Who this is for — Anyone who must turn a set of instruments or strategies into a governable portfolio while keeping assigned money separate from the risk each component creates.
Capital allocation is the process of distributing scarce resources across portfolio components: instruments, strategies, accounts, mandates and liquid reserves. Its output may be a set of weights, monetary amounts, notionals or limits. None of those objects is automatically a measure of risk. Two positions with the same market value may have very different volatility, liquidity, payoffs and scenario losses.
Asset allocation commonly refers to a distribution across asset classes; capital allocation may also cover strategies, desks or accounts. Both are resource decisions. A risk budget instead assigns shares of a chosen risk measure, while a risk contribution is an output calculated from the portfolio. Keeping these three levels separate prevents every percentage in a report from being labelled “allocation.”
Weights: an identity with a precise scope
In a long-only, unlevered and fully invested portfolio, a component weight may be written as:
wᵢ = Vᵢ / Vₚ; ∑ᵢ wᵢ = 1Here Vᵢ is the component’s market value and Vₚ is the
portfolio value in the same currency and at the same valuation time. This is an
accounting identity, not an investment method. With cash, short sales, futures,
options or swaps, both the sum and meaning of the weights depend on the chosen
convention. Market value, notional, delta-equivalent exposure, DV01 and gross
exposure answer different questions.
The constraint ∑ wᵢ = 1 is not universal either. A portfolio may
have gross exposure above its capital, net exposure near zero, or collateral
that is separate from notional exposure. Before comparing weights, specify the
entity, base currency, leverage, treatment of cash, fund look-through and the
metric used for derivatives.
From a decision to a portfolio
A verifiable process begins with the mandate, not with an isolated formula:
- Define the objective. Preservation, growth, hedging, relative return or liability management call for different criteria.
- Fix the scope. Eligible universe, currency, horizon, benchmark, leverage, minimum liquidity and operational capacity must be explicit.
- Choose the unit. Capital weights, sensitivities, volatility, Expected Shortfall and scenario loss cannot be mixed without a coherent conversion.
- Estimate the inputs. Expected returns, covariances, costs and liquidity are estimates exposed to sampling error and regime change.
- Apply the method and constraints. A mathematical result must pass concentration, turnover, tradable size, funding and mandate checks.
- Test beyond the central case. Scenarios, correlation shocks, gaps, margin changes and simultaneous exits reveal vulnerabilities that one historical matrix may not represent.
- Implement and reconcile. Rounding, execution prices, costs and market moves create differences between theoretical and actual weights.
Allocation is therefore a decision cycle. The weight vector is only one of its artifacts.
Different methods answer different questions
| Method | Dominant information | Limitation to disclose |
|---|---|---|
| Equal weights | number of components | ignores differences in risk and dependence |
| Strategic weights | mandate and economic judgement | depends on the decision-maker’s assumptions |
| Mean-variance | expected return and covariance | sensitive to estimation error |
| Minimum variance | estimated covariance | may concentrate in a few components and does not model every risk |
| Risk budgeting | contributions to a chosen measure | inherits that measure’s assumptions and limits |
| Scenario-based | losses under explicit shocks | depends on scenario choice and internal consistency |
Markowitz’s formulation shows how returns, variances and covariances can be combined in a portfolio-selection problem. It does not establish that the inputs are known, that variance exhausts risk, or that an estimated optimum will remain optimal out of sample. DeMiguel, Garlappi and Uppal document how estimation error can erode the theoretical advantages of many optimizers relative to a simple benchmark. That is evidence from specified samples and models, not proof that equal weights always win.
Similarly, a “more robust edge” does not determine a correct weight by itself. Allocation still needs a measurable definition of edge, estimation uncertainty, dependencies, capacity, costs and tolerable loss. A higher expected return may coexist with tail or liquidity risk that is incompatible with the mandate.
Rebalancing and drift
Rebalancing brings a portfolio back toward desired capital weights or risk levels. It may follow a calendar, a tolerance band, cash flows or a binding limit. There is no monthly, quarterly or daily frequency that fits every portfolio. Faster rebalancing may reduce some deviations while increasing turnover, spreads, market impact, taxes and reliance on noisy estimates.
Before trading, it helps to classify drift into four causes: price movement, changes in volatility or dependence, cash flows, and an intentional change in the mandate. Only the last is necessarily a new view. The other causes may require action or may be accepted within a documented band.
Common mistake — Treating 30% of capital assigned to a strategy as “30% of the risk.” Its contribution also depends on standalone risk, covariance with the rest, leverage and the selected measure.
A reading example, not a model portfolio
An account assigns 45% of its value to strategy A, 35% to B and 20% to cash. That describes capital weights. If A is relatively quiet while B reacts strongly to a factor already present elsewhere, B may contribute more risk despite receiving less capital. If the cash collateralizes futures, its 20% is not necessarily outside economic risk. Exposures, sensitivities, contributions, scenarios and liquidity constraints are needed to complete the picture.
The example does not recommend those weights. It shows why a percentage table is insufficient to evaluate an allocation.
Sources
- Harry Markowitz, Portfolio Selection, The Journal of Finance (1952) — the original portfolio-selection formulation using returns, variances and covariances.
- Victor DeMiguel, Lorenzo Garlappi and Raman Uppal, Optimal Versus Naive Diversification: How Inefficient Is the 1/N Portfolio Strategy?, The Review of Financial Studies — out-of-sample comparison and the role of estimation error.
- U.S. Securities and Exchange Commission, Division of Economic and Risk Analysis, Use of Derivatives by Registered Investment Companies — limitations of notional alone when comparing different exposures.
- Basel Committee on Banking Supervision, Principles for effective risk data aggregation and risk reporting — BCBS 239 — accuracy, completeness, reconciliation and governance of aggregated data.