Predictability of methods for selecting employees

In 1988, Professor John Hunter of Michigan State determined that the typical hiring interview is only 57% effective in predicting subsequent success in a job, meaning that the typical interview is only slightly better than flipping a coin.

In the July-August 1999 issue of the Harvard Business Review, an article titled “Hiring Without Firing” identified that 30% to 50% of all executive-level appointments end with firing or resignation. This turnover statistic is significant when you consider that management positions are not only the most important positions in the organization, but the positions that have the greatest number of face-to-face interview time. As such, one might expect that people who were hired for executive positions would have been the most-watched candidates, yet a third to half of those hires have a very short “tenure.”

The Harvard article and the study by Professor Hunter would certainly lead to the conclusion that better methods of assessment should be used, not just executive candidates, but all hiring candidates. The question is, “Which methods are the best?”

When searching for the best practices, I found a 1998 study (Schmidt, FL and Hunter, JE (1998), “The validity and applicability of selection methods in staff research: Practical and theoretical implications of 85 years of research results,” Psychological Bulletin, 124, 262-274), which helped focus my approach to interviewing. Based on meta-analytic findings, this study presented the validity (R) of 19 selection procedures for predicting job performance. The procedures with the highest validity for predicting performance at work were:

o Work test tests (R = .54)

o General psychic test (R = .51)

o Structured interviews (R = .51)

o Peer Rating (R = .49)

o Job proficiency test (R = .48)

o Training & experience of behavioral consistency (R = .45).

At the lower end of the validity scale were the following procedures:

o unstructured interviews (R = .38)

or traditional reference control (R = .26)

o Years of job experience (R = .18)

o Educational year (R = .10)

o Interests (R = .10)

or age (R = 0.01).

The best-known conclusion of this 1998 research project is that for companies that hire graduates who have no prior experience in the job, the most valid predictor of future performance and learning on the job is general mental ability (i.e., intelligence or general cognitive ability).

Here, it should be noted about the practical relevance of General Mental Ability (GMA) in this study. The predictive power of GMA listed above at R = .51 is the validity rating for jobs that rank in the middle complexity range. The current study from this GMA study revealed the following validity results for different levels of complexity by position:

o Professional and managerial jobs (R = .58)

o High level complex technical jobs (R = .56)

o Medium complexity jobs (R = .51) (This represents 62% of jobs in the US economy, including middle-class white collar jobs such as office and administrative positions and skilled blue collar jobs.)

o Semi-qualified jobs (R = .40)

or unskilled jobs (R = .23).

These data indicate that GMA becomes an important predictor of job performance when the level of complexity increases in one position. However, one cannot discount other factors such as behavior, experience, etc. and their importance in helping predict success in a job.

This study presents strong evidence to suggest that GMA, along with positive indicators from other assessment approaches, will provide a high correlation between success in higher-level complexity positions.

The truth is that no “silver bullet” selection method exists, and this research does not suggest one method compared to other methods. As with any decision-making process, a manager must collect as much data as possible about a candidate and then use his / her intuition and experience to make the best possible hiring decision.