Saturday, May 12, 2012

Performance ratings and the ‘above average effect’

“Performance ratings will be shared with the employees next week. We expect employee attrition to go up significantly in the next few months”, said the HR Manager.

It is a fact that in many organizations the attrition percentage goes up in the months after the annual performance ratings are announced. Some of this is because of the process linkages. Salary hikes and bonuses (that are linked to the performance ratings) usually follow soon after (or along with) the announcement of the performance ratings and it might make logical sense for employees to receive the bonus (after all one has worked for an entire year to get that) and the higher salary and then negotiate a better salary (with a  new company) based that. But some of the resignations are a direct emotional reaction to the performance ratings. Based on my experience across multiple companies (as an employee and as a consultant), I have often wondered why the sharing of performance ratings is such an unpleasant experience – both for the employees and for the Managers of the employees.

There could be many reasons for this. The performance objectives and targets might not have been properly defined or agreed upon. There might have been changes in the context or factors outside the employee’s control that made the targets unreasonable/impossible to achieve. The performance feedback might not have been given regularly and accurately (managers often try to ‘soften’ negative feedback) and hence the rating might have come as a surprise for the employee. But I feel that most of the unpleasantness of the situation is related to a psychological phenomenon known as ‘superiority illusion’ or the ‘above average effect’.

'Illusory superiority' is a cognitive bias that causes people to overestimate their positive qualities and abilities and to underestimate their negative qualities, relative to others. This manifests in a wide range of areas including intelligence, possession of desirable characteristics/personality traits, performance on tests and of course ‘on the job performance’ (for which performance rating is an indicator). While the exact percentages can vary based on the social/economic/cultural context, typically in a group at least 75-90% of the members rate themselves as 'above average'.

This fact (that at least 75% of the people rate themselves as 'above average') creates trouble when it comes to performance ratings. These days companies are keen on ‘differentiating based on performance’ (say ‘to build a performance driven culture’) and this would mean that when it comes to performance ratings, the relative performance of the employees becomes a critical factor apart from the absolute performance (performance against agreed upon targets). Whether or not a fixed percentage distribution of ratings are prescribed, some sort of a ‘normal curve’ emerges. Typically, the positively differentiated performance ratings (i.e. if we have a 1 to 5 scale with 1 being the lowest and 5 being the highest; ratings of 4 and 5 ) form about 25%. Thus only about 25% of the employees will get ‘above average’ performance ratings. The arithmetic is simple and the conclusion is inevitable. If at least 75% of the employees consider their performance to be ‘above average’ and only 25% of the employees will get ‘above average performance ratings’, then at least 50% of the employees will be disappointed with their performance ratings. Thus, sharing of performance ratings is likely to be an unpleasant experience – both for the employee and for the Manager.

Now let us look at this from the Manager’s point of view. Experienced people managers know that the problem described above will happen (though they might not be aware of the exact percentages/degree of the problem). But they can’t do much about it as the two critical factors (employee’s tendency to rate their performance as 'above average' and the maximum percentage/number of the ‘above average performance ratings’ that the Managers can give) are largely outside their control. Managers do what they can. This can range from ‘expectation management’ to ‘pushing for a higher percentage of above average ratings for their team’ to ‘providing other rewards and recognition to compensate for the unpleasantness created by lower than expected performance ratings’ to disowning the performance ratings (blaming it on HR and/or senior leadership). But these are of limited utility as they are not addressing the core problem. Also, this can lead to a situation where the employees lose confidence - in the Manager and in the Performance Management System. Another option for the Manager is to staff his/her team with people who have a low self-image (masochists are welcome!). But if the Manager wants the employees to have high self-belief/confidence when dealing with customers and low self-belief/confidence when interacting with the Manager, then it calls for a Janus-faced personality. While such personalities can be found in abundance in extremely hierarchical organizations (see Followership behaviors of leaders), it might not be a viable strategy for ‘normal’ organizations!

Logically speaking, grappling with this problem for an extended period of time and gaining insights and wisdom from the struggle should help the Manager to be more reasonable when estimating his/her own relative performance (and hence the performance rating he/she deserves) and to be more understanding when the Manager’s Manager tries to share and explain the Manager’s performance rating. But, as the studies in ‘Behavioral Economics’ have demonstrated, being aware of a ‘bias’ need not necessarily help one to overcome the bias! No wonder managers often dread the entire business of performance ratings – giving the performance ratings to their team and receiving their own performance ratings!!!!

The research done on the ‘above average effect’ has thrown up some interesting findings that might help us (at least to some extent) in dealing with this problem in the context of performance ratings. It has been found that the individuals who were worst at performing the tasks were also worst at estimating their relative performance/degree of skill in those tasks. It has also been found that given training, the worst subjects improved the accuracy of their estimate of their relative performance apart from getting better at the tasks.

