Elephants and camels

by John Quiggin on January 31, 2004

Via David Appell, I came across this marvellous quote from Freeman Dyson

In desperation I asked Fermi whether he was not impressed by the agreement between our calculated numbers and his measured numbers. He replied, “How many arbitrary parameters did you use for your calculations?” I thought for a moment about our cut-off procedures and said, “Four.” He said, “I remember my friend Johnny von Neumann used to say, with four parameters I can fit an elephant, and with five I can make him wiggle his trunk.”

It came to mind when I read this story in the NYT with the introductory claim What really stimulates economic growth is whether you believe in an afterlife — especially hell.The report is of some estimations done by Rachel M. McCleary and Robert J. Barro (the story notes that the two are married) published in American Sociological Review.

Barro is probably the biggest name in the field of cross-country growth regressions (a field in which I’ve also dabbled), and I’m sure he’s aware that thousands of these regressions have been run and that, with very limited exceptions, results that particular factors are conducive to growth have proved highly fragile. I haven’t read the paper, so for all I know, the results have been checked for robustness in every possible way. But my eyebrows went up when I saw this para

Oddly enough, the research also showed that at a certain point, increases in church, mosque and synagogue attendance tended to depress economic growth. Mr. Barro, a renowned economist, and Ms. McCleary, a lecturer in Harvard’s government department, theorized that larger attendance figures could mean that religious institutions were using up a disproportionate share of resources.

What this means is that at least two parameters have been used in fitting growth to religiosity and that the two have opposite signs – most likely it’s some sort of quadratic. In my experience, there’s always at least one arbitrary choice made in the pretesting of these models (for example once you have a quadratic, the scaling of variables becomes critical). That gives three free parameters, if not more.

I’m no John von Neumann, but with two parameters I can fit a dromedary and with three I can do a Bactrian camel.

{ 7 comments }

1

Andrew Boucher 01.31.04 at 8:50 am

Church attendance and growth: I presume there is a time-lag between the two parameters – increase in church attendance followed by decrease in growth. If it’s just correlation that’s being announced, the causal effect could of course be in the other direction – decrease in growth causes increase in church attendance, for the obvious reasons.

2

Dan Simon 01.31.04 at 9:00 am

No way–I say a healthy Bactrian camel requires at least four parameters.

3

Gavin 01.31.04 at 11:43 am

For all Barro’s good points, his applied work is his weakest work. Just have a look at his growth textbook with Sala-i-Martin: it’s a tour de force, until you hit the empirical chapters and then it’s just a joke.

4

dsquared 01.31.04 at 4:14 pm

and I’m sure he’s aware that thousands of these regressions have been run

An understatement, of course; Barro’s co-author on the above mentioned textbook has alone been responsible for the papers “I Just Ran A Million Regressions” and the sequel “I Just Ran Ten Million Regressions”.

5

humeidayer 01.31.04 at 5:16 pm

I’ve noticed that every single person I’ve known who has contracted cancer has also worn shoes. And, while I know very little about String Theory, I will say that “11 dimensions” has never sounded very promising to me. I wonder what von Neuman could do with 11 parameters.

6

Abiola Lapite 01.31.04 at 8:15 pm

“And, while I know very little about String Theory, I will say that “11 dimensions” has never sounded very promising to me.”

It doesn’t work quite the same way, I assure you. Of course, you don’t have to take my word for it; just larn yesself some quantum field theory and a dash of algebraic geometry and you’ll find out all on yer ownsome.

7

Backword Dave 02.01.04 at 12:37 am

>Oddly enough, the research also showed that at a certain point, increases in church, mosque and synagogue attendance tended to depress economic growth.
Surely this is totally post hoc ergo propter hoc?
Surely a the most parsimonious explanation is that people in church, etc, aren’t working or spending. Sunday in IKEA = growth. Sunday in church = no growth. (Adjust day to suit religion.) Both experiences are, of course, miserable.

Comments on this entry are closed.