I am planning to analyse the impact of two war periods on the development/growth of the country. Three variables (GDP, Capital formation and CPI) are included. Logic behind is simple. fiscal policy results in capital formation and monitory policy is depicted in CPI. Now included two dummy variables indicting each war time. Results are fine at 6 lags. Am I going the right way? what tests needed? Any suggestion to defend lags and theoretical framework? Please ask if you need more explanation.
This video might help:
https://www.youtube.com/watch?v=3CveW_7IDcc
In addition, although it could become manageable you might consider adding an interaction term and your dummy variable for war.
War
W = 1 if war
0 otherwise
You might add an independent variable Time (T)
T= 0 for the first year of data
1 is added for each year
(e.g., for the Gulf War that began in 1990, let's say your data begins in 1981, then T=0 in 1981)
Interaction Term = W*T
By including both W and W*T, you may be able to test the statistical significance of a change in the intercept and the slope by evaluating the t-statistics on W and W*T. If both of the calculated t-statistics exceed their critical value, then you should have a test that I consider to be a superior test to the (Gregory) Chow Test, which is a way to evaluate significant differences in regression equations.
The attachment is crudely drawn; however, it may give you an idea.
I hope this helps.
Rob
Including dummy variable into VAR framework is a good option, but only in case if you focus on the impact of other variables. The problem is that dummy variable doesnt change during the war period. Impact of two war periods on countries is better to model via Difference-in-difference estimation.
Thank you Rob and Svatopluk. Let me elaborate the scenario. the country fought first war from 1979 to 1988 and second war from 2001 to 2016 as a NATO allay. apparently, earlier war resulted in gain as compared to the latter one. My objective is to capture the impact of the wars on the growth. I am using two dummies for war=1 otherwise 0 for each war. the variables are GDP, Gross capital formation (GCF), and CPI. GCF is included because all fiscal measures are taken to increase GCF. Therefore, it is a good indicator of fiscal policy. Similarly, CPI is good indicator of monitory policy. Should I add any other variable?
I tested the lag length and it was 6. All tests are OK except stability test at first difference . When I reduce the lag length to 5, stability test is significant.
Further, my plan is to forecast the growth in absence of second war period (assuming if there was no war). Can I go for it?
Imagine I would like to correct 'structural breaks' within a time series ( Oil price ig).
Would it be licit to do so with a dummy variable?
What about two dummies within the same equation?
Thanks a lot.
Edu.
- Badges
- Science topic