Methodological Reconciliation of Mariel: Compositional Bias and the Role of Race
Following the publication of George Borjas's reappraisal of the Mariel Boatlift, a major methodological debate emerged over whether the reported 10% to 30% wage drop was a real economic effect or a statistical artifact of compositional bias in the survey data.
The Clemens and Hunt (2017/2019) Reconciliation
In a 2017 study (published in the ILR Review in 2019), economists Michael Clemens and Jennifer Hunt demonstrated that the wage decline identified by Borjas was spurious, driven by a sudden shift in the demographic composition of the Current Population Survey (CPS) sample in Miami.
During the 1970s and 1980s, the Census Bureau faced intense legislative and judicial pressure to correct its severe undercount of low-skilled Black men, particularly in Miami. Consequently, starting with the 1981 CPS (which reported wages earned in late 1980), the Census Bureau dramatically improved its survey coverage. In Miami's less-than-high-school sample, the proportion of Black men jumped from roughly one-third (33%) in 1979 to two-thirds (66%) in 1980, eventually reaching 91% in 1985. This shift did not occur in Borjas's control cities, where the Black fraction in the sample actually fell (sometimes to zero).
Because a substantial wage gap existed between Black and non-Black low-skilled workers in Miami at the time, this sudden demographic shift artificially dragged down the average wage of the sample. Clemens and Hunt showed that multiplying the increase in the Black sample share by the racial wage gap explained the entire 10% to 30% wage drop found by Borjas.
Furthermore, this explanation resolved why Borjas's findings were three times larger in the March CPS than in the Outgoing Rotation Group (ORG) dataset: the post-1980 increase in the coverage of low-wage Black men was three times larger in the March CPS than in the ORG.
Borjas's Defense: "Still More on Mariel"
Borjas responded to this critique in his 2017 working paper, "Still More on Mariel: The Role of Race" (also published in the ILR Review in 2019). He argued that the Clemens and Hunt thesis is "demonstrably false" because the timing of the post-Mariel wage decline in Miami does not perfectly coincide with the increase in the Black share of the low-skilled workforce in the survey. Borjas maintained that even after adjusting for racial composition, a significant wage decline persisted for low-skilled native workers in Miami, suggesting that the debate over how to adjust for compositional shifts in small survey samples remains active.