Rising Yields: The Future of Stock/ Bond Portfolios

When it comes to asset allocation, stock/ bond portfolios are front and center. For investors, these portfolios are appealing thanks to their simple construction, their balanced risk profile, and their low taxation. With a proven track record of almost 40 years, hundreds of publications recommend this type of portfolio. Now the tide is shifting, and the era of falling yields might come to an end. How will these portfolios hold up moving forward?

Construction of Stock/ Bond Portfolios

It is impossible to talk about stock/ bond portfolios without mentioning the ubiquitous “60/40” portfolio, the mother of all stock/ bond portfolios. This simple portfolio combines two ETFs, one representing the stock market and another the bond market. The portfolio statically assigns 60% of the assets to the stock market and the remainder into bonds, leading to a balanced behavior. This allocation can be altered to tailor the portfolio to the individual investor’s appetite for risk. Other than occasional rebalancing, the portfolio is maintenance-free.

Past Performance of Stock/ Bond Portfolios

To get a better understanding of the properties of stock/ bond portfolios, we construct a simple 60/40 portfolio, combining SPDR’s S&P 500 Index ETF with iShares’ 20+ Year Treasury Bond ETF.

performance of 60/40 portfolio of SPY/TLT

The chart above shows the performance over the full economic cycle, starting just before the 2008 recession. Over these 12 years, we achieve a total return quite similar to the S&P 500, however with much-reduced drawdowns. These gains are in the spirit of “making more by losing less”; the portfolio is leading during the recession, but slightly lagging during the bull market after that.

60/40 portfolio of SPY/TLT: monte-carlo analysis

While time-series charts are intuitive, they only paint a picture of the past. What we are most interested in, though, is a measure for future risks and returns. While we cannot predict the future, we can use Monte Carlo Simulations, to gain a better understanding of possible future outcomes. The critical assumption of these simulations is that the distribution of daily returns moving forward matches the past. By randomly drawing from the daily returns, we can simulate a set of several hundred “alternate realities,” for which we evaluate the gains and drawdowns separately. We visualize the set of simulations as Cumulative Distribution Functions. Using these distribution functions, we can make predictions for the future, and with a confidence level of our choice.

The upper half of the chart above shows the distribution of annual returns. We can see that the curve representing the 60/40 portfolio is much flatter, suggesting more stable returns. Over the observed period of 12 years and with 95% confidence, the annual return of our 60/40 portfolio will exceed 3.59%. Under the same conditions, the return of the S&P 500 will be no worse than a yearly loss of 2.21%.

The lower half of the chart shows the distribution of maximum drawdowns. Again, we see that the curve representing the 60/40 portfolio is much flatter, attesting to more predictable returns. Over the observed period and with 95% confidence, the maximum drawdown of our 60/40 portfolio will be less than -29.89%. Under the same conditions, the maximum drawdown of the S&P 500 will be less than -63.13%.

The chart clearly shows that the 60/40 portfolio, in comparison to the S&P 500, has a vastly improved distribution of returns.

Metric60/40 (SPY/TLT)TLTSPY
Simulation Start08/01/2007$1,000,000.00$59.34$114.36
Simulation End08/23/2019$2,777,557.59$145.70$284.85
Simulation Period12.1 years
Compound Annual Growth Rate8.84%7.73%7.86%
Stdev of Returns10.70%15.17%19.52%
Maximum Drawdown30.01%26.59%55.19%
Maximum Flat Days1,0121,0891,773
Sharpe Ratio0.780.440.40
Beta0.47-0.29n/a
Ulcer Index6.17%11.77%15.26%
Martin Ratio1.30.530.49

Looking at the individual components, the S&P 500 Index ETF, and the 20+ Year Treasury Bond ETF, reveals something surprising. While the naïve expectation would be that the performance metrics of the 60/40 land somewhere in the middle of its components, the 60/40 portfolio behaves much better than that. Its return is higher than the individual ETFs, and the standard deviation of returns is lower than the component ETFs. We achieve this result thanks to the magic of diversification and re-balancing.

