I have been asked to research some ways that we as a company can eliminate performance appraisal rater biases.  We are in the process of using performance appraisals to determine raises and bonuses for our employees.  One issue we face is that of rater biases.  We have a number of different offices and different managers conducting appraisals.  I would be very appreciative for information anyone can send me.

 Thank you in advance.

 

There are a lot of different biases in rating performance.  The most pronounced is the "what most easily comes to mind" bias.  For employees who do not have easily countable work their appraisals tend to be dominated by social performance or the last notable thing they did.  This is best avoided by supervisors keeping a performance diary on subordinates but supervisors rarely keep up with the diary.

Another is rating everyone in the middle.  That can be avoided by using scales with an even number of items ("forced chocie").

Another is the tendency to rate everyone very highly.  This is where you get into very different theoretical approaches to the purpose, value, and accuracy of ratings.

 

 

Are there any specific biases your company is experiencing in the appraisal process that you want to address? 

Im not even sure if this is a bias, but one of our departments rate all of their employees so much higher then other departments.  We are using a numeric system from one to ten and then the they take the average rating for that overall category of performance.  I was just looking for a way to take all departments and eliminate any biases before passing out raises and bonuses.

Perhaps you need to have an appraisal form that is tailored to each specific department then rather than just the overall core values of the company?

Otherwise I agree with TXHRGuy, backing up rating with specific examples of performance helps make the appraisal objective.

bebarrick:

Im not even sure if this is a bias, but one of our departments rate all of their employees so much higher then other departments.  We are using a numeric system from one to ten and then the they take the average rating for that overall category of performance.  I was just looking for a way to take all departments and eliminate any biases before passing out raises and bonuses.

 

That is a bias, and there are different ways to deal with it.  I would like to point out that research indicates that most people can't make meaningful distinctions after about 7 points on a scale: 10 is way too many.  If you are having a bias toward high ratings and assuming that the rater is acting in good faith, then either all the employees are awesome on all the rating factors or the rater is using some normative scale against which everyone compares favorably.  If you are not already telling them what to compare against, adding that would help.  There are companies that require raters to order their subordinates from first place to last in general or on some factor.  There are a lot of schemes for dealing with high rater bias and none of them are perfect.  Training can help if the raters take it seriously.  Generally, HR takes it more seriously than anybody else: they just want to get theirr job done.

Something that may seem simple, but has worked in the past for me is to require a second set of eyes (approval if you want to call it that) on the performance review before it is given to the employee.  This way the other person can say what about this and what about that and why did you put this down there.  At my last company, the general manager of the operation read all of the performance reviews before they were given out.  Each deprartment did the reviews at a different time so he wasn't bombarded with a lot to read.  The General Manager had me review many of them as well to make sure the review matched any issues we had had with the employee and were being fair across the board.

Another way is, where feasible, use multiple raters.

 

If you break it down to a four pount scale, "above expectations", "meets and occasionally exceeds expectations", "meets expectations minimally", "Does not meet expectations", that will help with rater confusion.  You can, if necessry, put a hard limit on how many can be "above expectations" without review by HR or other appropriate source.

You can also continue doing what you are doing but also do a forced ranking on each dimension if the groups are small enough.  "Please rank from greatest to least performer on the following dimension: <e.g., quality of work>."

 Use the forced rankings to put the other things into perspective.  At some poine, you can take the forced rankings back out when they get the idea.

 

"How can your lowest ranked person be constantly exceeding expectations?"