Hello everyone,
I'm currently facing a common issue in panel data regression. I'm doing empirical research on the effect of fiscal and monetary policies on output stabilisation in five different countries between 2000 and 2020. I have output gap as a dependent variable and several explanatory variables (government spending, tax revenues, consumption, interest rate, inflation). These are the steps I followed to estimate my model:
1) I log-transformed the variables government spending, tax revenues, consumption
2) I performed the Levin-Lin-Chu test to see if variable are stationary - variables are stationary at level.
3) As I am dealing with a dataset that has T>N, -xtregar- is more appropriate
4) I performed Hausman test to choose between FE and RE model - FE was the most appropriate.
5) I tested for heteroskedasticity and autocorrelation using -xttest3- and -xtserial- and found that my model suffers from both.
Now I can't resolve the issue of heteroskedasticity and autocorrelation—I tried to generate the first difference of each variable but it doesn't resolve the issue.
I would be extremely grateful for any kind help.
0 Response to Autocorrelation and Heteroskedasticity in panel data using fixed effect model
Post a Comment