A New Model for Improving Returns
After 12 years of modest investment results in the equity markets, investors and their advisors have ratcheted down their return expectations, and have focused energy and time on adjusting their portfolios to “anything but equities.” While equity allocations are near all time lows, and net flows to equity funds are still negative, the percentage of investor assets in equities remains one of the largest individual allocations within the typical portfolio. More important, the contribution to return, both positive and negative, provided by equity allocations continues to be the most significant of the major asset categories in most investor portfolios both due to the percentage allocation and the dispersion of return.
Traditional asset allocation models have proven disappointing in recent years due to periodic spikes in correlations between asset categories. Fortunately, these periods have proven short in duration and have self corrected. The same cannot be said for correlations within the equity markets. Since 2000, the equity markets have experienced a persistent rise in correlations between individual equities, economic sectors and broad market based indices. This rise has been both cyclical, reflecting periodic reactions to macro events, and secular, reflecting the proliferation of index based mutual funds and exchange traded funds (ETFs). Ironically, it is the very model that ETFs were designed to address, the tactical allocation model, that the rise in correlations has undermined.
In this paper, we examine the rise in correlations in the equity markets and its impact on traditional asset allocation models. We then explore a new model that seeks out areas in which correlations continue to be low, and, therefore, the opportunity to generate alpha remains. Finally, we examine a new approach to building the allocation that seeks to capture the alpha potential in equity markets.
The Rise in Correlations
The theoretical foundation for asset allocation is low correlations across asset categories. A portfolio of stocks and bonds earns a higher risk-adjusted return by virtue of the fact that the historical correlation of the series of returns is not only low, but persistently low. The search for “alternatives” in their various forms is a search for low correlations. Without low correlations, asset allocation as a model for investment portfolios loses all efficacy. Such was the case in 2008 during the financial crisis. As the crisis gained momentum, correlations across asset categories approached 1, sending even the most diversified portfolio downward.
Fortunately, the spike in correlations across broad categories proved short lived. Within three months of the end of the crisis, correlations had once again returned to within one standard deviation of their long-term average. The positive conclusion was that while the fear caused by the crisis drove correlations outside of long-term norms, the arbitrage opportunity it presented was quickly recognized. Subsequent mini spikes have also experienced sharp reversals. One conclusion to be drawn is that spikes in correlations between broad categories represent a unique opportunity to exploit.
The same conclusion cannot be drawn relative to equities. Since 2000, inter‐market correlations have steadily risen. This persistent rise in correlations has coincided with a series of macro events, beginning with the internet bubble, the terrorist attack on September 11, 2001, the financial crises of 2008 and the European debt crisis today. Each of these events has served to raise risk premiums within the market. Correlations have risen in tandem.
Much like the pattern observed relative to correlations across broad asset categories, spikes in correlations surrounding macro events are often short lived and self correcting. Different in this series, however, is that there is a persistent trend toward higher highs and higher lows. This suggests an underlying secular change affecting the series. We believe that the proliferation of index-based investing through mutual funds and ETFs is driving cross correlations higher. Significantly, there appears to be little chance that this trend will be reversed in any meaningful way in the near future.
High Correlations and Their Impact on Portfolio Management
As already noted, high correlations undermine the theoretical foundation of asset allocation. Level 1 decisions, those between broad asset classes, are still effective over time as the rise in correlations between groups has tended to be short lived. The impact of spikes in correlations is severe in the short run, so called fat tail results, but has tended to self correct. The impact and management of these risks is outside the scope of this analysis.
Level 2 portfolio decisions, those within each asset class, have been profoundly impacted by the secular rise in correlations in the equity space, however.
In the classic nine-box allocation method designed by Morningstar, higher Sharpe Ratios are nearly statistically impossible given the correlations between the style boxes. In the classic approach, each box is filled. Portfolio returns are equal to the weighted average returns of the components. Risk, or volatility, is a function of not only the volatility of each component, but also the correlation between the return series of each component. With lower correlations, overall volatility is reduced for the same level of return, resulting in higher risk-adjusted returns.
The example above looks at two scenarios. Each uses the same five assets at the same portfolio weights. In scenario 1, the cross correlation is .5, resulting in a portfolio standard deviation of 7.8%. In scenario 2, we raise the cross correlation to .8 while maintaining everything else the same. The portfolio’s standard deviation rises to 9.1%.
With pair wise correlations between style boxes currently at .986, volatility is reduced by only .1 versus a simple average of the component allocations. Asset allocation has been rendered ineffective.
The impact of higher correlations not only renders passive allocation ineffective, but also greatly reduces the opportunity and impact of tactical allocation strategies. Tactical strategies by their nature require divergence of return. High correlations require a level of success that is unattainable. Within the equity asset class, tactical allocation strategies lose efficacy as asset shifts cannot result in differentiated returns sufficient to justify the probability of failure.
