B. Technical standards for criterion-related validity studies—(1) Technical feasibility. Users choosing to validate a selection procedure by a criterion-related validity strategy should determine whether it is technically feasible (as defined in section 16) to conduct such a study in the particular employment context. The determination of the number of persons necessary to permit the conduct of a meaningful criterion-related study should be made by the user on the basis of all relevant information concerning the selection procedure, the potential sample and the employment situation. Where appropriate, jobs with substantially the same major work behaviors may be grouped together for validity studies, in order to obtain an adequate sample. These guidelines do not require a user to hire or promote persons for the purpose of making it possible to conduct a criterion-related study.
(2) Analysis of the job. There should be a review of job information to determine measures of work behavior(s) or performance that are relevant to the job or group of jobs in question. These measures or criteria are relevant to the extent that they represent critical or important job duties, work behaviors or work outcomes as developed from the review of job information. The possibility of bias should be considered both in selection of the criterion measures and their application. In view of the possibility of bias in subjective evaluations, supervisory rating techniques and instructions to raters should be carefully developed. All criterion measures and the methods for gathering data need to be examined for freedom from factors which would unfairly alter scores of members of any group. The relevance of criteria and their freedom from bias are of particular concern when there are significant differences in measures of job performance for different groups.
(3) Criterion measures. Proper safeguards should be taken to insure that scores on selection procedures do not enter into any judgments of employee adequacy that are to be used as criterion measures. Whatever criteria are used should represent important or critical work behavior(s) or work outcomes. Certain criteria may be used without a full job analysis if the user can show the importance of the criteria to the particular employment context. These criteria include but are not limited to production rate, error rate, tardiness, absenteeism, and length of service. A standardized rating of overall work performance may be used where a study of the job shows that it is an appropriate criterion. Where performance in training is used as a criterion, success in training should be properly measured and the relevance of the training should be shown either through a comparsion of the content of the training program with the critical or important work behavior(s) of the job(s), or through a demonstration of the relationship between measures of performance in training and measures of job performance. Measures of relative success in training include but are not limited to instructor evaluations, performance samples, or tests. Criterion measures consisting of paper and pencil tests will be closely reviewed for job relevance.
(4) Representativeness of the sample. Whether the study is predictive or concurrent, the sample subjects should insofar as feasible be representative of the candidates normally available in the relevant labor market for the job or group of jobs in question, and should insofar as feasible include the races, sexes, and ethnic groups normally available in the relevant job market. In determining the representativeness of the sample in a concurrent validity study, the user should take into account the extent to which the specific knowledges or skills which are the primary focus of the test are those which employees learn on the job.
Where samples are combined or compared, attention should be given to see that such samples are comparable in terms of the actual job they perform, the length of time on the job where time on the job is likely to affect performance, and other relevant factors likely to affect validity differences; or that these factors are included in the design of the study and their effects identified.
(5) Statistical relationships. The degree of relationship between selection procedure scores and criterion measures should be examined and computed, using professionally acceptable statistical procedures. Generally, a selection procedure is considered related to the criterion, for the purposes of these guidelines, when the relationship between performance on the procedure and performance on the criterion measure is statistically significant at the 0.05 level of significance, which means that it is sufficiently high as to have a probability of no more than one (1) in twenty (20) to have occurred by chance. Absence of a statistically significant relationship between a selection procedure and job performance should not necessarily discourage other investigations of the validity of that selection procedure.
(6) Operational use of selection procedures. Users should evaluate each selection procedure to assure that it is appropriate for operational use, including establishment of cutoff scores or rank ordering. Generally, if other factors reman the same, the greater the magnitude of the relationship (e.g., correlation coefficent) between performance on a selection procedure and one or more criteria of performance on the job, and the greater the importance and number of aspects of job performance covered by the criteria, the more likely it is that the procedure will be appropriate for use. Reliance upon a selection procedure which is significantly related to a criterion measure, but which is based upon a study involving a large number of subjects and has a low correlation coefficient will be subject to close review if it has a large adverse impact. Sole reliance upon a single selection instrument which is related to only one of many job duties or aspects of job performance will also be subject to close review. The appropriateness of a selection procedure is best evaluated in each particular situation and there are no minimum correlation coefficients applicable to all employment situations. In determining whether a selection procedure is appropriate for operational use the following considerations should also be taken into account: The degree of adverse impact of the procedure, the availability of other selection procedures of greater or substantially equal validity.
