December 30, 2006

Summer Reading

I've spent the last week working my way through The Time Traveller's Wife by Audrey Niffenegger. It's a shame that much of it doesn't live up to the intense breathlessness of the first hundred or so pages, but then, the first hundred or so pages more than make up for any other shortcomings it might have. Maybe it's just because of my own experience with a long distance relationship, but the feeling of love and longing that Niffenegger captures, right off the bat and without any apparent effort, is truly startling; and the way she expertly drip feeds you the exposition of her fairly kooky premise lets you very gently get to know the characters and their universe, at the same pace that they're discovering each other. It's the sort of opening that should make other writers hopelessly jealous.

And even though it does drag and digress in places, every now and then Niffenegger will just really nail something - a description, a feeling, an encounter - and it pulls you straight back in. Plus, the narrative device she's built into the story has enough potential for intrinsic interest that you never really want to put it down, anyway.

I haven't quite finished it yet, and I'm getting an ominous sense of a disappointing ending approaching, but unless it's truly horrific I can't see myself doing anything except recommending the book with much enthusiasm.

In other news, I also read Freakonomics, but my opinions on that are, predictably enough, fairly lengthy, so I've hidden them after the jump (just click on the post title).

In the meantime, happy new year!

So I finally sat down and read Freakonomics on the plane; I felt like, as a pithy and opinionated blogger educated in the social sciences, I really kind of had to.

It was, I will admit from the outset, an engagingly written piece of work. Having said that, I also found it at times to be simultaneously bland and infuriating, mainly because so many of the “mind-blowing” revelations that the authors are praised for peddling are pretty bog standard stuff if you’ve done even a little sociology (or economics, or whatever). I mean, they have whole chapters that basically boil down to “socioeconomic status has some effects on peoples’ lives”, and I have a hard time seeing why demonstrating this using some vaguely cute statistics has earned the authors such accolades. Of course socioeconomic status has some effects on peoples’ lives: that’s why we use it as an explanatory concept in the first place.

Worse, though, is the brutish way in which they take their narrow (by their own admission) conceptions of the issues and extrapolate from the results as if one set of numbers explains everything there is to know about a phenomenon.

A good example is the section they have comparing swimming pools to guns in terms of potential for accidental death. As it turns out, children are substantially more likely to die in a swimming pool accident than in a gun accident, and the authors use this ‘startling fact’ to overturn conventional wisdom and have a good laugh at anyone – chuckle – who has ever been uncomfortable with the idea of their chilrden playing with firearms (the simple-minded dolts).

And, sure, I guess it does prove something sort of useful if parents end up being a little more vigilant when their kids are near water. But it takes such an arbitrary slice of such a complex issue that it ends up just being a meaningless (if catchy) factoid. So what if more children die by water than by trigger? This really says very little (if anything) about the relative merits of swimming pools versus handguns, and I think I can safely say that, Freakonomics notwithstanding, we would all rather there was a pool in every backyard than a gun in every closet. It’s like arguing that bananas are more dangerous than rat poison, because you’re more likely to slip on a banana peel than on a rat poison box: true, in some odd sense, but only if you’re looking at the picture through a pinhole.

The swimming pool problem is actually a neat example of the duo’s undoing later on in the book, too, when they take great glee in demonstrating that there’s no such thing as “good parenting” – how well a kid does depends on things like parental education and income, rather than on any active parenting techniques (like museum visits, encouraging reading, and so forth).

Or so they say. Again, this feat of mind blowing is achieved largely through a bizarre transformation of the issue into something to which Levitt can easily take his sledgehammer of a statistical analysis. The entire argument hinges on a study of school attainment in the US, which shows, ceteris paribus, that there is no statistical relationship between high test scores and a whole host of ‘good parenting’ practices (like, for instance, a parent who stays at home to care for a child instead of taking paid employment; or, again, like museum visits and encouraging reading).

But it’s exceptionally poor statistics to then imply that parenting practices have no effect on anything: they may not have an effect on test scores, but we can’t say much more than that. The obvious counter example would be the parents who, following Levitt and Dubner’s earlier advice, build a fence around their swimming pool to prevent accidental drowning. This arguably constitutes excellent parenting; just not the kind of parenting that would be reflected in a child’s test scores (except insofar, perhaps, as that child actually has test scores). Indeed, not only are there any number of demonstrably positive parenting practices that are unlikely to affect test scores (cooking healthy meals every night instead of relying on microwave dinners, for instance); there are also plenty of other ways to assess how well a child has been raised, and how often a child visits museums (for instance) may well have an effect on these alternative measures. Besides, how often does somebody look at a straight-A report card and say, “My, what a well-bred young man”? High test scores are not something I would intuitively regard as the indicator par excellence of good parenting.

