BACKGROUND: The political and social transition in central and eastern Europe has been generally associated with widening educational differences in life expectancy. However, interpretation of these findings is complicated because the size of educational categories within the population has also changed. It is therefore important to disentangle these two phenomena. SETTING: The Czech Republic, Estonia, the Russian Federation and, as a western European reference, Finland, in two periods, 1988-89 and 1998-99. METHODS: Life tables were calculated in three categories: university; secondary; and less than secondary education. Changes in life expectancy were decomposed into contributions of population composition and within-category mortality. RESULTS: In Finland and the Czech Republic improvements are seen in all educational groups, with only a slight widening of the educational differences. Over 80% of the total life expectancy increase is attributable to improved mortality within educational categories. In Estonia and Russia, less favourable overall trends coincide with a dramatic widening of the educational gap. A decrease in life expectancy in those with low and middle education has been compensated for, to a small degree in Russia but a greater extent in Estonia, by improvements among those with higher education and by the improved population composition. For highly educated Estonians, the gains were seen at all ages, the greatest at age > or =60 years. In Russia mortality increased in those <60 years although compensated for by improvements at older ages. CONCLUSIONS: Russia and Estonia exhibit much less equitable transitions compared with the Czech Republic. Analyses of trends in health inequalities should capture the changing population composition. In Russia and Estonia an improved educational structure prevented an even greater decline in life expectancy. The highly educated Estonians can potentially catalyse a wider health progress.