Pronóstico de inflación en Costa Rica: el valor de variables exógenas en modelos desagregados por artículo

Abstract

This study proposes a methodology for forecasting inflation in Costa Rica based on a disaggregated analysis of the 289 items that comprise the Consumer Price Index (CPI). ARIMA models are used for most items, while ARIMAX models —which incorporate exogenous variables— are applied to those with higher weights and more volatile prices. The inclusion of specific information, such as the exchange rate, international commodity prices, and weekly agricultural prices, significantly improves forecast accuracy over short-term horizons. The disaggregated approach consistently outperforms more aggregated models or those without exogenous variables by reducing errors in sensitive items such as food, fuel, regulated goods, and products priced in U.S. dollars. The results highlight the value of integrating additional information into forecasting strategies based on disaggregated data and suggest that this methodology can effectively complement the short-term inflation forecasting models currently used by the Central Bank.