Greetings,
so let me explain what I mean in brief. I have n=2 ( two countries), t=30. the dependent variable is GDP growth and the explanatory are 4 continuous ones. the dummy variable ( the fifth) is inflation targeting, for the first country it takes 1 in the first ten years, and for the second country takes 1 in the last 8 years.
I want to compare the evaluation of the effectiveness of an inflation targeting strategy.
so, do a separate regression for each country alone maybe is inappropriate. Instead, I thought about panel data technique. but how to deal with that dummy one? should I insert 4 ones? does it make no sense to do so?
any suggestion.
A possibility could be to estimate:
Yit = b0 + b1*country_i + b2*time_t + b3*inflation_targeting_it + eit
country_i is a dummy for country (one for each country minus the reference category, tha is, one dummy)
time_t are time dummies (one for each period of time minus de the reference category, that is, 29 dummies)
inflation_targeting_it is a dummy that takes the value 1 when a country follows a monetary policy based on inflation targentin and 0 otherwise.
The remaining problem is to model serial correlation...
When you have n as 2 and t=30 it is not like a standard panel (where n greater than t). Here only advantage of using data of two countries would be the increase in the number of observations and the sample size. You may estimate the association between dummy and GDP growth after controlling for country dummy and time dummies. The country dummy will absorb the country specific institutional factors.
Use -xtreg, re- or -xthybrid-, you can get it from SSC.
If it is less linear relationship between your outcome , then enter as continuous variable,if not enter as a discrete variable (less efficient in monotone linear relationships).
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression