Informal workers, many of whom now work on gig-economy platforms in developing economies, lack formal contracts, social protection, and income stability. Can incentive design reduce risks and uncertainty in their earnings, while maintaining their preference for flexibility? Do such interventions improve productivity and workers' welfare? We will conduct a randomized controlled trial with drivers on a major ride-hailing platform in Kampala, Uganda, and randomly offer three bonus schemes: a fixed bonus, a conditional income guarantee, and additional flexibility to make-up bonuses. Using trip-level administrative data and survey measures, we will estimate the effects of exogenous bonus assignment on labor supply, earnings, consumption, and welfare. We will also conduct incentivized elicitation on bonus types to identify workers' preferences for risk and flexibility, and identify the welfare effects of matching contracts to workers' preferences. These findings will provide novel insights into incentive design to improve worker welfare in developing economies.
Experimental design
The research team will collaborate with a large ridesharing platform that offers motorcycle taxi (i.e., boda) services in Kampala, Uganda. The platform matches drivers with trips and charges commission fees. The platform also addresses safety and credit issues by mandating helmets and issuing loans for drivers’ motorcycles and telephones.
Rideshare drivers in Kampala face several risks and uncertainties in their earnings. First, as is the case in many other ridesharing contexts, driver earnings vary and depend on the demand conditions in the locations where they work. Second, the uncertainty around demand conditions is exacerbated by traffic and accident risk. Third, they face liquidity shocks and cash needs to meet their household needs; in many cases, boda drivers are connected to rural relatives who rely on them for financial assistance. A previous focus group conducted with a microcredit firm in 2019 also found that boda drivers still face liquidity constraints that can only be partially offset by adjusting supply hours on ridesharing apps.
Basic Methodological Framework
We will conduct a randomized controlled trial (RCT) to understand how drivers respond to financial incentives in the form of conditional bonuses, conditional earnings guarantees, and flexible bonuses. Doing so enables us to isolate how drivers make decisions and are impacted by the income effect, changes in uncertainty about earnings, and preference for flexibility. We will have two phases to the experiment; in the first phase, we will randomly assign drivers to one of three bonus incentive types or a control group. Drivers will receive daily bonus incentives for six weeks following the baseline survey and random assignment. In the second phase, which occurs after the endline survey, drivers bid for contract types and are assigned one of the contracts determined by the BDM method.
1. Labor Supply
Outcome 1A: Daily Active Hours on the Platform – The total number of hours a driver spends with the platform app open per day.
Outcome 1B: Weekly Hours Spent Working outside the Platform – The total weekly hours spent on all income-generating activities except for the ridesharing platform.
Outcome 1C: Daily Transaction Hours on the Platform – The total number of hours a driver spends actively making trips on the platform per day.
Outcome 1D: Weekly Hours Spent Working on and off the Platform – The total weekly hours spent on all income-generating activities, including the ridesharing platform.
2. Earnings
Outcome 2A: Daily Platform Earnings – The total post-commission income earned via platform trips, exclusive of bonuses.
Outcome 2B: Total Weekly Earnings – The total weekly earnings earned from all income-generating activities (survey data).
3. Consumption and Expenditures
Outcome 3A: Total Overall Monthly Expenditures – The total amount of money a household spent on "overall expenditures" in the last month.
Outcome 3B: Total Monthly Expenditures for Own Household Consumption – The total amount of money a household spent specifically on their own consumption (food, clothes, etc.) in the last month.
Outcome 3C: Income Needs – A binary indicator for whether or not a driver could meet their monthly income needs.
4. Preferences over Incentive Types
Outcome 4A: Preference for Reduced Earnings Risk – The bid (in UGX) a driver makes to receive Treatment 2 (earnings guarantee) over Treatment 1 (fixed bonus).
Outcome 4B: Preference for Flexibility – The bid (in UGX) a driver makes to receive Treatment 3 (fixed bonus with flexibility) over Treatment 1 (fixed bonus).
Would you like me to create a table comparing the data sources (Platform Data vs. Survey Data) for each of these outcomes?
Secondary outcomes
5. Labor Supply
We define additional measures of labor supply:
Outcome 5A: Proportion of Time on Platform - The proportion of total weekly hours spent on the platform relative to weekly hours spent on all income-generating activities, measured via survey data.
We will use survey data to collect the total number of self-reported hours a driver spends in a given week on each income-generating activity they do. We will then divide the time they spend on the platform by total hours spent working (sum of hours across all activities) to obtain the proportion of time spent on the platform.
This measure is useful as it can highlight switching behavior across different income-generating activities, highlighting how drivers respond to different types of financial incentives in relative terms.
This outcome will be part of the driver-level survey data set and captured during both baseline and endline.
Outcome 5B: Active Days per Week - The number of days each week a driver makes at least one trip.
This outcome will be part of the driver-week level panel data set.
Measured via app data.
Outcome 5C: Trips per Day - The number of trips completed each day.
This outcome will be part of the driver-day level panel data set.
Measured via app data.
Outcome 5C: Income Targeting Behavior - An indicator for whether or not a driver chooses how much to work by targeting a particular income.
This outcome will be part of the driver-level data set.
Measured via survey data.
Outcome 5D: Income Targets - The level of income a driver reports they target (conditional on income-targeting behavior).
This outcome will be part of the driver-level data set.
Measured via survey data.
6. Trip Changes
Drivers might respond to the intervention by selecting the trips they are willing to accept. We define a few measures to capture on what margins drivers might change the types of trips or the variety of trips they take:
Outcome 6A: Earnings Variance - The variance of post-commission, pre-bonus earnings across all days.
This outcome will be part of the driver-level data set.
Measured via app data.
Outcome 6B: Trip Duration - The number of minutes of each trip.
This outcome will be part of the driver-trip level panel data set.
Measured via app data.
Outcome 6C: Trip Distance - The number of Kilometers of each trip.
This outcome will be part of the driver-trip level panel data set.
Measured via app data.
Outcome 6D: Trips in the Morning - The proportion of trips per day between 7 am and 11 am.
This outcome will be part of the driver-day level panel data set.
Measured via app data.
Outcome 6E: Trips in the Afternoon - The proportion of trips per day between 11 am-4 pm.
This outcome will be part of the driver-day level panel data set.
Measured via app data.
Outcome 6F: Trips in the Evening - The proportion of trips per day between 4 pm and 8 pm.
This outcome will be part of the driver-day level panel data set.
Measured via app data.
Outcome 6G: Trips in the Night - The proportion of trips per day between 8 pm and 12 am.
This outcome will be part of the driver-day level panel data set.
Measured via app data.
Outcome 6H: Trips in the Early Morning - The proportion of trips per day between 12 am and 7 am.
This outcome will be part of the driver-day level panel data set.
Measured via app data.
Outcome 6I: Trips in Kampala City Center - The proportion of trips per day in the business district of Kampala.
This outcome will be part of the driver-day level panel data set.
Measured via app data.
7. Welfare
We define further measures captured in survey data that proxy for several other components of a driver's welfare:
Outcome 7A: Savings - Total savings a driver has this month.
This outcome will be part of the driver-level data set.
Measured via survey data.
Outcome 7B: Remittances - Total amount a driver has remitted in the past month.
This outcome will be part of the driver-level data set.
Measured via survey data.
Outcome 7C: Job Satisfaction - Binary indicator for whether or not the driver ranked their job satisfaction as satisfactory or higher.
This outcome will be part of the driver-level data set.
Measured via survey data.
Outcome 7D: Active Loans - Total amount a driver owes in loans this month.
This outcome will be part of the driver-level data set.
Measured via survey data.
Cited by
No citation data available for this registration yet.