Market Analysis

The Forecasting Abacus’ Flaws

What should one take from mathematical attempts at forecasting?

In the past few weeks, two media themes on markets have been relatively prevalent. First, that currently super-low government bond yields—particularly in the US—combined with recently expansionary economic data are a sure-fire sign a bond bear market looms. Second, many presume super-low Treasury yields foretell lower-than-average returns for stocks. While there may be some merit to aspects of these arguments, in our view, serious flaws exist with the rationale.

In many cases, it’s true the government bond market doesn’t look very attractive to a long-term investor today. Factoring in inflation (core CPI of 2.2%), yields on 10-year US Treasurys today (also about 2.2%) are basically nil. And to be sure, bond prices and yields move in opposing directions—rising yields depress prices (and vice versa). That Treasurys yield so little today could imply prices are high and set to fall if yields rise.

But note: This isn’t a timing tool, it’s really just a numerical reflection of where rates are today. When and how rates move—and to what extent different maturities, durations and types of bonds experience those moves—matter quite a lot. Even with expansionary US data of late, there’s little suggesting a material surge in long-term government bond yields or the Fed hiking the target rate is very likely in the here and now. And in fact, the last few years illustrate this nearly perfectly: Economic data, as is normal, have periodically accelerated and decelerated, yet no massive Treasury bear arrived in 2010 or 2011. Could it happen this year? Maybe, maybe not.

The second widely theorized point—stocks are likely in for a period of low returns ahead—is often similarly underpinned with low government bond yields as a key driver. Some cite fancy terminology for this, like the equity-risk premium (ERP), implying that stocks typically return a certain degree higher than what’s widely considered “a risk-free rate”—the 10-year Treasury rate. Since Treasury rates are far lower than their historical average, practitioners posit equity returns will be below average, too.

Now, we have myriad issues with this theory—the major one being the validity of ERPs in forecasting the far distant future. In the long run, shifts in equity supply are the primary determinant of equity returns, which ERPs make no attempt to foretell. What’s more, stocks’ long-run annualized return of about 10% (a figure that includes bear markets) is, after all, only an average. In bull markets, average annualized returns are more like 21%. Yet in the current bull market, global equity returns have annualized more than 25% since March 9, 2009. All the while, Treasury yields have been ultra-low. So it truly wouldn’t seem Treasury yields were very predictive of the three-year results in the present market cycle—no small issue for adherents to the theory.

More confusing yet, some argue for both dampened equity returns and a bond bear—seemingly an internally contradictory forecast, considering the former presumes low Treasury rates, while the latter, rising.

The broader picture, for fans of mathematical market forecasting, is similarly inconvenient: There is no perfect or even better-than-average forecasting calculation. This is one lesson of 1998’s Long-Term Capital Management failure—however fancy the math, if several Nobel laureates couldn’t determine a calculation that worked with a modicum of repeatability, how likely is it anyone does?

Forecasting is about much more than math. Yes, part of it is data—collection and interpretation. There, no doubt, math can help. But significantly, interpreting data and how they relate to expectations and sentiment, political factors, monetary policy and much more, aren’t likely to be easily distilled using an excel sheet, calculator or super computer. While anything from ERPs to Shiller’s CAPE to Tobin’s Q may occasionally have value, the chances a strategy hinging exclusively on such formulae or indicators generates repeatable, long-term success are low. Very low. Many mathematical indicators occasionally have value, but they should come with a warning label: This indicator or calculation is no forecasting holy grail. Use it exclusively at your own risk.

If you would like to contact the editors responsible for this article, please click here.

*The content contained in this article represents only the opinions and viewpoints of the Fisher Investments editorial staff.