Leading Article: Some more research that leads us nowhere

AN UNPUBLISHED research paper produced by two American academics, and now circulating among economists and criminologists, makes the claim that abortion, by removing the unwanted children of teenage, poor and minority women, removes a potential criminal cadre.

Steven Levitt and John Donohue argue that "abortion provides a way for the would-be mothers of those kids who are going to lead really tough lives to avoid bringing them into the world, the ones most likely to have been unloved by their mothers, to have faced intense poverty, to have had tough lives". Mr Levitt added that" "I don't think it's our job to withhold the truth because some people are not going to like it. I just think it's important to understand the impact of social policies."

So it is. But embedded in Mr Levitt's apparently high-minded remarks lies the assumption that their work provides all, or even part of, that understanding. We should be very sceptical about such claims, and not just because of the potentially unpalatable conclusions that their work has drawn.

We cannot know if their conclusions are statistically significant. Who knows what other factors the authors may - through no negligence on their part, but simply because of the complexity of the raw material - have failed to take into account? Research like this begs as many questions as it answers, and is rarely "conclusive". "Research shows..." is a phrase that crops up in arguments on everything from linking BSE to new variant CJD in humans to the timing of the next stock market crash. Typically, it settles nothing.

But even if - and it is a very big if - this research were in some sense perfect, it remains a very weak underpinning to derive policy from. We could, after all, end human hunger, poverty and misery as well as crime by making abortion compulsory. It is that absurd. Abortion, above all else, is a deeply moral issue ill-suited to balancing bits of social science research. Time to free the debate from the tyranny of statistics.