Advanced econometrics: panel data, double clustering and overlapping returns

[NHHnorway]

Hi,
I am writing a thesis about the relationship between long term returns (5 year) and P/E-, P/B-ratios etc. Because of short time series in the specific market, I have been forced to use overlapping returns, measured each month 5 years ahead. This will create serious issues regarding autocorrelation, which will lead to standard errors that are biased downwards. The majority of research on such panel data uses the Fam-Macbeth procedure to adjust for time effects. This technique only adjusts for time effects, not firm effects, thus still creates biased standard errors. (see for instans Petersen (2009), http://www.zaferyuksel.com/uploads/1/2/ ... n_2008.pdf). Thompson (2009, http://schwert.ssb.rochester.edu/f532/JFE11_ST.pdf) suggests using double clustering to adjust for time- and firm-effects. He also mention overlapping returns, which has not been covered in great detail in the literature.

Do someone know if this technique will create unbiased standard errors when using overlapping returns?

I really appreciate the help I could get on this topic.