Do shifts in SAT race gaps reflect who chooses to test?
Core argument: Setting Whites as the baseline with a mean of 100 and a standard deviation of 15, the author converts effect sizes (d) into SAT score gaps across racial groups, finding results consistent between years.
[I’ve] set Whites to the baseline for this graph and assume they have a mean of 100 with an SD of 15 and that gaps can be computed from d accordingly. Results are consistent between years, with most differences being within the range of expectations from sampling error. The exceptions are Amerindians and Asians, who, respectively, got much worse and much better relative to Whites. Why? I assume this reflects the continuing trend towards elite Asians selecting into SAT test-taking and elite Amerindians selecting out, rather than genuine population changes. To put these race differences in means and standard deviations into perspective, [the] graph of the data [is] scaled by population size.


