Basis Risk in Proxy and Cross-Hedging

Why a Hedge Can Leave a Material Residual

A hedge can reduce a market exposure without eliminating it. This is especially likely when the available instrument references a related market rather than the exact asset, liability, currency, or index being hedged. The exposure and hedge may usually move in similar directions, yet differ in the size or timing of their changes. That mismatch is basis risk.

A proxy hedge uses a more observable or tradable market measure in place of the exposure’s own market measure. A cross-hedge uses an instrument tied to a different, but related, underlying risk. These descriptions can overlap. For example, an insurer exposed to regional commercial property values might use a national property index instrument as a proxy. The hedge may be practical, but regional prices need not move exactly with the national index.

Definition: Basis risk is the risk that the value change in an exposure is not offset as intended by the value change in its hedge. For a linear exposure and hedge over a specified horizon, a useful representation of residual profit or loss is $Q R_X-HR_Z$, where $Q$ is the dollar exposure sensitivity, $R_X$ is the exposure return, $H$ is the notional of a short hedge, and $R_Z$ is the hedge-instrument return. The residual depends on the relationship between $R_X$ and $R_Z$, not merely on whether the two markets are related.

This sign convention assumes a positive exposure that gains when its market return rises and a short hedge that gains when the hedge instrument’s return falls. Thus, if the exposure is a liability whose value rises when rates fall, first express its profit or loss consistently; do not automatically treat it like a long asset. The simple formula also assumes a linear position over the chosen measurement horizon. Options and other nonlinear positions may not have a fixed dollar sensitivity, so a single return-based hedge calculation may not describe their full residual.

Sources of Basis Risk

The hedge instrument can fail to track the exposure for several distinct reasons. A good analysis names the mismatch rather than treating “correlation” as a complete explanation.

Source	How it creates residual risk
Different underlying market	A regional property exposure and a national property index can respond differently to local conditions.
Different location, grade, or instrument specification	Two commodity prices may diverge because they refer to different delivery points or product qualities.
Different timing or horizon	The exposure may reprice when contracts expire, cash flows occur, or an index is observed at a different date.
Different sensitivity or scale	Equal dollar notionals need not offset if the exposure and hedge respond by different percentages to a market move.
Changing relationship	A relationship estimated from ordinary observations may not persist during a market dislocation or a change in market structure.

In some markets, “basis” has a narrower conventional meaning, such as the spot price minus the futures price. In proxy and cross-hedging, the essential concern is broader: the exposure’s change and the hedge’s change do not offset in the required amount. Be explicit about the variables, units, horizon, and sign convention being used.

Worked Example: Residual Loss in a Regional Proxy Hedge

A hypothetical insurer has a positive sensitivity of $12,000,000 to a regional property index over a quarter. It shorts $10,000,000 of a national property-index instrument. In one market move, the regional index falls 4.5% and the national index falls 3.0%. Calculate the exposure profit or loss, hedge profit or loss, and residual. Then repeat the calculation if the national index instead falls 1.0%, with the regional fall unchanged.

Method. Use $Q R_X-HR_Z$, with $Q=\$12{,}000{,}000$, $H=\$10{,}000{,}000$, and a short hedge. A negative return on the hedge instrument produces a positive profit on the short position.

For the first move, the exposure loss is

\$12{,}000{,}000(-0.045)=-\$540{,}000.

The short hedge’s profit is

-\$10{,}000{,}000(-0.030)=+\$300{,}000.

Therefore, the residual is $-\$540{,}000+\$300{,}000=-\$240{,}000$, a loss of $240,000. Equivalently, the exposure return contribution is $-4.5\%$, while the hedge contribution per dollar of exposure is $-(10/12)(-3.0\%)=+2.5\%$; the net is $-2.0\%$ of $12,000,000, again $-\$240,000$.

If the national index instead falls 1.0%, the short hedge earns $-\$10{,}000{,}000(-0.010)=+\$100{,}000$. The regional exposure still loses $540,000, so the residual is $-\$540{,}000+\$100{,}000=-\$440{,}000$. The proxy fell in the same direction in both moves, but the amount of the hedge offset changed substantially.

Check and common mistake. The hedge gains when its instrument falls because the position is short; this sign is consistent with both calculations. A common error is to subtract the hedge’s loss as if the insurer were long the instrument, or to assume that a fall in the proxy automatically offsets the entire exposure. The correction is to calculate each position’s profit or loss separately before adding them.

