The costs (but not the benefits) of gambling

A one-sided review tells us nothing

As the crusade against gambling gathers pace in the UK it is becoming clear that there are two tactics being used to make it ‘the new tobacco’.

The first is to treat it as a ‘public health’ issue. ‘Public health’ activists/academics are always keen to find (a) new dragons to slay and (b) fresh pots of grant money. Once gambling is seen as a public health issue, the door is open to the kind of regulation that smokers, drinkers and, increasingly, eaters have been treated to over the years.

Unfortunately, ‘public health’ academics know even less about gambling than they do about the other issues upon which they pontificate. The articles they have started writing for medical journals in recent years spout the usual empty rhetoric about the ‘whole systems approach’ and dust down policies from tobacco control that don’t make any sense in the context of gambling. You can’t really put a sin tax on gambling, for instance, because people chose their own stakes and you certainly can’t put gambling in plain packaging, although that hasn’t stopped ‘public health’ chancers from suggesting it.

The second tactic is to pathologise and demonise gambling in general, rather than focus on problem gambling. This allows the ‘public health’ lobby to settle into the comfort of the ‘no safe level’ rhetoric it uses with tobacco and alcohol while justifying policies based on the ludicrous whole population approach. Rather than treating pathological gambling as a psychological disorder that affects a small number of people, gambling per se is to be treated as unhealthy.

Public Health England was all set to drive forward with this strategy. An early draft of a PHE press release in 2021 said:

‘Health leaders from organisations such as X and X are joining PHE in calling for a public health approach to gambling focusing on prevention, early intervention and treatment. This approach is similar to how we tackle tobacco consumption or unhealthy food consumption, and would require cross-government support’.¹

PHE was closed down for being incompetent shortly afterwards, but before it imploded it put on a series of reports on gambling, and its successor, the Office for Health Improvement and Disparities (OHID), has now produced some updates of them. Careful readers of the summary report will notice that the words ‘gambling’ and ‘gambler’ are often used when it should say ‘problem gambling’ and ‘problem gambler’. This is not an accident.

For example, the word ‘problem’ should prefix every word in bold below…

This suggests that gambling may trigger suicidal events in some people already prone to suicidal ideation. The link between gambling and suicide and self-harm was supported by qualitative studies.

In quantitative studies, anxiety and depression were the most commonly measured mental health disorders. Results were mixed in terms of showing whether gambling caused these outcomes. In qualitative studies, gamblers experienced emotions such as guilt, shame, loss of self-esteem, loneliness and sleep problems and neglected caring properly for themselves. Close associates of gamblers reported negative emotional, psychological and health impacts. These included anxiety, depression and sleep problems.

Studies also reported mixed findings on the link between gambling and various measures of alcohol, smoking and drug use. In qualitative studies, gamblers reported co-occurring alcohol and drug-related problems.

The qualitative studies described that adult gamblers had lost jobs, were demoted or resigned due to gambling. Gambling was linked to loss of concentration on work activities, showing up late, not turning up for work or turning up after no sleep.

Close associates of gamblers also reported their work performance being affected, and work colleagues and employers also suffered.

None of this is true of your average gambler. It only applies to (some) problem gamblers - and they only make up around 0.5% of the population. Ordinary gamblers tend to be well-adjusted, happy people, as OHID is forced to admit during a bit of throat-clearing towards the beginning of the summary:

The highest levels of gambling participation are reported by people who have better general psychological health and higher life satisfaction. And people who have poorer psychological health are less likely to report gambling participation.

The cornerstone of the report is a ‘cost of illness’ (COI) study which estimates that gambling costs British society between £1 billion and £1.8 billion a year. COI studies are often used to draw attention to the negative consequences of smoking, drinking and obesity. They are like a cost-benefit analysis but without any benefits and with costs that are often intangible, i.e. they are not financial costs to anyone. When you hear that alcohol costs England £21 billion a year or smoking costs Britain £13 billion, you have one of these studies to thank.

These studies have no merit as economic research. They are purely driven by advocacy. The hope is that the average person will wrongly assume that the costs are to taxpayers and agitate for change. In the case of gambling, a COI is particularly useful since it helps establish the issue as a matter of public health.

The bulk of the cost in the PHE/OHID report comes from ‘wider societal intangible costs’, but it doesn’t include the higher life satisfaction of ordinary gamblers as an intangible benefit, nor any economic benefits from the gambling industry, because benefits are explicitly ignored, as the minutes of a Public Health England meeting in 2019 confirm:

Benefits are well promoted and you do not need to include this in the analysis as not relevant to research question.²

If you ignore all the benefits, everything looks like a cost to society. The PHE report claimed that gambling cost the UK £1.27 billion a year. Nearly half of this came from intangible costs of gambling-related suicide based on an estimate of the number of gambling-related suicides that has no credibility.

I suspect that PHE were disappointed not to get a bigger number. The Gambling Commission were certainly disappointed. In a document from late 2021 released under the Freedom of Information Act, they said:

Given the (mostly acknowledged) limitations, it’s surprising that the £1.27bn cost of the gambling industry has been promoted by PHE as one of the main findings from the entire review. This may be due to its status as being one of the few conclusions that is ‘new’, a desire to emphasise the need for further research to strengthen the evidence base or intending to ensure gambling is considered as a public health issue.

However, the lack of data and underestimated costs mean that the estimated costs in this review are significantly below even the tax receipts from the gambling sector in 2017, which were estimated to be £2.7bn.³

This is a telling comment. The first rule of making a ‘public health’ cost-of-illness estimate is that the total must be more than the government gets in tax, otherwise economists will (quite rightly) say that the costs are internalised in a Pigouvian fashion.

The OHID’s new report manages to increase that figure a bit if you focus on the upper limit (as campaigners inevitably will), although £1.8 billion is still well short of what the state makes in gambling revenue. The main difference between the OHID report and the now-deleted PHE report is that they’ve come up with a bigger number for the cost of gambling-related depression and have changed the figure for gambling-related suicide from £619 million to an estimate ranging from £241 million to £962 million. Without these costs, which are neither financial nor external, the ‘costs’ of gambling are pitifully small, less than £300 million a year. This is less than a tenth of gambling duty revenue.

But coroners hardly ever declare a suicide to be gambling-related, and gambling-related depression goes largely unnoticed, so how can you even guess what the true figures are? Well, you can’t really. The whole report is finger-in-the-air stuff.

People who have a gambling problem are more likely to smoke, drink heavily and take drugs. Or, to put it another way, people who drink heavily are more likely to smoke, gamble and take drugs. To put it yet another way, people who take drugs are more likely to smoke, drink heavily and gamble. Such people tend to be risk-takers with low impulse control. They are disproportionately young men.

It would be silly to claim that they smoke, drink and gamble because they take drugs. Equally, it would be daft to say that they smoke, gamble and take drugs because they drink heavily. These behaviours may be related, but you can’t take one of them and portray it as the cause of all the others. Young problem gamblers are also more likely to play video games, have less parental supervision, be hyperactive, have lower mental well-being and be involved in criminal activity but it would be an extraordinary reach to claim that problem gambling was the root cause of all this, rather than being a consequence of a lot of it or merely incidental.

And yet that is essentially what OHID assumes. If there is a statistical association between, for example, problem gambling and drug-taking - and there is - OHID assumes that the person would not have taken drugs had he not been a problem gambler. They do the same thing with alcohol dependency, homelessness, imprisonment, unemployment, depression and suicide. They look for an association with problem gambling in the academic literature and then assume complete causality.

For example, they find a study that reports a relative risk of illicit drug use of 1.95 (95% CI 1.06 and 3.61) for people aged 17 to 24 who are moderate-risk gamblers or problem gamblers (i.e. they are 95% more likely to take illicit drugs). Applying this to the whole population and making various questionable asssumptions, they conclude that…

1,312 people aged 16 to 24 use illicit opiates and/or crack cocaine associated with at-risk and problem gambling in England.

Although they use the mild and more accurate phrase ‘associated with’, what they mean is that 1,312 people aged 16 to 24 use illicit opiates and/or crack cocaine because they are problem gamblers.

