Tuesday, 22 August 2017

Stop lying about problem gambling

If the case against fixed odds betting terminals (FOBTs) is so strong, why does everybody who campaigns against them resort to demonstrable lies?

I was going to blog about this article from Victoria Coren in The Observer. It pains me to have a pop at her because she is one of Britain's better writers and one of our finest poker players. Her article centres around a statistic that I have not been able to verify and which she is only able to support with a link to a Daily Mail story. It also contains the false claim that people can lose £500 a minute on FOBTs (which she has tried to justify on Twitter by saying it could happen if you played two machines at once!), but this is chicken feed compared to the outrageous lies that are usually made by anti-FOBT campaigners. Mainly, however, she relies on an argument that is so overtly snobbish that I'm going to pretend it was made tongue-in-cheek and pass over it.

I cannot be so generous towards Iain Duncan Smith and his Centre for Social Justice, however. Last week he was talking through his hat at Conservative Home under the headline 'Problem gambling is soaring – it’s time to cap the stakes on betting machines'. In the article, he claims:

The British Gambling Prevalence Survey identified 450,000 problem gamblers in 2010 with an average debt of £17,500. The number of problem gamblers has now increased to 593,000 in 2015.

IDS was promoting a report from the Centre for Social Justice, which begins by claiming...

In 2006, the Centre for Social Justice (CSJ) published its Breakthrough Britain report on addiction, which outlined the systemic issue of gambling and its pervasive effects on individuals and communities. At the time of publication, there were 250,000 problem gamblers in the UK, whereas estimates now exceed 593,000. [My italics]

Note the word 'exceed' there. The report cites the NHS Choices website as its source, but when you visit it you will see that it says:

There may be as many as 593,000 problem gamblers in Great Britain.[My italics]

Straight away, we have a problem. The NHS website says that 593,000 is the upper limit whereas as CSJ says it is the lower limit and IDS says it is the exact figure.

So who is right? The answer is none of them, although the NHS is slightly closely to the truth. The 593,000 figure is the upper limit of the higher of two estimates from the 2010 British Gambling Prevalence Survey, which reported:

This [problem gambling percentage estimate] equates to somewhere between 342,000 and 593,000 adults according to the DSM-IV and between 254,000 and 507,000 adults according to the PGSI.

That was seven years ago and there have been several more surveys since then, but none has reported an estimate as high as that found in the 2010 report. As I explained a few months ago:

The first three reports in 1999, 2007 and 2010 used two different methodologies and came up with the following estimates:

1999: 0.6 per cent (DSM-IV), N/A (PGSI)

2007: 0.6 per cent (DSM-IV), 0.6 per cent (PGSI)

2010: 0.9 per cent (DSM-IV), 0.7 per cent  (PGSI)

Responsibility for collecting the data was then handed to public health bodies who came up with the following estimate for England and Scotland (combined) for 2012:

2012: 0.5 per cent (DSM-IV), 0.4 per cent  (PGSI)

Since 2013, the figures have been collected by the Gambling Commission which only uses the PGSI methodology. Results are as follows:

2013: 0.5 per cent

2014: 0.5 per cent

2015: 0.5 per cent

2016: 0.7 per cent

Quite clearly, the estimates fall within a narrow range of 0.4% to 0.9% and have not risen over time. It is difficult to explain why the Centre for Social Justice would focus on figures from seven years ago and present them as current. A cynic might say that they are deliberately misleading the reader in their pursuit of a narrative that is not supported by the facts. Naughty cynics.

CSJ's Breakthrough Britain report on gambling (which was actually published in 2007, not 2006) is no longer online and I don't have a copy of it so I don't know whether it claimed that there were 250,000 problem gamblers. However, I am familiar with all the possible sources and the 250,000 figure seems to be plucked out of thin air. The British Gambling Prevalence Survey of 2007 reported:

Taking into account the 95% confidence intervals around the prevalence estimates, one can conclude that the number of adult problem gamblers in Britain is somewhere between 236,500 and 378,000 according to the DSM IV, and 189,000 and 378,000 according to the PGSI.

250,000 fits within these confidence intervals, albeit towards the lower end. It seems that CSJ is happy to take a (roughly) mid-point estimate for 2007 while taking the top end of the highest estimate for 2010 - and then claiming that the real figure exceeds even that!

At the very least, this is extremely sloppy. Now let's return to what IDS claimed at Conservative Home, because he uses a different number:

The British Gambling Prevalence Survey identified 450,000 problem gamblers in 2010 with an average debt of £17,500. The number of problem gamblers has now increased to 593,000 in 2015.

Where does the 450,000 figure come from? It is a mid-point estimate from the 2010 survey.

Where does the 593,000 figure come from? It is the top-end estimate from the 2010 survey.

In other words, IDS has taken two figures from the same estimate in the same survey in the same year and pretended that one is from 2010 and the other is from 2015! This is a squalid misuse of statistics.

In fact, the estimate of problem gambling prevalence in 2015 was 0.5 per cent, considerably lower than the 0.9 per cent IDS is relying on for 2010.

Neither IDS nor CSJ have bothered to look at any of the figures since 2010. The Gambling Commission's most recent mid-point estimate of the number of problem gamblers is 320,000. This is in line with estimates published back in 2000 when the population was smaller (and there were no FOBTs in the UK):

The likely number of problem gamblers in Britain is thus 370,000 according to the SOGS, and 275,000 according to the DSM-IV.

Of the 11 estimates made since 1999, the 0.9 per cent figure is the only one that exceeds 0.7 per cent. It is plainly an outlier, but if you took it as gospel you would have to conclude that problem gambling prevalence has fallen significantly since 2010.

A more reasonable interpretation is that problem gambling rates have been essentially static for as long as anyone in Britain has been measuring them, totally unaffected by FOBTs, advertising, casino deregulation and all the other bogeymen of anti-gambling crusaders.

These statistics are not difficult to find so can people please stop lying about them?

No comments: