Problem gambling figures misrepresented yet again

Back in May, I wondered why Phillip Blond had suddenly taken an interest in fixed odds betting terminals. It now transpires that his think tank, Respublica, had been commissioned by the Campaign for Fairer Gambling to write a report about them. It was published today and you won't be surprised to hear that it supports Derek Webb's longstanding goal to reduce the stakes to an unplayable £2.

I wouldn't bother mentioning it here if it weren't for the fact that it does what so many anti-FOBT campaigners do and lies about the problem gambling statistics... 

The latest available research has found that the number of problem gamblers has surged – from 280,000 in 2012 to 430,000 in 2015. 

Respublica provide two references for this: a Gambling Commission report about England and Scotland with statistics from 2012 and a Gambling Commission report about England, Scotland and Wales with statistics from 2015.

Respublica don't mention the fact that the latter report has an extra four million people in it (three million people live in Wales, plus UK population growth of around one million). In fact, they explicitly claim that both reports only look at England and Scotland. They then claim that there was a rising in problem gambling in these three years 'of over 50 per cent'.

This is implausible on the face of it and it is untrue. The 2012 report gives an estimate of the number of problem gamblers under the two usual measures: 

The confidence interval for the DSM-IV estimate was 0.3%–0.7%, for the PGSI estimate 0.2%–0.6% and for either screen 0.4%–0.9%.

And the 2015 report says:

The confidence interval for the DSM-IV estimate was 0.5% to 1.0%, for the PGSI estimate 0.4% to 0.9% and for either screen 0.6 % to 1.1%.

The figures from 2015 are higher, but there is not a statistically significant difference. All these estimates tell us is that there is a 95% probability that the real figure lies somewhere between the confidence intervals.

Even a naive reading of the stats does not imply a 50 per cent increase, however, and the Gambling Commission's most recent (absolute) number is 320,000 people, not 430,000 people.

If you look at the figures from 2010 you will see that they are higher than in either of the subsequent reports, being 0.7%-1.2% under the DSM-IV estimate and 0.5%-1.0% for the PGSI estimate. (The 2010 report didn't combine the two to come up with a third estimate.)

And the 2010 figures were slightly higher than the 2007 figures. So it goes. These estimates have wide confidence intervals and they fluctuate a bit but there is no visible trend in either direction. (See here for more details about this.)

There was not a significant increase between 2012 and 2015, just as there was not a significant decline between 2010 and 2015. Every problem gambling survey since 1999 has been consistent with the hypothesis that problem gambling prevalence has held steady at around 0.7%.

As I have said before, you can only pretend that there is a trend if you cherry-pick your reports and ignore the confidence intervals. That's why rates of problem gambling have appeared to be doubling and doubling in the last decade, if you believe the media, without the number of problem gamblers ever getting larger. 

There isn't much else in the Respublica report to discuss because it doesn't provide much in the way of evidence, but a dishonourable mention should go The Times for making this howler in its coverage...

In total about 1.5 million people play the machines, collectively losing more than £1.7 billion last year, almost £12,000 each on average.

£1.7 billion divided by 1.5 million people is £1,133, not £12,000. The Times is out by a factor of ten.