Bookmark and Share

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: Oaxaca decomposition for Tobit


From   Nils Braakmann <[email protected]>
To   [email protected]
Subject   Re: st: Oaxaca decomposition for Tobit
Date   Wed, 19 Jan 2011 06:56:46 +0000

Hi Marco,

regarding the detailed variable decompositions: Apart from the
familiar problem with dummy variables (explained share depends on the
category  used as the base alternative), I have never heard of the
problem. I have also seen quite a lot of papers published in good
journals that use detailed decompositions. To be honest, I also fail
to see the econometric reason why detailed decompositions should be
unstable, so I would also like to see some references backing Farida's
claim. Similarly, I don't get the issues with the output of -oaxaca-.
As far as I can tell, the output (at least if you use linear methods)
matches the results from by-hand-calculations using the standard
formulas, so I really don't know what Farida's problem is. On the
Tobit-decomposition: There is a chapter on decompositions methods by
Nicole Fortin, Thomas Lemieux and Sergio Firpo forthcoming in the new
Handbook of Labor Economics (Vol 4). I think it's available as NBER
Working Paper 16045. That might be worth a look (Stata files are
available on Nicole Fortin's website).

Cheers,
Nils

On Tue, Jan 18, 2011 at 11:18 AM, Marco Ercolani
<[email protected]> wrote:
> Dear Statalisters,
>
> 1. Thank you to Farida for his suggestions. In particular I am keen to know the reasons why the "detailed variable decompositions" are unstable. I teach (aggregate) Oaxaca decompositions so I am interested in specific references on this problem. My research decompositions use a moderately large number of observations, 200,000 to 2,000,000 and the "detailed variable decompositions" appear to work quite well and the results make sense. Perhaps the problem is just one of degrees of freedom? Or is it something else? Please let me know.
>
> 2. Thank you to Maarten for moderating our excesses. I work among economists and I am therefore used to a daily discordant tone. In time I have become tone deaf ;-)
>
> 3. I would still like to hear from anyone who knows how to deal with my technical questions of how to carry out the detailed Blinder-Oaxaca decomposition for Tobit in Stata. In particular I would like to know if the multitude of marginal effects in Tobit creates problems.
>
> All the best,
>  Marco.
>
>
>
> Dr Marco G. Ercolani
>  Senior Undergraduate Tutor
> Department of Economics
> University of Birmingham
> Edgbaston, B15 2TT
>
> ________________________________________
> From: [email protected] [[email protected]] On Behalf Of FARIDA HUSNA [[email protected]]
> Sent: 17 January 2011 20:30
> To: [email protected]
> Subject: Re: st: Oaxaca decomposition for Tobit
>
> Marco,
>
> You are not supposed to report for each variable.  They are unstable
> and not worth looking.  The experts in the field of discrimination -
> Blau, Oaxaca, Cotten, Blinder, Neumark, Khan, Polachek, Reimer,
> Goldin, O Neill - do not report them.
>
> The output created by "oaxaca.ado" is useless.  It was written by Ben
> Jann.  He is making outlandish claims on how to do this but he is not
> expert in this field.  "nldecompose.ado" is by Sinning, Hahn, and
> Bauer.  Two of them are active researchers in the filed of gender
> discrimination.  You should go back to "nldecompose.ado"
>
> Cheers,
>
> FARIDA
>
> On Mon, Jan 17, 2011 at 10:59 AM, Marco Ercolani
> <[email protected]> wrote:
>>
>> Dear All,
>>
>> I wish to carry out detailed Blinder-Oaxaca variable-level decompositions for a Tobit regression. Can anyone suggest a useful "ado" script?
>>
>> I have been using the excellent "oaxaca.ado" by Ben Jann but this does not accommodate tobit estimation. "oaxaca.ado" is excellent because it includes detailed decompositions for each variable and it will automatically bootstrap the confidence intervals.
>>
>> I also know that the very good "nldecompose.ado" can cope with Tobit but it does not provide individual variable decompositions.
>>
>> A very old version of oaxaca.do did seem to work for Tobit because it used stored estimates to carry out the decompositions but this version no longer exists.
>>
>> I suspect the problem is that in Tobit we can define more than one marginal effect per variable and it is therefore not obvious which marginal effect to choose when carrying out the Blinder-Oaxaca decomposition.
>>
>> I'd be grateful for any suggestions.
>>
>> Yours,
>>  Marco.
>>
>>
>> Dr Marco G. Ercolani
>> Department of Economics
>> University of Birmingham
>> Edgbaston, B15 2TT
>>
>> *
>> *   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/
> *
> *   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/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index