Can someone explain to me what adjusted betas are in the autoregressive context? I know how to calculate them by adjusting for meanreversion but not the autoregressive bit. Anyone??
*bump*
huh?? I only know adjusted beta have a formula to adjust them… somethign like 1/3 + 2/3 (real beta)… what are we talking abotu here? i am so confused.
So through empirical data, it’s been shown that beta approaches 1. Intuitively, as companies become larger they approach the sensitivity of the market - they become as volatile as the market in returns, sort of like how long-term growth rate g in DDM, should be in line with GDP. beta’s are apparently stationary over the long-term, so there is mean reversion to 1, so while historical betas use past regressions to estimate beta, forward looking beta includes the (1/3)*1 to account for future mean reversion of beta to 1. Remember that betas in a portfolio are the weighted average of each beta. I think that’s where the wa = 2/3 and 1-wa = 1/3 comes from.
I think that was pepp was refering to, but I agree - what does he mean by adjusting for the “autoregressive bit”? The 1/3*1 + 2/3(beta) equation is the only way I remember seeing to calculate it.
Its autoregressive, i.e. AR, because Beta(t)= .33+.6667Beta(t-1)
adavydov7 Wrote: ------------------------------------------------------- > Its autoregressive, i.e. AR, because > > Beta(t)= .33+.6667Beta(t-1) Oh boy … this means that we would have to test for ARCH and determine if the time series is covariance stationary … here we go quant.
well you know that if there is mean reversion, then it’s covariance stationary, since b1 is not equal to 1
Lol, good catch Ali!!!
If it is mean reverting, then it’s covariance stationary and hence there is no random walk and/or walk with drift. Equation clearly shows b1 != 0 and hence is well specified.