Monday, 5 December 2016

Problem gambling: doubling and doubling?

Consider these three claims:

In 2004, it was predicted that the number of problem gamblers in the UK would double to 750,000 if the Gambling Bill became law.

In 2013, it was reported that the number of problem gamblers had doubled in the previous six years and had reached 450,000.

Last month, it was reported that the number of problem gamblers had doubled in the last three years and had reached 336,000.

How does the number of problem gamblers keep doubling without it increasing? Let's take a look...

Back in 2004, when that nice Mr Blair was trying to liberalise the gambling market, a prediction was made...

Public Health Association chairman Geoff Rayner said the number of addicts could double if the plans went ahead... He highlighted research by The Henley Centre, a strategic marketing consultancy, which estimated the number of gamblers could double to 750,000 people.

You don't need to be Carol Vorderman to work out that there were approximately 375,000 problem gamblers in 2004.

Mr Blair got his liberalisation (or most of it) and we have had gambling advertising on TV and a somewhat more relaxed approach to casinos ever since. We have also seen the rise of fixed-odds betting terminals and online gambling, both of which were supposed to drag us to perdition.

What happened next? The popular narrative is that there was an epidemic of problem gambling. Take this, from the Independent in January 2013, for example...

A huge increase in gambling addicts will make Britain's obsession with online betting a £2bn business. New evidence reveals that the number of people in danger of becoming problem gamblers has reached nearly a million, while hardcore addicts have doubled in six years to almost 500,000.

What the Independent calls 'hardcore addicts', the British Gambling Prevalence Survey calls 'problem gamblers' and it is the number of problem gamblers that was estimated in the 2010 survey. The Independent was citing that survey, although they exaggerated the number. As they note later in the article, the survey's mid-point estimate was 450,000.

So not actually a doubling, but still a rise, right? Not necessarily. The only firm conclusion drawn in the 2010 British Gambling Prevalence Survey was that there were between 254,900 and 593,400 problem gamblers in the UK. The wide gap between the low and high ends of the estimate meant it was uncertain whether there had been any rise at all since the previous survey of 2007.

Nevertheless, a narrative had been set in motion that problem gambling had doubled and it was all the fault of online gambling/fixed odds betting terminals/advertising.

Last month, The Times reported (not for the first time) that the number of problem gamblers had doubled again.

The number of people with a severe gambling problem has almost doubled to 336,000 in the past three years, according to the Gambling Commission.
But hang on a moment. 336,000 is less than it was in 2004 when Geoff Rayner made his gloomy prediction - and the rate of problem gambling has supposedly doubled twice since then.

Fortunately, there is a simple explanation. There are not enough problem gamblers for surveys to estimate the number with any precision. Since 1999, when the first survey was conducted, the rate has ranged from 0.3 per cent to 0.9 per cent of the adult population. The high of 0.9 per cent was recorded in 2010 and the low of 0.3 per cent was recorded in 2013. There is no pattern, no trend, but there is random fluctuation.

So this is what happens: when the rate appears to go up, the media report it. When it appears to go down, they don't. The Independent compared the 2010 estimate to the 2007 estimate. The Times compared the 2016 estimate to the 2013 estimate. Nobody compared the 2013 estimate to the 2010 estimate, but if they had they could have claimed that the rate of problem gambling had more than halved.

None of it means anything. In practice, the estimates have such wide confidence intervals (or margins of error, if you like) that one year's data is statistically indistinguishable from another. Give or take a few fractions of a percentage point, the rate of problem gambling in Britain is 0.5 per cent and has been ever since we started trying to measure it.

Saturday, 3 December 2016

£52 billion cost of alcohol? The garbage keeps coming

The Public Health England alcohol report is so bad it would be hard to make it worse but the Guardian has managed it...

