Friday, November 30, 2012

Kevin Drum has an outstanding post on lead there.  I have a mediocre comment there and here.

When I finished the Steinglass quotes I was saying "lead lead lead" to myself.  But your post precedes this comment.   I agree both with your specific hypothesis and with the general argument that if the same thing happening all over suggests surprising simplicity not surprising complexity.  

I think that Steinglass did an OK job*, but has the problem of too much respect for conventional wisdom.  He notes that there seems to be something wrong with the conventional stories for New York and Boston, but then concludes that there is a deep puzzle and not just an odd failure for people to confront their hypotheses with any data except that which prompted them to form it (oh hell or for New Yorkers  to  really deep down believe that there are human beings outside of their city if they are New Yorkers).  

I think there are reasons why the lead hypothesis doesn't get the respect that your wonderful graphs show it deserves.  It is deeply offensively materialistic reducing human behavior to biology and chemicals.  This is a common problem with hypotheses which fit the facts (see also anti depressants do indeed work).   It's also painful to believe that so many people died and suffered because of something so almost exactly useless as putting lead in gasoline.

Anyway to add something, there is another common factor for Boston and New York and now that you oddly don't mention it Los Angeles.  William Bratton http://en.wikipedia.org/wiki/William_J._Bratton
(who I see is Sir William Bratton in countries which allow titles of nobility since Elisabeth II made him a Commander of the British Empire). I might add it is also odd that you don't mention the fact that one of the things he did in Boston, New York and LA was to uh use computers to allocate resources  and provide incentives (called CrimeStat).  Not really odd for a computer dude from the LA area.  Yes Bratton doesn't explain DC, Canada, Britain, Finland or the 1920s crime wave so your allocation of valuable pixels makes perfect sense.  I am typing about him, because pixels down in the comment thread are not so precious.

*Re-reading I do have to qualify the "OK job" attempt to be polite.  First he just guesses what results of experiments which weren't performed might have been.  It isn't really OK to discuss the proper analysis of made up data.   Also the only way that a study can show that something doesn't reduce crime is to provide statistically significant evidence that it increases crime.  The proposed interpretation of a hypothetical study in which an intervention did not have a statistically significant effect would be an incorrect interpretation of such a result.  Steinglass expresses skepticism about the experimental method and statistics.  He does so by imagining hypotheticial experiments and suggesting that statistical analysis must include the gross elementary error of rejecting the alternative hypothesis.  I think OK may have been a bit generous.

No comments: