The article discusses the convergence of unemployment rates at the county level in Poland in 1999-2006, on the basis of available statistical data. The authors examine both β- and δ-convergence; the former involves the relationship between the growth of the unemployment rate and its initial level, and the latter is based on an analysis of the dispersion of the rates and their changes over time. The authors use methods that enable them to examine changes in the distribution of the analyzed variables. These methods include transition matrices and a nonparametric kernel estimation method. Transition matrices make it possible to determine the likelihood of a county’s unemployment rate increasing, decreasing or remaining constant, while classifying the rates into several brackets. Kernel estimation, in turn, makes it possible to analyze the full conditional function of the density of the distribution of the unemployment rate at the county level and its changes over time. These methods were borrowed from research into regional convergence for income. They make it possible to detect the occurrence of polarization, or the so-called club convergence. The analysis of unemployment rates at the county level in 1999-2006 reveals a far-reaching stability of the regional distribution of unemployment rates—in terms of both monthly and yearly changes. Over the past seven years, no δ-convergence has occurred. The researchers have only detected slightly growing similarities between labor markets in counties with the highest relative unemployment rates. The analysis of β-convergence reveals a far-reaching divergence of unemployment levels in individual counties in Poland. This trend is less pronounced in counties with the lowest relative unemployment rates, while being markedly stronger on labor markets heavily affected by joblessness. Overall, the study places a question mark over the effectiveness of cohesion policies carried out in Poland through various channels since 1999.