The debilitating effect high correlations have on traditional asset allocation models within the equity space is also evident within active equity management. In 2011, correlations spiked and only 17% of active money managers beat their market benchmark. While it is well published that over the long term less than 50% of active equity managers exceed their benchmark each year, that percentage has steadily declined in recent years.
As in the case of tactical allocation, high correlations undermine the effectiveness of active bets by stock managers by eliminating differentiated results in the underlying opportunity set. A simple example is in order. In a market composed of two stocks, we look at the results of an active manager with perfect information under two scenarios: one with a wide dispersion of returns between the two stocks, and one scenario where the dispersion of returns is narrower. In both cases the “index return” is 20%. The investor makes the same trades based on perfect information. In both cases, the investor beats the passive index, but in the low correlation environment the incremental return is substantially larger due to the opportunity set that the wider dispersion of returns presents. In the absence of perfect information, a narrower dispersion of returns lowers the probability that the value of correct decisions outweighs the cost of poor decisions. The asset manager simply has less room for error.
For the active manager, successful stock picking is predicated on a dispersion of returns in the universe that the manager is picking from. In addition, high correlations across the equity universe are exacerbated as the universe narrows. The traditional model, designed to capture the benefits of low correlation across style and market cap universes, emphasized style purity at the stock manager level. The insistence that managers “remain in their box” was a risk avoidance approach. With a low correlation environment, the benefits of traditional asset allocation would only be undermined if the active manager “drifted” out of its box. Alpha generation at the stock manager level due to style or market cap bets undercuts the advisor's allocation strategy. Total weighted returns for the portfolio as a whole may rise, but such a move by the stock manager would undercut the covariance model designed by the advisor. Risk-adjusted returns declined. Active managers were therefore constrained to relatively narrow universes. These narrow pools of potential investment candidates were already experiencing rising correlations of returns, undermining the ability of the manager to add alpha. As larger percentages of active stock managers underperformed their benchmarks, advisors shifted more money to passive investments, augmenting the trend toward a further rise in correlations. Ironically, the move to passive investments within the equity space has helped to drive correlations to levels that undermine the very tactical model that modern portfolio theory was designed to exploit.
A New Model
The search for alpha, therefore, needs to start with a search for low correlations. Notwithstanding the rise in correlations already discussed, we note that investor preferences continue to shift as to the type and characteristics of stocks that are rewarded at different stages over a market cycle. This cyclical pattern is as it has been in previous cycles. While the order and the duration of each shift in preferences changes from cycle to cycle, we note that a defined pattern has developed. We can explore these shifting preferences by examining the performance attributed to factors or metrics such as book-to-market ratios (value preference), change in accruals (risk preference), return on invested capital (profitability preference), and change in nine-month return (momentum preference) as examples. If correlations between these preferences are “low,” then an opportunity to identify a model or process for exploiting these shifts can be developed, and there exists the potential for alpha production.
We examined the correlations of nine metrics from December 31, 1990 to April 30, 2012. The period was chosen for two reasons. First, this covers the time period where correlations started at a low point, and remained low before beginning their march upward. Therefore, our theory can be tested. Second, the quality and the consistency of the data over this period is sound.
Shown below are the pair‐wise cross correlations of these nine metrics for the period as a whole. As noted, these correlations tend to be very low.
Now we look at the time series of correlations to see if there is a trend to a materially higher level as we have seen in the equity market as a whole. The series is volatile, and while there is a modest overall trend to higher correlations, they remain under 0.2.
From the series of nine factors, we looked at one paired series for further insight. Shown below is the pair-wise correlation of book-to-market (valuation) versus 39-week return (momentum). Note that for most of the time series, the correlation between these two factors is negative. Importantly, the series remains negative today. This is the classic tug of war between valuation and momentum. This should not be confused with “growth versus value” in the traditional model. Valuation, for example, exists within the classic "growth" space in the form of Growth at a Reasonable Price, or GARP. GARP growth managers versus momentum managers tend to have very different return patterns.
In our examination of the 36 combinations of the nine metrics noted above, we note no consistent trend toward higher correlations. Factor correlations remain low. If, as postulated, a prerequisite of alpha generation is low correlations, factor, or metric-based decisions, provide an opportunity.
Long versus Long/Short
Our empirical research also explored the efficacy of adding a long/short component to the portfolio as an additional alpha driver. If top ranked stocks by a particular factor versus bottom ranked stocks by that same factor both generated positive alpha over time, and that alpha production also had a low correlation ratio, the portfolio would benefit through higher and more consistent returns, and have the potential for lower overall volatility.
First, we examine the return potential of the nine factors. Shown below is the cumulative performance by factor of a simulated long/short of the top decile ranked and bottom decile ranked stocks. While not all factors successfully generated alpha through a long/short construct, most did.
Correlation of the Metrics in a Long/Short Construct
Significantly the correlations on average are negative over the time series, and show no rising trend. Adding long/short to the long portfolio may not only add alpha but also reduce volatility.
