Freakonomics: Abortion Lowers Crime and the Iraq War is good for kids!

[Crispy]

It's an interesting book, mostly because it's an economist talking about driving factors that aren't normally considered by economists.

 

[kabbes]

(Also of interest: it was actually Alex B that recommended this book to me after he read a review of it when if first came out in the States. My version is the US version as it wasn't out in print in the UK when I ordered it.)

 

[Firky]

Going to see if I can it this afternoon

 

[zoltan]

Its a very good read and does allow you to think.

as EVERYONE seems to have commented, the NYC/ Crime link isnt projected as any kind of race issue, but the income demographics of who would benefit most by unfettered access to abortion speak for themselves

read it / then comment

 

[catrina]

I quite like the research they did, but I think it is an understatement to say they're pretty arrogant about their findings.

With the right sets of data, you can manipulate them to reveal what you want them to.

They seemed to think they'd solved all of America's problems right then and there.

So far they've predicted the American Democratic primary spot on, although now Obama's fundraising is exceeding Clinton's, it wasn't always. Their findings were that the most likeable candidate will win, regardless of how much money is spent. (With such few data on the subject, though, this prediction was a fair bit weaker).

 

[kropotkin]

I didn't think it was a good book- I thought it was reductionist and simplistic

 

[Poi E]

Does the book mention the removal of lead from petrol and the drop in crime rate in the US?

http://www.washingtonpost.com/wp-dyn/content/article/2007/07/07/AR2007070701073.html

 

[catinthehat]

I read it and found it interesting. Also rec it to my students who all declare hatred for their economics modules and it did spark some interest in terms of their "whats the point of it ness".

The sociologist in the chapter Why do Drug dealers live with their Mums" has a book himself about his research. Link:

http://www.powells.com/biblio?isbn=9781594201509

 

[Johnny Canuck3]

Been reading the blog in the NYTimes on and off now for some time and I am Quite interested in getting the book. So O have been reading some of the odd things it glues together with stats. Of course there's some contraversial topics on it such as the below:



Plus according the British Psychological Survey the Iraq War did wonders for kids:






There's some amazing stuff in there that has really changed the way I thought about things and probably would make many a liberal shit their pants in rage!

Has anyone read the book, if so is it worth the buy?

My kid wanted it for Christmas two or three years ago. She read it, and we talked about it some. I keep meaning to read it myself.

 

[pinkmonkey]

I hated the book. The author is so damn smug, some of the theories he puts forward are stating the bleeding obvious...and then he comes over all 'clever' for revealing them. Aren't I brilliant? No, not really, just incredibly smug.

 

[Poi E]

Aren't I brilliant? No, not really, just incredibly smug.

and rich.

 

[Chester Copperpot]

I've read it twice. Great book.

 

[co-op]

I can't believe that anyone takes this book seriously.

The abortion-crime case just isn't made at all. I have to admit I read it a few years ago so I can't remember the details of just how bad it is, but I was only a couple of years out of doing a psychology degree that was 35% statistics so I was fairly on top of my stuff and I thought it was woeful. He simply doesn't present any evidence whatsoever.

I finished the chapter (it's quite early in the book) put it down and didn't bother pick it up again, life's too short to read rubbish. The most extraordinary thing about it is the acclaim it receives.

Fwiw - I don't think that if he had made the case for Roe v Wade leading to a crime drop, that that would have been a fundamentally illiberal "fact" - good liberals tend towards environmental theories of crime causation so it doesn't seem necessarily conservative or racist or whatever else to me. It's just that he so utterly fails to make the case it's embarrassing. Or rather the embarrassment is the sense of complete smug he radiates when he reveals his "answer".

 

[kabbes]

I can't believe that anyone takes this book seriously.

The abortion-crime case just isn't made at all. I have to admit I read it a few years ago so I can't remember the details of just how bad it is, but I was only a couple of years out of doing a psychology degree that was 35% statistics so I was fairly on top of my stuff and I thought it was woeful. He simply doesn't present any evidence whatsoever.

I have studied and use day to day a hell of a lot more statistics than being "a couple of years out of doing a psychology degree that was 35% statistics". His case is perfectly cogent. And if you knew anything about generalised linear modelling then you would understand exactly how he went about analysing his data and why this is a valid technique well used across several financial industries.

If I were you, I'd suggest that you be a little less quick to assume that you are the great statistics I Am and consider the possibility that there may be things about the discipline of which you simply know nothing.

 

[co-op]

I have studied and use day to day a hell of a lot more statistics than being "a couple of years out of doing a psychology degree that was 35% statistics". His case is perfectly cogent. And if you knew anything about generalised linear modelling then you would understand exactly how he went about analysing his data and why this is a valid technique well used across several financial industries.

If I were you, I'd suggest that you be a little less quick to assume that you are the great statistics I Am and consider the possibility that there may be things about the discipline of which you simply know nothing.

I kind of knew it was going to be you who was going to quote me...

Easy tiger, no ones challenging you for the title of Stats Daddy.

I'm sure it lightens the dullness of a day at the office getting your wind-ups in, but it seems pretty obvious to me that I'm not claiming that I'm the "great statistics I Am".

That's why I used the line that you have quoted - "a couple of years out of doing a psychology degree that was 35% statistics". It means exactly what it says - I had a decent working knowledge at a reasonable standard.