The researchers behind the original studies attempt to control for confounding factors so that problem gambling is an independent variable. But that is very difficult. It’s not like adjusting for age. There are a lot of factors to adjust for and every adjustment creates more room for error. Some of the most important variables, such as personality type, are almost impossible to adjust for. Trying to isolate problem gambling as an independent variable is like trying to take the eggs out of a cake.

Even if you could adjust for everything perfectly, which way does causation run? It’s easy to see how problem gambling could lead to depression and heavy drinking, for example, but it’s also easy to see how depression and heavy drinking could lead to problem gambling. In the figures above, you could reverse the calculation and estimate how many people became problem gamblers because they are opiate/crack users. All it would take is a different assumption about causality.

As OHID admits, there isn’t enough evidence to assume a temporal relationship, let alone a causal one.

Most of the studies published on gambling and harm do not allow us to determine that gambling came before the harm.

But drugs are not OHID’s priority in this report so they take the 1,312 figure, estimate that 628 of them get NHS treatment and work out what that costs the taxpayer.

As it happens, the total (£1.8 million) is not very much in the great scheme of things, but they do the same thing with other health problems and arrive at much bigger figures.

Suicide and depression make up between 72% and 83% of the total ‘societal cost’ (depending on whether you use the low or high end of the estimate), so let’s focus on those.

I have written about the suicide figure before. It is based on the assumption that there are 409 gambling-related suicides per year in the UK. PHE/OHID then estimated how many years of life are lost, adjusted for quality of life and multiplied it by £70,000, which is the standard value of a quality-adjusted life year used by the Treasury.

The figure of 409 deaths is extrapolated from a Swedish study.

The study looked at 2,099 people who had been diagnosed with gambling disorder by a doctor while receiving inpatient or outpatient care (but not primary care) in the Swedish health system between 2005 and 2016. Of these 2,099 individuals, 67 subsequently died, including 21 who committed suicide. This suicide rate implied that these people were 15 times more likely to kill themselves than members of the general population.

PHE simply assumed that the suicide rate was the same among Britain’s problem gamblers as it was among these Swedish patients, but that is untenable. It is very likely that people diagnosed with gambling disorder in hospital are on the severe end of the spectrum. The authors of the Swedish study say this explicitly:

‘It is therefore likely that results may be skewed toward a population of individuals with more severe forms of GD [gambling disorder].’

You only have to look at the characteristics of the patients to see that they were rather unusual. As I said in the previous post…

51% of the Swedish patients were suffering from depression. 60% had an anxiety disorder. 41% had a substance-use disorder. 29% had an alcohol-use disorder. 12% were bipolar. 19% had a personality disorder. This was not a normal group of people, even by the standards of problem gamblers.

The OHID partially acknowledges that the PHE methodology was suspect, but since the only other study looking at the issue was even shakier, they decided to stick with it but use a high- and low-end estimate based on the severity of the gambling problem. The upper estimate uses the same measure of problem gambling that PHE used, i.e. people scoring 8 or more points (out of a possible 27) in this test. The lower estimate only includes people who scored 5 or more (out of a possible 10) in this test.

These tests are the PGSI and DSM-IV surveys respectively. They are the two main instruments used to diagnose gambling disorder internationally. A score of 8 or more in the PGSI indicates the sub-clinical classification of ‘problem gambling’ while a score of 5 or more in the DSM-IV indicates gambling disorder, which is a recognised psychiatric condition.

According to the designers of the PGSI, a ‘problem gambler’ classification describes “those who have experienced adverse consequences from their gambling, and may have lost control of their behaviour”. It does not denote psychiatric disorder. Gambling disorder on the other hand is defined as “persistent and recurrent problematic gambling behavior leading to clinically significant impairment or distress.” In short, gambling disorder typically describes greater dysregulation and harm than problem gambling.

Under the DSM-5 (which has a slightly different scale to DSM-IV), gambling disorder is sub-divided into three further categories.

Mild: 4–5 criteria met.
Moderate: 6–7 criteria met.
Severe: 8–9 criteria met.

A score of 5 under DSM-IV would therefore indicate a mild disorder under DSM-5.

Is OHID implying that someone who scores 5 or more on the DSM-IV test is in the same boat as the patients in the Swedish study? The Swedish cohort remains very different from people in the general population who would be classified as having a gambling disorder.

