The effect of response distortion on the accuracy of predictive inferences of personality inventories

Abstract

The effects of faking on the predictive ability of the CPI and social desirability (SD) scale were assessed. Three hundred and twenty-two subjects were asked to respond to the CPI and the SD scale twice, once responding honestly, and once faking to create socially desirable self-images. Results indicate that subjects could successfully distort their responses on all scales. Depending on the scale characteristics, faking could both reduce as well as increase scale variability and reliability. People who are highly intelligent did not distort their responses any better than people of lower intelligence. Furthermore, response distortion was found to have negative effects on valid scales. It did not affect scales that had no predictive ability. However, for certain scales, faking could improve the likelihood of those scales to predict performance when the amount of faked responses in the sample was between 10% and 30%. Implications for organizations and for further research were also discussed.