information for these TAs, and future empirical examination could delve into more disaggregated details for these areas. At the regional level, we find that Canterbury (-1.3 percentage points) and Nelson (-1 percentage points) exhibit the lowest values for in-work poverty prevalence. At the other end of the spectrum, we find Northland (+3.2 percentage points) and Gisborne (+2.3 percentage points) have particularly high above-average in-work poverty rates. For regions comprising more than one TA, sub-regional variation is evident and substantial in particular regions. This is highlighted with the example of Wellington. This region has an in-work poverty rate below the national average, at 6.4 percent, and ranges from 5.2 percent in Upper Hut to 8.0 percent in the Masterton District. Bay of Plenty also exhibits considerable sub-regional divergence in in-work poverty rates, between 6.9 percent in Tauranga City and 18.3 percent in the Opotiki District. Figure 3. Region and in-work poverty rate Distance from mean in-work poverty rate (in percentage points) 12 9 6 3 0 -3 AUK BOP CAN GIS HKB MBH MWT NSN NTL OTA STL TAS TKI WGN WKO WTC Source: IDI 2019. Notes: AUK = Auckland; BOP = Bay of Plenty; CAN = Canterbury; GIS = Gisborne; HKB = Hawkes Bay; MBH = Marlborough; MWT = Manawatu; NSN = Nelson; NTL = Northland; OTA = Otago; STL = Southland; TAS = Tasman; TKI = Taranaki; WGN = Wellington; WKO = Waikato; and WTC = West Coast. In-work poverty rate as defined in Section 3. 4.2 Ethnicity We next disaggregate our findings by ethnicity. To do so, we define the ethnicity of a household by looking at the prioritised ethnicity for each adult member in the household, based on 2013 Census data. 19 Households with multiple prioritised ethnicities (e.g., those with two adults of different prioritised ethnicities) are included in multiple categories. Ethnicity is prioritised based on the ordering of Māori, Pacific peoples, Asian, MELAA (Middle Eastern, Latin American and African), Other, and New Zealand European. 19 Page 23