Last week, Andrew Rhodes, CEO of the Gambling Commission, wrote an open letter regarding what he describes as ‘the misuse of statistics’. As someone who has worked with data and research all my adult life, the messages I took from it made for troubling reading. Most concerning is, in his letter, Mr Rhodes seems to reference three practises as though they were interchangeable (although he doesn’t cite specific examples): an interpretation of the data, data dredging, and confirmation bias. The first is paramount to scientific progression and curiosity – the other two are dangerous.
Consensus in science and data is a very hazardous practice. It exists by implication that it’s more important to agree and concede than it is to critically question and dissect the information in front of you. Often data can be interpreted in a myriad of ways, and therefore, just because some disagrees with your interpretation of a dataset, as long as its well-evidenced and statistically significant, it does not automatically mean your analysis is ‘inaccurate or misleading’. Scientific vernacular has long been guided by this core principle; we don’t have ‘certainties’ in science, we have ‘theories’, and we don’t have ‘this clearly shows that’, we have ‘this suggests that’.
Interpreting data in a different way is not the same as some of the other practises that are running rampant in the debate on gambling harm. Data dredging is a disingenuous method of presenting data as though coincidences in your data set are definitively correlated. Data should never be accepted as straight reading – that’s why researchers spend weeks, often months, mining it (identifying patterns and variables). An example of this is that in the UK, sales of baked beans happen to match the trajectory of sales of Lady Gaga records, but this does not mean the two are meaningfully linked.
Though that example is slightly comical, it can happen in circumstances where the subject matters are more compatible, but that still doesn’t mean there is a correlation. Gambling with Lives state in a graphic that there are “Up to 496 gambling suicides in England every year.” It is rather difficult to ascertain and pinpoint the exact cause of a suicide – hence why coroners are highly reluctant to classify suicide by motive (and why that figure is therefore based on anecdotal evidence rather than number of coroner reports, which for 2022 was 0). Even if it was the case that suicides happen to be higher in the gambling population than in the general population, as the Gambling with Lives figure tries to imply, until you have the smoking gun that links the two, you cannot demand correlation. Cause is often complex, and can happen for a variety of reasons. Indeed, specifically on the case of suicide, the Samaritans clearly state in their media guidance for reporting suicide that “Speculation about the ‘trigger’ or cause of a suicide can oversimplify the issue and should be avoided. Suicide is extremely complex and most of the time there is no single event or factor that leads someone to take their own life.” Usually, the motivation for researchers to data dredge is often to privilege a hypothesis they already have in their head ie they fit the data round the hypothesis, not the other way round. This leads me onto my next point: confirmation bias.
Research into gambling harm is often heavily biased before its even begun. Research examining the prevalence of gambling harm, conducted by those have previously extensively researched gambling harm, means the starting point for the investigation is that gambling is a fundamentally harmful activity. An example of this is that to apply for a research grant from the Gambling Commission/Gambling Research Exchange Ontario suicide research pot, you are required to involve those with lived experience of harm in the design and execution of the research, and cannot have undertaken any work with the gambling industry in the last five years, although none of this applies to researches who are part of an activist lobby group. This is why I have long argued that clinicians who specialise in addiction recovery should be undertaking the research: they are neither pro-gambling nor anti-gambling – their only interest is to help identify, treat, and manage addictions. In true scientific pursuits, you should have only a vague idea as to what your data is going to throw up,
such is the unpredictable nature of most things in the world. In gambling harm research, the monoculture of those involved often means that it’s possible to arrive at conclusions first, and then find the data to fit it.
It is therefore important when we are talking about misinformation that we do so fully aware that an interpretation is not automatically a misconception, and that to dare to question a dataset does not make you a peddler of misleading claims.
The GCF’s interpretation of official statistics:
Note – The GCF have already accepted that our 99.8% was an error in phrasing – we applied that figure to the general population, rather than the gambling population. We do not, however, accept we were being misleading with this statement, since the number of gamblers who suffer harm remains the same, regardless of the framing of population.
Andrew Rhodes states in his open letter:
“The same Health Survey for England 2021 indicated that 1.2 percent of people who gambled in the last 12 months were classed as being at a moderate risk of experiencing problems with their gambling leading to some negative consequences, with a further 4 percent being at low risk (gamblers who experience a low level of problems with their gambling with few or no identified negative consequences).”
Gambling disorders are the only cognitive condition I know of where those who have 1) self-diagnosed on a computer survey at home and 2) have been found to have no detectable signs of clinical harm ie ‘low risk’, are included in the statistic of harm.
Likewise, gambling disorders are the only health condition I know of where there is a ‘harm statistic’ where that harm is completely undefined, as the CEO of the Gambling Commission Andrew Rhodes himself acknowledges. Addiction is a cognitive and clinical condition, but it is not being treated as such. Accurate clinical diagnoses are the bedrock of genuine addicts being able to access proper treatment, and that limited treatment being given to those who are experiencing clinical harm. I am concerned The Gambling Commission seem to be endorsing an entirely non-clinical approach that essentially says ‘I am what a survey tells me I am’. In all other medical conditions, in order to ‘have’ something, you have to meet the clinical criteria.
The approach the Gambling Commission have taken shouldn’t be entirely unsurprising, given that the entire basis of affordability checks is entirely non-clinical. It wrongly assumes that big losses are somehow a symptom of addiction, in a way that would be ludicrous if we applied it to another addiction. If someone spends £500 on a bottle of champagne, without any other information provided, do we consider them to be an alcoholic? On the flip side, do we consider an individual who buys 90p cider and drinks 40 units a week not to be an alcoholic?
One of the most effective addiction recovery initiatives is Alcoholics Anonymous, and that is, in part, because of the clinical manner in which alcoholism is discussed. This is from the UK Addiction Treatment Centres.
“An illness of the brain, alcoholic addiction can be one of the most difficult addictions to recognise within yourself…there is no overnight miracle cure. However, the addiction can be effectively treated at alcohol rehab through a comprehensive recovery programme.”
Here, there is an understanding of the essence of addiction – the cognitive compulsion. Inevitably, there are people with dopamine deficits in their brain that make them susceptible to addiction who have yet to be diagnosed, or who are on their way to experiencing the negative consequences, but those people could and should be identified by harm markers, if they are written from a clinician basis. Those who don’t meet that clinical threshold, put simply, do not need treatment. This is nothing new or niche or controversial – this is way it’s always been. We do not, and should not, medicate for ‘possibility cases’. Therefore, every measure we use as a means of protection from potential harm, including affordability checks, must be guided by robust, clinical harm markers and nothing else.
The biggest take away from Rhodes’ letter is that ordinary gamblers are not being heard in the debate on gambling regulation, but also, that the addiction recovery clinicians are not being heard in the debate on gambling harm.