Skip to Content

The Science of Selection

Summarising almost 100 years of research, we outline the findings of major academic studies on employee selection, rank-ordering common employee selection techniques by effectiveness.

validity table

Employee Selection Procedures

  • Perfectly accurate hiring (100% accuracy)

    Imagine having a crystal ball which allows you to make the perfect hiring decision every single time, that would be perfectly accurate hiring.


    Effectively, a licence to print money.


    However, knowing exactly how well every candidate would perform with 100% accuracy is, unfortunately, beyond the scope of any employee selection method (other than that magic crystal ball!).


  • Cognitive ability tests (65% accuracy)

    General cognitive ability is defined as “the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience.” (Gottfredson, 1994).


    Almost 100 years of research finds cognitive ability to be the strongest single predictor of job performance. This research shows that cognitive ability tests are 65% as effective as perfectly accurate hiring in medium complexity jobs (and are even more accurate in highly complex jobs).


    All ability tests (verbal, numerical, inductive etc) measure specific facets of general cognitive ability, making them collectively the most important element of any employee selection process.


  • Face-to-face interviews (58% accuracy)

    Very rarely will someone be given a job without first sitting a face-to-face (F2F) interview.


    And with good reason, F2F interviews are the second most effective employee selection tool after cognitive ability tests.


    However, they typically take several hours to plan, book, conduct, score, and feedback to candidates, representing a considerable investment into prospective employees. As a result, a cost-effective shortlist is essential, ensuring that only high potential candidates make it to a F2F interview.


  • Peer ratings / 360 (49% accuracy)

    Naturally, because you spend a great deal of time around your colleagues, their ratings of your performance typically hold significant validity.


    Although this is likely to be useful in performance appraisal and personal development, rarely will employers have access to these data for employee selection purposes.


  • Job knowledge tests (48% accuracy)

    If a candidate simply doesn’t know how to do the job, you cannot expect them to perform well.


    However, one must always bear in mind that not everyone needs to start with this knowledge, especially if you plan on training people anyway. This is especially true for graduates and apprentices, who simply aren’t expected to hold significant job-related knowledge.


    In these instances, cognitive ability tests will be far more useful, as cognitive ability dictates the extent in which people gain knowledge, which is far more important than how much knowledge they have so far.


  • Integrity tests (46% accuracy)

    Integrity is the most important behavioural trait when predicting performance. This is because individuals that score very low on integrity tests display counterproductive work behaviour, whereas those scoring highly display organisational citizenship behaviour and pro-social behaviour.


    This enhances the contextual performance of the employee, making them valuable assets to the team.


    Our TPAQ-45 complete profile and our TPAQ-27 express profile both measure integrity.


  • Telephone interviews (46% accuracy)

    Telephone interviews typically show high validity, although not as much as face-to-face (F2F) interviews. This is because they lack media-richness, causing the interviewer to miss out on important social cues.


    Although telephone interviews are a reasonable substitute to F2F interviews, they are a very ineffective supplement to F2F interviews.


    Because they are less effective than F2F interviews (46% vs 58%) and they measure the same thing, they won’t add anything over and above F2F interviews. Instead, if you plan on interviewing candidates in person anyway, we recommend using alternative screening tools.


  • Job try-out procedure (44% accuracy)

    Of course, getting someone to do the job itself provides a useful indicator of how well they will perform in future.


    However, job try-out isn’t a perfect way to hire candidates, as past / present performance doesn’t always predict future performance.


    Moreover, paying someone a salary makes job try-out especially expensive as an employee selection tool.


  • Assessment centres (36% accuracy)

    Perhaps the most surprising result from these meta-analyses is mediocre validity of assessment centres.


    Surely, assessment centres should be the most effective way of hiring people!?!


    The reality is only certain assessment centre exercises have any validity at all, and assessment centre ratings are notoriously prone to biases and the subjective influences of assessors.


    Due to their high costs (1,000 / candidate) and mediocre validity (36%), assessment centres are among the least cost-effective employee selection tools commonly used by employers.


  • Work sample tests (33% accuracy)

    The relatively mediocre validity of work-samples tests is another surprise. In fact, until recently we thought that work-samples tests were among the most effective selection tools.


    The reality, however, is that the workplace isn’t something that can be effectively simulated, and any attempt to do so will only ever encompass a fraction of the role’s requirements.


    This is why most publishers have recently stopped selling or supporting in-tray and e-tray exercises.


  • Emotional intelligence (32% accuracy)

    Regardless of the role or the organisation, people are required to cooperate with one another.


    Emotional intelligence allows you to work more effectively with people, and thus navigate the complex interpersonal domain. Although this matters more in some roles than others, its very rare that it won’t play at least some role in employee performance.


    Our TPAQ-45 complete profile and our TPAQ-27 express profile both measure emotional intelligence.


