The evolution of portfolio construction has, over the years, undertaken different permutations.  The origins of investing were rooted in pure speculation, from the South Seas bubble to the financing of wars for Kings and Queens around the world.  Speculation is still the underlining element when one engages in investment activities but the understanding of the relationship between risk and potential reward has certainly improved since the early days of capital markets.

The first instinctive move of early investors was to diversify risk among different stocks; this qualitative process drove investors toward portfolios built of 20 to 30 names grouped together by rudimentary analysis and an embryonic understanding that more stocks in a portfolio were probably less risky than just one.

It was only in the 1950s that Dr. Markowitz codified a system that would quantify risk and reward and that would allow an investor to build a portfolio optimized by those two variables: risks and expected returns.  Dr. Markowitz ushered in the era of Modern Portfolio Theory which defined the potential reward of an investment as the long term historical average of its returns and defined the risk as the historical volatility of the asset versus its expected returns.  Dr. Markowitz also realized that different assets may move more or less in correlation to each other; therefore, the trick to an optimized portfolio was not to merely add more assets but to add assets that would have a positive expected rate of return and that would ideally be uncorrelated to one another. Mean-Variance optimized portfolios became all the rage and most money managers started spending endless hours figuring out Efficient Frontier portfolios for their clients.  The idea behind this process was that a portfolio built on the Efficient Frontier would “guarantee” that the return would be maximized for each unit of risk taken. 

While Dr. Markowitz’ framework provided a much needed  improvement in understanding risk and reward, the limitations of Modern Portfolio Theory became quickly manifest.  MPT was too reliant on constant expected returns and correlations.  In real life, forward looking returns are a function of starting asset valuations and structural issues which vary with time; additionally, correlations are notoriously unstable.  MPT also relied on total investors’ rationality and efficient markets; a history of behavioral irrationality by most investors made this assumption invalid from the start.  A problem with asymmetric information also made market efficiency a chimera as was the idea that markets would be frictionless.

Post-Modern Portfolio Theory Investing

A newer approach to portfolio construction takes into consideration that markets are indeed messy and that investors are generally rational but are subject to recurrent “irrational exuberance and despair.”  Because of these realizations, current insights on portfolio theory focus on multiple risk factors and not just the Beta factor, or the asset correlation to the market. This invalidates the neat but generally useless Capital Asset Pricing Model (CAPM) as a tool for calculating expected returns.  Risk premia on assets are also considered time varying and not static; in other words, current models will consider contingencies like present valuations to determine expected returns rather than considering historical averages.  Modern money managers also focus on skewness and liquidity preferences for assets. Skewness reflects the preference of investors for lottery tickets type of trades or in technical terms asymmetric return pay-offs while liquidity preferences reflect the premium investors are generally willing to pay for the ability to liquefy an asset at a fair value quickly and with low transactional costs.

New insights also help model the effect of supply and demand on asset prices and how market inefficiencies may influence the price discovery process.  One additional interesting element of analysis is concerned with the timing of an asset losses or co-variation with bad times. This kind of analysis tries to look beyond market declines and analyzes correlations with inputs like negative consumption growth rates, rising unemployment, rising inflation, unexpected monetary tightening, illiquidity spirals and increases in volatility.

What to Learn from Harvard and Yale

So what is an investor to do?  Building a portfolio today requires more work than just running a few historical averages in an Excel spreadsheet.  A process of multilayered quantitative work must be strengthened by qualitative analysis.  One element that MPT clearly pointed out was the need for smart diversification; the problem with MPT, however, was how it led an investor to forecast the proper blend of assets.  A solid portfolio should be built using different blocks, in terms of not only assets but strategies as well. 

One vicious trait of capital markets is that during significant downturns, the only thing that goes up is correlations. This factor invalidates the value of diversification exactly at the time when investors need it the most.  One way to mitigate this problem is to add strategies that may help de-correlate a portfolio.  Adding strategies and getting the weights right for assets and strategies in the portfolio also requires looking beyond historical averages and should force money managers into running scenario dependent analyses.  Making estimates of possible futures scenarios and checking how the selected assets and strategies performed in past similar contingencies should help in creating more efficient portfolios.

