Elroy Dimson, Paul Marsh, and Mike Staunton, Triumph of the Optimists: 101 Years of Global Investment Returns. Princeton, NJ: Princeton University Press, 2002. xii + 339 pp. $99.50 (cloth), ISBN: 0-691-09194-3.

Reviewed for EH.NET by John A. James, Department of Economics, University of Virginia.

First of all, I still don't get the title. It is alluded to only at the very end of the book. Things had not gone so great in the first half of the twentieth century -- two world wars, the Great Depression, hyperinflations, etc. At mid-century "who but the most rampant optimist would then have dreamt that over the next half-century the annualized real return on equities would be 9 percent..." (p. 224)? How can a book then that offers less optimistic estimates of long-run returns, filling in the historical record, extending rate of return series backward, to a period when equity returns were generally lower than those following World War II, be called "The Triumph of the Optimists"?

Now that I've vented for a bit about the title (but I shall come back to it again later), let me describe what the book involves. The authors, all affiliated with the London Business School, argue that the best-known long-term series on returns to equities, those based on US stocks from the Center for Research in Security Prices (CRSP) database beginning in 1926, might be potentially misleading. After all, the United States and US industry in particular had generally done pretty well over this period, so the picture one would get of the long-term return to equity based on such evidence might be too high. What we need are more data, extending farther back in time and covering more countries.

The authors have done an extraordinary job in assembling such a dataset. It encompasses annual real and nominal returns on equities, bonds, and bills, as well as GDP, inflation, and exchange rate data, over 101 years (1900-2000) for sixteen countries (Australia, Belgium, Canada, Denmark, France, Germany, Ireland, Italy, Japan, Netherlands, South Africa, Spain, Sweden, Switzerland, United Kingdom, and United States). It is assembled carefully with an eye to consistency over time and across countries. In constructing the equity indexes care is taken to avoid survivor bias, overweighting companies that last and/or become more important over time. Broader indexes are preferred to narrower ones. Where no satisfactory index existed, one was constructed, as for the UK between 1900 and 1954 (no mention is made, curiously, of Richard Grossman's index (2002) which overlaps part of this period). Long-term performance is measured based on total returns, dividends as well as capital gains. For almost all countries component return series are weighted by company market capitalization. Arithmetic rather than geometric averaging is used.

The countries selected are those for which a century run of financial data exists. The authors make a point of taking care to avoid easy data bias, that is a preference for data which is easy to obtain, steering clear of difficult periods such as wars and their aftermath or periods for which numbers are hard to get, usually earlier ones. Omitting such periods could bias average returns upward significantly. So there really is a bit more than a century of as-consistent-as-they-can-make-it data for most variables here (that said, in most of the book the experience of the German hyperinflation is almost always excluded since it would otherwise dominate the means and standard deviations of the variable in question). The focus on the long run, a century of data, however means that countries in which financial markets disappeared for a while -- Russia, China, etc. -- are excluded. Emerging markets and developing countries are similarly excluded. So we have North American and European countries plus Australia, South Africa, and Japan. The inclusion of other countries would have been interesting, but given the magnitude of the endeavor already, it would be querulous to ask for more. And the countries in the sample did account for 88 percent of world stock market value in 2000, perhaps more in 1900.

The most common time unit used for reporting summary statistics of variables is the century, although it is sometimes broken down into half-century periods. There is certainly something to be said for surveying the full historical record, encompassing the impacts of extraordinary shocks like wars and other disasters. Ignoring the tails of the distribution has had serious consequences for some hedge funds, for example, in recent years. While such a wholistic emphasis on long-run annual average means and standard deviations has the advantage that it captures the influence of lots of different institutions and events, it also obscures potentially interesting relationships. I, for one, would have liked to see more information broken by exchange rate regime, which is done to some degree in one chapter, but not generally.

We learn here that equity holders did better than bond holders over the twentieth century (one important word here -- inflation). Value stocks (those with higher dividend yields and/or higher ratios of book value to market value of equity) yielded higher returns in the long run than did growth stocks (Chapter 10). Generalizing from historical equity returns in the United States turned out to be not that misleading after all. While average returns in the US were higher than the sample average, they weren't extraordinarily higher. The authors spend substantial time (Chapters 13 and 14) on the equity-risk premium, the difference between the long-term average annual return on equities and that on default-free government bonds or bills, the additional compensation necessary for holding a risky asset. The average annual 1900-2000 equity-risk premium relative to bills in the US was 5.8 percent; for the weighted whole sample, 4.9 percent. Relative to long bonds it was 5.0 percent for the US, 4.6 percent for the world index. These figures run about 1.5 percentage points lower than previous estimates for the US and UK (pp. 174-175). (So is this the Triumph of the Optimists -- after we measure the equity premium more carefully, it's still a good bit larger than zero?)

While results such as the above will be the focus of many general readers, economic and financial historians will find that this volume has myriad charms. Perusing the multitudinous tables, graphs, and charts is really endlessly fascinating. How much effect would international diversification have had on average returns? How much effect has currency risk had on the returns to international investment? We see, among other things, that German and American equity returns were sizably negatively correlated during World War II (interestingly, while French and German returns were negatively correlated during World War I, they were positively correlated during World War II) (p. 116). And there's much, much more. Those seeking a respite from formal modeling and econometric tests will find a safe haven here. The principal instrument of assessment is the eyeball. There are lots and lots of colored figures to be studied, but very few formal statistical tests are reported. The graphs (pp. 96-97) certainly suggest that purchasing power parity holds in the long run, for example. But for a formal test of the proposition as well as an analysis of factors underlying deviations from PPP one must go to Taylor (2002) -- who is cited in the text. Another thing missing in addition to formal tests is any significant amount of historical or institutional detail. It would have been nice to have a bit of a story or some context as to why the graphs and charts look as they do. Why was the equity premium in France for instance three-and-a-half times that in Denmark? Is it the lingering influence of John Law?

Curiously, this volume has the look and feel of a textbook -- wide margins; glossy paper; bright, multicolor figures; excruciatingly detailed introductions to each chapter describing precisely what will be covered and long recapitulations at the end of each. But if it is a textbook, for what course or audience is it aimed? In an assessment of prospects for the future, for example, the fact that readers are told "if equities remain risky, as must certainly be the case, equity investors should continue to expect a positive risk premium" (p. 210) might seem to indicate not a very advanced one. The repetitive style makes reading it through from cover to cover, as I did, a bittersweet experience. Rather, it would seem much more felicitous to use it as a reference book (as the dust jacket itself notes), to be dipped into or browsed again and again.

Finally, the lessons of history. Now that the long-term equity premium has been properly measured, is this then a guide to future returns? Well, no. The authors argue (Chapter 13) that the future equity premium is going to be lower than it used to be. Watch for the sequel, "The Triumph of the Pessimists."

References:

Grossman, Richard S. (2002), "New Indices of British Equity Prices, 1870-1913," Journal of Economic History, 62, 121-146.

Taylor, Alan M. (2002), "A Century of Purchasing-Power Parity," Review of Economics and Statistics, 84, 139-150.

John A. James is a professor of economics at the University of Virginia. He continues to work on the operation of the interregional payments system before the Federal Reserve.