Goodhart's Law was used as an explanation of why those monetary targets used by Maggie et al in the 80s would never work: my, how the clever people mocked. Of course, those oh so clever people then went on to do exactly the same in spades: we ended up with targets for waiting times in emergency rooms, targets for child poverty and I'm sure it was only the intervention of an election which stopped us having targets for sheets of toilet paper successfully soiled.
For, while originally formulated to deal with certain problems in economics, Goodhart's Law really says that whenever you use a proxy as a target for what you really want to happen either the proxy will stop connecting to the target or people will game it. Thus we had patients waiting in ambulances so they wouldn't be in A&E, policies to bring those families just under the poverty level over it while ignoring the truly poor and, yes, those monetary targets going haywire two and three decades ago.
To see this on the hoof, taking all of the politics out of it, have a look at this:
By counting the cars in Wal-Mart’s parking lots month in and month out, Remote Sensing Metrics analysts were able to get a fix on the company’s customer flow. From there, they worked up a mathematical regression to come up with a prediction of the company’s quarterly revenue each month.
How excellent, we've a method of predicting the financial results from seeing how many people are going to the store. But of course as soon as this is known there will be gaming of the system. Sticking more cars into the parking lots will raise expectations of the financial performance. But no, you wouldn't want to do that, buy the shares and then stick cars in the lots. No, you stick the cars in the lots, wait for the shares to rise on the excellent results to come, then sell them short. As the satellite photos turn out to have been counting the cardboard models you've been putting up the shares will fall of course and you can buy back and retire on your ill got fortune.
Yes, it probably would be illegal, yes, it probably would be market manipulation but that's not the point here. Only that it's an example of what happens when we all decide to start using some glorious new metric to measure performance: as soon as we do someone starts gaming it and our new metric is now useless.
Sorry folks, but setting targets just doesn't work.
