Fig. 1

Local projection of a GDP growth shock on hedge fund strategies’ betas over the subprime crisis. Notes: The impulse response functions corresponding to the linear LP—i.e., the beta forecast following external shocks—are built using the Jordà’s (2004, 2005, 2009) algorithm provided by Eren Ocakverdi, consultant at Yapy Kredi. The local projection of a beta strategy is performed over the subprime crisis (2007Q3–2009Q4). We also provide the linear VAR estimated over the whole sample period selected as benchmark. The confidence interval of the LP impulses, computed at the threshold of 95%, is in dotted lines. As explained in the paper, our model comprises five endogenous shocks, placed in the following order to perform the Cholesky factorization (Sims 1980) required to build structural shocks: GDP growth, term spread, credit spread, beta (own shock) and VIX. As usual, these variables are ordered by increasing level of endogeneity. This order relies on the conventional information criteria: Akaike and Schwartz criteria. Only strategies whose impulse response functions are significant at a level of 10% or less are reported in this table. Note that the IRFs are computed for positive GDP shocks but the corresponding IRFs associated with negative shocks are the mirror images of the former since, for both methods, the IRFs being computed with linear techniques. The fixed income and futures strategies are excluded since their IRFs are not enough significant