Slot machine
Much of the moral panic about gambling in recent years has centred on the claim that the number of problem gamblers has “increased by 50% in three years” and that the UK has 450,000 gambling addicts.

These statistics were always very questionable. Three editions of the British Gambling Prevalence Survey (BGPS), published in 1999, 2007 and 2010 offered only ambiguous evidence of any increase. The surveys used two methodologies, the DSM-IV test and the PGSI test. The 2010 survey found a non-significant increase under the first and a borderline significant change under the second (compared to 2007). As the Department for Culture, Media and Sport notes:

“The proportions increased from 0.5% of the adult population in 2007 to 0.7% in 2010 (which is not statistically significant) on one measure and from 0.6% in 2007 to 0.9% in 2010 (which is at the margins of statistical significance) on the other measure used.” (DCMS, 2013: 29)

Unsurprisingly, campaigners focus on the positive finding and proceed to make post hoc ergo propter hoc assumptions about the cause of this putative rise in problem gambling, focusing on fixed-odds betting terminals (FOBTs). Rowenna Davis, for example, wrote in the Daily Mail: “According to a Gambling Association [sic] survey, problem gambling has increased 50 per cent since liberalisation was introduced in 2005.” Similarly, an anonymous source told the BBC’s Panorama programme: “After FOBT machines were introduced to betting shops, the number of problem gamblers mushroomed” (Panorama, 2012).

In fact, the only conclusion drawn with any confidence in the last British Gambling Prevalence Survey was that there were between 254,900 and 593,400 problem gamblers in the UK in 2010. There is a wide gap between the low and high ends of this estimate and it is uncertain whether the true figure has risen since 2007, let alone whether it has risen since 2010.

Responsibility for collecting problem gambling data has since been passed to public health authorities and are collected separately for England, Scotland and Wales. I mentioned last year that the Scottish data showed no rise in problem gambling between 2010 and 2012. Now we have the English data. Better still, we have the English and Scottish data combined in this report.

The results strongly suggest that (a) there are far fewer than 450,000 problem gamblers in Britain and (b) strongly suggest that there has not been a rise in problem gambling since 1999.

Under the DSM-IV measure, problem gambling prevalence is estimated to be 0.5 per cent. Under the PGSI measure, it is estimated to be 0.4 per cent. This compares to 0.9 per cent and 0.6 per cent in the 2010 British Gambling Prevalence Survey.

For some reason, the new report doesn’t give confidence intervals, but if we compare the mid-point estimates from the four surveys, this is the trend under the DSM methodology for England and Scotland:

1999: 0.6%

2007: 0.6%

2010: 0.9%

2012: 0.5% 

And this is how it looks under the PGSI methodology (which wasn’t used in the 1999 report):

2007: 0.6%

2010: 0.7%

2012: 0.4%

A few caveats need to be mentioned. Firstly, the latest survey doesn’t include Wales and so the authors remove Wales from previous surveys in order to make comparisons. This makes no difference except to the PGSI figure for 2010 which moves from 0.6 per cent to 0.7 per cent.

Secondly, as mentioned, we don’t have the confidence intervals. Confidence intervals tend to be quite wide in these studies. The mid-point estimates can act as a guide but should not be taken to be precise figures. There could be some changes going on below the limit of detection. However, there is nothing in the data to imply that problem gambling rates are higher now than in 1999. Indeed, if we were as literal-minded as the anti-gambling campaigners and looked only at the mid-point estimates, we would conclude that there has been a decline.

Thirdly, the most recent survey was carried out by a different organisation to that which carried out the previous three. Although the authors say comparisons should be “made with caution” as a result, they assure us that “the methods and questions used in each survey were the same” and they conclude that “problem gambling rates in Britain appear to be relatively stable” (ie. since 1999).

In short, the idea that problem gambling has increased in the last decade—let alone increased by 50 per cent—is looking weaker than ever. Like so many anti-gambling arguments, it seems to be a convenient myth.

Chris Snowdon Final

Head of Lifestyle Economics, IEA

Christopher Snowdon is the Head of Lifestyle Economics at the IEA. He is the author of The Art of Suppression, The Spirit Level Delusion and Velvet Glove; Iron Fist. His work focuses on pleasure, prohibition and dodgy statistics. He has authored a number of publications including Sock Puppets, Euro Puppets, The Proof of the Pudding, The Crack Cocaine of Gambling and Free Market Solutions in Health.

1 thought on “New data show no rise in problem gambling since 1999”

  1. Posted 01/05/2014 at 16:20 | Permalink

    Thanks for this but if you can let me examine post hoc ergo propter hoc a little more.

    The DSM-IV screen used in all these studies showed no rise at all 1999-2007 whereas the alleged crack of FOBTs were introduced in 2001 not as a result of legislation but of product innovation. By 2007 there were 27,500 of them up and running when the same screen done by the same people using the same sampling technique showed no change. The formalisation of the machines legal position in 2007 made little or no difference to their existence or operation as they were operating under an industry code that was incorporated to the legislation already, much as the new ABB code is going to be incorporated shortly.

    The potential blip, at the margins of significance in the period 2007-10 corresponds with the financial crash, falling real wages and an increase in unemployment of over 1m.

    A reasonable hypothesis would be that the screens which naturally include questions on the financial impact of gambling can be affected by changed financial circumstances. The adaptive theory in terms of gambling behaviour would suggest that worsened financial circumstances should lead to more meeting the problem gambling criteria and that over time people adapt their behaviour to the changed circumstances. I know this hypothesis is post hoc too but the theory matches the outcomes much better, is the result I predicted for the 2010 research and has a clear rationale. If this theory is right then the measures look to be solid, able to detect relatively small changes over time and the hypothesis is predictive for all the results over all the survey years.

    Essentially problem gambling is relatively rare and relatively stable, we can chuck FOBTs at gamblers – £1.5bn in bookies £300m in bingo halls, £150m in casinos, add in the Internet, now yielding about £2.4bn a year in the UK and despite this huge technological change in how people gamble since 1999, and the increased availability of gambling, the problem gambling rate remains low and stable. Exactly what you would expect of a pathological condition which needs help support and treatment.

Comments are closed.