What’s the one free lunch in investing that everyone agrees about?

**Diversification**

That’s one of the reasons people espouse a broad-based equity index fund like the S&P500. What better than owning 500 of the largest market-cap weighted firms in America! Amazing diversification!

Some people go a step further and say, no no, we’re not diversified enough, we need to own the whole market, why are we arbitrarily limiting ourselves to 500 firms? Okay, let’s buy something like the Vanguard Total Market Index Fund. Even better!

Most people stop there. Some go even further, venturing to question why we’re sticking to US stocks alone and exhibiting home-country bias when we should invest in the whole world. Enter the Vanguard Total World Market Index. Now we’re crushing it right?

The truth of the matter though, is that beyond a certain small number (some say 20, some 50, some 100), the additional benefits of adding more stocks from a diversification perspective diminish substantially. Picking 50 random stocks from the S&P500 in an equal-weighted manner would perform consistently better than the S&P500, but thats not because of being lucky, but because we held them equal weight. So even though it feels like we’re winning more by diversifying more, we’re not really because weighting schemes matter, and there is a limit to intra-asset class diversification.

What we really need, is another asset-class that has historically been un-correlated to equities over the long run. Bonds is one such powerful example. Most people recognize this and do suggest some allocation to bonds in their portfolios, thus leading to famous benchmarks like the 60/40 portfolio*.

There are other asset-classes that could add value. Gold is an example, usually marching to its own beat and doing well in inflationary / stagflationary environments. Bitcoin might just turn out to be another (though it’s too volatile for our tastes).

We’re lucky to have broad-based indexes for all these major asset classes, conveniently available in ETF wrappers like SPY (for the S&P500), TLT (for long-term treasury bonds) and GLD (for gold). All these come with the advantages of being super low cost. Now, what’s a portfolio we know that has stood the test of time that combines these asset classes?

**Levered Permanent Portfolio**

Harry Browne’s permanent portfolio (first suggested in the 1980s) is a simple equal-weight portfolio of stocks, bonds, gold and cash and rebalances periodically which has done quite well over long-periods of time, compounding away at a nice clip.

As we explored in our article about leverage being just a number, the inclusion of cash at 25% just means we’re running an equal weight portfolio of stocks, bonds and gold at 0.75 leverage. (0.75 * 33% -> 25% allocation to each stocks, bonds and gold in the permanent portfolio).

Let’s see what happens when we apply our volatility targeting trick to an equal weight portfolio of stocks (SPY), long-term treasury bonds (TLT) and gold (GLD) which we rebalance monthly to 33% allocation each.

The important thing to note here is that we don’t scale the volatility of the constituents, but rather the portfolio as a whole. This is important as these three constituents, being so un-correlated, combined together yield volatility that is lower than the-sum-of-its-parts, which enables us to scale them up more in a portfolio to match our required volatility.

Let’s say we want to match the volatility of the S&P500 (which is our benchmark), and that we are free to lever up and down while matching this volatility. At each rebalancing period at the end of the month, we simply calculate the trailing 63 trading days (3 months) annualized volatility of our equal-weight portfolio and lever it up or down to match our required volatility for the coming month. We clip our max leverage at 3.0 because we don’t want to go crazy. We assume the 3 month treasury as the borrowing rate.

The results below show the cumulative returns of such a portfolio which blows the S&P500 out of the water.

We improve our metrics on all measures, sometimes drastically. Since December 2004 (the inception of GLD, the ETF introduced most recently out of SPY, TLT and GLD), you would’ve ended up with 4.9 times the money in this strategy compared to if you’d invested in the S&P500 (We get a sharpe ratio of 0.92 and an Ulcer Performance Index of 1.9 over the whole time period) .

But where can retail investors access leverage cheaply and easily? One avenue is futures contracts, which have favourable tax treatment for high turnover strategies but still have relatively large starting capital requirements. Most easily though, there are 3x daily leverage variants for all three of these asset classes in the form of ETPs with ample liquidity, making it very easy for even little capital to follow along (namely UPRO, TMF and UGLD).

**Can we do even better?**

We can further improve performance by dynamically allocating based on volatility minimization techniques which use volatility and correlation estimates instead of statically / blindly allocating equal-weight to each of the three. Furthermore, we can use ensemble models of varying look-backs for correlation, volatility and leverage estimates to get even more robust results.

But let’s leave that for another day! 🙂

**Some people have recognized the power of levering asset class diversified portfolios, but unfortunately in a static always-levered sense. It’s still a powerful concept though. The ETF ticker NTSX, which is a 1.5x levered 60/40 portfolio does this. Similarly, a long thread on the boglehead’s forum does something similar with 3x leverage through UPRO/TMF.*

Pingback: Can we beat index funds? - Beat Passive

Thanks for the great article and website.

Is it possible to explain how one can use correlation also to adjust weights in permanent portfolio example?

You’re welcome!

So if you have both volatility and correlation estimates to work with, you can use mean-variance optimization methods (covariance matrix shrinkage) to come up with various risk parity strategies. A common one which uses only volatility and correlations (and not returns) is minimum variance. For python, you can use this pretty cool library to bypass a lot of the math (and gives a good explanation too): https://pyportfolioopt.readthedocs.io/en/latest/RiskModels.html

Thanks!

Do u think 3x etf are good for long time leveraging?

It’s interesting combining the stability of permanent portofolio with leverage…. I asked myself if it would be better leverage with futures… Also whats happens if bonds and shares breaking their negative correlation and moving down or lateral in a large decade. In the actual envoirement scary for leveraged strategies, without a very solid frame and drawndown protecion for tail risks… Leverage is short volatility and it’s dangerous if gold, stocks, have a flash crash together i don’s know what happens with 3x leverage

thanks