st: Re: Panel data and lagged dep variable

[Christopher Baum <kit.baum@bc.edu>]

<>
On Sep 25, 2010, at 2:33 AM, Fabio wrote:

> in a panel test, to what extent is it acceptable a model with no lagged dep variable among the indep variables? Is it always necessary to include a lag dep variable? Moreover: what do you think about a model that has all variables (both y and x) at the same year?
> Please: is there a reference that clarify these issues?

In terms of an economic model, if you write down y_{i,t} = f (X) where the Xs may contain current and past values, but not including lagged y, you are arguing that the data observed at each point in time represent equilibria. If you believe that it takes more than one of your time periods for economic agents to adjust to changes in the Xs (or to adjust to shocks), then the LDV is warranted. As you can always estimate the model (although not with fixed or random effects!) including the LDV, you can test whether it is warranted in your particular setup.  If you do that, it would probably be a good idea to include time dummies as well, as the absence of those factors could confound the analysis of the LDV's significance.

Kit



Kit Baum   |   Boston College Economics & DIW Berlin   |   http://ideas.repec.org/e/pba1.html
                              An Introduction to Stata Programming  |   http://www.stata-press.com/books/isp.html
   An Introduction to Modern Econometrics Using Stata  |   http://www.stata-press.com/books/imeus.html


*
*   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/