Replacing the employment interview with biodata : a written structures interview scored with modeled decisions

Three bodies of literature regarding employment interviews, biodata inventories, and decision modeling (or policy capturing) were reviewed, and an integration of major conclusions was presented. It was proposed that a structured interview could provide a good foundation for a biodata inventory. The result was expected to be a biodata inventory developed using theory, that would efficiently gather interview-type information, and could be scored statistically instead of clinically. Also, a scoring method was proposed as an alternative to the often-criticized empirical, rational, and intuitive methods: The decision model of a preferred expert could be used to score the inventory, especially if a well-defined or relevant criterion of performance were unavailable (e.g., for managerial selection).

These proposals were tested by having two experienced interviewers read and rate questionnaires (BSIs) completed by 166 retail store managers as part of a concurrent validation study. The BSI was a biodata inventory developed by one of the raters to simulate the interview he used to assess managerial applicants. The ratings were modeled consistently by small subsets of the 262 variables derived (a priori) from the BSI. The models were discussed in the context of Brunswick's Lens Model paradigm. Differences in prediction among double-crossvalidated and unit-weighted models were assessed using a variation of Efron's (1982) bootstrap method. Validity coefficients were reported for organizational criteria: Raters' global decisions about hiring and promotion were not valid, but statistical combinations of component (or dimension) ratings were moderately valid, and empirically-scored BSIs had strong validity coefficients.

Problems of criterion bias and relevance were discussed, as were suggestions for further research to clarify some of the findings.