5. CONCLUSION In this report we have explored a strategy for deriving the prevalence of in-work poverty in New Zealand, as well as provided an in-depth descriptive empirical examination of the affected groups and related characteristics. The unique nature of the linked administrative and survey data in the IDI has afforded this research the opportunity to extend analysis in areas that have not previously been fully examined in the international literature. In particular, the population-wide data has allowed disaggregation of information at a finer level relative to prior studies. We find an in-work poverty rate of 7.0 percent as at March 2013. Of substantive policy interest is that both Working for Families tax credits and the Accommodation Supplement make a sizable difference to the in-work poverty rate. Without these two income sources, the rate rises to 9.2 percent. Additionally, this rise is markedly greater for single-parent households, where the in-work poverty rate rises from 12.3 to 21.6 percent. With respect to gender, we find that females have an elevated in-work poverty rate compared to males (7.7 percent versus 6.6 percent, respectively). Additionally, one out of ten children living in working households is poor. The relevant in-work poverty rate also varies substantially across regions: Canterbury and Nelson experience below-average in-work poverty prevalence; Gisborne and Northland experience above-average prevalence; and the Bay of Plenty and Wellington exhibit large sub-regional divergence. The in-work poverty rate has been relatively stable between 2007 and 2017. In terms of characterising working households that are more likely to experience poverty, several expected patterns emerge from the empirical analysis. These include a negative relationship between in-work poverty prevalence and educational attainment, as well as the occupational hierarchy. In-work poor households were also more likely to work in agriculture, forestry and fishing, and in the accommodation and food service industries; to receive income from a benefit; to rent their home; to be disabled; and to have health difficulties (particularly around learning and communicating), all relative to their in-work non-poor counterparts. The importance of disaggregating findings at a finer level is evident in the majority of the empirical analysis. A useful example to illustrate this is migrant status. While, on average, having a migrant adult in the household is associated with the average rate of in-work poverty prevalence for New Zealand, individuals born in North-East Asia experienced a rate that was 7.4 percentage points higher relative to sample average, and in contrast, those from the United Kingdom experienced a rate that was 1.7 percentage points below the sample average. Two characteristics that stood out in terms of strong disparities across disaggregate sub-groups were ethnicity and household structure. Households with at least one adult with prioritised ethnicity of Pacific peoples (9.5 percent) or Māori (8.6 percent) experienced a substantially elevated in-work poverty rate relative to households of New Zealand European ethnicity (5.9 percent). This outcome is likely the result of a variety of factors: these households are underrepresented in occupations associated with low rates of poverty; they also have lower educational attainment relative to NZ European households; and they are, on average, larger households so earnings from employment must, in many cases, stretch further than for households with fewer members. In terms of household structure, the lowest in-work poverty rate is observed for households comprising couples without children (4.8 percent), couples with child(ren) (6.3 percent), and oneperson households (6.4 percent). In contrast, households with two or more families and single-parent Page 43