Re: st: Estimating the CES-Translog Production Function

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

I have been preoccupied with other business so I missed this thread 
until today.  Still, I would like to reinforce Nick's point but from 
a different perspective.

Why are you trying to estimate a mixture of a CES and a translog 
production function?  This makes little sense in theory - a 
(generalised) CES is a specific functional form, while a translog is 
usually viewed as a flexible approximation to some unknown production 
function.  Even if you got some results there is no reason to believe 
that the coefficients would make any sense.  You should either 
estimate a CES, for which NL is appropriate, or a full translog (with 
the linear terms restored) which can be done with normal linear 
regression.  It is hardly surprising that you cannot generate 
reliable results with this specification.

You may have the idea of testing the CES against the translog by 
nesting them, but this is not the way to do it.  In addition, you 
ought to be aware that just estimating production functions without 
thinking about how the input data is generated can be very 
misleading.  Hence, you should read up on production & cost functions 
and how they can be estimated before you go any further.  There are 
plenty of books available but you could look at the chapter on cost 
functions in Kenneth Stewart's "Introduction to Applied Econometrics" 
or Ernst Berndt's text "The Practice of Econometrics".

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/