to a household that experiences in-work poverty are labourers (+2.7 percentage points) and the group encompassing community and personal service workers (+3 percentage points). When we further break down the type of occupation (see Appendix Table A 3), we find that some groups are rather homogeneous: for example, among the different types of machinery operators and drivers, the in-poverty rate relative to the sample average ranges between -0.2 percentage points for road and rail drivers and -1.3 percentage points for machine and stationary plant operators. However, substantial heterogeneity can be found among most other groups: e.g., among labourers, households with an adult working as a factory process worker experience an in-work poverty rate of -0.4 percentage points lower than the sample average, while the respective statistic for households with cleaner and laundry workers is +6.6 percentage points. These examples suggest that understanding the relationship between occupation and in-work poverty prevalence requires examination of occupational categories in disaggregated detail. However, one drawback of this approach is that it is not clear whether the wage received from the occupation stems from the main or (if present) the secondary earner in the household. We therefore also identify for each household the individual with the highest monthly earnings from wages and salaries and consider the relationship between their occupation and in-work poverty. The sample size of households in this set-up (i.e., defined by occupation of main earner) is 698,628 observations, 25 and the average in-work poverty rate is 4.9 percent. The square markers in Figure 8 illustrates respective distributional numbers relative to the sample average. At the upper end of the occupational hierarchy, we find that the mean in-work poverty rate is lower when considering the particular occupation of the main household earner, relative to considering that occupation for any household earner. For example, this is the case for the categories of both professionals and managers. However, at the other end of the occupational spectrum (i.e., sales workers, community and personal service workers, and labourers), we find the opposite effect. The in-work poverty rate is higher if these occupations are associated with the main earner, relative to any earner in the household. For example, according to the first definition, community and personal service workers have an in-work poverty rate of 3 percentage points higher compared to the mean – and when this occupation was associated with the main earner in the household, the in-work poverty rate rose to 6.4 percentage points higher than the respective mean. Note this sample is lower than the full sample of 725,313 working households due to missing information and/or the identified main earner was working earlier in the year, but not during March 2013 when the Census was conducted, and hence had no occupation to report. 25 Page 33