When I first wrote about the permanent portfolio, Adam Butler of ReSolve Asset Management (@GestaltU on Twitter) pointed me to a couple of pieces he’d done on the concept. They are both worth reading:
Permanent Portfolio Shakedown I
Permanent Portfolio Shakedown II
Of particular interest to me was the second piece, which examines a permanent portfolio with a trend following and volatility targeting overlay. As I’ve written before, I am hardwired as a mean reversion guy psychologically. So getting on board with trend following was (and remains) really hard for me. For no good reason other than my own biases, I might add. But I’ve gradually come around to the idea.
The main reason is this: trend following ensures you incorporate market feedback into your investment process. As Jesse Livermore of Philosophical Economics writes in one of his exceptional pieces on trend following:
[T]he strategy has a beneficial propensity to self-correct. When it makes an incorrect call, the incorrectness of the call causes it to be on the wrong side of the total return trend. It’s then forced to get back on the right side of the total return trend, reversing the mistake. This propensity comes at a cost, but it’s beneficial in that prevents the strategy from languishing in error for extended periods of time. Other market timing approaches, such as approaches that try to time on valuation, do not exhibit the same built-in tendency. When they get calls wrong–for example, when they wrongly estimate the market’s correct valuation–nothing forces them to undo those calls. They get no feedback from the reality of their own performances. As a consequence, they have the potential to spend inordinately long periods of time–sometimes decades or longer–stuck out of the market, earning paltry returns.
The permanent portfolio concept works because it combines assets that are essentially uncorrelated across economic and market regimes (Treasury bonds, gold, equities). But within any given regime, assets can remain out of favor for extended periods of time.
Can a trend following and volatility targeting overlay help improve the return profile? I think the above linked blog posts provide compelling evidence that it can.
So I’d like to conduct a live experiment to test this out of sample. With my own money.
As I’ve mentioned before, the core of my portfolio* is now invested in a leveraged permanent portfolio:
There is nothing magical about either VMMSX or VINEX. These are just residual holdings in an old Roth IRA (we will revisit them in a bit down below). You may also recall that NTSX is allocated 90/60 S&P 500 and laddered Treasury bills. So the overall asset allocation looks like this:
29% S&P 500
19% Laddered Treasury Bonds
23% Emerging Markets
10% Ex-US SMID Cap Equity
(116% notional exposure a.k.a ~1.2x leverage)
The thing that keeps me up at night is the allocation to emerging markets and ex-US SMID cap equity. I am willing to place a bet on these market segments but I am also acutely aware that I could be wrong. Very wrong. For an extended period of time.
And this is where I think a trend and volatility management overlay can help. Rather than put my finger in the air to judge whether to double down or fold my hand, I’ll let feedback from market prices help me adjust the views expressed in my portfolio.
Here’s how it will work:
Step 1: First, check trailing volatility for the entire portfolio. If 12%, do nothing. If greater than 12%, proceed to Step 2. There’s nothing magical about 12%. I’m just trying to pick a high enough target so I’m biased toward remaining fully invested.
Step 2: Check trailing volatility for portfolio assets. For those with 12% or less, do nothing. For those with 12%+, proceed to Step 3.
Step 3: Check each asset’s price against its 200 day moving average. If above the 200 day moving average, do nothing. If below, trim positions to create cash such that overall trailing portfolio volatility falls falls to around 12% (transaction costs and taxes must be taken into account here).
Basically what we’re doing is volatility targeting by taking money from assets with poor price trends. If we were to find ourselves below target on overall volatility, we would check portfolio assets and add cash to the assets with higher volatility and strong price trends.
I ran an initial monthly rebalancing check on 8/21. Unsurprisingly, the portfolio was well above the 12% volatility target, at 17.27%. GLD, VMMSX and VINEX were all well above the 12% threshold. However, GLD is also trading well above its 200 day moving average. Thus, I trimmed significantly from VMMSX and VINEX to add cash and bring trailing volatility back to target. (In an ideal world we would actually risk-balance the portfolio as well, so that each asset held in the portfolio contributed the same amount of volatility. Unfortunately, at least as far as I am aware, I don’t have the tools available to do this in a small account)
You can compare the “before and after” portfolios here.
This is just a rebalancing mechanism for what is, on its own, a fairly well-balanced portfolio. Except here you are favoring the assets that are “working.” We are effectively mean-variance optimizing a highly diversified portfolio over short time horizons. Because we are optimizing more frequently, we are better positioned to adapt to regime changes than we are when using longer time periods.
Here are the results from my leveraged permanent portfolio since May. The timing is completely coincidental, and most definitely favors the permanent portfolio, but I think it’s compelling “live” evidence nonetheless (note that the first overlay-based rebalance did not take place until August 21).
*Ex-401(k). 401(k) investment options are literally the worst.
5 thoughts on “Permanent Portfolio + Trend”
How can one measure or see the volatility easily?
Would stockcharts Standard Deviation work? I’m guessing not because I can’t see a period when the SD was above the threshold. (e.g. currently, the 200 day SD is 82 and gold is $1529, so the 200 day volatility is only 5%?).
How would one easily check trailing volatility?
e.g. Using the Standard Deviation on Stockcharts, the 200 day on $gold is currently 82, which puts 200 day volatility at 82/1529 = 5%, no? I can’t see a recent time when it was above your benchmark of 12%, which suggests that one of us is doing it wrong (probably me!).
What’s the correct (and simple!) way of getting the volatility of an index?
You can do it (for free) with Portfolio Visualizer’s backtest tool. It is linked within the post.
I’ve figured out a way of doing it in R, so now I’m backtesting a variety of timing and volatility targets.
So far, there are many that beat buy ‘n hold (albeit over long periods – not day trading), but my suspicion at present is that they are specific to the stock / index being tested. i.e. the same period / volatility settings will not work for all indices.
In addition, they are also specific for the time pediod being tested.
In any case, it’s fun and I’m learning some code!
Do you know of anywhere to download historical data, other than Yahoo?
The best dataset I’ve found so far is OHLC for S&P500 going back to around 1950, but I’d like older datasets if possible, as well as other indices that go back that far.
Thanks also for this interesting take on the PP – I’ve long been interested in constructing a robust portfolio (after reading things like The Black Swan and Antifragile, as well as Harry Browne’s book) and the nerd in me can’t resist tinkering with graphs and indicators to see if there is a way of avoiding large drawdowns.