NEW HAVEN, Conn. -- Stock returns from 1815 through 1925 were slightly lower than those of the commonly used 1926-1999 period, according to a pioneering study still being completed.
This is the first time individual stock prices have been collected for all New York Stock Exchange equities over the 111 years that began in the year Andrew Jackson crushed the British in the Battle of New Orleans and Napoleon was defeated at Waterloo.
Roger G. Ibbotson and William N. Goetzmann, professors of finance, and Liang Peng, a Ph.D. candidate in finance, all at Yale University's School of Management, are collaborating on the study.
The three are compiling monthly prices for every stock. They include dividends for about half of the stocks.
The data will open new areas for more accurate research and analysis, said Mr. Ibbotson.
Modern researchers of pre-1926 stock returns have had available only the data of various stock indexes, such as the Cowles index covering 1871 to 1940, Mr. Ibbotson said. But the prices of individual stocks in those indexes have been lost and the indexes spanned only fragments of the 19th and early 20th centuries.
"The problem is prior to 1926, there has not been individual stock data available," Mr. Ibbotson said.
The study will be submitted for publication in the Journal of Financial Markets; the trio is working on an Aug. 1 deadline. Avanidhar Subrahmanyam, professor at the Anderson Graduate School of Management, University of California, Los Angeles, and an editor of the journal, said a publication date hasn't been set yet.
Publication will be the fruition of 11 years of paging through fragile old newspapers at the Beinecke Rare Book Library on the Yale campus in New Haven, Conn., to search for prices of individual stocks listed on the NYSE, Mr. Ibbotson said.
"All the raw stock prices have been collected over the last 11 years by students working every summer," he added. "You have to worry about the pages flaking off."
They began with 1815 because that was the earliest they could find prices. It was the year a weekly newspaper called New York Shipping and Commercial started reporting NYSE stock prices, which it reported through 1852. For the subsequent years, the researchers culled prices from other New York newspapers in the library, Mr. Ibbotson said.
The authors plan to make the database available to researchers.
"We might make it available free to some researchers, or sell it in some form. We don't expect to make a lot of money on it," Mr. Ibbotson said.
The study was mostly self-financed, he said. The three received some aid, but not from any consistent source.
One difficulty was getting dividends. Mr. Ibbotson said the researchers were able to find dividend histories on individuals stocks only from 1825 and then for only about half of the stocks. For these stocks, they compiled individual dividends through 1870. Then from 1871 to 1925, Mr. Ibbotson and his colleagues used aggregate dividend data from the Cowles series.
"We may go back and fill it in later," he said, referring to getting individual stock dividends from 1871 through 1925.
For stocks for which they couldn't find dividends, the three made low and high assumptions, either there were no dividends or dividends of these stocks were equivalent to dividends of the stocks for which they found histories.
"From 1815 through 1870, the capital appreciation is quite low, only 0.84%," Mr. Ibbotson said.
From 1825 through 1870, dividend return is estimated at a low of 3.77% and a high of 9.27%, he said.
For 1871 through 1925, capital appreciation is slightly higher at 1.65%. Dividend yield for that 55-year period, taken from the Cowles aggregate, was 5.33%.
The three reseachers haven't completed calculating results from 1815 through 1925.
The results so far indicate that returns over the 111-year period are a little lower than the Standard & Poor's 500 stock index return for 1926 to 1999, he said.
For the 74-year period ended in 1999, the total return, including dividends reinvested, averaged 11.3% a year, compounded annually, according to Mr. Ibbotson. Broken down, that modern return series consists of 6.62% in capital appreciation and 4.45% in dividends.
For 1825 to 1925, the total return is estimated between 6% and 8.5%, depending on the estimated dividend.
For 1825 to 1999, Mr. Ibbotson said, the total return on stocks is between 8.24% and 9.7%, including the researchers' low and high dividend estimates.
They also are still working on volatility data, he said. But for 1825 through 1999, the standard deviation, a measure of volatility, was 18.26%. By contrast, the 1926 to 1999 period had a standard deviation of 20.14%.
Understanding the past
Asked about the importance of the work, Gary P. Brinson, chairman of UBS Asset Management, which includes Brinson Partners Inc. and other investment units, Chicago, said, "The fuller and richer the history of the data, the better we should be able to understand how the markets and economy have evolved and developed. It's not clear it tells us anything about the future. But it helps us understand the past."
Robert D. Arnott, president and chief executive officer of First Quadrant LP, Pasadena, Calif., said, "It's very interesting work. But anytime you go back that far you have to ask how applicable is the data to today's system, when equity ownership has changed so much."
Mr. Arnott said he would expect the average annual returns on stocks in the 100 years prior to 1926 to be two percentage points lower than the modern return series, in part because price-earnings ratios average about 35 today, compared with about 12 in 1925.
Josef Lakonishok, professor of finance at the University of Illinois, Champaign, and chief investment officer of LSV Asset Management, Norwalk, Conn., said the new data sound interesting. But, he cautioned, "If I could choose years to analyze, I don't know if I would choose those years."
Thomas M. Richards, principal at Richards & Tierney Inc., a Chicago-based consulting firm, said of the study, "It's an interesting piece of information." He said its best use may be to "assess volatility and risk/return factors of the market."
But, he added, "Having data after 1945 is probably more important because it is a better representation of the world in which we live. Prior to that there were many institutional constraints" on the markets and investors.
Undetermined market value
Mr. Ibbotson said he and his colleagues are putting the prices into a stock-price-weighted index to study overall performance of the market during that period. They can't construct an S&P 500-like index weighted by market capitalization, because they were unable to find out how many shares of stock were outstanding and thus couldn't determine the market value of companies.
Mr. Ibbotson said they also ruled out creating an equal-weighted index to avoid biases, such as bad quotes and so-called bid/ask bounce. For example, a stock selling at 1 might have a bid of 1/2 and an ask of 11/2. Transactions could bounce between the two ranges, causing the price to go up 200% or down 66%, skewing results.
"The best way to do it we thought is a price-weighted index like the Dow Jones" industrial average, Mr. Ibbotson said.
In 1815, the data consist of about 10 stocks. In 1871, there were about 125 stocks, although there were earlier years when the number was higher. By 1925, there were about 500 stocks.
"Most of the time, there are about 100 stocks," he said.
"A lot of questions always arise in the research on what happened in the early period," that is pre-1926, Mr. Ibbotson said. "We've never been able to answer that."
The longer time series will reduce the estimated error rate in research of returns, he said, because the estimated error rate is related to time.
He compared a 25-year period to a 75-year period.
If over 25 years stocks return an average of 10% a year and the standard deviation is 20%, then the estimated error rate is 4. That is, the standard deviation divided by the square root of the years. So stocks over 25 years can be expected to return 10% plus or minus four percentage points.
But over 75 years, using the same 10% average annual return and 20% standard deviation, the estimated error rate is only about 2.3. So stocks over 75 years can be expected to return 10% plus or minus 2.3 percentage points.