Market information services established in 1999 were aimed at the promotion of market efficiency through provision of information across the nation. While the responsible bodies have improved the knowledge of prices, information exchange and flow, as a result of competition between markets, is not known and questions of market effectiveness still stand. This study examines market efficiency based upon response to price signals across Ugandan markets. We focus on information exchange for maize and beans among 16 key markets. We study weekly price data from the first week of 2000 to the last week of 2003 from each of the sixteen markets. Each commodity is studied separately using Vector Autoregessions (VARs) and Directed Acyclic Graphs (DAGs). The two techniques are widely used to show market risk and causal relations in time series data. While results are presented individually for each commodity, the markets are comparable. In determining market efficiency, we test for stationarity of the data, explore the magnitude of forecast error decompositions over time across markets, and observe the patterns of communication based on DAGs. We find that markets are more efficient in exchanging information on maize than beans. Communication of data is mostly between markets in eastern, western, and central parts of Uganda. Overall, markets are very slow in reacting to information in the short run.Information from the Mbale and Iganga markets, which are located in areas of high production, is very valuable in the maize trade. However, of the two markets, it is data from the Mbale market, located near the border with Kenya, which is of paramount importance. Specifically, price is discovered in Mbale in the maize trade. Our results also show the Gulu market, which is situated in an insecure zone, to be very responsive to price signals over the long run. In the case of beans, it is the price signals from Tororo and Jinja that cause more disruption in most of the markets. Price is discovered in these two markets. A majority of the markets is more affected by data from Jinja than Tororo. This segmentation in market price discovery suggests an existing market failure. Arua and Gulu are found to be the least responding markets in regards to price signals for beans. We do not find information from the Kampala market to be important in either the maize or beans trade.