  • Interests (31% accuracy)

    Generally speaking, people tend to develop interests in line with their abilities.


    When someone discovers they are bad at something, more often than not, it serves as a deterrent, limiting their interest in that topic.


    However, gauging whether someone is genuinely interested in particular topic is difficult, as it takes almost no effort to simply feign interest.


  • Reference checks (26% accuracy)

    I suspect (although I don’t know for sure), that the mediocre validity of refence checks is due to how truthful people are when giving references.


    Most people believe if you can’t say something nice, don’t say anything at all.


    Moreover, when likable people underperform, people are more likely to give good references in spite of their performance.


    Overall, fairly mediocre results for reference checks.


  • Situational judgement tests (26% accuracy)

    Situational judgement tests (SJTs) show fairly mediocre levels of validity, but their true strength is what we call “incremental validity”.


    Although SJTs aren’t much on their own, research suggests they complement ability very well, providing unique insight.


    This is because decision-making has both behavioural and cognitive components, hiring managers well advised to assess candidates on both.


    Our Perceptions suite of situational judgement tests are examples of these assessments.


  • Conscientiousness (22% accuracy)

    The extent in which conscientiousness it predicts performance varies considerably between roles.


    Conscientiousness is most important in very process driven work, as being organised and diligent is especially important.


    However, research shows that in highly complex and innovative work, conscientiousness ceases to be important and cognitive ability takes-over.


    Our TPAQ-45 complete profile and our TPAQ-27 express profile both measure conscientiousness


  • Job experience (16% accuracy)

    Although in the early years, relevant work-experience can make a significant difference, after a few years in the role it ceases to be relevant.


    For example, if you haven’t learned how sales works after 5 years in a sales role, you simply never will, it doesn’t matter how long you stay thereafter.


    As a result, its rarely worth using job-experience as a selection criterion.


  • Emotional stability (12% accuracy)

    Emotional stability, although overall a fairly insignificant predictor of performance, can be quite important in highly stressful work.


    In stressful work, it becomes a useful indicator of turnover and employee retention, whereby those scoring low on emotional stability may be at risk of leaving.


    Naturally, the more stressful and high pressure the work environment, the more salient the effect of emotional stability


    Our TPAQ-45 complete profile and our TPAQ-27 express profile both measure emotional stability.


  • Years of education (10% accuracy)

    Simply staying in education isn’t likely to influence a person’s future job performance, as made evident by its low validity.


    Of course, smarter people are more likely to pursue advanced degrees and spend more time in education, but this correlation isn’t perfect.


    Plus, if you are interested in hiring smart people, you should be using ability tests, not length of education.


  • Extraversion (9% accuracy)

    For the most part, extraversion isn’t a relevant predictor of performance in the majority of roles.


    The exception to this rule, however, is highly interpersonal roles such as sales, where extraversion becomes quite important.


    This is especially true from a job-fit perspective and can have a major impact on employee retention and turnover.


    Our TPAQ-45 complete profile and our TPAQ-27 express profile both measure extraversion.


  • Agreeableness (8% accuracy)

    Agreeableness for the most part isn’t very predictive of performance in the workplace, especially in roles that require a thick skin or frequent confrontation.


    Indeed, research suggests that agreeableness is negatively correlated with performance in managerial roles, as highly agreeable managers get taken advantage of by their staff.


    However, agreeableness is important in caring roles and those with high emotional load, encouraging people to go the extra mile for those in need.


    Our TPAQ-45 complete profile and our TPAQ-27 express profile both measure agreeableness.


  • Openness to experience (4% accuracy)

    Openness to experience is perhaps the least important personality trait when it comes to performance.


    Although being open to experience is important when it comes to accepting change or being creative, only in certain roles does this translate to performance.


    As with many personality traits, openness to experience is more important when it comes to job-fit, employee engagement, and length of tenure.


    Our TPAQ-45 complete profile and our TPAQ-27 express profile both measure openness to experience.


  • Graphology (2% accuracy)

    Graphology is a pseudoscientific study of handwriting, whereby graphologists claim they can describe a person’s character based on their style of handwriting.


    Needless to say, the evidence does not support the use of graphology in employee selection.


  • Age (0% accuracy)

    Not only is it illegal to use age as an employee screening tool, it doesn’t work either.


    Best to steer clear of this one.


  • Random hiring (0% accuracy)

    And finally, picking people entirely at random.


    Although this seems ridiculous, many common employee selection tools have no validity either, and require a significant investment in either time, money, or both. For example, research shows that manual CV sifting is no better than random hiring but requires a significant investment in time and administrative effort.


    Although random hiring presents a massive opportunity cost, it doesn’t actually cost anything and so this option is actually more cost-effective than buying an assessment with no validity and hiring candidates with it.


Back to top