Highest Since Whenever: How to Read a Government Statistic

This week the Government announced that school attendance had reached its highest Term 2 level since 2016. The number itself was 64.2 percent. Both halves of that sentence are true, and only one of them made the headline, which tells you nearly everything you need to know about how official statistics reach the public.

Start with the baseline trick, because it is the workhorse of the genre. “Highest since 2016” is a comparison, and every comparison has a hidden second character: the year that was chosen to compare against. Choose a bad enough baseline and almost any present-day number becomes a triumph. Attendance genuinely is up 5.8 points on last year, which is real progress and worth reporting. But the framing invites you to feel the way you would about a good result, when the underlying figure means roughly one in three students is not regularly in school. The announcement is accurate. The emotion it is engineered to produce is not.

The second trick of the week was subtraction. Annual inflation came in at 4.1 percent, driven by petrol up 27.5 percent and diesel up 71 percent. The Government noted that excluding fuel, inflation was 2.9 percent. Now, economists do this for legitimate reasons, because volatile items can mask the underlying trend, and there is nothing dishonest about the calculation itself. The dishonesty risk lives in the presentation. “Excluding fuel” is a technical adjustment when an analyst does it and a consolation prize when a press release does it, because no household gets to exclude fuel. The test is simple: is the adjusted number being used to understand the problem, or to shrink it?

There is a third habit worth naming, the percentage-of-what problem. A number like “199,000 more students attending regularly than in 2022” sounds enormous, and 2022 was a year of pandemic disruption, teacher shortages, and general chaos, which is precisely why it was picked. Raw counts with flattering baselines are the compound form of the first trick, and they travel well because big numbers feel like evidence.

None of this means the statistics are cooked. Stats NZ and the agencies producing these figures are generally straight. The numbers are real. What gets bent is the frame around them: the baseline, the exclusion, the choice of which figure leads. So when an announcement arrives, ask the three questions in order. Compared to what? Excluding what? And what does the raw number say without its handler standing next to it? The answers are usually in the release itself, a few paragraphs down, past the part that was written to be quoted.