Betting Against Correlation

Betting against beta buys low-beta stocks and shorts high-beta stocks, exploiting the empirical fact that low-beta names earn higher risk-adjusted returns than the capital asset pricing model predicts. Beta itself is the product of a stock's volatility and its correlation with the market, $\beta_i = \rho_{i,m} \frac{\sigma_i}{\sigma_m}$, so a "betting against correlation" strategy isolates just the correlation piece: it goes long stocks with low correlation to the market (holding volatility roughly fixed) and shorts stocks with high market correlation, rather than sorting on beta or volatility directly.

The rationale mirrors betting-against-beta: investors who are leverage-constrained bid up high-beta, high-correlation assets to get more market exposure without borrowing, since they cannot use leverage to scale up a safer low-correlation portfolio to the same expected return. That demand compresses the expected return on high-correlation stocks relative to what their risk would otherwise justify, leaving a return premium on the low-correlation side.

Decomposing beta into its volatility and correlation components lets researchers check whether a beta-based anomaly is really being driven by the volatility piece (the low-volatility anomaly) or the correlation piece specifically — the two often overlap but are not identical, and separating them clarifies which underlying risk is actually being compensated. For instance, two stocks can share the same volatility but very different market correlation: a stock with idiosyncratic, stock-specific swings has low correlation and would sit on the long side of this strategy, while an equally volatile stock whose swings track the index closely would sit on the short side, even though a plain low-volatility sort would treat them identically.