Precautionary Principle Again

More Sorry Than Safe ... [Sir Colin Berry] cites the controversial issue of SIDS - Sudden Infant Death Syndrome, aka cot death - about which parents are given lots of often contradictory advice. Berry says that in the 1980s, the favoured precautionary measure to guard against the possibility of your baby falling victim to SIDS was to lay her on her side or front. 'We tended to consider babies and young infants as being rather like the unconscious patient', he says, 'where it is not clear that all the reflexes around the nose and mouth, for breathing and swallowing and so on, are finely tuned. So parents were told to put babies on their side or front, as you would do with an unconscious or stroke-troubled patient. It seemed like a reasonable, precautionary measure to take. Now we know that, in fact, it cost lives.' Berry says that subsequent observations made in Australia and New Zealand, and a case-controlled study in Britain in the 1990s, showed that reversing this policy and putting babies to sleep on their backs instead reduced the death rate from SIDS. In the UK, it fell from about 1,300 to 1,400 a year to about 300 to 400, he says. 'With the best intentions the precautionary measure of putting babies on their sides or fronts caused misery; a great many precious baby lives were lost because of what seemed like a reasonable precaution. It was one of those things that just happened to be wrong. This shows that we need data - that being precautionary, taking safety measures without testing the evidence, is not enough.'

This is pretty much a standard anti-precautionary principle article. The PP debate is generally fairly unproductive, as you get people like Berry bashing the idea of "take no risks at all without 100% certainty" and PP advocates suggesting that the only alternative is a reckless taking of any risk, and nobody able to offer a clear guide to what middle ground they think is appropriate (pieties about the need for "good science" and "empirical evidence," such as those Berry offers, do not suffice).

But it's still possible to see good and bad arguments made. Let's start with the bad one in the bit of the article quoted above, in which Berry suggests that the PP led to SIDS deaths. To me, it sounds like a good example of prudence. Initially, we had uncertainty -- we didn't know what sleeping position was most likely to prevent SIDS. By analogy with stroke patients, it seemed like sleeping on their side would be good. Since that was the best evidence available at the time, it made sense to tentatively recommend side-sleeping. Then, while most parents took the most promising-seeming course of action, a subset did a controlled test to gain more information, testing the initial hunch about side-sleeping. In this particular case, it turned out that the hunch was wrong. Armed with this stronger evidence, we were able to recommend having babies sleep on their backs. Would Berry have recommended recklessly sleeping any which-way from the start, despite some evidence suggesting side-sleeping was better? Now, if the PP had been used to stymie the experiment -- arguing that it's too risky to ask test subjects to sleep on their backs -- then he would have a case, but that doesn't seem to have happened with SIDS.

Unfortunately, this sort of "play it safe, but do some tests" strategy doesn't work in many cases. We only have one global climate system, for example. If there had only been one SIDS-prone baby in the world, it would have made sense to have it sleep on its side, even though with additional information we know that's a bad idea.

He says that those who opted to travel by road rather than rail following the Hatfield train crash of October 2000, which killed four passengers and injured 30, had in fact exposed themselves to an increased risk of injury or death. 'Road accidents kill more people than railway accidents do', he says. 'Yet because there is a perception that rail travel is unacceptably risky, some people opt to go by car instead. But the death rate on the road per billion person miles travelled is about 12 times that of the railways.'

I know it's always fun to point out that driving in a car is more dangerous than whatever other activity people are afraid of, but this doesn't seem to be a case of the PP at all. The PP here would advise travel by train, as that's less risky. The people avoiding the train are miscalculating their risks. Naturally the PP isn't immune to GIGO. It's also possible that, to these commuters, all deaths are not alike -- perhaps they, for whatever reason, fear death in a train wreck more than they fear death in a car wreck. Given that utility function, they may be perfectly rational in deciding to drive.

Later on he gives a somewhat better example:

It has since been discovered that the [cholera] epidemic was, in part, a result of the Peruvian authorities' decision to stop chlorinating drinking water supplies - and that one reason they stopped doing this was because reports issued by the American Environmental Protection Agency had claimed there was a link between drinking chlorinated water and an increased risk of cancer (a link which the EPA has since admitted is not 'scientifically supportable'). 'Chlorinated water would have prevented the outbreak', says Berry. 'The water production and cleaning system had gone wrong before the outbreak, so it wasn't just that they stopped chlorinating water and then, bang, cholera arrived. But in a deteriorating situation, the failure to chlorinate - based on the principles of precaution and bad science - helped to make things a whole lot worse than they might have been.'

It's possible that this is just another story of a time that things would have worked out better if we had taken a risk. The PP doesn't deny that such instances exist -- it argues that such instances are outweighed by the times that taking a risk makes things worse. No amount of anecdotes can tell us whether the precautionary principle or its reverse systematically leads to worse outcomes (I'm sure PP advocates have a load of stories about new chemicals being introduced that wound up killing more people than they saved due to inadequate pre-release testing). Depending on how assertive the EPA document was, it may have been perfectly reasonable, given the information at hand, for the Peruvian authorities to believe that not chlorinating gave the highest expected outcome -- they certainly didn't have time to go back and double-check the EPA's studies to find out that the original claim about cancer was wrong (which fact is the crux of the anti-PP argument -- if chlorination was carcinogenic, the PP could very well have saved lives).

Nevertheless, the cholera example points to an important assumption underlying many strong applications of the PP. It's often taken as a rule that leads to committing sins of omission in order to avoid sins of commission. The Peruvian authorities seemed more concerned with not actively creating new problems (cancer) than they were about alleviating old problems (cholera). Ceteris paribus, we should weigh the two sins equally.