Debate over the equity risk premium is unlikely to be covered in the supermarket tabloids anytime soon, but a war of words among financial literati is getting pretty sensational.

Roger Ibbotson, chairman of Ibbotson Associates and a professor of finance at Yale School of Management, recently fired back at those who have disagreed with his way of forecasting long-term equity returns. With Peng Chen, the director of research at Ibbotson Associates, Ibbotson recently wrote “Stock Market Returns in the Long Run: Participating in the Real Economy,” a paper that will be published in coming months in Financial Analysts Journal. In it, Ibbotson revises downward his forecast for long-term stock returns by revising his forecasting method.

Financial advisors who give advice every day to retail investors sometimes dismiss the academic literature as, well, ivory tower discussions with little real-world effects. But the debate over the equity risk premium is important to advisors. It provides a framework for assumptions that you need to make when you tell clients how to invest so they can retire securely and not run out of money.

In addition, since the debate over the equity risk premium began to heat up about two years ago with the release of a sobering paper entitled, “The Equity Premium,” by professors Eugene Fama of the University of Chicago and Kenneth French of the Sloan School of Business at MIT, the stock market has suffered the worst bear market in decades.

So the fact that Ibbotson, the granddaddy of stock return forecasters, has reworked his forecast and revised it downward, if only slightly, is big news.

Extrapolating Returns

In 1976, when Ibbotson was a doctoral candidate, he co-authored a paper with one of his graduate students, Rex Sinquefield, in which they deconstructed historical stock returns dating back to 1926 and then extrapolated that data forward to make a long-term forecast. Ibbotson and Sinquefield forever changed the way long-term investors would look at the market by breaking a stock return into its component parts. Stock returns, they said, were composed of the rate you get on U.S. Treasury bills, which implicitly includes inflation, plus an extra chunk of return called the Equity Risk Premium (ERP).

Based on their historical study, Ibbotson and Sinquefield made a 25-year projection. They forecast an ERP through the end of 2000 of 7.9% compounded annually, which was not far off from the ERP of 6% actually realized during this spectacular bull market period for the stock market.

Keep in mind that in 1976, when the forecast was made, the excess return above the risk-free rate of return was low. Stocks were in the middle of a terrible slump. In 1965, the Dow was at 1,000 and would not top that mark until 1983. As late as 1979, three years after the Ibbotson-Sinquefield forecast, Business Week’s cover called the market era of the time “The Death of Equities.”

However, the actual forecast arguably was not as important as the way Ibbotson looked at things. Prior to this, stock forecasts were made based on total return. It was only after this 1976 paper that financial economists began to focus on its components.

Ibbotson’s Ideas Questioned

Ibbotson capitalized on his academic success by founding Ibbotson Associates. This Chicago-based consulting firm explains the behavior of stocks, bonds, inflation, diversification, and other issues of central importance to investors, and makes software for financial professionals. His method of extrapolating long-term history into the future became the industry standard in the 1980s and 1990s.

But in the last couple of years, Ibbotson’s ideas about the future of the stock market have come under attack by other respected academics and money managers who have published forecasts of their own, and who calculate the ERP differently from Ibbotson. They also predict lower returns over the long haul.

In 2000, Fama and French published “The Equity Premium.” In the article, the two men argue that the ERP is 2.55% over bonds, much lower than the long-term equity risk premium of more than 7% that you’d get by simply extrapolating history the way Ibbotson did with Sinquefield in 1976.

Fama and French used a dividend and earnings model to forecast long-term equity returns. Basically, what Fama and French show in their paper is that stocks have paid out more in returns than companies can make on their money. The return on equity of corporate America is lower than the returns on stocks. Having lately become accustomed to being paid too well, stock investors, according to Fama and French, should expect lower returns in the future. This differed sharply with forecasts from Ibbotson.

Jeremy Siegel’s Critique

Circulation of Fama and French’s working paper had followed an article published in The Journal of Portfolio Management in 1999 by Jeremy Siegel. In the article, Siegel, of the Wharton School, said Ibbotson’s 1976 forecast was off the mark when his inflation forecast was considered. Siegel said that Ibbotson had forecast an inflation rate of 6.4% when the actual inflation rate was 4.8%. Siegel also argued that Ibbotson’s forecast underestimated real returns, especially on fixed income assets.

More recently, Rob Arnott, CEO of First Quadrant, an institutional money management firm in Pasadena, California, co-authored a paper with Peter L. Bernstein, consulting editor at The Journal of Portfolio Management, about the equity risk premium. They dispute Ibbotson’s argument that when P/E ratios are high and the market is expecting higher earnings growth, that the market is right and earnings will grow. They also disagree with Ibbotson’s assertion that corporate earnings not paid out in dividends will come back to the investor in the form of increased growth, an idea relying on the Nobel Prize-winning research of Miller and Modigliani in the early 1960s. Arnott and Bernstein found that retained earnings do not improve future earnings growth and in fact degrade future earnings growth.

Here’s what Ibbotson said about all of this in a recent interview.

You’ve moved away from simply extrapolating history to come up with an equity premium forecast that predicts future stock market returns. Is that approach now outmoded? It gives you a good indication for what kind of returns you’ll get for taking risk into the future. It’s a good way [to do it] but not the only way. Another good guide is what is happening in corporations and in the economy itself, and you could argue that historical returns have been unusual in some respects. Maybe the economy is generally stable in the long run, while stock returns may have been unusual, especially during the 1980s and 1990s’ long bull market, and that may have raised equity returns. In other words, the [more recent] historical period may have higher returns than would be reasonable to forecast for the future.

Your recent equity risk premium forecast makes a significant adjustment to the historical approach by not assuming that price/earnings growth will continue into the future. Please explain what you did and why. The underlying earnings of corporations have grown at basically 1.25% less per year than the returns of corporations. The reason is that the P/E ratio, which started in 1926 at 10, went to 25 or so, and in the supply side approach, that’s considered a one-time windfall. You should not expect that to happen again.

If returns are higher than earnings because returns had a boost from the P/E ratio, you should not expect another P/E ratio boost looking forward. The P/E ratio would have to go from 25 to 62.5. That’s not in these predictions.

In the earnings approach, we extrapolate the earnings historically and not the return history.

Why did the P/E ratio go up? Presumably, it was because growth in earnings was increasing. But it would have to keep accelerating for the P/E ratio to continue to accelerate. So, if you take that out, that’s essentially the difference between the historical ERP forecast that Ibbotson-Sinquefield made and the supply side forecasts that Ibbotson-Chen makes based on earnings.

The Ibbotson-Sinquefield approach was just extrapolating returns, which would be automatically extrapolating the P/E ratio growth from 10 to 25, and would now say the P/E ratio should go from 25 to 62.5.

Now we are saying that the growth in earnings is supposed to be higher but we don’t think the growth in earnings will accelerate more. The growth in earnings is the first derivative and the change in growth in earnings is the second derivative. We believe there is a change in growth of earnings but not an acceleration in growth of earnings. So this is how we depart from historical equity risk premiums by that 1.25% per year.

If instead of getting an expected return in the high 10′s [using an] historical approach, you might expect it to be in the 9′s. This is geometric, or compounded. So if you take a 10.7% historical return and remove the 1.25%, you get about a 9.4% return as the forecast. If you express it as an ERP, the historical figure was about 5.25%. That’s the historical return of stocks minus long-term bond yields. And that would fall to a little below 4% compounded.

Stocks will give you almost 4% more than long-term U.S. Government bonds. So if the yield is today 5.5%, add 4% to that and that would be your expected return going forward over the next 25-year period.