One approach that has proven successful over time and attracted numerous attempts at replicating its results is the Endowment Investing Model.  This model was popularized by the successes of Yale and Harvard in managing their endowments.  The problem in replicating this model was the inability of most investors to access many of the deals available to Yale and Harvard.  However, the focus of a proactive investor should not have been on strict replication of Yale or Harvard but in analyzing the replicable characteristics that made this approach successful.  We found six elements in the endowment model which we think are the backbone of the strategy and that are actually replicable:

• Long term investment horizon
• Contrarian thinking
• Significant diversification of asset classes
• Uncorrelated strategies (factor diversification)
• Illiquidity premium capture
• Avoidance of active management in efficiently priced markets
• Real assets focus.

Most non-professional investors have a tendency of being swayed by much of the noise that surrounds capital markets.  This is unfortunate as noise and trading friction often result in significantly underwhelming investment results.  The longer the time horizon an investor has, the higher the chances of capitalizing on the true earning capabilities of an asset.  Additionally, long term investing horizons allow an investor to be a provider of liquidity when liquidity commands a high premium.  In layman terms, this means an investor can be contrarian and arbitrage market inefficiencies due to short term dislocations.  Baron Rothschild, an 18th century nobleman and banker famously said: “Buy when there’s blood in the streets…even if the blood is your own…” We could not agree more.

When constructing a portfolio, an investor should be paying attention to the dichotomy between active and passive management.  Certain markets are very efficient and unless a money manager can truly add performance by exploiting an unobserved anomaly, a passive tracking vehicle is a better choice.  Active management is appealing and sought after but it is often elusive.  Inefficient markets provide opportunities but time, talent and analytical skills are required to single out the worthy active managers from the merely expensive ones.  Yale produced a report which showed the return disparity by asset class in the top and bottom quartile of money managers per annum over the last ten years:

• US Treasuries: 0.8%
• Large Cap Domestic Equities: 1.5%
• Domestic Small Cap Equities: 2.3%
• Developed Markets Foreign Equities: 2.7%
• Emerging Market Equities: 2.8%
• Leveraged Buy-Outs: 13.8%
• Natural Resource: 17.4%
• Real Estate 19.1%
• Venture Capital: 19.8%.

Asset classes with the highest spread between top and bottom quintiles indicate markets that are relatively inefficient and that therefore warrant looking for a talented active manager.

Engineering Portfolios

The most important element in engineering solid portfolios is the estimation of expected returns and, to a degree, of future correlations.  Needless to say this is also the most difficult step.  While historical data is an important guideline, to take history at face value can cause enormous pain to investors.  The process of forecasting future returns should take into consideration different factors and not only static historical premia.  Earlier, we mentioned value as one defining input. Especially over long time horizons, value provides a significant margin of safety in most asset classes; the long term expected return of an asset is generally inversed to its current relative valuation.  Value arbitrage, however, does need time to unwind and therefore it requires discipline and patience. 

In fact, on a shorter term framework, the opposite of value, or momentum, tends to gain the upper hand.  Momentum is the tendency of an asset to continue to move in the direction of previous moves.  Structural issues and behavioral biases ensure that this anomaly continues to exist even in generally efficient markets.  A way to capitalize on this would require for example to build an equity portfolio split between a long duration value strategy and a short duration momentum component.  To this effect, Wilmer Henao of ING published a report in 2009 showing the information ration of a value strategy built on book-to-price ratios at 0.32 and the information ratio of a momentum strategy at 0.50.  However, when he blended the two strategies in a 50/50 portfolio the information ratio of the new portfolio increased to 0.86.  The higher the information ratio, the more efficient the portfolio.

Another element that may be useful in estimating returns is the size factor or the tendency of small caps to outperform over certain intervals and at specific segments in the economic cycle. Additionally, we can look at the momentum of earnings per share as a forecasting tool.  Volatility also can have a great impact in determining futures prices but it is rather difficult to estimate. Many studies have been performed on the subject, from simple stochastic analysis to more sophisticated GARCH models but improvements are still needed on this subject.

Another element that needs to be incorporated into any portfolio engineering process is inflationary expectations.  Forecasting the proper inflation rate for the investment period in question will allow for proper fine tuning of asset class weights.  For instance, if the historical annualized rate of return of bonds is 6% but the probabilities of inflation for the future are higher than average, it is reasonable to expect much lower returns going forward and therefore a lower weight in the portfolio.  Higher or lower rates of inflation have also influence on the correlation of assets.  For instance, during normal and stable times, bonds and equities tend to be correlated but during times of high inflation the correlation breaks down as equities tend to hold up while bonds significantly re-price at lower levels. 

