Equal Risk, Not Equal Dollars: The Math Behind Risk Parity

Here is an uncomfortable fact about the classic 60/40 portfolio. You put 60 percent of your dollars in stocks and 40 percent in bonds, and it feels balanced. It is not. Stocks are far more volatile than bonds, so when you measure where the portfolio's risk comes from rather than where the money goes, the stock sleeve typically accounts for the overwhelming majority of it. Your "balanced" portfolio is, in risk terms, an equity portfolio with a small bond garnish.

Risk parity is the response: forget equalizing dollars, equalize risk contributions. The idea had been floating around the industry, notably at Bridgewater, for years. What Maillard, Roncalli and Teiletche did in 2010 was give it a proper mathematical treatment: define it precisely, prove it exists and is unique, work out where it sits relative to the alternatives, and show what it actually does.

The problem: dollar weights hide where your risk really is

The trouble starts with the fact that "how much of my portfolio is in asset X" has two completely different answers.

The dollar answer: the fraction of capital allocated to it. Easy to compute, easy to explain, and almost useless for understanding risk.

The risk answer: how much of the portfolio's total volatility is caused by that asset. This depends on the asset's own volatility and on how it correlates with everything else. A volatile asset that moves with the rest of the portfolio contributes a lot of risk. A calm asset that hedges the others contributes little, and can even contribute negatively.

Once you compute the risk answer for a standard 60/40 portfolio, the illusion of balance evaporates. And it gets worse: because equities dominate the risk, the portfolio's fate is essentially the fate of equities. In an equity crash, which is precisely when you want your diversification to work, you find out you did not have much.

Two obvious alternatives each have problems. Equal weighting ignores risk entirely, so it is even more exposed to the fact that some assets are much wilder than others. Minimum variance goes to the opposite extreme, piling into whichever assets look calmest, which produces highly concentrated portfolios that are exquisitely sensitive to estimation error in the covariance matrix.

Risk parity aims to sit between them.

The key idea via analogy: the rowing team

Picture a rowing boat with four rowers. Equal weighting is giving each rower an oar of the same size, regardless of their strength: the strongest rower will pull the boat off course. Minimum variance is handing almost all the rowing to whoever seems steadiest and letting the others rest, which works right up until that rower has an off day and you have no backup.

Equal risk contribution is sizing each rower's oar so that every rower is putting in the same amount of effort and exerting the same influence on the boat's direction. No single rower can steer the boat by themselves, and no single rower having a bad day sinks you.

Concretely, the recipe is: size each asset so that its contribution to total portfolio volatility is the same as every other asset's. Since a wilder asset contributes more risk per dollar, this means you hold less of it. Bonds, being calm, get a large dollar weight. Equities, being wild, get a small one. The dollar allocation looks bizarre next to 60/40, but the risk allocation is even.

There is a special case worth knowing because it is so intuitive: if all assets have the same correlation with each other, the equal-risk-contribution weights are simply proportional to one divided by each asset's volatility. Inverse volatility weighting. That simple rule is where a lot of practical risk parity implementations start.

What the paper actually established

The contribution here is rigor, not the idea. The authors:

• Defined risk contributions formally, using the fact that portfolio volatility decomposes cleanly into per-asset pieces that sum to the total.
• Proved the equal-risk-contribution portfolio exists and is unique under long-only constraints. This matters. If the target could be met by many different portfolios, or by none, the concept would be practically useless.
• Located it on the risk spectrum. They showed the volatility of the ERC portfolio sits between the volatility of the minimum-variance portfolio and the volatility of the equally weighted portfolio. That is a satisfying result: ERC is a genuine middle path, more risk-aware than naive equal weighting, less extreme and less concentrated than minimum variance.
• Characterized the weights. An asset's weight ends up inversely related to its risk contribution per unit of exposure, which is why lower-volatility and lower-beta assets get bigger allocations.

Why it mattered

• It turned an industry practice into a defensible method. Risk parity was being sold before it was properly understood. This paper gave it a definition, an existence proof, and a place in the taxonomy of portfolio construction rules, which is what allowed it to be studied and criticized seriously.
• It needs no expected returns. Like minimum variance, ERC requires only the covariance matrix. Given how catastrophically bad expected return estimates are, and how much damage they do inside an optimizer, a rule that simply refuses to use them dodges the single largest source of estimation error in portfolio construction.
• It generalizes to risk budgeting. Once you can equalize risk contributions, you can just as easily target unequal ones: 40 percent of risk here, 30 percent there. That framework, risk budgeting, is now how a great many institutional portfolios are described and governed.
• It made "where is my risk" a standard question. The permanent contribution of this line of work may simply be that no serious allocator now looks only at dollar weights.

The honest limitations

• Volatility is not risk. ERC equalizes contributions to volatility, which treats upside and downside symmetrically and assumes the world is reasonably well described by a covariance matrix. It says nothing about tail risk, skew, or the possibility that a "safe" asset has a small chance of a very large loss.
• It usually needs leverage to be interesting. A risk parity portfolio is naturally low-volatility, because it is dominated in dollar terms by calm assets like bonds. To get equity-like returns you have to lever it up. That imports funding risk, margin risk, and the possibility of forced deleveraging at the worst moment. Asness, Frazzini and Pedersen argued that the ability to use leverage is precisely the source of risk parity's edge, which is an honest way of saying the edge is not free.
• It is a bet on the covariance matrix. No expected returns are needed, but correlations and volatilities still are, and both are unstable. Correlations in particular have a nasty habit of converging toward one in a crisis, which is exactly when your risk-balanced portfolio turns out to be less balanced than you thought.
• The bond exposure is a hidden regime bet. Risk parity portfolios built over the last few decades held large levered bond positions during one of the greatest bond bull markets in history. Backtests over that period look wonderful. Whether the approach performs as well in a sustained rising-rate or inflationary regime is a genuinely open question, and 2022 was an uncomfortable reminder.

The one-line takeaway

Maillard, Roncalli and Teiletche formalized the portfolio in which every asset contributes an equal share of total risk rather than an equal share of capital, proved it exists, is unique, and sits neatly between minimum variance and equal weighting, and in doing so gave risk parity the theoretical foundation it had been operating without.