A meta-analysis is perfonned to review empirical estimates of technical efficiency (TE) in developing country agriculture. The objective of the study is to contribute to a better understanding of the factors that influence estimates of mean TE. A data set of 51 observations of TE from 32 studies is used in order to test if specific characteristics of the data and econometric specifications account for systematic differences in the efficiency estimates. Results using the two-limit Tobit procedure indicate that factors such as primal versus dual, number of fixed inputs and number of variable inputs increase average TE estimates. On the other hand, using the Cobb-Douglas functional form and cross-sectional data yields a lower level of TE. Other factors, including the number of variables in the model, crop type, stochastic versus deterministic frontiers and sample size, do not seem to significantly affect estimates of TE across studies. © 200 l Elsevier Science B. V. All rights reserved.