Risk Parity

Risk parity is a portfolio allocation approach that seeks to give each chosen asset or asset class an equal share of modeled portfolio risk, rather than an equal share of invested money. In its common volatility-based form, the calculation depends on both each asset’s variability and how its returns move with other holdings.

Equalizing estimated risk does not guarantee the highest return, the lowest possible volatility, or protection from losses. The risk measure and the assets being balanced must be specified.

Key Takeaways

• A 60% stock and 40% bond allocation can derive much more than 60% of its modeled volatility from stocks.
• Risk parity equalizes component risk contributions, not dollar amounts or marginal contributions alone.
• Lower-volatility assets may need larger capital allocations to contribute the same amount of risk.
• Leverage is a separate implementation decision, not a requirement for an unleveraged risk-parity portfolio.
• Estimates can fail as correlations, volatility, or funding conditions change.

The CFA Institute Research Foundation’s alternative-investments primer, section 16.4 explains this distinction between allocating capital and balancing contributions to risk.

Worked Example: Dollars Are Not Risk Shares

Assume a hypothetical portfolio has only a stock fund and a bond fund. Use annualized volatility estimates of 20% for stocks and 5% for bonds, with zero correlation. These are invented model inputs, not current market estimates. Both funds are measured in the same currency over the same horizon.

For a 60/40 allocation, the weighted stand-alone volatilities are 12% for stocks and 2% for bonds. With zero correlation, the portfolio’s variance is the sum of their squares:

The stock share of modeled volatility is 0.0144 / 0.0148 = 97.30%; the bond share is 2.70%. These percentages divide total portfolio volatility between the two components. They are not probabilities of loss or forecasts of which asset will lose money.

Now change the allocation to 20% stocks and 80% bonds. Each weighted stand-alone volatility becomes 4%:

In the second portfolio, each component contributes approximately 2.83 percentage points to the 5.66% total volatility. On $100,000 of unleveraged capital, that means $20,000 in stocks and $80,000 in bonds, not $50,000 in each.

The lower modeled volatility is a result of these particular inputs and weights. The calculation contains no expected-return estimate, so it cannot establish which portfolio offers the better future return.

Component vs. Marginal Risk Contribution

For fixed linear exposures, define portfolio volatility using the covariance matrix:

Here, w contains the position weights, Sigma is the return covariance matrix, MRC is marginal risk contribution, and RC is component risk contribution. Portfolio volatility must be positive.

MRC measures sensitivity to a small change in one exposure while holding the others fixed. RC multiplies that sensitivity by the existing weight. It is RC that adds up to total volatility and is equalized in this model:

In the 20/80 example, stock MRC is approximately 14.14% and bond MRC is 3.54%. Multiplying by their respective 20% and 80% weights gives the same 2.83 percentage-point component contribution. Equal component contributions therefore do not require equal marginal contributions.

MOSEK’s risk-budgeting documentation sets out this decomposition and the more general case of unequal risk budgets. Additional portfolio constraints can prevent exact equality.

Why Inverse Volatility Is Not Always Risk Parity

Weighting assets in inverse proportion to volatility gives lower-volatility assets more capital. For the two-asset example:

(1 / 20%) : (1 / 5%) = 5 : 20, which normalizes to 20% : 80%.

For two positive-volatility assets with positive weights, this equalizes volatility contributions as long as portfolio volatility is nonzero. Correlation still changes the total risk. Holding the same weights and volatilities but raising correlation from zero to 0.75 increases modeled portfolio volatility from 5.66% to 7.48%; the two risk shares remain equal.

With three or more assets, inverse-volatility weights need not equalize risk contributions. Consider three assets with the same 10% volatility. Suppose A and B have correlation 0.8, while C has zero correlation with each. Inverse volatility gives all three equal capital weights, but the modeled risk shares are approximately:

A and B reinforce each other’s movements. C does not, under these assumptions. Ignoring that relationship misses the reason their risk contributions differ.

Leverage Example: Balanced Risk Can Still Mean Large Losses

Some implementations scale a risk-balanced portfolio to a higher volatility target using borrowing or derivatives. Scaling all linear exposures equally preserves their relative risk shares in an unchanged model, but increases the amount of risk borne by the investor.

Suppose an account has $100,000 of its own capital and borrows $50,000. It invests the $150,000 in the same 20/80 proportions:

• Stocks: $30,000, or 30% of the account’s starting equity.
• Bonds: $120,000, or 120% of starting equity.
• Borrowing: $50,000, leaving net starting equity of $100,000.

Assume that over one year stocks lose 20% and bonds lose 10%, with no distributions, cash flows, rebalancing, or forced sales.

If the borrowing costs a hypothetical 5% for the year, another $2,500 is due. The leveraged account’s loss becomes $20,500, or 20.5%, before other fees or taxes. Its ending investments are worth $132,000; subtracting $52,500 of principal and interest leaves $79,500.

This is an arithmetic stress illustration, not a model prediction. Real financing can require collateral or position reductions before year-end. The SEC’s margin-account investor bulletin explains forced-sale and borrowing risks for U.S. securities margin accounts; futures and other instruments have their own contractual mechanics.

How to Evaluate a Risk-Parity Portfolio

Identify the risk being balanced. This article uses volatility. A model based on a tail-loss measure or underlying economic factors may produce different weights. Balancing named asset classes does not necessarily balance inflation, interest-rate, credit, or currency exposures.

Inspect the estimates and grouping. Check the data window, return frequency, currency treatment, correlation assumptions, and rebalance rule. Splitting one stock allocation into several highly similar funds does not create several independent sources of risk.

Separate the risk budget from the risk level. Two portfolios can both have 50/50 risk shares while taking very different amounts of leverage. Report total estimated volatility and exposures alongside the contribution percentages.

Test simultaneous losses and funding stress. Model stocks and bonds falling together, changed correlations, higher financing costs, collateral calls, and limited ability to trade. Volatility is not a maximum-loss estimate.

Compare like with like. A low-volatility unleveraged portfolio and a leveraged version have different risk levels and costs. Risk parity is not the same objective as minimum variance or return optimization, and equal estimated contributions alone do not prove superiority.

Related Terms

• Portfolio Variance: Combines weights, variances, and covariances to calculate portfolio dispersion.
• Correlation: Measures how return series move together, affecting diversification.
• Leverage: Increases exposure relative to capital and can amplify losses and funding demands.
• Portfolio Rebalancing: Adjusts holdings when prices or target allocations change.
• Smart Beta: Uses alternative index construction rules; equal weighting is not necessarily risk parity.

Knowledge Check

FAQs

This article provides general financial education, not personalized investment or portfolio-construction advice. Model estimates can fail, investments can lose value, and leveraged losses can exceed the capital initially committed.