George Borjas on Immigration and the Minimun Wage: "Am I the Only One Who Finds the Contradictory Inferences Disturbing?"

George Borjas is puzzled:

The Minimum Wage And Immigration: A Puzzle, by George Borjas: The federal minimum wage is rising today for the first time in a decade, from $5.15 to $5.85 an hour. And this reminds me of an important puzzle in labor economics that remains unresolved ... between the studies that examine the impact of a rising minimum wage on employment and those that examine the impact of immigration on wages.

Many studies in each of these literatures calculate correlations between wages and employment across cities or geographic regions. The minimum wage studies, for instance, relate changes in employment across states to changes in the minimum wage across states. The immigration studies relate changes in wages across regions to immigration-induced changes in supply. ...

The ... two sets of studies draw completely contradictory inferences from the data. On the one hand, the minimum wage literature often finds that a regression of employment on wages reveals a near-zero coefficient on the wage, and this is interpreted as saying that changes in the minimum wage have little effect on employment. On the other hand, the immigration literature often finds that a regression of wages on employment reveals a near-zero coefficient on employment, and this is interpreted as saying that immigration has little effect on the wage.

In the minimum wage literature, the inference is that the labor demand curve is almost vertical (or very inelastic), while in the immigration literature, the inference is that the labor demand curve is almost horizontal (or very elastic).

Let me rephrase the puzzle another way. If one were to believe the zero-effect result in the minimum wage literature, one would be forced to conclude that immigration must have huge adverse effects on the wage of native workers (at least in the short run). But if one were to believe the zero-effect result in the immigration literature, one would then have to conclude that minimum wage increases would have huge disemployment effects.

The only thing in common between the two sets of studies is that there is a zero correlation between wages and employment across geographic areas. What one puts on the left-hand side and the right-hand side of the regression model doesn't change this fundamental empirical fact.

Am I the only one who finds the contradictory inferences disturbing? It seems to me that at least one (and perhaps both) of the inferences economists draw from these cross-region correlations must be wrong.