The linchpins of science

I don’t read The Wall Street Journal very often but yesterday, during a lull in the action of the election (there were lots of them given the pace of voting—about 12 voters per hour), I read a copy that the local precinct captain had brought in.  It had a front-page article (subscribers only), “Is an Economist Qualified to Solve Puzzle of Autism?” which I thought deserved additional comment.

The article described the work of economist Michael Waldman in which Dr. Waldman used rainfall as an instrumental variable in analyzing the causes of autism.  The article characterized the use of instruumental variables as

a statistical method that, by introducing some random or natural influence, helps economists sort out questions of cause and effect.

you can find a more complete description here and in Dr. Waldman’s paper (you can find it here) he analyzes patterns of rainfall in Oregon, Washington, and California and the incidence of autism.  As I read Dr. Waldman’s reasoning goes roughly like this:

1. When it rains people watch more television.
2. In the areas of Oregon, Washington, and California where it rains more people watch more television.
3. In the areas of Oregon, Washington, and California where it rains mor there are more diagnoses of autism.
4. Consequently, watching more television causes autism.

There are all sorts of problems with this line of reasoning and some have been pointed out by others.  They include the fact that television watching isn’t the only thing that’s different between areas where it rains more and areas where it doesn’t, for example such areas also have more household mold, people have less exposure to natural sunlight, they run less, and a host of others which could, presumably, be equally causative.  In addition, I suspect that there are significant genetic differences between the people living in those areas and people who don’t, due to different prevailing ethnic backgrounds.  But the point I want to make is somewhat different.

When I studied economics (before the glaciers receded) it was immediately before the point at which econometrics (the use of mathematics, particularly statistics, in economic analysis) overwhelmed the field (which apparently occurred at some point when I wasn’t watching closely).  I had surmised this was going to be the case and, as someone who was studying applied mathematics, I thought that economics was a good place to use my skills and that it was pretty obvious that econometrics would be increasingly important in economics and I might be able to get ahead of the curve.

I was rather appalled by economists.  I didn’t find them to be particularly good mathematicians and, worse, I didn’t find them to be particularly good scientists.  Science actually has two linchpins,  observation and prediction, and I found economists to be lousy at both, much more akin to the ancient Greek philosophers who believed that anything that couldn’t be determined that anything that couldn’t be determined by pure reason and the amount of observation you can do from your armchair wasn’t worth studying.  Perhaps that’s changed in the years that have passed.
In the particular case of this study, all observable phenomena are either positively or negatively correlated and, as has been observed, correlation is not causation.   The study does not determine that the children with autism watched more television than those who did not only that children who lived in certain areas did.  Sometimes you’ve got to get out of the armchair and count the teeth.  It doesn” make a damn bit of difference if kids, generally, watch more television in those areas if the kids with autism in those areas didn’t.

The focus of the WSJ article isn’t just about whether television causes autism it’s about the contributions that economists might or might not be able to make to the study of autism and other complex phenomena, generally.  Genuine expertise isn’t just in knowing the ropes it’s in knowing the ropes to skip, as well.  In the particular context of autism there is a pretty well-established genetic link:  siblings of kids with autism are more likely to have autism than risk in the general population, identical twins of kids with autism are highly more likely to have autism.  Non-identical twins of kids with autism are actually less likely than other siblings to have autism themselves.  Explain that by television watching habits.

My own view of autism BTW is that it doesn’t have a single cause.  It’s probably some combination of genetic predisposition, developmental disability, perinatal or other environmental conditions, etc.  There might even be a kid out there somewhere whose autism was actually caused by the thimerosal used as a preservative in the innoculations they received.

I believe that mathematical methods and genuine empiricism, rooted in observation and proven by predictive ability, have a lot to contribute to all of the sciences.  I don’t think we should be confused into thinking that because Pasteur was a chemist and contributed enormously to medicine (to the dismay, rejection, and condemnation of medical doctors of the time) that all chemists know more about medicine than any doctor.