الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 206 |  |  |
| | | from | 206 | | |
Abstract
Statistical arbitrage, which is the subject of this study, covers a variety of short-term trading and investment strategies that employs numerous parametric and nonparametric statistical techniques to identify possible relative mispricing between pairs of stocks thereby generating positive and market neutral returns. This study aims to uncover potential arbitrage opportunities present in the Egyptian stock market (EGX) by constructing trading algorithms based on combining different statistical techniques and investigating whether the risk and returns generated using these techniques can be improved by creating and implementing a dynamic trading threshold. It is worth mentioning that this is the first study to investigate the performance of such strategies in the Egyptian stock market.
The study covers the period from January 2019 to December 2020 using the daily adjusted closing prices of stocks that are eligible by EGX for short selling. The in-sample formation period in which the arbitrage portfolios are constructed is set to 12 months and the out-of-sample trading period (testing period) in which the portfolios are traded is set to 6 months. The formation and trading periods are rolled over every month, thereby creating 7 trading cycles. The approaches applied are the nonparametric distance approach pioneered by Gatev et al. 1999, Johansen cointegration approach, and Principal component analysis combined with hierarchical clustering analysis using Ward’s method (PCA/HCA).
In all approaches, fixed and dynamic trading thresholds for generating trading signals are constructed and implemented. The fixed threshold is based on a standard deviation metric. The dynamic threshold is based on the conditional standard deviation that is estimated recursively using a GARCH model. In addition, sensitivity analysis is performed to investigate the effect of the trading threshold value on trading statistics and returns generated by the different approaches. All algorithms, analysis, and model building are performed using the R programming language.
The results show that all approaches generate positive excess returns when applied to the Egyptian stock market. The cointegration approach outperforms other approaches in terms of excess returns. In terms of risk and market neutrality of the returns, the PCA/HCA approach is superior to the other approaches. The excess returns generated show a low level of risk with market neutral returns, as evidenced by the insignificant low market beta value. The study also finds that generating trading signals based on the dynamic threshold does not improve returns but achieves a risk reduction in most of the approaches. Sensitivity analysis shows that higher trading threshold values generate higher excess returns but on the other hand, substantially reduces the percentage of pairs traded and increases the strategy’s risk. Thus, more consideration should be given to the choice of the trading threshold value.
The trading algorithms created, and approaches applied can be exploited by traders, investors, and fund managers to generate positive short-term, market neutral returns while hedging risks.