Straight Statistics frequently focuses on statistical claims that are too bad to be true – but today there’s one that’s almost too good to be true.

The NHS Information Centre has published a survey carried out for it by the National Centre for Social Research and the National Foundation for Educational Research, showing declines in smoking, drinking and drug use among 11 to 15 year-olds. The eye-catching figure is a sharp fall in the numbers of pupils saying that they had ever drunk alcohol, or that they had drunk it in the last week.

The general trend in both these measures has been a steady decline for the past decade (see Table) but the 2010 figures suggest this decline is accelerating. The numbers saying that they had drunk alcohol in the past week fell from 18 per cent in 2009 to 13 per cent in 2010 – five percentage points in a single year.

This result is further evidence that public anxiety about drinking is exaggerated, as frequently argued here. It would be good to believe that underage drinking is in terminal decline, but that would be premature.

The authors of this section of the report, Rosie Sutton and Sally Bridges, clearly have their doubts. The big fall is matched by a fall in the numbers of pupils saying that they had ever drunk alcohol, which fell from 51 per cent in 2009 to 45 per cent in 2010. In both cases, they say, the fall continues the downward trend seen in recent years, but represents a greater fall than would be expected, and that future surveys will establish how it fits into the trend.

There is one possible explanation, beyond the obvious one that pupils bombarded with advice about the evils of drink have become more inclined to lie about it. The sampling was changed in 2010. Until then, schools to be sampled were selected by school type and sex of intake, and the distribution across regions was in proportion to the distribution of 11 to 15 year-olds.

In 2010 the sample was selected by Strategic Health Authority, with each SHA contributing an equal number of schools. While this may provide a more up-to-date regional analysis, in hindsight it might have been better not to base the sample on bodies that face extinction under the Coalition Government’s health reforms. However, they weren’t to know that at the time, and the same geographical areas can be used in future even if SHAs cease to exist.

The team did indeed wonder if the change of sample could be responsible for the sharp drop in the prevalence of drinking alcohol, or if it could be accounted for by the fall in the school response rate, down from 54 per cent in 2009 to 48 per cent in 2010. (The reasons schools gave for refusing to respond were lack of time, burden on staff or pupils, or simply survey fatigue – schools are evidently worn out by the number of surveys being conducted.)

The team made a series of checks, including checking sample sizes and response rates, comparing 2009 and 2010 while excluding London schools in case there was a “London effect”, and comparing the 12-15 age group in 2010 with the 11-14 age group in 2009. None of these provided any reason to believe that the results had been affected by the change in sampling.

However, it’s not impossible. Those who carry out surveys are constantly tempted to make them better by small changes, and tend to disregard the fact that for most users consistency of the time series matters more than any improvements actually achieved. We’ll have to wait until next year to find out if the 2010 results are a fluke.

But we’ll know, which may no longer be the case for older people, given the announcement that the NHS IC will no longer support the section of the General Lifestyle Survey that measures statistics on smoking, drinking and health. It seems extraordinary that this long-running time series can be under threat because of cuts to the NHS IC’s budget, but such is the case. Earlier this month Sir Michael Scholar, chair of the UK Statistics Authority, wrote a brisk letter to Health Secretary Andrew Lansley asking him to reconsider. No reply has yet been made.