CrossBoundary Mini-Grid Innovation Lab · January 2023
Which appliances actually grow mini-grid demand?
A multi-year field experiment across 25 mini-grid sites in Kenya, Nigeria and Tanzania testing whether appliance financing raises electricity consumption. Productive-use appliances did; household appliances didn't.
- consumption per user with productive-use financing
- +48%
- consumption per user with productive-use financing
- consumption of top-20% productive users vs peers
- 16×
- consumption of top-20% productive users vs peers
- connections across 25 sites
- 12,000+
- connections across 25 sites
Mini-grids live or die on demand. Their costs are largely fixed, so the more electricity each customer uses, the better the economics. A popular idea is to finance appliances for customers, on the theory that more appliances means more consumption. But which appliances?
The CrossBoundary Mini-Grid Innovation Lab tested this over three phases, working with 10+ developers across 25 sites and more than 12,000 connections in Kenya, Nigeria and Tanzania. I worked on the study as Associate Principal for Data.
- Phase 1.0 financed household appliances such as TVs, fridges and speakers.
- Phase 2.0 financed productive-use appliances such as grain mills and welders.
- Phase 3.0 focused on energy-efficient productive-use appliances.
The result
Average consumption per user (ACPU) moved in opposite directions depending on what was financed.
Source: CrossBoundary Mini-Grid Innovation Lab, Appliance Financing 1.0–3.0
Financing household appliances did not raise consumption; it fell slightly. Sites offering productive-use appliance financing saw ACPU 48% higher. The gains were also highly concentrated: the top fifth of productive-use customers consumed 16 times more than their peers.
What it showed
The lesson for developers is to target financing at income-generating uses of power, because those are what grow demand. It also matters who you finance, since a small group of productive users drives most of the uplift. The approach is now being scaled through a consortium-led roadshow, starting in Ethiopia.
The analytical work
Field experiments on live infrastructure are messy. Sites differ, customers self-select into financing, and seasonality can swamp small effects. Separating the effect of financing from everything else called for regression, propensity-score matching and random-forest models on customer consumption data.