Shamas245/nigeria-rainfall-replication

Replication of Amare et al. (2018) — Rainfall Shocks and Agricultural Productivity, Nigeria LSMS-ISA Panel Data # Replication: Amare et al. (2018) ## Rainfall Shocks and Agricultural Productivity: ## Implications for Rural Household Consumption --- ## 📄 Paper Reference Amare, M., Mariara, J., Larsens, R., & Passarelli, S. (2018). Rainfall shocks and agricultural productivity: Implication for rural household consumption. *Agricultural Systems*, 166, 79–89. doi.org **Replication by:** Shamas Liaqat **Institution:** University of Agriculture Faisalabad **Date:** March 2026 **Language:** Python 3.12 --- ## 📋 Project Overview This repository contains a partial replication of Amare et al. (2018), which studies how rainfall shocks affect agricultural productivity and household consumption in rural Nigeria using three waves of World Bank LSMS-ISA panel data. ### What I Replicate - Panel dataset construction from 3 survey waves (2010/2012/2015) - Rainfall shock calculation using georeferenced GPS data - Household and time fixed effects regression - Asset-based heterogeneity analysis (poor vs. non-poor) - Regional heterogeneity analysis (North vs. South Nigeria) ### What I Do Not Replicate (and Why) The paper's main specification uses a full IV-FE model where agricultural productivity is the endogenous variable instrumented by rainfall shock. I run the **reduced form** specification (rainfall shock directly on consumption) for the following reason: Wave 1 (2010) and Wave 2 (2012) of the Nigeria GHS survey have severe crop value data limitations — only 1.1% of crop harvest observations have usable monetary values. This is a known limitation of early Nigeria GHS survey design, corrected in Wave 3 (2015) which has 85.6% coverage. Without reliable crop values across all three waves, constructing a consistent panel measure of agricultural productivity is not feasible without strong assumptions. I document this limitation transparently rather than imputing unreliable productivity estimates. --- ## 📊 Main Results | Model | β | SE | Sig | N | |- …