The study is designed to inform the Government of Ghana’s energy policy efforts, which are currently focused on improving the country’s Liquified Petroleum Gas (LPG) distribution system. Ghana is transitioning to a cylinder recirculation model, and many aspects of the eventual market evolution and regulation remain undecided. The core idea for GasPay is to harness ubiquitous cell phone and mobile money usage in Ghana to streamline and incentivize clean household fuel purchases. GasPay will create a simple, direct mechanism for consumers to purchase fuel from LPG marketing companies.

This pilot is intended to fine‐tune and evaluate the feasibility of different components of the digital platform and financing, as well as marketing approaches that best encourage take up by low income users, to assess equity in access. Relevant features that the research team hypothesizes will increase LPG adoption and usage include:

1. A dedicated LPG savings wallet augmented with savings encouragement mechanisms (e.g., cash rewards for consistent contributions to the wallet).
2. A credit mechanism whereby consumers can borrow small sums if they run out of fuel but lack the money needed to purchase LPG. Optimization of the digital credit product design will also be a main focus of the proposed pilot.
3. The ability to request home deliveries of LPG via GasPay for a small fee (as opposed to cylinder exchange at an LPG depot).
4. Group discounts that will be available to groups of 3 or more households that coordinate simultaneous deliveries.

First, the research team will work with the partner to deploy different versions of GasPay, with variation in (i) the price of home delivery, (ii) scheduling ahead discount, (iii) savings incentives, (iv) minimum balance on the savings – allowing for negative balance as credit along with credit terms, (v) group discounts for delivery costs, and (vi) marketing approaches to bringing potential users into the system.

During the pilot, the team will use Zeepay’s administrative data to evaluate outcomes including take up of the application, scheduling and purchasing of LPG, and use of application features such as credit. Because users may substitute between LPG purchased with and without GasPay, the team will supplement administrative data with surveys to measure LPG use and disadoption of traditional fuels.

The pilot will focus on a sample of 1000 households in greater Accra, including existing Zeepay customers (n = 500) and new enrollees from an area‐based random sample in the vicinity of the initial sample (n = 500). The pilot will use Zeepay administrative data from the full sample of customers targeted for take up of the application. For a subset of n = 400 (including both existing and prospective Zeepay customers), researchers will conduct two household visits. During the first, provided marketing scripts will vary both how the application is described and which features are offered, and a baseline survey will measure basic demographic information and fuel use. This will allow researchers to test the effect of features and marketing on take up. Conditional on take up, customers will be randomly offered additional features, which will further allow measurement of effects on use. A second household survey will be conducted roughly 6 months after the first to measure outcomes not observed in the administrative data. For the sample of existing customers not covered by the household visits, application features will be randomized across customers, and the team will rely exclusively on administrative data for outcomes.

The team will measure three main types of outcomes during the pilot: (i) Take up of a payment platform along with a digital credit and savings account that will be provided on participants’ phones; (ii) LPG use, measured through transactions conducted on the payment platform and through surveys conducted six months after the intervention; and (iii) Household level outcomes including use of traditional fuels for cooking, attitudes toward LPG, health, and understanding of the payment platform features.