CCLYX/DigitalInfrastructure-and-TaxCompliance

Digital Infrastructure and Tax Compliance: Evidence from the Mobile Finance Expansion in Nigeria # Impact Evaluation of Digital Infrastructure on Tax Compliance ## Executive Summary This project utilizes **Causal Inference techniques (PSM-DID)** to evaluate whether the expansion of mobile financial infrastructure (a proxy for digitalization) effectively drives corporate tax compliance. Using a quasi-natural experiment in Nigeria (2009-2014), the analysis reveals a **Null Effect** (p=0.970). The findings challenge the techno-optimist assumption that "digital tools automatically lead to compliance," suggesting that without enforcement mechanisms, infrastructure alone provides zero ROI for tax authorities. ## Key Data Science Challenges & Solutions Evaluating policy impact in real-world settings is difficult due to **Selection Bias**: regions with high mobile growth are often economically different from low-growth regions initially. | Challenge | Solution Implemented | | :--- | :--- | | **Non-Random Assignment** |Constructed a quasi-natural experiment using uneven infrastructure rollout. | | **Selection Bias** |Applied **Propensity Score Matching (PSM)** (Nearest Neighbor) to create a balanced synthetic control group. | | **Confounding Trends** |Combined PSM with **Difference-in-Differences (DID)** to isolate the treatment effect from time-invariant factors. | | **Robustness** |Validated results using **Placebo Tests** (on Managerial Experience) and Kernel Matching. | ## Methodology ### 1. Data Engineering * **Sources**: Merged firm-level panel data (World Bank Enterprise Surveys) with state-level infrastructure data (EFInA). * **Feature Engineering**: * Constructed a "Treatment" variable based on the top 30th percentile of mobile growth. * Log-transformed skewed features (Sales, Employees) to normalize distribution. ### 2. Causal Model (PSM-DID) The model estimates the Average Treatment Effect on the Treated (ATT) using the following specification: $$TaxCompliance_{it} = \beta_0 + \beta_1 Treat_i + \beta_2 Post_t + \delta (Treat_i \times Post_t) + \Gamm …