Bourquin et al. (2019) show for the UK that the increase of the in-work poverty rate in the past 25 years is partly due to the “catch up of pensioner incomes (driven by higher state and private pensions)” which “has pushed up median income” (p. 1). The authors conclude that “this highlights one potential oddity of a relative measure of poverty, in which rising pensioner incomes can lead to higher relative poverty for working-age families, as some working-age families subsequently fall behind the rising poverty line” (Bourquin et al., 2019, p. 14). 7 For our circumstances, inclusion of the pensioner households in the overall income distribution to derive the poverty threshold may have the opposite effect. This is because our monthly data for these households does not include income from private superannuation and investments, and we may likely be underestimating income for these housholds, which will pull down the median income threshold and result in a downward bias for inwork poverty prevalence. Therefore, in our sensitivity results in Section 4, we also provide the estimated prevalence of in-work poverty using a poverty threshold that excludes pensioner and selfemployed households. 8 Finally, to generate the population spine for this study, we focus on households where at least one adult in that household is part of the working-age population, i.e., aged 18 to 65 inclusive. We then define a household as ‘working’ if at least one working-age adult in that household is in employment for seven or more months of the year (April 2012 to March 2013 inclusive), which is identified by the months receiving positive wages and salaries. There is no assumption that the months in employment are consecutive or that they are employed in the final month of the year in question (i.e., March 2013). 9 In subsequent tables and figures, we denote these households as ‘working’ or ‘inwork’, and if the converse is true, then the household is denoted as ‘non-work’. Using the above definitions of working and poverty status for the household: In-work poverty = proportion of working households that fall below the poverty threshold Data To study in-work poverty in New Zealand, we use several data sources within the Stats NZ framework of the Integrated Data Infrastructure (IDI). The IDI is a large research database containing microdata about individuals and households in New Zealand. It includes Stats NZ surveys as well as information from government and non-government agencies. The first major dataset is the 2013 Census, which collects information on numerous individual and household characteristics, including the relationship structure of household members. This enables us to identify different types of household structures, such as couples with children or one-person households. The 2013 Census also has a rich set of characteristics, including information on The former UK Prime Minister David Cameron has pointed at this aspect in a speech by stating that “[B]ecause of the way it is measured, we are in the absurd situation where if we increase the state pension, child poverty actually goes up”. See: https://www.gov.uk/government/speeches/pm-speech-on-opportunity 8 See Bourquin et al. (2019), who follow a similar approach by excluding pensioner households. 9 There is also no minimum threshold regarding the level of wages and salaries required to be considered ‘working’ in a particular month. 7 Page 15