In an attempt to identify price stabilization strategies and rationalize public intervention in buffering markets, this article investigates the intertemporal dynamics of commodity prices in Ethiopia. A classical rational expectation model is modified to account for seasonal correlation of shocks. Model predictions are reduced to computable periodic threshold autoregression. Several nonlinearity tests are applied to detect threshold effects. A regime-switching normalized maximum likelihood method is formulated to estimate thresholds and threshold autoregression parameters using monthly data from Ethiopia for the period 1996–2006. The result indicates the presence of periodic price thresholds that could be formed as a result of speculative storage. Comparison of price movements below and above thresholds indicates that prices are more correlated below the thresholds than above them. However, the effect on error variance is not very strong. Temporal arbitrage, which is the gross return from speculative storage, appears to be modest. The long- and short-term implications of the findings are discussed within the context of ongoing policy debates.