It’s the same problem, over and over again: the authors reduce a complex issue to fit the limits of their data, then (undeservedly) expand their conclusions back to the problem’s original terms. And, yes, they do acknowledge that their data is often just “a proxy” for what they’re really trying to study, but without any intelligent discussion of what this practically means for their results (namely, that they don’t directly measure what the authors blithely trumpet as the object of analysis).

And this brings us to the nitty-gritty: the somewhat inflammatory statement made in Freakonomics that legalised abortion is what caused the fall in crime in the mid-Nineties (because, the reasoning goes, those most likely to have been aborted are also those who would have been most likely to become criminals had they instead been born).

I have a couple of problems with this argument, not because of any moral indignance at the idea, but because it is, again, an excrutiatingly narrow (and methodologically messy) approach.

First of all, let’s talk about crime rates. Crime rates take the total number of crimes committed over a given period of time and in a specified population, and then express that figure relative to that same population. So a crime rate of, say, 20% means that for every ten people in a population, two crimes were committed over the period being studied. If the crime rate then falls to 10%, it means that for every ten people only one crime was committed over the period being studied. It doesn’t necessarily mean that there were fewer criminals, because that’s not what crime rates actually measure (though of course the two are highly correlated). They also don’t measure the likelihood of a particular person committing a crime, or the likelihood of a particular person being the victim of crime, even though these are common intuitive leaps when thinking about crime rates (that’s why governments are so fond of publicising their fall).

I point all this out not to dispute the truth in L&D’s numbers, but to try and make clear how little crime rates really tell us about reality; once we start to question the ironclad "truth" of the statistic, we can start to appreciate some of the deeper problems in the abortion-crime argument. Crime rates, for instance, tell us nothing about the qualitative nature of crime. Consider: if ten crimes are committed on a given day, these might be ten completely unrelated incidents, or they might have been concerted as part of a organised crime syndicate. The former is, of course, lamentable, but the latter is arguably the more troublesome situation; organisations, as any good sociologist will tell you, have a strong tendency towards self-preservation, and a well-established crime ring will be far harder to eliminate than so-called ‘opportunistic’ crime. (Organised crime syndicates, once ‘busted’, often simply find a new way to make money illegally, rather than disbanding altogether; that’s why the Mafia continues to exist decade after decade, informant after informant.)

In fact, Freakonomics itself provides a good example of this in its chapter on crack dealers. The Eighties and Nineties saw a big jump in numbers of crack gangs in the US. This was undoubtedly reflected in the skyrocketing crime rates leading up to the mid-Nineties, but when those crime rates began to fall again, crack gangs didn’t cease to exist. On the contrary, even while New York’s homicide rate fell by almost seventy-five percent, cocaine related arrests fell by barely fifteen. So the much-ballyhooed fall in crime rates masks the fact that a greater proportion of all crime is now linked with well-organised criminal gangs that have a propensity to self-replicate. Can we really then say that abortion has "reduced" crime?

The answer is no, because abortion rates, of course, tell us nothing much at all about crime at all. Why would they? That we can demonstrate a statistical relationship between abortion rates and crime rates is interesting, but it is hardly conclusive. Nor is it instructive: pointing out large scale demographic trends is not a useful way to generate anti-crime ideas, unless you're an advocate of pre-emptively aborting or locking up low income individuals, just in case they commit a crime at some point in the future.

Even supposing we simply let the figures guide our anti-crime efforts towards the specific chunk of the population identified by L&D's abortion figures (low-income, single parent families), we run into the same problem outlined above: the abortion-crime link tells us little about the qualitative nature of the crimes that would have been committed by the aborted children of the Seventies. Murder? Rape? Aggravated assault? Burglary? Grand theft auto? Admittedly Levitt’s original study does differentiate between property crime and violent crime, but only to show that they are both affected by the abortion rate, and what good does that do? There is still a large qualitative difference between burglary and murder, and the two obviously require very different prophylactic strategies (it’s a trite example, but closing your windows when you leave the house will reduce your chances of being burgled; it won’t reduce your chances of being murdered).

The abortion-crime link also tells us nothing about why people might commit crimes. It’s not enough to say that they’re from low-income, single parent families: that may make an individual statistically more likely to have committed a crime, but it’s not a motive. Its laughable that L&D omit this from their discussion-- how can anyone claim to explain why fewer people are committing crime without reference to motive? It's abusrd!

In summary: boo on mindless use of statistics. Analyses like Levitt's can be useful in making people think about topics in new ways, but they are never and should never the faites accomplis that Freakonomics makes them out to be. Any social scientists who use statistics in their work need to be able to back up their numbers with a rich understanding of the reality that those numbers attempt to describe. Frankly I don't get the impression that Levitt really wants (or even knows how) to do that.

1 comment:

Anonymous said...

I loved this book! I too was worried about the ending because there is such a sense of inevitability that you can't imagine how she can find a creative out. Let me know what you think.

Happy New Year! See you at Vaganza!

Post a Comment