A Global Passive Benchmark with ETFs and Factor Tilts

In Miscellaneous on

“One way to test our thinking would be to ask the question in reverse: If your index manager reliably delivered the full market return with no more than market risk for a fee of just 5 bps, would you be willing to switch to active performance managers who charge exponentially more and produce unpredictably varying results, falling short of their chosen benchmarks nearly twice as often as they outperform—and when they fall short, losing 50% more than they gain when they outperform? The question answers itself.” – Charles Ellis, “The Rise and Fall of Performance Investing

Passive Aggressive

In a recent article published in Financial Analysts Journal, Charles Ellis makes an excellent case for the death of active management. Ellis asserts that the efficiency of a market is a function of the number and quality of active, informed investors at work in the market at any time. As more investors with increasingly deep educational backgrounds armed with mountains of data and obscene amounts of computational horsepower enter the market seeking inefficiencies, they will eventually eliminate all of the inefficiencies they so diligently pursue.

Plenty of literature supports this view. Ellis himself cites a seminal study by Fama which concluded that,

“Active management in aggregate is a zero-sum game—before costs. . . . After costs, only the top 3% of managers produce a return that indicates they have sufficient skill to just cover their costs, which means that going for- ward, and despite extraordinary past returns, even the top performers are expected to be only about as good as a low-cost passive index fund. The other 97% can be expected to do worse.”

Two recent studies (here and here) by Blake et. al. sponsored by the Pensions Institute at Cass Business School in London further bolster the results from Fama.  They applied a more rigorous methodology called bootstrapping, which allowed the authors to compare actual mutual fund returns to a distribution of returns which might have been expected purely as a result of random chance. Their results are in Figure 1.

Figure 1.


To interpret this chart note the green and blue curves. The blue curve charts the results of the robust bootstrap test, and describes the distribution of returns that would be expected purely due to random chance. The green curve describes the observed distribution of results for mutual funds which were active during the entire period 1998 – 2008. The blue dotted vertical lines bookend the 5th and 95th percentile performance (measured as the t-score of alpha) which might have been expected from random chance. Note that the green line is to the left of the blue line over the entire distribution, leading Blake to remark:

“…there is no evidence that even the best performing mutual fund managers can beat the benchmark when allowance is made for the costs of fund management.”

In fact, the authors conclude that on average, investors would accrue an extra 1.44% per year in alpha from investing in passive benchmarks. We would encourage more technical readers to refer to section 2.2 in Blake for a more detailed explanation of the bootstrap methodology.

Interestingly, the authors also studied the impact of mutual fund size on performance, and found that smaller funds outperform larger funds. In fact, this is a very economically significant effect. Specifically, Blake et. al found that a doubling in fund assets results in an average 0.9% per year reduction in fund alpha.

Let’s think about these two facts for a second. First, there is no evidence that any mutual fund managers outperform after accounting for fees and luck effects. Second, larger funds lag smaller funds. How might this help to explain the chronic and egregious underperformance of private investors described by perennial Dalbar studies, per Figure 2 (note: red bar is average private investor returns)?


Certainly there are many factors that have contributed to this dismal reality, such as performance chasing behaviour, poor advice, and emotionally driven decision making. That said, retail investors very often gravitate toward, or are directed into, behemoth funds operated by large, well-known investment firms. Perhaps investors (and Advisors) feel that a large institution with a long history is more likely to have investors’ best interests at heart. Almost certainly there is a feeling of ‘safety in numbers’; as Keynes famously said, “Worldly wisdom teaches that it is better for reputation to fail conventionally than to succeed unconventionally”. The sad reality, however, is that most investors, chasing the wrong kinds of funds on the basis of precisely wrong evaluation methods, will continue to fall far short of their goals.

