This thesis presents research into the development and calibration of relative value fixed income trading models. The first chapter provides some background into the models studied, chapters two and three focus on calibration problems relating to an earlier version of the model: the relative value Nelson Siegel and Svensson model (rv-NSS). Chapter four introduces a more advanced version of the model, the relative value Dynamic Nelson and Siegel model (rv-DNS). Chapter five draws overall conclusions and discusses avenues for further research. Contributions to the literature: * Chapter 2 - Shows that Differential Evolution could be successfully applied to calibrate the rv-NSS * Chapter 3 - Compares the widest set of ridge regression estimators ever assembled - Modified (r-k) Class Ridge Regression (MCRR) did not specify how to estimate all of its parameters, two methods to address this were introduced - Improved Ridge Estimators (IRE) had convergence problems, chapter three tries to address these succeeding in the majority of conditions tested - Linearized Ridge Regression Estimator (LRRE) had estimation problems at the lowest volatility levels, an attempt was made to fix this - Although no one estimator dominated in every scenario tested the LRRE came closest to fulfilling that goal * Chapter 4 - Introduces a dynamic relative value trading model based on the Dynamic Nelson Siegel Model (DNS) introduced by Diebold and Li (2006) - This is the only relative value trading model based on the DNS - Successfully tests the model on simulated and real data Overall the thesis successfully introduces a functioning relative value fixed income trading model based on the Nelson and Siegel approach.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:716093 |
| Date | January 2015 |
| Creators | Patience, H. A. |
| Publisher | City, University of London |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://openaccess.city.ac.uk/17570/ |
Page generated in 0.0014 seconds