People have a bad habit of wanting to tell stories about random low-probability events. Whether it's a few dog attacks that happened in the same couple of weeks, or a blip in youth drinking rates, people want to know why.
Often, the best answer is that we should ignore individual monthly data points and look to the longer term trend.
The most recent holiday road toll statistics provided fodder for a lot of armchair reckonings over the past week. There were 15 fatal crashes this holiday season; last year, there were six.
Nobody has been keen to take credit for the increase, though we can be fairly sure that if the figure had halved instead of doubling, the police would have credited tough speed enforcement measures and the drop in the blood alcohol limit.
And so we all look for stories to explain the difference. Perhaps everyone had their eyes on the speedometer rather than on the road for fear of breaching the tighter speed enforcement rules. Maybe strict enforcement of the 100km/h limit led to more dangerous overtaking situations.
But the best explanation is probably that events like this can fluctuate a lot from year to year. Accident rates are very low. New Zealand has about eight traffic fatalities for every billion kilometres driven.
So if you drove a billion kilometres this holiday season, you could expect to have died eight times. If you drove a more normal, but still high, thousand kilometres, you would have about a 0.0008 per cent chance of being in a fatal accident.
Or, for every 125,000 vehicles travelling 1000km over the holiday season, we would expect one death.
When events are infrequent, like fatal traffic accidents, small things can affect the figures.
Blogger James K, at Ordinary Times, points out that this year's holiday season was two days longer than last year's, so we should have expected a 20 per cent rise in accident rates solely because of that.
He also shows that when we look at longer-term trends in the daily number of crashes, this year's figures remain within the expected range.
It is a bit odd that this year's figures are higher than last year's despite changes to the drink-driving limit and police speed enforcement, but accident rates in any one season are likely to depend more on the holiday weather than on recent policy changes.
This summer's been particularly nice in Wellington. If the weather had been worse, more people would have stayed home and fewer people would have died in car accidents.
But again, that is just a story. We would need to check longer term accident rates against the weather to be able to tell.
And plausible-sounding stories often fail to bear up. Having been frustrated in a couple of queues this season, I wondered whether tougher enforcement of the 100km/h limit had led to more dangerous overtaking manoeuvres.
But the Ministry of Transport told me that only one of the 15 crashes resulted from overtaking. It would be a stretch to say that one out of 15 is higher or lower than expected - you have to look to longer-term trends to be able to draw conclusions.
Whether it's dog attacks or other random, low-frequency events, it's always best to look to the longer-term trends rather than tell stories based on statistical blips.
Dr Eric Crampton is the head of research at The New Zealand Initiative.