Recently, BigERN and Frank Vasquez of Risk Parity Radio got into a public feud. It was a silly one. Both are quantitative guys, so you’d expect them to argue about numbers. Instead, the conversation slid into personal attacks and side issues.

Part of the problem is that they’re using the same two words, “risk parity,” to mean different things. Both are partly to blame, though I think Vasquez has the stronger case. The mix-up is a good excuse to pin down what the term technically means and what retail investors usually mean by it. I’ll borrow heavily from Vasquez’s own writing, because he explained it well. That makes it more frustrating that BigERN, who is supposed to be good at abstracting a concept, didn’t quite get there (you wonder why).

Both uses of risk parity start from one idea: own assets that don’t move together. They part ways on three questions: how to pick the weights, whether to use leverage, and to a lesser extent what counts as success.

The first use is the technical one, and it comes from professional portfolio-construction material like the CFA curriculum. You weight asset classes so that each contributes a similar amount of risk, usually measured as volatility. The second is looser. It took off after 2014, when Ray Dalio explained his “Holy Grail” principle in an interview in Tony Robbins’s book Money: Master the Game.

The principle says a robust portfolio combines uncorrelated or low-correlated assets. Dalio sketched a simplified mix of stocks, Treasury bonds, gold and commodities, and the public started calling it “the risk parity portfolio.” The label then spread to similar simplified portfolios. To keep the two meanings apart, I’ll follow Vasquez and call that family “Holy Grail–style” from here on. You can probably guess why the sloppy term stuck: it’s easier to say.

CFA-style risk parity

In the technical sense, risk parity is a construction technique. You estimate each asset class’s volatility and its correlations with the others, then set weights so that each one contributes a balanced share of the portfolio’s risk. Four features define it.

A calculated allocation. The weights come out of volatility and correlation estimates, so they move when those estimates move. I’m stressing “move” because it’s the most misunderstood part of the framework. 99% of the people who invoke risk parity, professionals included, don’t have in mind that the weights aren’t static.

A bond-heavy base. Bonds are less volatile than stocks, so balancing risk instead of capital puts a lot of capital into bonds. Not always, though. See the previous point.

Leverage as a built-in tool. A risk-balanced portfolio of low-volatility assets returns too little for most goals, so leverage lifts it to a target return. For 99% of the people in this conversation, that’s obvious. For BigERN, somehow, it wasn’t. There is no risk parity without leverage.

Institutional implementation. Estimating risk, trading futures and managing financing costs usually takes a professional team. The “estimating risk” part is the crux. Risk parity stayed in the professional world because almost nobody else could run a model to estimate volatilities and correlations (and sometimes returns), let alone build a backtesting framework to trust the strategy. Maybe AI will make this easier for ordinary investors. So far, it hasn’t.

What it doesn’t help here is that even an ETF, called RPAR, is not managed according to the above.

Holy Grail–style portfolios

Here the starting point is the principle itself: combine uncorrelated or low-correlated assets so the portfolio holds up under different economic conditions. A common example is the four-quadrant model, which sorts assets by how they behave when growth and inflation each rise or fall. It’s a way of reasoning about correlations, and it isn’t unique to risk parity.

Dalio is also the man behind the All-Weather concept, which helps explain why the two ideas got tangled together. Widen the set of risk factors beyond growth and inflation, and you arrive at the Total Portfolio Approach. So rather than a set of cleanly separated investing philosophies, we have a spectrum, with each definition overlapping the others in places.

Typical retail model portfolios in this family hold four to eight funds. Two well-known examples:

  • The Golden Butterfly: 20% each in a total stock market fund, small-cap value, long-term Treasuries, short-term Treasuries and gold.
  • The Robbins–Dalio “All Seasons” mix: roughly 30% stocks, 55% Treasuries, 7.5% gold and 7.5% commodities.

It isn’t hard to see that these portfolios have more in common than not. The one genuinely new tool to reach retail investors in the last three decades is trend following. Frank Vasquez was among the first to put an allocation to trend into this kind of exercise, in his Golden Ratio portfolio.

The main takeaway is that the investable universe is much bigger than stocks and bonds (and that their private siblings are not alternatives). Alternatives can bring something relevant to the table: positive expected returns and low correlation to both stocks and bonds.

By now, most people know why everyone suddenly cares about these portfolios. For forty years, stocks and bonds were inversely correlated, and a two-asset portfolio worked beautifully. Why spend time learning about other assets when the simple version was working and life has more interesting things to offer? There was a second reason, too. Any portfolio that diversified beyond stocks and bonds without using leverage did worse than plain 60/40 in terms of raw returns. Diversifying meant complicating your life in order to underperform. Or, as someone might put it, a profoundly stupid idea.

The differences that matter

How weights are chosen. The CFA-style approach recalculates risk, so its weights move with the data. The Holy Grail–style approach sets allocations, sometimes ranges, by category and judgment, which is simpler and more transparent. The cost is that nothing guarantees balanced risk. Or, more precisely, it can look as though all the risk comes from one source, if you only examine a short window.

Take the Permanent Portfolio. After 2022, almost everyone would agree that the stock, long-bond and gold allocations each contribute a similar amount of risk, which is why equal weights make sense. The common belief that nearly all of a portfolio’s risk comes from its stocks was a product of four decades in which bonds were unusually calm.

Now treat the portfolio’s cash allocation as what it is, inverse leverage. The gap between the Permanent Portfolio and a simplified risk parity portfolio gets a lot narrower, doesn’t it?

Leverage. Leverage is what lets the textbook design reach an acceptable return. Remember the line “nobody eats risk-adjusted returns”? Holy Grail–style portfolios usually rely on their stock and alternative sleeves for growth. Without leverage, they’re well-balanced allocations that fail the test most investors care about, because they don’t earn enough. The problem got worse because stocks returned roughly double their historical average for almost two decades.

