Impact Assessment of Intelligent Learning Systems on Credit Discipline and Micro-Enterprise Development

Abstract

Timely repayment and small firm performance are central to enterprise development and microfinance sustainability because predictable payments support liquidity and continued credit access. Existing fintech work lacks borrower-level causal evidence and standardized outcome measures, and this protocol confines causal claims to sampled partner microfinance branches. This pre-registered protocol defines a cluster-randomized field experiment that assigns partner microfinance branches and records the randomization script and its documented seed. Primary outcomes are borrower-level on-time repayment over a six-month follow-up and enterprise revenue change over a twelve-month window. The pre-analysis plan fixes the estimands and reports intention-to-treat contrasts using covariate adjustment, cluster-robust variance estimates, and cluster bootstrap inference with fixed seeds. These design and quality checks defined in advance aim to support credible operational decisions by microfinance institutions in partner microfinance branches.