Risk budgets: allocate uncertainty, not just dollars

Capital allocation and risk allocation differ

An equal-dollar portfolio does not generally divide risk equally. A volatile asset can dominate fluctuations despite receiving the same cash allocation as a calmer asset. Conversely, a small derivative margin payment can support a large exposure. Capital weights answer where money is allocated; risk contributions ask how positions affect a chosen portfolio risk measure.

Begin by naming that measure. Volatility contributions, scenario-loss contributions and drawdown contributions are not interchangeable. A portfolio optimized for one can look concentrated under another. An educational risk budget should identify the objective and the risks deliberately left outside it, rather than treating equal risk as a universally balanced state.

Marginal contribution explains the interaction

For volatility-based analysis, a position's marginal contribution reflects how a small change in its weight changes total portfolio volatility. Multiplying that sensitivity by its weight gives a component contribution under the usual smooth, homogeneous setup. These components sum to portfolio volatility, while simple standalone volatilities generally do not.

The covariance matrix drives the result. A position can have a negative contribution if it offsets other exposures under the model. That does not mean the instrument is inherently safe or always profitable; it means it reduces this particular estimated measure around the current portfolio. Large changes in weight or stressed relationships can alter the conclusion.

The volatility calculation can be made explicit. Write portfolio volatility as σp = √(wᵀΣw), where w contains position weights and Σ is the covariance matrix. For nonzero portfolio volatility, the marginal contribution of position i is (Σw)i / σp, and its component contribution is wi × (Σw)i / σp. Dividing that component by σp expresses its share of total volatility. These are local sensitivities to exposure, before imposing a particular financing or rebalancing rule; a funded switch between two assets changes both weights.

Inverse volatility is only a special shortcut

Allocating inversely to volatility gives larger capital weights to lower-volatility assets. For two uncorrelated assets it equalizes their volatility contributions when weights are chosen appropriately. With a larger correlated portfolio, inverse-volatility weights need not produce equal risk contributions because the pattern of relationships matters.

The shortcut also assumes volatility is measured meaningfully. Stale valuations, limited history or a period of suppressed variability can produce deceptively low estimates and excessive allocation. A minimum observation requirement, exposure constraints and alternative stress assumptions can reveal fragility, but they do not make the underlying estimation problem disappear.

Leverage and rebalancing create additional risks

A low-volatility mix may be levered to reach a target level of modeled risk. This adds financing costs, collateral requirements and possible forced deleveraging. The portfolio can become vulnerable precisely when volatility rises and its models demand a smaller position. Selling during market stress may then be both expensive and operationally difficult.

Rebalancing is not free. Frequent adjustments to noisy volatility estimates can generate turnover without meaningfully improving risk control. Define the schedule or tolerance bands before evaluation, and include the costs of moving between allocations. A smooth backtest based on frictionless daily weights is not evidence that the same risk budget can be maintained in live markets.

Maintain multiple limits rather than one optimizer

A practical framework can combine a modeled risk budget with concentration, leverage, liquidity and counterparty limits. These constraints may prevent mathematically equal contributions, which is not necessarily a defect. They represent risks the primary objective does not capture or uncertainty about the inputs.

Report both target and realized contributions, along with changes in the estimates that caused rebalancing. Test scenarios in which a major diversifier stops diversifying. The useful outcome is an explainable allocation process with visible trade-offs, not an aesthetically equal set of percentages. An optimizer cannot decide how much uncertainty a person can afford; it only translates the objective and constraints it is given.

Read a risk budget as an account of assumptions

Return to the uncorrelated 50/50 portfolio with 20% and 10% asset volatilities. Portfolio volatility is √0.0125, or approximately 11.18%. The first component is 0.0100 / 0.1118 ≈ 8.94 percentage points of volatility; the second is 0.0025 / 0.1118 ≈ 2.24 points. Together they produce 11.18 points. Dividing each by the total gives shares of 80% and 20%. A percentage-point contribution and a percentage share are different quantities, even when both appear in a report with a percent sign.

Suppose the first asset’s estimated volatility doubles while weights remain unchanged and correlation stays at zero. Its variance contribution numerator becomes 0.0400, compared with 0.0025 for the second asset. It now supplies about 94.1% of modeled variance rather than 80%. The allocation did not change, but the risk budget did. That distinction helps separate drift caused by market estimates from drift caused by trades or price-driven changes in weights.

A target-volatility overlay creates another moving part. If a mix has estimated volatility of 8% and a hypothetical target of 12%, an unconstrained model scales its exposure by 12 / 8 = 1.5. If the estimate later jumps to 16%, the same rule wants only 0.75 times exposure. Bringing the portfolio from 1.5 to 0.75 times exposure requires selling or closing positions; it is not an instantaneous change in a spreadsheet. Costs, available collateral and the permitted leverage range can prevent the target from being reached.

Use these calculations to ask better questions about the allocation. Which input caused a contribution to rise? Is a hedge’s negative contribution stable across alternative scenarios? Would restoring the target concentrate exposure in an asset whose price is merely stale? A budget is informative when the explanation survives such questions. Equal-looking risk shares are not an independent reason to ignore an instrument’s settlement, liquidity or counterparty constraints.