Re: st: Juhn-Murphy-Pierce (1993)

[Vora Nakavachara <nakavach@usc.edu>]

Thank you Ben.

But should one really use the survey weights 
with the estimation? I have not seen any literature
that mentioned anything regarding the weights in 
their Juhn-Murphy-Pierce (1993) estimation.

Also, what about DFL? 
DiNardo-Fortin-Lemieux (1996)
My gutts feeling says that I should be using
the survey weights (include into the estimation 
when they do the reweighing method). 
However, I also don't see any literature 
says that they use weights in their
DiNardo-Fortin-Lemieux (1996) estimation.

I'm quite confuse. Can you or anyone help me with this?

Sincerely,
Vora N


----- Original Message ----- 
From: "Ben Jann" <ben.jann@gmail.com>
To: <statalist@hsphsun2.harvard.edu>
Sent: Tuesday, November 28, 2006 12:45 AM
Subject: Re: st: Juhn-Murphy-Pierce (1993)



> It is absolutely okay to use weights with -jmpierce-. Simply apply the
> weights while estimating the models that serve as input to -jmpierce-.
> -jmpierce- will then pick up the weights and appropriately use them in
> its calculations.
> ben
>
> On 11/27/06, Vora Nakavachara <nakavach@usc.edu> wrote:
> > Hi All,
> >
> > I'm trying to follow Juhn-Murphy-Pierce (1993)
> > wage decomposition and have some questions.
> >
> > If the data I'm using is a survey data (labor force data)
> > and it has a weight (sampling weight) attached to each
> > observation, then when I do jmpierce do I need to care
> > about the weight?
> >
> > I have read some literature but I don't seem to see anyone
> > mentioning anything about this weight.
> > (Maybe I'm missing something -- can you please point out
> > the literature that discuss it?)
> >
> > Basically, if I want to calcuate p90-p10 of log wage of people
> > in 2000 (and 1995) -- and to describe it as a measure of inequality
> > of the country then I need to use weight, right?
> >
> > So when I decompose how p90-p10 change from 1995 to 2000,
> > I should also consider weight?
> >
> > But, (from the help file of jmpierce)
> >
> >  y1_i1 = x_i1b + F[-1](p_i1|x_i1) --- for year 2000
> >  y1_i2 = x_i2b + F[-1](p_i2|x_i2) --- for year 1995
> >
> > are "hypothetical" distributions, thus if I am to calculate p90-p10
> > (to be used in future steps of the decomposition) from these and
> > want to use weight then the weight attached to the observation i
> > wouldn't be correct anyway?
> >
> > So should I just ignore the weight in the decomposition?
> > But what about when I want to discuss the summary statistics,
> > (as mentioned before) like p90-p10 inequality measure in 2000
> > then I believe the weight is important?
> >
> > Could anyone please help me clarify this?
> >
> > Thank you very much.
> >
> > Vora N
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/