Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: Technical Efficiency with Stata sfpanel command
From
Federico Belotti <[email protected]>
To
Stata List <[email protected]>
Subject
Re: st: Technical Efficiency with Stata sfpanel command
Date
Tue, 25 Feb 2014 12:07:48 +0100
Dear Aristeidis
as noted by Nick, I would need a copy of your data and code (both in Stata and Frontier 4.1) in order to understand what is going on. If you believe this is appropriate, you can send (privately) them to me.
In the meantime, let me say something about your model specification. As discussed in Kumbhakar and Lovell (2000, p.93), you can estimate an input distance function by exploiting the fact that it is linear homogeneous in the inputs and by normalizing the input vector in some way. You can surely select different normalizing strategies but from what you wrote in your post my guess is that you have not chosen at all. As far as I know, the most used strategy is that proposed in Lovell et al. (1994) in which the input distance function is normalized by one of the inputs (the so called ``ratio'' model). In this model, the dependent variable should be the reciprocal of the normalizing input, and the regressors are the outputs and the normalized inputs.
Moreover, you wrote “...I tried the exact model specifications on Frontier 4.1 and the model produced results”. In my experience, the fact that Frontier 4.1 have produced results does not ensure that the estimation process has reached the convergence.
Hope this helps
Federico
References
Kumbhakar and Lovell (2000) Stochastic Frontier Analysis, Cambridge University Press
Lovell CAK, Richardson S, Travers P, Wood LL (1994) Resources and functioning: a new view of inequality in Australia. In: Eichhorn W (ed) Models and measurement of welfare and inequality. Springer, Berlin
On Feb 24, 2014, at 10:26 PM, Aristeidis dadoukis <[email protected]> wrote:
> Dear Nick,
>
> Thank you for the comments. I tried the exact model specifications on Frontier 4.1 and the model produced results. This leads me to believe that I am not doing something right when using Stata. I tried retreating to a simpler model (ie removing in turn 1 input and 1 output) as you suggested, however it still did not work on Stata. It seems that no matter what I try in Stata the MLE does not converge. I would really appreciate any ideas as to where I should start looking into.
>
> Thanks and regards
> Aristeidis Dadoukis
>
>
>
>> ________________________________
>> From: Nick Cox <[email protected]>
>> To: "[email protected]" <[email protected]>
>> Sent: Friday, 21 February 2014, 15:09
>> Subject: Re: st: Technical Efficiency with Stata sfpanel command
>>
>>
>> -sfpanel- is user-written, as you are asked to explain:
>>
>> SJ-13-4 st0315 . . . . . . . . . . . Stochastic frontier analysis using Stata
>> . . . . . . . . . . . F. Belotti, S. Daidone, G. Ilardi, and V. Atella
>> (help sfcross, sfcross_postestimation, sfpanel,
>> sfpanel_postestimation if installed)
>> Q4/13 SJ 13(4):719--758
>> estimates cross-sectional and panel-data stochastic frontier
>> models
>>
>> The authors may (should) have good answers, but otherwise
>>
>> 1. Difficulties fitting a model to your data could just stem from your
>> data being unsuitable for your model, and without a sight of your data
>> it's difficult for anyone to comment.
>>
>> 2. People who do this kind of thing (not me) would surely be
>> interested in whether you have success with fitting this model in
>> other software, which does not seem to be spelled out here.
>>
>> 3. The usual advice if a model doesn't fit is to retreat to trying a
>> simpler model, although I have no idea whether that even makes sense
>> in this case.
>>
>> 4. Regardless of whether Battese and Coelli (1995) is utterly
>> standard in this field, you are asked to avoid minimal name (date)
>> references on Statalist.
>> Nick
>> [email protected]
>>
>>
>> On 21 February 2014 14:53, Aristeidis dadoukis <[email protected]> wrote:
>>> Dear Statalist members,
>>>
>>> I am new to Stata, so far I was using the Frontier 4.1 software. I am interested in estimating an input oriented distance function using the Battese and Coelli (1995) model specifications, and then deriving scale efficiencies for individual banks.
>>>
>>> My model includes 3 inputs (x), 3 outputs(y) and 3 exogenous variables(z). I have prepared the data for the translog function in the same way as when using the Frontier 4.1 software (ie arithmetically mean adjusting the inputs/outputs and generating the translog terms). My sample covers 8 years and 15 obs per year.
>>>
>>> However, when using the sfpanel command the MLE does not converge. The command that I used is the following (please note that I also included the translog terms in the command)
>>>
>>> sfpanel lnx1 lnx2 lnx3 lny1 lny2 lny3, model(bc95) emean(z1, z2, z3)
>>>
>>> I would appreciate any advice on what might be wrong.
>>>
>>> Aslo, what are the steps for estimating the scale efficiencies through Stata?
>>>
>>> Thanks and regards
>>> Aristeidis Dadoukis
>>>
>>>
>>> *
>>> * For searches and help try:
>>> * http://www.stata.com/help.cgi?search
>>> * http://www.stata.com/support/faqs/resources/statalist-faq/
>>> * http://www.ats.ucla.edu/stat/stata/
>>
>>
>> *
>> * For searches and help try:
>> * http://www.stata.com/help.cgi?search
>> * http://www.stata.com/support/faqs/resources/statalist-faq/
>> * http://www.ats.ucla.edu/stat/stata/
>>
>>
>>
>
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/faqs/resources/statalist-faq/
> * http://www.ats.ucla.edu/stat/stata/
--
Federico Belotti, PhD
Research Fellow
Centre for Economics and International Studies
University of Rome Tor Vergata
tel/fax: +39 06 7259 5627
e-mail: [email protected]
web: http://www.econometrics.it
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/