An ARCH-GARCH Approach to Paddy Production Volatility: A Seasonal Analysis of Monaragala District, Sri Lanka

Paddy cultivation forms the backbone of Sri Lanka’s agricultural economy and remains a vital pillar of national food security. However, fluctuations in paddy production, driven by climatic variability and other external factors, pose significant challenges to agricultural planning and policy formulation. This study investigates the volatility of paddy production in the Monaragala District, Sri Lanka, using annual production data for the Maha and Yala cultivation seasons spanning the period from 1960 to 2024. The analysis employs autoregressive (AR) models to estimate the conditional mean, followed by Autoregressive Conditional Heteroskedasticity (ARCH) and Generalized Autoregressive Conditional Heteroskedasticity (GARCH) models to capture and forecast time-varying production volatility. The stationarity of the log-return series was confirmed using the Augmented Dickey–Fuller (ADF) test, while the ARCH–Lagrange Multiplier (ARCH-LM) test verified the presence of conditional heteroskedasticity in both seasonal series. The estimated results indicate that the two cultivation seasons exhibit distinct volatility characteristics. The Maha season is best represented by an AR(1)–ARCH(1) model, indicating short-term volatility persistence, whereas the Yala season is more appropriately modelled using an AR(2)–GARCH(1,1) framework, demonstrating substantial long-term volatility persistence (α + β = 0.8827). These findings suggest that short-term crop insurance schemes may be effective in mitigating temporary climate-related risks during the Maha season, whereas reducing the highly persistent risks associated with the Yala season requires long-term structural interventions, including reservoir modernization and the rehabilitation of village tank cascade systems.