Brad DeLong explains the birth of an early pro-NAFTA number:

Here’s where I state that the “200,000 net jobs projected from NAFTA” number was mine: we took an estimate of overall economic efficiency gains from tariff reductions and an employment elasticity with respect to the real wage from the Labor Department, and estimated that in the long run stable-inflation employment would grow by 0.14 percent as a result of the deal. I think it was the right answer to the question asked in 1993; I don’t think that was the right question to be asking.

I naively assumed numbers like this came from the aggregation of lots of detailed industry studies. Apparently I was way off – Clinton’s economists just combined an estimate of the effect of NAFTA on the real wage with an estimate of the effect of the real wage on employment.

Which frankly leaves me a little puzzled: Does the substitution effect exceed the income effect after all? The history of the workweek in the 20th century makes that hard to believe. If it doesn’t, then NAFTA destroys jobs – by making people so rich that they don’t want to work as much.