WINTER 2009 THE JOURNAL OF INVESTING 1 M ost investors appear to believe that the long-term return on U.S. stocks is a known quantity—after all, it can be looked up in any Ibbotson yearbook (e.g., Ibbotson [2007]). The Ibbotson estimates, derived from the database maintained by the Center for Research on Security Prices at the University of Chicago (CRSP), provide the foundation for most asset allocation systems. Thus, according to Ibbotson, for purposes of asset allocation, stocks can be expected to return about 10% annualized, or 7% after infla- tion. More particularly, on an annual basis stocks have an arithmetic mean return of about 12.5% and a standard deviation of about 20%. It’s a rare investor who has not encoun- tered these figures. And the ordinary investor generally sees these numbers presented in the language of historical fact—e.g., “The histor- ical record shows …”, or “Over the long term stocks have returned …”, etc. There is no more authoritative source than CRSP, and no better- known distributor of CRSP data than the Ibbotson yearbooks (now owned by Morningstar). The purpose of this article is to high- light the limitations and biases of the CRSP data that underwrite both the Ibbotson find- ings and other academic accounts of long-term U.S. stock returns (e.g., Fama and French [2006]). Once acquainted with the possible biases in CRSP data, investors can decide for themselves how much faith to put in such shib- boleths as the Ibbotson yearbook returns or Fama and French’s finding that small stocks outperform large stocks. All depend on the comprehensiveness of CRSP coverage, which turns out to be more limited than many investors might suspect. CRSP DATABASE Introduced in the mid-1960s (Fisher and Lorie [1964]), the CRSP database at present includes complete return information on essentially all exchange-traded stocks in the United States. Unfortunately, the comprehen- siveness of current CRSP data has tended to obscure the very real limitations of historical CRSP data. The key limitations to understand are: 1) the CRSP timeframe, which begins in 1926, excludes more than 50% of the historical record of widespread, large-scale stock trading in the United States, which goes back almost 200 years; and 2) for more than 50% of its timeframe, the CRSP dataset excludes the majority of stocks trading in the United States, especially the smaller and more vulnerable enterprises. Putting these two facts together, we may say that CRSP provides comprehensive price series data for less than 20% of the total U.S. stock trading record, aggregating across time period and type of stock. EDWARD F. MCQUARRIE is a professor of marketing & associate dean for assessment at Leavey School of Business, Santa Clara University in Santa Clara, CA. emcquarrie@scu.edu The Myth of 1926: How Much Do We Know About Long-Term Returns on U.S. Stocks? EDWARD F. MCQUARRIE JOI-McQuarrie:JOI-McQuarrie 10/24/09 12:16 AM Page 1 Author Draft For Review Only