adverse impact analyses, disparate impact, mantel-haenszel, chi-square, fisher exact, Breslow-day, OFCCP analyses

ai toolkit online > availability comparison

Adverse Impact Toolkit Online - Availability Comparison

This section of the Program is designed for comparing one group's representation (i.e., the percentage of incumbents in a given position who belong to the gender or ethnic group) to that group's availability in the relevant labor market (using availability data from inside or outside the organization). This type of comparison is useful for determining whether a group is underutilized in a particular position, or group of positions.

This analysis should be differentiated from the "Selection Rate" comparison because (under most circumstances) a statistically significant underutilization does not automatically constitute a finding of adverse impact. The reason for this is straight-forward: a selection rate comparison directly evaluates how two groups fared on a particular employment practice, so if one group significantly outperforms the other, direct evidence is gathered regarding the impact of a particular employment practice on the group of interest. Then the attention can shift toward evaluating that particular employment practice for job relatedness (i.e., validity).

By contrast, the Availability Comparison does not (necessarily) consider the impact of one employment practice. Because the comparison is an overall evaluation that considers one group's makeup in a given position compared to their availability outside of position, it does not consider all of the practices, procedures, or tests that may have been used to select or promote individuals for that position. Further, it does not take into consideration other factors such as "job interest" or qualification levels of the at-issue group. For example, if outside availability data shows that men are statistically significantly underutilized for a group of clerical jobs, the underutilization could possibly be explained by either lack of interest on the part of men to pursue these positions, or the fact that men performed poorly on the multitude of qualification screens required for entry into the position (or likely some combination of these two factors and others). For these reasons, the Availability Comparison should be considered as a "threshold" or "initial inquiry test."

Men Women White Black Hispanic Asian Native Amer. Total Min. Gender Ethnicity
# Selected/Represented
% Available*
% Represented (1)
% Over/Under Represented (2) Red = Underutilization
Stat. Test Results-EXACT (3) Red = Statistical Significance
Stat. Test Results-ESTIMATED (4)     Red = Statistical Significance

Interpretation of EXACT Statistical Test (5)
Likelihood (One Chance In):
Probability as Std. Deviations:

Interpretation of ESTIMATED Statistical Test (5)
Likelihood (One Chance In):
Probability as Std. Deviations:

* Cells are highlighted if the total exceeds 100.0%


(1) % Represented: This row indicates each group's representation as a percentage of the total pool.

(2) Over/Under Represented: This row shows the group's representation compared to their availability. Underutilization at any level is indicated using red cells. Negative values show underutilization.

(3) Statistical Test Results (EXACT): This test uses the (two-tail) exact Binomial Probability Test (with mid-p adjustment that mitigates the effects of conservatism of exact methods while continuing to use the exact probabilities from the distribution being analyzed) to assess whether the degree of underutilization is extreme enough to be considered "beyond chance." Values less than .05 (in red) are "statistically significant"; values between .05 and .10 (in orange) are "close" to significance. Because this test compares one group's representation against their availability (rather than comparing the selection rates of two groups, like the Statistical Test on the "Selection Rate Comparison" page), statistically significant findings (without other evidence) should not be considered as direct evidence of adverse impact (because both discriminatory and non-discriminatory reasons can possibly account for the group's underutilization).

(4) Statistical Test Results (ESTIMATED): This test uses an estimator technique for the Binomial Probability Test that usually produces values that are similar to the Exact Test (see note above). When the probability value output by this test approaches .05 or the sample sizes are small (<30), only the Exact Test should be used because the results from this test can overestimate probability values.

(5) Interpretation of Statistical Test: These outputs describe the degree of the Statistical Test findings. For example, if the output shows the likelihood of the statistical test value is "1 in 20," this means that the group's underutilization is so extreme that the odds of it occurring by chance is only 1 in 20, or about 5%. In other words, this result indicates that chance can be "ruled out" as a reason for this difference. The "Probability as Std. Deviations" describes the probability value (from the Statistical Test) in terms of standard deviations units, which are sometimes easier to interpret than small probability values. A standard deviation of 1.96 corresponds with a probability value of .05, and a likelihood of 1 chance in 20.