Leverage closes that gap. Portfolios can be efficient and still generate enough return to fund a decent retirement or grow wealth. The trouble is that every financial crisis has involved leverage, so people learned to treat leverage itself as the culprit. That’s like blaming knives for stabbings. Every stabbing involves a knife, but most uses of a knife don’t end in one. Leverage becomes dangerous when it’s combined with concentration. Remove the concentration (and the illiquidity, if you’re in that game), and the problems with leverage almost entirely disappear.

What success means. Holy Grail–style portfolios used for retirement spending are often judged by how much you can safely withdraw through the worst historical sequences of returns. Unlevered versions come up short because their risk profile is too low. That’s the whole issue.

BigERN is right that leverage doesn’t make an inefficient portfolio efficient. It only amplifies the inefficiency. What he’s missing, for reasons I can’t explain, is that Holy Grail portfolios are efficient, and more efficient than his 75/25 solution. There’s a large body of ex-post, out-of-sample data showing it.

Was I lucky to choose DBMF for the model portfolio five years ago? Maybe. That risk is exactly why the portfolio also holds an allocation to COM. Will DBMF stop beating the index it’s supposed to track? It might. But should we erase the outperformance it has already delivered because it “wasn’t supposed to” happen? Someone did it. That matters less than whether the asset keeps its diversification properties, which in trend following are both statistical AND mechanical. The link goes to an “amateur” study, but there are plenty of peer-reviewed studies on trend following. At this point, ignoring them can only be a deliberate choice.

There are many good reasons someone might not want to allocate to trend, or to other risk premia. “It has been proven not to work” isn’t one of them, because it hasn’t.

A related lens: convexity and the stock–bond correlation

A recent article by quant researcher Yoav Git, “Convexity You Can Afford,” examines two things this strategy depends on: bonds as the diversifier, and risk-based sizing with leverage.

Git defines convexity as how much better a strategy does per unit of absolute move in the underlying asset. If equities move 5% in either direction and a trend-following fund gains 2.5%, its equity convexity is 0.5.

Across 1961–2026, in low or falling inflation regimes, a 10% fall or rise in equities was followed on average by an extra 0.8% or 1.8% bond gain. That’s a “smile” shape. In high-inflation regimes the convexity vanishes, and the stock–bond correlation turns positive, so bonds diversify less.

Trend shorts whatever keeps falling, so it profits from sustained moves in either direction. It comes with two costs. It loses money when nothing happens, and when an asset does well, trend is on average not fully invested in it. For an asset with a Sharpe ratio of 1 over two to five years, $100 of fast trend carries about $30 of exposure, mid-speed about $50 and slow about $60.

Trend funds size positions by risk, not dollars. With a 15% risk target, a roughly 20%-volatility equity position gets scaled down, while a roughly 6%-volatility 10-year bond position gets scaled up to 15%. That’s the same volatility-balancing logic as CFA-style risk parity, and it’s why Git finds about three times as much bond convexity as equity convexity per dollar.

10-year bondsEquities
Typical volatility~6%~20%
Mid-speed trend convexity, 1961–2026~2.1~0.7
Convexity bought by a 10% allocation to a typical multi-asset trend fund~9.4%~2.8%

Moving 10% from equities into such a fund trims equity exposure only from 50% to about 47.2% after a sustained decline, which Git considers poor value for “crisis alpha.” By his estimate, the same allocation shifts bond exposure from 30% to just over 20% in a bond downturn, and to nearly 40% if bonds rally. His conclusion is that trend is better used to protect the bond sleeve against inflation and fiscal shocks than as equity-crash insurance. A fund with no equity trend and a 50% bond allocation could do that job with an allocation of about 5%.

This is a short illustration of how a trend allocation can support the other two sleeves, stocks and bonds, on a Total Portfolio Approach basis. An inflation shock hits both stocks and bonds, so you care less about which sleeve the trend strategy protects and more about whether it protects anything at all.

Additional considerations on HG portfolios

Cullen Roche has probably written more than anyone else on asset-liability matching for retail investors, an approach more commonly called goal-based investing. He treats stocks as a long-duration asset, around 15 years if memory serves, which means they should be used to cover liabilities of equal or longer duration. It’s an elegant way of saying: don’t put money in stocks if you need it within five years.

Diversification lets you take long-duration assets and, after some cooking, turn them into shorter-duration ones. The beauty of the recipe is that you give up less return than you gain in shorter duration. That is probably the most misunderstood part of the strategy.

Classical economic theory says this shouldn’t be possible. The market would notice the mix had those properties and bid it up until its risk-adjusted return matched everything else. In that world, corporate bonds wouldn’t exist. Nobody would buy them, because a blend of stocks and government bonds would be strictly better. And yet corporate bonds exist. (A lot of financial regulation helps explain why: many market participants are not allowed to hold a stock–government bond mix, or are heavily penalized for it.)

Another way to see it is as single-line-item bias. The portfolio as a whole is less volatile, so its risk-adjusted return is higher. But many investors stop at the volatility of each component, see that risk and react accordingly.

The data is available to everyone, and yet the market doesn’t close the gap. Most market participants don’t see it, which is very convenient for those of us who do.

A few months ago, someone posted a link on our subreddit to the quarterly update from Return Stacked, specifically the Spanish chapter. These are funds available to retail investors in Spain, living in UCITS land.

Sure, they’re only three years old, and the figures may be gross of their roughly 80bps management fee. But still, if you don’t see it here, you probably never will.

What I am reading now:

Follow me on Bluesky @nprotasoni.bsky.social


0 Comments

Leave a Reply

Avatar placeholder

Your email address will not be published. Required fields are marked *