Beta Hedging and Residual Alpha

Beta hedging removes deliberate exposure to a benchmark or factor so that portfolio P&L more closely reflects the signal you intend to own. It is essential for interpreting long-short strategies. A book that appears to have alpha during a rising market may simply have positive market beta; a “market-neutral” book can still carry sector, value, momentum, or volatility exposure.

Residual alpha is the return remaining after a specified factor model. It is model-relative, not a metaphysical property of a trade. Change the benchmark or estimate beta poorly and the residual changes too.

Start with the return decomposition

The single-factor model is:

r_p,t = alpha + beta × r_m,t + epsilon_t

With multiple factors:

r_p,t = alpha + B' × f_t + epsilon_t

The CAPM beta is a useful first diagnostic, but equity portfolios usually need sector, style, and country factors. Use an appropriate benchmark return in the same currency and calendar as the portfolio.

import numpy as np
import pandas as pd

def rolling_beta(asset: pd.Series, market: pd.Series, window=63):
    aligned = pd.concat([asset, market], axis=1).dropna()
    aligned.columns = ["asset", "market"]
    cov = aligned["asset"].rolling(window).cov(aligned["market"])
    var = aligned["market"].rolling(window).var()
    return cov.div(var.replace(0, np.nan))

def hedge_notional(portfolio_value, beta, hedge_price, multiplier=1):
    return -portfolio_value * beta / (hedge_price * multiplier)

The contract calculation needs an FX conversion and an integer rounding rule in production. Small portfolios may be under-hedged because futures contract granularity dominates.

Estimate beta for the decision horizon

Beta is unstable. A 252-day estimate may be statistically smoother but stale after a portfolio rotation; a 20-day estimate is reactive but noisy. Select the window based on rebalance frequency and instrument behavior, then shrink extreme estimates toward sector or market averages. Estimate beta from portfolio holdings as well as realized returns: holdings-based beta updates immediately, while return-based beta reveals what actually happened.

Method	Strength	Weakness
Historical regression	Measures realized co-movement	Lags changes
Holdings-based risk model	Immediate and decomposable	Depends on factor model
Futures overlay	Liquid hedge execution	Basis and roll risk
ETF overlay	Simple for small accounts	Fees and tracking error

Hedge exposure, not every fluctuation

If the signal deliberately captures market timing, hedging beta destroys its intended return. If the mandate is stock selection, beta hedge consistently and report both gross and hedged results. A daily full hedge can create unnecessary turnover when beta is small or noisy. Establish a no-trade band and rebalance when predicted beta exposure exceeds a meaningful threshold.

def rebalance_hedge(current_contracts, desired_contracts, band=2):
    delta = desired_contracts - current_contracts
    return current_contracts if abs(delta) <= band else desired_contracts

The same cost-aware logic appears in cost-aware portfolio optimization: a perfect frictionless hedge is often inferior to a stable, cost-aware hedge.

Residual alpha diagnostics

After hedging, run a factor regression with point-in-time factors and inspect:

1. Alpha annualized with a robust standard error.
2. Factor loadings before and after the overlay.
3. Residual volatility, drawdown, skew, and autocorrelation.
4. Alpha by market regime—not only the full sample.
5. Costs, financing, borrow, and futures roll return.

If alpha disappears after adding momentum and value factors, the “stock selection” signal was likely an unhedged factor tilt. That is not necessarily bad; it is bad labeling. Compare it honestly with factor investing, which may deliver the same exposure more cheaply.

Common implementation failures

Using same-day closing beta to hedge at that close is lookahead. Applying beta to gross rather than net exposure can double hedge a long-short book. Hedging an international portfolio with a domestic future ignores currency and time-zone effects. Finally, a hedge can create collateral and margin stress exactly when the unhedged book is losing money.

Stress beta jumps, futures basis moves, and factor correlation convergence. The aim is not zero realized daily correlation; it is a transparent risk budget in which remaining P&L has a credible relationship to the research signal.

Key takeaways

• Residual alpha exists only relative to an explicit factor model and benchmark.
• Combine holdings-based and realized-return beta estimates for responsive hedging.
• Use futures or ETFs with threshold rebalancing to control overlay turnover.
• Test post-hedge P&L against style factors before calling it stock-selection alpha.
• Include basis, financing, currency, and margin risk in hedge performance.