A comprehensive psychometric audit of an existing selection procedure

[摘要] ENGLISH ABSTRACT:Selection represents a critical human resource intervention by virtue of its ability to regulate themovement of employees into, through and out of the organisation. Selection thus represents arelatively visible mechanism through which access to employment opportunities can be regulated.From the perspectives of both affirmative action and fairness, as well as utility, selection hastherefore been under intense scrutiny. This implies that there are two substantial criteria in terms ofwhich selection procedures need to be evaluated, namely equity and efficiency. Should the humanresource function be challenged to defend its selection procedure, it should be able to assemblecredible evidence to show the efficiency and equity of the disputed intervention by means of areasoned justification. The problem is, however, that most selection procedures being operated inSouth Africa would probably not be able to successfully meet this burden of persuasion. The searchfor equitable and efficient selection procedures thus necessitates the need for psychometric audits toprovide the feedback required to adjust selection procedures towards greater efficiency and equity,and to provide the evidence required for the vindication of organisations should they be challengedin terms of the South African anti-discriminatory labour legislation.The Guidelines for the Validation and Use of Selection Procedures developed by the Society forIndustrial Psychology (1998) represents an attempt to illustrate the ideal process according to whichselection procedures should be developed and validated. Conditional on the acceptance that theGuidelines (1998) set out the most justifiable methodology for the development and justification ofselection procedures, it becomes a necessity for organisations to periodically evaluate (i.e.periodically psychometrically audit) their current selection procedures and its developmental historyto determine whether the human resource function can convincingly demonstrate:.:. The business necessity of the selection procedure;.:. The validity of the performance theory on which the selection procedure is based; and.:. That the selection strategy combines applicant information fairly.A checklist was developed from relevant psychometric literature for the purpose of thepsychometric audit representing a structured list of activities required to justify the use of aselection procedure. A psychometric audit was conducted on a selection procedure for call centre staff of a large SA insurance company. The audit uncovered a number of deficiencies in the callcenter selection procedure and its developmental history.The performance hypothesis, in which the choice of operational predictor measures is grounded,was neither developed, nor argued, nor documented with sufficient clarity to indicateunambiguously the presumed nature of the nomological network of performance determinants andperformance constructs. Problems were found with the external validity of the validation design.No reliability, validity, fairness or utility analyses had been performed at the time of the audit.Subsequent correlation analysis indicates low statistically insignificant correlations between themajority of the chosen predictors and the developed criteria. Nonetheless, linear combinations ofpredictors were found for each of the three call center positions that significantly explain moderateproportions of criterion variance. The fairness of the use of the CSR multiple regression equationacross black and white applicants was examined and found to be acceptable. Due to practicalconstraints, the utility of the selection procedure has not been evaluated.It is recommended that the current selection procedure be re-examined in detail by the company tobring about positive changes in the performance hypothesis and the operational criterion measures.Thereafter, concrete evidence of reliably generated methodological research needs to be obtainedagain in order to verify the appropriateness, reliability and the meaningfulness of the inferencesmade from predictor assessments, thereby limiting, ifnot eliminating, possible cases oflitigation.

[发布机构] Stellenbosch University