A meta-analysis on the validity of personality tests used for personnel selection : a thesis presented in partial fulfilment of the requirements for the degree of Master of Science in psychology at Massey University
For decades, personality tests have been commonly used as one of the tools for personnel selection. However, through the use of various validity generalization techniques, researchers have claimed that they have very poor validity in predicting job performance. These claims were made by Guion and Gottier (1965), Ghiselli (1973), and Schmitt, Gooding, Noe and Kirsch (1984). Each of these researchers used different statistical techniques and methodologies to reach their conclusions. The latest study by Schmitt et al. (1984), used a statistical validity generalization technique called meta-analysis. Based on data collected from only two journal publications they claimed that personality tests had a validity of .15. The present study tested the conclusions of the Schmitt et al. (1984) study, by re-analysing the same data using a more accurate meta-analysis technique and by incorporating a larger data base. In addition to this, any new data from 1952 up to 1990 was included in an overall analysis to find out the current validity of personality tests. A flexible coding technique which interacted with a computerised data base allowed any combination of data to be separately analysed. This made it possible to discover which types of personality tests worked best in differing situations such as different sample types and criterion measures. Results of the Schmitt et al. (1984) re-analysis showed that by correcting coefficients for unreliability, the overall validity was significantly higher than the Schmitt et al. (1984) result. A separate analysis revealed that vocational tests had the highest validity of the six personality test types. The sample-types with the highest validities were Supervisory and Skilled workers. The best criterion-types were in the "Other" category whereby measures were developed specifically for the type of job. The overall analysis incorporating 38 years of research showed that personality tests had a validity of .22. This was significantly higher than the figure quoted by Schmitt et al. (1984). Results showed that personality tests in their present state are generally poor predictors of job performance, however when they are modified to become more job specific, their validity improves. It is suggested that in the future, personality tests should be specifically designed for the purpose of personnel selection and for specific jobs.