by the Sandwichman
Tim Worstall at TCS Daily gets all dewy-eyed about the "American Social Model", based on the TRUTH ("There's only one small problem with this idea. It turns out not to be true."), FACTS ("there's one uncomfortable little fact...") and PROOF ("The latest empirical proof") contained in the Gary Becker-inspired data torturing exercise performed by Mark Aguiar and Erik Hurst for the Federal Reserve Bank of Boston. I suspect from its eager reception by Worstall, and earlier from the Economist , that the study's conclusions will soon become an article of faith for the you-never-had-so-good crowd.
There's only one small problem with Aguiar and Hurst's truefactproof: it's all a question of judgment and data quality. The authors at least have the integrity to say so in their conclusions, even if their abstract hypes the extent to which their interpretation "documents" what they suppose it does.
Page 30: "Any definition that distinguishes “leisure” from “work” is a matter of judgment." Check.
Page 33: "The ability to examine different patterns in time use over four decades hinges critically on the quality of data within each of the time‐use surveys." Check.
Let's start with data quality, since that's no doubt the last thing that would occur to a booster like Mr. Worstall. Aguiar and Hurst compared data from five surveys, conducted by three different organizations over the course of nearly 40 years -- in 1965, 1975, 1985, 1993 and 2003. Just to give some sense of the subtlety of what's involved in such an exercise, the 1965 survey consisted of interviews with 2,001 individuals, 776 of whom were from Jackson, Michigan.
Now, I'm sure Jackson, Michigan is a wonderful place to conduct a survey. But whether a survey conducted there is comparable with one that is nationally representative is another matter. Also, it turns out that the 1965 sample included only 17 non-working men. You can weight your demographic categories all you want, but you'll never be able to weight a very small, unrepresentative sample into a representative one. Garbage in, garbage out.
One does have to admire the daring with which the researchers benchmarked their apple, orange, peach, pear and banana comparisons, though. They benchmarked the market work reports from the surveys to market work data in larger studies. In other words, the results of these surveys were fairly robust on questions that required much less of a subjective judgment. Cool. And unpersuasive.
Speaking of judgments, how does one deal with the fact that time spent traveling to an activity is included in the activity? For example, driving to Mickey D's for a happy meal counts as leisure, while driving to the store to shop for groceries would be non-market work. Notice the difference? I didn't think so.
Caring for ill or elderly family members is, of course, a leisure activity because it would be too complicated to count it otherwise. Ditto for child care. Likewise, watching TV probably accounts for a sizable chunk of the increased leisure of those less-educated adults who have been especially blessed with increased leisure over the past 40 years. Three cheers for TV! It sets the underemployed, on-call, contract, just-in-time, precarious, contingent workers free!
It short, while the study may provide some interesting food for thought, it is in no way conclusive. It is proof only that you can massage some counter-intuitive and headline-grabbing conclusions out of an odd assortment of loosely-similar surveys.
Posted by sandwichman at February 7, 2006 07:05 AM