Got data too poor for time series analysis? Can cluster analysis be a remedy? Studying wheat market integration in Ethiopia

Recent global food price developments have spurred renewed interest in analyzing integration of local markets to global markets. A popular approach to quantify market integration is cointegration analysis. However, local market price data often has missing values, outliers, or short and incomplete series, making cointegration analysis impossible. Instead of imputing missing data, this paper proposes cluster analysis as an alternative methodological approach for analyzing market integration. In particular, we perform cluster analyses on a set of statistical indicators of eight Ethiopian local price series to analyze how they relate to world market prices. Moreover, recognizing several policy regimes in the period 2007-2010 we investigate how market clusters change over time. Results show that in periods with wheat imports via the private sector, several local markets form a common cluster with the world market. In periods with government controlled imports and exchange rate collapse, domestic prices measured by a comprehensive set of characteristics were strongly dissimilar from those of world market prices.