But I digress. The point is, Ellis claims active management is a mug’s game, and the research strongly supports this view. And this fact is complicated further by that fact that, while some managers will inevitably outperform in any given period purely as a result of good luck, it is virtually impossible to identify these managers in advance. Worse, traditional methods of selecting managers based on 3 to 5 year track records are a near certain recipe for disaster. Figure 3 describes the proportion of institutions which evaluate and terminate managers at various horizons. Observe that while most institutions evaluate managers on a quarterly basis, they base termination decisions on 3 to 5 year evaluation periods. Yet, as Figure 4. makes clear, managers that are fired, presumably because of poor 3 to 5 year performance, go on to outperform replacement managers over the next 1, 2, and 3 year periods.

Figure 3. Proportion of institutions that evaluate and terminate managers at various horizons.

Source: Employee Benefit Research Institute

Figure 4. Excess returns to terminated and newly hired managers in the 3 years prior to, and subsequent to, termination

Source: Goyal and Wahal, 2008

Whatever method these institutions – and their consultant advisors – are using to evaluate, terminate and hire managers, it doesn’t appear to work very well on a 3 to 5 year evaluation period. Here we have a situation where the vast majority of active managers underperform, exacerbated by the fact that the managers who are expected to outperform typically go on to underperform the managers who are expected to underperform. Quite a conundrum.

As I was writing this section, a new paper from Vanguard hit my inbox which further bolsters the point that chasing top performing managers is a surefire way to underperform. Figure 5 from the paper compares the results of two simulations for each traditional mutual fund ‘style box’. First, the authors randomly selected each year from 2004 to 2013 a portfolio of funds from the universe of funds in each style box. They performed this procedure many times to generate a distribution of performance across all possible portfolios during the period. Next they simulated a ‘performance chasing’ portfolio by randomly selecting from only the top performing funds over the previous three year period. They chose this evaluation horizon because this is the typical mutual fund holding period.

Figure 5. Distribution of returns for all funds vs. performance chasing strategy by style box, 2004 – 2013


It’s easy to see that, in every style box, top mutual funds by three year returns underperformed the average mutual fund by a wide margin: about 12 Sharpe points on average. That’s a lot of Sharpe points when average Sharpe is about 0.4. In the context of these seemingly insurmountable hurdles for active management Ellis advises that, “…investors would benefit by switching from active performance investing to low-cost indexing.” It’s tough to argue with this conclusion. Unfortunately however, this raises as many questions as it answers.

So Now What?

While Ellis’ prescription to eschew active management for low-cost indexing appears to solve some important problems, his article falls remarkably short on how to implement such an approach. He seems to favour low-cost Exchange Traded Funds as the most appropriate instruments to gain exposure to passive returns. However, the reader is left to determine how best to assemble such instruments to meet client goals.

I sought answers in the 6th edition of Ellis’ book, Winning the Loser’s Game, which has an introduction from none other than Yale CIO legend David Swensen, and echoes many of the themes David has trumpeted over the years. This is unsurprising because Charles served on the Yale endowment board for many years alongside David.

After a thorough read, I was still flummoxed. Ellis cites a great deal of data on the long-run performance of passive strategies, and even more data on the failure of active management, but he offers no meaningful prescriptions for implementation. Instead, he implores investors to get educated about estate planning and the fundamentals of asset allocation, and to take charge of their own affairs. This is undoubtedly excellent advice.

Advisors can play a role in what Ellis calls, “values discovery”, which is, “the process of determining each client’s realistic objectives with respect to various factors—including wealth, income, time horizon, age, obligations and responsibilities, investment knowledge, and personal financial history—and designing the appropriate strategy.”

Again, we support this conclusion, and Advisors do not always take this part of their role as seriously as they should. Certainly, each client should be thoroughly advised in the context of their objectives and constraints. But it is not obvious how to link Ellis’ vision of a purely passive approach to the idea of custom advice, and commensurately a custom asset allocation. Our inclination would be to invoke the Capital Market Line which would acknowledge the existence of one optimal portfolio, where risk is scaled up and down by introducing cash or leverage.

The vast majority of investable assets for both private individuals and institutions is ‘long-term money’, with a time horizon in excess of five years. This kind of capital will generally benefit from full exposure to a diversified portfolio of risky assets in order to maximize the opportunity for excess returns above what might be earned from cash. The question is, what might this portfolio look like?

