Chief Statistician Liz MacPherson resigned this morning, saying she took full responsibility for the "shortfalls identified in the report".
The independent review concluded that Census 2018 was adequately funded, but problems arose with IT systems and staff displacement following the November 2016 Kaikōura earthquake.
The official results were initially slated to be released in August 2018, but they will now be released in September 2019.
The SMC gathered expert comments on the review.
Professor Richard Arnold, Victoria University of Wellington, comments:
"Statistics New Zealand has just received its independent review of the Census 2018.
"The census was a 'digital-first' collection, in which 70% of respondents completed an online census form. The goal of this new version of the census was to save money for the government, in what is an extremely costly five yearly exercise to count the New Zealand population. It was also hoped that the online census would reverse the (slow) decline in response rates.
"However, the census had a catastrophically low response rate. Instead of a undercount of 2-3% (the trend of recent censuses), the undercount was 9% for partial completion of the census, and 14% for full completion.
"Statistics NZ has tried to repair the data set – by imputing data from administrative and other data sources. They have used other data sources, including the 2013 Census, to estimate the properties of the people who did not respond to the census, and claim that in doing so they have reduced the undercount to only 1.4%.
"For key outcomes – the drawing of the electoral boundaries and DHB funding – the Census has met its requirements. However, they admit that for a host of other variables, such as iwi affiliation, the results will be insufficiently accurate.
"This has also led to a delay of more than a year in the release of the first census results: they were planned to be released in August 2018, and they will now finally be released on 23 September 2019.
"What went wrong?
"The review found that 'many aspects of the census
went well', and that:
• The goal of the digital-first census was laudable
• The approach that Stats NZ took to it was sound
• The resources Stats NZ had were sufficient.
• There was lack of 'overall governance and strategic leadership';
• There was an over-optimistic attitude that didn’t recognise the complexity of the task;
• There was with an overemphasis on the online dimension of the census, and insufficient attention paid to the respondents who would find the online census difficult to complete;
• Testing was insufficient;
• Some poor decisions were made;
• The field force was too small for the task that it faced;
• Decisions were made late, and in particular remedial actions were made too late to fix the response rate when it was clear it was too low.
"Key among the decisions that put the response rate at risk were the decisions not to leave a paper form and not to make personal contact with households that had individuals who had not responded and who needed an access code to begin the census.
"What seems clear is that the lack of personal contact with a census field operative, which was the backbone of the previous census model, was a barrier to participation for too many people. Moreover a decision was made not to follow up partially responding households, where only some of the residents had completed their personal forms.
"The review also found accounts of long waiting times for helpline calls, and insufficient staff for community liaison, especially for Māori people.
"From the outside, it looks as if the appropriate technical and statistical expertise was lacking at the highest levels of decision making. The Review concludes that Statistics NZ should continue to follow this model for the next census in 2023, but that there are a number of serious lessons it needs to learn from this experience."
No conflict of interest.