Another possibility here is to make the performance ratings less dependent on relative performance and more dependent on absolute performance (performance against agreed upon targets) or to increase the percentage of 'above average ratings'. But these kind of steps can go against the performance management philosophy of the organization (of differentiation based on performance) and hence impractical. If the context so permits, standardization of performance objectives/targets for a particular role and making the information on the performance of employees on the objectives/targets available to all can also be looked at. Research shows that self-evaluation (especially in comparative contexts) is driven primarily by an intuitive ‘heuristic process’ as opposed to a logical/effortful ‘evidence-based process’. However, by making valid & reliable data on relative performance available and by encouraging the employees to look at it (and may be even participate in an open discussion about it) before they do the self-evaluation (and evaluation of their relative performance), the influence of the ‘evidence-based’ part on the decision making-process might increase.

Again, we can make discussions on the challenges related to the 'above average effect' part of the performance management related communication and training  for employees and Managers. May be, we can even build in some 'nudges' (like asking the employees to write down three things that their peers have done better than them - as part of the self  assessment) that will prompt them to deal with their cognitive bias (of superiority illusion) in a more rational manner.

Apart from this, ensuring that basics of performance management - performance planning, coaching, feedback and review - are done well also helps, though they don’t directly address the problem we are discussing. It is similar to ‘taking antibiotics for dealing with a viral infection’. While they don’t solve the core problem (virus) they do help in preventing secondary infections and hence has some utility in some cases (especially when effective anti-viral drugs are not available and the possibility of secondary infections are high)! The problem we are dealing with here is too 'human' to be completely solved by 'performance management techniques' and we have to live with it to some extent as the price for being human!

Any ideas/comments?


John T said...

Very thought provoking article. One aspect (that I have seen as a HR professional myself) is the tendency of employees to compare negatively - "he is worse off than me in coding / QA still gets higher rating than me, so why didnt I?". Its the same rhetoric we see from motorists on Indian roads - "he also went through the no entry and did not get punished so why should I?" when caught by a cop!! Until this mentality goes away and introspection is based on absolute benchmarks, it will be difficult to deal with such employees.

bombay dosti said...

I knew of a line manager who made the performance appraisal system more transparent to deal with this issue. He asked the team members to present their accomplishments/achievements to everyone in the team and then asked for their anonymous ratings for each of their team members. The final ratings were still based on the discretion of the manager. The manager believes that this mechanism helped create a group buy in on who their relative performances. Somehow, I feel, even when 75% people think that their performance is above average, the best performers stand out and would be accepted by all??? I also tend to believe that people know when they have not achieved as per expectation. So, can we assume that the major issue would be for those who are at the border?

Although, one need not make absolute achievements as a mechanism, I am reminded of another strategy that a line manager used. He would have minimum mandatory targets and some stretch targets. If you dont achieve the minimum targets, you are below par and if you achieve the stretch targets you are above par. The trick then would lie in designing the right goals. However, the sector plays a major role in the ease of goal setting. A sales organization might be able to set goals more easily than a research organization.

Prasad Kurian said...

@ John and @ bombay dosti – Thank you very much for your comments

One of the factors that lead to ‘superiority illusion’ is the desire to maintain a favorable self-image. Even if we provide data on relative performance on pre-defined performance parameters (e.g. person x has sold more as compared to y), individuals can bring in other parameters (that are not in the standard/pre-defined list) and/or context specific factors (unique to their performance context) and conclude that their relative performance is better and hence they ‘deserve’ a better rating. Hence, we have a more fundamental issue here – people using a definition of ‘good’ (i.e. what good performance means) that is convenient for them (i.e. aligned to what they are good at/what they have done better than others). Also, this ‘definition of good’ (since this is essentially an intuitive one), can vary across time/context to suit the particular employee better!!

GoRo said...

What if you have a manager who takes the 'above average bias' to heart, and deducts 2 points from every single category?
Wouldn't you call that some sort of bias? Or reverse bias?
So, after being as honest as possible with his numbers, a worker should ADD 2 to every category in anticipation of his manager lowering by 2?
Somewhere along here, with BOTH management and worker being somewhat forced to be dishonest, any true worth of the process is LOST.

Prasad Kurian said...

@GoRo Thank you. To me, the best way to deal with a possible bias is to be aware of it and also create behavioral nudges to reduce the risk of it affecting the quality of the decisions. Mathematically, 2-2 = 0 and hence the biases can cancel out. In practice, there is no way of knowing if the bias on the part of a particular employee is going to add a positive offset of 4, 3, 2, 1, 0.5, 0, -1 etc. Remember, the above average bias is a statistical phenomenon in a group of employees and it can't predict what exactly will happen in the case of a particular employee. Of course, there are variations across groups also. So we can't bank on two wrongs acting in perfect harmony to make the answer right! So it is much better to manage the possible bias (and hence prevent it from happening to the best possible extent) than to assume it and correct for it using a fixed algorithm. Actually, this is the kind of approach that often coverts budgeting to such a comical exercise (you add buffers thinking that Finance will any way cut it)!