The Only True Passive Benchmark: The Global Market Portfolio

In 1964, Bill Sharpe demonstrated that, at equilibrium, the portfolio which promises the greatest excess return per unit of risk is the Global Market Portfolio, which is composed of all risky assets in proportion to their market capitalization. Many investors will be familiar with this concept from their experience with market cap weighted indexes like the S&P 500. These are the ultimate passive investments within an asset class. However, it is not as obvious how to apply this concept across asset classes.

Importantly, since the Global Market Portfolio represents the aggregate holdings of all investors, it is the only true passive strategy. It is also the truest expression of faith in efficient markets. All other portfolios, including the ubiquitous 60/40 ‘balanced’ portfolio of (mostly domestic) stocks and bonds, represent very substantial active bets relative to this global passive benchmark.

Doeswijk et. al. recently published a paper on the evolution of the global multi-asset portfolio, where they examined the relative dollar proportions of all financial assets around the world from 1959 through 2012. There were roughly $90.6 trillion in tradeable financial assets globally as of the end of 2012, divided up as described in Figure 6.

Figure 6. The Global Market Portfolio, 2012



Source: : Doeswijk, Ronald Q. and Lam, Trevin W. and Swinkels, Laurens, The Global Multi-Asset Market Portfolio 1959-2012 (January 2014). Financial Analysts Journal

You will note that bonds represent about 55% of total financial assets while equity-like assets represent 45%. It’s well documented that Private Equity is just equity and real estate with a lag factor; furthermore, unless you are an Ivy League school endowment, or a member of the global elite, you don’t have access to quality private equity, so you might as well assume it doesn’t exist. We also wondered whether the authors include infrastructure investments under equity, and whether there is a place for commodities, though they aren’t strictly a financial asset. But in our opinion, this framework is 99% complete.

An ETF Proxy Global Liquid Market Portfolio

The proliferation of ultra low-cost index tracking mutual funds and Exchange Traded Funds (ETFs) makes it easier than ever for private and institutional investors alike to express a global passive bet via the Global Market Portfolio. Figure 7. illustrates our best effort at recreating the proportional exposures described in Figure 6 with liquid ETFs.

Figure 7. The Global Market Portfolio, 2012 ETFs


It should be simple to link the allocations in Figure 7 with the allocations in Figure 6. The one exception relates to Private Equity, which we have subsumed into roughly equal allocations to equities and real estate. Note that the total annual Management Expense Ratio (MER) for this portfolio on a weighted average basis is under 30 basis points, or 0.3%, and ETF MERs are dropping all the time.

It’s interesting to note that this portfolio requires no rebalancing because the weights will drift according to the relative performance of each asset class. However, a passive investment in these ETFs will not account for relative issuance and retirement of securities. This has a large impact on weights over longer periods, so investors will need to consult the literature periodically to ensure weights are still aligned. That said, this portfolio has the lowest theoretical turnover of any portfolio.

While the Global Market Portfolio is the only true passive benchmark, there are some simple ways to improve on the concept without introducing traditional forms of active management.

An ETF Proxy Global Market Portfolio with Factor Tilts

Even the most ardent believers in efficient markets acknowledge the existence of persistent risk factors which give rise to returns in excess of what is achievable from a purely market capitalization based benchmark. While enthusiastic finance PhDs and practitioners have identified hundreds of possible equity anomalies, only three stand up to rigorous statistical scrutiny (see here and here): value, momentum, and low beta (or low volatility) [Note: the illiquidity premium is also significant, but for obvious reasons is not very investable.] The so-called SMB or ‘size’ premium was discredited many years ago for US stocks (see here), and no evidence exists for this anomaly outside US stocks (see here). That said, small-cap value shows enduring promise.

