|Attribution bias and risk management performance benchmarking|
Wednesday, October 26, 2011
Performance Benchmarking and Attribution Bias
One of the challenges for measuring the effectiveness of risk management (or any type of management system for that matter) is a little glitch in human perception known as attribution bias. Attribution bias is simply our tendency to invent explanations and to attribute things to a particular cause (whether real or imagined). These attributions serve to help us understand the world and give us reasons for a particular event.
For example, let’s say that Bill gets sacked from his job. He will attribute his sacking to his boss being an asshole and that he is better off not working for that company anyway. He can thus make sense of his misfortune without having to accept any responsibility. The fact that Bill has been sacked from a series of jobs for poor performance can be easily overlooked. Bill goes even further and attributes having a series of bad bosses to bad luck - despite the fact that he is the only common denominator each time. Without the attributional explanations, Bill will be very embarrassed and discomforted to believe that his performance at work has been the cause of his sacking(s).
Psychologists describe attributional biases as a class of cognitive errors which are triggered when people evaluate the dispositions or qualities of others based on incomplete evidence. The key element here is that biases are 'errors' that are hardwired into us. There are many good reasons for us to have a range of biases. Attribution bias for example, allows us to maintain self-esteem and a sense of purpose without having to face our own behavior. When it comes to risk management benchmarking however, attribution bias is simply another way to introduce errors.
For example, when that crime in the United States started to decline in 1992, in complete contrast to the projections of criminologists, many politicians were quick to attribute this drop to their policies of 'zero tolerance' policies or extra police, etc. Steven Levitt of the University of Chicago and John Donohue of Yale University meanwhile, suggest that the drop was due to a completely different and unexpected cause.
They suggest that the the cause was the landmark legal case of Roe v. Wade, in which the United States Supreme Court controversially legalized abortion. Levitt and Donohue persuasively argue with detailed statistical analysis, that the absence of unwanted aborted children, following legalization in 1973, led to a reduction in crime 18 years later. The years from 18 to 25 would have been the peak crime-committing years of the unborn children and hence resulted in lower crime. For many years criminologists and politicians alike took credit for various risk management treatments as being the cause of this drop in crime. It probably led to a lot of wasted resources and unwarranted bonuses or at lease congratulatory back-slapping.
The inference however is clear - attribution bias can easily have you wasting your precious risk management resources. At the very least you'll end up with a risk management benchmarking which is based on the wrong performance indicators and at the very worst, you can make things worse.