Data Collection
2018 Post-Enumeration Survey
##Sample Design
The 2018 PES sample design was purpose built to balance the desired accuracy targets with the cost and sample size constraints. The was a change from previous PESs which have had sample designs based on existing Stats NZ household surveys such as the New Zealand General Social Survey (NZGSS, the base for 2013 PES) and the Household Labour Force Survey (HLFS, the base for 2006 PES). As mentioned in the previous section, the first stage of the sample design and selection is based on the PSUs which are formed from meshblocks. At the time of the 2018 PES sample design and selection, the NZ geographic framework consisted of 53,589 meshblocks of which 52,973 are within the household survey frame grouped into 23,174 PSUs (excluding PSU 999999 Area Outside Scope). These PSUs get further grouped into strata to improve sampling efficiency. The strata groups are created from a combination of geographic concepts that are included to manage cost of field and interview operations, and concepts that are highly related to the survey objectives (potential to be missed or miscounted by the census). The 2018 PES sample design grouped the PSUs into 102 strata based on:
- Regions
- Mix of urban and rural
- Planned mode of delivery of census forms
- Socio-economic characteristics captured in the NZ Deprivation index Territorial Authority and Auckland Local Board Areas (TALB) were unable to be explicitly included int he strata due to sample size constraints. However, the PSU selection process is designed to implicitly account for TALBs and to maximise the probability of selecting at least one PSU in each TALB. The final stage of the sample design was to boost the sample representation of Pacific people. This was done using the probability proportional to size (PPS) sampling strategy, where the probability of selecting a PSU depends on its size. The PPS proportion was transformed by multiplying it by a factor reflecting the proportion of adults of Pacific ethnicity in the PSU. The targeting of Pacific people was done through the PPS (and not through stratification) because it allowed increasing the survey precision for this group without decreasing the survey precision for other groups.