Has problem gambling amongst children quadrupled?

The Gambling Commission published its annual figures on gambling amongst 11-16 year olds today. The media have understandably picked up on one particular statistic.

From the BBC...

Number of child gamblers quadruples in just two years

The number of children classed as having a gambling problem has quadrupled to more than 50,000 in just two years, a report has claimed.

Anti-gambling campaigners have been quick to blame this apparent rise on advertising and online gaming. However, as I explained in a post earlier this year, participation in online gambling is almost non-existent in this age group - not surprising given how difficult it is for a child to set up an account. Insofar as 11-16 year olds use the internet for gambling, it is mostly to buy lottery tickets with their parents' consent.

The most common forms of gambling by children are private bets with friends, wagers on private card games, Category D amusement games (with stake limits of 10p) and the National Lottery. Everything else has a prevalence rate of one per cent or less.

Nevertheless, a quadrupling of problem gambling - from 0.4 per cent to 1.7 per cent - is striking. So striking that it almost defies belief. It is rare for anything to quadruple in two years without a change in methodology. Sure enough, the methodology of the survey that supplies the statistics has been changed significantly.

You have to read the technical report to see what's happened and that report isn't easy to find (I had to ask the Gambling Commission for it). Strangely, it does not mention one of the most important methodological changes which is that the 2016 figure (0.4%) was based on 11-15 year olds whereas the 2018 figure (1.7%) is based on 11-16 year olds.

This makes a big difference because 16 year olds have the highest rate, as you can see by comparing the two charts below.

2016...

2018...

The Gambling Commission may have decided to not bother mentioning this because the switch to 11-16 year olds happened last year, not in 2018. The change probably explains why the rate appeared to rise from 0.4 per cent in 2016 to 0.9 per cent in 2017, but it cannot explain the apparent rise to 1.7 per cent this year.

But there have been other changes in 2018 which have led to a bigger number. Significantly, the 2018 survey is the first in which the children were given the option of completing the survey online. Previously it had been a written survey only, but this year the schools which participated were given a choice between paper and online - and 70 per cent opted for online.

This doesn't sound like it should make much difference but it does. As the report explains...

One of the key metrics captured in the Gambling Commission’s study is the proportion of young people who are ‘problem’ or ‘at risk’ gamblers, as measured by the DSM-MR-IV-J, a screen consisting of 9 domains (see Chapter 3 for more about the screen). Responses across these domains are aggregated to form an overall score; respondents scoring 4 or more are classified as ‘problem gamblers’ and those scoring 2-3 are ‘at risk’ gamblers. The problem gambling rate among those answering on paper is in line with rates in 2017, while the aggregate 2018 rate (including the online and paper samples) is higher than 2017.

Only those who gamble are given the problem gambling questions. This year, many more children were given these questions because gamblers were screened on the basis of whether they had gambled in the past year rather than - as before - in the past week. The rate of problem gambling is obviously higher among gamblers than among the entire cohort and, again, the online survey produced much higher numbers than the paper survey.

.. the proportion of those screened who were classified as problem gamblers was higher among those responding online (5.6%) than on paper (1.6%), despite similar screening rates (35% and 34%, respectively). This is because online respondents were more likely to indicate problematic behaviour at 7 of the 9 domains on the screen. As such, it appears that differences by mode as well as the improved screening rate are factors in the increased problem gambling rate seen in 2018.

It is not entirely clear why the online version elicits higher rates, nor is it clear which version produces the more accurate responses. The Gambling Commission says that 'respondents are more likely to give responses indicating problematic or socially undesirable behaviour when answering a survey online rather than on paper' and cites this study as evidence.

Whatever the reason, the results from the 2018 survey are simply not comparable to the 2017 survey, let alone the 2016 survey. The 2016 survey was paper only and used a younger cohort.

The 2017 survey had the same age group but it was all paper whereas this year's study was mostly online. The difference this makes becomes crystal clear on page 10 of the technical report. If you compare the paper responses in 2018 to the paper responses in 2017, the rate of problem gambling has actually declined! I've circled the relevant figures below - click to enlarge. The apparent rise is due solely to the online responses.

It is fair to say that the Gambling Commission could have done more to flag up the apples and oranges problem here. They have a section in the technical report titled 'The impact of moving the survey online' but the technical report is not available from their webpage and is only briefly referenced on page 32 of the main report. I doubt any journalists would have read it.

Even if they had, few journalists would pass up the chance to report a quadrupling of problem gambling in the space of just two years. It must have seemed almost too good to be true. As it happens, it isn't true.