It is a very common procedure when working with panel data to take into account trends that are common to all the panels, by introducing as a predictor a linear and a quadratic term. Some econometricians go further and introduce linear and quadratic terms for each panel. These approaches make sense if the trends are likely to be smooth. An alternative is to introduce dummy variables for each year but the reference year. This controls for common changes regardless of whether they have a linear or quadratic functional dependence. I don't think it is relevant that your data set is not balanced. Most of the time, in adopting these strategies, one doesn't think of time as causing change. It is simply a way of taking change due to unmeasured causes into account. David Greenberg, Sociology Department, New York University
----- Original Message -----
From: Steven Archambault <[email protected]>
Date: Thursday, July 16, 2009 0:58 am
Subject: st: Panel data unbalanced--time as indep variable?
To: [email protected]
> Hi all,
>
> Would it ever make sense to have time as an independent variable on an
> unbalanced panel regression (fixed or random effects)?
> I could see doing it for balanced time series data, but what about in
> my case with 8 years and about 90 panels? I'd appreciate some
> feedback, or some resources to read.
>
> Thanks,
> Steve
> *
> * 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/
>
*
* 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/
