pick & twist part 1

The point made by Sean in an earlier post is well-taken. I offer the following, possibly apocryphal quote from David Romer:

Good papers in economics have three characteristics: A viewpoint. A lever. A result.

I think this is not limited to economics. And it is a useful lens through which to understand research based on the “pick the data carefully” approach. In such papers (in addition to Isabel’s magnum opus, I also think of Roberto Fernandez’s study of skill-biased technological change in a single plant), the careful picking *is* the lever.

A world in which the only type of acceptable lever is a valid instrumental variable would be an intellectually poorer world. At the same time, picking and twisting strategies are potentially complements, not substitutes.

An orgtheory-relevant illustration of the pick & twist approach is Scott Stern‘s analysis of wage-setting in the labor market for pharmaceutical scientists. Scott not only picks the data carefully; he also chooses to conduct the study at the job offer level, rather than the individual level. And he demonstrates how fully controlling for individual heterogeneity overturns the conclusions that would have been reached in an individual-level research design [to this day, this paper remains one of my all-time favorites].

Written by Pierre

August 23, 2009 at 3:38 pm

%d bloggers like this: