The Science Behind Predicting Job Performance at Recruitment
Wyn Davies, Global Product Manager, Pearson TalentLens Angus McDonald, Chartered Psychologist The Science Behind Predicting Job Performance at Recruitment
Recruitment is a game of risk, where the chances of selecting a candidate who will turn out to be an effective employee are balanced against the risks of identifying someone who will not work out as expected. Recruiters use a wide variety of tools and processes to help them manage the risks associated with hiring decisions. If risks are effectively managed, recruiters are more likely than not to identify candidates who perform well in the job, though even under optimal conditions this process is far from perfect.
The complexity of understanding candidates and evaluating whether they really have the capabilities, There is now over 100 years of scientific characteristics, and motivations necessary for any evidence on the effectiveness of different tools given job means that incorrect decisions are easy to that can be used in the recruitment process. make. Factor into this issues such as a candidate’s deliberate deception, and the difficulties for recruiters only increase. If recruiters get it wrong, they risk costs including the need to quickly replace employees and the potential disruption that poorly fitting employees may cause to the organization. The central concept of validity In response to this recruitment challenge, there is now In the context of recruitment, validity concerns the over 100 years of scientific evidence on the effectiveness propriety of a specific assessment tool for a defined of different tools that can be used in the recruitment purpose. Tools with higher validity enable recruiters to process. This evidence is based on the association more accurately identify those candidates who are likely between evaluations of candidates made during the to go on to perform well in the job. In doing so, they recruitment process, and how they subsequently substantially aid in the management of risk that is an perform in the role for which they are hired. In turn, inevitable part of each hiring decision. this evidence should feed back into the hiring process, guiding recruiters as to which tools will most effectively allow them to manage risk. This paper summarizes scientific research on the effectiveness of different recruitment tools, explains key concepts, and identifies considerations to be taken when using recruitment tools. It is recognized that the processes and legislative frameworks recruiters work within will vary between organizations and regions, so the information given here will need to be applied in a way that is consistent with those local frameworks.
888.298.6227 | US.TalentLens.com 2 The Science Behind Predicting Job Performance at Recruitment
Correlation and validity Many different statistical techniques can be used to determine how effective any assessment tool is for candidate selection, but the majority of these are based on the idea of correlation. Correlation is a statistical technique that looks at the strength of association between two variables, typically the score or scores obtained on an assessment tool, and some measure of job performance. The stronger the correlation, the more closely associated with each other the two measures are said to be, as illustrated in the two diagrams on the right. The charts in figure 1, on the right, show the association between tests scores and job performance. The one at the top shows there is a lower association between tests scores and job performance, whereas the chart below in figure 1 shows a much stronger association. A strong association is indicated by the majority of data Figure 1. Sample of weak (chart at top) points falling on or close to a “line of best fit”. The importance of this vs strong (chart at bottom) correlation association is that it more accurately allows us to identify candidates (validity) likely to perform well in the job. Having strong validity means that an assessment tool is able to predict job performance with a good degree of accuracy. By knowing this, it is possible to use assessment scores to identify those candidates most likely to perform well in the job. Consider the charts in figure 2 which plot the scores of two different assessments against job performance for 30 candidates. The horizontal green line shows the expected level of job performance, meaning we want to select those candidates whose performance will be above the green line. In the top chart in figure 2 which shows a weak association between assessment scores and job performance—or low validity—it is difficult to know which score to use to identify good performers. No matter how high you set your cut score on the assessment, illustrated on the horizontal axis, you will identify some people who go on to perform below the level expected. Conversely, low pass marks will also identify some good performers. The lower chart in figure 2 shows a stronger association between the assessment scores and subsequent job performance. This makes it much easier to set a cut score that identifies where the majority of candidates will perform at the required level. Having strong validity makes it easier to manage the risk associated with recruitment decisions, as assessment scores more accurately predict performance. It is worth noting that performance is affected Figure 2. Sample of weak (chart above) by a number of factors and there is no absolute measure or vs (strong)—chart below and the ideal test that predicts success in the job with 100% certainty. You can, level of job performance however, use a range of relevant and valid assessments to greatly increase the chances of predicting likely performance and reduce the risk of hiring the wrong person.
