Gendered Differences in Mobility and the Demand for Transport: Evidence from a Field Experiment in Urban Ethiopia

As part of the intervention, participants will receive transport credit in the form of a personalized transport voucher, which they may use over a two-month period to reimburse the cost of any trips that they take with a team of 10 hired taxis. The total value of the voucher that is presented to each participant will be randomized to vary between “low” and “high” amount – 1540 ETB ($30) and 4100 ETB ($80) respectively, such that the average amount received across couples will be 2800 ETB ($55 USD). In addition, the voucher and transport intervention package will be randomized to either be presented to the eligible woman alone (T1), the eligible man alone (T2), or the couple jointly (T3).
If a participant chooses to schedule a trip with a taxi, they will be instructed to call one of the program staff members, whose numbers will be provided on the voucher document. Participants will be asked to confirm their identity and will be able to provide details of their intended trip (the pick-up and drop-off locations as well as time, number of other people who will be accompanying them if applicable). Each taxi driver will be accompanied by an enumerator, who will be responsible for: (1) confirming eligibility of passengers by checking their voucher and ID details; (2) collecting the data on trips’ origin, stops, duration, purpose, accompanying travelers, etc; (3) handling the reimbursement process upon the trip completion with the driver directly; (4) registering the trip cost data on the voucher and electronically; (5) informing participants of the travel credit balance that is remaining following the trip.

2022-07-01

2022-11-30

Out primary outcomes of interest are: (1) intervention uptake; (2) length of an average trip (km); (3) proportion of trips taken with partner; (4) incidence of trip chaining; (5) women empowerment.

Our measures of women's empowerment cover a number of components, including: 1) measures of freedom of mobility and access to services outside the home (markets, health facilities, etc.); 2) self-reported measures of agency and decision-making; 3) self-reported measures of self-efficacy; 4) mental health and well-being; and 5) measures of safety during travel.

Our study is a three-armed randomized controlled trial with an estimated 1,000 households in Addis Ababa, Ethiopia that consists of a baseline survey followed by implementation of the transport intervention over a two-month period. Following a baseline survey, each household will be randomly assigned to one of three treatment arms: woman arm, man arm or couple arm. For households that are randomly assigned to the woman treatment arm, the taxi service will be offered to eligible women for their private use. For households that are randomly assigned to the man treatment arm, the service will be offered to eligible men for their private use. Finally, for households that are randomly assigned to the couples treatment arm, the service will be offered to men and women jointly.
Following randomization, all participants will be visited by the field team to be introduced to the intervention. A follow-up survey will be conducted with all participants after completion of the intervention.

We will randomize households to intervention arms such that intervention assignment is balanced according to the following baseline characteristics: household wealth, employment status, average travel frequency and access to transport modes (distance to public transport stop, bicycle ownership, etc.).

Our unit of randomization is a couple.

No

1,000 couples

1,000 couples

T1: "Woman" treatment arm, where the taxi service will be offered to the eligible women for their private use (N = 400)
T2: "Man" treatment arm, where the taxi service will be offered to the eligible men for their private use (N = 300)
T3: "Couple" treatment arm, where the taxi service will be offered to the eligible couples jointly (N = 300)

Moreover, we randomize the voucher amount to vary between “low” and “high” amount, so that half of respective couples will fall into each category:
T1.30: "Woman low" sub-arm, where transport credit amount equals 1540 ETB ($30) (N = 200)
T1.80: "Woman high" sub-arm, where transport credit amount equals 4100 ETB ($80) (N = 200)
T2.30: "Man low" sub-arm, where transport credit amount equals 1540 ETB ($30) (N = 150)
T2.80: "Man high" sub-arm, where transport credit amount equals 4100 ETB ($80) (N = 150)
T3.30: "Couple low" sub-arm, where transport credit amount equals 1540 ETB ($30) (N = 150)
T3.80: "Couple high" sub-arm, where transport credit amount equals 4100 ETB ($80) (N = 150)

To infer a potential effect size for the intervention uptake, we look for evidence from Franklin (2018) study in Addis Ababa, which had about 66% of treated individuals taking up the intervention, i.e. collecting transport subsidy. Because women are more likely to experience additional constraints such as safety issues, requiring a permission to travel, we expect the Woman arm to end up with lower uptake, i.e. proportion of treated couples using the taxi service at least once. Moreover, given randomization of travel credit, we expect participants receiving “high” amount to be more likely to perform at least one trip. Assuming a 10 percent attrition-adjusted sample of 630 couples (360 couples in Woman arm (T1) and 270 couples in Man arm (T2), from which 180 couples in Woman arm (T1.30) and 135 couples in Man arm (T2.30) receive the “low” amount), we will have 80 percent power to detect 14 percentage points difference in intervention uptake (alpha = 0.05). Data from NHTS (1995) performed in United States suggests that more than 40% of the women in two-adult households with small children chained non-work trips into their commutes compared to 24% of men. Assuming a 10 percent attrition-adjusted sample of 630 couples (360 couples in Woman arm (T1) and 270 couples in Man arm (T2), we will have 80 percent power to detect about 11 percentage points difference in incidence of trip-chaining between two arms (alpha = 0.05). Moreover, based on the evidence of trip chaining, we expect women to perform trips of longer distances compared to men, hence associated outcome in Woman arm (T1) would be higher than in Man arm (T2). Christensen and Osman (2021) study provides estimate of 11-15 km length of a trip. Our study differs from Christensen and Osman (2021) Uber experiment in Cairo, Egypt in that we do not include a pure control group but rather compare extensive and intensive travel outcomes of treatment arms varying gender of the recipient. Assuming a length of 11 km for Man arm and a 10 percent attrition-adjusted sample of 630 couples, we are powered to detect at least 2 km difference between two arms (alpha = 0.05). Our prior is that women are more likely to require partner’s permission to travel: according to EDHS (2016), only 18 percent of married women in Addis Ababa mainly make decision to visit their friends and family on their own. As we vary the main recipient of the taxi services, we might observe a gradual decrease in the incidence of traveling with partner from Women (T1) to Couple (T2) to Men (T3) arm, since women, even being main recipient, might need their partner’s presence to perform the trip. Moreover, we might expect women in “high” arm to travel more often alone, since increased travel credit amount allows for more trip, which, after performing several trips with their partner, will build trust in the service and make women more comfortable in traveling alone. Assuming a 10 percent attrition-adjusted total sample of 630 couples (360 couples in Women arm (T1), from which 180 couples will receive the “low” amount (T1.30) and 180 couples will receive the “high” amount (T1.80)), we will have 80 percent power to detect about 10 percentage points difference in incidence of travel with partner between two sub-arms (alpha = 0.05).