How to Segregate Time Series Effects of Explanatory Variables with Panel Data

[Tyler Box]

Hi everyone,

I am running a fixed-effects regression on a panel data set spanning the years 2003-2013. I would like to know how the impact of the explanatory variables on my DV changes between time period 1 (2003-2008) and time period 2 (2009-2013). So far I have generated two time series dummies (time1=1 if the observation is from 2003-2008 and 0 otherwise while time2=1 if the observation is from 2009-2013 and 0 otherwise) and then interacted each with every explanatory variable. Then I put all of those interacted variables in the model. This gives me:

DV= β0 +β1(time1*variableA) + β2(time2*variableA) + β3(time1*variableB) + β4(time2*variableB)... + βx2003… βy2013 + µ

Is this a statistically valid way to specify my regression model?

[Carlo Lazzaro]

Tyler:
you shoud probably be better off with one categorical variable only (see -help fvvarlist-):

g time_var=0 if year<=2008
replace time_var=1 if year>2008
label define time_var 0 "2003-2008" 1 "2009-2013"
label val time_var time_var

Then you can interact -time_var- with other predictors (again, see -help fvvarlist- for creating interactions).
In general, I would shy away from creating too many interactions, as their results are often difficult to disseminate.
I would rather choose one or two variables to interact with -time_var-.
As an aside, for the future please post what you typed and what Stata gave you back via CODE delimiters (see FAQ on this tiopic). Thanks.