Properly Using Selection Procedure Scores: Employee Retention
Although Kunze Analytics frequently references published guidelines for the use of assessments in the workplace originating from the US Department of Labor, another very helpful document to obtain direction is published by the Society for Industrial and Organizational Psychology, Inc. (SIOP). In 2003, the fourth edition of Principles for the Validation and Use of Personnel Selection Procedures became available. One may download the document from SIOP’s site here.
In this article, I’d like to think out-loud about the topic of employee retention. There’s been a lot written surrounding this topic over the years: it’s cost to companies, changes to employee morale, company loyalty, generational differences in typical tenure among other subjects. On page 57 of the 84-page pdf from SIOP, one reads the following admonitions:
The researcher must communicate how selection procedure scores are to be reported and used. Results should be reported in language likely to be interpreted correctly by persons who receive them. The administration guide should also indicate who has access to selection procedure scores.
Administrators should be cautioned about using selection procedure information for uses other than those intended. For example, although selection procedure data may have some validity in determining later retention decisions [the emphasis is mine], more potentially relevant measures such as performance ratings may be available. Furthermore, if the pattern of selection procedure scores is used to make differential job assignments, evidence is required demonstrating that the scores are linked to, or predictive of different performance levels in the various assignments of job groupings.
There a lot of information in these two paragraphs that should be unpacked carefully. Let’s give it a try.
The initial paragraph quoted above contains common sense indications due to the highly sensitive and confidential nature of assessment scores. Companies exercise care in protecting personnel data. The new General Data Protection Regulations (DGPR) for Europe recently came into effect in on May 25th, 2018 with European Union companies obliged to hire a Data Protection Officer. See the Wikipedia article here.
The second paragraph cautioning those who administrate assessments, surveys, or tests to stick to the intended purpose of their use should not go unnoticed. When companies prepare to implement an assessment, they either select a model from the vendor’s library or they do a local validation study by testing incumbents in the role or roles for which the assessment will be sued to assisting executive hiring decisions.
If the intended purpose of the solution is to make a better hiring decision when a position becomes vacant, whether the position is filled from an external or internal candidate, then this is the proper use of the psychometric instrument and any scores coming from the system. Referring to assessment scores from a freshly-validated assessment performance model as justification of involuntary separation of employees is an extraneous purpose not originally intended from the outset. With great circumspection and caution is it to be carried out. One reason for the grave admonition against this practice is because, as I see it, assessments might only explain 5%, 10%, 30% of variability in job performance in any role. That means that 95%, 90% or 70% of the cause of success in the job is opaque and hidden. One needs to be honest. If a company has decided to involuntarily terminate one or more employees, it should not be because of a low job fit score.
This does not mean personnel decisions to promote certain employees and keep others in their current role cannot be a reasonable use of assessment scores. It just means that when such decisions are to be made, then absolutely the most recent job performance metrics needs to be used in the validation of the performance model or models deployed to create the job fit scores. The researcher should tie assessment job fit scores back to each of the key employee performance metrics that are used to make personnel decisions. This adds a new level of integrity and transparency to the use of psychometric data. It also interprets scores in a way that everyone understands. What does a 5% increase in job fit score mean? Well, if the validation study shows through trendline analysis that it is equivalent to a 17.3% increase in annual store profitability, then the entire assessment has more meaning and become a great deal more relevant.
At Kunze Analytics, we know how to do the appropriate research surrounding the day-to-day usefulness test data. As an added plus, there are occasions that when assessment scores go up, not only does job performance increase, but also tenure in the position. Happy, productive employees tend to stay longer. This is not always the case, but it makes us smile everything the relationship exhibits a correlation within a statistically-significant range and is, according to the US Department of Labor, very beneficial to know.