The Social Affairs Unit

Print Version • Website Home • Weblog Home

Use the buttons below to change the style and font size of our site.
Screen version     Print version:   
May 25, 2005

Why reducing drink driving limits may not save 65 lives per year

Posted by John Maloney

The Parliamentary Advisory Council for Transport Safety (PACTS) - a charity set up "to promote transport safety legislation to protect human life" - is calling on Parliament to lower the drink driving limit from a blood alcohol count (BAC) of 80mg per 100ml to one of 50mg per 100ml. PACTS - and their expert Prof. Richard E. Allsop - acknowledge that such a change will "do little to save the 400 and more lives lost each year from driving way over the 80 limit" but argue that a reduction to 50mg could save around 65 lives per year. They are calling for the Road Safety Bill announced in this year's Queens Speech to implement such a change.

Dr John Maloney - Reader in Economics, University of Exeter - has examined the case made by PACTS and Prof. Allsop. Dr Maloney argues that the assumptions on which their case is made are questionable. The PACTS figures assume that everyone driving around with a BAC between 80mg and 100mg - i.e. already above the old limits - will drop by 30 points if a new lower limit were introduced, i.e. drive at the same point above the new limit as they were above the old limit. Dr Maloney argues that this is quite an assumption to make. If reducing the drink driving limit would not alter the behaviour of this group, the number of lives saved drops to 23.

In 1967 it became illegal to drive with a blood alcohol count (BAC) above 80mg per 100ml. For the average person this corresponds to just over two pints of beer. But as soon as alcohol starts entering your bloodstream you start losing it as well. If (again) you're average, this will be at the rate of half a pint of beer's worth per hour. So the formula for those of standard biochemistry is P = 2 + H/2, where P is the number of pints you can drink legally and H is the number of hours the session has lasted. But, because everyone's metabolism is so different, you would be highly unwise to go up to this limit.

And, quite apart from the law, your risk of an accident rises quite sharply before you reach the current limit. How sharply, again, will depend on who you are - some drivers are more easily affected by alcohol than others, some face a drive back along safer roads. But in 1997 Geoff Maycock of the Transport Research Laboratory concluded that a driver on the current limit of 80 was on average 5.6 times as likely to be involved in an accident in which someone is injured, and 12.4 times as likely to die in a crash, as a driver who had drunk nothing [G. Maycock, Drinking and Driving in Great Britain - A Review, Transport Research Laboratory 232, (1997)].

Mr Maycock's figures led Professor Richard Allsop of University College London to estimate that cutting the limit to 50mg would save 65 lives a year. The Parliamentary Advisory Council for Transport Safety took this figure up, and is now advocating a cut in the limit to 50mg. So how reliable is this estimate of 65 fewer deaths?

Prof. Allsop reaches his figure in three stages. First, he assumes that, with a change in the law, everyone driving around at 50-80, i.e. between the proposed & existing limits, would go down to 50. Second he assumes that everyone currently driving around between 80 and 110 would drop 30 points, i.e. drive at the same point above the new limit as they were above the old one. Finally, he applies these assumptions to Mr Maycock's analysis of risk ratios.**

Risk ratios work by comparing the BAC's of drivers stopped in a random check with those of drivers involved in an accident. Suppose your control panel of randomly stopped motorists includes 40 people with a blood alcohol level over 80mg, and your accident-involved "panel" contains 100. Then the "crash rate" for this category is 100/40 = 2.5. Suppose now that the control group and the accident group respectively contain 2000 people and 1000 people who have drunk nothing at all. The crash rate for the abstainers is thus 0.5. The risk ratio for the over-80's is 2.5/0.5 = 5. So being over the 80mg limit will make you five times as likely to crash.

By working out risk ratios for various, overlapping, ranges of BAC, Mr Maycock derived two continuous exponential functions, one for the risk of drivers crashing (with injury involved) at different alcohol levels, the other for their risk of being killed. The second formula (based on a number of roadside surveys plus coroners' reports on the BAC of drivers killed in accidents) was:

RR = e 0.072BrAC

where RR (risk ratio) is your risk of dying as a multiple of your risk if you drink nothing. BrAC is breath, not blood, alcohol count. But it translates into BAC by the accepted formula that BrAC = 0.4375 BAC. Hence:

RR = e (0.072BAC)0.4375 = 1.032BAC

This produces a stronger link between drinking and accidents than previous surveys (and, in particular, a stronger link than the Grand Rapids survey of 1962, which found a statistically detectable effect starting only at 80 mg, and indeed was used as part of the case for fixing the limit at 80 back in 1967.) But let us now look in more detail at the estimate that lowering the limit to 50 mg would save 65 lives.

This figure is derived in table 1.

  50-60 60-70 70-80 80-90 90-100 100-110
(1)Risk ratio 5.65 7.75 10.61 14.55 19.93 27.32
(2a) Risk ratio at BAC =50 4.83 4.83 4.83      
(2b) Risk ratio at BAC 30 pts lower       5.65 7.75 10.61
(3) Relative risk ratio after change in law ((2)/(1)) 0.855 0.623 0.455 0.389 0.389 0.389
(4) No. of drivers killed in this decile as percentage of no. of drivers killed at >80 mg. 3.91 3.91 3.91 3.97 3.97 3.97
(5) Total no. of fatalities involving drivers in this decile [550 x (4)] 21.5 21.5 21.5 21.8 21.8 21.8
(6) No. of fatalities after change in law [(5) x (3)] 18.4 13.4 9.8 8.5 8.5 8.5

Here we take drivers with BAC's in each of the six deciles which Prof. Allsop assumes would be affected. His assumption, to repeat, is that lowering the limit to 50 would make drivers at 50-80 cut their BAC to 50, and drivers at 80-110 cut their BAC by 30 points. It is no good saying this assumption sounds a bit arbitrary unless one can come up with one that is less so, so let us follow its consequences. The third row in the table gives, for each decile, the relative risk ratio as between drivers changing their behaviour as assumed and drivers carrying on as now. So it says, for example, that drivers currently at 70-80 will only die 45.5% as often as now if they cut their BAC to 50.

But it is, of course, total deaths in drink-related accidents that concern us. Prof. Allsop now assumes that if you do have a fatal accident, the total number of people you kill is unrelated to your BAC. Very drunk drivers, in other words, have more fatal accidents, but the same kind of fatal accidents, as slightly drunk ones and sober ones.

This seems reasonable - in fact it is hard to think of reasons why it should not be so. And it allows us to estimate how many people, in total, are being killed by drivers in each BAC category.

If drivers at different BAC's do have the same kind of accidents as one another, then the percentage of total deaths caused by drivers in each BAC decile will be the same as the percentage of driver fatalities that fall into that decile. Thus the 4th row (driver deaths in each decile as a percentage of driver deaths at over 80 mg) also gives us the figures for total deaths in each decile as a percentage of total deaths when the driver was over 80 mg. Multiplying these percentages by 550, the current annual total of deaths caused by drivers over the existing limit of 80, thus gives us total deaths caused by drivers in each of the critical BAC deciles (row 5).

But the assumption that more and less drunk drivers nonetheless have the same kind of accidents has a second consequence. As drivers drink less, risk ratios will come down for innocent victims in the same proportions as they do for the drivers themselves. Row (3) thus also gives the relative risk ratios, after a change in the law, for all fatalities. (5) can now simply be multiplied by (3) to get the numbers of total deaths after a lowering of the limit to 50. Subtracting (6) from (5) and aggregating across all six deciles gives a figure of (to the nearest integer) 63 lives saved.

There are, however, at least three reasons for treating these figures with caution.

(1) First, and most important, 40 of the 63 lives allegedly saved are drivers, and the victims of drivers, who were already over the old limit. Professor Allsop assumes that that everyone driving around between 80 & 110 drops 30 points i.e. drives at the same point above the new limit as they were above the old one. The DETR however concluded in 1998 that "international evidence is divided" on whether cutting the limit has any detectable effect on those who were already above the old one. If it doesn't, the number of lives saved drops from 63 to 23.

(2) Some motorists refuse to give breath samples at roadside surveys; others disappear down a side street when they see a random check ahead of them. Both categories are disproportionately likely to be drunk. If drunks disappear from the control group but not from those who crash their cars, the link between drink and accidents is going to be exaggerated. The surveys used by Mr Maycock have, in fact, coped with both these problems relatively well. As he says, "The sites were not announced in advance and were carefully sited so as not to be visible too far ahead so as to prevent drivers re-routing to avoid them." The refusal rate was also low - in the main survey used by Mr Maycock, only 0.8% refused to give a breath sample. This does not mean, however, that estimates of risk will only be 0.8% out - far from it! The 0.8% of refusers is only just below the 1.16% found to be over the limit. Suppose all those who refused were over the limit. Then the true proportion of drivers over the limit nearly doubles (1.16% to 1.96%), and the relative risk for drivers in this category falls by 41%.

The idea that all the non-cooperators are over the limit is an extreme one - though no more so, perhaps, than Mr Maycock's belief that, because drivers were offered an amnesty against prosecution, drinkers were no more likely to refuse than non-drinkers. As he says himself, they faced the penalty of not being allowed to drive any further - quite apart from the discomfiture, ranging from mild embarrassment to severe shame, of being revealed to any stranger as a drink-driver.

(3) Counteracting the last factor, however, is the likelihood that drivers who leave the scene of an accident are also disproportionately likely to be drunk. Assuming anyone well enough to drive (or run) away isn't going to die of his injuries, this won't affect the coroners' BAC statistics. It will, however, lower the apparent number of victims of drink-related accidents and thus bias risk ratios downwards in the same kind of way that drink-drivers opting out of random surveys biases them upwards.

It is arguable (and implicitly Mr Maycock does argue) that neither (2) nor (3) will cause much bias in the context of the specific question: "What would happen if the limit were lowered from 80 to 50?" It is the relative risk ratio between those at 50-80 and those at 80-110 that we need to know. If drivers in these two categories are equally likely to refuse to give a test, and equally likely to leave the scene of an accident, the two risk ratios will be biased upwards or downwards in the same proportion, and the relative risk ratio will be unchanged. But it is hard to believe that drivers below, and drivers above, the current limit have been behaving in the same way as each other in either of these contexts. If, as everyone accepts, drinkers are more likely to make themselves scarce, surely illegal drinkers will make themselves scarcer than legal ones, even allowing for the uncertainty, in both camps, as to exactly where one is vis-a-vis the limit.

We thus still have two reasons for thinking the figure of 63 lives saved may be too high, and one reason for thinking it may be too low. Can we play (2) and (3) off against each other? Which is likely to be the stronger bias? On the one hand, drivers over the limit have a much stronger motive to leave an accident than to avoid a mere survey. On the other, their opportunity to disappear when they have just killed someone is, fortunately, often limited. If we did take (2) and (3) as cancelling out, we are left with (1).

But it is (1) which gives the greatest cause for concern. If, in fact, a cut in the limit has no effect on those already exceeding the old one, the estimate of 63 lives saved falls to 23. To some people this may still be more than enough to justify a change in the law. Most people, however, whatever their views, would be happier knowing if the true figure were really 23, 63, or somewhere in between. So how likely are people currently driving around at 80-110 mg to change their behaviour if the limit falls to 50?

It will depend on whether they know they are over the existing limit. If they do, it seems unlikely that they will fine-tune their drinking to make sure they are breaking the law exactly as badly as they were before. Why should they? But presumably some of those only just over the 80mg limit mistakenly think they are beneath it. Lower the limit and these people will be trying, and failing, to stay below 50mg instead of 80. Here, a 30mg cut in their intake does seem to be the best assumption.

So how many people driving around at over 80 fail to realise this? When I asked 18 drivers of all ages and sexes how many pints of bitter or lager they thought they could have in a 2-hour pub session before going over the limit, every single one came up with a lower figure than the true one (3.25 pints of standard bitter if you have average metabolism, just over 2 pints of strong lager). On that basis, there are far more people wrongly thinking that they will be over the limit if they have another drink than wrongly thinking they are under the limit as they climb into their cars. But pubs serve standard measures and advertise the alcoholic content of their drinks. One of the biggest themes of doctors and road safety campaigners in recent years is how much more alcohol a glass of wine contains than it used to. The old rule that a glass of wine equalled half a pint of beer was based on a 125 ml glass of wine with 9% alcohol content. Nowadays a glass of wine generally means 175 ml and at least 12% alcohol, i.e. nearly double the alcohol in the old measure. People who go round to friends' houses and drink their wine, or accept whiskies or gins on the usual scale (most hosts agree with Kingsley Amis that a single Scotch is just a dirty glass) may very well go home over the limit without realising it.

Perhaps the Department of Transport might look into the question of how different people over the limit got to be like that. It might well shed some light on how far their habits would change if the limit were lowered. Until then it would be best to take the figure of 63 lives saved not as a mean estimate but as an upper limit which might be far too high. Maybe this is one of the reasons why the Government is still holding back and making cautious noises about any change in the law.

** A note on terminology: what I am calling risk ratios Mr Maycock & Prof Allsop call relative risk ratios. I have reserved the term relative risk ratio to describe any ratio of such ratios (in order to avoid talking about relative relative risk ratios!) So "risk ratio" in this article means (e.g.) how much more likely you are to crash at 80 mg compared with 0 mg. "Relative risk ratio" means (e.g.) the risk ratio at 50mg as a proportion of the risk ratio at 80 mg.

Dr John Maloney is Reader in Economics, University of Exeter.

Comments Notice
This comments facility is the property of the Social Affairs Unit.
We reserve the right to edit, amend or remove comments for legal reasons, policy reasons or any other reasons we judge fit.

By posting comments here you accept and acknowledge the Social Affairs Unit's absolute and unfettered right to edit your comments as set out above.

This is a long and wholly theoretical discussion. If Dr. Maloney had examined the experience of jurisdictions which have lowered the drink drive limit he would have found it to be consistent with Professor Allsop's arguments, and consistent also with what knowledge of the effects of alcohol on psychomotor performance implies - lowering the limit saves lives. Of course, it is impossible to predict with total accuracy exactly how many lives - it would depend on, for example, enforcement practices. That doesn't alter the fact that lowering the limit is a good policy, as well as being one supported by the large majority.

Posted by: A.McNeill at May 26, 2005 12:31 PM

Readers may find it helpful to know that the comment by A McNeill above - as his e-mail clearly shows - is by Andrew McNeill, Director of the Institute of Alcohol Studies. A solid academic organisation you might think. This may be the case today, but the organisation has an interesting history. Its origins are in an attempt to impose prohibition on the UK. Interested readers may care to take a look at the website of the National Archive and read up the history of the Institute of Alcohol Studies

I have pasted the material below for the convenience of readers:
"The United Kingdom Alliance was founded in 1853 in Manchester to work for prohibition of alcohol in the UK. This occurred in a context of support for the type of law passed by General Neal Dow in Maine, USA, in 1846, prohibiting the sale of intoxicants.

It was initiated by Nathaniel Card (1805-1856), an Irish cotton manufacturer and member of the Society of Friends. He had also been a member of the Manchester and Salford Temperance Society since 1852, and was interested in what was coming to be known as the Maine Law. At a private meeting at Card's house on 20 July 1852, the National League for the Total and Legal Suppression of Intemperance was formed. Other members included Alderman William Hervey of Salford and Joseph Brotherton MP (Salford). At the third meeting of the League a Provisional Committee was formed, based in Manchester.

Their objectives were openly political, to form and enlighten public opinion nationally, believing that the self-denying and benevolent efforts of temperance societies would never be able to end the liquor trade while legalised temptation to drink and get drunk was permitted. They aimed for total and immediate legislative suppression of traffic in intoxicating beverages.

The name of the League was changed on 14 Feb 1853, to the UK Alliance for the Suppression of the Traffic in all Intoxicating Liquors, and Sir Walter C Trevelyan, became the first president in June the same year, with a General Council holding its first meeting on October.

A weekly newspaper Alliance News was begun in 1854, a journal of moral and social reform, and sold for one penny. Since 1980 it has been published as a bi-monthly magazine.

They were not a total abstinence society, and membership was open to teetotallers and drinkers alike, by 1858 membership had risen to 4500, and £3000 was raised by subscription for their work. Their chief public spokesman was Sir Wilfrid Lawson, MP (1829-1906).

In 1862, the London Union of Alliance members changed to the London Auxiliary of the Alliance, and appointed it's the first London agent, Rev John Hanson. The Alliance had occupied premised in Victoria St, London, until the decision was made to build a new headquarters. A site in Caxton St was purchased in 1937, the new building - Alliance House - being opened in 1938, at a cost of £75,000.

In 1942, the Alliance became a limited company, the UK Temperance Alliance Ltd. By the 1970s the main role of the Alliance was educational work and its interest had broadened to other areas of addiction besides alcohol (much of which is undertaken by the Institute of Alcohol Studies (IAS), a trading arm of the Alliance. In 2003, the UK Temperance Alliance was renamed the Alliance House Foundation."

Posted by: Anonymous at May 26, 2005 02:52 PM

John Maloney has commented very thoughtfully on my estimate, taken up by PACTS, the Parliamentary Advisory Council for Transport Safety, that cutting the alcohol limit for driving to 50mg/100ml of blood would save about 65 lives per year. Neither I nor PACTS envisage such a change being made in isolation, but rather accompanied by improved technology for enforcement (already provided for in the Organised Crime Act), increased allocation of police resources to enforcement, firmer treatment of those who drive way over the limit, and a fresh programme of public information to refresh drivers’ awareness of the risks of driving after drinking and help them to comply with the reduced limit. In such a context there are several reasons for regarding my estimate as cautious, to set alongside what John has rightly identified as a quite optimistic assumption on my part about the response to a lower limit of drivers currently exceeding somewhat the existing limit – his main reason for judging my estimate as "an upper limit which might be far too high".

The first reason is that whilst the data we have both used indicate nearly 600 deaths per year in accidents where a drinking driver has an alcohol level below 50mg, I have counted no reduction in these. In fact some of these drivers would be likely to moderate their drinking in the light of a reduced limit, and it would only take a 7 per cent reduction in these deaths to replace all the 40 of my 65 that John has suggested might not be saved. The second is that I have counted no reduction in deaths among those currently exceeding 110mg, because so many of these would be unlikely to change their behaviour because of a reduced limit. But some might well be affected – people on a one-off binge who are conscious of the limit but have kidded themselves that they are still within it; enough of these might reduce the leeway they give themselves to save a number of lives each year. The third is that the combination of measures that PACTS and I envisage could reasonably be expected to have enough effect on the culture of drinking and driving to achieve an appreciable long term reduction in the proportion of each age-cohort who ever turn into people who persistently drive way over the limit; this would help to tackle the stubbornest core of the problem of death and injury from driving after drinking.

Yes, my estimate is less cautious than some previous ones, but it would be unwise to regard it either as an upper limit or as one which is likely to be far too high.

Prof. Richard E. Allsop

Posted by: Prof. Richard E. Allsop at May 26, 2005 04:08 PM

This reads like a send-up a la C. Northcote Parkinson, but I take it Dr Maloney's analysis is for real. Very impressive, too, although I can't claim the mathematical knowledge needed to fully understand it.

The real question, though, is: even if the estimated saving of 65 lives per year turns out to be correct, would that justify the change in the legal limit? Obviously preventing 65 deaths from drink-driving accidents would be a good thing. But if that is the only criterion, then all pubs should be outlawed and booze completely banned, because doing so would presumably save even more lives.

What? You say that's been tried (in the United States) and there were unintended consequences?

That aside, I can't help wondering if lowering the blood alcohol limit beyond what is arguably a reasonable level at which no-one becomes blind drunk is sending the wrong message.

It implies that the drinking-and-driving issue is at best a matter of adding up ounces, pints or liters, and at worst a sort of cat-and-mouse game in which the drinker's goal is to imbibe up to just under the limit and the law's goal is to catch those who've slipped over it. In either case, it helps implant the idea that road safety is the government's responsibility, rather than the responsibility of the person who is closest to the action: the individual.

I don't want drinkers I might be sharing the highway with asking themselves, "how many is that?" I want them to be asking, "How are my brain and muscular coordination working tonight, seeing as how whether I kill myself or somebody else is down to me?"

Posted by: Rick Darby at May 26, 2005 10:03 PM

I see from the link to Dr John Maloney's website that he claims to last year have delivered "Ned Welch memorial lecture on the subject of 'Merrie England'." Might that not put a new angle on this piece? Those who don't get the reference should read Lucky Jim. Dr Maloney clearly has a fondness for Kingsley Amis.

Posted by: Jane at May 27, 2005 01:38 PM

I’m indeed fond of Kingsley Amis – an unlikely but entertaining character invented by an extremely unsatisfactory junior lecturer in the late Neddy Welch’s history department. ‘Martin Amis’, by the way, is clearly based on Prof Welch’s artistic son Bertrand.

More seriously, I think Rick Darby has a point about the law encouraging cat-and-mouse games, and of course we’d all like people to be so responsible that no laws were needed. Clearly a law is needed on this issue however &, as with any other road safety measure, it’s essential to at any rate try & estimate how much good a stricter law would do.

Posted by: John Maloney at June 9, 2005 05:45 PM

Well done for a thoughtful and dispassionate piece of analysis.

It is all too easy to conclude that cutting the drink-drive limit must save lives, but when we look at the likely effect on real-world behaviour the verdict is much less clear-cut.

Also bear in mind the number of drink-drive offenders who have been caught on the morning after, when a lower limit would probably have made no difference at all to their behaviour.

Posted by: Peter at July 17, 2005 07:06 PM

Agreeing with Rick Darby that we should avoid the cat-and-mouse game. One way of acheiving that is by setting the limit so low that drinking *anything* before driving is seen as unacceptable. Several European countries have done just that, with limits of 0.02% or less.

Having lived in both France when they changed from 0.08 to 0.05 and in Norway which went from 0.05 to 0.02, I can tell you there is an immense difference in social acceptance for driving after a beer in the two cases (0.05 and 0.02).

The change from 0.08 to 0.05 is merely an adjustment to the cat and mouse game, while 0.02 does something to peoples attitude.

Posted by: Johan Borg at April 8, 2006 02:42 PM

yea man ... me sum times av a drink n drive but me as gotta get round innit

Posted by: james at July 20, 2007 11:57 PM
Post a comment

Anti-spambot Turing code

Creative Commons License
Except where otherwise noted, this site is licensed under a Creative Commons License.

The Social Affairs Unit's weblog Privacy Statement