Friday, November 12, 2010

Observer ratings of personality: An exciting possibility?

Using measures of personality to predict job performance continues to be one of the most active areas in I/O psychology. This is due to several things, including research showing that personality measures can be used to usefully predict job performance, and a persistent interest on the part of managers in tools that go beyond cognitive measures.

Historically the bulk of research on personality measures used for personnel assessment has been done using self-report measures—i.e., questionnaires that individuals fill out themselves. But there are other ways of measuring personality, and a prominent method involves having other people rate one’s personality. This is known as “other-rating” or “observer rating.”

Observer ratings are in some sense similar to self-report measures that ask the rater to describe their reputation (such as the Hogan Personality Inventory) in that the focus is on visible behavior rather than internal processes; in the case of observer ratings this takes on even greater fidelity since there is no need to “guess” at how behavior is perceived.

Research on observer ratings exists, but historically has not received nearly the attention given to self-ratings. Fortunately, in the November 2010 issue of Psychological Bulletin, Connelly and Ones present the results of several meta-analyses with the intent of clarifying some of the questions surrounding the utility of observer ratings and, with some caveats, largely succeed. The study was organized around the five-factor model of personality, namely Openness to Experience, Conscientiousness, Extraversion, Agreeableness, and Emotional Stability.

The results of three separate meta-analyses with a combined 263 independent samples of nearly 45,000 individuals yielded several interesting results, including:

- Self- and other-ratings correlate strongly, but not perfectly, and are stronger for the more visible traits, such as Extraversion and Conscientiousness, compared to those less visible (e.g., Emotional Stability).

- Each source of rating seems to contribute unique variance, which makes sense given that you’re in large part measuring two different things—how someone sees themselves, their thoughts and feelings, compared to how they behave. Which is more important for recruitment and assessment? Arguably other-ratings, since we are concerned primarily not with inner processes but with behavioral results. And it certainly is not uncommon to find someone who rates themselves low on a trait (e.g., Extraversion) but is able to “turn it on” when need be.

- Self/other agreement is impacted by the intimacy between rater and ratee, primarily the quality of observations one has had, not simply the quantity. This is especially true for traits low in visibility, such as Emotional Stability.

- Inter-rater reliability is not (even close to) perfect, so multiple observers is necessary to overcome individual idiosyncrasies. The authors suggest gathering data from five observers to achieve a reliability of .80.

- Observer ratings showed an increased ability to predict job performance compared to self-ratings. This was particularly true for Openness, but for all traits except Extraversion, observer ratings out-predicted self-ratings.

- Just as interesting, observer ratings added incremental validity to self-ratings, but the opposite did not hold true. This was especially true for Conscientiousness, Agreeableness, and Openness.

- Not only were observer ratings better at predicting job performance, the corrected validities were considerably higher than have been documented for self-report measures, and in one case—Conscientiousness—the validity value (.55) was higher than that reported previously for cognitive ability (.50), although it should be noted that the corrections were slightly different. Values before correcting for unreliability in the predictor were significantly lower.

Why do observer ratings out-perform self-ratings? Perhaps other-ratings are more related to observable job performance. Perhaps they do not contain as much error in the form of bias (such as self-management tendencies). Or perhaps because the measure is completely contextualized in terms of work behavior, whereas the questions in most self-report measures are not restricted to the workplace.

Despite these exciting results, some caution is in order before getting too carried away. First, the authors were unable to obtain a very large sample for observer ratings—sample sizes were around 1,000 for each trait. Second, there are currently a limited number of assessment tools that offer an observer rating option—the NEO-PI being a prominent example. Finally, there is an obvious logistical hurdle in obtaining observer ratings of personality. It is difficult to see how in most cases an employer would obtain this type of data to use in assisting with selection decisions.

However, even with that said the results are promising and we know of at least one way that technology may enable us to take advantage of this technique. What we need is a system where trait information can be gathered from a large number of raters relatively quickly and easily and the information is readily available to employers. One technology is an obvious solution, and it’s one of my favorite topics: social networking websites. Sites like Honestly (in particular), LinkedIn, and (to a lesser extent) Facebook have enormous potential to allow for trait ratings, although the assessment method would have to be carefully designed and some of the important moderators (such as degree of intimacy) would have to be built in.

This is an intriguing possibility, and raises hope that employers could soon have easy access to data that would allow them to add substantial validity to their selection decisions. But remember; let’s not put the cart before the horse. The question of what assessment method to use should always come after an analysis of what the job requires. Don’t jump headlong into other-ratings of personality when the job primarily requires analytical and writing skill.

This is just another tool to put in your belt, although it’s one that has a lot of exciting potential.

1 comment:

Anonymous said...

This is a great summary. Observer ratings of personality have a lot of potential further contributing to personality assessment beyond self-reports. The MTMM principle has been long forgotten in personality research, but with this kind of research available, the personality assessment will bring a more promising outlook to selection researchers and practitioners. In particular, you can embed observer ratings of personality into recommendation forms, application forms, self-introduction forms, background checks... The interent will make it a more doable option.