Expanding Financial Protection through Microfinance: Evidence from a Randomized Evaluation of Bundled Microfinance and Health Insurance in Nigeria

Intervention
This study employs a randomised encouragement design in which randomisation applies to the offer of the bundled product rather than to actual enrolment. The unit of randomisation is the beat-week – the 24 operational units of the MFB (each assigned to a dedicated credit and recovery officer) crossed with the 16 weeks of the recruitment period, yielding approximately 384 beat-week clusters. For each beat, weeks during the recruitment period will be randomly assigned to either encouragement (treatment) or no-encouragement (control) conditions. During encouragement weeks, all eligible clients visiting the beats will be offered the bundled product with full product information. During control weeks, no offer will be made, and clients receive the status quo product only. Clients remain free to enrol or not; only those visiting during randomly assigned encouragement weeks can access the bundled product. Randomisation will be stratified by branches of the MFB (4 branches) to ensure balance across study arms.

2026-08-20

2026-12-20

Primary outcomes: 1) insurance take-up (2) out-of-pocket health expenditure

1) insurance take-up measured as a binary indicator equal to one if the client has an active EDOHIC enrolment record arising from the study's encouragement period; and (2) out-of-pocket health expenditure measured as the total household spending on health care in the 30 days prior to the survey, net of any amounts reimbursed by insurance, from the endline survey.

Secondary outcomes are (1) catastrophic health spending (2) health-seeking behaviour, (3) financial health score (4) business profit, and (5) Loan default

Secondary outcomes are (1) catastrophic health spending, measured as a binary indicator equal to one if out-of-pocket health expenditure exceeds 10 per cent of total household expenditure in the same 30-day reference period; (2) health-seeking behaviour, defined as the type of care sought at the most recent health event, timeliness of care-seeking, and utilisation of the EDOHIC-assigned contracted facility among enrolled clients; (3) Financial health score, defined as a composite index adapted from the EFInA Financial Health Framework; (4) Business profit measured as the net monthly business income defined as total revenue minus total costs in the 30 days prior to the survey, self-reported at the endline; (5) Loan default measured as a binary indicator equal to one if the client's loan is classified as non-performing in Zeemat's internal loan management system
The financial health score is a composite measure captures day-to-day financial management, resilience to shocks, and capacity to pursue financial opportunities.

This study implements a randomised encouragement design to evaluate the causal impact of bundling health insurance with microfinance loans on the financial outcomes of informal micro- and small-business owners in Edo State, Nigeria. The study is conducted in partnership with a microfinance bank (MFB) operating 24 beats across 4 branches and the Edo State Health Insurance Commission.
The unit of randomisation is the beat-week. The 24 operational beats of the MFB are each assigned to encouragement (treatment) or no-encouragement (control) conditions on a week-by-week basis across 16 weeks of recruitment, yielding approximately 384 beat-week clusters. During encouragement weeks, all eligible clients visiting a beat are offered a bundled product comprising a microfinance loan and voluntary health insurance coverage (Bronze plan, ₦18,000 annual premium). During control weeks, clients receive the microfinance loan only. Randomisation is stratified by branch. Clients are never compelled to enrol; take-up is entirely voluntary.
Eligible participants are active loan clients and clients presenting for loan renewal who operate informal micro or small businesses. Two rounds of data collection – baseline and endline – will be conducted via structured surveys. The endline data is collected across two waves to enable short-run and long-run analysis of effects. Wave 1 is administered approximately two months after the end of the recruitment period and focuses on outcomes expected to respond quickly to insurance access, specifically insurance take-up, out-of-pocket health expenditure, and catastrophic health spending. Wave 2 is administered approximately four to five months after the end of the recruitment period and captures outcomes that require sustained exposure to health insurance coverage exposure, including health-seeking behaviour, the financial health score, and business profit. The study will estimate the Intent-to-Treat (ITT) effect of the offer and the Local Average Treatment Effect (LATE) among compliers using Two-Stage Least Squares with the randomised offer as an instrumental variable for actual enrolment.

Not available

Randomisation will be conducted using a computer-generated random number procedure in Stata. For each of the 24 operational beats, the 16 recruitment weeks will be randomly assigned to either encouragement (treatment) or no-encouragement (control) conditions using a block randomisation procedure, stratified by branch (4 branches). Within each branch stratum, weeks will be assigned with equal probability to treatment and control, ensuring balance across arms within each branch. The assignment will be implemented by the research team and communicated to MFB branch coordinators on a weekly basis to minimise anticipation effects among loan officers and clients.

The unit of randomisation is the beat-week (the intersection of one of the 24 operational beats of the MFB and one of the 16 weeks of the recruitment period), yielding approximately 384 beat-week clusters. A beat is the smallest operational unit of the MFI, each managed by a dedicated Credit and Recovery Officer (CRO) responsible for a defined client portfolio within a branch. Weeks within each beat are randomly assigned to encouragement or no-encouragement conditions; all eligible clients visiting a beat during an encouragement week are offered the bundled product, while no offer is made during control weeks.
The unit of observation and analysis is the individual loan client.

Yes

384 beat-weeks (24 beats × 16 recruitment weeks)

Approximately 1152 to 1536 individual loan clients (384 beat-week clusters × 3 to 4 clients per cluster on average), based on MFB administrative data showing an average of 329 combined new and renewal loans per month across 24 beats. The exact number of observations will depend on realised client turnout during encouragement and control weeks over the 16-week recruitment period.

192 beat-weeks assigned to encouragement (treatment) - approximately 576 to 768 individual loan clients
192 beat-weeks assigned to no-encouragement (control) - approximately 576 to 768 individual loan clients
Total: 384 beat-weeks, approximately 1152 to 1536 individual loan clients across both arms

Minimum detectable effect sizes are calculated assuming J = 384 beat-week clusters (24 beats × 16 weeks), n = 3-4 clients per cluster, ICC (ρ) = 0.29, R² = 0.10, α = 0.05 (two-tailed), and 80% power. The design effect ranges from 1.64 (n = 3) to 1.56 (n = 4). Primary outcome - Financial health score (continuous): MDE ≈ 0.08 standard deviations (Cohen's d). Using an illustrative baseline standard deviation of 17 points on a 0-100 financial health scale – consistent with EFInA's financial health measurement framework for Nigerian informal sector workers – this corresponds to a detectable difference of approximately 1.4 points on the financial health index. The exact MDE in index points will be updated following baseline data collection once the observed standard deviation is available. Secondary outcome - Loan default (binary): Assuming a baseline default rate of 6% per MFB administrative records, the MDE is approximately 2.4–2.5 percentage points, representing a detectable reduction from 6% to approximately 3.5%. This is expressed as a percentage point change in the probability of default. Secondary outcome - Business profit (continuous): MDE ≈ 0.08 standard deviations. The MDE in absolute Naira terms will be calculated after baseline data collection once the standard deviation of monthly business profit is observed. Secondary outcome - Out-of-pocket health expenditure (continuous): MDE ≈ 0.08 standard deviations. The MDE in Naira terms will be determined at baseline.