Choosing a Hedge Notional by Minimizing Variance

When the exposure and hedge returns are uncertain, a minimum-variance hedge chooses the notional that minimizes the variance of the linear residual. Let $X$ denote the exposure return and $Z$ the return on the hedge instrument. For a positive exposure sensitivity $Q$ and a short hedge notional $H$, the residual is $QX-HZ$. Assume the returns have finite variances, the hedge return has positive variance, and the notional is unrestricted. This is a one-period linear calculation; it does not include transaction costs, limits, or nonlinear payoff effects.

Minimum-variance result: The notional of the short hedge that minimizes $\operatorname{Var}(QX-HZ)$ is $H^*=Q\operatorname{Cov}(X,Z)/\operatorname{Var}(Z)$. If $h=H/Q$ is the hedge notional per dollar of exposure, then $h^*=\operatorname{Cov}(X,Z)/\operatorname{Var}(Z)$. The result can also be written $h^*=\rho_{XZ}\sigma_X/\sigma_Z$, where $\rho_{XZ}$ is the correlation and $\sigma_X,\sigma_Z$ are the return standard deviations.

The formula is about variance, not a guarantee that the residual will be small in every scenario. The optimal notional depends on the estimated covariance and variance for the selected horizon. If the relationship changes, the hedge chosen from those estimates may no longer minimize variance.

To justify the result, expand the variance of the residual. Since $Q$ and $H$ are fixed,

\operatorname{Var}(QX-HZ) =Q^2\operatorname{Var}(X)+H^2\operatorname{Var}(Z)-2QH\operatorname{Cov}(X,Z).

Differentiate this quadratic expression with respect to $H$, set the derivative equal to zero, and solve:

2H\operatorname{Var}(Z)-2Q\operatorname{Cov}(X,Z)=0 \quad\Longrightarrow\quad H^*=\frac{Q\operatorname{Cov}(X,Z)}{\operatorname{Var}(Z)}.

The second derivative is $2\operatorname{Var}(Z)>0$, so this stationary point is a minimum. Substituting $H^*=Qh^*$ into the residual variance gives

\operatorname{Var}(QX-H^*Z) =Q^2\left(\operatorname{Var}(X)- \frac{\operatorname{Cov}(X,Z)^2}{\operatorname{Var}(Z)}\right).

This expression is nonnegative because the covariance cannot exceed the product of the standard deviations in absolute value. It also shows that imperfect co-movement leaves residual variance. If covariance is zero, the minimum-variance hedge is zero under these assumptions. If covariance is negative, the formula gives a negative $H^*$, meaning the variance-minimizing position is in the opposite direction to the short position defined here.

Worked Example: Minimum-Variance Notional and Residual Volatility

A hypothetical insurer has a $3,000,000 linear exposure sensitivity. For the selected horizon, the exposure return $X$ has variance 0.0016, the proxy return $Z$ has variance 0.0025, and their covariance is 0.001. Calculate the minimum-variance short notional and the standard deviation of residual profit or loss.

Method. Apply the covariance-variance formula, then substitute the optimal hedge ratio into the residual variance. The inputs are return moments in squared-return units.

First calculate the hedge ratio and notional:

h^*=\frac{0.001}{0.0025}=0.4, \qquad H^*=\$3{,}000{,}000(0.4)=\$1{,}200{,}000.

The minimum-variance position is therefore short $1,200,000 of the proxy instrument. The residual return per dollar of exposure is $X-0.4Z$, with variance

\operatorname{Var}(X-0.4Z) =0.0016+(0.4)^2(0.0025)-2(0.4)(0.001) =0.0016+0.0004-0.0008 =0.0012.

Thus the residual profit-or-loss standard deviation is

\$3{,}000{,}000\sqrt{0.0012} =\$3{,}000{,}000(0.034641\ldots) \approx \$103{,}923.

Check and common mistake. The residual variance is also $0.0016-(0.001)^2/0.0025=0.0012$, which agrees with the direct expansion. A common mistake is to use $H=Q\rho$ without adjusting for the relative volatilities. Here $\sigma_X=0.04$, $\sigma_Z=0.05$, and $\rho=0.001/(0.04\cdot0.05)=0.5$; hence $h^*=0.5(0.04/0.05)=0.4$, not 0.5. The two methods agree.

Basis Risk Under Alternative Market Moves

A variance-minimizing hedge is selected using a distributional summary of co-movement. A stress calculation instead asks what the residual would be under a specified pair of market moves. These answer different questions: a hedge can minimize variance across the modeled observations and still perform poorly when the exposure and proxy separate in a particular scenario.