Alcohol-related crime, lost output and ill health costs UK £52bn a year 

Treasury urged to set minimum pricing to reduce alcohol-related harm as research finds cost to taxpayer is twice old estimate

No doubt the government is being lobbied to bring in minimum pricing, but apart from that every word of this is untrue. There is no new research in the PHE report, there is no £52 billion estimate, and the cost of crime, lost output and ill health do not come close to that amount.

The review was undertaken by Public Health England (PHE) and leading academic and medical experts on alcohol. It found that the true cost of alcohol-related harm, which had usually been cited as £21bn a year across the UK, has been “generally underestimated”.

The overall economic burden is due to be between £27bn and £52bn in 2016 (1.3%-2.7% of GDP), the researchers said.

This a rough approximation of what the PHE report says but, as I explained yesterday, the PHE report is riddled with misinformation.

The authors do indeed say that estimates of the external costs of alcohol range from 1.3% and 2.7% of GDP. In fact, there are only two estimates that are ever cited, both of which are more than a decade old. The most commonly cited study is a 2001 report for the cabinet office by Rannia Leontaridi which is the source of the £21 billion figure. This equates to roughly 1.3 per cent of GDP and PHE reference it correctly.

Leontaridi's estimate is often portrayed as the cost of alcohol to the taxpayer, but - as she made clear clear in the text - it is actually a combination of costs to public services and costs to individuals, including drinkers themselves. Moreover, some of these are 'emotional' costs, ie. non-financial. For example, she includes £12 billion as the cost of alcohol-related crime but this includes £4.7 billion of ‘emotional impact’ costs, £1 billion of lost productivity, £2.5 billion of costs borne by victims and £1.5 billion spent in anticipation of crime (eg. insurance, security systems). None of these are costs to the taxpayer.

Leontaridi's study is fine so long as you understand what she's measuring. Unfortunately, it has been willfully misinterpreted for the last fifteen years.

The other estimate that is occasionally cited is £55 billion. This must be the figure PHE are citing when they talk about 2.7 per cent of GDP, but it is impossible to be certain because they give no reference for it. The PHE report is so inept that it wouldn't surprise me if the authors don't know where it comes from.

They can almost be forgiven their ignorance. The source is so obscure that I doubt even many alcohol researchers have ever read it. It comes from a 2006 rapid review by the National Social Marketing Centre. It's never been in the public domain but I got the NSMC to send me it when I was writing The Wages of Sin Taxes a few years ago and I have put it online today so you can see what I'm talking about.

It's a shame that the NSMC report isn't better known because it contains moments of unintentional hilarity. My favourite bit is when include the UK's entire annual expenditure on'sports & fitness related products' and 'fat and low calorie food' as costs of obesity.

It is difficult to determine the proportion of expenditure on sports and fitness and on diet foods that can be attributed to obesity or fear of obesity. However, overall fitness and in particular a desire to lose weight or maintain weight loss may be a driving factor in the decision to join a gym or purchase diet food.

In 1999, the total value of the market for sports & fitness related products in the UK was estimated at £6.3bn (FCO 1999). Adjusting for inflation and country, this equates to £6.31bn across England in 2005. The market for reduced fat and low calorie food was estimated at £5.2bn in 2000, with an estimated growth rate (in cash terms) of 17.4% between 2001 and 2003 (Mintel International Group Ltd 2003). This equates to an annual growth rate of 8.4%. Adjusting for country, this means the market for diet foods is estimated at £6.70bn for England in 2005. Thus total household expenditure on sports and fitness and diet foods is estimated at £13.1bn across England in 2005.

This gives you a flavour of the whole report. Any study that portrays £13 billion of sports equipment and Diet Coke as a cost of obesity should not be taken too seriously.

The NSMC's cost of alcohol estimate isn't much better. It arrived at a figure of £48-50 billion but at some point this was upwardly adjusted for inflation and has been cited as £55 billion ever since.

Included in the NSMC figure is £8 billion that drinkers spend on alcohol! Their rationale, such as it is, is this:

The cost of alcohol consumption in England is £32 billion, approximately one third of total household expenditure on food and drink and about 6% of total consumer spending. Drinking of more than the guideline levels accounts for about 25% of all alcohol consumption. Heavy drinking men consume 2.5 times the mean level of male alcohol and heavy drinking women consume 4.5 times the mean level of female consumption. This means that families with one or more heavy drinker are likely to spend a high proportion of disposable income on alcohol. Estimating these costs as 25% of expenditure on alcohol amounts to £8billion.

If that doesn't make any sense to you then are you are thinking clearly. There is no justification in economics in including any part of private expenditure on a product as being a societal cost. Private expenditure is always exceeded by private benefits. You can't count one without counting the other.

And why stop at 25% of alcohol expenditure? Why not go wild and include it all? (Sure enough, when they get to their estimate of the cost of smoking, they include 100% of expenditure on tobacco.)
For reasons that I cannot understand at all, they then add £3 billion to their estimate for the tax paid by drinkers on 'excess alcohol'. This is wrong for so many reasons. Firstly, it's a transfer, not a cost. Secondly, it's private expenditure by drinkers. Thirdly, it's a Pigouvian tax that offsets external costs, it doesn't add to them. Fourthly, it's double counting because the tax is already included in the £8 billion of expenditure they've already (inappropriately) included.

Most of the rest of the NSMC estimate consists of 'intangible' (ie. non-financial) of £16 billion and private costs to individuals, including the drinker (£23 billion) and the authors are upfront about the fact that the study is not of external costs.

We have taken as wide a societal perspective on costs as possible. The costs primarily fall on individuals/families, although these are more difficult to measure, sometimes because of lack of any contact with formal agencies. 

So there we have it. A ten year old rapid review which includes costs that shouldn't be included and which is explicitly not an estimate of negative externalities.

And yet this is how the Guardian presents it...

Doctors are urging Philip Hammond to raise the price of alcohol to tackle the “scourge” of drink-related harm after it emerged that crime, ill health and lost productivity cost up to £52bn a year, far more than previously thought.


The review was undertaken by Public Health England (PHE) and leading academic and medical experts on alcohol. It found that the true cost of alcohol-related harm, which had usually been cited as £21bn a year across the UK, has been “generally underestimated”.

No one who has read the NSMC publication would suggest that it underestimated anything. PHE's claim that cost-of-alcohol estimates are too low is based on a lie, as I said yesterday. The PHE report says:

Few studies report costs on the magnitude of harm to people other than the drinker, so the economic burden of alcohol consumption is generally underestimated.

This could not not be less true. I seriously doubt whether the authors have read either of the studies they are obliquely referring to. They count every legitimate cost 'people other than the drinker' and then pile on a bunch of ineligible costs to both drinkers and and non-drinkers to arrive at a greatly inflated figure. 

In any case, there is no new study and no new estimate. Insofar as the Guardian's £52 billion figure has a source (PHE do not use this figure), it does not say what the Guardian thinks it says.

Friday, 2 December 2016

PHE's alcohol review: part 1 - the lies

Public Health England (PHE) published their review of alcohol control policies today. It is a hodge-podge of wishful thinking, blinkered analysis and outright deception. As you might expect, it takes the 'public health' view that all human behaviour is dictated by advertising, affordability and availability, and that the correct response is therefore to tax and ban.

Alas, the world refuses to bend to neo-temperance dogma and so when observable evidence does not fit the narrative, PHE rely on modelling, modelling and more modelling, mostly from the prolific Sheffield University alcohol team, supplemented by opinions from a handful of anti-alcohol fanatics.

A blog post covering all the tricks used in the PHE report would be too long so I'm going to split it into several sections, starting today with some of the outright lies.

Part one: the lies

1. Lying about the price of alcohol

Over the last 30 years, the affordability of alcohol in the UK has steadily increased and alcohol is now 60% more affordable today than it was in 1980 (182). Relatively speaking, disposable incomes have increased and real-term alcohol prices have decreased.

The real-term price is the price adjusted for inflation. Since 1980, the price of alcohol has risen by 23 per cent more than the rate of inflation, therefore the real price of alcohol has increased, not decreased.

2. Lying about the effect of minimum pricing

Unlike tax increases, where the price increase may not necessarily be passed through to the point-of-sale, this policy ensures that a minimum price is paid by the consumer. In principle, this applies to all alcohol, however this policy typically affects the high-strength, cheap products that are predominantly sold in the off-trade.

Minimum pricing may have a greater effect on the price of high-strength, cheap products but it does not 'typically' affect them. Recent evidence from Scotland shows that, at 50p per unit, minimum pricing will affect half of all the alcohol sold in off licences, including two-thirds of the beer. That is not a targeted measure.

3. Lying about Canada

A 10% increase in average minimum price for all alcoholic beverages in British Columbia was associated with a 32% (95% confidence interval [CI]: ±25.7%) reduction in wholly alcohol-related deaths within nine months, a 9% reduction in acute alcohol-related hospital admissions and a 9% reduction in chronic alcohol-related hospital admissions two to three years after the policy was implemented.

Regular readers will know that this is flagrant lie. Alcohol-related deaths did not fall during this period and alcohol-related hospital admissions continued to rise.

The best that can be said about Public Health England here is that they didn't make this factoid up themselves but got it from the minimum pricing crusader Tim Stockwell.

4. Lying about advertising

Marketing is a commercial strategy with the goal of increasing sales of alcohol by increasing market size (new sales from consumers who would not have purchased or purchased less of a product) and market share (new sales from consumers who would have purchased rival products). 

The goal of advertising, whether for alcohol or anything else, is not necessarily to increase market size. In fact, that is seldom the goal and certainly not the effect. If it were, most advertising would be considered a failure. Beer sales have been in decline for many years, for example, but millions of pounds are spent advertising beer. The overwhelming goal of advertising for any established product is market share, not market share and market size, as PHE claim.

5. Lying about costs

The economic burden of alcohol use is substantial, with estimates placing the annual cost to be between 1.3% and 2.7% of annual GDP. Few studies report costs on the magnitude of harm to people other than the drinker, so the economic burden of alcohol consumption is generally underestimated. 

Bunkum. All the studies PHE refer to look at harm to other people (ie. negative externalities). Not only do they include negative externalities, but they inflate their figures by portraying many internal costs as external costs. It would be impossible to claim that the cost of alcohol is between 1.3% and 2.7% of annual GDP without including a vast range of alleged externalities.

The 1.3% actually comes from the most-cited English study which claimed the costs of alcohol were £21 billion per annum. That study explicitly set out to look only at negative externalities, including physical and financial harm to others. In practice, however, it included various internal costs such as lost productivity which should not have been included but which inflated the estimate. Either way, it is a lie to claim that it didn't 'report costs on the magnitude of harm to people other than the drinker'. On the contrary, it exaggerated them. 

6. Lying about costs again

Crucially, the financial burden which alcohol-related harm places on society is not reflected in its market price, with taxpayers picking up a larger amount of the overall cost of harm compared to the individual drinkers.

This may be sheer ignorance rather than a deliberate lie, but it is certainly untrue. By any measure, the costs to non-drinking taxpayers are far lower than the £10.4 billion drinkers in England pay in alcohol duty. The most charitable way to interpret PHE's statement is that the authors don't know the difference between societal/intangible costs and financial costs to the state.

7. Lying about alcohol consumption

This is a bonus lie that appeared in Public Health England's press release and in Duncan Selbie's weekly e-mail...

As a nation we are drinking twice as much as we did 40 years ago

No, we're not.

8. The big lie

The dishonesty in (7) points to a wider lie that runs through the whole PHE document. When looking at alcohol consumption over time it never refers to per capita consumption. Instead it talks about overall consumption. That might not seem very important, but the population of England and Wales has grown significantly in the last 40 years, with a large share of the increase taking place in the last ten years. The graph above shows alcohol consumption per capita, but PHE choose to show overall consumption like this...

Even if you ignore population growth and look at gross alcohol sales, PHE are still lying when they claim that we are drinking twice as much as we did 40 years ago, but there is a very specific reason why they have picked this inappropriate measure of consumption.

If you look at per capita consumption, the recent peak occurred in 2004, but if you look at overall consumption the peak occurred in 2008. This seemingly minor discrepancy has major implications for PHE's tax-and-ban approach to alcohol because if you accept that consumption started falling after 2004, there is no obvious explanation in the neo-temperance handbook. In 2005, incomes were still rising, alcohol was still becoming more 'affordable' and the licensing laws had just been relaxed. According to the beliefs of the 'public health' lobby, alcohol consumption should have kept rising.

But if you pretend that alcohol consumption was still rising during this period and only began to fall after 2008, you can claim that it was the result of alcohol becoming less affordable thanks to the alcohol duty escalator and the global economic downturn (both of which began in 2008). Nick Sheron and Ian Gilmore tried this trick out in a BMJ article earlier this year and it has now been embraced by PHE (unsurprisingly, since Sheron is one of the authors of the PHE report).

Measuring consumption by overall volume rather by per capita also has the effect of making the drop in consumption look smaller than it is. The decline per person in the last twelve years has been nearly 20 per cent, but by ignoring population growth PHE can portray the drop as being relatively insignificant. This gets them off the hook because it means they don't have to give an explanation for why alcohol-related hospital admissions and deaths have failed to decline (as they should under the total consumption model).

Despite the recent small declines in overall alcohol consumption, many indicators of alcohol-related harm continue to rise.

A twenty per cent reduction in drinking since 2004 is not as recent as PHE are claiming and it is certainly not small. I call that a lie.


Thursday, 1 December 2016

Brazilian smoking ban miracle

This execrable study didn't get any media attention and when I saw it tweeted by Tobacco Control magazine I didn't even bother reading the abstract, so bored am I of fraudulent smoking ban miracles.

Anyone who's followed these scams over the last twelve years knows what to expect. Typically, the activist-researchers take heart attack figures from before and after a smoking ban, create a counterfactual that suits their purpose and then claim that there were fewer heart attacks than there would have been without a smoking ban. It helps if heart attacks are already in decline (they often are).

That's why I didn't read the study. I thought I'd seen it all before. But my interest was piqued by Michael Siegel who blogged about it yesterday. The authors attempted the method described above but were faced with the awkward fact that the heart attack rate absolutely skyrocketed after Sao Paolo introduced its smoking ban in early August 2009.

The authors helpfully provide all the numbers in the study - for this a crime carried out in plain sight - and I have graphed them below. The red line shows the date of the smoking ban.

You can see the number of deaths from myocardial infarction decline after September, but you can see the same seasonal trend in previous years with a fall at the end of the year and a rise at the start of the next. Early in 2010, it rises sharply - much more sharply than in previous years - and stays high for the rest of the period covered. Unless the population of Sao Paulo increased by half at the start of 2010, this is not a decline in the heart attack rate.

Before the ban, the number of deaths hardly ever exceeded 600 per month and was often below 500. Within a few months of the ban, there were never fewer than 700 deaths per month.

In the bizarro world of Tobacco Control, this shows that the 'mortality rate for myocardial infarction reduced following the smoking ban in Sao Paulo'.

Things are not much better if you look at hospital admissions for heart attacks. There is a year-on-year increase almost every month after the smoking ban...

These inconvenient facts would have deterred lesser mortals, but our intrepid researchers pressed on regardless, making so many unexplained adjustments to the data that they were able to conclude their study as follows:

Mortality rate and hospital admission rate for myocardial infarction decreased after the comprehensive smoking ban law in Sao Paulo city.

And so history is re-written. It can join the growing list of 'public health' facts that are not facts, such as 'minimum pricing reduced alcohol-related deaths in British Colombia by a third' or 'Rotten teeth in toddlers are at crisis level’ or 'Sugar tax reduced soda sales by 12 per cent in Mexico'. Or, indeed, 'Scottish smoking ban reduced heart attacks by 17 per cent'.

It's not that these guys have failed to prove cause and effect. There is no effect to find a cause for. The campaigners have created their own fantasy land where the facts are whatever they want them to be and words mean whatever they want them to mean.

Wednesday, 30 November 2016

The IQOS cometh

Today sees the UK launch of Philip Morris's heat-not-burn product, IQOS, and the Today programme carried an item on it which you can listen to here. Today went to PMI's Geneva headquarters to hear about the science behind it and also interviewed ASH's Deborah Arnott who did her usual moaning about industry but said that if IQOS helps smokers quit that was 'all well and good'.

She made the facile point was made that if PMI wanted to reduce harm they would stop selling cigarettes. The reality is that no harm would be reduced by one tobacco company ceasing production of cigarettes and if even if every tobacco company stopped selling them it is doubtful whether harm would be reduced by driving the market underground. But the more important point is that tobacco companies can only sell cigarettes because people want to buy them.

A more rational way of looking at it is to say that if smokers want to reduce harm, they would stop buying cigarettes. The fact that they don't suggests that they think smoking is worth the risk, or plan to give up at a later date, or have not found a suitable substitute.

That is where e-cigarettes, snus and new nicotine products such as IQOS come in. A couple of striking claims were made by PMI on the Today programme. Firstly, that IQOS delivers 90-95 per cent fewer harmful constituents than conventional smoking and, secondly, that 70 per cent of Japanese smokers who have tried it have switched to IQOS permanently. The other striking figure is that PMI have spent $3 billion on developing IQOS to date. In other words, they are taking this seriously. It is no harm reduction gimmick. 

If the degree of risk reduction can be independently confirmed and if the Japanese experience can be verified and replicated, this could be a significant development. Much depends on whether smokers know the product exists (no easy task given the UK's extreme ban on tobacco marketing) and whether they like it when they try it. As I discussed in Free Market Solutions in Health: The Case of Nicotine, there are unintended consequences from hyper-regulation that work against harm reduction.

We must wait and see, but UK smokers now have a chance to find out. PMI have opened an IQOS store in Soho. I'll be popping up there in a couple of weeks. I've tried the product before and although I switched to e-cigarettes a few years ago, I can see why smokers who have never fully embraced vaping—like my IEA colleague Mark Littlewood—like it.

Tuesday, 29 November 2016

Last Orders podcast with Martin Durkin

There's a new Spiked Last Orders podcast available. This month's guest is the great Martin Durkin (director of Brexit: The Movie, Britain's Trillion Pound Horror Story and more).

We talk about booze curtains in Ireland, the shrinking of chocolate bars, the war on fizzy drinks and go off on other tangents. It's fun. Have a listen.

Monday, 28 November 2016

Still no evidence of a slippery slope

As we all know, the idea that the anti-smoking movement is a precursor to nanny state regulation of what we eat and drink is an evil myth spread by the tobacco industry. As Simon Chapman, Deborah Arnott and many good, honest tobacco control experts have tirelessly explained, it is a baseless scare story. Tobacco is a unique product, etc.

Now that we've cleared all that up, let's take a look at what was said at the WHO's recent anti-tobacco conference by Dr Vera da Costa e Silva, the head of the Convention Secretariat on Tobacco Control...

We are also watched by sugar and alcohol products manufacturers, who see the tobacco control movement as a precursor to threats they now face from public health campaigns. These industries fear a united international community acting on behalf of consumers. In the coming days, I hope their fears will be fully justified as we take further steps to end the tobacco epidemic.


As Johnny Rotten once said, do you ever get the feeling you're being cheated?