This paper deals with structural changes that are observed in farm sectors in many developed economies: the increase in farm size and in farm specialization. Using panel data on Israeli farm communities for the years 1992-2001, we estimate a system of simultaneous equations in which these variables are determined jointly in a dynamic setting. We employ the Arellano and Bond dynamic panel GMM algorithm for each of the equations, treating the other variables as endogenous and allowing for unobserved heterogeneity and for time trends that depend on geographical and institutional factors. The results exhibit positive and statistically significant autoregressive effects in both size and specialization. Farm size depends negatively on specialization, while specialization does not depend significantly on size. This implies that opposite to what one may conclude from aggregate data, Israeli farms expand by diversifying into additional crops and/or livestock rather than by expanding existing enterprises. Simulations show that while specialization increases in all types of communities, farm size is increasing over time in cooperative villages and decreasing over time in collective farms. This implies a gradual process that leads, in the long run, to concentration of agricultural production in a small number of large, business-oriented, family farm enterprises.