We revisit a key finding of the labour market discrimination literature: that conditional on ability, there is no earnings gap between college-educated white and black men. Drawing on UK administrative data linking school records, university records, and tax records, we confirm this earlier finding from the US for British university-educated men born in the late 1980s. However, a substantial gap re-appears with the inclusion of additional controls for university course choice and commuting zone of residence. This is because all else equal, men from all ethnic minorities study higher-return subjects and live in higher-wage labour markets than white men. Conditional on the full set of controls, gaps are large for all ethnic minorities and across a variety of subgroups, with the exception of those with the highest skills and those who attended the most selective universities.

Our findings imply that the previous literature may have missed substantial ethnic earnings gaps among graduates due to the absence of controls for university subject and commuting zone of residence. The fact that black men study higher-return subjects and work in higher-wage labour markets than their white peers seems to offset underlying earnings gaps, potentially confounding earlier estimates. The common assumption in the economics literature that worker productivity for university graduates is perfectly observed appears to hold at most for a small subset of highly skilled graduates.