Re: st: Reverse Causation Solutions?

[Suryadipta Roy <sroy2138@gmail.com>]

There are a variety of techniques to address the statistical problem
that you have mentioned. Please read the Stata Journal article by
Austin Nichols available here:
http://www.stata-journal.com/article.html?article=st0136
 to have an idea about the umbrella of identification issues of which
reverse causation is one. For example, you may use a Two Stage
Least squares method for this purpose. The paper will tell you about
most of the important methods in practice.


Best wishes,
Suryadipta.

On Thu, Jun 20, 2013 at 12:18 PM, Mai Anh Ngo <maianh.ngo@gmail.com> wrote:
> I have a problem of reverse causation in my regression model for a
> panel data set with unobserved heterogeneity. My regression equation
> is as follows:
>
> Y1it = a + b*Y2it + c*Xit + ui + eit
>
> where Y1it and Y2it are binary and it is suspected that there is
> possibly reverse causation from Y2it to Y1it or from Y2i,t-1 to Y1it
>
> I could not find a good instrument to use the instrumental variable
> method and I am thinking of estimating a bivariate model:
>
> Y1it = a + b*Y2it + c*Xit + ui + eit
>
> Y2it= d +e*Y1it  +  f*Zit + vi +  rit
>
> where the error term eit and rit are allowed to be correlated.
> However, I was told that in general, a bivariate model would not solve
> reverse causation/simultaneity problem. Is that true?
>
> What are the methods that can be used to solve reverse causation or
> simultaneity problem?  Thank you for your consideration.
>
> Mai Ngo
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/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/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/