Why I think diversification is done wrong

Mix assets that don't move together and you can cut risk faster than you cut return.

They say diversification is the only free lunch. However, in my opinion, diversification today is done wrong: when a downturn hits, uncorrelated stocks become correlated and all tank together. So yes, you are reducing idiosyncratic and sector risks, but not market risk.

While it is true that market risk can't be eliminated completely, it can be reduced further than what we see with popular indices such as the S&P 500. I can't promise it will keep working, but I believe so.

Why? Because I believe it's structural. The property that makes this possible is that when mixing assets together, returns are the linear weighted average, but volatility is non-linear.

Notice how the portfolio variance has this additional term at the end, showing that volatility is also affected by how assets move together.

Let's take a simple example. Asset A returns 14% a year with 20% volatility, and asset B returns 8% a year with 20% volatility. We build a 50-50 portfolio. Let's assume, for fun, a covariance of 0, meaning the two assets are completely uncorrelated.

Wait a minute... did you see this? Because the assets don't move together, the return is about 21% lower than asset A alone, but the volatility is about 29% lower. We just built a portfolio with better risk-adjusted returns.

Same two assets, different correlation. If they move together (correlation of 1), the 50-50 portfolio has 20% volatility. At 0 it has about 14%. At -0.5 it has 10%. Same returns, less risk, only from how the assets move together.

Now, the obvious problem is that covariance is not fixed, as we discussed at the beginning. It changes over time, and assets become increasingly correlated in downturns, exactly when we need correlation the least. That's where orthogonal assets come into the picture.

Orthogonal assets are assets that tend not to move together, even during stress periods. When it rains you sell umbrellas, but you also carry sunscreen in your store for sunny days. That way you make money either way.

Here are real numbers. Since May 2019, the portfolio I track is up 277% vs 185% for the S&P 500. It's tracked live on the homepage at orthogonalportfolio.com. In 2022, when stocks and bonds fell together, the portfolio I track fell 8.5% vs 18.2% for the S&P 500. In 2008 it fell 29.9% vs 36.8%. That one is a backtest, 2022 is actual results. It lost money both times, just less.

Orthogonality is not just a matter of a backtest. Backtests play a role, but the assets need to be structurally different. For example, gold's correlation with equities is very low, and it has stayed low in most crashes (not all: it fell with stocks for a while in March 2020). More recently, we have also seen the rise of managed futures. And there are alternative assets and bonds.

Now, I know what you're thinking: in the example above, the return was lower than asset A, so we are penalized on the return side with an orthogonal portfolio. Well, put your hat on, that's where it gets crazy.

Imagine for a moment that you leverage asset A by 2x, so instead of 14% a year, you get 28% a year, at 40% volatility (simplified: this ignores financing costs and the drag that volatility puts on compounded returns, so real numbers would be a bit lower). Let's run the calculation again:

So now, instead of asset A's 14%, you get roughly 18% a year, with a volatility close to what asset A had on its own (20%). You can start to see that if you play with the weights and the leverage, you can come up with orthogonal portfolios that beat each of the initial assets individually.

How much leverage to use depends on the use case. I've been investing this way for the last seven years. In my emergency fund I use no leverage and the equity sleeve is very low. In my retirement account it's risk on. These goals vary for everyone, so I'm only telling you mine. The point is that an orthogonal portfolio can be tailored to a specific risk and return objective.

With this strategy I personally love dollar cost averaging (DCA). I trust the portfolio because of the orthogonal assets, so I'm happy to buy on schedule, and to buy more when I see a correction. I still avoid dumping big lump sums all at once and prefer to invest over time.

I also rebalance the portfolio every quarter. In most of my tests the difference between annual and quarterly rebalancing was very small, but I personally prefer to reset every quarter.

I found three major obstacles to constructing orthogonal portfolios:

Finding orthogonal assets is easier said than done. Today there are so many ETFs and mutual funds that scanning all of them for candidates, and then analyzing the relationship from a structural perspective, takes a lot of time.

Optimizing comes next. Once you have the orthogonal assets, you need the right asset mix and leverage for a specific objective. For example, if I want to beat the S&P 500, which assets do I mix, at what weights, and how much leverage should I include? What's the optimal point? The weights play a large role, and an efficient frontier model that uses bootstrapping to infer different paths from historical data can be a great tool to help choose them.

Monitoring is crucial, because orthogonality can break. In 2022, for example, the long bond, often considered mostly orthogonal to equities, fell together with equities. Those relying on the long bond as a hedge for stocks were disappointed. By monitoring the statistical relationship of the portfolio, it's possible to find out early when orthogonality doesn't seem to be holding, and to enrich or change the diversification sleeves. This takes a lot of time too.

If this is so good, why don't people talk about it more? Why are we told it's impossible to beat the S&P 500? There are many reasons. It's fairly new, and with the many new ETFs we can now construct more complex portfolios. It's also not easy to construct and monitor. It's not for the average person who just wants to buy VOO and chill.

But for me, it has worked so well. It generated an extra hundreds of thousands of dollars over the last seven years. So I want to share the theory behind it so anyone can see if it's a fit for them too.

If you prefer a done-for-me approach, paid members get the exact holdings, the orthogonal assets, the optimization research, and monitoring with a portfolio update every week. It's $299 a year, with a 14-day money-back guarantee, no questions asked.

This is what I do with my own money, not personal advice. Past results don't promise future ones.