Betting against beta: An empirical study

Sonja Sørensen

Student thesis: Master thesis


I have analyzed the Betting against beta theory, presented by Frazzini and Pedersen in 2014, on 8 different countries and their national stock index as well as US treasury bonds. I found that in five of the countries, namely Germany, France, Spain, Switzerland and Belgium, I was able to accept the proposition that the alphas of the portfolios are declining, when the portfolios moves from being a low-­‐beta portfolio, to being a high-­‐beta portfolio. Further more I found that the Sharpe ratio is also declining with the beta for these five countries. I was not able to prove the alpha-­‐beta relation for the American S&P 100 index and British FTSE 100 index. My analysis shows that both the US and the UK stock indices, displays an increase in alpha when moving from the low-­‐beta portfolio to the high beta portfolio. The second proposition treats the BAB portfolio in two parts. The first says that the excess return of the BAB portfolio should be positive, and the second part states that the BAB portfolio return should increase in the beta spread and in case of funding tightness. Both parts of this proposition are met for all 8 counties, because all the BAB portfolios deliver positive abnormal returns, also when adding factor returns based on size (SMB), book-­‐to-­‐market (HML) and momentum (UMD). I was also able to accept the second part of the proposition, because all countries shows an upwards-­‐sloping trend line in plot of the BAB factor excess return and the beta spread. I am not able to accept any of the propositions, when they are applied to US treasury bonds. Lastly I use time series regression to test, how a tighter portfolio constraint affects the BAB factor. I only find the results for the lagged level of the TED spread statistically significant in all countries, and therefore I cannot draw a definitive conclusion of the third proposition for all countries as a whole, but the proposition is fully accepted for Germany, Sweden, Switzerland and US treasury bonds. Although the remaining countries are not statistically significant at the 5% level, they all show the tendencies predicted by the model with regards to the operational sign. The estimated coefficient for the lagged level of the TED spread should be positive and the estimate for the change in TED spread ought to be negative. The model further predicts that the coefficient for the beta spread should be positive. This prediction is fulfilled in all the counties, and the coefficient is statistically significant for all except Sweden, Switzerland and the US treasury bonds.

EducationsMSc in Applied Economics and Finance, (Graduate Programme) Final Thesis
Publication date2016
Number of pages75