Equal weight, risk-aware weighting, and risk parity

Weighting decides which component controls the portfolio

Portfolio weights translate a set of assets or strategies into a combined exposure. If four components each receive 25% of capital, the portfolio is equal weight. That is simple and transparent, but not equal risk: the most volatile component can dominate the portfolio’s gains and losses.

Inverse-volatility weighting gives less capital to components with higher estimated standalone volatility. Risk parity goes further. It uses both volatility and correlation to target contributions to total portfolio risk.

These rules do not forecast returns. They decide how strongly the portfolio expresses each component under a risk model.

Capital weight and risk contribution are different

Illustrative example. Suppose two strategies each receive half of capital. Strategy A has 5% volatility and Strategy B has 20%. If their correlations and notionals are otherwise comparable, B contributes far more variability despite the equal capital split.

Inverse-volatility weights would allocate roughly four times as much capital to A as B before normalization. That balances standalone volatility but ignores whether A and B move together.

Risk contribution includes interaction. In simplified terms, a component contributes according to its weight and how it covaries with the whole portfolio. Two individually quiet assets can jointly create substantial risk if they are highly correlated.

For portfolio covariance matrix $\Sigma$ and weights $w$, component $i$'s contribution to modeled variance is:

$$\mathrm{VC}_i=w_i(\Sigma w)_i$$

The contributions sum to total portfolio variance. This compact expression is useful because it shows why standalone volatility is insufficient: the term $(\Sigma w)_i$ includes how component $i$ moves with every other position.

Why simple rules can be competitive

Expected returns are difficult to estimate. Equal weight makes no return or covariance forecast and often provides a difficult baseline for complex optimization. Risk-aware rules add estimates, but avoid relying directly on precise expected returns. In broad out-of-sample comparisons, equal weight can remain a demanding benchmark once estimation error is included.

That does not make them assumption-free. Volatility and covariance are backward-looking and change through time. Low measured risk can cause large notional positions, leverage, or concentration in similar defensive assets.

The design choices

After applying constraints, recompute the actual contributions. A constrained solution may no longer achieve equal risk even if that was the initial target.

A three-component example

Illustrative example. Assume estimated annualized volatilities of:

• Strategy A: 8%;
• Strategy B: 12%; and
• Strategy C: 24%.

Equal capital assigns 33.3% to each. Ignoring correlation for this first illustration, inverse-volatility raw weights are $1/8$, $1/12$, and $1/24$. After normalization they are approximately 50%, 33%, and 17%.

This is not yet risk parity. If A and B are highly correlated while C is independent, their marginal contributions depend on the covariance matrix. A risk-parity solution may reduce one of A or B even though both have lower standalone volatility.

If the final portfolio is scaled to a 12% volatility target, any leverage above one introduces financing, margin, and loss-amplification assumptions. Show the unscaled result first.

Designing a fair comparison

Run equal capital, inverse volatility, and any risk-parity construction on:

• the same components and dates;
• the same net return streams;
• the same estimation window;
• the same rebalance opportunities;
• the same leverage and concentration limits; and
• the same cost model.

Report weight paths, turnover, ex-ante and realized risk contribution, concentration, leverage, and drawdown. Perturb the risk window and covariance method. A robust result should not flip from one extreme allocation to another after a small data change.

Compare scheduled rebalancing with a threshold policy that trades only when weights drift materially. Lower turnover may come at the cost of larger temporary risk imbalance.

What risk parity does not promise

Equal risk contribution does not mean equal loss in every scenario. The model summarizes historical covariance, while crises can change volatility and correlation together.

It also does not create diversification when all components share one economic risk. Allocating equally across five equity strategies can remain a concentrated equity portfolio.

Important failure modes

• Volatility and covariance estimates lag a new regime.
• Quiet assets receive large notionals just before volatility rises.
• Correlation error makes modeled contributions unstable.
• Frequent risk rebalancing creates turnover and procyclical deleveraging.
• Constraints dominate the solution but the portfolio is still described as pure risk parity.
• Volatility targeting assumes financing and liquidity that disappear in stress.

Try it in Arizmic

Further reading

• Maillard, Roncalli, and Teïletche, “The Properties of Equally Weighted Risk Contribution Portfolios” (2010) — Derives the equal-risk-contribution portfolio and distinguishes it from equal capital, inverse volatility, and minimum variance under an explicit covariance model.
• DeMiguel, Garlappi, and Uppal, “Optimal Versus Naive Diversification” (2009) — Finds that a simple 1/N rule remains a demanding out-of-sample benchmark for estimated allocation models once parameter error is included.
• Engle, “Dynamic Conditional Correlation” (2002) — Models correlation as time-varying, making clear that an equal modeled risk contribution is conditional on the covariance estimate and date at which it was computed.