To measure the effects of discrimination, one has to rule out skills, education, or other factors that could explain between-group differences.
To understand this, go to LearnRegression.com and click on the "Differences between groups" page. Then click on the X,Y graph to create data that shows two things: (i) a statistically significant mean difference in Y for the blue and black groups:And (ii): once we account for the differences in X, the between-group differences lose statistical significance:
The author finds the same pattern in data for "unemployment, teen pregnancy, incarceration and other outcomes."
The solution isn’t to look away from discrimination. It does exist. But we also can’t point at every gap in outcomes and instantly conclude it’s racism. Prejudice must be measured rigorously. Statistically. Disparity doesn’t necessarily imply racism.