# Ethiopia Financial Inclusion — Forecasting System
## The Problem
Ethiopia is in the middle of a fast digital financial transformation. Telebirr has grown to
over 54 million users since 2021. M-Pesa entered the market in 2023 and already has 10+
million users. Interoperable P2P digital transfers have overtaken ATM cash withdrawals for
the first time. And yet, per the 2024 Global Findex survey, only **49%** of Ethiopian adults
have a financial account — just 3 percentage points higher than in 2021.
Selam Analytics was engaged by a consortium of development finance institutions, mobile money
operators, and the National Bank of Ethiopia to answer three questions:
1. **What drives financial inclusion in Ethiopia?**
2. **How do events** — product launches, policy changes, infrastructure investments — **affect
inclusion outcomes?**
3. **How did inclusion change in 2025, and where is it headed in 2026-2027?**
This repo builds a forecasting system that answers all three, using the World Bank's Global
Findex framework's two core dimensions:
- **Access** — Account Ownership Rate
- **Usage** — Digital Payment Adoption Rate
## Approach
Rather than a flat spreadsheet of numbers, the dataset uses a unified `observation` / `event`
/ `impact_link` / `target` schema, so that policies, product launches, and infrastructure
milestones can be modeled as *drivers* of Access and Usage, not just listed alongside them.
Events deliberately never get a pillar (Access/Usage/etc.) pre-assigned — Telebirr's launch,
for example, affects both, so labeling it as one up front would bias the analysis before it
starts. `impact_link` records carry that interpretation instead, each one joined back to its
parent event via `parent_id`, with a direction, magnitude, lag, and evidence basis.
### Schema at a glance
| record_type | `category` filled? | `pillar` filled? | What it represents |
|---------------|---------------------|-------------------|---------------------|
| `observation` | no …