Dear All,
Thanks a lot for your engaged response, I really appreciate. I did
not elaborate because I thought there are some obvious references that
I am missing. It turns out that it is unlikely to be true.
I agree with most of the concerns (definition of capital, TFP increase
vs. mark up increase, finished goods as intermediates). I don't think
the extant literature deal very precisely with these, mostly due to
data limitations, at least in developing countries. Now I also agree
with Nick that some if not all are relevant for mfg firms. Questions
are 1)how do parameters differ if we use the same functions and same
definitions of variables for mfg and service firms and 2) how do
parameters differ if we have different definitions of variables and 3)
should we use different definitions of variables?
I disagree with one aspect of Nick's example. While certain natural
phenomena like growth of trees can be exactly measurable, things like
capital stock are more elusive.
My data comes from balance sheets of firms. Hence it has sales,
assets, wage bill, capital stock, investments, fuel, material and a
bunch of other cost items. No plant level information.
Thanks again for taking your time to engage in this.
Best,
Prabal
On 10/27/09, Nick Cox <[email protected]> wrote:
> Thanks for this. Perhaps my question was too cryptic, but I don't see an
> answer here, or elsewhere in this thread. I can fit power functions to
> raspberry bushes and redwood trees and I am not surprised that the
> variables' values differ and possibly the parameter values too. But I
> don't need different software in the two cases. The equations are the
> same. What differs in your case?
>
> Nick
> [email protected]
>
> Prabal De
>
> My Bad. It does stand for Total Factor Productivity which is
> contribution of the 'residual term' A after factoring out
> contributions of labor and capital in a production function. For a log
> Cobb-Douglas production function
>
> logY = logA + (alpha)logL + (1-alpha)logK
>
> Nick:
> Since this is essentially an "accounting" procedure, TFP will be
> mechanically high if you have low labor and capital. And
> Levinsohn-Petrin procedure controls for endogeneity in capital stock
> by instrumenting with intermediate inputs like fuel.
> Now intuitively, for service sector firms both physical capital and
> fuel are much less important. Then one argument is that they DO have
> very high TFP. I haven't found a logic contrary to this myself except
> toying with other intermediate inputs like communication expenses(nor
> any reference), but then there are smarter economists around and in
> Statalist. I still hope someone can shed more light on this issue.
>
> On 10/27/09, Martin Weiss <[email protected]> wrote:
>>
>> Prabal may also want to let statalisters know what "TFP" stands for...
> Let
>> me guess: "Total Factor Productivity"?
>>
>> As far as I can tell, not even the -rather comprehensive- article
>> introducing -levpet-
>> http://www.stata-journal.com/sjpdf.html?articlenum=st0060 mentions
> this
>> term....
>
> Nick Cox
>
>> Just curious, as I only understand some of this and it's not my field:
>> why do different numbers require a different logic?
>
> Prabal De
>
>> I am trying to estimate a production function for service sector
>> firms using <levpet>. However, the usual method for manufacturing
>> sector is giving very high TFPs as naturally the service sector firms
>> use less physical capital. Is there is variation of the levpet
>> procedure for service sector firms?
>
> *
> * 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/
>
*
* 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/
