Immigrant earnings,relative to what? The importance of earnings function specification and comparison points

Immigrant relative earnings estimates are sensitive to the choice of comparison point and the specification of earnings. Non-sample mean comparisons (Borjas, 1985) understate relative earnings. Simple earnings specifications (linear education, quadratic experience) overstate relative earnings for both poorly and well-educated immigrants. Specifications which ignore omitted variables understate the relative earnings of poorly educated immigrants and overstate those of well-educated ones. Although measures of assimilation and changes in immigrant quality are insensitive to earnings specification, they indicate strong earnings growth for post-1964 immigrants, an overall decrease in immigrant quality, and an increase in Mexican immigrant quality.