Blume and Vasicek Beta Adjustments

Beta, estimated by regressing a stock's returns against the market's, is a noisy statistic — it varies a lot from one sample period to the next, and empirically, extreme betas (very high or very low) tend to move back toward 1.0 in later periods as a business matures or its historical extremity turns out to have been partly noise. Two adjustments — Blume's and Vasicek's — both correct for this by shrinking the raw estimate toward 1.0, but they do it differently.

Blume's adjustment uses a fixed formula regardless of the stock: $\beta_{adj} = \frac{2}{3}\beta_{raw} + \frac{1}{3}(1.0)$, applying the same two-thirds/one-third blend to every company.

Vasicek's adjustment instead weights the blend by how statistically reliable the raw estimate was: a beta estimated with a small standard error (a tight, confident regression) gets shrunk only a little, while a beta estimated with a large standard error gets pulled much closer to the market average or to a peer-group average.

Worked example. A stock's raw beta is 1.8. Blume's adjustment gives $\frac{2}{3}(1.8) + \frac{1}{3}(1.0) = 1.53$, regardless of how precisely that 1.8 was estimated. If Vasicek's method judges that this particular beta was estimated with unusually low precision (a wide standard error), it might shrink further, to something like 1.35 — a bigger pull toward 1.0 than Blume's uniform rule would apply.

Practitioners use these adjustments because forecasting cost of capital with a raw, unadjusted beta tends to overstate how extreme a stock's systematic risk will remain going forward.