How much do differences in endowments explain ethnic and racial earnings differences?

This paper’s objective is to examine differences in earnings between ethnic groups to find how much of the earnings differences between two groups can be explained by endowments, versus rewards for those endowments.  Differential rewards could be interpreted as discrimination.  We use Current Population Survey data on over a million households from 2007 – 2014.  The ethnic groups are whites, blacks, Asians, Hispanics, and mixed race.  To explain earnings, we include variables for unionization, occupation, education, part – time work, age, marital status, metropolitan area residence, government employment, and receipt of transfer benefits. We use the Oaxaca – Blinder decompositions (that employ a separately estimated regression for each group).    Using this model with our data, we can explain over 95% of the variation in earnings between groups from differences in endowments.  This leaves very little to be explained by differences in rewards, thus suggesting that discrimination has diminished greatly over time.  In addition, since the variables explain a very large percentage of the earnings variation (over 97%), there are extremely small differences between the two decompositions, unlike some previous work. 

In addition, we compare results with our full model to a simplified model with many fewer explanatory variables.  For this simplified model, the different decompositions vary much more.    

We also compute the Neumark decomposition which uses regression coefficients from a pooled sample.  Previous work has suggested that this decomposition overestimates the amount of earnings differences explained by endowments.  However, we find that this pooled decomposition does not overestimate the explained variation, since the results from the pooled decomposition are normally very close to the results from the decompositions computed using separately estimated regressions.