This study analyses the regional convergence of the levels of labor productivity in the agricultural sector of the state of Rio Grande do Sul, through a Markov matrix. It also examines the variables that explain this convergence, through the statistical technique of principal components followed by statistical regression. The study is based on the data from Agricultural Censuses of the Instituto Brasileiro de Geografia e Estatística (IBGE) referring to the years of 1975 and 1995/96. The study considers four classes of levels of labor productivity. The regions in the upper class, containing the large cattle raising farms, show a tendency to migrate, in the long run, to the lower classes. The reason seems to be the reduction in the area of land available per unit of labor. The lower class tends to maintain, in the long run, approximately the same percentage of the total regions, which will not evolve in the level of labor productivity, maybe because the farms are too small to get mechanized, or do not change the production system to products that aggregate more value, or do not adopt technological innovations.