Re: st: stochastic frontier model with technical efficiency determining factors

[Gordon Hughes <G.A.Hughes@ed.ac.uk>]

Nathalie,

In Stata you have to use -frontier- if you want to make the 
efficiency term a function of explanatory variables - this is 
referred to as the conditional means model.  There is a cost to this 
because frontier does not offer a cluster option for estimating the 
variance-covariance matrix which is where the panel nature of the 
data would have an effect.

Outside Stata it is possible to carry out panel estimation of 
frontier models with conditional means using Limdep, since much of 
the literature on this approach originates from work by William 
Greene, the primary author of Limdep.  For my own purposes, I wrote a 
script to estimate this type of model in gretl, which is an open 
source econometrics package.

Since any such generalisation is basically a matter of modifying the 
likelihood function, it would not be hugely difficult to write an 
extended version of -xtfrontier- to incorporate conditional means of 
the truncated normal distribution.  However, I would warn that it can 
be very difficult to get such estimators to converge and they are 
often quite sensitive to the starting values.

Gordon Hughes
g.a.hughes@ed.ac.uk

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/