Selection is a complex decision-making science. Major factors which impact the outcome include:
- The base level of current performance. What do we have to improve on?
- How selective can we be?
- What is the validity of various selection methods?
- What is the cost of different aspects of the process?
Assessment Centres are considered very valid and fair but are resource intensive and costly, so they tend to be put at the end of a procedure. Similarly, management interviews are costly in terms of time.
Telephone interviewing can have a role in screening out ‘gross negative disqualifiers’, such as no citizenship, but they are probably a less reliable and valid measure of behaviour and ability than psychometric tests. Telephone interviewing is not necessarily cheap and in addition, the quality varies.
Resume screening suffers from similar problems as telephone screening. It is actually quite expensive and time-consuming to deploy resources for this purpose and the gain or return on investment is very low indeed. Even online applications have a cost and probably offer very little discriminative capability.
In general, it is considered best practice to use methods of good validity but low cost as close to the start of the process as possible. This is where psychometric tests come into their own. The closer up the funnel you can use them, the greater the return on investment.
Say you wanted to select 50 graduates from a pool of 1500 applications, in other words, people with limited experience and track record. You may be able to make some coarse screening decisions from their initial application, however, the time spent doing this is hardly worth it.
If you cull significantly from applications and resumes alone, you may easily be missing some of the better candidates and bringing through others who turn out to be not that good. Once you have rejected a candidate at some point, you can’t get them back.
So, a better method is to test all applicants online. If you just use an aptitude battery, you immediately get a cost-benefit by screening out candidates with a lower likelihood of job success. However, if you also test early on behavioural dimensions, then you can screen on other highly important variables as well, both quickly and efficiently, eg are they good at influencing outcomes, or do they prefer not to do this? How good are they at delivering results or do they prefer to spend more time on other things? Research from both the UK and the US has now consistently shown that combining a valid behavioural questionnaire with aptitude testing produces incremental validity ie better prediction of successful candidates (Saville et al 1996, Ruhfus 2001). Not only that, but using behavioural questionnaires as well is fairer – it produces less negative impact against minority groups for example.
Results can be managed very efficiently to produce a merit list against key criteria, using for example, Saville’s Job Profiler.
At this point you have a merit list based on high validity, without having to deploy expensive human resources.
You could at this point go straight to an assessment centre, by picking the top 200. However, before doing that, now is a good time to check those applications and resumes to see if there are any problems. If you have to reject someone at this point on a gross negative disqualifier, you can easily choose the next one down from the merit list, and so on. Far better to do this than to try and physically process 1500 applications.
Because you now have highly reliable and standardised information on all candidates, the assessment centre or interview can focus on clarifying any question marks and seeing that the tested behaviour translates into observable behaviour.
In fact, this process could be producing a result where applicants are on average 1 standard deviation above the mean in terms of performance – a considerable competitive advantage.
Notes
Saville, P et al, Journal of Applied Psychology, Wiley 1996.
Ruhfus, S, Paper presented to the Industrial and Organisational Psychology Conference, Sydney 2001, citing Professor R Jacobs, Penn State University.