Measuring the demand for public transport in Lagos, Nigeria

How to cite

Tsivanidis, Nick; Duhaut, Alice; Björkegren, Daniel; Nagpal, Geetika. 2022. "Measuring the demand for public transport in Lagos, Nigeria." AEA RCT Registry. AEARCTR-0010283. https://doi.org/10.1257/rct.10283-1.4

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Nick Tsivanidis* , Alice Duhaut , Daniel Björkegren , Geetika Nagpal

* Primary investigator

Registry ID
AEARCTR-0010283
DOI
https://doi.org/10.1257/rct.10283-1.4
First registered
Last updated
Status
Completed
Trial dates
2023-05-29 – 2023-07-31
Countries
Nigeria (Lagos)
Registry record
http://www.socialscienceregistry.org/trials/10283

Abstract

This study will measure demand for public transport in Lagos.

The first study consists of a field experiment to measure participants value of wait time. The experiment hinges around an SMS-based app (playable on all cellphones) in which participants arrive at a bus stop and are offered a payment amount to wait for a number of minutes before boarding their bus. Enumerators stationed at the bus stops have a tablet which displays a secret code that changes each minute; participants text these codes to our shortwhen they arrive to register their arrival time and receive their offer. If they accept the offer, they send the code displayed after the specified amount of time has elapsed to verify they waited. We randomize the offers of payments and waits; participants' decisions to accept or reject the offers reveal their value of wait time.

The second study consists of a field experiment that measures the price elasticity of demand for transport services. In collaboration with the transit regulator in Lagos, we have developed a way in which price subsidies can be administered via the swipe cards used to the pay for the formal bus system in Lagos.

Intervention

This study will measure demand for public transport in Lagos.

The first study consists of a field experiment to measure participants value of wait time. The experiment hinges around an SMS-based app (playable on all cellphones) in which participants arrive at a bus stop and are offered a payment amount to wait for a number of minutes before boarding their bus. Enumerators stationed at the bus stops have a tablet which displays a secret code that changes each minute; participants text these codes to our shortwhen they arrive to register their arrival time and receive their offer. If they accept the offer, they send the code displayed after the specified amount of time has elapsed to verify they waited. We randomize the offers of payments and waits; participants' decisions to accept or reject the offers reveal their value of wait time. We have two samples, a larger sample of 200 individuals who undergo the field experiment for two weeks and smaller sample of 100 who do it for 5 weeks. The latter allows us to estimate the distribution of individual-level parameters, while the former is sufficient for estimates of population-level parameters.

The second study consists of a field experiment that measures the price elasticity of demand for transport services. In collaboration with the transit regulator in Lagos, we have developed a way in which price subsidies can be administered via the swipe cards used to the pay for the formal bus system in Lagos. Subsidies will be administered for a period of 12 weeks to gauge medium-run behavioral responses to cheaper ticket prices. Participants in the control group will receive a smart card but without subsidized fares. They will receive compensation at the end of the study if they use the card during the study.

Experimental design

Already covered in the abstract and description of intervention.

Design details

Randomization method
The wait time randomization is done by the Python script that runs the backend, randomly selecting the particular offer to be given from a pre-specified distribution. The price subsidy randomization is done in survey CTO at the time of recruitment. Recruited participants are assigned an id by the enumerator based on the order in which they were recruited, and a database loaded into CTO randomly assigns each id to treatment or control.
Randomization unit
Individuals.
Clusters
Wait Time Study: 300 individuals. 200 of these will be participants in a shorter experiment that lasts for 10 days. 100 of these will participate in a longer version which lasts for 25-30 days. This will allow us to measure individual-level parameters, and their distribution, for model extensions that allow for heterogeneity in parameters across individuals. Price Subsidy Study: 800 individuals. We will recruit from 25 neighborhoods for the wait time study and 40 for the price study, but the treatment is at the individual-level. Our power calculations account for this clustering.
Observations
Wait Time Study: 300 individuals. Price Subsidy Study: 800 individuals.
Arms
Wait Time Study: Everyone is being treated. Price Subsidy Study: 400 Treatment, 400 Control. 20 clusters in each treatment arm and a minimum of 20 individuals in each cluster (i.e. 800 individuals)
Primary outcomes
For the Wait Time RCT, the outcome is simply whether they accept or reject the offer. The aim is to structural estimate the disutility of waiting relative to the marginal utility of income. The experiment allows us to structurally estimate these parameters via MLE. We will also estimate the parameters across low- and high-educated individuals, as well as an alternative model of demand which allows the value of time to depend linearly on income. For the Price Subsidy RCT, the primary outcomes are (i) probability of taking a bus trip, (ii) number of trips taken. We will also look at types of trips taken, and heterogeneity by gender and income. We will measure outcomes both from primary surveys, as well as metadata from the smartcards.

Cited by

No citation data available for this registration yet.

Keywords