Introduction
If you have studied automated advisor platforms, you have probably encountered commentary trumpeting the virtues of Modern Portfolio Theory (MPT) and such related verbiage as mean-variance optimization (MVO), efficient frontier, capital market line, and so forth. It is also likely that you encountered reference to a Nobel Prize, which was, indeed, awarded on the basis of the model about which they speak.
This protocol, standard fare in investment-related finance classes, was a bold contribution to our understanding of how return, risk and diversification work together to enhance investment outcomes. Much of what we understand today about managing return-risk tradeoff was inspired by that model (also referred to as the Markowitz model) as well as another Nobel Prize winning concept known as the Capital Asset Pricing Model. Today’s best practices in asset management would probably not be possible without the foundation of knowledge those sources provide.
That said, there is a big difference between a set of foundational theories and real-world implementation.
The idea of MPT is as profound as the Nobel Committee believed. A portfolio is not merely the sum of its constituent parts. The way those parts combine is also important. For example, a portfolio containing two extremely risky securities may itself carry very little risk if the securities, however volatile each may be on its own, move opposite that of the other. In other words, negative correlation may operate to offset extreme risk in each component.
For any collection of assets, MPT computes the overall return and risk of the portfolio (the combined package) based on the return and risk of each component and the way each component correlates to each other component. The end result is a set of possible combinations each of which is said to be “efficient;” in other words for each level of overall risk, MPT identifies the combination of assets (i.e. the weighting of each) that will produce the highest possible return. Or put another way, for each level of return, MPT will identify the combination that results in the lowest possible risk.
So there is no single perfect-best-optimal portfolio. Instead, we get a series of best-possible (efficient) portfolios one for each possible choice regarding risk. (If we plot all of these combinations on a graph and draw a line connecting them, that line is referred to as the “efficient frontier.”) The level of risk appropriate for a specific investor is identified and voila, MPT tells us the components and weights of the portfolio that is best for them. All the investor need do is establish their risk preference. In other words, the investor chooses the one portfolio from the efficient frontier that matches their risk profile.
That sounds wonderful, arguably irresistible. That’s the standard robo adviser marketing theme. Why try to have humans pick investments when it’s impossible for them to compete successfully against objective computer-generated “efficient” or optimal portfolios, which they supply.
Here’s the problem: It doesn’t work in the real world.
All of the inputs to the model (the data that must be fed into it) come from the past. Accordingly, we cannot naively assume any of it is representative of a genuine characteristic of the security during the period(s) studied, as opposed to a temporary aberration and even to the extent the historical observations were legitimate, there would be no reason to automatically presume they would continue that way into the future. Practitioners are, therefore, forced to massage the data with very active human judgments of the sort they claim should give way to automation. They do this by the historical time periods they choose to study and how they combine-weight different numbers and by out-and-out manual over-ride in some cases through use of something known as Black-Litterman, or similar protocols which were developed by academicians to address the problems found in real-world applications of MVO (where results were seen to have been highly sensitive even to modest errors in forecasting impossible-to-predict items).
And even with such corrections, we still need more. Bona fide MVO computations often lead to negative weights (i.e. short selling, sometimes aggressive short selling) and/or “dominance” by one asset which can overpower the others through an exorbitantly high weighting. So the ultimate fail-safe protocols are human-generated policy-based constraints; minimum (typically at or above zero) and maximum permissible weights that will be used whenever an MVO computation (with or without a corrective adjustment based on Black-Litterman, etc.) results in numbers outside the boundaries.
And by the way, nothing in MPT suggests which assets should be chosen to be subjected to the MVO computations. That is a purely human judgment.
Bottom line: There is no such thing as an objective optimal portfolio. There are:
- Human-generated portfolios in which the human roles are buried for purposes of marketing and instances of unsuccessful decision making are blamed on the model thus shielding the humans from accountability, or
- Human-generated portfolios in which the human roles are openly explained and for which humans are accountable
We reject the former and embrace the latter. Our human-generated allocations reflect the following core beliefs:
We believe a human brain is a terrible thing to waste and use what we have openly and proudly.
Instead of trying to summarize the views of passive indexing adherents, we’ll defer to the words of one of its leading practitioners, Vanguard. On page 14 of a document known as “Vanguard’s Principles For Investment Success,” that firm states as follows:
In practice, diversification is a rigorously tested application of common sense: Markets will often behave differently from each other—sometimes marginally, sometimes greatly—at any given time. Owning a portfolio with at least some exposure to many or all key market components ensures the investor of some participation in stronger areas while also mitigating the impact of weaker areas . . . .
