Several studies have shown a positive relationship between mortality and episodes of income decline, unemployment, or poverty shortly before death or in the more distant past. Our objective was to analyse the mortality effects of earlier income changes more generally, net of the overall level. We used Norwegian register data that included individual histories of annual labour income and focused on mortality among men aged 50-69 in 1990-2002. Men in this age group who, during the preceding 15 years, had experienced at least two substantial falls in income as well as at least one substantial increase, or vice versa, experienced an excess mortality of 17 per cent. For men who experienced fewer changes, there were only weak indications of excess mortality. Variation dominated by falls in income did not have a more adverse effect than variation dominated by rises.