The change in methodology seems to have been in response to criticism of the PHE report, but they haven’t addressed the root problem. OHID even goes so far as to suggest that the Swedish patients could be less likely to kill themselves than the average problem gambler…

It could be argued people receiving treatment are less likely to be suicidal due to the care they are receiving, but it could also be argued that these are more severe cases so are more likely to result in suicide. It is not possible to say with confidence either way. 

On the contrary, I think we can say with some confidence that people who have gambling problems so severe that they seek treatment are more likely to be suicidal than those who don’t. Only around 0.2% of England’s problem gamblers have a hospital admission related to a gambling disorder diagnosis.

We only read about the most severe cases in the newspapers so it is easy to forget that most problem gamblers are young men who basically grow out of it. They might lose more than they can afford but for most of them there is no lasting harm and they don’t require professional help.

By contrast, the people who seek help have very high scores on the tests. In 2013, the NHS London Clinic reported a mean PGSI score for patients entering treatment of 19.8 (out of 27). The National Gambling Treatment Service reports a mean PGSI score (on entering treatment) of 19. The average PGSI score for GamCare patients is 19.4.

But you only need to score 8 to be classified as a problem gambler. Common sense tells us that the kind of people diagnosed in the Swedish study were more at risk of suicide than the milder cases, regardless of treatment. In any case, the study doesn’t say the patients received treatment for gambling disorder! For half of them, gambling disorder wasn’t even their primary diagnosis.

What about depression, which OHID reckons creates £508 million? Unlike the suicide ‘costs’, some of this cost is external, namely £114.2 million to the NHS service.

OHID scoured the literature and found several studies looking at problem gambling and depression, such as this one…

Results showed that gambling was associated with increased odds of major depressive disorder (adjusted odds ratio (AOR) 1.98, 95% CI 1.14 and 3.44)

Notice that OHID once again forgets to put the word ‘problem’ before ‘gambling’.

In the end they settled for using this study, which happens to come up with the largest relative risk for depression: 3.08 (95% CI 2.60 to 3.65), meaning that problem gamblers are around three times more likely to have depression.

Assuming, as always, that causality only runs one way, i.e. that the depression was caused by the problem gambling, OHID estimate that 69,099 people in the UK have depression because of their problem gambling. A value is put on their lost quality of life and the cost of treatment and prescription drugs and at a figure of £508 million is arrived at.

Most of this is what calls OHID calls an ‘intangible societal cost’. More accurately, it could be described as an intangible internal (or personal) cost. It does not cost the rest of society anything and should properly be seen as part of a trade off between risk and pleasure. Having acknowledged that non-problematic gamblers have ‘better general psychological health and higher life satisfaction’, OHID could have estimated how many cases of depression are prevented by gambling. They should have done this. Even in COI studies, it is normal to include health benefits.

What is the point of an estimate of intangible costs if you don’t know the intangible benefits? What is the point of making estimates at all when you don’t have enough data for them to be credible? Is it simply that - as Marguerite Regan, who led the PHE review, said - “more research and evidence are needed to support advocacy and action”?

I think that is the key. This is policy-based evidence designed to spur action from the government and yet the report has no meaningful lessons for policy. There is clearly insufficient data to estimate how many suicides are caused by problem gambling. OHID’s cost estimate for suicide is so wide as to be useless and could still be wrong. The studies OHID used to make their estimates are riddled with residual confounding and there are massive questions about the direction of causality.

The lessson is that if you don’t have the data, don’t make an estimate. A bad estimate is not better than no estimate. The OHID report is basically agenda-driven guesswork. It doesn’t measure anything and it doesn’t tell us anything useful. All we can conclude is that problem gambling leads, in some cases, to cases of depression and occasionally suicide, but we have no idea how many. It probably also leads to cases of alcoholism, drug use and homelessness, but it is impossible to put even a rough figure on any of this.

We already knew that there are negative consequences from problem gambling and the government earns ample money from gambling duty to provide world class treatment to anyone who asks for it. Conjuring up a spurious cost estimate might be useful to anti-gambling campaigners but it is useful for nothing else.

1. Document released under the Freedom of Information Act to Dan Waugh.

2. Ditto.

3. Ditto.