Mon: New course: Street-Fighting Math
Tues: Paxil: What went wrong?
Wed: Pro-PACE, anti-PACE
Thurs: My namesake doesn’t seem to understand the principles of decision analysis
Fri: Risk aversion is a two-way street
Sat: A reanalysis of data from a Psychological Science paper
Sun: The devil really is in the details; or, You’ll be able to guess who I think are the good guys and who I think are the bad guys in this story, but I think it’s still worth telling because it provides some insight into how (some) scientists view statistics
Hey Andrew, this might grab your interest. In 2014, Greenwood, Jeremy, Nezih Guner, Georgi Kocharkov and Cezar Santos publish “Marry Your Like: Assortative Mating and Income Inequality”in American Economic Review: Papers & Proceedings, 104 (5): 348-353, claiming that assortative mating is a huge driver of the growth in inequality, massively raising the gini from 0.34 to 0.43. This gets very wide press (search for “assortative mating inequality”).
ungated version of original: http://www.econstor.eu/bitstream/10419/93282/1/dp7895.pdf
Turns out those numbers are wrong, after correcting errors they now think the difference is more like 0.42 to 0.43, or even 0.429 to 0.430. (“corrigendum”: http://pareto.uab.es/nguner/ggks_corrigendum.pdf). However, they point to a different paper which shows rising divorce rates among the poor “explain” a large share of inequality.
Maybe this is just a cost of AER P&P which presumably has a lower quality standard than a regular journal — not that the media treats it this way. It just happens to be a very consequential issue at the center of a large amount of public debate, so incorrect “facts” like these can be harmful. I was a little bothered by the approach of the authors, in that the tone of their non-retraction appears to be “we were mostly right.” They reported a huge number when the true number is basically zero. The divorce question is a very different question and plausibly caused by inequality or other factors and hardly validates the initial erroneous analysis.