Portfolio Construction: Exploiting Metric Dispersion
As discussed earlier, the traditional construction methodology had the advisor capturing the benefits of low correlations by making equity designated allocations across style and market cap buckets, i.e., Morningstar nine box. Asset managers were given the mandate to remain “stylistically pure,” focusing solely on stock picking as their contribution to alpha. High correlations undermine both the advisor and the asset manager. To capture the alpha potential in factors or metrics, the responsibility for portfolio construction within the equity allocation must shift more toward the asset manager. Whether fundamental or quantitative, the asset manager has the ability to build into the process the evaluation of trends and relative valuation of various factors or metrics. By evaluating metrics across the style and market cap buckets, the asset manager is given the opportunity to add alpha not just through successful stock picking, but also through timely identification of trends in investor preferences in the characteristics of companies and their stocks based on the environment that currently exists. The advisor simply does not have the tools to effectively implement the strategy. For example, ETF’s do not exist that would allow an advisor to position the portfolio to take advantage of the superior performance of stocks that are successfully improving profitability this quarter, while shifting to stocks that display superior valuation characteristics for next quarter. For the asset manager with a process sensitive to these changing trends, the portfolio is built from the bottom up, with these changing characteristics reflected in the individual equities in the portfolio.
The Portfolio Construction Model Would Look Like This:
To exploit the opportunity that low correlations in metrics provide, the advisor gives the asset manager a broader mandate. The asset manager is thus charged with the task of seeking out multiple sources of alpha, adding to stock picking, style, market cap, and most importantly, factor or metric tilts that reflect the changing preferences in the market. Note in the diagram, that the center of the allocation is materially broadened, giving the asset manager the latitude to pursue the alpha opportunities discussed; yet we continue to include allocations, albeit smaller, to the traditional style box. This is to recognize that even in a high correlation environment, temporary opportunities exist that should be exploited by the advisor through a direct style/market cap bet. One of the more recent examples was small cap stocks in the summer of 2010.
The shift in the way that equity portfolios are constructed and managed is not new to other asset categories. Advisors have often readily shifted responsibility in the alternative category to the asset manager in recognition of the advisor's lack of tools necessary to evaluate the opportunities available. Fixed income has also seen this trend, as asset managers often are given the ability to allocate across investment grade versus high yield, domestic versus international, developed versus developing, as their bottom-up analysis dictates. In the equity space, the proliferation of the Morningstar model in the 1980’s and 1990’s eliminated the critical analysis of the model even as the efficacy of such an approach eroded.
We began by evaluating the traditional model for portfolio construction and its necessary foundation of low cross correlations. Despite periodic spikes, we note that across broad asset categories, low correlations continue to exist, thereby supporting the efficacy of asset allocation. We then reviewed the correlations within the equity market. We observed both a cyclical as well as a secular rise. This has served to undermine the traditional approach to portfolio construction within the asset group.
With the postulation that alpha is first a function of dispersion, it was necessary to seek out where, within the equity universe, cross correlations remain low. We discovered that factor dispersion remains low. With a lack of the tools necessary to exploit low factor dispersion at advisors' disposal, we present a portfolio construction model that shifts the allocation responsibility within the equity category away from the advisor to the asset manager. This broader mandate gives the asset manager the ability to pursue multiple sources of alpha by exploiting factor dispersion. Further we noted that factor correlation in a long/short environment can add to portfolio risk-adjusted returns through generation of incremental return, and through low correlations to market returns.
This report is limited to the dissemination of general information pertaining to Convergence Investment Partners, LLC's services and general economic market conditions. The information contained herein is not intended to be personal legal or investment advice or a solicitation to buy or sell any security or engage in a particular investment strategy. Nothing herein should be relied upon as such. The views expressed are for informational purposes only and do not take into account any individual personal, financial, or tax considerations. There is no guarantee that any claims made will come to pass. The opinions and forecasts herein are based on information and sources of information deemed to be reliable, but Convergence Investment Partners, LLC does not warrant the accuracy of the information that this opinion and forecast is based upon. Opinions expressed are subject to change without notice. Past performance does not guarantee future results. You cannot invest directly in an index. Consult your financial professional before making any investment decision.
Convergence Investment Partners, LLC (“Convergence”) is an SEC registered investment adviser with its principal place of business in the State of Wisconsin. Convergence and its representatives are in compliance with the current registration and notice filing requirements imposed upon registered investment advisers by those states in which Convergence maintains clients. Convergence may only transact business in those states in which it is notice filed, or qualifies for an exemption or exclusion from notice filing requirements. Any subsequent, direct communication by Convergence with a prospective client shall be conducted by a representative that is either registered or qualifies for an exemption or exclusion from registration in the state where the prospective client resides. For additional information about Convergence, including fees and services, please contact Convergence or refer to the Investment Adviser Public Disclosure website (www.adviserinfo.sec.gov). Please read the disclosure statement carefully before you invest or send money.