International agricultural technology.: Comparative study with nonlinear panel models

The most common approach used in international comparisons of agricultural productivity and technological change is generally based on Cobb-Douglas production functions. Similarly, the technological level is usually modelled as a linear deterministic trend. In this paper, we propose more general and less restrictive specifications. In particular, we consider a translog production function as well as a dynamic stochastic representation of the technological level, which produces a non-linear dynamic panel data model. The model is estimated for a sample of 104 countries and various subsamples, over the period 1961-1991. We find that the proposed model outperforms the alternative specifications considered and we use its results to characterize the agricultural technological processes and to make comparisons among groups of countries. A striking result is that the technological gap between developed and less developed countries has considerably increased over this period of time.