Diversity shrinkage of pareto-optimal solutions in hiring practice: Simulation, shrinkage formula, and a regularization technique

Adverse Impact, Diversity, Personnel Selection, Pareto-Optimal Weighting

Song, Qianqi

关键词: Adverse Impact, Diversity, Personnel Selection, Pareto-Optimal Weighting;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/101350/SONG-DISSERTATION-2018.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

To reduce adverse impact potential and improve diversity outcomes from personnel selection, one promising technique is De Corte, Lievens, and Sackett’s (2007) Pareto-optimal weighting strategy. De Corte et al.’s strategy has been demonstrated on: (a) a composite of cognitive and noncognitive (e.g., personality) tests (De Corte, Lievens, & Sackett, 2008), and (b) a composite of specific cognitive ability subtests (Wee, Newman, & Joseph, 2014). Both studies illustrated how Pareto-optimal weighting (in contrast to unit weighting) could lead to substantial improvement in diversity outcomes (i.e., diversity improvement), sometimes more than doubling the number of job offers for minority applicants without changing the job performance outcome in personnel selection. The current dissertation investigates topics related to a key limitation of the technique—the possibility of shrinkage, especially diversity shrinkage, in the Pareto-optimal solutions. The dissertation consists of three studies. Study 1 attempts to study diversity shrinkage and job performance validity shrinkage using Monte-Carlo simulation. Using Monte Carlo simulation, sample size and predictor combinations are varied and cross-validated Pareto-optimal solutions are obtained. Study 2 derives approximate mathematical formulae to directly correct for job performance validity shrinkage and diversity shrinkage when using Pareto-optimal weights. These shrinkage formulae for Pareto-optimal weighting are evaluated using simulation. Finally, Study 3 attempts to develop a Pareto-optimal weighting algorithm that achieves both optimization and regularization (similar to ridge regression, LASSO regression, or elastic nets; in the context of Pareto-optimal weighting with two criteria). An R package is developed to estimate Pareto-optimal solutions in personnel selection (i.e., ParetoR package), which includes: (a) De Corte et al.’s (2007) Pareto-optimization method (i.e., based on the NBI algorithm; used in Study 1), (b) Pareto-optimal shrinkage formula corrections (i.e., as introduced in Study 2), and (c) a regularized Pareto-optimal method (i.e., as introduced in Study 3). In sum, the current dissertation aims to contribute to the field of diversity selection by investigating job performance validity shrinkage and diversity shrinkage under the Pareto-optimization method to simultaneously optimize both the job performance and diversity of new hires.