Not logged in.
Quick Search - Contribution
Contribution Details
| Type | Bachelor's Thesis |
| Scope | Discipline-based scholarship |
| Title | The Profitability of Technical Trading Rules on the Swiss Stock Market |
| Organization Unit | |
| Authors |
|
| Supervisors |
|
| Language |
|
| Institution | University of Zurich |
| Faculty | Faculty of Business, Economics and Informatics |
| Number of Pages | 60 |
| Date | 2016 |
| Zusammenfassung | Since the findings of Alexander (1961) that trends, once established, tend to persist more often than not academics and practitioner have searched for a set of rules to benefit from this market irregularity. While practitioners implemented technical trading rules from the early stages on academics were more skeptical towards any potential additional profit technical rules generate. Numberless authors examined the effectiveness of potential technical trading rules resulting in little support for any worthwhile implementation. One of the most cited research was done by Brock, Lakonishok and LeBaron (1992) who tested different common trading rules on the Dow Jones Industrial Index. They conducted standard statistical tests (double sided t-test) resulting in significant different daily rule returns compared to a benchmark buy-and-hold return of the index. As dependencies within the used return time series may have had altered their findings, they further back tested their results with bootstrap tests under various null models. Thereby they identified the influence of false distributional assumptions of standard tests. This revealed that examined rules do have predictive power over their data sample. Profitability however, was not granted as assumed transaction costs devoured the earned excessive return, making technical trading rules inferior to a simple buy-and-hold return. The suggested rules of Brock, Lakonishok and LeBaron (1992) were the main focus of numerous research papers considering different asset classes, different time periods and different markets. Academics tried to replicate the obtained results and to support the effectiveness of technical trading rules. Further information can be gathered through the work of Fama and Blume (1966), Lukac et al. (1988), Lo et al. (2000), Fernández-Rodríguez et al. (2000), Bessembinder and Chan (1995) and Isakov and Hollistein (1999). Many of these authors found support for the findings of Brock, Lakonishok and LeBaron (1992). Nonetheless, statistical errors were a possibility for obtained results as claimed by various studies. This is heavily supported by the findings of Fang et al. (2014) which challenged the results of Brock, Lakonishok and LeBaron (1992). They replicated the study of Brock, Lakonishok and LeBaron (1992) for two out-of-sample tests resulting in no statistical evidence of technical trading rule predictability/profitability. Hence, to ensure the quality of reported results studies should be back tested over out-of-sample test. This gets supported by numerous authors proposing different solution to potential statistical biases, e.g. White (2000) proposed a bootstrap method to identify the influence of statistical dependencies. Further Fama (1991) and Lakonishok and Smidt (1988) suggested the usage of long term new data sets. The purpose of this thesis is to reproduce the results of Isakov and Hollistein (1999). They replicated the research of BLL for the Swiss market and found support for the results of Brock, Lakonishok and LeBaron (1992); technical trading rules were predictive but not profitable for private investors. Institutional investors may be able to draw an excessive return from these rules, as their abilities excel compared to private investors. Hence, this thesis orientates on the work of Isakov and Hollistein (1999) enabling a comparison between their results and the results of this thesis. Therefore, similar technical trading rules have to be used. Isakov and Hollistein (1999) mainly tested the suggested rules of BLL on the Swiss Bank Cooperation General Index. Consequently, the same technical trading rules are analyzed in this thesis. However, to conduct an out-of-sample test the evaluation of set rules is taken place over the SMI™, SPI™ and UBS100™ indices. These replicate the Swiss market to its full extent and do provide long return data sets. Additionally, the three indices are tested in both price and total return notation to capture the influence of dividends and other repayments. Statistical tests on the chosen technical trading rules follow the methodology of Brock, Lakonishok and LeBaron (1992) and Isakov and Hollistein (1999). First statistical standard tests were conducted (double sided t-test) to analyze the daily conditional return difference. Second, dependencies within the dataset required bootstrap tests with a random walk null model. Profitability was tested through implementing various transaction costs levels depending on the assumed abilities of the investors. Private investors faced higher transaction costs than institutional investors based on their market access, trading volume and trading frequency. Further, opportunity costs of out of the market days were minimized by investing in a risk free asset. Statistical tests revealed that certain rules return statistical significant different returns than a buy-and-hold return (benchmark). Sell signals hereby tend to be more significant than respective Buy signals. Especially double method strategies (combined Moving Average rules and relative strength oscillators) yield significant different returns, as predicted by various authors. Differences between individual indices are not noteworthy, as similar technical trading rule specifications performed well over all three indices. However, the combined method approach provided unsteady results due to the major influence of risk free asset returns. Bootstrap tests implied that the distribution assumption of standard tests had small influences on the resulting t-statistics. However, as only random walk models were considered as null models these results have to be viewed with care. Testings with different null models may show deviant results, as they replicate the basic population to a better extent. All tested rules combined with a risk-free investment yielded a higher return over the benchmark both over the full sample and the two individual sub periods. However, if transaction costs were considered profitability of analyzed rules was mostly not sufficient enough to provide an excessive return compared to a benchmark. Institutional investors would have been able to achieve excessive profit due to their low fee levels. Individual rules (only compared method approaches) were even profitable for private investors. The resulting profit was however mainly due to the risk-free return, as combined method strategies established a negligible amount of trading signals. Results for various tested indices are again similar for price return notated indices. Total return indices yielded inferior returns than price return indices. As total return notation includes dividends and other repayments index losses are weakened whereas index gains strengthen. This results in slight higher gains in Buy signals but inferior conditional Sell returns leading to an overall inferior application of technical trading strategies. This is linked to the higher significance of Sell signals compared to Buy signals. Obtained results for standard statistical tests are in favor of the research conducted by Brock, Lakonishok and LeBaron (1992). Further there are similarities between the findings of Isakov and Hollistein (1999) and this thesis. Bootstrap tests however are returning slightly different results. This is probably due to either a not sufficient null model or a calculation error. Different null model analysis may clarify the results for the applied bootstrap. Stated results may not only come from additional benefit from technical trading rule application. Statistical, methodological or calculation errors are not excludable. As mentioned are results of standard statistical tests influenced by the basic normal distribution assumption of t-tests. Therefore a bootstrap analysis gives insight about the impact of biased assumptions. The application of the bootstrap method in this content gets suggested by Brock, Lakonishok and LeBaron (1992) as well as by White (2000). Methodological biases may occur due to the limited theoretical knowledge of the author. Therefore biased applications which stand out to an expert were unnoticed. Further errors may arise through incorrect calculations. As all calculations are based on Matlab 2016b, which uses programming scripts in order to perform calculations, simple errors can have major influences. Considering the presented results, programming errors are unlikely, as they are similar to previous research. |
| Export |
BibTeX
|