A key debate in investing today is whether to buy US stocks or Emerging Markets (EM) stocks. US stocks have looked historically expensive for several years now, but have continued to outperform on higher earnings growth, technology leadership, and rising valuations. Firms such as Research Affiliates¹ and GMO² forecast using their valuation models that large-cap US stocks will have near-zero real returns over the next 5 to 10 years. They favor underweighting the US and instead overweighting Emerging Markets, which they project will return 5% to 10% per year.

Emerging Markets stocks have looked historically cheap for several years now, but have lagged on lower earnings growth, sector differences, and falling valuations. China is about 30% of the MSCI EM Index, so the trade dispute has further hurt performance. Hedge fund managers such as Kyle Bass and Russell Clark argue that EM will suffer more due to currency and political instability, while innovation and a “winner-take-all” marketplace still favors the US.

Both arguments are convincing and their proponents are persuasive. This creates a hard problem – should you invest now in the US or EM?

One solution is to invest equally in both. Historically, this has been a reasonable approach as each has gone through long periods of outperformance and underperformance. Exhibit 1 compares the US, EM, and a 50/50 blend over the last 30 years.

Exhibit 1: S&P 500®, MSCI EM, 50/50 blend (total returns)

S&P 500, MSCI EM, 50/50 blend (total returns)

Source: Bloomberg, Janus Henderson, June 1989 to August 2019. Note: Monthly data from index series, 50/50 blend rebalanced monthly, excludes transaction costs. Past performance is not a guide to future performance. See disclaimers for additional information on simulated performance.

 
Another solution is to assess whether the US or EM is in favor and allocate accordingly. This is impossible to do perfectly, but a simple trend model can be surprisingly helpful.

Knowing that individuals and institutions frequently allocate new capital based on trailing one-year returns, we start by comparing the relative performance of US vs EM over that period. Every month, we calculate whether the US or EM has done better over the prior 12 months.

If the US has outperformed, we guess that the current environment probably favors the US, so the switching model holds US stocks for the next month. Else the opposite is true and the model holds EM stocks for the next month. Either way, we recalculate this trailing 12-month signal a month later and continue with the same position or reallocate. To be realistic, the model also deducts a 1% annual switching cost which is well above estimated market impact for any reasonably sized portfolio.

Exhibit 2 shows how this hypothetical switching model might have performed versus the S&P, the EM index and a 50/50 blend of both, with the time periods in which the model favored EM stocks highlighted in gray. Looking back at the last 30 years, this model switched positions a total of 28 times (about once per year). The switching model was also quite balanced over time – holding EM stocks in 52% of months and US stocks in the remaining 48% of months. There were some fast switches (between gray and white or vice-versa in Exhibit 2), but usually either the US or EM stayed in favor for extended periods.

Exhibit 2: S&P 500/MSCI EM Hypothetical Switching Model Results

S&P 500/MSCI EM Hypothetical switching model results

Annualized return S&P 500/MSCI EM

Source: Bloomberg, Janus Henderson, June 1989 to August 2019. Note: Monthly data from index series, 50/50 blend rebalanced monthly, excludes transaction costs. Switching model assumes a complete portfolio turnover at time of switch and incorporates a hypothetical 1%/year cost.

Past performance is not a guide to future performance. See disclaimers for additional information on simulated performance.

 
Most recently, the US vs Emerging Markets switching model has favored the US every month from June 2018 onward. Since then (July 2018 to August 2019), the S&P 500 Index has returned +10.2% while the MSCI EM index has returned -4.5%.

There is nothing magic about using the 12-month trailing relative return to assess the current environment. We tested different trailing periods from 8 to 13 months and all behaved similarly and outperformed the static 50/50 blend (Exhibit 3). The 11-month model performed best but the difference was not statistically significant. We stick with the 12-month trailing model for simplicity.

Exhibit 3: S&P 500/MSCI EM Hypothetical Switching Model (various trailing periods)

S&P/MSCI EM Hypothetical switching (various trailing periods)

Source: Bloomberg, Janus Henderson, June 1989 to August 2019. Note: Monthly data from index series, excludes transaction costs. Switching models incorporate 1%/year cost. Past performance is not a guide to future performance. See disclaimers for additional information on simulated performance.

