7. APPENDIX Potential caveats regarding income sources To construct our population of interest, we use 2013 Census information that enables us to identify New Zealand households for March 2013. This enables analysis at a very granular level (i.e., comparing differences across a wide range of variables, as shown in the sub-sections of the results, such as birthplace, educational achievement, household structure, occupation, industry, etc.), which has often been lacking in past literature, due to usually relying on survey data. Importantly, this means that we require income information also on a granular level (specifically for the month of March 2013) to determine poverty status of a household. The standard approach for deriving the poverty threshold (which then determines the in-work poverty rate) is to account for all income sources for the whole population. The Integrated Data Infrastructure (IDI), which is the gateway to the administrative data we use, hosts several datasets on different income sources. Some information is provided on the monthly level and some on the annual level. For instance, on the one hand, the employer monthly schedule (EMS) provides monthly information on taxable income where PAYE is deducted at source. This includes wages and salaries, New Zealand superannuation, government benefits, paid parental leave, etc. on the monthly level. On the other hand, data within the IR3 which includes income from “non-zero partnership, self-employment, or shareholder salary income, as well as rental income” 31 is available at the annual level. It would not be suitable to simply convert these annual figures to a monthly (March) figure (e.g., by dividing the annual figure by 12), so these annual sources are not counted in the income calculation. By restricting our focus to data available at the monthly level, we expect to underestimate the importance of income from self-employment and from investments. Potential effects therefore include: • • • Overestimating the number of households without any income. According to Table 2, we exclude 168,057 households (the difference between the second and third line) that do not have any income records for the sample period. This will potentially include households that receive income solely from self-employment or investments. Underestimating the income of households where one or more members are self-employed. Underestimating the income of households that receive income from investments and those receiving private superannuation. The distribution of the net equivalised income for the whole population can be found in Table A 1. According to the income distribution, we find for our whole population that 19.9 percent of the households are identified as poor when setting the poverty threshold at 60 percent of the median. Interestingly, despite the caveats described above, our overall poverty rate is comparable to OECD estimates. The OECD database finds the share of poor households after taxes and transfer (using a 60 percent poverty threshold) was 19.1 percent in 2012 and 19.6 percent in 2014 (no numbers were available for 2013). 32 Another dimension to test the comparability of the income distribution utilised is the overall child poverty rate. According to Stats NZ (2019b), the child poverty rate (children living in households below 60 31 32 See: http://archive.stats.govt.nz/methods/research-papers/topss/comp-income-info-census-idi/data-sources.aspx Numbers were retrieved from https://stats.oecd.org (25 September 2019). Page 48