Galleries, gambling and self-harm
On the perils of mistaking cause and correlation
Last week, a British think tank was forced to admit that it had based proposals for gambling tax increases on a misreading of Government figures. The episode highlights once again, the need for greater rigour in the use of evidence in policy-making. More significantly, it raises a question about the point of state bodies producing cost estimates of unclear meaning. In this article, we examine the problems with mistaking correlation for causation where health harms are concerned.
During the last week in July, the Social Market Foundation (‘SMF’) published proposals for swingeing tax increases on betting shops and online gambling websites in Great Britain. It justified these hikes (of variously 66% and 138%) by stating that social and economic costs of gambling may be as high as £7.2bn a year - upweighting figures produced by the Office for Health Improvement and Disparities (‘OHID’) in 2023. Following an investigation by the Office for Statistics Regulation, the SMF has now admitted that this claim was based on a misreading of the OHID report. Its mistake was to assume that the OHID’s estimate of costs associated with gambling harms was the same as costs caused by gambling. This, as the SMF seems to accept, is more than a semantic difference.
The significance of the SMF’s error, is made clear by the OHID’s treatment of suicide mortality, which constitutes 54% of the cost calculations (£962m a year for OHID; £4.8bn a year for the SMF). In its 2023 report, the OHID claimed that between 117 and 496 suicides in England each year are associated with harmful gambling (SMF: 585 to 2,480). These figures rely on a 2018 study by researchers at Lund University, who found that hospital patients in Sweden with a gambling disorder diagnosis were, on average, 15.1 times as likely to die by suicide as the general population. OHID incorrectly assumed that ‘problem gamblers’ (a sub-clinical classification and not the same as gambling disorder) in England are at similar risk; and ignored warnings from the Lund researchers about the generalisability of the findings. In subsequent analyses (Karlsson et al., 2025), the Swedish researchers revealed that gambling disorder was not in fact, a statistically significant risk factor for suicide, once comorbid health problems and economic status had been controlled for (see A Very Public Deception – Cieo for a more detailed examination of the myriad problems with the OHID report).
In response to the more nuanced findings from the team at Lund, the OHID has claimed, unconvincingly, that it never intended to suggest a causal relationship between gambling and suicide. This raises a question about what exactly the OHID’s mortality figures (and the wider cost estimates) do show. It is clear from the Swedish research that they can tell us nothing meaningful about the harmfulness of gambling - or indeed very much about anything. While the elision of correlation and causation is clearly tempting for those with an a priori indisposition towards gambling, it leads to difficulties, as findings from the Gambling Survey for Great Britain (‘GSGB’) illustrate.
The GSGB collects data in relation to a wide range of socio-demographic characteristics and lifestyle choices. One previously unpublished finding from the survey is that people who visit museums and art galleries are more likely to experience suicidal thoughts (13.5%) compared to people who do not (11.1%); and almost twice as likely to report a past-year suicide attempt (1.34% for museum-goers compared to 0.75% for others). According to the GSGB, visiting art galleries is more strongly associated with self-harm than gambling is. Those unwise enough to follow the OHID methodology, might therefore conclude that around 62,000 suicide attempts in Britain each year are associated with visits to museums and galleries – or claim that there is a cost to society of £1.2bn a year arising from ‘museum-related suicides’. Such figures would, however, be about as meaningful as observations about the sale of ice-cream and death by drowning.
Other, less trivial examples highlight the same issue. GSGB data reveal elevated risks of suicidal thoughts and attempts for some non-white ethnic groups as well as for non-heterosexual people. These data ought to be a cause for concern and investigation (indeed, there is a rich research literature on similar findings) but it would be morally repugnant to use them to claim an economic burden on society associated with ethnicity or sexual orientation.
Some have attempted to argue that because we know that excessive or disordered gambling can be a contributory factor to self-harm, it is acceptable to produce spurious statistics. In 2022, the chair of the Gambling Commission’s Advisory Board for Safer Gambling advised OHID that criticism of its cost figures was valid – but then intimated that accuracy in such matters was beside the point and that scrutiny of public health mortality claims was “a distraction from what matters to people and families harmed by gambling”.
The SMF upweighted the OHID figures by a factor of five to reflect the fact that the GSGB (which has itself been beset by controversy) provides substantially higher estimates of ‘problem gambling’ than seen in NHS surveys. The SMF therefore assumes that around 2,500 suicides in England each year (or about 50% of all such deaths) are caused entirely and exclusively by ‘problem gambling’, with no other contributory factors; but this flatly contradicted by the Swedish research that these calculations rely upon. If the Swedish data provide a reliable guide to gambling harms in England, then the SMF and OHID estimates cannot hold; if they do not, then the estimates crumble into dust.
That gambling gives rise to social and economic costs is not in dispute - nor is that fact that excessive gambling may contribute to self-harm (something that deserves more intelligent and serious discussion than it is usually afforded). Gambling also confers benefits on consumers, including sociability, entertainment and mental stimulation. Binde (2013) likened the process of betting on horseraces (including study of the form and bet selection) to the task of completing a crossword puzzle; and several studies have indicated that playing bingo may reduce risk of dementia. Data from the GSGB, the NHS Health Surveys and the NHS Adult Psychiatric Morbidity Survey reveal that non-problem gamblers have on average, better mental health than non-gamblers and are at lower risk of suicidal behaviour. It would be speculative to assume a causal relationship here (or to assign a positive economic and social value on this basis) – but no more so than is the case with the SMF’s and OHID’s treatment of costs.
The SMF deserves credit for acknowledging its error – although its report has not yet been corrected and it remains to be seen whether a change in the ‘facts’ leads to a softening in its tax proposals. The real question raised by the episode is why the OHID – at considerable public expense – bothered to produce cost estimates that have no obvious meaning. Perhaps in recognition of this, the Office for Statistics Regulation has asked the OHID to make clearer the limitations of its calculations and to act where figures are misused. The extent to which the OHID does so may be an indication of its fitness to serve as the new commissioner of gambling harm prevention.