Two Technical Analysis Rules Pass Academic Muster
Technical analysis has long been looked down upon in many influential circles. Academicians publish countless articles “proving” that technical analysis lacks merit. Technical analysis has also not been highly thought of by modern Wall Street analysts trained at the leading business schools, and skeptical journalists have not exactly been favorable to technical analysis, either.
The concept behind technical analysis is that the study of the market itself can lead the investor to form expectations regarding the future course of prices. By studying past prices, an accurate appraisal of the demand/supply equation for stocks can be achieved and forecasts can be made regarding their future course.
“Baloney,” say the efficient market theorists. A stock price at any point in time effectively discounts all pertinent information and any future price action will tend to be random or will reflect new information, rather than any pattern suggested by technical analysis.
So, it came as more than just a mild surprise when an article in the December issue of The Journal of Finance, “Simple Technical Trading Rules and the Stochastic Properties of Stock Returns,” reported some study results favorable to technical analysis.
The objective of the authors, William Brock and Blake LeBaron, both from the University of Wisconsin, and Josef Lakonishok of the University of Illinois, was to test two basic indicators in technical analysis—moving averages and trading-range break—as to whether these indicators were able to forecast future price changes.
The authors were also interested in examining alternative statistical testing techniques. However, for most AAII readers, primary interest in the article is whether the two techniques of technical analysis being tested—moving averages and support and resistance—can indeed be useful in predicting future returns.
Technical analysis is often associated with “trading,” as the authors have done. But any light that might be shed on being able to “predict” stock prices should be of interest to both short-term traders and long-term investors alike.
The database used for the study was the Dow Jones industrial average from 1897-1986, broken down into four subperiods, 1897-1914; 1915-1938; 1939-1962; and 1963-1986.
The theory behind moving averages is that a moving average will “smooth out” to some degree the volatility of stock prices associated with shorter time periods, thereby spotlighting the more important trends of the market. The popular 200-day moving average, for example, is in effect a proxy for the primary trend of the market, smoothing out the minor and intermediate trends. When a primary trend indicator such as the 200-day moving average is penetrated one direction or another, change of primary trend inferences can be made.
Their first rule was to test variations of short moving averages (one-, two-, or five-day) moving above or below longer moving averages (50-, 150-, or 200-day). A filter or band of 1% was introduced to lessen “whiplash” effects of the market changing momentarily and then quickly reversing; in other words, no trades were made if the change was less than 1%.
The first moving average rule, coined Variable-Length by the authors, was designed for the trader to buy when a short moving average moved above the long moving average; to stay in the market until the short moving average moved below the long moving average; and then either move out of the market or sell short.
For each of the variations the results were positive, with the average for the buy rules being an average one-day return of 0.042% (close to 12% at an annual rate) and the average one-day return for the sell rule being –0.025% (about –7% at an annual rate). These returns compare favorably with the unconditional one-day return of 0.017% (about 4% at an annual rate). For all intents and purposes, crossing the moving average does have forecasting implications.
A second moving average rule, coined Fixed-Length, measures the returns for the 10 days following the penetration of the long moving average by the short.
Again, the results for the different variations were all positive. The average 10-day buy return was 0.53%, while the average 10-day sell returns were all negative and averaged –0.40%. The average unconditional 10-day return was 0.17%.
Noteworthy in both moving average tests was that the returns using a 1% band were greater than using no band. For example, the average buy-sell difference for the Fixed-Length without a band is 0.77%; with a band the difference increased to 1.09%. The effects of a band were to blank out the effects of “whiplash” from any day-to-day “backing-and-filling” of the market at key junctures.
Also noteworthy is that about 50% more buy signals are generated than sells, consistent with a market that tends to be rising. On the other hand, stock returns are significantly more volatile following sell than buy signals.
There were some marked differences between the buy signals and the sell signals within each time frame. Nonetheless, the traditional proxy of the primary trend of the market, the 200-day moving average, with a 1% band, in the Variable-Length test, generated a buy-sell figure of 0.070%, which compares very favorably with the unconditional return of 0.017%. Under the Fixed-Length test, a buy-sell signal of 1.35% was generated, which again compared favorably with the 10-day unconditional return of 0.17%.
The moving average results strongly imply that most investors should at least consider the relation of the current market to the 200-day moving average. While horizontal trading periods—trendless markets—such as 1990, are the “bearbug” of any trend-following system based on the 200-day average, the authors’ results suggest that, overall, it pays to be in the market when the Dow is above its 200-day moving average, and to be more conservative when the Dow has moved below its 200-day moving average.
The trading-range break is a version of support and resistance, where lows indicate support (a price “floor”) and highs indicate resistance (a price “ceiling”).
Since reversal areas such as previous highs and lows are normally a level of some concentration of previous supply and demand, they are by definition support and resistance levels. Once a price is able to move above a previous high, or resistance level, then the demand for the stock has absorbed the supply and theoretically the issue will move higher. Or, once a price is able to move below a previous low, or support, then the supply for a stock is greater than demand and theoretically the issue will move lower.
While the trading-range break was a trading rule test, it was also in effect a test of trend. The classic Dow theory definition of trend, as summarized by Robert Rhea in The Dow Theory, is as follows: “Successive rallies penetration preceding high points, with ensuing decline terminating above preceding low points, offer a bullish indication. Conversely, failure of the rallies to penetrate previous high points, with ensuing declines carrying below former low points, is bearish.”
For the trading range test, the buy and sell signals were generated when the price level moved above or below local highs and lows, local being defined by the high and low of the last 50, 150, and 200 days. For each test a 10-day holding period return was calculated following buy and sell signals.
In all six rules tested, the average buy return of 0.63% and the average sell return of –0.24% compared favorably with the average 10-day unconditional return of 0.17%. Like the moving average test, the 1% band improved the returns substantially. In all the tests, however, when the local highs or lows were penetrated, the market followed through as it should be expected to do under the traditional concept of trend.
In the conclusion the authors stated that the results “provide strong support” for the technical strategies they tested—moving averages and trading-range breaks—and that the results are “consistent with technical rules having predictive power.” They caution, however, that transaction costs have to be considered in the implementation of such rules.
The authors suggest that the “returns-generating process of stocks” (technical analysis) is probably more complicated than suggested by some statistical models, and that technical analysis rules can pick up on some of the “hidden” patterns. They state that they tested only the simplest trading rules and that more elaborate trading rules may generate even more significant returns.
Readers of the AAII Journal may or may not choose to read the article for themselves. However, I wanted to bring the article to the attention of readers since it is an academic review that suggests the underlying rationale of technical analysis has validity—the underlying rationale being that investors do indeed react in predictable fashion to changes in stock prices.
AAII readers will not become technicians as a result of articles like “Simple Technical Trading Rules and the Stochastic Properties of Stock Returns,” but some awareness that trends exist, can be measured and classified, and exploited, should be of interest to all investors. At the least, some knowledge of technical analysis can reduce risk.
For instance, why buy stock when the market moves below the moving average or the trading ranges, if the facts suggest lower prices ahead? Might it not be more prudent to buy when one of the moving average models or the trading range model suggests higher prices?
Next time you are thinking about buying or selling stocks, or mutual funds for that matter, ask yourself the question: “How’s the market?” Technical analysis—an understanding of how the markets work—may well provide some important insights.