The premier source of historical data used by the academic and corporate community comes from the Center for Research in Security Prices (CRSP). CRSP, which is housed at the University of Chicago Graduate School of Business was established in 1960 with the goal of building and maintaining historical databases for stock (NASDAQ, AMEX, NYSE), indexes, bond, and mutual fund securities. Part of the goal of the center was to unite the common interests between the academic and financial communities by providing a better understanding of the operations of the market. Since computer technology was in its infancy, no machine-readable, historical stock data files were in existence at the time CRSP was launched. Initially, CRSP was formed to accurately measure the returns from investing in common stocks listed on the New York Stock Exchange for the period 1926 to 1960. It took the researchers at CRSP four years to complete this initial study. Since its inception CRSP has developed a host of new data resources. The data housed at CRSP is used extensively for financial, economic, and accounting research. Currently, Eugene Fama, a well respected professor of finance at the Graduate School of Business at the University of Chicago and director of research at DFA, is the chairman of CRSP. DFA bases several of its investment products on Fama’s findings from that database.
The prestigious University of Chicago publication, The Journal of Business, caused the academic equivalent of an earthquake in an article published in January 1964 titled “Rates of Return on Investments in Common Stocks.” In the article, James Lorie and Lawrence Fisher, two business professors at the school, made the first comprehensive measurement of the performance of all common stocks listed on the New York Stock Exchange from 1926 through 1960. They obtained and compiled their data from CRSP.

The study presented a mind-boggling accumulation of statistical calculations. Both academic researchers and investment professionals were astonished at Fisher and Lorie’s discoveries. For instance, the study showed that an investor who invested $1,000 in the stock market in 1926, reinvested all dividends, paid no taxes, and remained fully invested until the end of 1960 would have accumulated nearly $30,000 or a gain of about 9% a year. In light of the fact that many investors in 1964 still had vivid memories of the Great Depression and its stock market crash, 9% a year was a great deal of money. In addition, this return was far greater than the amount an investor would have earned from bonds or savings bank deposits during that time period. For the first time, investors had comprehensive historical investment data that gave them a sense of how common stocks performed compared to other investments.

Roger G. Ibbotson and Rex A. Sinquefield, two graduates of the business school at the University of Chicago, released a study that was published in The Journal of Business titled “Stocks, Bonds, Bills and Inflation.” The two researchers were the first to compile and present in an organized way historical investment data that covered not only stocks, but bonds as well. They even reported data on inflation. As was the case with Lorie and Fisher’s study, their data went back to 1926, and was obtained from CRSP. The Ibbotson-Sinquefield data, now updated annually in what has come to be known as the “Stocks, Bonds, Bills and Inflation (SBBI) Yearbook,” is widely used in the investment world.

In 1990, G. William Schwert of the University of Rochester published an article in The Journal of Business titled “Indexes of U.S. Stock Prices from 1802 to 1987.” Schwert pointed out that the data compiled by CRSP launched an explosion of research in finance in the 1960s to 1970s. However, notes Schwert, a major drawback of the CRSP database is that it starts in 1926, a time right before the Great Depression. Consequently, the behavior of the stock market and stock returns was unusual in the 1929 to 1939 decade. Therefore, an empirical study based on the data could be “suspect.” So, Schwert set his sights on pre-CRSP stock return data. His article compares and contrasts all of the major indexes of stock prices or returns that were available monthly from 1802 to 1925 or daily from 1885 to 1962. The outcome of the comparison is a series of monthly stock portfolio returns from 1802 to 1925, and daily returns from 1885 to 1962. This important study included many refinements of the concept of “stitching” together several different index data series to obtain a longer term prospective.

9.2.2 Time Series Construction

A time series construction is the stitching together of indexes through history so that researchers and investors can better characterize the risk and return of their investments. See Figure 9-1 for an example of the time series construction of indexes. As seen in the graph, indexes are stitched together to increase the sample size of the data. All indexes have been taken back to 1927 through this process. A substitution process is used to extend current indexes back in time. This process is far from perfect, but provides the best information available for extending current indexes and mixes of those indexes back in time. Table 9-1 is the annual returns of these stitched together indexes with corresponding color buttons, and a total market index for comparisons. This is an interesting assembly of the per-year annual returns for each index going all the way back to 1927! These indexes are described in further detail in Appendix B.

The first problem investors are faced with relative to history of stock market returns is the lack of quality long-term data. Secondly, they are not aware that long-term data has more value to them than does short-term data. When looking at 80 years of data many investors think it is irrelevant because they do not have 80 years to live. This point of view overlooks the importance of sample size and the concern for sample error. When gathering information to characterize the risk and return of capitalism, the more quality data you have, the more accurate your conclusions. Any subset of the data, such as five years worth of data, is bound to contain significant errors in its attempt to describe the risk and return of an index. For example, for the five-year period from 2000 to 2004, the S&P 500 had a total loss of 12%. Based on that negative total return, many investors would conclude that the S&P 500 was not a good investment. But when considering 80 years from 1927 to 2006, we see that the annualized return over that period is about 10% per year, and it would be within normal limits for it to fluctuate that much over five year periods. Therefore, it is still an important component of diversified index portfolios. That is a very different conclusion and is far more accurate than the conclusion many investors make based on the last five years.