How to correct raw stock returns using Fama-French?
Posted: Tue Apr 12, 2016 10:50 am
Hi all,
I've got a large dataset with monthly stock returns of 350 companies, which I want to correct for risk using Fama-French's method (or Carhart, for that matter). I know how to do regressions, but what would happen is that I run a regression per company, leaving me with the alpha and the betas in the regression report. I am interested in the alpha, but I don't want the overall intercept of my regression, but the out/underperformance PER MONTH.
So my data looks somewhat like this. I've got a big matrix with as said 350 companies, every column is company. Every row represents a month (360 in total), so every cell is the stock return for that month for that company. I also have the four Carhart factors per month (market return, small-minus-big, high-minus-low, and momentum). Depending on a stock's beta exposure to these factors, this forms the benchmark unique for every stock. All under/overperformance wrt this benchmark, per month, is the risk corrected return per stock.
But like I say, just running regressions will leave me with one alpha per stock, instead of the outperformance per month per stock. Any ideas how I could do this?
Lots of thanks! You'd save my life!
I've got a large dataset with monthly stock returns of 350 companies, which I want to correct for risk using Fama-French's method (or Carhart, for that matter). I know how to do regressions, but what would happen is that I run a regression per company, leaving me with the alpha and the betas in the regression report. I am interested in the alpha, but I don't want the overall intercept of my regression, but the out/underperformance PER MONTH.
So my data looks somewhat like this. I've got a big matrix with as said 350 companies, every column is company. Every row represents a month (360 in total), so every cell is the stock return for that month for that company. I also have the four Carhart factors per month (market return, small-minus-big, high-minus-low, and momentum). Depending on a stock's beta exposure to these factors, this forms the benchmark unique for every stock. All under/overperformance wrt this benchmark, per month, is the risk corrected return per stock.
But like I say, just running regressions will leave me with one alpha per stock, instead of the outperformance per month per stock. Any ideas how I could do this?
Lots of thanks! You'd save my life!