Construct valdity relates to whether a particular psychometric assessment...
Lead consultant at Test Partnership, Ben Schwencke, explains what is adverse impact.
In the context of employee selection, adverse impact occurs when the selection ratio of a specific group is disproportionally lower than expected. Typically, adverse impact is most concerning when it pertains to legally protected groups, which can include ethnic groups, genders, age groups, and sexualities. Discrimination on these grounds is illegal, and affected job candidates have the right to legal representation and protection. Employers therefore, are obligated to minimise any adverse impact within it selection processes, tracking and rectifying any potential issues as soon as possible.
Here is a practical example: An employing organisation is recruiting for a graduate scheme, and plans to interview 100 candidates. 50 of those candidates are men, and 50 of those candidates are women. If all 50 of those male candidates are given offers, but all of the female candidates are rejected, the interview process can be said to show significant adverse impact on the basis of gender.
In this same selection process, 80 candidates are ethnically white, and 20 candidates come from Black, Asian, and Minority Ethnic backgrounds (BAME). Of the white candidates, 40 are given offers, compared to 10 offers for the BAME candidates. Because 50% of both white and BAME candidates were given offers, this selection process does not appear to show adverse impact on the basis of ethnicity.
Quantifying adverse impact can be done in several ways. Some employers use the “four-fifths” rule, requiring that a minority group must show at least four fifths of the selection ratio of the majority group, i.e. if 50% of white candidates are selected, then at least 40% of BAME candidates must be selected. Alternatively, Chi-Square statistics can be used to determine whether the difference in selection ratio is statistically significant. Lastly, adverse impact in psychometric assessments can be evaluated using Cohen’s d effect sizes, highlighting the size of differences in score between groups.