Census 2018: Review shows you could almost count on census failure
Wednesday, 14 August 2019
ANALYSIS: A mad scramble after the Kaikōura earthquake, letters that were never printed, and 'aggressively' cut boots on the ground.
An independent review of Census 2018 shows disaster could almost be counted on by March 6, 2018.
Statistics New Zealand chief executive Liz MacPherson resigned on Tuesday, at the release of the review into the national survey. It has been extensively delayed after failing to count one-in-10 New Zealanders.
Experts have speculated about the failure for more than a year, and Stats NZ remained mostly quiet. But the review detailed the debacle.
**READ MORE:
* National Party still plans to dispute electorate boundaries after census botch up
* James Shaw says resignation over bungled census is appropriate, but not forced
* Census 2018: Chief Statistician Liz MacPherson has resigned
* More than census data is compromised after resignation of chief statistician**
Two minutes past midnight on November 14, 2016, and the Statistics Building partially collapsed in the 7.8 magnitude Kaikōura Earthquake. So began a chain of events which would ultimately derail the census.
Census staff were in crisis mode after the permanent evacuation of the building; recovering and reinstating IT systems was a massive task.
The prior census, 2011, had been delayed due to the Christchurch Earthquake. This was front of mind for Stats NZ officials as they discussed whether or not to postpone Census 2018 due to the Kaikōura earthquake.
On Tuesday, MacPherson said they believed the decision not to postpone was right at the time. The review said the decision raised risk levels and a $2 million contingency fund provided by the Government did not mitigate this risk.
The disruption caused substantial delays. Decisions on what questions to ask in the census came 10 months late. End-to-end testing of the census system was significantly reduced and came a month late.
Risks mounted and concerns were raised, but there was no adjustment to the plan.
The letters Stats NZ planned to send were too complex, and not enough were sent. The plan was to have 1.8 million letters in English and 250,000 bilingual letters, but the contracted printer did not have the capacity.
Instead there were 1.3 million English letters and 53,000 bilingual letter. Of a certain type of letter sent to low-particiation area, only 47 per cent of 4800 bilingual letters were sent.
Some communities waited a long time to receive the paper, and some never did. The lack of paper, and follow up, meant many in Māori and Pacific communities weren't reached.
Then came the field work. Stats NZ had planned to use 3000 ground staff to knock on doors and gather responses; 40 per cent of the staff used in 2013. With an online census, less boots on the ground were required.
This reduction was 'too aggressive', the reviewers said. It was further reduced to 2300, a decision apparently made on the basis of a mathematical model without the risk being considered.
Compounding this was a failure to hire staff. There were 1500 field workers needed for following up with people who hadn't responded to the census, but only 900 were deployed.
MacPherson on Tuesday said it was harder to recruit and retain, as people were asked to work 30 hour weeks for Census 2018. The previous larger workforce staff were hired on a more casual basis, allowing them to fit the work around their existing life.
Following up with hard-to-reach people was 'disconcerting' to the staff, she said.
Throughout this, the reviewers said, Stats NZ had an 'optimism bias'. An initial press release celebrated a successful survey after census day.
MacPherson said it took a while to figure out the responses weren't coming in, and for some time Stats NZ thought it just needed to 'work a bit harder'.
Follow-up was planned for 10 days after the census, and the reviewers said management were slow to react.
She notified Minister James Shaw 'shortly before' going public with the low response rate in June 2018.
The rest would come out in time.