Incentives for Safety: A Field Experiment in the Ugandan Transit Industry

2023-02-09

2023-02-17

Driving behavior:
- Dummy variable indicating whether a speeding violation (driving above 50 kph) has been committed in a given day
- Average excess speed in a given day

Labor market outcomes:
- earnings
- labor supply
- productivity

Driving behaviors and labor market outcomes are respondent-by-day measures. The frequency reflects the unit of observation (respondent by day).

Demand for commitment:
- Dummy variable indicating whether respondent picks a dominated contract
- Willingness to pay for the contract

Labor market outcomes:
- share of recurrent customers
- number of new contacts from customers

Driving behavior:
- Air pollutant emissions: carbon footprint
- Expected average fine

Demand for commitment of a respondent is elicited at the end of the study period. The commitment device is a financial contract that pays based on speeding behavior (details about the contract in the design section). The demand elicitation follows the same method adopted in AEARCTR-0010635.
Demand for commitment of a respondent is elicited at the end of the study period.
Labor market outcomes and carbon footprint of a respondent are calculated at daily level.

Expected average fine based on fines for speeding declared at baseline by the study sample.

360 motor-taxi drivers, all operating from different taxi stations in the metropolitan area of Kampala, Uganda, are randomly assigned to receive an offer to access financial incentives for ten days. Each respondent is asked to answer a phone survey every two days for a time period of two weeks.

Participants were randomly assigned to three treatment groups with the following goals. First, a sub-sample is offered financial incentives to reduce speed excess; to control for income effects, the remaining part of the sample is offered a flat payment regardless of the driving behavior. Second, to study the impact of financial incentives that can be used to counter peer pressure, for a subset of respondents, the financial incentive offer is publicly disclosed to the peers working at the same taxi station.

Data on participants' behavior is collected in two ways: a) a brief phone survey is administered right before, during, and after exposure to treatment; b) a free-of-charge GPS tracker was installed in the motorbikes of all participants approximately two months before the intervention.

Not available

Randomization done in office by computer

Individual (respondent)

Yes

360 respondents (cluster)

3600 observations (respondent-day)

550 observations (55 clusters) in arm "Flat payment"
1800 observations (180 clusters) in arm "Private contract"
1250 observations (125 clusters) in arm "Public excuse contract"

I collected GPS data for 40 bikes for a period of about 6 months. I used this data to simulate treatment effects and calculate the necessary sample size for each treatment arm. I randomly extracted 360 strikes over 10 consecutive days to construct a simulated panel of 360 drivers receiving the treatment for 10 days. I focus on the speeding violation dummy as an outcome. Within-subject correlation ranges between 0.26 and 0.41 across strikes. At baseline, the mean and standard deviation are (.85,.32) for the speeding indicator. Controlling for outcome values pre-treatment, I can detect a minimum treatment effect of .1 standard deviation with .8 power with the following sample size: 520 obs in control (52 clusters), 1750 obs in the private contract arm (175 clusters), and 1210 obs (121 clusters) in the public excuse contract arm. The private contract arm serves to test both the effect of the private contract compared to flat payment and to test the differential effect of private vs. public incentives.