households experience elevated in-work poverty rates of 9.6 and 12.3 percent, respectively. For these latter household types, it was also evident that if at least one child in the household was under five, the in-work poverty rate was higher relative to similar household types without this circumstance. This further demonstrates the strong inter-relationship between child poverty and in-work poverty and signals that tackling one will likely result in improvements in the other. Relatedly, the number of workers in a household also has an important bearing on a household’s inwork poverty rate. Amongst households comprising a couple with children, the in-work poverty rate for those with only one adult working was 13.5 percent; and this falls to 1.9 percent when both adults in the household meet the working threshold. This is a substantial difference. These findings converge with the international literature which finds that the number of workers in a household is a very strong predictor of in-work poverty. With recent examples in the international literature giving prominence to the role of housing costs, and constructing in-work poverty prevalence pre- and post-housing costs (see for instance Hick & Lanau, 2017), we also examined the relationship between home ownership and housing costs to inwork poverty in New Zealand. We observed a strong relationship between home ownership status and in-work poverty, with renters facing an in-work poverty rate of 9.2 percent while those who own their own home (in whole or in part) have a lower rate of 5.2 percent. Furthermore, for renters we found an average BHC in-work poverty rate of 8.1 percent, which rises to 12.8 percent after accounting for housing costs. The empirical analysis in this report concluded with an investigation of within-year variation in poverty duration for both in-work and non-work households. We found that while 7.0 percent of working households are poor based on information for March 2013, just 1 percent of these households experience poverty in all twelve months (and 6 percent for at least seven months) of the preceding year. These sub-populations, particularly those that experience being below the poverty line every month in a twelve-month timeframe, are the most vulnerable with respect to ensuring their income meets the basic needs of the household. Finally, there are a few caveats of our analysis that are worth discussing. This report provides an exploratory understanding of measuring in-work poverty with administrative data in New Zealand. Due to requiring monthly income information, we may potentially be underestimating selfemployment income, as well as income from private superannuation and investments, which are only available at the annual level. Of note is that the sensitivity analysis indicated that if, when deriving the poverty threshold, the income distribution used as the reference excluded self-employed households and pensioner households (the two groups most likely affected by this missing income), then the inwork poverty prevalence rose to 12.8 percent. Another caveat to note is that our descriptive profile was hampered by lack of data in some domains, such as housing costs and hours worked. Our analysis also relied on the assumption that household composition remains unchanged in the year ending March 2013. We were also not able to follow population-wide patterns of in-work poverty over time. Future research could follow individuals from the 2013 Census to future Census waves to assess longitudinal changes in in-work poverty status. 30 Further empirical analysis could also model accumulated in-work poverty rates from multiple risk factors, as well as delve into a qualitative examination of households that are most vulnerable in this respect. It is important to note here that the signal from Stats NZ is household information may be of poor quality in the 2018 Census, so linking between 2013 and 2018 Census will potentially be problematic. 30 Page 44