Panel regression with country fixed effect and panel-robust standard errors
Dear Statalist Members:
I have some questions about the panel regression with country fixed effect and panel-robust standard errors. The data sample is attached.
First, description about data:
-time: daily time series (As you can see, I have a sample with small N, but large T)
-countryid: two countries. France(countryid==1); Italy(countryid==2)
-individual: local stock market returns (France and Italy)---this is a transformed data, so it is independent of market variable.
-market: European stock market returns (Same for both countries)
-bond: log-returns of 10-year bond yields (France and Italy)
What I want to test:
- whether individual(local stock returns) is an important factor for explaining bond(local bond yield returns)? Since I have commonly shared market returns, I should leave out time-fixed effect and control for country-fixed effect.
I want to perfrom:
(1) Panel regression with robust standard error and country fixed effect. The panel-robust standard errors(clustering at country level) to control for heteroskedasticity and autocorrelation.
-try1:
xtreg bond individual market, fe vce(cluster countryid)
-try2:
xtscc bond individual market, fe lag(4)
Question:
Why the results of individual are so different(insignificant in "try1", but, significant in "try2")? Am I making any mistakes, if so, what should be the right code to use(to have the panel-robust standard errors(clustering at country level) to control for heteroskedasticity and autocorrelation)? (My confusion is : If you run each country separately by using: regress bond individual market if countryid==1, robust, and, regress bond individual market if countryid==2, robust. Results show that individual is significant in each regressions. So, why individual becomes insignificant in try1?)
France:
Italy:
-My second (small) question is that: if I want to do:
Panel regression with robust standard error and country fixed effect. The standard errors are corrected for both heteroskedasticity and serial correlation.
Is it correct to do
- regress bond individual market i.countryid, robust cluster(time)???
Any suggestions are greatly appreciated!!!
Many Thanks,
Tonya
I have some questions about the panel regression with country fixed effect and panel-robust standard errors. The data sample is attached.
First, description about data:
-time: daily time series (As you can see, I have a sample with small N, but large T)
-countryid: two countries. France(countryid==1); Italy(countryid==2)
-individual: local stock market returns (France and Italy)---this is a transformed data, so it is independent of market variable.
-market: European stock market returns (Same for both countries)
-bond: log-returns of 10-year bond yields (France and Italy)
What I want to test:
- whether individual(local stock returns) is an important factor for explaining bond(local bond yield returns)? Since I have commonly shared market returns, I should leave out time-fixed effect and control for country-fixed effect.
I want to perfrom:
(1) Panel regression with robust standard error and country fixed effect. The panel-robust standard errors(clustering at country level) to control for heteroskedasticity and autocorrelation.
-try1:
xtreg bond individual market, fe vce(cluster countryid)
-try2:
xtscc bond individual market, fe lag(4)
Question:
Why the results of individual are so different(insignificant in "try1", but, significant in "try2")? Am I making any mistakes, if so, what should be the right code to use(to have the panel-robust standard errors(clustering at country level) to control for heteroskedasticity and autocorrelation)? (My confusion is : If you run each country separately by using: regress bond individual market if countryid==1, robust, and, regress bond individual market if countryid==2, robust. Results show that individual is significant in each regressions. So, why individual becomes insignificant in try1?)
France:
Italy:
-My second (small) question is that: if I want to do:
Panel regression with robust standard error and country fixed effect. The standard errors are corrected for both heteroskedasticity and serial correlation.
Is it correct to do
- regress bond individual market i.countryid, robust cluster(time)???
Any suggestions are greatly appreciated!!!
Many Thanks,
Tonya
