Speeding as a Signal to Peers in the Ugandan Transit Industry: A Survey Experiment

Last registered on January 03, 2023


Trial Information

General Information

Speeding as a Signal to Peers in the Ugandan Transit Industry: A Survey Experiment
Initial registration date
December 21, 2022

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
January 03, 2023, 5:02 PM EST

First published corresponds to when the trial was first made public on the Registry after being reviewed.



Primary Investigator

University of Zurich Econ Dep

Other Primary Investigator(s)

Additional Trial Information

In development
Start date
End date
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
With 1.35 million road fatalities worldwide every year, of which 90% occur in developing countries, the epidemic of road safety has become an absolute priority for economic development. The World Bank defined “tackling the road safety crisis as both a moral imperative and an economic necessity”. Traditional economics suggests that speeding behaviors, a highly consequential risk-taking choice, trade off an increase in profits for serving more customers with higher fuel, maintenance and (potentially) healthcare costs. Yet, we have limited empirical evidence about why suppliers of transportation in low-income settings systematically adopt risky behaviors.

This project considers beliefs about driving behaviors among motor-taxi drivers, the most common supplier of transportation in the metropolitan area of Kampala, Uganda. Drivers operate in semiformal clusters. As a result, peer cultures related to the driving behavior may find fertile ground and perhaps affect productivity.
In this survey experiment, I test whether motor-taxi drivers perceive speeding behavior as a signal for earnings, ability, and social status among peers. To get to the causal impact of speeding on perceived social and monetary returns, I create triplets of identical profiles based on real drivers, except for randomly-manipulated speeding behavior. I ask motor-taxi drivers to rate a random subset of profiles of other motor-taxi drivers along monetary and social status characteristics, as well as their willingness to refer the drivers to their work organization.
This setting allows testing whether speeding increases the likelihood of being perceived as a high earner and as an individual with high social status within the working station.

External Link(s)

Registration Citation

Raisaro, Claude. 2023. "Speeding as a Signal to Peers in the Ugandan Transit Industry: A Survey Experiment." AEA RCT Registry. January 03. https://doi.org/10.1257/rct.10625-1.0
Experimental Details


Intervention Start Date
Intervention End Date

Primary Outcomes

Primary Outcomes (end points)
First order beliefs about:
- earnings and costs;
- driving abilities;
- capacity to influence decisions at the workplace;
- measure of "social coolness";
- Willingness to make a referral
Primary Outcomes (explanation)
Average daily profits:
- What is your best guess for his average daily earnings in the last 4 weeks of work?
- What is your best guess for his motorbike cost in the last 4 weeks of work? Fuel, service, reparation costs

Driving ability:
All respondents are reminded of the ability exercise, consisting of an S-shaped path with 4 cones. They are also informed that the pilot sample whose profiles are used in the survey went through this task, as they did.
- Out of 4 cones, how many cones did the driver touch in executing the task?

All respondents are asked to rate the following social traits: capacity to influence decisions at the station, coolness.
- We ask drivers in the pilot to promote one of their contact to join the group. Please give your best guess about whether the contact joined the group.
- How would you rate how cool he is as a person? Please respond on a scale from 1 to 4, where 1 is not at all and 4 is very.

Make a Referral:
All respondents are asked if they are willing to make an anonymous referral to their own taxi station. The answer to this question is incentivized.
- Would you like to refer ANONYMOUSLY this profile to your stage? If so, we will share the contact with your stage manager?

Secondary Outcomes

Secondary Outcomes (end points)
Secondary Outcomes (explanation)
Beliefs on others’ beliefs:
- On average, how did other drivers rate the profiles ABC?
ABC variables: average daily profits, driving ability, status. This set of outcomes are asked only to a for one random profile proposed out of four.

Experimental Design

Experimental Design
This is a survey experiment. Respondents are asked to answer a set of questions about a series of 4 randomly chosen profiles of other drivers.

Each profile contains the following information: age, years of working experiences as a motor-taxi driver, location of current work station, days with recorded speeding violations within 7 days. Speeding violation is defined as driving faster than 50kmh in the urban area of Kampala. A day is defined as a speeding day when at least a speeding violation is recorded. Speed is recorded by the GPS technology also installed in the respondents’ motorbikes.
Experimental Design Details
In the listing exercise of 1400 drivers, I ask respondent to rate whether a driver is perceived as slow, medium speed or fast based on the number of days out of 7 working days in which they commit at least one speeding violation. Speeding violation is defined as passing the speed limits of 50kmh in urban areas and 90kmh in main national roads in the metropolitan area. I classify as "slow" when days with speeding are 1,2 or 3; medium if 4,5; fast if 6 or 7. This classification is based on the distribution of perceived speeding behavior according to the drivers in the list.

Each respondent is shown 4 profiles, randomly selected from a list of 45. The list is the result of a random manipulation of speeding behavior of 15 real drivers. For each driver, I create 3 types: a slow, medium and fast version of it where one is the real profile and two are manipulated versions. The classification of speeding types follows the categories reported above. Each profile shown to the respondent includes information on age, years of work experience as drivers, location of the current work station and speeding behavior. Respondent receives extensive explanations about the technology used to trace speeding behavior, which is also been offered to them in the frame of the same project. Respondents have also already gained experience with the technology prior to being exposed to this survey experiment: the goal is to establish trust in the technology and make the information shown to the driver to be perceived as reliable.
The 15 real profiles have been exposed the following during the pilot phase: ability S-shaped test, task to get a new driver accepted in his taxi station.

I exploit random variation within profile across respondent to estimate the causal effect of speeding behavior on the perceived social and monetary returns. The order the profiles are show to each respondent is also random. In the regression analysis, I control for ordering, profile fixed effects and respondent fixed effects.

Heterogeneity of interest based on respondents’ characteristics: experience, earnings, social connectedness with peers (measured as number of friends among coworkers in the same taxis station), own speeding behavior. I also plan to test if the share of drivers perceived as fast drivers in the taxi station prediction the experiment’s outcomes.

The set of profiles is constructed as follows. I use 15 real drivers from the pilot of this project. Each driver has been interviewed and offered and GPS tracker that allows them to monitor their driving behavior and locate their bike position from the day before. The same technology is also offered to the subjects of this project. I classified their driving behavior of the 4 weeks of works before the beginning of this project and classify each driver's as slow, medium or fast according to rating distribution from the listing exercise. For each of the driver, I then created two manipulated versions corresponding to the two missing speeding types. Conditional on the type, the number of days with speeding violations is assigned randomly.

Respondents are not informed about which of the profiles they are shown are real. As I have information about the real profiles, I incentivize answers for using them. Beliefs on others beliefs are also incentivized using the answers from pilot data.
Randomization Method
In office by a computer
Randomization Unit
Across subject and within profiles randomization
Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters
360 subjects (clusters). Respondent's eligibility criteria: 18 years old or above; belonging to a taxi station; current primary work motor-taxi driver; working as a driver a minimum of 4 days a week in the last month. Sample is stratified by age groups, experience, access to smartphones in relation to the listing exercise of 1400 drivers performed in November 2021.
Sample size: planned number of observations
1440 profiles: 4 profiles rated by each subject.
Sample size (or number of clusters) by treatment arms
Each respondent is assigned to view either the slow, medium or fast version of a given profile. Each respondent sees at least one profile for each type of speed classification, in random order. 1/3 of the sample is assigned to slow, 1/3 to medium and 1/3 to fast driving types.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Exploratory analysis during the listing exercise suggests that fast drivers are perceive as more capable to influence decision by 0.25 standard deviation compared to medium speed drivers. They are also perceived 0.15 standard deviation less profitable. Controlling for profiles fixed affect and accounting for intra-cluster correlation of 0.5, power calculations point at a sample of 1200 observations to detect a 0.2 standard deviation effect.

Institutional Review Boards (IRBs)

IRB Name
MUREC Research Ethics Committee
IRB Approval Date
IRB Approval Number


Post Trial Information

Study Withdrawal

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information


Is the intervention completed?
Data Collection Complete
Data Publication

Data Publication

Is public data available?

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials