Wednesday, 19 November 2014

The Wages of French Sin Taxes

If you wanted to mislead without actually lying you could do worse than create a list of ‘social costs’ and present them to the media as the price paid by society for any given activity. No matter how explicitly you state that your study only looks at costs - not benefits or savings - and no matter how clearly you state that many of your costs are paid by individual consumers - not by the government - you can be sure that politicians and journalists will assume that you are talking about net costs to the taxpayer.

I have written at length about the various claims about smoking, drinking and obesity ‘costing’ the UK and US governments eye-watering sums of money. Upon closer inspection, the underlying figures invariably show that the bulk of these multi-billion dollar ‘costs to society’ are dominated by direct costs to the consumer, such as lost productivity and forgone income, but neglect to offset the costs with benefits, such as the heavy taxes paid by consumers. They are squeamish about the cold fact that premature mortality often results in significant savings to the government (by reducing pension payments and elderly healthcare, for example). In short, they are a one-sided assessment of a balance sheet, focusing on the debit side without acknowledging the credit side, and they treat costs to individual consumers as costs to society.

There is nothing technically wrong with compiling a list of costs that excludes savings, benefits and consumer surpluses. Nor is there anything wrong with including private costs to individuals as part of a ‘cost to society’ estimate. The authors of such studies usually make it plain that this is what they are doing. The problem is that very few people will ever read the study. They will only hear that alcohol, for example, ‘costs society’ $100 billion a year and assume that this is a direct cost to the state that should be paid for by drinkers. The conclusion is obvious: alcohol duty should be increased until it can raise $100 billion a year. Obvious, but wrong.

This confusion is not confined to English-speaking countries. Last month, a group of French MPs claimed, contrary to all evidence, that taxes on tobacco do not cover the costs of smoking and even a peer-reviewed journal has published the insupportable assertion that ‘France spends €47 billion a year treating smoking-related illnesses’. Both claims are based on a misreading of a 2006 study by Fenoglio et al. which estimated that the ‘social cost’ of tobacco was €47 billion, the ‘social cost’ of alcohol was €37 billion and the ‘social cost’ of drugs was €2.8 billion.

The Fenoglio study was based on data from 2000 when smoking prevalence was higher than it is today, but that is the least of the problems with using it as a guide to tobacco’s cost to the government in 2014. The €47 billion grand total is dominated by an €18 billion sum for healthcare and another €18 billion sum for lost productivity. The €18 billion cited in this study is much higher than the £2.7 billion (€3 billion) that smoking is said to cost the British healthcare system, but it is true that there are significant costs incurred by smoking-related diseases on publicly funded healthcare systems. The question is whether smokers require more healthcare - or more expensive healthcare - than nonsmokers.

This question has been repeatedly answered in the academic literature over the last thirty years. Van Baal et al. (2008) found that the lifetime healthcare costs of smokers were 21 per cent lower than those of nonsmokers. Similar conclusions have been drawn in many other studies from around the world, most recently by Kampen et al. who confirmed that ‘Elimination of diseases that reduce life expectancy considerably increase lifetime health care costs.’ This is true of smoking-related and obesity-related diseases, although there is much less evidence about alcohol-related diseases.

Clearly, there are many reasons why governments would want to reduce the prevalence of disease, but saving money is not one of them. It may be true that the treatment of smoking-related diseases costs France €18 billion, but the evidence strongly suggests that healthcare costs would be even higher in the absence of smoking. Smoking cannot, therefore, be said to create an excess cost that requires a Pigouvian tax.

The €18 billion cost of lost productivity would also not feature in a calculation of net costs to government. In the case of premature (working age) mortality, the government loses a taxpayer who was paying into the communal pot, but also loses someone who was taking from the communal pot. There is, on average, neither a net gain nor a net loss.

In terms of absenteeism or sub-standard work (as a result of alcohol or drug use), there is a cost to employers, but it is a cost that is ultimately paid by the employee. As Crampton et al. note:

‘Employer and employee are bound by a contractual nexus; the worker’s reduced productivity is internal to his relationship with his employer. A less productive employee is less likely to receive future promotions and salary increases; he bears the burden of his reduced productivity. Firms that fail to detect worker productivity and promote workers beyond their worth will eventually go under.’

Since wages are closely linked to productivity, the cost of lost productivity is already dealt with by normal economic mechanisms without the need for Pigouvian taxes which would, in any case, go to the government rather than the employer.

It could be argued that lost productivity incurs a cost on government by lowering wages and therefore lowering income tax. The logical extension of this argument is that if citizens do not work as hard as they can for the longest possible hours, they are a burden on the state. This is a morally dubious claim, but even if it were true the cost of lost productivity would be a tiny fraction of the figures cited in the Fenoglio study (€18 billion for tobacco, €16 billion for alcohol and €812 million for drugs).

The other major cost of tobacco and alcohol in Fenoglio et al., which makes up €8 billion and €7 billion respectively, is ‘loss of consumer revenues’. The authors provide no explanation of what this consists of but it seems that most of it is incurred by the users of the product and therefore cannot be an externality. Other costs cited, including ‘insurance spending’, ‘private associations’ and ‘other private costs’, are obviously private costs which are not paid by the government and, in most cases, are paid by the consumer of the product.

The study also includes substantial figures under ‘loss of tax’ (€3.7 billion for tobacco, €3.5 billion for alcohol). It is unclear what this refers to, but it is wrong to count lost tax without balancing the ledger with received tax. Tax revenues from alcohol and tobacco (but not, of course, from drugs) are very substantial. France receives €14 billion in tobacco duty and €3 billion in alcohol duty. Tobacco duty certainly exceeds any reasonable estimate of the net costs to the government from smoking and it is possible that alcohol duty does likewise.

This leaves only a small number of costs to the state. €1.78 million is cited as the cost of smoking-related fires, but most of these take place in the smoker's own home and are therefore private, internal costs. The external cost of smoking-related fires to the state is confined to the cost of publicly funded fire services. A further €87 million for alcohol and €740 million for drugs is listed as costs of administration. Although these figures are not explained in the text, they may involve the public sector and could therefore be considered worthy of Pigouvian taxation.

€2.8 million is listed as the cost of tobacco prevention. This could be interpreted as a legitimate cost of smoking, but could equally be seen as a cost of anti-smoking policy.  Similarly, the cost of enforcing drug laws is given as €145 million. The authors note that this is higher than the cost of enforcing alcohol laws (€56 million) and it could be argued that these higher costs are a consequence of drugs being illegal, rather than drugs being consumed. Regardless of whether these costs are seen as genuine costs of tobacco and drug use, they remain a very small part of the overall cost estimates.

It summary, many of the costs listed in the Fenoglio study are not externalities incurred on government by the users of tobacco, alcohol and drugs. Some of the costs, such as healthcare, are genuine costs to public services, but they disappear once savings, benefits and existing Pigouvian taxes are accounted for.

To be clear, there is nothing necessarily wrong with the way the authors have conducted their research. They explicitly state that they are studying gross costs, not net costs, and they make it plain that they are including costs to smokers/drinkers/drug users as costs to society. The problem lies with the way the study has been interpreted. Politicians and journalists routinely confuse ‘gross societal costs’ with ‘net costs to the taxpayer’. Whilst their confusion is understandable, it cannot be said too often that there are vast differences between the two.


Christopher Snowdon said...

The first point that should always be made in any discussion of 'smoking costs' is that there is no such thing: smoking provides a massive boost to the economy, with large revenues and savings for government, and significant revenues for multiple industries. The term 'smoking wealth' should be used instead.

I would also like to see an economics expert discuss the gains surrounding specific parts of smoking-created wealth. For example, fires caused by smoking. Are these a cost or a gain? I suggest they are a gain (and certainly not a 'cost to the nation' as is implicit in many of these 'cost estimates'). Let's say fires cost £2m in damage. Insurance pays, or the individual if uninsured. Therefore these costs are internalised, as you point out - they are not costs to the state. In fact the opposite is true: if insurance pays for the fires, it has created employment, and created more taxes for the state. In fact these fires are a net gain for the economy as they produce movement of funds and therefore employment and taxes. Nobody pays except the consumer. The consumer foots all bills in the end, and contributes to state funds in one way or another.

Christopher Snowdon said...

A good example of 'Lost productivity' foolishness is to suggest the lost productivity from having an eight-hour working day, not a 12-hour shift. A 50% loss of productivity.

Or take weekends. Based on a five-day working week, there is 20% lost productivity by having two whole days away from the coal-face, not just the single day required for the worker's essential recuperation from the 12-hour work days.

Christopher Snowdon said...

However, you've partially fallen into their trap: "dominated by direct costs to the consumer, such as lost productivity and forgone income, but neglect to offset the costs with benefits, such as the heavy taxes paid by consumers." Heavy taxes are not a benefit. Enjoyment of what is being taxed is a benefit; taxes are yet another cost.