Lessons from census failure
Chief Statistician Liz MacPherson had little choice but to resign yesterday in the wake of a report that said the leadership of Census 2018 lacked strategic direction, had governance that was ‘‘overly complex and ultimately ineffective’’, and focused too much on the online census.
‘‘We were too optimistic, placed too much emphasis on the online census, and did not have robust contingency plans in place for when things started to go wrong. When that happened, problems were not escalated to a higher level,’’ MacPherson admitted.
Those seem like failures of leadership, the optimism misguided, and understandably the buck stops with her, though an independent report into the 2018 census, released yesterday, does identify some successes, including partnering with external vendors like NZ Post, and an Internet Collection System that was ‘‘secure, stable, and easy to use with over 80 per cent of forms (dwelling and individual) completed online’’.
It’s difficult not to form the view that, if there had been a different outcome to discussions after the 2016 Kaiko¯ ura earthquake about deferring the census, the fallout might have been reduced. We’ll never know for sure, and there is probably not a lot to be gained from that bit of hindsight as Statistics NZ seeks to bounce back from an exercise that failed to count one in 10 New Zealanders, and from which the first results are not due until next month, almost a year behind schedule.
However, the decision to proceed, despite the massive disruption caused by 500 staff having to be moved from Statistics House in Wellington, and the months required to stabilise all IT systems, plainly factored into a lack of sufficient testing of
new systems and inadequate contingency planning identified in the report. It points out that, unlike the 2006 and 2013 editions (the latter delayed two years by the Canterbury earthquakes), which were ‘‘minimum-change’’ censuses, ‘‘Statistics New Zealand embarked on a significant change journey with the 2018 census’’.
The online focus meant the implementation of a ‘‘new suite of interdependent IT systems’’, along with a much-reduced field workforce, only 40 per cent of the size of that used in 2013, a cut identified in the report as too aggressive.
Of course, the use of new IT systems for an exercise as vital as a national census, with the resultant data to be used for such important functions as the drawing of electoral boundaries and decisions about funding for district health boards, requires comprehensive testing. However, the report, by New Zealander Murray Jack and Connie Graziadei, a former assistant chief statistician at Statistics Canada, says the development of the new systems was ‘‘already behind schedule’’ when the 7.8-magnitude quake hit Kaiko¯ ura.
Not surprisingly, the report records that ‘‘the decision to postpone the 2011 census due to the earthquake in Christchurch weighed heavily in the discussion’’. There’s not much that can be done now about the failures that have emerged, though plainly the remediation work still being overseen by MacPherson to get the delayed 2018 data out is important.
However, with an eye on the next census, which the report recommends be held in 2023, those involved in 2018 can at least take comfort from the recommendation that the model planned for last year be maintained, with changes made based on the numerous lessons learned.
Those seem like failures of leadership, the optimism misguided, and the buck stops with her.