The Cato Institute's Alex Nowrasteh published a new analysis last week of the famous Mariel boatlift. The boatlift was a massive inflow of Cuban refugees to Miami over the course of a single summer in 1980. To make a long story short, the recent work of economist George Borjas indicates that the boatlift probably caused a decline in wages for Miami workers who did not have a high school degree.
In contrast, Nowrasteh says that Borjas's own methods indicate that the boatlift "raised the wages of low-skill Miamians." The reason is that "low-skill" for Nowrasteh means both high school dropouts and people with only a high school degree (hereafter "HS-and-below") – a definition that covers more than half the workers in Miami at the time. Nowrasteh combines the wage decline for dropouts and the wage increase for high school graduates and argues that Mariel's overall effect on "low-skill" natives is positive.
The point is not especially interesting, since the standard immigration narrative has always been that efficiency gains come at the expense of the natives with whom immigrants most directly compete – high school dropouts, in the case of Mariel. From a policy perspective, it is not clear that we should welcome immigration that generates a net economic gain by redistributing away from the least-skilled Americans, even if some of the gains are going to the skill group right above them.
Policy arguments aside, I am unconvinced by Nowrasteh's empirical argument that the HS-and-below group enjoyed a net wage increase from Mariel. He analyzed two separate datasets each with four groups of control cities, for a total of eight estimates. Of those eight estimates, four show a positive wage effect for HS-and-below natives, and four show a negative effect.* We can debate which combination of dataset and control group gives the best estimate, but obviously there is a lot of uncertainty here, and taking an average across the estimates is not an appropriate way to deal with it. Note in contrast that Borjas's finding of a wage decline for dropouts is robust across all eight scenarios.
More uncertainty comes when we consider how past analyses of the same question led to different answers. The economist Joan Monras previously investigated the impact of Mariel on the HS-and-below category using the same datasets and one of the same control groups (the "Card cities") as Nowrasteh. The combination of the larger of the two datasets and the Card cities gave Nowrasteh his highest positive estimate of the wage impact on HS-and-below natives. However, using the same dataset and control group, Monras estimated the wage impact on HS-and-below to be essentially zero. This discrepancy could be due to any number of differences in sample construction or methods, but it is not reassuring that the results can change so much.
Furthermore, a working paper by Borjas and Monras uses a regression-based method to determine the wage impact of Mariel on dropouts and graduates separately. Nowrasteh cites that paper's finding of a negative impact on dropouts and a positive impact on graduates as motivation for his own analysis of HS-and-below. But it would have been simple to determine the HS-and-below impact directly from the Borjas and Monras paper: Just take a weighted average of the dropout and graduate wage effects reported in their Table 4. Doing that yields a negative number that does not reach statistical significance.
Overall, Nowrasteh frames his argument as the following: Borjas may or may not be right about Mariel lowering the wages of dropouts, but if he is, then we must also conclude that Mariel raised the wages of dropouts and high-school-only natives put together. The argument doesn't work. Across several different methodological scenarios that do not alter Borjas's conclusion, Nowrasteh's numbers – both his own and those of economists who previously studied the same question – do not tell a consistent story. His claim that accepting Borjas requires accepting that HS-and-below natives benefited from Mariel is therefore unconvincing.
* Nowrasteh's estimates are reported not as average wage changes, but as average wage changes multiplied by the size of the populations, making interpretation difficult. For example, the first column of Table 1 reports a net wage change for HS-and-below natives of -4.31. Is that a big effect or a small effect? Furthermore, no standard errors are provided, so the precision of the estimates is unknown.