Background: Although many different factors have been identified to contribute to excess male mortality, it is still unclear which path of the complex cause-effect chain is the decisive driver of the life expectancy gap between women and men. Objective: The question behind this study is whether these sex differences are caused primarily by factors leading to low female mortality or rather by factors causing high male mortality. We hypothesise that they are to a large extent caused by specific subpopulations of men with particularly high mortality levels that decrease the average life expectancy of men. Methods: To test this hypothesis, we investigate in a meta-analysis the variability in mortality (VM) in women and men - defined as the range of death rates prevailing among subpopulations - in empirical studies analysing specific phenomena of mortality differentials. We used the data of 72 empirical studies, including 146 total effects (TE) and 1,718 single effects (SE) for 21 different risk factors. Results: In 85% of TE and three quarters of SE the VM was higher in men than in women, taking into account men's higher overall mortality. The corresponding figures for the direct differences in the VM between women and men are 92 and 82%, respectively. Cases with higher female VM are rare exceptions and appear in particular in the highest age groups. Conclusions: We find support for our hypothesis that the disproportionate high mortality levels of specific male subpopulations are the central cause of the current extent of sex differences in life expectancy. Thus, public health programmes should be targeted toward these disadvantaged subpopulations among men which seem to be related primarily to socioeconomic characteristics.