Definition
Beta hedging is the practice of removing a portfolio's exposure to broad market movements by taking an offsetting short position whose size is calibrated to match the portfolio's beta — its sensitivity, in a CAPM regression sense, to a market benchmark. After the hedge, gains or losses driven by the market as a whole net to roughly zero, leaving only the stock-selection or factor-specific component of return.
The technique is a workhorse of long-short equity, statistical arbitrage, and many event-driven strategies. It assumes a stable enough relationship between portfolio and benchmark that the regression-estimated beta holds prospectively; in practice betas drift and the hedge must be rebalanced. The same logic extends to currency, rate, and sector hedges — any factor whose contribution you want to subtract from the residual you care about.
Why it matters
How it works
The standard recipe estimates portfolio beta by regressing portfolio returns against benchmark returns over a rolling window, then sells short an amount of the benchmark (or its futures or an equivalent ETF) equal to portfolio market value times beta. If the portfolio has a 1.2 beta to the S&P 500 and is worth ten million dollars, the hedge is a short S&P 500 futures position with a notional of twelve million. The book is now beta-neutral: a one percent rise in the market loses one percent on the hedge for every one and a fifth percent gained on the long book, and vice versa.
The complications are where the craft lives. Beta is not constant — it shifts with regime, with the portfolio's composition, and with the liquidity of the underlying names — so the hedge ratio must be recomputed and the futures position rebalanced periodically, with each rebalance incurring transaction cost. The choice of benchmark matters: a long book of US small-caps hedged with an S&P 500 short still carries unhedged small-cap factor risk. And the assumption that residual returns are uncorrelated with the market often fails in tail events, when correlations spike toward one and the hedge stops working precisely when it is most needed. Sophisticated programs hedge multiple factors at once (beta, sector, size, value) using a covariance-matrix-driven approach rather than single-factor regression.