Saturday, 10 July 2010

Lies and Statistics

I never said I'd only be reviewing newly-released books; here's one from 2005*.

Freakonomics, by Steven D Levitt and Stephen Dubner.
I am by no means an economics buff. In fact I'm mildly hostile to most economics, parading its theories as science, often justifying the worst excesses of government economic policy with ideas that soon go out of fashion.
This book is definitely not mainstream economics, though, and earns its subtitle - 'A rogue economist looks at the hidden side of everything'. This is a book which challenges accepted wisdom.

Some of the results of Levitt's analyses are predictable. For instance, who doubted that estate agents lie, not only to buyers but also to sellers? Similarly, it comes as no surprise that the career structure in crack gangs is exactly analogous to that in McDonalds - a few people make serious money, the vast majority make almost nothing. The personalities of those at the top of the heap in either organization is no doubt also comparable; fans of The Wire will probably recognize the similarities, in Levitt and Dubner's profile of a business-studies-educated crack boss, to the Wire character Stringer Bell, probably as nice a sociopath as any investment banker you'd ever care to meet.

Other analyses do not, on reflection, surprise, but it's interesting to see them demonstrated in action. The '9/11 effect' is named for the spike in ordinary human decency that followed the attack on the Twin Towers; people feel mutually aligned in a condition of common threat.

Others are decidedly counter-intuitive. For instance, it seems that the death penalty has no deterrent effect on would-be murderers. Levitt and Dubner write: 'Even in a death penalty advocate's best-case scenario, capital punishment could explain only one twenty-fifth of the drop in homicides in the 1990s.' (p125). So if we want judicial killing, we'll have to come up with other reasons or base punishment policy on a lie.

Which brings me to the argument that makes up the core of the book. Many of Levitt's studies use regression analysis, the careful parsing of groups of statistics to freeze all but one of the many variables that apply to the raw data of socio-economic research. This is the chief tool Levitt used to approach the vast, multi-variable question of the spectacular drop in crime in 90s USA. To cut a long story very short, it seems it was not due in the main to new policing strategies (such as New York mayor Giuliani's programme of 'zero tolerance') but to the increased availability of abortions from the late 1970s onwards.

The chain of causality and its underlying social tragedy are easy to see: More women who did not want, and could not bring up, a child, had abortions. For obvious reasons, the main distinguishing socio-economic factor was poverty. Combined with, or as the main cause of, a mother's inability to raise a child, is it any surprise that these conditions predict behaviour in later life that attracts the attention of the law. Another study, applied to the years of the Ceausescu regime in Romania, supports this connection.

There is a grim satisfaction to the discovery of a truth that is going to be hated by such a wide spectrum of people; and hated it was - the campaigns against Levitt became angry and personal. The public reaction to certain kinds of uncomfortable facts or well-researched theories tells us why politicians lie so much. Most of them seem to take the view that it's their responsibility to keep the public away from uncomfortable truths, and this is one of the weaknesses of democracy - if you're a politician, and you don't lie about some issue, your opponents will, and it will be they who get elected. This leads to a consensus of misinformation, a political culture of mendacity. A recent UK example was the disgraceful sacking of David Nutt, the Government's chief scientific advisor on drugs. He published and defended a well-evidenced scale of the levels of damage caused by drugs both legal and illegal. The best knowledge that could be gathered about this subject did not agree with the official picture, the consensus lie, and Prof Nutt was given the boot.

So this book is good, because it helps us to think. It is, ipso facto, rather bad for the stupid and those who would lead them.

*On loan from my local library. Apparently, library book borrowing increased by 1% this year, the first increase in a decade. Get some in now, before the government gets down to its real function and kills off this too.

1 comment:

  1. Thanks for this review, and reminding me about this book. As an economist (a heterodox one, I promise) I usually sigh with boredom as correlations between events gets wheeled out as causality by orthodox economists waving their regression analysis like some kind of infallible truth wand. But these authors make a fine effort in arguing for potential cause and effect, and even better don't parade themselves as the last word on the subjects.

    I agree that some of the examples they investigate seem a little obvious but I think that even these serve a useful purpose: to demonstrate to the reader the authors' logic, and so the more interesting issues and the ones we may not have thought of ourselves get lent more credibility as a result.

    Definitely worth reading, imo - interesting for non-economists and probably incredibly annoying for mainstream economists. Good, I hope they splutter their tea and short their orthodox regression analysis software :)