Some Implementation Ideas

The following approaches are three frameworks that an investor can use to implement the investment concepts discussed in this analysis.

Endowment/Diversified/Mean-Variance Framework

This process starts with the creation of an efficient frontier based on expected returns, expected volatilities and expected correlations.  The second step is called “portfolio torturing” and consists in the addition or manipulation of constraints to achieve a mean-variance optimized portfolio that is also able to respond to practical implementation.  In simple words, at this stage, we create constraints that make the actual portfolio more practical and not subject to just quantitative analysis.  This step also attempts to mitigate the risk associated with non-standard asset defined as “dragon risk.”  In medieval times, “dragon risk” referred to the mapmakers’ characterization of uncharted territories as areas where dragons might exist; in today’s investment lingo, it refers to a number of issues associated with alternative assets such as the risk of asymmetries and fat tails or the potential danger of relatively unproven investment vehicles.

Core Satellite

A core-satellite or core-plus structure allows an investor to take advantage of the dichotomy between beta driven investments and alpha producers.  Traditional core-plus portfolios build the core (the larger percentage of the portfolio) around beta investments or vehicles that earn the asset class premium passively.  Usually this repartition is dedicated to highly efficient markets and is implemented with low cost vehicles such as ETFs.  The satellite or plus component (the smaller percentage of allocation) takes care of the alpha.  Alpha, or skill based return, is found in active management vehicles and usually comes with a higher management cost.  Ideally, the alpha component will de-correlate the portfolio and provide excess returns as well.  Successful satellites require the ability to single out successful and accessible active money managers.

A twist on the traditional approach is the core-plus portfolio with an alpha core as opposed to a beta core as devised by Martin Leibowitz and Anthony Bova.  The rationale behind this approach is the implied correlation of many alpha strategies to traditional beta.  True highly de-correlating alpha strategies are hard to come by but it is easier to find products or managers that can produce superior returns at some lower correlation level.  In fact, many alpha strategies may have low correlation but also have high beta sensitivity.  If the perception of available alpha is as such, the core-plus strategy may start from the alpha component, the alpha core.  In this process, the allocation should focus first on the potential returns of the non-standard assets being considered rather than on their volatility risk.  Once the alpha component is decided, the implied beta risk can be estimated and traditional assets (equities, cash and bonds) can be added as satellites in order to reach the needed level of total portfolio volatility.

Alpha Transport and Leverage

Alpha transport and leverage are not comprehensive portfolio engineering frameworks but represent two additional options that can be layered onto an asset allocation. 

Transportable alpha implies the ability to short an entire asset class or generalized strategy and go long an alpha producing manager. This process will allow an investor to arbitrage the pure skill based outperformance provided by the manager. The interesting element of true portable alpha is that it can be layered on top of the portfolio and it remains independent of the underlying allocation structure.  Unfortunately, alpha portability is not very widely available as most active alphas usually are found in inefficient and generally illiquid markets where the ability to efficiently short the entire asset class is often unavailable.

The concept of leverage is linked to the ideas discussed above. In the case of access to true alpha portability, given its independence of the portfolio’s allocation structure, an investor could employ leverage to increase exposure to the manager’s excess return up to a given constraint usually determined by the repartition of the total risk budget.

Leverage can also help engineer tailored risk/reward profiles.  Once an investor decides on which point on the risk curve she wants to be, an ideal portfolio may be constructed based on its blended volatility and with correspondent expected returns.  The process can also be inverted and start with a desired expected return which would come with its own specific volatility.  The resulting allocation (which would be the product of the processes we discussed in this article) would include different asset classes and strategies some of which may have undesirable liquidity profiles, valuation opacity and so on.  An alternative would be to rebuild the allocation by underweighting these assets - or eliminating them altogether - and by overweighting using leverage lower yielding asset which should come with lower volatility and better liquidity profiles. The targeted return and correspondent volatility of the portfolio would remain the same but it would be achieved with leverage and a different mix of assets.

Conclusions

In this study we took a journey throughout the history of portfolio construction.  The days of unchecked speculation may never be completely behind us since markets will always be swayed by the “madness or wisdom of crowds” but our understanding of risk and returns and what drives these two defining elements of financial assets has undergone quantum leaps in the last few years.  It is up to the individual investor to treasure this knowledge and leverage it into more efficient investment portfolios.