 
We also tested the effect of delays in implementing the signal. The switching model above assumes that the environment is determined and allocation updated on the last day of each month. This is reasonable as the signal changes slowly, can be computed quickly, and both the S&P 500 and MSCI EM have deep and liquid ETF/futures markets.

However, there is often a lag before an individual checks their monthly/quarterly statements and reacts to recent returns. Institutions also take time to notice relative outperformance, decide to adjust allocations, and deploy cash. This waiting is often costly. To show this, we built models with the adjustment done one to five months after the trend signal (indicating that the environment has changed) was triggered. Exhibit 4 compares these lagged models to the no-lag and static 50/50 blend.

Each month of lag degraded performance. After three months, returns were indistinguishable from the static 50/50 model. The excess returns of the switching model came primarily from being a fast follower, as shown in Exhibit 4 – adjusting promptly to a change in environment.

Exhibit 4: S&P 500/MSCI EM Hypothetical Switching Model (various implementation lags)

S&P/MSCI hypothetical switching (various implementation lags)

Source: Bloomberg, Janus Henderson, June 1989 to August 2019. Note: Monthly data from index series, excludes transaction costs. Switching models incorporate 1%/year cost. Past performance is not a guide to future performance. See disclaimers for additional information on simulated performance.

 
As is typical of momentum-based investment trends, subsequent participant flows help to create the excess returns captured by the early adopter. Haghani and McBride (2016)³ have highlighted that the difference between trend following and return chasing is when and how you adjust to new conditions – early and quickly usually beats late and gradually.

There are multiple solutions to the US vs EM allocation problem. One is to maintain a constant long-term allocation like the 50/50 split. Another is to emphasize valuation like Research Affiliates/GMO suggest. Others use political and macroeconomic assessments or fundamental data to determine allocations. Finally, a trend model like the switching solution shown above uses price movement to decide positioning.

There is no perfect solution that always outperforms.

A constant allocation model might continually rebalance into a highly overvalued and declining asset. Valuation models are notoriously bad as a timing signal and prices often overshoot in both directions. Market assessments are invariably subjective, while fundamental approaches generally depend on stable causal relationships. Trend models assume the current environment will persist for some time so a rapidly oscillating market can cause unnecessary position flips.

However, the trend-switching approach has three positive characteristics that make it an attractive and potentially simple solution to the US vs EM problem. First, it’s unlikely to stay wrongly positioned over time as the model periodically adapts to relative price moves. Next, it’s easy to calculate and doesn’t require any predictions of economic conditions or political actions. Finally, it can benefit from the flows of other market participants who are slower to enter and exit an asset class due to behavioral or structural reasons.

 

Footnotes

1 Research Affiliates, Asset Allocation Interactive Website, Long Run Expected Returns as of 31 August 2019

2 GMO, 7-Year Asset Class Forecast, as of 31 August m 2019

3 Haghani, Victor and McBride, Samantha, Return Chasing and Trend Following: Superficial Similarities Mask Fundamental Differences, January 2016

Note on simulated returns: The hypothetical, back-tested performance shown is for illustrative purposes only and does not represent actual performance of any client account. No accounts were managed using the portfolio composition for the periods shown and no representation is made that the hypothetical returns would be similar to actual performance had accounts actually been managed in this manner.

Hypothetical, back-tested or simulated performance has many inherent limitations only some of which are described herein. The hypothetical performance shown herein has been constructed with the benefit of hindsight and does not reflect the impact that certain economic and market factors might have had on the decision making-process. No hypothetical, back-tested or simulated performance can completely account for the impact of financial risk in actual performance. Therefore, it will invariably show better rates of return. The hypothetical performance results herein may not be realized in the actual management of accounts. No representation or warranty is made as to the reasonableness of the assumptions made or that all assumptions used in construction the hypothetical returns have been stated or fully considered. Assumption changes may have a material impact on the returns presented. This material is not representative of any particular client’s experience. Investors should not assume that they will have an investment experience similar to the hypothetical, back-tested or simulated performance shown. There are frequently material differences between hypothetical, back-tested or simulated performance results and actual results subsequently achieved by any investment strategy. Prospective investors are encouraged to contact the investment manager to discuss the methodologies and assumptions used to calculate the hypothetical performance shown herein.