How to Pick ETFs for a Stock/ Bond Portfolio

We can construct stock/ bond portfolios by combining any stock with any bond ETF, and there are hundreds of products to choose from. We are confronted with the Paradox of Choice: which ETFs should we pick and why?

Stock ETFs by Market

The chart above compares stock ETFs by market. We find that developed and emerging markets have lower returns and higher volatility than the US market. Even worse, in times of crisis, these ETFs have a high correlation to the S&P 500. It is, therefore, an easy choice to stick with the U.S. Markets.

The chart above shows how the total U.S. Market is virtually indistinguishable from the S&P 500. There are many other stock ETFs available, including value and growth indices, and equal-weighted indices. For this exercise, we have ruled them out as we prefer cap-weighted indices, due to their passive nature and low requirements for rebalancing. The S&P 500 remains a logical choice, with the many available substitutes being a bonus.

Bond ETFs by Issuer: Return

Next, we need to decide on a bond ETF. The chart above compares the returns of bonds by various issuers to the S&P 500. We find that returns and volatility of them are (or can be) very similar. It turns out that return is not the most critical aspect to look at when choosing a bond ETF.

Bond ETFs by Issuer: Beta

The next chart compares the same bonds by their 3-months rolling beta to the S&P 500. This chart shows us a key differentiator. High-yield bonds have a strong positive correlation to the stock market, investment-grade corporate bonds have a low correlation to the stock market, and treasury bonds are negatively correlated to the stock market. Due to this negative correlation, treasury bonds provide the best diversification for our portfolio and are our clear preference.

Treasury ETFs by Maturity

Treasury bonds come in many different maturities. The chart above shows that longer maturities both increase the return and the negative correlation.

We are now able to construct a successful 60/40 portfolio, relying on a proven track record back to 1982. However, how about the decade ahead?

Stock/ Bond Portfolios Moving Forward

To understand why the future might be different from the past, we need to know why our 60/40 performed so well in the past. There are mainly two factors contributing to the portfolio’s success: First, we enjoyed bond returns on par with the stock market. Secondly, the negative correlation between our portfolio components reduced our overall volatility.

The negative correlation between stocks and bonds comes a bit as a surprise. With bonds paying a fixed interest, this is only possible due to fluctuations in bond prices.

Bond Interest vs Total Return

The chart above shows the 30-year Treasury yield versus the Federal Funds Rate. The Federal Reserve sets the latter rate, and it is conceivable that the 30-year yield has to approach the 30-day rate over time. The chart shows us a long trend of declining rates. Starting around 13% in late 1982, rates dropped to a mere 2.5% in mid-2019, resulting in an average decline of 0.28% per year.

The chart further shows the total return of a hypothetical 30-year treasury bond, compared to the interest-only return of the same bond. We see that these curves differ by a wide margin. Bond valuation dictates that bond prices go up, as yields go down. Since 1982, the steady decline of yields has almost doubled the return of our hypothetical 30-year bond.

The chart also shows that yields are quite volatile. This behavior is driven by the market’s “flight to quality” and is the driving force behind the negative correlation between stocks and treasury bonds.

The following table shows the sensitivity to yield changes of a 20-year bond, based on its term:

Yield Change2.0%3.0%4.0%5.0%
+0.25%-4.01%-3.66%-3.35%-3.07%
-0.25%+4.20%+3.83%+3.50%+3.20%

We can see that dropping the yield by 0.25% will lead to a price increase somewhere between 3.2% and 4.2%. This number aligns well with our chart, suggesting that total bond returns have about doubled thanks to ever falling yields.

While we have seen interest rates drop for almost 40 years, it is clear that this can’t go on forever. Even though rates can reach zero or even become negative, there has to be a limit somewhere. Assuming that 20-year rates can decline to almost zero, bond prices can probably climb by another 40%. With this outlook, finding a replacement for bonds is not an urgent matter as of mid-2019.

However, at some point after the next recession, interest rates will probably stop falling. When this happens, the performance of long-term bonds will suffer significantly. Should rates begin to rise, the total return of long-term bonds can quickly become negative, due to the falling bond prices. This scenario is not only a drag on the portfolio, but it also has undesirable tax consequences: the interest paid is taxed immediately at the income tax rate, while we might not be able or willing to realize the capital losses from the falling bond prices.

The immediate remedy is to rotate towards shorter maturities. Doing so will make sure that we can at least enjoy the interest return and suffer less from capital losses. However, this is certainly no cure-all. The yield for the shorter-term bond ETF is lower because they contain a higher percentage of the more-recent lower yields. Further, the shorter-term ETFs have a less negative correlation to the stock market, resulting in higher volatility of our portfolio.

Rising Yields: Is There an Alternative to Bonds?

With these drawbacks in mind, it is time to think of a new plan. We want to keep the tax-deferred gains of a passive 60% stock partition, as this is an excellent benefit for those investors, holding significant assets in taxable accounts. In an environment of rising yields, bonds will not create capital gains, but only interest income. As the tax treatment of short-term capital gains is identical to that of interest income, we are free to manage the 40% partition actively, without adverse tax consequences.

We decided to use a mean-reversion strategy for this 40% partition. The central concept of these strategies is the expectation that markets overreact, and that in the short term prices revert to their historical mean or long-term trend. These strategies typically hold their assets for only a few days and provide returns which are mostly uncorrelated to the market.

In a nutshell, this is the logic of these strategies:

  • Only take positions in the direction of the current trend
  • Enter position, when the market becomes oversold
  • Increase position size, should the market become even more oversold
  • Exit position, when the market becomes overbought

There are many strategies of this kind published. For the sake of this post, we chose the “TPS” strategy from Larry Connors‘ book “High Probability ETF Trading.”

Larry Connors: High Probability ETF Trading

We have implemented the strategy verbatim as published in the book, and only applied some basic optimization of the entry and exit criteria.

Connors' TPS: Equity Curve

We find that this strategy works beautifully since 1997. It provides very steady returns with only low drawdowns. Even better, the strategy has performed very well during the last two recessions, namely the dot-com crash in 2000, and the housing crisis in 2008. With these properties, the strategy appears to be an excellent candidate to replace bonds.

Connors' TPS: Exposure

The chart above shows how the strategy enters and exits positions frequently, and gradually increases position sizes. Further, we see how the strategy switches its direction during the 2008 recession, but also during other times of market stress, most recently in late 2018.

Approaches to Synthesis Having decided on the mean-reversion strategy, we need to find a proper method of overlaying this strategy with the 60% buy and hold partition. The goal is to achieve a varying exposure, controlled by our mean-reversion strategy, like this:

  • 60% exposure to the S&P 500 as the baseline
  • 100% exposure with a bullish outlook
  • 20% exposure with a bearish outlook

Strategy Overlay: Naive

The naïve approach to achieving this dynamic exposure is, to buy and sell S&P 500 shares accordingly. However, doing so reduces our buy and hold portion to only 20% of the assets. We actively trade the remaining 80%, which becomes taxable as short-term capital gains. Consequently, the new portfolio is much less tax-efficient than a classic 60/40.

Strategy Overlay: Long/Short

In an attempt to overcome this shortfall, we could overlay the mean-reversion strategy using long and short positions. And while we cannot hold long and short positions of the same security simultaneously, we can mimic the behavior by using one of the many substitutes for S&P 500 Index ETFs. Unfortunately, the IRS will consider these substitute ETFs “substantially identical securities” and combine the long and short positions under the Wash Sale Rule. Consequently, 80% of the portfolio is taxable as short-term capital gains, even though we continuously held 60% of the assets.

Strategy Overlay: Long/Inverse

Luckily, there is an alternative. Instead of taking a short position in the Index ETF, we can also take a long position in an Inverse ETF tracking the same index. These inverse ETFs create an inverse exposure matching the amount invested. As prices fluctuate, these ETFs perform a daily reset, during which they adjust the exposure. This mechanism makes them quite complex products, which is why the IRS does not consider them substantially identical, even though they track the index. Using inverse ETFs, we are now able to achieve our goal: 60% of the portfolio grow tax-deferred while only 40% of our assets are subject to short-term capital gains.

