Bond yields are highly correlated to future returns. In fact, if you select a specific maturity from a sound issuer, there is little doubt what the return will be. Between now and 2025, you don’t know what the bond will be worth on any future date, but if you buy a five-year Treasury today, in five years you’ll get exactly what you signed up for, roughly 1.63% as of this writing.
Equity returns are tougher. There aren’t any great predictors of future returns. The one valuation measure that has received the most attention for its predictive value is Shiller’s P/E 10 ratio. But the P/E 10, while it has more explanatory power than other valuation measures, is not a great predictor.
A Vanguard study, “Forecasting Stock Returns: What Signals Matter, And What Do They Say Now?” found that P/E ratios only explained about 40% of future returns from 1926 to 2011. The Shiller P/E 10 ratio had an R squared of 0.43 to subsequent 10-year returns, a slight improvement from the one-year P/E ratio, which had an R squared of 0.38.
If you are looking at valuation levels as a guide before moving in or out of markets, you are likely to be disappointed. The best predictor of one-year returns had an R squared of a mere 0.12. As the study pointed out, “Stock returns are essentially unpredictable at short horizons,” Vanguard wrote. “Quite frankly, this lack of predictability is not surprising given the poor track record of market-timing and related tactical asset allocation strategies.”
This also means that clients who read about high valuations and turn those statistics into an anxiety-riddled fear of a crash are likely working themselves up for no reason. A good financial planner may be able to bring some perspective to the table. The Vanguard study was written in 2012. Valuations at that time were high and yields were low, yet returns have been good for equity investors and more than adequate for reasonably constructed balanced portfolios.
The last decade or so corroborates the main conclusion of the Vanguard study: “Expected stock returns are best stated in a probabilistic framework, not as a ‘point forecast,’” and should not be forecast over short horizons.
That gets us back to Monte Carlo simulations. If we are going to use estimates of future returns that are less favorable than the historic record, what do we use? The most popular method on the discussion board was to simply reduce the mean by a couple of percent or so. This method was likely the top choice because it was a choice that could be made given the software’s limitations.
I have read a few studies that use different approaches. In one (made by Morningstar’s David Blanchett), low yields were used for the first part of simulated retirements to reflect today’s low-yield environment. Instead of assuming that would persist, latter portions of the retirement horizons used yields more in line with historic yield levels.
But these are issues of how. The question that sparked the online debate was, “Should we use historic returns or our estimates of future returns when running a simulation?” Maybe we should do both. Run one with history-based parameters and one with a reasonably thought out good faith estimate. The whole point of Monte Carlo is that it should show a range of outcomes, not serve as some sort of crystal ball.
One knock on using historic returns is that they can give clients a false sense of security. Regardless of whether you use one of the methods I just mentioned or some other approach that uses below-historic average returns, the simulations will show more failures. This can lead clients into a false sense of dread. Use both and clients may get the idea that the future is full of uncertainty. That’s a good thing for them to believe because it’s true.