Greetings.
I'm new to this forum, so I hope you could please kindly tell me if I break any rules somewhere.
Anyway, onward to the question.
For my assignment, I'm testing the impact of several macroeconomic aspects (GDP, foreign reserves, debt to GDP ratio, and exchange rate) to Credit Default Swap (CDS) spread.
The data is an unbalanced data panel with 4 countries (Indonesia, Malaysia, Philippine, and Thailand).
Total N is 59.
For each macroeconomic aspects, there are two independent variables used (one for the nominal itself, while the other one is the changes. For example, for GDP aspect, I make a variable for the nominal of GDP (log-transformed) and GDP growth). And thus, there are a total of 8 independent variables.
The models are homoscedastic (tested using White test) but are autocorrelated (tested using xtserial command).
And thus, here are the questions:
1) Is using Prais-Winsten in this case appropriate? I was also contemplating about using first difference / xtregar / xtreg with clustered standard error, but truthfully (and I'm kinda ashamed to admit this), but I don't really get the difference between all these methods.
2) Is Prais-Winsten a Random Effects / Fixed Effects / Pooled Least Square model? I tested using Chow & Hausman and found that PLS fits my needs the best.
Thank you for reading through this.
Best Regards.
I'm new to this forum, so I hope you could please kindly tell me if I break any rules somewhere.
Anyway, onward to the question.
For my assignment, I'm testing the impact of several macroeconomic aspects (GDP, foreign reserves, debt to GDP ratio, and exchange rate) to Credit Default Swap (CDS) spread.
The data is an unbalanced data panel with 4 countries (Indonesia, Malaysia, Philippine, and Thailand).
Total N is 59.
For each macroeconomic aspects, there are two independent variables used (one for the nominal itself, while the other one is the changes. For example, for GDP aspect, I make a variable for the nominal of GDP (log-transformed) and GDP growth). And thus, there are a total of 8 independent variables.
The models are homoscedastic (tested using White test) but are autocorrelated (tested using xtserial command).
And thus, here are the questions:
1) Is using Prais-Winsten in this case appropriate? I was also contemplating about using first difference / xtregar / xtreg with clustered standard error, but truthfully (and I'm kinda ashamed to admit this), but I don't really get the difference between all these methods.
2) Is Prais-Winsten a Random Effects / Fixed Effects / Pooled Least Square model? I tested using Chow & Hausman and found that PLS fits my needs the best.
Thank you for reading through this.
Best Regards.
Comment