Performance leadership is quick to change, and a portfolio that diversifies across markets is less vulnerable to the impact of significant swings in performance by any one segment. Investments that are concentrated or specialized, such as REITs, commodities, or emerging markets, also tend to be the most volatile. This is why we believe that most investors are best served by significant allocations to investments that represent broad markets such as U.S. stocks, U.S. bonds, international stocks, and international bonds.
Generally speaking, that is a valid point and it’s why we offer asset allocation portfolios. But elevating it to the status of dogma requires us to assume something we find objectionable: that human advisers are so incapable of rational thought as to justify the suppression of any attempt to exercise any judgment, no matter how clear or high-probability a scenario might be.
Consider the present and the question of interest rates, which has much to do with how investors are best served approaching the bond market. Benchmark rates remain extremely low even after having bounced up off recent near-zero levels (after a generation-long march downward). Are rational advisers allowed to assume that what we saw from the bond market in the past (returns boosted by asset values that rose as interest rates fell) cannot occur in the future and that reasonably probable bond-market scenarios range from neutral (interest rates move sideways – aside from modes ups and downs of interest mainly to short-term traders) to negative (interest rates rise). Other automated investment advisory platforms follow the above-quoted Vanguard approach, refuse to take any position on the future of the bond market, and passively invest clients’ money in a broad-based capitalization-weighted index of the entire bond market, including so much of the longest-term (most vulnerable to market risk) issues as happen to be required by passive capitalization weighting.
We do not follow suit. While we are not inclined to be the stereotypical active adviser who lauds its own unique genius or hunches, we consider ourselves obligated to act to address such clear-cut probabilities as must be recognized by any diligent, thoughtful observer. The other course, the one we reject, is best illustrated by the 1950 Kurt Vonnegut dystopian short story “Harrison Bergeron” from the author’s Welcome to the Monkey House collection. Consider the following passage from the third paragraph of the story:
Hazel couldn’t think about it very hard. Hazel had a perfectly average intelligence, which meant she couldn’t think about anything except in short bursts. And George, while his intelligence was way above normal, had a little mental handicap radio in his ear. He was required by law to wear it at all times. It was tuned to a government transmitter. Every twenty seconds or so, the transmitter would send out some sharp noise to keep people like George from taking unfair advantage of their brains.
This, essentially, is the sort of world to which advocates of dogmatic indexing aspire. The ruling authorities in Harrison Bergeron’s world thought it was unfair of people with above-average intelligence to benefit from their metal capabilities. Dogmatic indexers act in accordance with that view when they suggest that a money manager who thinks he or she has above average intelligence is probably deluded and that their clients would be better off if the manager was distracted by a metaphorical beep every 20 seconds to prevent them from trying to figure how to do something other than buy index funds.
We do not and will not wear “handicaps” and our analytic judgment is not compromised by politically-mandated beeps. We recognize what is unpredictable and do not try to predict where we know full well we can’t. However, when there are things we can and do know, if not to perfect certainty than at least with a very high degree of probability, we will not don mental handicap earpieces to obscure our perception of that which is reasonably clear.
This is why our fixed-income choices as of this writing emphasize ETFs with short- and intermediate-term targeted maturities.
Diversification is desirable but should not be carried out in a naive way.
If one knows the future with complete certainty, diversification would be improper: There would be no reason to do anything other than invest 100% of one’s funds into one’s favorite asset. That we don’t do this reflects the reality that we do not know the future.
Widespread suggestions that diversification is used to reduce correlation are oversimplified, perhaps dangerously so. Correlation is a statistical report card telling us what happens to have happened during the specific past period(s) that were studied. Whether it can be used to guide our decisions about investing with an eye toward the future depends entirely on judgment as to whether what we saw in the past will or will not persist. Also, correlation cannot be allowed to trump expected return and risk.
Consider, as an example, three assets, A, B, and C. Expected returns for each are 4%, 7% and minus 15% respectively. Assume, too, that risk in each case is appropriate in light of expected return.
Dogmatic diversification would require us to own all three assets.
That would be a legitimate approach if those expectations were merely being pulled out of thin air. There are, however, times when our ability to assume, albeit far from perfect, is sound enough to warrant use of judgment in developing portfolios. There is no magic formula for recognizing such situations. While we never want to overestimate the application of common sense and sound professional judgment, we do not believe investors are well served by those who deny, based on rigid policy, the possibility that anything can ever be known.
Given the expected return of the three hypothetical assets mentioned above, +4%, +7% and -15%, we consider it improper to force the -15% asset into the portfolio for no reason other than to satisfy a rigid policy requiring complete diversification.
