BACKGROUND: Many studies have found socioeconomic differentials in cancer survival. Previous studies have generally demonstrated poorer cancer survival with decreasing socioeconomic status but mostly used only ecological measures of status and analytical methods estimating simple survival. This study investigate socio-economic differentials in cancer survival using four indicators of socioeconomic status; three individual and one ecological. It uses a relative survival method which gives a measure of excess mortality due to cancer. METHODS: This study uses prospective record linkage data from The Office for National Statistics Longitudinal Study for England and Wales. The participants are Longitudinal Study members, recorded at census in 1971 and 1981 and with a primary malignant cancer diagnosed at age 45 or above, between 1981 and 1997, with follow-up until end 2000. The outcome measure is relative survival/excess mortality, compared with age and sex adjusted survival of the general population. Relative survival and Poisson regression analyses are presented, giving models of relative excess mortality, adjusted for covariates. RESULTS: Different socioeconomic indicators detect survival differentials of varying magnitude and definition. For all cancers combined, the four indicators show similar effects. For individual cancers there are differences between indicators. Where there is an association, all indicators show poorer survival with lower socioeconomic status. CONCLUSION: Cancer survival differs markedly by socio-economic status. The commonly used ecological measure, the Carstairs Index, is adequate at demonstrating socioeconomic differentials in survival for combined cancers and some individual cancers. A combination of car access and housing tenure is more sensitive than the ecological Carstairs measure at detecting socioeconomic effects on survival--confirming Carstairs effects where they occur but additionally identifying effects for other cancers. Social class is a relatively weak indicator of survival differentials.