Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: one way/two way fixed effects


From   "Austin Nichols" <austinnichols@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: one way/two way fixed effects
Date   Wed, 5 Dec 2007 10:03:39 -0500

You should read the manual entry [R] xtreg first, and note that your
specification using 15 regions with fixed effects is unlikely to
identify the impact of immigration on per capita income. Immigrants
are not randomly assigned to region; at least in part, they migrate to
regions that they anticipate will have high income growth.  This means
that using changes over time within region is not a good
identification strategy.  If you have policies that create
discontinuities in immigration allowed by region over time, you might
be able to use an RD approach (-ssc install rd- and -help rd-) or if
you can measure existing concentrations of immigrants by origin and
total emigration from those origins, and you hypothesize that
immigrants are attracted to regions with like populations, you could
formulate an IV approach (-ssc install ivreg2- and read the help),
though you would need more than 15 regions.  As it is, the strategy is
a bit like interviewing 15 people and asking whether they would earn
more if they moved, and whether their neighbor would have higher
earnings wherever they moved--it's a small sample size with all kinds
of selection problems.

As for interpreting year effects omitting year1, the coefs measure the
change in mean income relative to the base year.  To analyze
residuals, you can graph the residuals versus predicted values; if you
see a pattern of any kind, you may want to rethink your specification.

On Dec 4, 2007 5:46 PM, valentina p <valendena@hotmail.it> wrote:
>
> Hi to everyone,I have data of 15 regions over 15 years and I am trying to understand which model to use to estimate the impact of immigration on per capita income. Now I have to choose between a one-way and a two-way fixed effects model.
>
> 1)To control for regions and time I suppose that a two-way fixed effects model can be useful, but I do not know the commands for stata.
> 2) Apart from this, if used a one way fixed effect model like xtreg y x1 x2 ... ,fe would it be necessary to make an analysis of the residuals? If yes, through which commands?
*
*   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/



© Copyright 1996–2021 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index