Has the rapid inflation of the nation's belly been curbed? After several years of agonised debate over whether Britain's children were becoming the chubbiest in the world — or merely the second or third chubbiest — a study published in October calmed nerves by suggesting some corner or other had been turned. It was vague about what had actually changed, but offered the view that far fewer children will be overweight or obese in 2020 than previously expected. Cue cheerful headlines and preening ministers congratulating everybody, not excluding themselves, for this turnabout.

The curious feature of the new prediction, omitted from media reports, was that it came from exactly the same team as the old one it replaced. In 2007, the Cabinet Office Foresight team had painted a picture of unrelieved gloom, forecasting obesity for several decades hence by a heroic extrapolation from very limited data. The new predictions, from the National Heart Forum, added a modest three more years of data (2004-7) and came to a very different conclusion. 

In 2020, rather than 19 per cent of teenage boys being obese, it forecasts that just six per cent of them will be. Instead of 30 per cent of teenage girls, it will be only nine per cent. In fact, we will have roughly the same number of overweight and obese teenagers in 2020 as we do today. Among males between the ages of 11 and 30, "the predicted prevalence of obesity is actually falling", says the report in wonderment.

This is not a case of two academic teams arguing over conclusions. Both reports were produced by the same team, led by the Oxford Professor Klim McPherson. So what's going on? It's obvious that the model the team is using to predict our children's future weight is ultra-sensitive to the input data. That means it's virtually useless as a forecasting device.

Somewhere in the maw of the government machine, there is a report saying exactly that. Completed last March for the Department of Children, Schools and Families by a committee chaired by David Buckingham, Professor of Education at London University, it mocked the pretensions of the 2007 Foresight report. In a technical appendix it argued acutely that the Foresight projections were based on extrapolating 12 years' data for a further 46 years "with apparently no theoretical justification for the particular extrapolation model". The report concludes: "It might be better simply to acknowledge that we currently have no scientific basis for making such long-term projections."