The Information Inaccuracy of Stock Market Forecasts: Some New Evidence of Dependence on the New York Stock Exchange

Some recent research into the short-term behavior of speculative prices has combined the methodology of information theory with the analysis of aggregate market data on the daily proportions of securities advancing, declining, and remaining unchanged in price. For example, Theil and Leenders [8] employed the measure of average information inaccuracy to analyze aggregate data on the Amsterdam Stock Exchange and concluded that there was a strong dependence between the proportions of prices advancing, declining, and remaining unchanged today, and the respective changes tomorrow. The evidence of dependence on the Amsterdam Exchange prompted Fama [4] to apply similar methodology on the NYSE data, for the period 6/2/52 to 10/29/62, in search of similar patterns. Professor Fama, after proper adjustment for nonstationarity in the time series, concluded that the evidence of dependence on the NYSE was meager and unreliable for forecasting short-term security prices. However, Dryden [3], utilizing the same methodology on data from the London Stock Exchange, found significant dependence between the proportions advancing, declining, and remaining unchanged today, and the respective changes tomorrow. Dryden's results yielded a higher dependence for the London Exchange than for the Amsterdam and New York Exchanges, and were later verified by the straightforward application of a Markovian approach.