This study examined the impact of different scoring methods for situational judgment tests (SJTs) on validity and group differences. It found that shape-based scoring methods produced higher validities and smaller differences in scores between minority and majority groups compared to distance-based scoring methods, partially supporting the hypotheses. The results provide further evidence that shape-based scoring should be considered for SJTs used in personnel selection.
SJT Scoring Methods and their Impact on Group Differences and Validity
1. Prepared by Aon Hewitt | Consulting
Performance, Reward, & Talent | Assessment & Selection
The Impact of SJT Scoring Methods on
Validities and Group Differences
Kate LaPort, Christopher Huynh, & Ernest Paskey
3. Aon Hewitt | Consulting | Performance, Reward, & Talent 3
Background and Objective
SJTs are valid predictors of job performance, but additional research is needed to find
SJT scoring methods which maximize scale validity and decrease adverse impact
(McDaniel & Weekley, 2012)
McDaniel and colleagues (2011) compared distance and shape-based scoring
algorithms for SJTs that require respondents to rate each response option
– Validity: Shape-based scores control for elevation and scatter tendencies - producing
higher criterion-related validities than distance-based scoring algorithms
– Group Differences: Black respondents endorse extreme ratings on Likert scales
more frequently than White respondents - resulting in lower scores for Blacks relative
to Whites when those scores use distance-based scoring algorithms
Objective
Replicate McDaniel et al’s (2011) findings that SJT shape-based scoring methods are
associated with higher validities and lower majority-minority score differences using a
diverse sample of candidates in a high stakes testing environment.
4. Aon Hewitt | Consulting | Performance, Reward, & Talent 4
Hypotheses
Hypotheses
1
SJT shape-based scoring methods will yield higher validities than distance-based
scoring methods
2
SJT shape-based scoring methods will yield lower majority-minority mean
differences than distance-based scoring methods
6. Aon Hewitt | Consulting | Performance, Reward, & Talent 6
Participants and Procedure
Participants
2,611 employees in a large, nationwide organization
57.0% Female, 57.1% White
Procedure
Predictive validation
Participants responded to computer-based SJT items as part of application process
Supervisor ratings of performance collected for employees with at least 4 months’
experience
7. Aon Hewitt | Consulting | Performance, Reward, & Talent 7
SJT Measures
Simulation
Candidates completed a high-fidelity, computer-based simulation as part of their
employment application
Two scenarios (Angry Customer, Multitasking) featured 4 SJT prompts assessing three
competencies:
– Stress Tolerance (2 prompts)
– Multitasking
– Customer Centricity
Candidates were asked to rate between 7 and 9 behavioral response options on a 5-
point Likert scale (1 = Least Effective, 5 = Most Effective), and identify the most effective
response option
Distance and shape scores were computed for each competency
8. Aon Hewitt | Consulting | Performance, Reward, & Talent 8
Performance Ratings
Supervisors provided performance ratings of the participants’:
– Total Performance Ratings
– Overall Performance
– Adaptability
– Member Interaction
– Technology
13. Aon Hewitt | Consulting | Performance, Reward, & Talent 13
Summary
Designed to replicate McDaniel et al. (2011) findings in a diverse sample of candidates in
a high stakes testing environment
Results provide support for further consideration of SJT shape-based scoring methods
The validities associated with these short and efficient (i.e., 7-9 item) SJT prompts lends
further support to the development and utilization of Likert-style response scales for SJT-
items in the use of personnel selection
Hypotheses Results
1
SJT shape-based scoring methods will yield higher validities
than distance-based scoring methods
Partially
Supported
2
SJT shape-based scoring methods will yield lower majority-
minority mean differences than distance-based scoring
methods
Supported