Examining the Effectiveness of the Identify and Removal Faking Correction Strategy that Developed on the Response Pattern Based Faking Detection Methods in Selection Settings

Abstract

In selection research, many studies have attempted to correct for the impact of faking on personality assessment. One of the often-used correction strategies is to identify and remove faking job applicants from the job applicant pool from consideration for employment. In the current study, a simulation study was conducted to demonstrate the identify and removal strategies developed upon two response pattern based faking detection (machine learning method and covariance index method) methods could reduce the propotion of fakers being selected but could not improve the group mean criterion score in the selected group that vary across faker to honest respondent ratios (FHRs) and selection ratios (SRs). The simulation conditions that used the identify and removal strategies developed based on either the machine learning method or the covariance index method hire a significantly lower the proportion of fakers than the no correction strategy condition across all FHRs and SRs. The simulation condition that used the identify and removal correction strategy developed based on the machine learning method had a higher correlation coefficient between emotional stability and the criterion than the no correction condition as long as the FHRs were equal or bigger to 0.15. The simulation condition that used the identify and removal correction strategy developed based on the covariance index method had a higher correlation coefficient between emotional stability and the criterion than the no correction condition as long as the FHRs were equal or larger than to 0.3. On the other hand, the simulation condition that used the identify and removal correction strategy developed based on the machine learning method had a higher correlation coefficient between conscientiousness and the criterion than the no correction condition only when the FHR was equal to 0.5. The simulation condition that used the identify and removal correction strategy developed based on the covariance index method did not have a higher correlation coefficient between conscientiousness and the criterion than the no correction condition across FHRs. The simulation conditions that used the identify and removal correction strategies developed based on either the machine learning method or the covariance index method did not improve group mean criterion score in the final selected group than the no correction condition across all FHRs and SRs. These findings suggest that if the main objective is to reduce the proportion of fakers being selected, the use of the identify and removal strategies developed based on the two response pattern based faking detection methods is effective. On the other hand, if the main objective is to improve the group mean criterion score in the selected group by maximizing the criterion related validity, the use of this correction strategy will produce a negative effect.