This blog post is available as a PDF here.

Summary

  • It is common for large asset management firms to publish their capital market assumptions: long-term global asset expected return and covariance assumptions.
  • Yet many firms do not draw the link between what published capital market assumptions say and what they mean when carried through the portfolio construction process.
  • We find several interesting results when applying simple portfolio optimization techniques to J.P. Morgan’s current published capital market assumptions.
    • Traditional, market-cap weighted stock/bond allocations are suboptimal.
    • Achieving better risk-adjusted returns than these traditional portfolios requires possessing the strength to stomach significant tracking error.
    • Core stocks and bonds are both materially underweight in the optimized portfolios. Satellite fixed income (e.g. high yield, bank loans, emerging market debt) receives large allocations across the board, taking on a portion of the return generating responsibility of stocks and a portion of the risk mitigating responsibility of bonds.
  • In reality, relying entirely on any set of capital market assumptions to build an optimized portfolio is probably unwise. We suggest how investors may utilize these types of optimization-based results while also addressing model/process risk.

J.P. Morgan Long-Term Capital Market Assumptions

A number of asset managers publish their long-term capital market assumptions on at least an annual basis.  For the unfamiliar, capital market assumptions present an individual’s or firm’s view on asset class returns, volatilities, and correlations.

Typically, we tend to be skeptical of these types of forecasts.  Not because we think capital market assumptions are useless, but rather because they are often presented as truths instead of assumptions.

That being said, we are big fans of J.P. Morgan’s capital market assumption piece.  They published the 20th anniversary version of their outlook in October of last year.

There are three things that really set J.P. Morgan’s work apart.

  1. They do not suffer from black box syndrome. The assumptions are not presented devoid of any context as to where they came from.  Instead, they go into quite a bit of detail on their methodology and the beliefs that form the foundation of their forecasts.  This transparency is crucial.  It allows investors to use the assumptions in a thoughtful, informed way.For example, the following graphic presents their forecasts for the evolution of cash interest rates in the United States, United Kingdom, Eurozone, and Japan over the next fifteen years.  If you agree with their beliefs, great.  If you don’t, you don’t have to throw away their capital market assumptions.  Instead, you could adjust the assumptions based on how you think cash rates with evolve.Screen Shot 2016-09-05 at 10.39.28 AM
  2. They recognize the limitations of normality. Post-2008, most investors seem to now be aware that many asset class returns are not normal, but instead exhibit “fat tails.” By summarizing asset class behavior with expected returns, volatilities, and correlations, most sets of capital market assumptions either ignore this reality or try to adjust for it in their volatility/correlation estimates in some non-transparent way. J.P. Morgan tackles the problem head-on, including a six-page article titled “Modelling and managing fat-tailed market risks.” The article starts on page 36 of the full document.Picture1
  3. They hold themselves accountable by comparing their past projections to the market reality that actually unfolded subsequent to publication.  To date, actual market returns have been statistically consistent with J.P. Morgan projections.Picture2

All of this being said, we do find J.P. Morgan’s piece to be a bit lacking when it comes to laying out simple, actionable takeaways for portfolio construction. Now, this very well could be intentional and so is not meant as a criticism.  However, we thought it would make sense to explore what J.P. Morgan’s asset class views imply for portfolio construction.

Quick sidebar: This commentary is not an endorsement by Newfound of J.P. Morgan’s views.  While we agree with much of what they say, we have disagreements as well.  Instead, this commentary is meant to explore the implications for portfolio construction if J.P. Morgan’s assumptions are accurate.  All capital market assumptions discussed come from J.P. Morgan Asset Management as of September 30, 2015 except for hedge fund returns and indices, which are as of June 30, 2015.  Capital market assumptions are – as the name states – assumptions and are not a guarantee of future performance. 

Traditional Stock/Bond Risk Profiles Are Suboptimal...

Constraining portfolios to only stocks and bonds is costly in today’s market environment.

Using J.P. Morgan’s assumptions, we can significantly improve the efficient frontier by removing asset class constraints and considering a full palette of investment options.  For example, J.P. Morgan estimates a 6.0% expected return and 10.6% volatility for a 60/40 portfolio of global stocks (MSCI ACWI) and U.S. core bonds (Barclays Aggregate).

For the same 10.6% target volatility, the expected return rises by 1.1% to 7.1% when we consider unconstrained access to the full universe.

Picture3Source: Capital market assumptions from J.P. Morgan.  Optimization performed by Newfound Research using a simulation-based process to account for parameter uncertainty.  Certain asset classes listed in J.P. Morgan’s capital market assumptions were not considered because they were either (i) redundant due to other asset classes that were included or (ii) difficult to access outside of private or non-liquid investment vehicles.   

