Estimate Jensen`s alpha for a mutual fund in a specific year

[RBechsgaard]

Hello guys

I have the following issue. Normally when evaluting mutual fund performance you either use a 1 factor (CAPM) or a 4 factor model (Fama French + mom) to see whether the fund are able to generate positive alpha. If you want to examine the performance of a particular fund in a time period between 01-01-2000 to 01-01-2012 you simple use a time serie regression and regress excess return upon market excess return if are considering a 1 factor model (CAPM).

If the obersvations in this period on montly basis, then the intercept is the average monthly alpha over the estimation period. This is relative easy and quite understandable.

However my issue is, how do I estimate Jensen`s alpha for only the lartest year i.e. from 01-01-2011 to 01-01-2012? Simply I am not interest in the alpha of the whole period but only the alpha for the latest period.

Can anybody help? Or provide my with some good litteratur, where they explain how to do it? I have googlet for like 3 days now.

Hope you can help

//René

[Roberto Liebscher]

Why not use the if condition in your regression command?

regress excret mom smb hml markret if date > date("01.01.2011", "DMY") & date < date("01.01.2012", "DMY")

Then the intercept gives you the monthly alpha for this fund in the given period. Allowing for autocorrelation in the residuals using the newey command might also be a good idea.

[RBechsgaard]

Hi Roberto.
Thank you for your reply. It is definitely a way to do it.
Another question comes to mind; if I want alpha (i.e. the intercept) for a given year, I can do what you recommend. However, I only have monthly data, so that would only give me 12 data points in the regression, which is hardly enough. If I increase the number of data points I’m going to get return data from more than that year year, and thus I do not end up with the alpha for that given year.
Do any of you have ideas on how to such an estimation? I know it is properly more of econometric interest than "Stata-technical".
// Rene

[krishnakumar]

Would a Bayesian regression be appropriate in this context? A prior with \beta = 1 and \alpha = 0 (which indicates that the mutual fund is nothing more than an index fund) with some appropriately chosen prior variance might yield interesting results.