But suppose we’re wrong about Asset C. Maybe it won’t return -15%. Maybe it will ultimately return -20%, -5%, -12%, etc. So how is the investor hurt by having acted on an erroneous forecast and eliminated C? Suppose, on the other hand, C returns +45%. Then a significant error would have occurred. But errors are inevitable when investing. If one isn’t willing to tolerate and be accountable for that, one should not be an investor. Competent diligent investing requires one to sometimes go out on a limb and say a return of +45% is very unlikely and that potential returns of -5% to -20% are much more probable.
One who cannot or will not do that should not be choosing investments for one’s self or others, and should certainly not hide behind naive gibberish relating to diversification.
Fixed Income as a whole offers prospects for return and risk that have been, are, and we expect will continue to be lower than that available for equities.
This is inherent in the nature of fixed income, where expected return derives from contractual obligation (regarding payment of interest and principal). Adverse developments can and do still occur but the probable range of bad outcomes is less than that applicable to stocks, where there are no such contractual obligations. Therefore, we believe that absent extreme degrees of risk taking or risk avoidance, all portfolios should reflect a combination of equity and fixed income with the weighting toward each being determined by the client’s risk profile.
Commodities and Real Estate, traditionally recognized as important asset classes also worthy of inclusion in a core portfolio are, we believe, less autonomous than historic data might suggest and hence can and should be subsumed by the broader equities asset class.
Low correlation (viz. equities) has tended to be the primary argument advanced by those who advocate autonomy for these asset classes.
We must note, however, that past outcomes do not assure future performance. Much historical data has covered periods during which global economies and interconnectedness were considerably less established than is now the case. There is much that can be debated regarding the newer markets, but some things are clear. They feature large populations that tend to be young. This predisposes them to particular growth scenarios. That will likely spur demand for “things,” which in turn is expected to drive demand for commodities (used to make or power things). We believe this will move to the forefront of commodities markets, to the detriment of speculator-driven dynamics. Given that, and complexities involving the structure of commodity markets (contango, backwardation, rollover, delivery, leverage), we believe the benefits of owning commodities can best be captured by ownership of equities, shares of companies that produce and distribute them, shares of companies that make and sell the things made using commodities, and others that relate to the economy that governs this entire eco-system. We do not think there is enough to be gained by zeroing in on the commodities themselves to offset the structural challenges posed by doing so.
Real estate too is driven by bona fide economic factors (the cost of debt, potential incomes of commercial tenants, potential purchasing power of home owners or renters which are driven by economic considerations, etc.). We believe these forces can also be harnessed by the equity markets as a whole including, where a special focus is desired, Real Estate Investment Trusts (REITS). Traditional arguments addressing non correlation refer often to direct ownership of real property; the proliferation of REITS has given most investors far easier, more liquid, ways to participate in this area. And better data and platforms provide better insights into the relationships between real estate and driving factors.
Equities should be weighted more heavily by those who are more willing to pursue higher potential returns — fixed income should be weighted by those who are more conservative and less willing to tolerate large losses and even willing to sacrifice opportunities to earn top returns in order to pursue relative stability.
This is a well-established set of notions that many are likely to find self-evident.
While we diversify geographically, the manner and extent to which we do so reflects what many refer to as a “home country bias.”
All investments bear basic risk. Investments outside the U.S. add completely separate classes of risk, including currency-fluctuation risk.
As is the case any time one considers taking on risk, such a course of action needs to be balanced against the potential rewards. We believe non-U.S. investments offer enough potential return to justify some measure of exposure but not nearly enough to justify going out of our way to act as if currency risk doesn't exist and to treat non-U.S. Investments the same way we treat U.S. investments. Because of this extra risk, and especially because it’s a risk we and all but a few dedicated specialists find it especially challenging to analyze, we believe each non-U.S. allocation should receive a lesser weight than it would if it were a domestic allocation of the same class.
This is not a carved-in-stone position. We study all kinds of non-U.S. investments including those that are currency hedged. What we say as of this writing could change in the future if we come to believe currency-hedged vehicles become more attractively priced relative to non-hedged vehicles, or if circumstances change to the point where we believe non-U.S. investments offer sufficient expected returns relative to domestic vehicles to justify the extra risk factor.
This is not to suggest we anticipate becoming currency-motivated traders. But we do need to acknowledge the potential for the situation to evolve such that the proper course of action becomes generally noticeable.
Emerging markets offer the potential for very high returns but also feature very high risk, which cannot be competently measured by mere statistical computations.
