Placement and Classification

A Conceptual Example

The idea of classification can be illustrated by an example. An organization has 50 openings in four entry-level jobs: Word processor has 10 openings, administrative assistant has 12 openings, accounting clerk has 8 openings, and receptionist has 20 openings. Sixty people apply for a job at this organization and each completes three employment tests: word processing, basic accounting, and interpersonal skills.

Generally, the goal of classification is to use each applicant's predicted performance score for each job to fill all the openings and maximize the overall predicted performance across all four jobs. Linear computer programming approaches have been developed that make such assignments within the constraints of a given classification situation such as the number of jobs, openings or quotas for each job, and applicants. Note that in the example, 50 applicants would get assigned to one of the four jobs and 10 applicants would get assigned to not hired.

Using past scores on the three tests and measures of performance, formulas can be developed to estimate predicted performance for each applicant in each job. The tests differ in how well they predict performance in each job. For example, the basic accounting test is fairly predictive of performance in the accounting clerk job, but is less predictive of performance in the receptionist job. Additionally, the word processing test is very predictive of performance in the word processor job but is less predictive of performance in the receptionist job. This means that the equations for calculating predicted performance for each job give different weights to each test. For example, the equation for accounting clerk gives its largest weight to basic accounting test scores, whereas the receptionist equation gives its largest weight to interpersonal skill test scores and little weight to accounting test scores. Additionally, scores vary across applicants within each test and across tests within each individual. This means that each individual will have a different predicted performance score for each job.

One way to assign applicants to these jobs would be to calculate a single predicted performance score for each applicant, select all applicants who have scores above some cutoff, and randomly assign [Page 627]applicants to jobs within the constraints of the quotas. However, random assignment would not take advantage of the possibility that each selected applicant will not perform equally well on all available jobs. Classification takes advantage of this possibility. Classification efficiency can be viewed as the difference in overall predicted performance between this univariate (one score per applicant) strategy and the multivariate (one score per applicant per job) classification approach that uses a different equation to predict performance for each job.

...