-treatment 1 (supervisor notified): control (93), treatment (93)

Sub-treatment 2 (top-up level): low (63), high (63), not included in top-up level randomization (60)

N/A