(Another) Bad Day for Economists

One interesting project for future intellectual historians will be figuring out how economics became the queen of the social scientists when virtually none of their predictions have come true and so much of their empirical work is downright shoddy.  Perhaps it will lie in the way ideology can take over the discipline because of data specifications and econometric choices.  But sometimes it’s just simple ignorance.

Earlier today, we discovered that the famous work by Kenneth Rogoff and Carmen Reinhardt, purporting to show that nations whose public debt exceeds 90% suffer reduced growth, has essentially been a sham.  R & R’s attempts to respond have sunk them further in the mire.

Looking through the blog of the Shrill One, however, I discovered something even more disturbing over the long term for a land use scholar: the commonly-used measure of population density might be completely misleading.  The Census looked at population density as a function of census tracts, e.g. what is the density of this census tract etc. etc. Fair enough.

But then when researchers looked at population density over several census tracts, or calculated them through regions, they essentially treated each census tract as equal.  You have one census tract with density X, and another with density 3X, and that gives you an average density of 2X.  Notice the problem?  Not all census tracts have the same population: they are geographical, not demographic.  It’s nice to know that the Census is now trying to weight larger and overall density numbers by population: the question isn’t the average density across census tracts, but what is the density that the average person lives in.

But why in the world has that not been done before, or at least made clearer to cavemen such as myself.  It’s as if you have one guy with 400 at-bats who hits .250, and another guy with 100 at-bats who hits .400, and you say that the team overall has a .325 batting average.  It’s totally deceptive.  And — to the point of this blog post’s title — it appears to be precisely the same sleight-of-hand that Rogoff and Reinhardt performed in their now-debunked study: in calculating countries’ growth rates, they counted each country as one data point even if some countries had many more years of data.  One year of New Zealand, and 19 years of Great Britain, in their scheme, equals two data points.  That seems insane.

Two great morals of the story at this point:

1.  We are really going to need to be extremely careful in assessing land use, planning, and growth policies based upon density numbers if there are these distortions in the data; and

2.  People love citing economists and love trashing lawyers, but the lawyers might be better at what they do than economists are.

Not a bad couple of lessons for the day.

(Over to you, Matt!).

Carmen Reinhardt, census, Kenneth Rogoff, Land Use, Paul Krugman, Smart Growth, urban economics