(7) Overstatement of validity findings. Users should avoid reliance upon techniques which tend to overestimate validity findings as a result of capitalization on chance unless an appropriate safeguard is taken. Reliance upon a few selection procedures or criteria of successful job performance when many selection procedures or criteria of performance have been studied, or the use of optimal statistical weights for selection procedures computed in one sample, are techniques which tend to inflate validity estimates as a result of chance. Use of a large sample is one safeguard: cross-validation is another.
(8) Fairness. This section generally calls for studies of unfairness where technically feasible. The concept of fairness or unfairness of selection procedures is a developing concept. In addition, fairness studies generally require substantial numbers of employees in the job or group of jobs being studied. For these reasons, the Federal enforcement agencies recognize that the obligation to conduct studies of fairness imposed by the guidelines generally will be upon users or groups of users with a large number of persons in a job class, or test developers; and that small users utilizing their own selection procedures will generally not be obligated to conduct such studies because it will be technically infeasible for them to do so.
(a) Unfairness defined. When members of one race, sex, or ethnic group characteristically obtain lower scores on a selection procedure than members of another group, and the differences in scores are not reflected in differences in a measure of job performance, use of the selection procedure may unfairly deny opportunities to members of the group that obtains the lower scores.
(b) Investigation of fairness. Where a selection procedure results in an adverse impact on a race, sex, or ethnic group identified in accordance with the classifications set forth in section 4 above and that group is a significant factor in the relevant labor market, the user generally should investigate the possible existence of unfairness for that group if it is technically feasible to do so. The greater the severity of the adverse impact on a group, the greater the need to investigate the possible existence of unfairness. Where the weight of evidence from other studies shows that the selection procedure predicts fairly for the group in question and for the same or similar jobs, such evidence may be relied on in connection with the selection procedure at issue.
(c) General considerations in fairness investigations. Users conducting a study of fairness should review the A.P.A. Standards regarding investigation of possible bias in testing. An investigation of fairness of a selection procedure depends on both evidence of validity and the manner in which the selection procedure is to be used in a particular employment context. Fairness of a selection procedure cannot necessarily be specified in advance without investigating these factors. Investigation of fairness of a selection procedure in samples where the range of scores on selection procedures or criterion measures is severely restricted for any subgroup sample (as compared to other subgroup samples) may produce misleading evidence of unfairness. That factor should accordingly be taken into account in conducting such studies and before reliance is placed on the results.
(d) When unfairness is shown. If unfairness is demonstrated through a showing that members of a particular group perform better or poorer on the job than their scores on the selection procedure would indicate through comparison with how members of other groups perform, the user may either revise or replace the selection instrument in accordance with these guidelines, or may continue to use the selection instrument operationally with appropriate revisions in its use to assure compatibility between the probability of successful job performance and the probability of being selected.
(e) Technical feasibility of fairness studies. In addition to the general conditions needed for technical feasibility for the conduct of a criterion-related study (see section 16, below) an investigation of fairness requires the following:
(i) An adequate sample of persons in each group available for the study to achieve findings of statistical significance. Guidelines do not require a user to hire or promote persons on the basis of group classifications for the purpose of making it possible to conduct a study of fairness; but the user has the obligation otherwise to comply with these guidelines.
(ii) The samples for each group should be comparable in terms of the actual job they perform, length of time on the job where time on the job is likely to affect performance, and other relevant factors likely to affect validity differences; or such factors should be included in the design of the study and their effects identified.
(f) Continued use of selection procedures when fairness studies not feasible. If a study of fairness should otherwise be performed, but is not technically feasible, a selection procedure may be used which has otherwise met the validity standards of these guidelines, unless the technical infeasibility resulted from discriminatory employment practices which are demonstrated by facts other than past failure to conform with requirements for validation of selection procedures. However, when it becomes technically feasible for the user to perform a study of fairness and such a study is otherwise called for, the user should conduct the study of fairness.
User Comments/Questions
Add Comment/Question