BUNK 20

EQUITY RISK PREMIUMS—FORECASTING FUTURE RETURNS WITH EASE

Wish you could know where stocks will be in 10 years? Me too! I don’t think it’s possible. Still, people try. There’s a notion in academia about the equity risk premium (ERP). The ERP is, literally, the premium you get, expressed as a percentage over some supposed risk-free rate—like the 10-year US Treasury—from holding stocks. Some folks use the T-bill rate. Either way—same basic concept.

And there’s nothing wrong with that! It’s true—most often investors are rewarded long-term for taking extra volatility risk (done right). Since 1926, the average annualized ERP (using 10-year Treasuries) has been 4.4 percent—Treasuries have annualized 5.3 percent, and the S&P 500 9.7 percent¹—a huge spread! And theoretically, investors should be rewarded for suffering through stock market swings. (Although, they hate the volatility. And the more the volatility, the more they hate it when they rationally should love it since, in the long run, they usually get paid handsomely for it.) If you weren’t likely to get higher reward for higher risk, why would anyone want the higher risk—whether measured by volatility or otherwise?

The problem with ERPs is some academics try to model future ERPs—predicting future stock returns. Bunk. I’ve never seen any ERP model stand up to historical back-testing. Not one! Yet, every year, we get a new wave of them.

Long-Distant Future Supply

When I say future, I mean most ERPs attempt forecasting far into the future—usually 7 to 10 years (10 is most common). Yet, you know from Bunk 10 that stock returns in the near term—over the next 12 to 24 months—are driven mostly by shifts in demand. And even those are devilishly difficult to forecast. And further out, supply pressures swamp all. And there is absolutely no way—none—to predict stock market direction 7 or 10 years out unless you can somehow predict future stock supply shifts. Yet I’ve never seen anyone even attempt that, nor can I fathom a way to do it myself. (Though maybe one day someone can do it—and then a future ERP might be possible.) But not a single ERP model I’ve ever seen has even addressed the issue of predicting long-term supply flows. And if you can’t address future supply, your model is worthless. Because, in the long term with securities, supply is all that matters.

Instead, most ERP models make forward-looking assumptions based on cobbled-together current or even past conditions. Except, right away you know you can’t make 10-year projections based on current or past conditions. Past performance is never, by itself, indicative of future results.

An example of an ERP model might be some nonsense like: Take the current dividend yield, the average earnings per share over the last 10 years, plus the current inflation rate, and subtract the bond yield. Add or subtract a few components. Mix that together with a guesstimate for some percent stocks are supposed to beat bonds by over the next 10 years, based on what Treasuries are yielding now.

Except what does today’s dividend yield, inflation yield, earnings, or anything have to say about what will happen 7 or 10 years from now? Or even three years? No one can answer that.

Plus, none of these ERPs stand up to historical back-testing that I’ve seen. A scientist would say you must test historically—picking multiple past periods and inserting historic data for the ERP assumptions. A model that “worked” would spit out the actual forward 7- or 10-year returns from that past period pretty darn closely. One that didn’t would be broken—no exceptions. And none do, or if they do it’s merely accidental—they work for a couple of past periods, but not consistently.

Academics who are prone to bearishness—surprise!—produce bearish ERPs, and bearishly biased press promotes them. They say, “The ERP will be below average for the next 10 years, just 1.5 percent!” If the 10-year Treasury is now 3.0 percent, for example, that means stocks will average just 4.5 percent over the next 10 years. Yuck! On the flip side, bullish academics (who are fewer in my observation) produce bullish ERPs—they bake in their own biases. Still, bullish or bearish, all ERP projections are as much bunk as anyone else’s long-term forecasts—bias-based guesstimates, nothing more.

We could dissect any number of popular ERP models. And it’s a worthwhile exercise to do on your own (straight debunkery)—but the alpha and the omega and the fatty acid reality is they don’t address future supply, so creating an ERP is a fruitless fatheaded academic flexercise. (Flexercise is technically a comberation—see this book’s Introduction.)

Meager Assumptions

Another ERP red flag. A huge one! ERP models usually predict 2 percent or 2.5 percent or 3 percent—some small number. Why academics fiddle-faddle over 2 percent versus 3 percent is beyond me. They can easily check history. Looking backward, ERPs are very wildly variable. After all, normal stock returns are extreme—not average. (See Bunk 5.) Table 20.1 shows historic ERPs by decade. The 1960s and 1980s ERPs were darn close to the long-term average ERP of 4.4 percent. Amazing! Other than that, they’ve been extreme—18.9 percent in the 1950s. And 10.2 percent in the 1990s.

There have been negative ERPs—the 1930s, of course. The ERP was just flat in the 1970s, while stocks overall were positive (though below average). And the 2000s were another. (Without forecasting long-distance stock returns, see how long ERPs were hugely positive following past negative periods. Folks who say that the 2000s being overall down for stocks means we’re in some long-term “new normal” of malaise don’t know history. Such a thing has never happened in the past.)

Table 20.1 Historic Equity Risk Premiums by Decade—Not Predictive

Source: Global Financial Data, Inc., USA 10-Year Government Bond Total Return Index, S&P 500 total return from 12/31/1929 to 12/31/2009

Simply, academic ERPs are usually too bearish, don’t address past wide variability, don’t stand up to back-testing, and can’t address future stock supply shifts. Yet, academics still produce them, the press promotes them, and the investing world laps them up—because they sound quantitative, academic, sophisticated, and rigorous! Wall Street “wisdom” at its finest. Near as I can tell, sophisticated has screwed up more investors than bubbles have.

Typically, an academic will pontificate about the multiple complex variables in his ERP and why they combined with his formulaic approach, leading to a vision of the future. (Sophisticated!) Few will say, “My ERP model is a fancy-dancy but useless way to express my basic optimism or pessimism about the next 10 years.” Yet, ironically, that’s what it really is more often than not.

ERP models almost never predict it right. Maybe academics think overt “optimism” or “pessimism” is unseemly—unprofessorial. But it also runs contrary to empirical evidence. Stocks historically rise more than fall. Stocks historically do pretty well in the long term compared to cash or bonds, but in a widely varying path. That’s the truth. That’s hard for most folks to get in their bones.

Maybe you can use what you’ve learned in this book (or maybe read my 2006 book, The Only Three Questions That Count) and do a debunkery that uncovers how to do long-term stock supply forecasting. Then you might be able to build a credible, forward-looking ERP! That would give you huge power (at least until the rest of the world figured it out). I’d love to be able to do it, and I’d love for you to be able to do it too. But until then, don’t bother with ERPs.