Getting a new perspective

In portfolio management, there is a widespread belief that maximizing long-term returns and minimizing volatility are conflicting goals. While diversification has proven to be an effective tool for balancing both, a lesser-known but perhaps equally impactful approach that has shown merit is managing portfolio exposures relative to their realized volatility.

Edward Thorp, a former math professor and hedge fund manager known for his book, Beat the Dealer, argued that utilizing a technique known as the Kelly criterion offered the most efficient path to maximizing long-run rates of compounding. Also known as the “growth-optimal” portfolio, the technique aims to maximize the expected log value of a portfolio by making bets based on their expected volatility-adjusted outcomes.¹

To understand how managing exposures relative to volatility works, let’s first consider a portfolio that grows 10 percent in one period and declines by 10 percent in the next. The ending portfolio is not flat: it has effectively lost 1 percent of its starting balance even though the average return over the two periods was zero. 

Let’s consider a portfolio that returns 5 percent before falling by the same amount in the following period. That portfolio will only lose 0.25 percent relative to its starting balance. Thus, by reducing the magnitude of the return swing (the volatility) by a factor of two, the magnitude of the loss attributable to volatility falls by a factor of four. In other words, the optimal exposure to risky assets in a portfolio declines with the square of expected volatility.

Putting it into practice

But how could this be applied to long-term investment portfolios in real life? The answer lies in understanding that return volatility tends to be autocorrelated and to cluster in time, which means that models can use trailing realized volatility to forecast future volatility. Robert Engle won the Nobel Prize in Economics in 2003 for demonstrating that, in essence, realized volatility is a good predictor of future volatility.²

To that end, we can create a simple backtest to understand the impact of managing portfolios relative to their realized volatility. In this case, we will use well-known exchange-traded funds as proxies for the various elements of the backtest, where “SPY” will represent the S&P 500, “TLT” will represent riskless bonds, and a lagged 60-day simple moving average of VIX will be used as the volatility measure (which we will refer to as “VOL”). The time period covered will start on January 1, 2002, and end on December 31, 2022.

The rules for the backtest are as follows: At the end of every month, if VOL is rising (measured as the month-over-month rate of change in VOL), the portfolio allocation to SPY for the following month will be 50 percent, with the other 50 percent allocated to TLT. If VOL is declining, then 100 percent of capital is allocated to SPY.

The results are intriguing. Over the 21-year testing period, the “buy-and-hold” SPY portfolio generated a total return of 399 percent, or 8.0 percent compound annual growth rate (CAGR), with an annualized standard deviation of 15.1 percent. The volatility-adjusted portfolio, on the other hand, generated a total return of 725 percent, or 10.6 percent CAGR, with an annualized standard deviation of 11.8 percent. What’s even more interesting is that even if 50 percent of the capital was not allocated towards TLT (long-term bonds), and was simply left in cash, that portfolio would have generated a total return of 379 percent, or a 7.7 percent CAGR, with an annualized standard deviation of 11.3 percent. In other words, simply bringing down risk exposure after periods of heightened realized volatility has a material impact on the long-term growth of capital.

Performance and Risk Metrics Overview (Jan 2002 to Dec 2022)

Taking a closer look

Looking deeper into the results, we could use rolling ten-year CAGRs to segment how volatility management impacts returns over time. This is an important exercise because the investor experience (and psychology) is not that of investing at a single discrete point in time, but rather that of “buying” multiple times through the life of an investment.

Distribution of Rolling Annualized Ten-Year Returns (Jan 2002 to Dec 2022)

We see that the impact of the volatility-managed portfolio is felt across most rolling ten-year periods, with 0 out of 132 showing a CAGR of less than 8 percent. Even the volatility-managed portfolio that allocates 50 percent of capital toward cash saw an improvement, with only 15 periods realizing a CAGR of less than 8 percent, versus 64 periods for the “buy-and-hold” portfolio.

The backtest presented here is quite simplistic, and further testing could be performed across time horizons and asset proxies used. It does suggest, however, that managing portfolios relative to their realized volatility offers a compelling approach towards improving long-term risk-adjusted portfolio outcomes.

This article was written by guest contributor Daniel Nieto, CFA.