This report mm•ins lbe reaults of five metbodok,gicel lbldies carried oat ia different COllllbies in Africa during lbe c:oune of 1987. The participating CIDlllllries were: Benin, Centrel Africu Republic, Keay,. N°Jger and z.imbabwe. Each COlllllry carried out lbe study, using a common experimenteJ design, with the objective of comparing crop esti -mates bued on the classic 'crop-cut' method with estimates -obtained by asking fanners direc:tly to state their production. Both types of estimates were compared egainst a com -plete barvest and weighing of the sampled plots. The objective of the study was to test the hypothesis that estimates of production obtained by interviewing farmers could be at least es accurate as estimates obtained through physical measurement involving crop-cutting on sample subplots. If this could be demonstrated, it would make posul>le the introduction of a number of cheaper ud more efficient improvements to the sample design ud to the manner in which large-scale production surveys arc executed. The data from all five studies were pooled ud analyzed together. They gave the following results: 1) Estimates based on square -cuts ap -pear to result in serious over-esti -mates - by around 30%. There is considerable variability among countries however with the over-estimate ranging from 15-40%. 2) Farmers estimates are n:marbbly dOIC to ac:tua1 production figures in all countries, varying from -8% to + 7%. They also display considerably smaller variances than the crop-cut estimates. 3) Farmers estimates obtained before harvest. (forecasts) arc also good at pmlic::ting production levels but are subject to significantly greater variance. 4) Farmers estimates of planted area do not appear to provide a viable alternative to objective area measurements and gave very considerable over-estimates. When analyzed separately, each of the five country studies confirmed ud supported the overall conclusions noted above. The results, thus, strongly support the study hypothesis and indicate that, under a wide range of ge -ograpbical, soc:ial, ud administrative condi -lions, farmers' post harvest estimates performed in a superior fCl$hion to the so -called objective method both in terms of predicting the average value ud hawig a smaller variance.