Table 1. Historical Equity Factor Premia


Table 1 from Robeco shows the historical returns to these equity market factor premiums. A statistically significant anomaly might be expected to deliver 2 or 3% alpha per year; given that 30% of the portfolio is exposed to factor tilts, investors might expect 0.6 – 0.9% per year in excess returns. The MER of the portfolio is 0.35%, so this would essentially cover fees, plus a little extra. Furthermore, the diversification properties among the assets and factors might be expected to lower volatility by 0.25% to 0.5%, so the boost to risk adjusted performance from this portfolio could be meaningful, at least in the context of a passive framework.

To our knowledge, bond anomalies are fewer in number, and only two types of risk offer persistent excess returns: duration and credit. Duration risk is simply the risk of lending money at a fixed rate for a longer period, and the empirical evidence is weak for any material premium above maturities of about 10 years. Rather, the best we can say is that longer duration bonds outperform during declining inflation regimes while shorter duration bonds outperform during rising inflation regimes. Hardly a consistent anomaly. Credit risk is the return that investors demand in order to be compensated for the risk of bond default. After accounting for default risk and recoveries, the only credit spread with a significant positive risk premium is the BBB-AAA spread, also called the ‘Crossover premium’.

It is a relatively simple task to assemble the equity factor exposures to approximate the market-cap and geographical distribution of the global market portfolio. Figure 8 is an attempt to do just that.

Figure 8. ETF Proxy Global Market Portfolio with Factor Tilts

Global Market Factor Tilt Portfolioa

At the margin, it would be advantageous to hold a diversified exposure to commodities. However, there is little evidence that commodities exhibit a positive risk premium over the long-term. Rather than passive commodity exposure, sophisticated investors might contemplate a 5% strategic investment in CTAs. These funds have positive expectancy, largely because they harness the momentum factor across assets, but their real strength is structural diversification. This class of investment is really the only alternative asset class (except short equity) with persistent negligible correlation to equities. They also tend to deliver their strongest performance during equity bear markets, making them a compelling tail hedge.

The Global Market Model would almost certainly be further improved by the introduction of systematic factor exposures across asset classes as well as within them, as part of a Global Tactical Asset Allocation overlay. For example, there are well documented value and momentum factors which might be systematically applied as a portable alpha strategy to improve absolute and risk-adjusted returns, as described in Table 2 from Asness, Moskowitz and Pedersen (2013) (see also here). The statistical significance of these systematic tactical alpha premiums is actually higher than what is observed among analogous equity factors, so if you acknowledge one there is no logical reason why you wouldn’t adopt both.

Table 2. Global Tactical Asset Allocation Momentum and Value Return Premia


Source: Asness, Moskowitz, and Pedersen: Value and Momentum Everywhere (2013)

Table 2 illustrates that simple systematic exposures to momentum and value factors across asset classes have delivered 2.6% and 2.9% annualized alphas (t-scores > 3), respectively  over the past 40 years. Furthermore, these factors are excellent mutual diversifiers at the portfolio level, offering the opportunity to further lower aggregate risk. There is little doubt that institutions and private investors alike would benefit from these kinds of tactical alpha overlays, especially in today’s low-yield environment.

In summary, investors are starting to acknowledge the overwhelming evidence that active security selection is a loser’s game. This realization has caused a massive exodus from traditional mutual funds and Separately Managed Accounts and into passive Exchange Traded Funds. Investors who choose to follow this trend face a new set of challenges related to the expression of a passive view in their asset allocation. The Global Market Portfolio represents the most coherent expression of this view, and any deviation from this portfolio represents an active bet. Thus most investors who think they are passive are actually active; worse, they are making large concentrated bets unintentionally.

A thoughtful conception of the Global Market Portfolio would seek ways to gain exposure to the most persistent systematic market anomalies, while preserving the core capitalization and geographic exposures of the original model. Excess returns from factor exposures might net investors an extra 0.25% to 0.5% per year, with slightly lower risk.

In our opinion, the Global Market Portfolio with Factor Tilts represents the ultimate passive policy portfolio benchmark for institutions and private investors alike, as it represents the average expectations of all participants in markets. It should be the starting point for most long-term investment policies, and investors should thoroughly question the merits of any deviation in the absence of a carefully scrutinized and statistically significant long-term edge.