Worked Example: Applying a Cross-Hedge to Two Stress Moves

A hypothetical insurer has a positive $20,000,000 sensitivity to a regional energy-market index. It considers a short hedge using a national benchmark instrument. Over the selected horizon, suppose the exposure return standard deviation is 6%, the benchmark return standard deviation is 4%, and their correlation is 0.5. Calculate the minimum-variance hedge notional. Then calculate residual profit or loss if (a) the regional index rises 8% and the benchmark rises 2%, and (b) the regional index rises 8% while the benchmark falls 3%.

Method. First calculate covariance from correlation and standard deviations; then use the minimum-variance formula. For each stress, calculate the exposure and short-hedge results separately.

The covariance and benchmark variance are

\operatorname{Cov}(X,Z)=0.5(0.06)(0.04)=0.0012, \qquad \operatorname{Var}(Z)=(0.04)^2=0.0016.

The optimal hedge ratio and notional are

h^*=\frac{0.0012}{0.0016}=0.75, \qquad H^*=\$20{,}000{,}000(0.75)=\$15{,}000{,}000.

In move (a), the exposure gains $\$20{,}000{,}000(0.08)=\$1{,}600{,}000$. The short hedge loses $\$15{,}000{,}000(0.02)=\$300{,}000$. The residual gain is therefore $\$1{,}600{,}000-\$300{,}000=\$1{,}300{,}000$.

In move (b), the exposure again gains $1,600,000. The benchmark falls 3%, so the short hedge gains $-\$15{,}000{,}000(-0.03)=\$450{,}000$. The residual gain is $\$1{,}600{,}000+\$450{,}000=\$2{,}050{,}000$. In the second stress, the hedge adds to the exposure’s gain rather than offsetting it.

Check and common mistake. The exposure is positive and the hedge is short, so a rising benchmark causes a hedge loss and a falling benchmark causes a hedge gain. A common mistake is to describe the second stress as “more protection” solely because the benchmark fell. Here both positions gain, leaving a larger net gain; whether that is favorable depends on the exposure being managed. The calculation confirms that a hedge is an offsetting position, not an unconditional reduction in every possible change in total value.

Interpretation and Practical Limits

The hedge notional should be interpreted alongside the assumed exposure measure, horizon, and market relationship. A ratio estimated using monthly returns, for example, need not be suitable for a short horizon if the co-movement differs at that frequency. Differences in contract maturity, settlement dates, or the market observations used to measure returns can also leave timing basis risk.

A useful basis-risk assessment therefore separates three questions:

What is being exposed?
Specify the market variable, direction of exposure, dollar sensitivity, and measurement horizon.

What is being used as the hedge?
Name its underlying market, notional, position direction, and any differences in location, grade, timing, or specification.

What remains after the hedge?
Calculate the residual for plausible alternative moves, including moves where the exposure and proxy diverge rather than merely move together.

Do not confuse a close historical relationship with a contractual offset. Unless the exposure and hedge payoff are matched by their terms, the residual remains exposed to their changing relationship. Nor should an estimated minimum-variance ratio be treated as a universal hedge ratio: it is conditional on the selected data, horizon, linear approximation, and unconstrained-position assumptions.

Key takeaway: A proxy or cross-hedge reduces the exposure only to the extent that its return offsets the exposure’s return in the relevant amount. Quantify the residual using a consistent sign convention, and test alternative market moves because basis risk is precisely the risk that the relationship does not behave as expected.

Check Your Understanding

For each problem, use the stated sign convention and treat each exposure as a linear positive sensitivity unless specified otherwise.

A hypothetical insurer has a $4,000,000 exposure sensitivity and shorts $2,500,000 of a proxy. The exposure return is −3% and the proxy return is −2%. Calculate the residual profit or loss. (Answer: $70,000 loss)
A $10,000,000 positive exposure is hedged with a short notional equal to 60% of the exposure. The exposure rises 5% and the proxy rises 4%. Calculate the residual. (Answer: $260,000 gain)
An exposure sensitivity is $8,000,000, $\operatorname{Cov}(X,Z)=0.0009$, and $\operatorname{Var}(Z)=0.0036$. Find the minimum-variance short hedge notional. (Answer: $2,000,000)
A hypothetical insurer has a $6,000,000 positive exposure and a $4,000,000 short proxy hedge. Calculate the residual when the exposure falls 7% and the proxy falls 5%; then calculate it when the exposure still falls 7% but the proxy instead rises 2%. (Answer: $220,000 loss in the first move; $500,000 loss in the second)
A $5,000,000 exposure has return standard deviation 5%; its proxy has standard deviation 2.5%, and their correlation is 0.4. Find the minimum-variance short notional and residual profit-or-loss standard deviation. (Answer: $4,000,000 short; approximately $229,129)

What You'll Learn