It is worth mentioning that the daily reset causes an undesirable decay, depending on the market volatility. Therefore, we cannot recommend using these products for holding periods exceeding a few days. Fortunately, our mean-reversion strategy fits that requirement.

SPY 60 MR 40

Results

The chart above shows the performance of our newly created portfolio. As we can see, the return is very similar to the S&P 500, but with much-reduced volatility. Further, we can see how our TPS strategy is negatively correlated to the S&P 500, especially during the 2008 crisis, and in late 2018.

SPY 60 MR 40 Benchmark TLT

Comparing our new portfolio to the “classic” 60/40 we started with, we see that the two portfolios indeed behave quite similarly. Thanks to its faster recovery, the new portfolio is handling the 2008 recession slightly more gracefully. In mid-2011, the portfolio shows a noteworthy drawdown, caused by the “TPS” strategy declining in tandem with the S&P 500. In the remaining periods, the two strategies are almost indistinguishable.

Metric.60/40 (SPY/”TPS”)60/40 (SPY/TLT)SPY
Simulation Start08/01/2007$1,000,000.00$1,000,000.00$114.36
Simulation End08/23/2019$2,789,907.45$2,777,557.59$284.85
Simulation Period12.1 years
Compound Annual Growth Rate8.88%8.84%7.86%
Stdev of Returns10.73%10.70%19.52%
Maximum Drawdown26.05%30.01%55.19%
Maximum Flat Days7071,0121,773
Sharpe Ratio0.810.780.40
Beta0.540.47n/a
Ulcer Index5.09%6.17%15.26%
Martin Ratio1.641.300.49

The metrics confirm that both strategies are having very similar returns, as well as an almost identical standard deviation of returns.

SPY 60 MR 40 MonteCarlo TLT

To gain more detailed insight into the distribution of returns, we run a Monte-Carlo Analysis. This chart confirms once more that the overall returns between the portfolios are almost identical. However, this chart also reveals that our SPY/TPS portfolio has slightly more tail risk than SPY/TLT. This tail risk stems from the strategy having up to 100% exposure to the S&P 500, even if only for a few days. The time-series chart above showed a glimpse of that risk with the fluke in mid-2012.

performance chart SPY/"TPS" vs SPY/AGG Monte-Carlo analysis SPY/"TPS" vs SPY/AGG

In summary, we are holding up very well. It is worth pointing out that the SPY/TLT we are comparing to has much better characteristics than a plain vanilla SPY/AGG portfolio. Therefore, it makes sense to compare against such a portfolio as well.

When comparing our SPY/”TPS” portfolio to SPY/AGG, we find that our new portfolio has better returns with an almost identical downside.

Conclusion

We started with the observation that the performance of classic stock/bond performance has been boosted by an almost 40-year-long trend of declining interest rates. We believe that this trend will come to an end in the mid-term, probably at some point after the next recession. When that happens, the performance of classic stock/bond portfolios will sharply decline. We are attempting to address this problem by replacing the bond partition with a short-term trading strategy. We have been able to identify a structure that will protect 60% of the portfolio from taxable events, leading to overall taxation very similar to that of a classic 60/40. Using this structure, we have shown that returns and risk profiles are superior to a vanilla SPY/AGG portfolio, and almost on par with a SPY/TLT portfolio.

As of fall 2019, we still find ourselves in an environment of falling rates, making these extra steps unnecessary. However, it is always useful to analyze these questions in advance and be prepared for the future.

Moving forward, there are multiple possible improvements:

  • For once, we like to combine technical analysis with macro-economic indicators. Doing so can make a portfolio “recession aware” and further reduce the downside during turbulent times.
  • Also, we want to increase the trading activity of the mean-reversion strategy to make better use of available capital.
  • Lastly, we want to run the mean-reversion strategy in multiple markets to improve diversification.

We very much hope that we’ve been able to provide some insight into the future of stock/ bond portfolios. At Bertram Solutions, we pride ourselves in doing independent research, and this post is a good example of our approach to analyzing investments.