Predictor Validity as Related to Criterion Relevance, Restriction of Range, and Ethnicity

Abstract

Abstract Among 139 Western Union operators, a job-oriented spelling test correlated highly (-.53) with spelling errors in messages. Findings were attributed to criterion relevance (the use of actual work samples) and to relatively good criterion reliability (.70). There was no evidence of single-group or differential validity. The spelling test also correlated significantly with the performance of new operators during training. A numerical test correlated significantly with training performance and with errors in messages sent by experienced operators. The trait of cooperativeness was significantly related to customer complaints. Emphasis was given to the importance of correcting for range restriction (and of developing relevant and stable criteria) to ascertain the true value of a predictor.