Dear Statalisters,
I am writing to ask for a suggestion on estimating CAPM/market models.
I have monthly returns (60 months) for a sample of almost 10k mutual funds. I would like to estimate the average alpha for the entire sample and for some subsets (the aim is to partially replicate European Green Mutual Fund Performance: A Comparative Analysis with their Conventional and Black Peers | SpringerLink). I am wondering which is the most appropriate empirical strategy. The paper states that they used a "random effects panel least squares seemingly unrelated regressions (SUR)".
After searching the forum, I see the following possibilities:
1) Use xtsur, which seems to be the closest command to replicate what the paper did.
2) Keeping the time series dataset (each fund returns with its own variable) and running sur.
3) Building a panel dataset and running a panel random effect model
However, I have the following doubts on each of those choices:
1) Xtsur does not seem to work, and by looking at the forum past discussion it looks like there were some bugs:
mreldif(): 3200 conformability error
_xtsurub(): - function returned error
<istmt>: - function returned error
2) Sur seems to work, but it I am not sure about how I can run all equations at once and calculate the significance of the average alpha since the command would just provide me results for each equation
3) This model seems to work so far, providing me with realistic estimates - however I wonder if it is the most appropriate choice.
I hope that I have been clear enough, and thank you for any advice you might provide.
I am writing to ask for a suggestion on estimating CAPM/market models.
I have monthly returns (60 months) for a sample of almost 10k mutual funds. I would like to estimate the average alpha for the entire sample and for some subsets (the aim is to partially replicate European Green Mutual Fund Performance: A Comparative Analysis with their Conventional and Black Peers | SpringerLink). I am wondering which is the most appropriate empirical strategy. The paper states that they used a "random effects panel least squares seemingly unrelated regressions (SUR)".
After searching the forum, I see the following possibilities:
1) Use xtsur, which seems to be the closest command to replicate what the paper did.
2) Keeping the time series dataset (each fund returns with its own variable) and running sur.
3) Building a panel dataset and running a panel random effect model
However, I have the following doubts on each of those choices:
1) Xtsur does not seem to work, and by looking at the forum past discussion it looks like there were some bugs:
mreldif(): 3200 conformability error
_xtsurub(): - function returned error
<istmt>: - function returned error
2) Sur seems to work, but it I am not sure about how I can run all equations at once and calculate the significance of the average alpha since the command would just provide me results for each equation
3) This model seems to work so far, providing me with realistic estimates - however I wonder if it is the most appropriate choice.
I hope that I have been clear enough, and thank you for any advice you might provide.
