Prof. Don Moore on improving the accuracy of personnel selection
One of the clearest lessons to emerge from decades of research on personnel selection is that the traditional face-to-face job interview is terrible for predicting future job performance.
Authoritative reviews of the research literature put the correlation between face-to-face job interviews, as they are usually conducted, and subsequent job performance, at around 0.4. While that’s better than nothing, it leaves much to be desired. For example, weight and height have about the same correlation. There are plenty of tall people who weigh less than short people.
So how can your company improve how it selects people? First, if you are going to continue using interviews, structure them. The same interviewers should ask all candidates the same set of questions. Research shows that structuring interviews this way is the single most important thing organizations can do to increase their predictive validity. Frank L. Schmidt and J.E. Hunter’s definitive review of the research literature suggest they could improve the correlation between interviews and work performance to 0.52—nearly as good as more costly tools like job tryouts.
One assessment tool that’s less biased and more accurate than interviews is an intelligence test. It turns out that intelligent people perform better at their jobs.
Another key is to determine which skills and abilities are actually necessary to succeed on the job and then identify hard-to-fake ways of assessing them. Each interview question should have a point and should assess some work-relevant ability or behavior. As such, it should be easy for interviewers to specify beforehand what a good answer might look like. Actual responses should be scored against these criteria. These scores should then be averaged to establish some numerical assessment of interview performance.
If you’re concerned that the scored questions seem to omit an intuitive holistic assessment of the candidate, then you can consider adding one. This can be averaged in with the scores for the specific interview questions.
Another assessment tool that’s less biased and more accurate than interviews is an intelligence test. It turns out that intelligent people perform better at their jobs, whether they’re designing microchips or mopping floors. And there are intelligence tests that do not discriminate on the basis of language, culture, or ethnicity. These tests, such as Raven’s Progressive Matrices, are simple to administer and considerably less costly in time, effort, and organizational resources than are face-to-face interviews.
Many people worry—justifiably—that intelligence tests fail to take into account many things that they value in a work colleague, including kindness, conscientiousness, and wisdom. That’s true, but the evidence suggests that these things are also hard to diagnose based on an interview. Moreover, intelligence tests omit biases of current employees–conscious or unconscious–about gender, ethnicity, or age. Evidence suggests that it’s difficult for individuals to completely set aside their cultural attitudes when assessing others’ potential cultural fit in the organization. As a result, hiring decisions perpetuate the current demographic makeup of the organization.
When choosing a candidate, don’t hold lengthy meetings discussing the merits of each. Simply hire the one with the highest overall score.
These recommendations will likely improve the predictive validity of your company’s selection process and also reduce the time and effort you put into recruiting.
A hiring process that is both cheaper and more effective at the same time? That’s a contender worth selecting.
A fuller account of this research appeared in the Nov. 2017 issue of California Management Review.