Monday, October 15, 2007

Incentives Matter

In the 1970s union leaders were concerned about the stress that air traffic controllers were under. They asserted that a sign of stress was if a controller nearly caused a collision. So a policy was enacted that if a controller had 2 or more near misses in a month then the worker would have a two week paid vacation. This graph illustrates the effects of the policy:


Anonymous said...

a) Citation?

b) post hoc ergo propter hoc

c) all libertarians are idiots

Anonymous said...

d) not including other factors like deregulation of the airlines and changes in the defination of "near miss"

e) lack of data before the red line

Jason (not the previous poster, just appending to the 2/3 valid reasons given)

David said...

I can't answer these all of these to their fullest extent but I can say that (b) only works if there's no logical story to connect them, much like how correlation doesn't automatically mean causation. But a good story exists here so unless you have a better one, simply repeating Latin phrases is not a persuasive argument.

I'm not aware of any changing of definition, but it's hard to believe effectively paying people for near misses wouldn't affect how they do their jobs.

The data goes back two months before the change. More would be better, but that's reasonable unless there's some systematic reason to believe it shouldn't be (i.e. an institutional change 2+ months back).

Jenny said...

So, near misses aren’t a sign of incompetence?