888.298.6227 | US.TalentLens.com 3 The Science Behind Predicting Job Performance at Recruitment
WHAT IS META-ANALYSIS? What does science tell us? For many years, organizations and academic In practice, recruiters do not know how employees researchers have been trying to understand will go on to perform in a job. This makes selecting the association between assessments made valid assessment tools and applying results from them as part of the recruitment process and appropriately prior to hiring critical. subsequent aspects of job performance. We know the principles behind validity and why it is Many thousands of studies exploring important, but which assessment tools are most valid? this link have been conducted with There is extensive research on the use of different inconsistent results. There are many assessment tools and techniques. More recently, this challenges in conducting this “real world” research has been synthesized to allow comparison of research including being able to follow the effectiveness of different methods. One of the main people from recruitment to a time when ways this has been done is through a technique called job performance can be meaningfully meta-analysis, which is a statistical technique that draws measured, while obtaining accurate together the findings. indicators of job performance and small A recent paper by Frank Schmidt and colleagues1 sample sizes. summarizes the effectiveness of 31 different assessment The last of these points is a particular issue methods for the prediction of job performance and 16 when assessing the validity of recruitment for the prediction of performance in job-related training. processes. For research results to be robust, The key finding from this research was that tests of it is important that sufficient people are general mental ability (GMA) were identified as having the included. However, many organizations highest predictive validity of any selection method. They have only a limited number of people in were also seen to be the best predictor of performance each role, so collecting adequate samples on job-related training. The extensive research takes a long time, if it is even possible. underpinning the use of GMA means a high level of Meta-analysis is a statistical technique that confidence can be placed in these findings. allows the findings from many individual studies to be combined. In doing so, issues General Mental Ability (GMA) Tests such as measurement errors and small As Schmidt and colleagues observe, “when an employer sample sizes are allowed for, meaning uses GMA to select employees who will have a high more robust estimates of the associations level of performance on the job, that employer is between different assessment methods and also selecting those who will learn the most from job job performance can be established. training programs and will acquire job knowledge faster Meta-analyses have consistently identified from experience on the job”. This last point touches that some types of assessment are far more on the primary reason why candidates higher in predictive of job performance than others. GMA perform better in the job; they acquire job In the absence of good local validation for knowledge more rapidly and more deeply than those a specific job role, these findings provide lower in GMA. recruiters with a sound basis for identifying Many tests have been constructed to assess aspects assessments that are likely to be the of mental ability, and there are numerous examples most effective. where these have been developed primarily for use in recruitment settings. Examples include the Watson- Glaser™ Critical Thinking Appraisal and RAVEN’s™ Progressive Matrices. As versions of these tests can be delivered unsupervised online or administered to groups of applicants at a time under supervised conditions, they have a high level of utility in recruitment settings. They
888.298.6227 | US.TalentLens.com 4 The Science Behind Predicting Job Performance at Recruitment
are relatively easy to use and interpret, do not take long It would be rare to be offered a job without some form to complete, and are generally inexpensive. of interview. Research supports the effectiveness of Another notable finding relating to GMA is the wide interviews, and after tests of GMA, they are one of the range of job roles for which it has validity. Typically most effective selection methods. Interestingly, there is these tests have been used in areas such as graduate little evidence to suggest that structured interviews are recruitment, where jobs are assumed to place more effective than unstructured interviews, however considerable cognitive demands on the job holders. structured phone-based interviews do tend to come out Tests of GMA are known to effectively predict graduate- as slightly less valid. level roles, but they also have considerable predictive Other methods that show at least modest predictive validity for virtually any role, including unskilled. GMA’s validity in Schmidt and colleagues’ research include ability to predict performance tends to increase as the integrity tests, biographical data, assessment centers, cognitive demands of jobs increase. Highest validities are grade point average (i.e. academic results at graduate seen for professional and managerial roles, where GMA level), peer ratings, work sample tests, job tryouts, is able to account for over 50 percent of the variability in behavioral consistency method, and job knowledge tests. job performance on average. However, not all of these methods can be readily applied in all recruitment settings. For example, job tryouts How do other measures compare to the cannot easily be used in high-risk jobs or jobs where effectiveness of GMA? considerable training is required, and peer ratings can There are many recruitment tools apart from GMA, only be used with internal candidates. and most selection processes use combinations of these Even when other assessment methods can be used, (See figure 3). it does not necessarily mean they will reduce the risk
S Ts
I te ests
G a uate Post a e u at o a es
B o a al ata