Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: constant variable as IV in panel?


From   Austin Nichols <austinnichols@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: constant variable as IV in panel?
Date   Wed, 10 Aug 2011 11:06:55 -0400

Abhimanyu Arora <abhimanyu.arora1987@gmail.com>:
You report having variation across years within panel, so (a) should work:
"For all months in a year, take the same (annual) value of the yearly variables"
but include year and month dummies and cluster by both panel and year
using -ivreg2- on SSC.
See refs in that program's help file.

On Wed, Aug 10, 2011 at 6:41 AM, Abhimanyu Arora
<abhimanyu.arora1987@gmail.com> wrote:
> Dear statalist
> I think this is related to the posts by Matthias Opfinger and Austin
> Nichols with an addundum and would like request your intellectually
> stimulating views/references. The panel model I am estimating has a
> monthly frequency (no IV). But I also have yearly variables (in my
> case I have almost 20 years) for each cross section, which I would
> definitely like to include. The options that I am considering are the
> following
>
> (a) Related to Matthias' idea-For all months in a year, take the same
> (annual) value of the yearly variables
> (b) Year dummies-subsumes all (extraneous) yearly variables. But I am
> taking a trend.
> (c) Take the yearly value for the 6th month and impute the monthly
> variables from the series...a risky one
>
> Would be great to have your views and concerns or better alternatives.
> (Could hierarchical models come to the rescue?)
>
> With best regards
> Abhimanyu Arora
>
>
>
>
>
> On Thu, Jun 23, 2011 at 2:29 PM, Austin Nichols <austinnichols@gmail.com> wrote:
>> Matthias Opfinger <opfinger@vwl.uni-hannover.de>:
>> In such a case, you are using only between variation, and it is more
>> intellectually honest to run a cross-sectional regression for one
>> point in time, or possibly to compute a long difference using a pair
>> of years (estimating a cross-sectional regression in changes rather
>> than in levels), depending on which assumptions are more plausible
>> (plausibility comes from theory not from the data).
>>
>> On Thu, Jun 23, 2011 at 5:22 AM, Matthias Opfinger
>> <opfinger@vwl.uni-hannover.de> wrote:
>>> Hi everyone,
>>>
>>> I want to estimate a panel model with observations on approx. 90 countries
>>> for five points in time between 1980 and 2005. However, one of my
>>> explanatory variables is endogeneous. The only possible instrument is a
>>> variable on climatic zones. So it would be the same for each country over
>>> all points in time. How can I put this into my model? Do I just put the same
>>> value for each country over all the points in time and then use xtivreg?
>>>
>>> Best
>>> Matthias
*
*   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 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index