Over the last 30 years the share of individuals in the Belgian working-age population without employment (‘individual joblessness’) has fallen continuously, while the share of households with no working-age member in employment (‘household joblessness’) remained fairly stable. In this paper we examine why individual joblessness and household joblessness diverge. The growing gap between both measures of joblessness reflects changes in household composition and changes in the distribution of individual employment over households. We describe the latter phenomenon by constructing a measure of ‘polarisation of employment over households’, which is based on the difference between observed and ‘expected’ household joblessness. On the basis of changes in household formation and changes in individual joblessness one would have expected household joblessness in Belgium to decrease. However, increasing polarisation, i.e. an increasingly unequal distribution of jobs over households, has counteracted this ‘expected’ evolution. Singles constitute households that are most vulnerable to the polarisation we describe, but a shift towards such more vulnerable households offers only a small share of the explanation of the evolution over time. We observe rising levels of polarisation, both in single adult households and couples. Within these household groups, changes in polarisation are similar, but increasing polarisation among couples is the most important factor in the overall increase in polarisation. The personal characteristics associated with individual joblessness (gender, education, age, region, origin) explain a significant part of this polarisation on the level of households. In 2012, almost half of polarisation in single adult households can be explained through typical individual characteristics of singles. In couples almost one third of polarisation is related to individual characteristics and marital selection. However, this also means that a substantial part of household joblessness cannot be explained by these ...