I see that the Fama-French library offers annual factors for their models, but everyone seems to exclusively use monthly returns of stocks in their regressions involving Fama-French factors. I am very fresh to this so excuse me if I'm asking a severely dumb question, but why doesn't anyone use annual returns over say 20 years. Is it simply because the regression probably wouldn't have much data to go on?
Eric, please, check again their website. It is possible to find monthly or daily returns for some model versions in this website.
In general, there are some advantages of using monthly returns:
a. You don't need very long-term data series as we have much more data points for running time-series regressions;
b. It is possible to cover new assets with shorter returns history;
c. Stock's Sensitivities (betas or exposures) may be updated every month, then adjusting faster for company characteristics changes;
d. Any other model derived data (like estimated stock's alpha) will have new a data point every month, better adjusting for current market conditions.