As I also said it was long enough ago that I can't remember exactly how bad Freakononmics was, just that it was pretty striking. In all fairness to you and to this debate I suppose I ought to re-read it and tell you exactly why I think what I think - but I'm pretty sure the book has been dumped by now so I'm going to let myself off the hook. My recollection was that there wasn't much to argue over - it wasn't so much that he said anything wrong, it's just that the case was so thin, almost Correlation, therefore Causation.

However if the figures in http://www.isteve.com/abortion.htm are right it would seem that there isn't even a correlation (thanks to Idris2002 for the link) so he is just plain wrong.

Congratulations on using statistics a "hell of a lot" every day etc though.

 

[magneze]

Cracking good read and some fascinating theories although you do wonder if you could make the figures mean anything with the right interpretation.

 

[Idris2002]

I'll probably read it at some point, but it sounds very iffy to me.

 

[magneze]

I'll probably read it at some point, but it sounds very iffy to me.

Well worth reading ... certainly before dismissing out of hand ...

 

[Bob_the_lost]

That critique site is standing on some rather dodgy statistical ground in places.

Legal abortion is a major cause of what it was supposed to cure -- unwanted pregnancies. Levitt himself notes that following Roe, "Conceptions rose by nearly 30 percent, but births actually fell by 6 percent …" So for every six fetuses aborted in the 1970s, five would never have been conceived except for Roe!

WTF?

 

[kabbes]

He wasn't arguing causation at all. He was looking for provably significant rating factors and interactions for the metric "crime rate". He managed to isolate each of the various inputs claimed to be rating factors and show that these actually had no relationship with crime rate. He did find, however, that access to abortion had a strong significance as a predictor for crime rate movement.

The point is that access to abortion is therefore unarguably a statistically significant predictor of crime rate. How you interpret this link is up to you -- he suggested a possible (and likely, IMO) mechanism; namely that the aborted foetuses are precisely those that would have been more likely to commit crimes. However there are other possible mechanisms too.

An analogy would be car insurance. Age of driver is a strong predictor of accident frequency. It is therefore used by insurers as a rating factor for premiums. The insurers do not need to care about the mechanism that links age to accident frequency, however -- they merely need to know it exists.

 

[Maurice Picarda]

The book contains a handful of moderately interesting anecdotes about statistics which together would make perhaps two good magazine articles. Padding it out into a slim paperback was a dreadful waste of words.

 

[co-op]

He wasn't arguing causation at all.

Really?


I had such a strong memory that he was indeed making quite a strong claim of causation that as I walked past a bookshop this morning I had to duck in and leaf through a copy off the shelf.

So on pages 108-9 of the paperback he lists several popular media "explanations" for the drop in crime and then invites the reader to guess which one(s) of them have merit and which don't. He then writes:

"...one of the greatest measurable causes of the crime drop does not appear on the list at all, for it didn't receive a single newspaper mention" - and a quick skim-read forward reveals that this "cause" is indeed, as one suspects it will be, the legalisation of abortion.

Now obviously we've established that I'm just a miserable worm crawling in the fetid mud of statistical ignorance here but to me that sounds awfully like a causation claim.

 

[kabbes]

He also, elsewhere, is at pains to point out that what he is measuring are not causations. He repeats that multiple times.

Besides which, nothing that it could subsequently claim would invalidate the original analysis that shows that access to abortion is undoubtledly a statistically significant predictor of crime rate. That is what you were taking issue with before.

 

[co-op]

He also, elsewhere, is at pains to point out that what he is measuring are not causations. He repeats that multiple times..

Perhaps he does my lucky dip reading didn't pick that up (perhaps you have a reference handy?) - it looks to me as if he also makes a pretty bald claim of causation. Maybe he's just a bit confused as to what exactly he's saying?

Besides which, nothing that it could subsequently claim would invalidate the original analysis that shows that access to abortion is undoubtledly a statistically significant predictor of crime rate. That is what you were taking issue with before.

Well I suppose figures revealing that there was no correlation between crime rates and abortion rates 18 years earlier would tend to invalidate that claim (which is what the website linked to by Idris2002 suggests).

What I was taking issue with was not - as you say - the ability of 'access to abortion' to predict crime (I know little of criminology and have none of the figures to hand). It was the incredible thinness of the case - he had found a correlation and almost without bothering with much else he decided that that would do, thank you and good night.

You say that he analyses the various "ratings factors", yet what he actually does is take the various news-media stories and arrange them by popularity (ie number of stories). Well - that the news-media is bullshit on crime might come into play on that one. I'm also intriguied straight away for example by why there would be no similar pathway of "unwantedness of child" = 'increased likelihood of criminality' effect by the almost universal introduction of the contraceptive pill about 15 years before Roe v Wade and questions like that (in fact that massive reduction in conceptions seems to have co-incided with the huge growth in crime during the 70s and 80s, the reduction in which Levitt claims to have explained via legalised abortion). Never mind the total absence of anything relating to relative-poverty (degrees of wealth differentials - which also have fantastic correlations with violent crime in the US). And so on and so on. He either doesn't know about this stuff, or he just ignored it.

To be clear; I'm not saying the theory is wrong, it's just that it's so thin. I was left intrigued by the evidence but totally unconvinced and fairly sceptical about someone who was so ready to sound off so loudly in public on such a flimsy foundation.

And especially when this is hailed as terribly daring "iconoclasm" etc. Looks more like highly effective hype to me, to me his publishers PR dept come out of the exercise far better than he does.

 

[zoltan]

I think yer being a bit too picky here - the idea of the book is to think about things outside the conventional way of doing things
yes , its a bit hypey, but still interesting

anyway, hes an economist - economists are by definition unable to produce a definative answer!