Downside Beta

Ordinary beta measures how much a stock tends to move for every 1% move in the market, averaged over all days — up days and down days alike. But investors mostly care about a more specific fear: how much will this stock hurt me when the market is actually falling? A stock could have a modest overall beta of 1.0 while behaving very differently depending on the market's direction — surging on up days but crashing even harder on down days, a pattern ordinary beta blends away into a single average number.

Downside beta answers this directly by restricting the same regression to only the days the market's excess return is negative:

$$\beta^- = \frac{\mathrm{Cov}(R_i, R_m \mid R_m < 0)}{\mathrm{Var}(R_m \mid R_m < 0)}.$$

In plain English: run the ordinary market-beta calculation, but throw out every day the market was up first, so the number is a pure read on down-day sensitivity.

Suppose a stock's overall beta is 1.0, but on the roughly half of days the market fell, the stock's downside beta comes out to 1.4 — it drops about 40% more than the market on the days that matter most for a drawdown, even though its average sensitivity across all days looks ordinary. That asymmetry is exactly what a portfolio manager building downside protection needs to know and what a single overall beta hides.