The last two posts have had me thinking about potential output. I had an idea that I'd like to think I may do something with eventually, but I probably won't. This semester in my microeconometrics seminar a lot of people interested in economic development have lead the discussion and presented their own papers, and many have used stochastic frontier analysis (SFA), an efficiency estimation technique common in the development literature. SFA essentially has two error terms - one is a classical mean-zero error term and the other is the "technical inefficiency component" that is the departure from the technical frontier. Obviously assumptions around the error structure are going to influence the resulting estimate of inefficiency.
My understanding of potential GDP estimates is that they use things like medium- to long-term growth rates, and some standard macro relations (Okun's law for example) with capacity utilization and unemployment statistics to estimate economic potential.
It seems to me another way you could do this is to take state-level data from the BEA (goes back to the 1960s only, but that's OK) and do an SFA estimate of potential output. It would be akin to the capacity utilization approach conceptually, but you'd actually be getting the production function parameters from the data rather than by (what I assume is) assumption.
Any thoughts on the idea?
I haven't heard about SFA for a long time -- I'm glad you brought it up, because back then I didn't have the experience to understand how it worked. Based on the Wikipedia article you link to, it looks like SFA is essentially adding arbitrary structure to TFP, which is what you would get with just a mean zero error term. Seems pretty strange to me. How do people who use it justify the normal 'random shock' component vs the arbitrary mean zero 'technical inefficiency' part of the error?
ReplyDelete