Wednesday 13 January 2010

Causality revisited

A while back I whinged that the press does a bad job of distinguishing correlation from causality. The latest headlines on correlations between TV-watching and mortality risk are no different: absent some kind of natural experiment or clever IV strategy, all we really know is that sedentary folks are more likely to die than folks who aren't.

For even more fun, check the shifting standards in the following paragraph:
Alcohol factors significantly in violent crime statistics: one third of all offences and half of all serious violent offences in NZ are committed by offenders who have been drinking (NZ Police); alcohol is associated with 46 percent of incidents of sexual violence (National Survey of Crime Victims, 2001). By contrast, nothing in the NZ Drug Statistics indicates cannabis use a cause of violent crime. Any association between cannabis and violence is linked to the black market and is therefore a direct result of prohibition itself.
So alcohol is bad because there may be correlations between drinking and criminal activity, but marijuana is good because nobody has established a causal relationship between weed and crime. As I'd previously noted:
...high proportions of arrests and crimes that are "related" to alcohol. But what I've never seen numbers on is what proportion those "alcohol related" crimes would have taken place in the absence of alcohol. You could just as easily say that 100% of crimes are "Oxygen related", 50% are mullet-related and 25% involve the wearing of t-shirts that say "No Fear". Ok, without the oxygen, none of those oxygen-related crimes would have taken place. But would a policy getting rid of No Fear t-shirts really prevent crime? Mightn't criminals switch to even riskier shirts? This is actually a reasonably serious point. If you ramp up the price of alcohol, how many criminals switch to P before going out to do nasty stuff? One of the country's more prominent criminologists suggests in informal chats that he reckons about 60% of these "alcohol-related" crimes would have taken place anyway even if alcohol disappeared; the actual crime-costs of alcohol are then about 60% lower than typically reported as alcohol in those cases is more like wearing a No Fear t-shirt: just something criminals like to do while offending but wouldn't stop them from offending in its absence.
I always find it depressing how folks like NORML think beating up on alcohol makes marijuana legalization more likely rather than just making for tighter regs. The hospitality industry lobbies for more restrictions on supermarket-bought alcohol to boost sales in bars; small brewers push for more punitive tax rates on big brewers... the only winners are the healthists who get support bit by bit for more regulations on everything. It's like a bunch of folks on the scaffolds complaining that the other guy's noose isn't quite tight enough. Y'all might instead direct your attention to the hangman sometime and try helping each other cut those ropes.


  1. Yeah, it's the old one: one-third of road accidents involve alcohol, so two-thirds must be caused by plain old stupidity. We'd have a much better road accident rate if we cracked down on stupidity.

    Your shifting standards paragraph reminds me of this complete bollocks by a serving MP and PC.

    Note the line in there:

    "People have died as a result of taking ecstasy and committed crimes under the influence of cannabis."

    I can only think of one crime someone under the influence of cannabis might commit, and that's possession of cannabis.

  2. Yeah, I saw this article on TV last night, and mused it wouldn't be long to pop up on this blog.

    Your points on causuality is on the button though. Because you can't have data on everything to answer every question, policy decisions (public & private) usually get made by influential anecdotal evidence, or formed through a consensus process of people 'directly involved' in the problem policy is meant to address. Those discussions probably start off with everybody convincing each other about issue causality when the skant available information is purely correlation. Its worse in consensus decision-making, with the potential for peer-pressure to slip in.

  3. I sort of take your point, but there is shoals of evidence that alcohol _causes_ road accidents and violence; and indeed this is precisely what would be expected from a knowledge of its psychopharmacology.

    Taking all the causal evidence, alcohol really is one of the worst drugs - harming and killing users (medically and by accidents and violence) in very larger numbers, and also large numbers of bystanders, and causing a huge amount of social pathology - especially in the UK and Ireland.

  4. @bgc: I certainly wouldn't disagree that drink driving is far riskier than sober driving, or that alcohol is causal in some offences. But some (small) proportion of drink-driving accidents would have taken place even if the driver had been sober and some (much larger) proportion of crimes would have taken place even if the criminal had been sober. So, in the costs of alcohol report that I previously fisked, where the consultants figured that each and every crime where the criminals, on survey, said that alcohol had contributed at least "somewhat" to offending would disappear in the absence of alcohol, I pushed the bar up a bit to ones where offenders said it contributed at least "a lot". Both are very imperfect standards, but I think mine is more realistic.

    I could pretty easily be convinced that marijuana is safer than alcohol. But I worry that the strategy of demonizing the other intoxicant results only in the regulation and prohibition of both.

  5. Their submission pushes on the whole for greater regulation rather than the status quo or lesser regulation: a ban on advertising and "glorification" of alcohol is hardly liberal.

  6. I worry less about the alcohol submission and more about the overall strategy of "look how bad alcohol is and it's legal, why can't marijuana be legal too since it's safer?" when the pols' ears shut before the comma....

  7. I think the alcohol comparison is something we're not going to get rid of, and NORML do need a different strategy. I think there's a greater need amongst the politicians to pay attention to public perceptions and the Single Convention on Narcotic Drugs than to common sense.

    I do think that alcohol advertising and promotion needs to be subject to tighter regulations - actually, advertising and standards on television and in the media altogether. There's a key standard in the advertising rules on alcohol in a lot of European countries - you can't advertise alcohol like it's going to make you socially or sexually successful, or more attractive. In the UK, you also can't feature people apparently under the age of 25. There's a number of commercials in circulation in NZ that break this standard. From a social norms perspective, advertising can be a great influence on behaviour, but banning it outright won't change a culture, just like banning cannabis hasn't changed that part of NZ culture.

    The NORML argument is confused because it has no clear outcome - you're right, they make the comparison, but they don't give a clear direction to lead out of the comparison and into the point that harm reduction isn't achieved by banning a substance.

  8. I wonder what would happen if one of our brewers started airing ads that prominently featured some of the research that suggests social moderate drinkers earn more than abstainers, complete with footnotes....