CONSUMPTION EFFECTS OF ECONOMIC POLICY

Evaluation covers the period 6/78-3/80 and was based upon a review panel discussion with the Michigan State University (MSU) project team. Although the quality of production data was good, consumption data was poor. In particular: (1) Consumption data was estimated as a residual. (2) Too many men were surveyed; women would have provided more accurate data regarding the consumption of non-cash crops. (3) Time series data concerning changes in income and prices were not available; the results from cross-sectional data should be qualified accordingly. The panel also noted that: (1) The elimination from the Sierra Leone survey of data included in the original project sample raises the possibility of a systematic bias. MSU should determine the significance of the omitted data. (2) The use of clustered samples was not a problem. Account should be taken, however, of the characteristics of the original clustered sample before drawing possibly erroneous statistical inferences -- as already happened in Working Paper No. 2. (WP). (3) Kano State data should be used to analyze the determinants of food consumption at the household level. MSU should recommend survey designs that can be replicated. WP does not contain a detailed analysis of food consumption factors in relation to different household groups. The latter should be classified according to socioeconomic factors and conclusions drawn from each group. Multiple regression analysis should be used for both Kano State and Sierra Leone data to determine behavorial relationships, linking such variables as religion and ethnic origins. This analysis should be submitted to AID/W before proceeding with the micro model. MSU should continue with the modelling of behavior of rural households as a producing and consuming unit after completing the food consumption analysis. A formal macro model should not be developed for Sierra Leone; a food accounting matrix should be constructed instead.