The four-fifths rule (a.k.a. 4/5ths rule or 80% rule) is the simplest and most common way of estimating adverse impact. The Uniform Guidelines indicate that the 4/5ths rule is the preferred method for determining adverse impact (see Questions and Answers #18, 23, and 24) unless samples are very small (see Questions and Answers #20 and 21) or very large (see Questions and Answers #20 and 22). The 4/5ths rule can be computed according to the four steps shown below (see Question & Answer #12):
1) Calculate the selection rate for each protected group that makes up more than 2% of the applicant pool. The selection rate is equal to the total number or applicants within the group that are hired divided by the total number of applicants within that group.
2) Observe which group has the highest selection rate. This is not always the white, male, or "majority" group.
3) Calculate the impact ratio by dividing the selection rate for each group by the selection rate of the group with the highest selection rate.
4) Determine if the selection rates are substantially different. If the impact ratio is less than .8, there is a 4/5ths rule violation.
The impact ratio could be considered a test of practical significance because it focuses on an effect size (i.e., the ratio of selection rates). The advantages of the impact ratio include (a) it is easy to use and does not require statistical software or training in statistics, (b) it describes the magnitude of the selection rate difference between the groups that are being compared, and (c) it is more powerful than statistical tests (Collins & Morris, 2008; Morris, 2001).
However, the most notable disadvantage of the impact ratio is that it is subject to considerable sampling errors (especially when the sample size and selection ratio are small) and is prone to making a Type I error (Collins & Morris, 2008; Roth, Bobko & Switzer, 2006; see also Ironson, Guion & Ostrander, 1982; Lawshe, 1987; Morris, 2001; Morris & Lobsenz, 2000). This is problematic because, if the impact ratio is less than .8, it cannot be certain if adverse impact truly exists or if the result is due to chance. Roth, Bobko and Switzer (2006) demonstrated that the impact ratio incorrectly indicates that adverse impact exists (i.e., makes a Type I error) 20% or more of the time when there are 50 or fewer hires. Lawshe (1987) compared impact ratio results using the same test in the same manner for the same job across two consecutive years and found that adverse impact for race changed significantly in 6/16 comparisons and that in 9/21 comparisons, the 4/5ths rule was satisfied in one year, but not in the other.
Use Adverse Impact Analysis to estimate adverse impact using the four-fifths rule now.