Statistical testing of technical trading rule profitability

The profitability of technical trading rules has been widely studied in the literature and many researchers have found evidence that technical trading rules are able to generate profits in excess of those available from a simple buy-and-hold strategy. Taken at face value, this contradicts the Efficient Market Hypothesis and suggests that much of the classical theory of finance needs to be revised. However, despite the large volume of empirical literature on the subject, no clear consensus on the profitability of trading rules has emerged. Studies of US equity markets tend to find that trading rules were profitable prior to the mid-1980s, but there is little evidence of profitability after that time. Studies of other countries present mixed results – with evidence of profitability being found for some countries by some authors, but no evidence of profitability being found for many markets. There are a number of potential shortcomings of the existing literature which may explain its failure to provide consistent conclusions about the profitability of technical trading rules. Previously published papers have almost all focussed on a narrow range of technical trading rules. In fact, the majority of trading rules that feature in the practicioner’s literature on technical trading have received little or no attention in the academic literature. Similarly, the academic literature has focussed on a narrow range of markets. Not surprisingly, the US equity markets, and the other major markets, have received the largest share of academic attention. Several studies of non-major markets also exist, but they tend to concentrate on relatively small numbers of Asian and Latin American markets. Authors often neglect to report precisely which parameterizations of which rules are found to be profitable, and there are a number of variations of methodologyused, making it difficult to draw firm conclusions from the literature. Furthermore, the fact that the academic literature has considered only a small subset of all trading rules over narrow ranges of markets raises the possibility that interesting results are still waiting to be discovered. The credibility of many of the findings of trading rule profitability in the literature is compromised by frequent poor selection of statistical methodology. There exist many different classes of technical trading rules, and most rules can be paramaterised in any number of ways. Financial theory gives little guidance on which paramterisations of which rules should be profitable, so it is inevitable that empirical research must consider a large number of different rules. This raises statistical challenges. The classical methods of hypothesis testing control the probability of rejecting a true null hypothesis and are appropriate for testing a single hypothesis. When applied to multiple hypotheses, theprobability of rejecting at least one true null hypothesis is likely to be greater than the nominal significance level of the test. In cases in which a large number of hypotheses are tested, the probability of rejecting at least one true null hypothesis is unknown, but may be close to 3. This is referred to as data snooping. Its consequence is that studies that apply classical hypothesis testing to a large number of different parameterizations of different trading rules, perhaps for different assets and/or in different countries, and report the rejection of some hypotheses are not statistically valid. With the above comments in mind, the main objective of this thesis is to provide a comprehensive study of the profitability of technical trading rules which covers a broad set of rules over a wide range of markets, employing a consistent methodology and utilizing recent advances in multiple hypothesis testing that provide control of well defined error rates when testing large numbers of hypotheses. The thesis is structured as follows. Chapter 3 provides the background and outline of this thesis. Chapter 4 then surveys the relevant literature and discusses the motivations behind the development of the three key research questions addressed in Chapter 5 through 7, respectively. Chapter 3 examines the the profitability of technical trading rules in Australian financial markets (stock, currency and interest market) using tests which provide weak control of the family-wise error rate. In Chapter 4, we consider the US equity market over a period of more than a century using 54different classes of technical trading rule. I use statistical methods which provide strong control of the family-wise error rate and allow us to identify sets of profitable trading rules. Chapter 5 presents a cross-country study of technical trading rule profitability, in which the 54 trading rules introduced in Chapter 4 are applied to the equity markets of 39 different countries, 21 of which are classed as emerging markets by MCSI. In order to enhance power, we employ techniques that control generalized family-wise error rates and the false discovery proportion. Chapter 6 summarizes the key findings of this thesis along with some recommendations for future research. This thesis makes two distinct contributions to the literature on technical trading rules. Firstly, to my knowledge, it is the most comprehensive empirical study of technical trading rule profitability to date. It employs a wide range of technical trading rules, most of which have received little or no attention in the prior academic literature. These are applied to US data dating back as far as the late 1800s and to more recent data from 39 different equity markets. Secondly, I provide an illustration of the application of a number of recently developed statistical techniques for multiple hypothesis testing which are appropriate for the empirical analysis of technical trading rules. In particular, to my knowledge, this is the first study of technical trading rule profitability to control generalized family-wise error rates and the false discovery proportion. In each chapter of the thesis, I also produce results using standard hypothesis testing procedures and find many apparently spurious positive results. This illustrates the dangers of data-snooping in the analysis of trading rule profitability, and demonstrates the importance of the selection of appropriate testing methodologies.