1.1% of extra annual return is meaningful.  However, flipping the perspective gives even better context.

On the full frontier where all asset classes are considered, how much volatility do we need to take on to achieve the same 6.0% expected return as the traditional 60/40 portfolio?  The answer: 6.5%, a nearly 40% reduction in volatility just from taking off the asset class handcuffs.

The Takeaway: Passive, market-cap weighted stock/bond allocation profiles ignore macroeconomics, fundamentals, valuations, and risk and so may be non-optimal.   

…But Doing Better Than Traditional Stock/Bond Risk Profiles Means Stomaching Tracking Error

Like anything in life, this potential improvement does not come free.  In this case, the most stomach-churning cost – assuming that the capital market assumptions are accurate – likely comes from tracking error since the unconstrained portfolios would almost surely be benchmarked against their traditional counterparts.

We computed the tracking error between unconstrained portfolios on the efficient frontier and traditional MSCI ACWI / Barclays Aggregate portfolios with the same target volatility.  In this context, positive (negative) numbers imply that the unconstrained portfolio outperforms (underperforms).  Below, we use this data to plot a 90% confidence interval for the annual return differential between the two approaches.

Picture4

Source: Capital market assumptions from J.P. Morgan.  Optimization performed by Newfound Research using a simulation-based process to account for parameter uncertainty.  Certain asset classes listed in J.P. Morgan’s capital market assumptions were not considered because they were either (i) redundant due to other asset classes that were included or (ii) difficult to access outside of private or non-liquid investment vehicles.   

Sidebar: You might be wondering why in the graph above the leftmost column is labeled as a 4/96 stock/bond portfolio instead of a 0/100 stock/bond portfolio.  We use the 4/96 portfolio since this is the minimum volatility profile assuming that the allocation decision is only between the MSCI ACWI and Barclays Aggregate Bond.  In other words, a 0/100 portfolio is actually more volatility using J.P. Morgan’s assumptions than a 0/100, so we use the 4/96 as our lowest risk comparison point.    

 

Let’s return to our 60% MSCI ACWI / 40% Barclays Aggregate Bond example to understand exactly what this means.  As we mentioned before, the volatility of this portfolio – using J.P. Morgan’s capital market assumptions – is 10.6%.  An unconstrained portfolio, targeting the same 10.6% volatility, is expected to outperform by 1.1%.

However, this expected outperformance comes with tracking error of 3.4%.  The 90% confidence interval ranges from -4.5% to 5.6%.  In non-statistical language, this means that while we expect the unconstrained portfolio to outperform by 1.1%, this is by no means a guarantee.  The actual performance differential will vary year-by-year.  The confidence interval implies that we can be 90% confident that the actual outcome will range from the unconstrained model underperforming by 4.5% to the unconstrained model outperforming by 5.6%.

Quite a range of possibilities!

Another way to think about this issue is through the lens of underperformance probability.  Assuming that J.P. Morgan’s assumptions are correct – i.e. the unconstrained portfolio really is expected to outperform on average – what is the likelihood that the unconstrained portfolio underperforms its traditional counterpart over various holding periods?

For a 1-year holding period, the probability of underperforming is not much less than the probability of losing a coin flip.  For longer holding periods, the chances of underperforming declines, but still remains meaningful.  The unconstrained portfolio has a 29%, 23%, and 15% probability of underperforming the traditional 60/40 over 3-year, 5-year, and 10-year holding periods, respectively.

Picture5

Source: Capital market assumptions from J.P. Morgan.  Optimization performed by Newfound Research using a simulation-based process to account for parameter uncertainty.  Certain asset classes listed in J.P. Morgan’s capital market assumptions were not considered because they were either (i) redundant due to other asset classes that were included or (ii) difficult to access outside of private or non-liquid investment vehicles.   

The Takeaway: For optimized portfolios to be better than traditional stock/bond risk profiles, they must be different.  Allocation differences create tracking error and the possibility of underperformance, especially over shorter time horizons.   

Satellite Fixed Income Is King

The reason for the tracking error discussed in the last section becomes obvious when we look under the hood of the unconstrained frontier.

Picture6

Source: Capital market assumptions from J.P. Morgan.  Optimization performed by Newfound Research using a simulation-based process to account for parameter uncertainty.  Certain asset classes listed in J.P. Morgan’s capital market assumptions were not considered because they were either (i) redundant due to other asset classes that were included or (ii) difficult to access outside of private or non-liquid investment vehicles.   

