We are about to end a dismal decade for stocks: The past 10 years will likely provide the first negative return for stocks since the 1930s. Many investors rightfully question whether the lessons of the past are still relevant and whether stocks are still the best long-term investment.
Last summer Jason Zweig, a well-known journalist and financial writer for the "Wall Street Journal" questioned the quality of the early 19th century stock data that I used to support the long-run case for stocks. Although he has no problems with my data or my analysis of the stock market since 1871, he claims that the data from 1802 through 1871 is "rotten with methodological flaws." Furthermore, he claims that I inexplicably raised the dividend yield during that period from 5 percent, when my data were first published in academic journals in 1992, to 6.4 percent two years later. This unwarranted increase, Zweig claims, juiced my stocks returns in the early period and gave a much more favorable cast to long-run stock returns.
One could argue whether any of my (or other researchers') conclusions about the superiority of stocks as long-term investments is at all dependent on data that are nearly two centuries old. But I take his challenge to my data and research seriously, and I believe it very important to set the record straight.
Early Returns
My first studies were based on the path-breaking research of Prof. William Schwert of the University of Rochester, who published a paper in 1991 titled "Index of U.S. stocks prices from 1802 to 1897". In that paper, Schwert assumes a 5 percent dividend yield on stocks, borrowing the yield that researchers found in later data, and he admits that he has no evidence to prove whether 5 percent is correct for the early sample.
My first published article on long-term returns in 1992 used Schwert's 5 percent dividend yield. But when I began sampling the dividend yield on stocks from that period, I found that many stocks had a higher return, and I used an average of those returns to make my case.
New Data
Admittedly, Schwert's early data, first compiled in the 1930s by Professors Walter Smith and Arthur Cole, have flaws. But since then we have been graced with some superb research that strongly supports the returns that I used. Two of the top researchers in the field of U.S. stock returns, professors Will Goetzmann and Roger Ibbotson of Yale University. published an article in 2001 titled "A New Historical Database for the NYSE 1815 to 1925: Performance and Predictability". This work is by far the most thoroughly documented research on early U.S. stock returns, collecting monthly price and dividend data on more than 600 individual securities over more than a century of data.
This was a prodigious effort. They reported that it took their research team more than a decade of effort to track down individual share prices and dividends, mostly from original publications found in Yale's Beinecke Rare Book Library. The data that they collected is free from the survivorship bias and other problems that Zweig cites in his critique of Schwert's data.
Ibbotson and Goetzmann determined that the biggest source of uncertainty in these early stock returns is the dividend yield, since many of the sources from which they obtained stock prices did not report dividends. As a result, they formed two series of dividend yields, one assuming that those stocks for which they could not find dividends had zero dividends (their "low income" estimate), and another which uses the dividend yield of those stocks for which they could find dividends (their "high income estimate"). They write:
"The low income returns from the pre-1871 period is 3.77 percent per year. .... When we consider only the dividend paying stock during that era, however, we estimate much higher income returns -- 9.27 percent per year. This higher income return estimate is consistent with the practice of paying out profits to keep stock prices in the early period trading near par values. The true dividend return to a capital-weighted investment in all NYSE stocks is undoubtedly somewhere in between these two extremes."
This midpoint of their high and low estimates is 6.52 percent, higher than the dividend which Zweig criticizes as too high. Furthermore, their estimate of the capital gains for stocks during the period is actually slightly higher than the 0.3 percent per year estimate that I used. So recent research suggests that my estimate of stock returns in the early period is actually quite conservative.
Long-Run Outperformance
Zweig sharply criticizes my statements about long-term stock returns. He points out that "U.S. stocks have underperformed long-term Treasury bonds for the past 5, 10, 15, 20, and 25 years" and it is likely that 30-year under-performance is near. Well, Zweig can scratch 25 years as the market rally has pushed stocks ahead of bonds over that period. And if stocks return only 4 percentage points more than treasury bonds next year (which I consider extremely likely), he can scratch 20 years from his list as well.
The last 30-year period in which bonds beat stocks was from 1831 through 1861. Furthermore, stocks, in sharp contrast to bonds, have never suffered negative after-inflation returns over any 20 year period or longer. That is quite a record, and Zweig does not disagree with either of these statements. Nor does he disagree with any of my analysis of the data over the past 130 years. Nevertheless, he claims that history cannot tell us whether stocks will beat bonds over the long run.
Final World
With a final knock on my research, Zweig proclaims, "Another emperor [Jeremy Siegel] of the late bull market, it seems, has turned out to have no clothes."
On the contrary, I will be most happy to bet my wardrobe against his that stocks' 30-year returns will keep their century and a half record of outperformance over bonds intact in future years. If he takes the bet, I have no doubt that Jason, not I, will be the one running around naked.
