By Pattrick Smellie
April 29 (BusinessDesk) - The five-yearly national census could become an annual affair as the official statistics
agency uses more of the data constantly collected by government agencies rather than rely on declining response rates
from individual citizens.
Declining participation compromised the results of the 2018 census and those from 2013. Government Statistician Liz
McPherson today outlined two key areas where last year's census failed to collect sufficiently robust data: household
and family information, and iwi affiliation data.
Lower than anticipated response and survey completion rates for the 2018 census forced McPherson to twice delay the
usual pattern of data releases. She outlined today how Stats NZ had plugged those gaps by fast-tracking plans to make
far greater use of government-collected 'administrative data' to complete population-wide counts in the future.
The agency had been testing and refining models for use of administrative data for seven years already. It had intended
to use an increasing amount of such data from the 2023 census onwards and instead accelerated its modelling processes to
create a statistically robust 2018 census result, McPherson said.
Some 1.2 percent fewer people participated in the census than anticipated. Data gaps left by people not completely
filling in their forms meant partial information equivalent to around 500,000 citizens was drawn from administrative
data sources rather than census forms filled in on census night, March 6 last year.
“The team at Stats NZ has risen to the challenge and delivered a new way of confidently combining the strengths of
census and administrative records to create the 2018 census dataset.
“There are now records for approximately 4.7 million people in the census dataset. The number of records is 1.2 percent,
or 58,000 people, less than our best estimate of the population on Census Day 6 March 2018. In 2013, the official census
undercount was 2.4 percent, or 103,800 people.
"Eighty-nine percent of the total number of records comes from 2018 census forms and 11 percent comes from other
government data."
“We are confident that we are including genuine information about people we are sure were in New Zealand on Census Day,
to help us provide as complete a picture as we can. For example, data on Maori ethnicity and Maori descent is likely to
be more comprehensive than what was released from the 2013 Census,” MacPherson said.
McPherson said the lack of robust data on iwi affiliation was "a significant loss", but that Stats NZ was taking the
failure across government agencies to collect data on individual Maori peoples' iwi affiliations as "an opportunity for
a completely different way forward" in collection of Maori iwi statistics.
McPherson said the 2018 data was robust enough to allow the re-setting of electoral boundaries for the 2020 election and
the population funding models used by public hospitals to determine their budgets, contrary to speculation from critics
of the census process.