Friday, July 1, 2011

When good people do bad things with data - wealth and children

I used to think the "Freakonomics" folks had a good thing going. The first book was relatively well researched, and, let's face it, the book was fun to read.

However, in the past five years I have been perpetually frustrated by the authors' repeated mis-application of very basic statistical/logical priciples. Namely the authors blatantly ignore the difference between correlation and causality. If my Intro to Sociology students can pick apart the difference between correlation and causality, you would think trained economists could do the same.

The latest blunder strikes me as particularly egregious: implying that children are an "inferior good" based on oversimplified interpretation of a couple of charts. Now, whatever your take on pint-sized people may be, I suggest that a chart showing a negative correlation between GDP-per-capita and children-per-woman only shows correlation, not causality.

Even the most casual observer can see that the countries with the low income/high fertility combination are still largely agricultural or in the very early stages of industrialization. Conversely, the countries with high income/low fertility are largely industrialized or post-industrial in their development. Basic demographics teaches us that fertility is higher in agricultural societies than in industrial ones, regardless of wealth. Fertility is as much about social and cultural structure as about income.

But even more fundamental to my argument: Something really fascinating happens to those post-industrial countries... At a certain point, the fertility rate starts going back up as income increases.

Let me put this another way. Income has continued to rise in the United States from the mid-1990s to today (recession aside) and through that entire period of time, with rising income fertility has increased.

You can see this trend in the screen show above by following the course of the yellow bubble (U.S.) from the early 1990s to 2008. And note, further, that the screen shot captured by the freakonomists shows only 2008, not the upward trend over the prior 15 years. (Or the drop in birth rate during the recession, which implies a positive, not negative, correlation between income and fertility in the U.S.)

Why is that trend not shown in their analysis?

Perhaps because it would ruin their headline.

(But maybe I'm just seeing a correlation between bad analysis and eye-catching headlines, and assuming causality...)

To read the original post see:
Or a useful summary by the NY Times.
Or build your own Gapminder plot.

1 comment:

  1. Good stuff, Beth. That was a wretched article and I skewered it just for fun in my intermediate micro class when it came out. Justin Wolfers is a very smart guy--certainly he's familiar with Becker's quantity-quality tradeoff story or he ought to be if he's going to write about the topic, but it doesn't even warrant a mention.


your insights?