Your paper takes aim at recent studies relying on a dividend growth model to make long-term forecasts. Fama and French and Arnott-Bernstein use that method. Why? Fama and French–I’ve talked with them about the issue. I think that in their earlier version, their paper violated Miller and Modigliani. I pointed it out to them and they straightened that out in the published version.

My complaint about dividend yield approach is that dividend yields were 4% historically but only are a little over 1% today. Therefore, the ERP is 3% lower than it has been historically if you rely on this approach without adjusting for that [decline in yields].

I’ve said such thinking violates Miller and Modigliani. If yield is lower and companies are retaining earnings and reinvesting them, the money is not lost. That money is being used by the company and the returns should reflect this. That’s what Miller and Modigliani said in 1962 when they wrote their original dividend paper and won the Nobel Prize for it. It doesn’t matter if the dividend is retained or paid out. Investors will benefit from it in the same way.

Now Arnott says he understands all this but he believes corporations waste the money they retain. I feel Fama and French is legitimate, good work. However, Arnott and his colleagues are not even estimating the ERP; they’re really timing the market and saying the market is way overvalued today and, because of that, it’s going to lower stock returns in the long run.

My starting assumption, like Fama and French, is that the markets are fairly priced today. You must make that assumption to figure out a long-term return on the market. You need to have that number in order to know how to discount a cash flow. What Arnott and Bernstein and others are doing is really market timing. They’re long-run market timers, but they are market timers. So I’m actually reluctant to even be drawn into a debate with them.

I’m sure my theory has been under attack. But the timers are discussing another topic. Conceptually, the ERP cannot be negative. Some may say it is low but not negative. Arnott and Bernstein say it is negative. They say the market is mispriced and, under some long-term reasonable horizon, we need to work it out [the mispricing]. If you’re a market timer, you may say that over the next five years, the stock market will go down and not up. An ERP assumes a fair valuation exists to begin with and at the end point of the forecast. Fama and French are not bears. They’re trying to calculate the long-term equilibrium ERP. You could classify Arnott as a bear.

I was speaking with Harold Evensky recently about your article, and he mentioned Siegel’s article published a couple of years ago in which Siegel basically said that your 1976 forecast was actually off the mark after inflation is considered. How do you respond to the points made by Siegel in that article? We were forecasting the ERP and the stock market. Siegel characterizes everything in terms of inflation. We separately made an inflation forecast. We said that yield in 1976 was a combination of inflation and real interest rates. But the starting point was the yield curve and then we built on it from there to get an inflation rate. From there we get the ERP. It is true we underestimated the stock market and overestimated inflation. Now, Siegel’s taken the inflation part of it and focused on that. I’m not saying it is totally unfair to do that. He is saying if you take the real returns, they were off quite a bit. But we were forecasting first nominal return based on a nominal yield curve. I don’t want to quibble with the point entirely by saying that our method came out in nominal terms. We did make an inflation forecast, so if you recast everything some of our forecasts are off by quite a bit. Our forecast of the stock market was too low and our forecast of inflation was too high.

But you really should look at these things through 2000, as our forecast was made to span that 25-year period [starting in 1976]. When Siegel published that paper it was 1999, and he used data through 1998. Our forecast was for the period from 1976 to the end of 2000. For that [time period], our forecasts would not be off very much. For instance, the real ERP from 1976 through the end of 2000 was 7.9%, which is not that far off from our prediction in 1976 for a 6% real compounded ERP. If you look at all this in real terms, as Siegel has characterized us, that’s where we did the worst. But that’s like taking a picture of you in your least favorable light and saying that is what you look like. In terms of our ERP forecast made in 1976, we were not that far off.

Keep in mind that as part of that 1976 forecast, we made a Dow forecast for 10,000 in 1999, and that was right on the money. These numbers were in the right ballpark. You don’t expect them to be on the button. You can cast them one way and you can look like a genius and another way [and you can] look like a fool. While we underestimated the stock market and overestimated inflation, we were pretty close on the ERP.