The purpose of this thesis is to contribute to theoretical knowledge and help investors, fund administrators, financial regulators and database vendors. Its chapters examine hedge fund performance attribution and fund persistence under changing market conditions, and issues of hedge fund index engineering, using a rigorously constructed unified database (U.S. dataset, from 1990 to 2014). The core of my modelling approach is a custom piece-wise parsimonious multifactor model with predefined and non-defined structural breaks, flexible enough to capture differences in asset and portfolio allocations. This is implemented on a strategy, fundamental, and a mixed level. Concerning funds’ persistence, I use several parametric and non-parametric techniques whereas I develop a framework with mixed trading strategies for investors’ conditional high returns. I examine the classification problem of hedge funds by implementing several classification techniques used by database vendors, on the same dataset. The findings are robust, showing that during stressful market conditions most hedge fund strategies do not provide significant alphas to investors as fund managers are more concerned about minimizing their systematic risk and there is a switch from equity to commodity asset classes. Directional strategies have more common exposures than non-directional strategies under all market conditions. Falling stock markets are harsher than recessions for hedge funds. Moreover, during stressful conditions, small funds suffer more than large funds, young funds outperform old ones and funds that do not impose restrictions (and survive) outperform funds with no lockups. There are cases where funds can deliver significant negative alpha to investors conditional on stressful market conditions. In general, stressful market conditions have a negative impact on all types of funds’ persistence whereas my zero investment “synthetic” trading strategy can bring conditional high returns to investors. Furthermore, I found that the differences between index vendors are mainly due to the use of different selection criteria and different datasets.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:696083 |
| Date | January 2016 |
| Creators | Stafylas, Dimitrios |
| Contributors | Anderson, Keith ; Uddin, Moshfique |
| Publisher | University of York |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://etheses.whiterose.ac.uk/15252/ |
Page generated in 0.0026 seconds