Beyond 60/40

60/40 or 60% equities and 40% bonds has long been the standard benchmark to which portfolio performances are measured against. One typical low-cost implementation uses the total stock market (VTI/VTSMX/VTSAX) and the total bond market (AGG/BND/VBMFX/VBTLX) funds. Other acceptable variations include substituting the total stock market can be substituted with an S&P 500 index fund (SPY/VOO/VFINX/VFIAX) with little impact on the total returns. Indeed I use SPY/AGG as the benchmark in my monthly performance tracking. Another variation is to use intermediate treasuries (IEI/IEF/VGIT/VSIGX) for the bond allocation due to better credit and a more negative correlation with equities during crisis. Weightings other than 60/40, such as 70/30, 50/50 introduce different risk-return trade-offs but derive returns from the same sources. These portfolios constitute the standard “boiler-plate” recommendations to most investors, along with some kind of “glide path” of equity percentage reduction with age. These are eminently reasonable recommendations and for many people, they serve the purpose. However, as I have come to emphasize risk control and view return on a risk-adjusted rather than absolute basis, I became more and more drawn to alternative assets.

Portfolio Theory

The Modern Portfolio Theory (MPT) revolutionized our understanding of portfolio construction. One of its key teachings is the advantage of holding low or negatively correlated portfolio components ā€“ simply put, we are told to select such asset classes that also have high enough expected returns. Unfortunately, MPT is ex post, i.e., backward looking based on past returns and correlations. It’s for good reason that the warning “past return does not guarantee future results” is dotted around financial ads like the surgeon genera’s warning on cigarette packs. Future returns are notoriously difficult to forecast; in contrast, cross asset correlations, while still time varying, are more stable and predictable. More recent portfolio construction methods: risk-parity, maximum diversification, and minimum variance (reference) rely on correlations and other risk factors rather than future returns (the expected portfolio returns are still calculated from the expected future returns of each asset but the allocations are not). It is beyond the scope of this blog post to discuss the intricacies of these approaches other than saying they affirmed the MPT teaching of selecting asset classes that have low/negative cross-correlations. In my view, the most significant impact of these work is negating the notion that the recommended portfolios (typically containing a significant amount of alternative assets) as based on past performance data mining. As long as we believe the alternative assets will have low/negative correlations to the mainstream equity/bonds in 60/40 in the future, they have a place in our portfolios.

Global Asset Composition

The global investible universe is much greater than publicly traded US equities and bonds. The chart below provides a comprehensive picture (as of 2013) where global investible assets total $101 billion. US ex-REIT equities only account for 18% of the global total, roughly equal to the share of international developed (13.7%) and emerging markets (4%) combined. The share of US IG bonds is 15.2%, considerably less than international development market bonds (22.4%), while EM bonds make up 2.7%. These 4 asset classes (US/ex-US equities/bonds) are 76% (77.2% including public REITs) of the global investible universe. One key take-away is that ex-US is really important if you want to gain exposure to a more diverse set of economic drivers, although ex-US bonds give me pause given their negative interest rates.

The rest of the asset classes are usually referred to as “alternatives”, the access to which is getting easier as asset managers try to increase AUM. However, care is definitely warranted.

global_aa_20161016

Source

Princeton Endowment Model

Among major university endowments Yale and Harvard’s draw the most attention. The former, due to the legend of its long term CIO David Swensen

, and the latter due to its sheer size (~$30+ billion). Yet unbeknown to many, Princeton amassed one of the best records in recent years. Looking at the allocation of its “policy portfolio”, the thing that immediately jumps out is the low “straight” equity allocation of only 26% consisting of 10% each of domestic and emerging and 6% of international. Within this slice, the overweight of EM is noteworthy. Private equity and independent return (i.e. absolute return) each takes up another quarter. Lastly, real assets take up 19% and fixed income only 5%. Princeton’s 2014-15 financial report is worth a read as it goes over sources of over-performance (it helps to be a $22 billion endowment to have access to some great managers) as well as the evolution of its policy portfolio since 1996. Its approach is very active ā€“ the policy portfolio appears to be constantly tweaked and it uses outside active managers for each of its allocation slices. The approach has obviously worked out for them as the rolling 10 year returns have been nothing short of amazing. Though as the equity bull aged the out-performance has declined to “only” 3% a year from a stratospheric 7-10% a year. So what are the key take-aways for a retail investor who has no chance of accessing the same managers that Princeton has access to? First of all, good active managers do out-perform. Secondly, it behooves us to consider alternatives that are not correlated to the classical 60/40 but also has equity-like returns.

princeton_aa_20161016
princeton_returns_20161016

Source

Asset Correlations

Before factor investing was all the rage, there was “tilting” to small and value based on the Fama-French 3-factor model. I think of tilting as a qualitative and low cost way of getting exposure to those factors. One of the best known example is the Merriman buy-and-hold portfolios. Though from a portfolio construction point of view, most equity sub-classes are very tightly correlated to each other. The correlations are 0.9 and above amongst VTI (total), SPY (large blend), VB (small blend), VBR (small value) and VXUS (international). They’re >0.8 with VWO (emerging) and >0.7 with VNQ (REITs).

equity_corr_20161016

Source

On the other hand, correlations among various components of the bond market are all over the place. Treasuries of various maturities are negatively correlated with equities, as are treasuries dominated AGG (total bond). Interestingly so is MBS (mortgage backed securities). Munis (MUB) are uncorrelated; while HYG (high yield) is quite equity-like.

fi_corr_20161016

Source

Examples

As written before, gold is the foremost asset that has essentially zero correlation to both equities and bonds. It has been a major part in both my active and passive portfolios. There may be additional candidates within the fixed income space according to the correlation matrix above. Typically, I would replace equities with the alternatives resulting in a portfolio with much lower traditional equity percentage than most age based formula would recommend. While equities have a high expected return, this approach will likely raise risk-adjusted returns.

There are some readily accessible examples of portfolios than goes beyond the traditional 60/40. Meb Farber has been writing on this topic for a long time. Paul Novell from Investing for a living has been writing nice follow-ups to Farber with additional performance data. Below are various portfolios that he tracked. By many metrics, the 60/40 is in the bottom half of the list.

Source

Leave a Reply

Your email address will not be published.