This research work uses stochastic frontier analysis (SFA) and data envelopment analysis (DEA) to examine the sources of agricultural productivity growth over time and of productivity differences among countries and regions in European Union over the period 1980-1998. A comparison of the mean productivity scores obtained by the two approaches show that DEA results are higher than in SFA results, because DEA fits a tighter (more flexible) frontier than the translog frontier. This study is a valuable warning for people to be carefully about the effects of the methodology choice upon their results and to use more than one approach if they suspect that it may have some influence.