Recalling the Past: Memory and Forecasting in Burundi
Abstract Many smallholder farmers in developing countries rely solely on memory and experience to make decisions about planting and investing. For farmers with limited access to advanced technology and financial services, errors resulting from memory inaccuracy and faulty forecasts can severely impact harvests. Using a recall experiment, survey data, and secondary data from various sources, I examine farmers' recall and its relationship to their predictions about rainfall, agricultural output, and crop prices in Burundi. I find that farmers' recall accuracy of prices tends to deteriorate or remain unchanged over time and is influenced by events unrelated to agriculture. I also establish a link between recalled and forecasted values, providing evidence of long-lasting effects and recency bias. Finally, I estimate farmers' price forecast errors, the portion attributable to recall inaccuracy, and the type of loss function farmers likely use when making forecasts. My findings suggest that policymakers could improve farmers' decision-making by providing timely, accessible information to supplement memory-based forecasts.
