1 / 18

Decision making: Utility and assessing return on investment

Decision making: Utility and assessing return on investment. Efficiency of linear models Alternative prediction models: multiple regression, multiple cut offs, multiple hurdle approaches Classical validity approach: selection ratio, base rate Decision-making accuracy

locke
Download Presentation

Decision making: Utility and assessing return on investment

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Decision making: Utility and assessing return on investment • Efficiency of linear models • Alternative prediction models: multiple regression, multiple cut offs, multiple hurdle approaches • Classical validity approach: selection ratio, base rate • Decision-making accuracy • Utility models: Taylor-Russell, Naylor-Shine, Brogden-Cronbach-Gleser models • Alternative methods of estimating SDy in Rands: the 40% rule, Schmidt-Hunter Global Estimation, Cascio-Ramos Procedure • Integration of selection utility with capital budgeting models

  2. Classical validity approach • Primary emphasis on measurement accuracy (reliability) and predictive efficiency (predictive validity) – correlation between predictor and criterion; emphasis on psychometric criteria • Goal of selection process is to capitalise on individual differences and to select those individuals who possess the greatest amount/quantity of the particular characteristics judged important for job success • Simple regression: predicting each individual’s criterion status based on predictor information • Multiple regression – compensatory as low scores on one predictor offset by high scores on another

  3. Efficiency of linear models in prediction • General linear model: y = a + bx • Weights (b) derived by least squares regression procedure, subjective/intuitive, unit weights • These models appropriate and accurate for a wide range of situations

  4. Moderator variables • Differential prediction exists: the correlation between predictor (assessment/test) and criterion varies as a function of classification on a third variable (gender, race, education, age) • Problems: (i) utility rarely assessed – cross validation - show that selecting based on moderator combined with assessment score superior to assessment scores alone (ii) subgroups often small – statistical power not enough to detect significant differences (iii) meta- analyses – sample sizes – Type I and II errors (iv) sub-grouping/-clustering – similar profiles enhances understanding of kinds of individuals for whom specific assessment predicts accurately

  5. Suppressor variables • Lack of correlation with criterion but high inter-correlation with one/more of the predictor variables – increases multiple R but not prediction • Model of multiple regression assumes that each predictor variable correlates with criterion but not with each other, thus predicts unique part of criterion, contributes to multiple R • In practice suppressor variables found in complex model – particularly when using aggregate data

  6. Criticism of classical validity approach • Largely ignores external parameters (minimum requirements, costs) of the situation that largely determine the overall worth of the selection instrument • Makes unwarranted utility assumptions • Fails to consider the systemic nature of the selection process (linear, mechanical)

  7. Alternative prediction models • Although multiple regression basis of traditional prediction approach, it has to be compared with alternative – depending on chosen strategy, different employment decisions may result • Consider advantages/disadvantages of multiple regression and compare with multiple cut-off and multiple hurdle approaches

  8. Multiple regression approach • Y’ = a + b1X1 + b2X2,+ … + bkXk • Assumptions of linearity, trait additivity • Values of predictors (X1, X2, X3,…Xk) vary across individuals, regression weights (b1,b2,…bk) constant for sample • Thus possible for individuals with different predictor scores to obtain identical predicted criterion scores – compensatory model (high scores on highly weighted predictor can compensate for low scores on predictors with lower weights

  9. Advantages of the multiple regression model • If assumptions are met and sample is large: errors in prediction minimised; predictors are combined optimally, to yield most efficient estimate of criterion status • Model flexible: mathematically can be modified to handle nominal data and non-linear relationships, linear and non-linear interactions; equations for number of jobs generated using same predictors, different weights, or different predictors

  10. Multiple cut-off approach • In some selections proficiency in one predictor cannot make up for deficiency in another (e.g. visual acuity drivers, pilots) • Minimal level of proficiency in one/more variables essential for job success - no substitutes allowed, failure on these predictors disqualifies applicant completely • As this approach non-compensatory, assumes curvilinearity in predictor-criterion relationships (above cut-off, higher predictor score not necessarily predicts higher criterion score)

  11. Multiple cut-off approach … • Curvilinear relationships can be accommodated in regression approach, in practice multiple regression and multiple cut-off approaches may lead to different decisions (e.g. Cascio (1991), p. 287) • No satisfactory solution to finding cut-off scores • If organisation knows number of posts available and number of applicants, can determine hiring rate

  12. Multiple Regression and Multiple Cutoff Model X1 cutoff A D R B X2 cutoff Predictor X2 R C R Multiple regression cutoff Predictor X1

  13. Setting cut-off scores • No single best way of setting cut-off scores for all situations • Begin with job analysis that identify critical competencies and relative levels of proficiency • Validity and job-relatedness of assessment critical • Where possible data on actual relation between assessment and criterion considered carefully • Cut-off scores high enough to meet minimum standards of job performance • Cut-off scores consistent with normal expectations of proficiency within the work force

  14. Expectancy tables • Depicts likelihood of successful criterion performance to be expected for any given level of predictor scores for an organisation (Cascio (1991), Figure 13.5, p. 288) • Also for individuals • Expectancy charts useful way of illustrating the effect of the validity coefficient on future hiring decisions • Optimal: initially use multiple cut-off scores to identify individuals who meet minimum requirements, then multiple regression with remaining predictors

  15. Organisational expectancy chart

  16. Multiple hurdle approach • Multiple regression and multiple cut-off models are single stage (not sequential) decision strategies • In multiple hurdle (sequential) cut-off scores on predictors used to make investigatory decisions • Cut-off scores on assessments or composite multiple regression used to make decisions • Applicants provisionally accepted and assessed further • Several staged approach appropriate when subsequent training is long, complex and expensive

  17. Multiple hurdle approach … • In best interest of individual and organisation to reach a decision as quickly as possible • Essential decisions as accurately as possible given available information • Forecasting accuracy increases as candidate clear the various hurdles but so do costs • As candidates are eliminated range restricted and validity underestimated

  18. Evaluating selection efficiency • Comparison of decision based on assessment to choosing candidates at random • Index should indicate proportion of saving by using assessment to predict success

More Related