We have accepted certain principles over time. If one feels there is a mispricing within a security or an asset, then they could gently exploit such difference through a dynamic asset allocation model; one that concurrently shifts all asset weights, based upon the new optimization.
Where do these “mispricing” ideas come from? Maybe from a fundamental research analyst seeing what appears to be misvaluation, or from a technician studying the general levels across many asset classes. But what gets lost through history is how difficult it is to own quite-differentiated and statistically significant signals, and how dramatically pricey it is to execute on most that a portfolio manager might come across. Historical proxies also rarely reveal much about the future, other than by chance.
This is why most active investors are — despite a flood of modern resources — still unable to outperform their benchmarks by a high enough amount to warrant skill. Never mind the add-on gravitational forces of paying advisory fees and taxes, expensive investment vehicles and other inefficiencies. In a crowded field of investors, asset prices are just too efficient to exploit.
It's also easy to see people fool themselves with randomness. This randomness of prices also eclipses the underlying signal most analysts can really have on their asset class view. On a macro view, we see across hundreds of years of some financial data that we can probabilistically partition into only a handful of super-regimes: decade(s)-long cycles that can occur among asset classes. And no one ever knows a priori which strategy will dominate in the future, thus rendering discomfort in tilting one's portfolio for the next decade, into specific equity or into fixed income, or into momentum or mean-reversion styles. Beyond getting lucky with a single asset class, one still wouldn't know the other components of investment random error, such as how other asset classes would correlate to it. Correlations can always change and problematically weigh against long-term performance.
Let's take an example from psychology and probability theory to demonstrate the values we might conclude from a small data sample. Say that a triple coin flip results in the same face (e.g., three heads). This might be interesting, although that streak happens quite a bit. We now psychologically hinge off of this mundane result, and we model a high-return streak in, say, bonds to be represented by this coin flip series. Some will argue that more heads will now be flipped (greater strength in bonds), while others will suggest that tails will subsequently dominate to balance things (less strength in bonds). Over the next triple coin flips, there is also a 100% chance that one will see either two heads or two tails or a more extreme streak. So we're always validating the rationale for half the analysts. And meanwhile it was essentially random to begin with.
In a recent article, we showed a series of investment charts, asking readers to suggest the best one in which to invest, using styles such as momentum or mean-reversion. Similar to coin flips, people were outright adamant in their conclusions. Investment professionals often sit on opposing sides of any specific investment market theme (but for the rare times they incorrectly crowd into a bad idea). The cumulative friction of constantly debating and changing portfolio weights is an inefficient way of market timing, which never works over the long run and certainly is troublesome for an industry aggregate.
Our mathematical research showed the investment rules people followed were only “signs” with symmetrical performance conclusions. The same investment decisions were being made, even as subsequent market changes were random. This idea is critical in probability theory, evidencing why —over hundreds of years — investment styles doggedly and unpredictably come and go. Most can't predict this nor the correlations (including during financial stress); it is simply outside the scope of most fund manager or consultant conversations. Even with ex ante, equal investment statistics, different discretionary asset allocation decisions en route lead to obvious mistimings, and as we'll see, worse performance outcomes.