Today’s opinion poll is from those wild and crazy folks at Galaxy. Published in the Murdoch tabloids, its results are unremarkable: Labor leads by 55-45 two-party preferred, off a primary vote of 46% to 41%. That’s a swing of just under 8%, slightly less than other polls in the last fortnight.

As always, one poll on its own means practically nothing. But this one is consistent with the general trend that shows the Coalition has recovered some ground in the last two months, while Labor still enjoys a huge lead.

More interesting is the way the media coverage, which tries to make every poll significant, has been hoist by its own petard.

The last Galaxy poll, published four weeks ago, showed an unexpectedly narrow Labor lead, 53-47. Some of us warned that it could well be a rogue sample – a view confirmed by subsequent results. But reports at the time seized on it eagerly as evidence of a Coalition resurgence

So now they have to backtrack. Today’s reports depict a swing back to
Labor: the Courier-Mail, for example, says “Labor has opened up a big lead in the past month”.

Since the government seems to have had a pretty good couple of weeks, it’s odd to think that it’s lost ground. Much more likely that things are basically stable, and the previous poll was an outlier.

Statisticians call this “regression toward the mean“: if the outcome of some random process is unusually high (or low), the next one is likely to be closer to the average.

The papers all like to hide the random element in their precious polls, but shoddy reporting eventually exacts its own price.