The Biggest Myths in Investing, Part 10 – Forecasts are Useless

This is the tenth instalment of a ten part series similar to what I did with “The Biggest Myths in Economics”. Many of these will be familiar to regular readers, but I hope to consolidate them when I am done to make for easier reading. I hope you enjoy and please don’t forget to use the forum for feedback, questions, angry ranting or adding myths that you think are important. 

Smart asset allocation is really about establishing an intelligent set of probability distributions. When we build a diversified and operationally grounded portfolio we don’t have to be exactly right. Instead, what we largely avoid is being precisely wrong. This improves our odds of financial success by creating a high probability of an asymmetric outcome. As I’ve described before, great investors think in terms of probabilities. And while it’s become fashionable in recent years to shun forecasts and expert opinions, I believe this is a dangerous misunderstanding about how we should approach the process of portfolio construction.¹

As we learned in Myth 4, we’re all active investors allocating assets by making implicit forecasts about the future. For instance, a Vanguard Balanced Index investor is explicitly saying that a combo of stocks and bonds in a 60/40 allocation is likely to meet their financial goals. But the difference between this investor succeeding and failing comes down not only to understanding that this portfolio has performed well in the past, but also understanding why it might perform well in the future. This involves not only a reasonable forecast of its likely future returns, but also having the confidence that its underlying instruments are structured to generate those returns.²

The rise of technology and data driven information has allowed us to better test portfolio outcomes. But testing these outcomes is only half the battle. Using past evidence to forecast the future (called extrapolative expectations) is a useful means of portfolio testing, but does not necessarily mean we are utilizing an operationally sound methodology. In order for these implicit forecasts to have a high probability of success they should also be grounded in a sound operational understanding. An approach such as Bayes’ theorem is a sensible approach that involves the use of past information that is corrected to account for new information such as potential future changes. Such an approach does not solely rely on past data, but attempts to correct for future events that will alter past results.

When we think about the asset allocation process and creating high probability outcomes it’s useful to think in terms of operational realities. Building a portfolio isn’t exactly like engineering, but it has similarities. For instance, when we build a plane we engineer it so that it is designed to take advantage of certain operational realities. If we build the plane in a certain way to create a certain amount of thrust it will create lift which will create flight. Once an engineer has a sound operational theory for creating flight they can construct the plane and test it thereby giving it a historical track record of evidence supporting the underlying operational engineering. This is how we establish high probability forecasts about their performance.

The financial markets are not that dissimilar. By thinking of our process in terms of operational asset allocation we can create higher probability outcomes by undergoing the same process that the engineer does when building their plane. This is achieved via a rigorous two step process:

1. Understanding if this asset allocation is consistent with an operational understanding of how these instruments are structured and designed to perform.
2. Testing the structure to see if there is evidence supporting the underlying operational structure showing that these instruments actually perform the way we expect.

In order to better understand this process we can look at a simple example of stocks and bonds. In building our operational asset allocation of stocks and bonds we must first understand what these instruments are and how they are structured to perform:

• Stocks are a contractual obligation structured in such a way that they give the owner access to the residual profit earned by the underlying entity.
• Bonds are a contractual obligation structured in such a way that they give the owner access to what is usually a fixed income over a defined period of time.

Stocks earn what is called an equity risk premium because it does not make sense to own stocks if one can earn an equivalent return in a lower risk instrument like a bond. Stocks, therefore, are operationally structured to outperform bonds. When we study the evidence testing if this operational structure is supported by historical testing it confirms our underlying thesis that stocks are operationally designed to generate higher returns than bonds. This is an exceedingly simple example, but one could expand on this to better understand how any asset class fits into this methodology.³

Of course, nothing in life is guaranteed. Planes don’t always fly and stocks don’t always outperform bonds. But by taking this sort of operational approach to asset allocation we can improve our odds of financial success because we are making high probability forecasts about the future that are grounded in operational realities.

¹ – In fairness to the anti-forecasters, they are usually saying that short-term forecasts are worthless. This, I agree with. 

² – From a more specific financial planning perspective the forecasting of future returns is actually essential to good portfolio construction as a good planner will match likely future returns to meet necessary withdrawal rates.  

³ – For instance, if we wanted to dive further into fixed income we can conclude that all fixed income products are not created equal. Some, such as US government bonds, are unique within the scope of the global monetary system and serve as unique safehavens in a world where the US government is the largest income generating entity of them all.