In a blog post about the minimum wage, Bryan Caplan makes note of a tension in some of David Card's research. For those who don't know him, Card is one of the great empirical labor economists of his generation. In some of his research, joint with Alan Krueger, Card finds that increases in the minimum wage have negligible effects on employment. In other research, on the Mariel boatlift, Card finds that increases in the supply of unskilled workers have negligible effects on wages and employment of existing workers.
Caplan notes that these results are hard to reconcile: The former suggests that labor demand is highly inelastic, whereas the latter suggests it is highly elastic. That is a very good point. While I was well aware of Card's famous research on these topics, I had not put them together and recognized the tension between them.
To me, this brought to mind an econometrics class I took from Frank Fisher many years ago. Frank used to say that the "iron law of econometrics" is that coefficients are biased toward zero. He meant that various problems, such as measurement error and misspecification, tend to make it hard for researchers to find an effect even when one really exists. Fisher's "law" overstates things, which was Frank's style, but his basic point is often right.