Saturday, December 12, 2009

A cautionary tale: Where numbers come from

Megan Carpentier posted on AirAmerica suggesting that the magic number 30 appears much more frequently than one would expect in reports of collateral damage of air strikes in Afghanistan. She attributes that to a rule that field commanders are required to obtain Washington permission for an air strike that would have more than 30 killed as collateral damage. Field commanders probably don't like bothering top officials in Washington nor waiting for permission, and so decide that strikes are unlikely to cause more than the magic number of deaths and injuries. And of course, in post reporting of actual casualties military bureaucrats are not likely to tell Washington that the rule of 30 had not been followed. Do they round down the estimates? It certainly seems possible.

It is always a good idea to look at the distribution of repeated numerical estimates, and if that distribution appears unlikely, to question the accuracy of the estimation process!

No comments: