An Analysis of Non-Linear Stock Price Dynamics with Point & Figure Charts

Abstract

Chart-based technical analysis is primarily concerned with the recognition of non-linear patterns in the time series of past prices. Financial economists have traditionally regarded pattern recognition a highly subjective task, and, therefore, not subjected this area of finance to their usual scrutiny. We propose the use of point and figure charts to study the non-linear dynamics of stock prices. Pattern recognition in point and figure charts is an objective task, and it only requires algorithms of low (i.e. linear) computational complexity. An application of this technique to the 1962-2003 CRSP daily stock files yields a number of interesting findings. First, certain well-known chart patterns are shown to be indeed useful in predicting future stock returns. Shares whose stock prices completed 'buy-signals' systematically outperform shares that completed 'sell-signals' over subsequent holding periods of three to twelve months. Second, this effect appears to be attributable exclusively to chart patterns that we classify as 'trend-confirming'. 'Trend-reversing' chart patterns appear to be less reliable and contain little useful information. This finding provides some empirical support for the popular notion that stock prices tend to move in trends. Finally, an investigation of stock returns immediately following the completion of various chart patterns suggest that some investors are aware of the above findings, and that their trading activities affect stock prices at least in the short run. Overall, our study suggests that point and figure charts may be a useful tool for empirical research in the areas of market efficiency and behavioral finance.