What Equity Risk Premium?

One of the fundamental tenets of finance is that stocks do better than bonds over the long haul. The difference is known as the equity risk premium. In other words, it’s the amount that investors are paid to take on the extra risk of owning stocks. (Small but important note: the equity risk premium most often refers to the gain stocks have over short-term T-bills, in this article I’m referring to the gain stocks have over long-term government and corporate bonds.)

The reason for the equity risk premium has puzzled economists for a long time. In fact, Jim Glassman and Kevin Hassett went so far as to say that it shouldn’t exist, and that’s how they got their Dow 36,000 hypothesis. I did some data-crunching today to add in the market’s recent performance and found that there really hasn’t been much of a premium for a long time. So were Glassman and Hassett correct in their theory except they wrote the book 20 years too late?? (Well, no…but I’ll get to that).

The best source for long-term investment information is Ibbotson Associates, now a part of Morningstar. Each year, Ibbotson releases its yearbook for historical returns of stocks, bonds, bills and inflation going back to 1926. I often refer to their work on this blog.

From the end of 1968 to the end of 2007, stocks’ advantage over bonds has been quite modest. Over 39 years, stocks have basically doubled both Treasuries and corporates. Doubling in 39 years may sound nice, but it really isn’t that impressive. It works out to about 1.85% a year for corporate bonds and 1.89% for government bonds. Given how much more volatile stocks are, I don’t think you’re being paid a lot.

Since 2008 has been a horrible year for stocks, I was curious how these data sets have changed. I called Ibbotson but unfortunately, they don’t do any mid-year updates. So I went to see if I could find a reasonable estimate. Obviously using different data sources can alter your results, but I was looking for data that’s broadly considered fair.

For the S&P 500, Ibbotson uses the dividend reinvested S&P 500. For the first three quarters of 2008, that index is down 19.3%. For corporate bonds, they use Citigroup’s Long-Term High-Grade Corporate Bond Index and for the government bond, they use the 6.375 Treasury that matures in August 2027 (I assume they’ll use a 2028 bond for this year). I couldn’t find the stats on either of these but I called Vanguard to see how some of their index funds were doing.

The Vanguard Long-Term Investment-Grade [VWESX] fund is down 7.39% this year, and the Vanguard Long-Term U.S. Treasury [VUSTX] is up 6.69% this year. I think both of these funds can serve as proxies (VWESX has a current yield of 6.69%, an average rating of A1 and an average maturity of 22.5 years; VUSTX has a yield of 3.86% and an average maturity of 17.5 years).

Tacking these numbers onto the data series, that makes the 39.75-year advantage stocks have over corporate bonds just 1.50%, and only 1.10% for government bonds. If we do a little data picking, we can see that long-term Treasury bonds have outperformed stocks since the summer of 1987, and come in just behind stocks since late 1980. Reasonable people can disagree but that certainly sounds like the long-term to me. This means that you could have sat out the entire stock market over the last 28 years, parked your money in long-term T-bonds and done just as well as the stock market, which we know beats the vast majority of fund managers.

Disclosure: None

Comments

[kuvholt:]

I think you're missing something big here. The equity risk premium does get exhibited in long term outperformance of stocks vs. bonds, but there is a cause of this effect - that earnings of companies grow, and that earnings yields of stocks tend to be higher than interest yields of bonds. In other words, you're correct in saying that stocks in hindsight look bad and bonds look good. Follow that and you'll buy growth stocks in 2000 or banks in 2006. An investor's job is to determine what should perform better than other options in the future, rather than the past, and to act accordingly. Today stock earnings are in a trough, which provides opportunities as people extrapolate poor earnings into the future. Today the earnings yield on the S&P 500 is over 70% higher than the interest yield on the 30 year Treasury. If history offers any lessons, the exact time to buy a quality asset class (like large cap domestic equities) is when it looks like everyone who purchased them in the past was wrong - and that's the point of your article - anyone who purchased stocks since the 1970's appears to have been foolish. I take this as a massive buy signal.

2008 Oct 07 10:26 AM

[JasonC:]

Both ideas are right for different reasons. The comment is correct that stocks are buys exactly at the moments when their past returns do not look attractive compared to bonds, and are sells at the exact moments their past returns seem to trounce bonds. (Measure to peaks e.g.). The other problem with the original analysis is that 1968 is an epic market peak in real terms, akin to measuring from the 1929 high.

But corporate bonds at present spreads are attractive compared to stocks. Treasuries aren't. You aren't going to get the past performance of treasuries that started at 9% yields and moved to 4% yields, starting from 4% yields. Mathematically impossible. But corporate spreads are at record levels.

The second reason is the place bonds actually shine is not in absolute returns, but in Sharp ratio. There is no reason to make a fetish of absolute returns, since for a given level of risk they are available in either asset class. If you can tolerate stock levels of risk, you can tolerate the risks in carrying bonds with 2 to 1 leverage, because they are the same. Bonds have an excellent historical Sharp ratio. Large cap value stocks and commercial real estate are the only other assets in the same league, even. Small value and large cap indexes aren't that far behind, but are behind those three.

You can do very well long term with a 50-50 mix of large cap value and corporate bonds, rebalancing to average in or out based on the relative performance, in contrarian fashion. With great robustness in period to period terms, etc.

2008 Oct 07 11:26 AM