Sending out an SOS: using start of rainy season indicators for market price forecasting to support famine early warning

Abstract

We examine relationships between the start of rainy season (SOS) and sub-national grain (white maize) market price movements in five African countries. Our work is motivated by three factors: (a) some regions are seeing increasing volatility SOS timing; (b) SOS represents the first observable occurrence in the agricultural season and starts a chain reaction of decisions that influence planting, labor allocation, and harvest—all of which can have direct impacts on local food prices and availability; and (c) pre- and post-harvest price movements provide key insights into supply-and-demand issues related to food insecurity. We start by exploring a number of different SOS definitions using varying reference periods to define whether an SOS is ‘on-time’ or ‘late’. We then compare how those different definitions perform in seasonal price forecasting models. Specifically, we examine if SOS indicators can predict price means over 6 and 9 month periods, or roughly the length of time from planting to market. We use different reference periods for defining ‘early’ versus ‘late’ seasonal starts based on the previous year’s start date, or median start dates over the past 3, 5, and 10 year periods. We then compare the out-of-sample forecast performance of univariate time-series models (autoregressive integrated moving average (ARIMA)) with time-series (ARIMAX) models that include various SOS definitions as exogenous predictors. We find that using some form of SOS indicator (either an SOS anomaly or 1st month’s rainfall anomaly) leads to increased predictive power when examining prices over a 6 months window. However, the results vary considerably by country. We find the strongest performance of SOS indicators in central Ethiopia, southern Kenya, and southern Somalia. We find less evidence in support of the use of SOS indicators for price forecasting in Malawi and Mozambique.

Keywords

• food security
• forecasting
• climate change