Table 2. Preparation of the population spine from the 2013 Census 1. Full Census dataset 2. Dropping those without household ID 3. Dropping households without any income records for the sample period → calculation of the poverty threshold 4. Dropping households that contain a self-employed member 5. Dropping households where no member is 18 or above 6. Dropping pensioner households 7. Dropping households where one member was outside of NZ for more than 90 days in the sample period → population of interest Individuals 4,354,686 4,029,348 3,661,971 Households 1,549,890 1,381,833 2,925,468 2,924,832 2,609,091 1,142,634 1,142,181 920,823 2,438,994 874,797 Source: IDI 2019. Notes: Sample period: April 2012 – March 2013. Self-employment status is derived from self-reported information in the 2013 Census. Income records refer to income information in the EMS. We then employ a number of further exclusions to create our population of interest, from which we can separate households into ‘work’ and ‘non-work’. These exclusions are rows 4 to 7 in Table 2, and include dropping: households that have a self-employed member; households containing only individuals aged below 18; pensioner households; and households where a member of the household was out of the country for more than 90 days in our twelve-month period of interest. Our final sample of households is 874,797. We match this sample to Inland Revenue records to identify the employment status of individuals within the household, categorising them as ‘in-work’ if an individual of working age in the household earned positive wages and salaries for at least seven months (i.e., non-zero earnings for each month) in the year ending March 2013. We are also interested in investigating how the prevalence of in-work poverty has changed over time. Unfortunately, we cannot follow household units over time in the IDI (via the 2013 Census), and we therefore rely on the Household Labour force Survey (HLFS) for this supplemental analysis. Identification of working households is based on information provided by the HLFS, and income information is derived from matching these individuals and their respective households to their corresponding records available from Inland Revenue. The HLFS is a survey that interviews about 15,000 households (approximately 30,000 individuals) on a quarterly basis and is available in the IDI from 2007 onwards. We also utilise the HLFS to derive some additional labour market related statistics that are not available at the population-level, such as whether the individual working is a union member or on a permanent employment contract. One limitation of the HLFS is that due to its small sample size, descriptives cannot be provided at the same level of disaggregation that is done when using 2013 Census data. Another drawback is that the dataset holding the monthly information on WfF and AS ends in March 2013. Wherever we state in-work poverty numbers that exclude WfF and AS, it is clearly indicated in the report. Page 17