All OECD countries are faced with substantial inequalities in health status between socioeconomic groups within their populations. One aspect of these inequalities for which data are routinely available in many countries is inequalities in mortality by level of education: people with a lower level of education typically have considerably higher death rates and lower life expectancy than people with a higher level of education. The OECD recently started a project to generate measures of the distributions of ages at death by educational level, gender and cause of death for as many countries as possible. This working paper aims to highlight the most important methodological issues to be faced when trying to create valid statistics on mortality by level of education, and to highlight how different methodologies may affect results and comparisons. Topics covered include study designs (e.g. use of cross-sectional census-unlinked versus longitudinal census-linked data), data harmonization issues (e.g. use of a common educational classification scheme), and data analysis issues (e.g. choice of a summary measure of inequalities in mortality). The paper ends with a number of recommendations for data analysts.