Adverse impact is a question about rates, not head counts. Two groups can produce very different hire numbers simply because different numbers applied, so the Uniform Guidelines compare the proportion selected.
Selection rate
- Selection rate (UGESP section 16R) is the proportion of applicants or candidates who are hired, promoted, or otherwise selected. Formula: selection rate = number selected in the group / number of applicants in that group.
- Section 16B defines adverse impact as a substantially different rate of selection that works to the disadvantage of a race, sex, or ethnic group. Because the definition is written in rates, rates are the comparable quantity. Raw hire counts are not.
- Rates do not adjust for differences in job requirements. They only tell you the proportion selected.
The impact ratio
Impact ratio = the group's selection rate / the highest group's selection rate. The denominator is always the group with the highest rate, whichever group that is.
Work the example the questions use: 320 applicants with 48 selected gives 48/320 = 0.15, or 15 percent. 180 applicants with 36 selected gives 36/180 = 0.20, or 20 percent. The highest rate is 20 percent, so the impact ratio is 0.15/0.20 = 0.75, or 75 percent, below four-fifths and generally regarded as evidence of adverse impact.
Second example: 27 of 90 is 30 percent, 14 of 70 is 20 percent, so the ratio is 0.20/0.30 = about 67 percent.
Three ways the wrong answer is built
- Dividing the higher rate by the lower one, which inverts the ratio (133 percent instead of 75).
- Treating selection counts as rates, so 48 and 36 become percentages.
- Reporting the percentage-point gap (20 minus 15 = 5) instead of the ratio.
Not absolute
Section 4(D) says rates below four-fifths generally evidence adverse impact and greater ratios generally do not, and then qualifies itself in both directions. A smaller difference may still be adverse impact where it is significant in statistical and practical terms; a greater difference may not be where it rests on small, statistically insignificant numbers. With 22 applicants in total, a 78 percent ratio is a flag to test for significance, not a conclusion. Report it and analyse; do not declare, and do not suppress it.