Various international studies have demonstrated socio-economic differences in health. Linking the 1991 Census to the National Register and using the Health Interview Survey 1997 has enabled assessment of the association between the level of education and health in Belgium using the composite indicator 'health expectancy'. The Sullivan method was used to calculate health expectancy on the basis of current probability of death and prevalence of perceived health. Two measures of educational attainment were used: absolute educational attainment and the position on a relative hierarchical educational scale obtained by a regression-based method. The latter measure enables international comparisons. Differences in health expectancy by education were spread over the whole range of the educational hierarchy, and were consistently larger among females than males. At 25 years of age, the difference in health expectancy between different levels of education reached up to 17.8 and 24.7 years in males and females, respectively. Compared with people with the highest educational attainment, males and females at the lowest level of education spent more than 10 and 20 additional years in poor perceived health, respectively. Between ages 25 and 75 years, the difference in health expectancy between people with the lowest and highest levels of education was 17 years among males and 21 years among females. Compared with people at the top of the relative educational scale, males and females at the bottom of the scale had 13.6 and 19.7 additional years in poor perceived health, respectively. The conclusions of this study in Belgium are consistent with studies in other countries. People with a low level of education have shorter lives than people with a higher level of education. They also have fewer years in good perceived health, and can expect more years in poor health in their shorter lives. The inequality in health expectancy seems to be greater in females than males.