E-hailing and last mile connectivity in Sao Paulo.
Last registered on November 02, 2018


Trial Information
General Information
E-hailing and last mile connectivity in Sao Paulo.
Initial registration date
October 31, 2018
Last updated
November 02, 2018 5:50 PM EDT
Primary Investigator
Fundação Getúlio Vargas
Other Primary Investigator(s)
PI Affiliation
99 App
PI Affiliation
PI Affiliation
World Bank
PI Affiliation
99 App
PI Affiliation
World Bank
Additional Trial Information
On going
Start date
End date
Secondary IDs
This study aims at understanding the price elasticity of demand for e-hailing services for first and last mile connectivity using a random sample of users of one of Brazil’s main e-hailing apps in Sao Paulo. Every day close to 21 million residents of the Sao Paulo Metropolitan Region (SPMR) make around 44 million trips, with the average trip in public transport lasting more than an hour. Mixing new transport technologies such as e-hailing apps with available public transit infrastructure is a possible sustainable solution to congestion in large cities like Sao Paulo. E-hailing apps can provide flexibility and on-demand services that can easily be combined with rapid mass public transportation. While the study focuses on understanding the price at which people are motivated to exchange a car-only trip for a car and metro one, it will also try elucidating how income-levels, gender and age affect these choices; and what complementary factors are part of the decisions (i.e. personal security, for instance). The study will provide unique evidence of the potential of e-hailing for first and last-mile connectivity across one of Latin America’s largest city.
External Link(s)
Registration Citation
Bianchi, Bianca et al. 2018. "E-hailing and last mile connectivity in Sao Paulo.." AEA RCT Registry. November 02. https://www.socialscienceregistry.org/trials/3518/history/36698
Experimental Details
We planned two interventions (2 Treatment intensities):

1) A 25% discount on e-hailing trips that end or start at subway and train stations. Subjects can use this discount twice a day, for one week.
2) A 75% discount on e-hailing trips that end or start at subway and train stations. . Subjects can use this discount twice a day, for one week.

For both treatments a message will be sent stating that the discount is for trips to and from subway and train stations exclusively. While we can expect that the treatment may include trips to reach points of interest around the stations, researchers will be able to distinguish the origin and destination of each trip, and will have information about residential location and point of work/study, allowing to distinguish which users applied the discount for the intended purpose. In this sense we could retrieve not only the Intention-to-Treat but also the Average-Treatment-Effect on the Treated.

The hypothesis behind these two interventions is that e-hailing discounts can potentially influence modal choices. One example is that people could now substitute “car only” trips by alternative trips that mix mass public transportation and e-hailing (a first/last-mile behavior change).
Intervention Start Date
Intervention End Date
Primary Outcomes
Primary Outcomes (end points)
Share of e-hailing trips to subway and train over the total number of e-hailing trips made during the period of treatment.
#/% of e-hailing trips that are first/last mile to metro/train stations for work/school/others.
Price at which e-hailing trips are first/last mile to metro/train station.
Distance of total trips (if trip is combined e-hailing-metro).
Primary Outcomes (explanation)
Additionally, as the e-hailing app can monitor all the trips made by their users, a big concern of the study is to measure whether treated units present behavior change over a longer period of time, after the end of the discounts. Do treated units start having more e-hailing last mile trips?
Secondary Outcomes
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
Gender, age and income/education heterogeneity for primary outcomes; as well as heterogeneity by type of commuting mode (most used mode), purpose of trips (job/school trips); combining with security data by block of origin of a trip we can also measure whether insecurity increases the use of e-hailing for first/last mile.
An additional dimension of interest is the generalized cost of transport. If there is indeed behavioral change, it might as well impact transport costs.
Experimental Design
Experimental Design
Anonymous information of the e-hailing app users’ trips is tracked by the company to understand patterns and customers profiles across the city. This database is anonymous and contains all historical points for each customer.

The experiment will be based on random sample of users for which the researchers would have access to the historical database of regular trips. The final sample will further be built from respondents of a survey (sent randomly trough the app to 60 thousand users ~ about 1.2 million trips in 2018) that would allow to complete users socio-demographic information and regular commuting behaviors. Information such as income, age, gender, level of schooling and, most importantly, regular transport expenses and modal choices will be collected from the surveys. The closest metro station from residential address and place of work/study will also be collected. Respondents were given a financial incentive to participate in the survey. Historical experience suggests close to 5-6 % of the sample will respond to the surveys.

The analysis will be carried out using georeferenced data – i.e. it is possible to know precisely the origin and destination of each user’s trips – combined with survey information, such as distance to workplace from train stations. Keeping up with each subjects’ trips through the app we can infer their behavior during the treatment period and the subsequent periods.
Experimental Design Details
Randomization Method
The treated and control groups will be randomly defined among the e-hailing randomly chosen sample of users (60 thousand) that responded to the survey sent through the app (N=4788). This sample allows to complete users socio-demographic information and regular commuting behaviors. The first randomization was done by the e-hailing app operations team by a computer; the second randomization (T1,T2 and C) will also be done by a computer.
Randomization Unit
Individuals (users of e-hailing app).
Was the treatment clustered?
Experiment Characteristics
Sample size: planned number of clusters
4788 users of e-hailing app.
Sample size: planned number of observations
4788 users of e-hailing app.
Sample size (or number of clusters) by treatment arms
1/3 Treatment 1, 1/3 Treatment 2 and 1/3 Control Group.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Main outcome: share of e-hailing rides that start or end at a metro station (individual for a 2 week period). The MDES was estimated on a pre-baseline sample (the 60 thousand randomly selected users) and later, on the sample of respondents (baseline survey, N=4788) for a targeted power of [0.8, 0.9]. With the first sample (N=60 thousand with 2 treatment groups) the mean was 0.1507 with a standard deviation of 0.25027. The minimum detectable effect was 0.2 SD. For the final sample (N=4788, with 2 treatment groups), the mean is 0.17136 with a standard deviation of 0.24362. The minimum detectable effect is 0.1 SD.
IRB Name
IRB Approval Date
IRB Approval Number
Post Trial Information
Study Withdrawal
Is the intervention completed?
Is data collection complete?
Data Publication
Data Publication
Is public data available?
Program Files
Program Files
Reports and Papers
Preliminary Reports
Relevant Papers