The optimized portfolio with 60/40 equivalent volatility has only 35% core stocks and 3% core bonds with the remaining 62% split between satellite fixed income (e.g. high yield corporates, high yield municipals, emerging market bonds, levered loans) and alternatives.  A similar pattern emerges for other volatility levels.

Interesting results include:

  • Core fixed income is significantly underweight relative to traditional risk profiles.
  • For most conservative risk profiles, alternatives serve as a risk management complement to core fixed income without the return drag from historically low interest rates.
  • Equities are also significantly underweight. When equities do enter the portfolios, there is a major bias towards international equities and away from U.S. equities.
  • Satellite fixed income is king.  These asset classes, many of which exhibit hybrid fixed income/equity characteristics, have decidedly un-satellite-like allocations.  They help carry the risk mitigation burden of core bonds and the return generation burden of core stocks.

The Takeaway: Most investors need both return generators and risk mitigators in their portfolio.  Historically, core stocks and bonds have slotted nicely into these roles.  Going forward, it is becoming increasingly important to look to other asset classes (e.g. satellite fixed income) and strategies (e.g. managed futures, equity long/short) to fulfill these roles. 

Practical Considerations

In reality, a moderate allocation portfolio with 2% invested in U.S. equities is problematic.  Almost all investors exhibit favoritism towards their home markets, and U.S. investors are no different.  This home market bias means that the considerable tracking error that the unconstrained portfolio would have to U.S. equities could create behavioral problems.

Even if we were 100% sure of the accuracy of these assumptions – and therefore the long-term efficacy of our portfolio – having U.S. equities as a core reference point may create considerable short-run anxiety during periods our optimized portfolio underperforms.  Ultimately, a portfolio is only long-term optimal if we can have the discipline to stick with it in the short-run.

This is, of course, all before we even consider we can never actually be sure that any set of capital market assumptions, no matter how thoughtfully constructed, fully and accurately represent asset class behavior.  Uncertainty around these assumptions is just another form of model risk.

Managing these behavioral and model risks is a combination of art and science.  On the science front, we can employ optimization techniques that explicitly account for parameter uncertainty.  On the art front, we may wish to blend a number of complementary allocation approaches.  For a traditional “60/40” investor, this may mean something like this:

  1. 1/3rd allocation to market-cap weighted, passive “60/40” reference portfolio.
  2. 1/3rd allocation to unconstrained, long-term optimized portfolio using capital market assumptions (here we continue to use J.P. Morgan assumptions).
  3. 1/3rd allocation to tactical asset allocation (for the example, we use an equally weighted combination of our Risk Managed Global Sectors, Multi-Asset Income, and Total Return strategies).

As of today, the individual sleeve and overall portfolio allocations would look something like this:

Picture7

Allocations are presented as an illustrative example of how an investor may think of portfolio construction.  Allocations are subject to change and should not be considered investment advice. 

Client Talking Points

  • Based on J.P. Morgan’s outlook, investors looking to maximize success over the next decade must look beyond core stocks and bonds.
  • For these forward-looking investors, satellite fixed income asset classes like high yield bonds may offer valuable diversification without necessarily sacrificing return since they offer attractive yields.
  • Alternatives, especially those strategies with a focus on risk management, can be attractive risk mitigators in a low-interest rate environment where large allocations to core bonds, the classic risk mitigator, will likely drag down overall portfolio returns.
  • Employing forward-looking asset allocations generally means deviating from traditional stock/bond risk profiles. To be better, we have to be different.  Allocation differences will inevitably lead to periods of short-term underperformance.  Successfully implementing a forward-looking asset allocation means having appropriate expectations as well as the discipline to stick with it through periods of underperformance.

Justin is a Managing Director and Portfolio Manager at Newfound Research, a quantitative asset manager offering a suite of separately managed accounts and mutual funds. At Newfound, Justin is responsible for portfolio management, investment research, strategy development, and communication of the firm's views to clients.

Justin is a frequent speaker on industry panels and is a contributor to ETF Trends.

Prior to Newfound, Justin worked for J.P. Morgan and Deutsche Bank. At J.P. Morgan, he structured and syndicated ABS transactions while also managing risk on a proprietary ABS portfolio. At Deutsche Bank, Justin spent time on the event‐driven, high‐yield debt, and mortgage derivative trading desks.

Justin holds a Master of Science in Computational Finance and a Master of Business Administration from Carnegie Mellon University as a well as a BBA in Mathematics and Finance from the University of Notre Dame.