Scaling digital payments: A market-level randomized controlled trial on cross-side referrals

Note that all interventions are implemented at the market-level.

The main intervention provides incentives for merchants and customers in a market to adopt digital retail payments and for merchants to communicate with their customers about digital retail payments.

The low-saturation main intervention is identical but engages fewer merchants.

The adoption incentives only intervention includes only the adoption incentives from the main intervention.

The referral incentives only intervention includes only the referral incentives from the main intervention.

2024-03-01

2024-05-01

Adoption of digital retail payments; Regular user of digital retail payments; Volume of digital retail payments

A merchant will be considered to have adopted digital retail payments if they accepted a digital retail payment at least once after intervention launch.

A customer will be considered to have adopted digital retail payments if they made a digital retail payment at least once after intervention launch.

A merchant will be considered to be a regular user if they accepted digital retail payments >=4 times per month.

A customer will be considered to be a regular user if they made digital retail payments >=1.5 times per month.

Volume defined as number of digital retail payments accepted/made over a given period.

Main secondary outcomes: Frequency of cash out; Usage of other digital financial services; Willingness to accept mobile money
Mechanisms: Conversations about digital retail payments; Awareness of digital retail payments; Perceptions of digital retail payments
Auxiliary outcomes: Crime; Government revenue; Merchant formalization; Merchant profits; Frequency of remote payments; Multi-homing
Alternative outcome measures: Value of digital retail payments

Cash out will be measured based on volume and value of cash withdrawals from mobile money agents in the market in time since intervention.

A merchant will be considered to be multi-homing if they accepted digital retail payments using two different products since intervention launch.

A customer will be considered to be multi-homing if they made digital retail payments using two different products since intervention launch.

Usage of other digital financial services will be captured by an index. The index will include usage of mobile money, digital loans, digital savings, and other relevant digital financial services.

Perceptions of digital retail payments will be captured by an index. The index will include overall ratings of digital retail payments, and ratings on specific sub-categories such as cost, convenience, security, privacy, speed, and other relevant sub-categories.

Willingness to accept mobile money will be elicited through the Becker-DeGroot-Mashak mechanism.

Crime will be captured by whether the respondent was robbed since intervention launch.

Taxation will be captured by the amount of transaction fees and tax payments made by the respondent since intervention launch.

Merchant formalization will be captured by an index. The index will include whether the firm has a tax identification number, whether the firm is registered with the government, and other relevant measures of formalization.

Merchant profits will be captured by an estimate of the merchant’s revenue subtracting out their costs since launch of the intervention.

All indices will be constructed in the style of “Multiple Inference and Gender Differences in the Effects of Early Intervention: A Reevaluation of the Abecedarian, Perry Preschool, and Early Training Projects” (Anderson, 2008). If there are substantial amounts of missing values in an index, then item response theory may be used instead.

A baseline census and survey was conducted across 300 markets between March and May 2023. We sampled 4 merchants per market and 6 customers per market for the baseline survey.

The 300 markets in our sample will be randomized into 5 groups:

60 control markets
60 main intervention markets
60 low-saturation main intervention markets
60 adoption incentives only markets
60 referral incentives only markets

We plan to conduct the endline survey 30 days after the end of the intervention. We plan to use administrative data from one year prior to the intervention to one year after the intervention.

Not available

Randomization by computer

Market

Yes

300 markets

For market-level outcomes: 300 (1 per cluster) For merchant-only outcomes based on endline survey responses: 1200 (4 per cluster) For customer-only outcomes based on endline survey responses: 1800 (6 per cluster) For merchant and customer outcomes based on endline survey responses: 3000 (10 per cluster) For administrative data outcomes relevant for target merchants: 3600 (12 per cluster) For administrative data outcomes relevant for all merchants in the merchant census: Approximately 12000 (40 per cluster)

60 markets per treatment arm

Assuming the baseline survey standard deviations and intra-cluster correlations for adoption of digital retail payments match the endline values for these parameters, we would be able to identify a minimum detectable effect size of 16.4 pps in market-level adoption, 8.9 pps in surveyed merchant-level adoption, 7.6 pps in surveyed consumer-level adoption, 6.4 pps in surveyed respondent-level adoption, 6.1 pps for target merchant-level adoption, and 4.7 pps for enumerated merchant-level adoption with .8 power at the 95% significance level.