Assessing the Developmental Impact of Private DFIs on SME Performance: Evidence from Matched Firm-Level Data in Kenya

Research Hypothesis, Dataset Description, and Interpretation Research Hypothesis The dataset was constructed to test the hypothesis that: H1: SMEs receiving financing from Private Development Finance Institutions (PDFIs) experience higher employment growth, stronger revenue performance, and higher survival probability compared with similar firms that do not receive PDFI financing. This hypothesis is grounded in Financial Market Imperfection Theory, which argues that SMEs face credit rationing due to information asymmetries and limited collateral, and Development Finance Theory, which posits that mission-driven financiers can alleviate these constraints through catalytic capital and technical support. What the Data Show The dataset contains 657 firm-level observations, of which 268 are PDFI-financed firms and 389 are matched non-PDFI firms, reflecting realistic Kenyan SME characteristics. Key variables include: • Employment Growth (percentage change) • Revenue Growth (percentage change) • Survival Probability (0–1) • Matching weights (CEM_Weight and PSM_Weight) Preliminary analysis generated from this dataset indicates that: • PDFI-financed firms exhibit higher average employment growth (≈12.4%) vs. control firms (≈5.8%). • PDFI-backed firms show stronger revenue growth (≈18.7% vs. 9.5%). • PDFI-financed SMEs demonstrate a higher probability of survival (0.89 vs. 0.78). These trends support the hypothesis that catalytic financing improves SME performance. Interpretation and Usage The dataset is structured to enable replication of: • Quasi-experimental causal inference methods (Coarsened Exact Matching—CEM; Propensity Score Matching—PSM) • Weighted regressions for employment, revenue, and survival outcomes • Robustness checks across multiple matching estimators Users can interpret the variables as representing realistic SME performance metrics commonly found in African enterprise datasets. The matching weights allow researchers to directly apply causal inference models without reconstructing the matching process. The dataset is ideal for: • Teaching econometrics • Demonstrating causal inference methods • Simulating development finance impacts • Replicating tables in the associated manuscript (Tables 4–7) Because the dataset is hypothetical but empirically structured, users should not treat results as representative of actual Kenyan firms; instead, it should be used for methodological or training purposes.