The potential for high returns comes from the nature of emerging, less mature, markets in general. Growth-development cycles are such that growth can be and often is more rapid when things are less developed.
That is a widely accepted notion. Less clear is the idea of risk.
It is obvious even to the most casual observer that the capacity for things to go wrong in emerging markets is great. All the risks faced by younger smaller companies in the U.S. are present. But there’s more. There is extra economic-instability risk; not only are those companies at risk due to the business cycle, they are also at risk due to the less established nature of the economic institutions and cultures within which they operate and the greater degree of uncertainty as to whether those economies will constructively adapt and correct. There is also greater political risk; government institutions in such countries tend to be less established and/or less committed to market-based economic policies that are so important to us as investors. Many also face greater military risks from outside and inside their borders. There is also more risk, involving the integrity of financial markets; the nature and effectiveness of regulatory oversight, the nature and effectiveness of accounting policies, standards, and compliance, etc.
In contrast to the approaches taken by other automated advisory platforms, we are not willing to boil all these risks down to numerical computations of standard deviation and correlation. Such numbers are simply statistical “report cards” summarizing what occurred over one or more specific periods of time in the past. They tell us nothing about other potential scenarios that could have (but did not) occur at those times, and more importantly, nothing about what can be expected in the future. We believe reliance on such flawed data (as we’ve seen in the case of one well-known “robo” firm that allocated 14% of its most conservative clients’ portfolios to emerging markets presumably — we have no choice but to assume —because of low historically computed correlation) is highly inappropriate.
We do not provide separate sets of investment recommendations depending on whether the account enjoys tax deferred status (such as with IRA accounts).
First things first: We do believe in use of tax deferred investment vehicles and encourage all to use them as best they can in light of their individual situations. Income received and capital gains will benefit from whatever advantages the law confers upon such accounts. Where we differ from some other robo advisors is our unwillingness to pretend we can effectively take personalized tax planning into account in the composition of the portfolios we recommend.
The nature of the information we have on users of our site and the relationships we have with our users does not lend itself to tax planning which, by its very nature, is highly individualized. Typically, robo advisors handle this in a simplistic manner. They recommend standard fixed income vehicles (ETFs) but switch all or part of their recommended fixed-income allocations to the iShares National Municipal Bond ETF (MUB) if the account is not tax deferred. Their idea is to sacrifice a bit of current yield (MUB usually yields less than taxable equivalent fixed income funds) believing the client’s current tax-free accumulation of income will result in an effective after-tax yield relative to what the client could have had through a higher taxable yield.
In a simple world, that would work. Actually, however, our world is highly complex. The difference between a pre-tax yield and an after-tax yield varies from one individual to the next and is best evaluated in consultation with a tax planner or a detailed one-on-one interaction with a skilled financial planner or adviser (i.e. someone who can and will render far more sophisticated service than what one might expect from a telephone support rep). One layer of complexity is introduced by the obvious reality that there are multiple tax brackets (i.e., percentages of taxable income that are paid in taxes). That could in theory be manageable assuming the client discloses his or her income to the adviser. Actually, though, that’s not the end. It’s barely a scratching of the surface.
We can’t just accept the supposedly standard federal rates. We also have to consider the state and/or city in which the client lives or is otherwise taxable. These local rates can vary quite a bit and depending on which ones apply to a client, MUB may be a much less effective choice as compared to other ETFs or mutual funds that aim for deferral of local taxes as well as federal levies.
More perplexing, however, is computation of taxable income, the level of income to which applicable tax rates apply. What exemptions are available to the client? What deductions? What credits (these are setoffs against taxes rather than taxable income). The 2017 tax overhaul added complexity since it now becomes more important to weigh the client’s situation with state and local taxes (the deductibility thereof against federal taxes), mortgage interest, standard deduction and so forth. And even for those who deeply understand their own tax profiles, there’s more to consider than just applying a tax rate to a yield. MUB is supposed to represent the entirely municipal market and, hence includes a passive rather than thoughtful mixture of maturities and credit risks.
Also, do not give short shrift to questions of cash receipts risk. Unlike the federal government, municipalities cannot “print” money and in contrast to businesses, their ability to earn what they need to service their debt is limited at best. Many localities are already struggling with financial, burdens and changes in federal funding policies may bring more into jeopardy going forward.
We don’t suggest the municipal market as a whole is doomed, or even undesirable. We do, however, suggest that it’s not appropriate for naive simplistic passive solutions. Investors for whom professional tax planning can make a difference should get the advice they need from qualified sources, and that, in our opinion, excludes robo advisers.