Tuesday, January 3, 2012

Randomization at work

The story of a firm that used random assignment to study the effects of allowing telecommuting on worker outcomes such as productivity and absenteeism.

My sense is that firms do remarkably little of this sort of thing, though I have heard stories of catalog merchants randomly assigning prices across catalogs as a way of estimating their demand curves. And, of course, direct mailers of other sorts do a lot of experimentation with their mailings.

I think part of the reason that firms do not do this as much as they might is that business schools tend not to teach econometrics in a way that would encourage students to think this way. Put differently, my sense is, first of all, that most MBA programs do not require much in the way of statistics or econometrics to begin with and, second of all, that what they do require is often not taught in a way to emphasize the identification of causal effects.

Hat tip: Sue Dynarski on google+

No comments: