On the non-linear dynamics of financial market risk and liquidity.

Reusch, Christian (2008) On the non-linear dynamics of financial market risk and liquidity. PhD thesis, London School of Economics and Political Science.

Preview

PDF Download (14MB) | Preview

Abstract

This thesis provides a novel empirical treatment of the dynamics of financial market risk and liquidity, two very important areas both for financial research as well as to practitioners in the financial markets: We devise empirical non-linear time series models of the two concepts that specifically take into account 'explosive', self-reinforcing dynamic patterns. While 'conventional' empirical models are often 'linear' and tend to neglect these effects, real-life evidence such as e.g. the 1987 crash, the large stock market drops on February 27th, 2007 or the huge losses posted by investment banks and hedge funds during July and August 2007, suggest that such an approach is warranted: In the first part of the thesis we extend a time series model of Value-at-Risk (VaR) with non-linear multiplicative features and endogenous regime thresholds. When estimated with a Markov Chain Monte Carlo (MCMC) method against real data, the resulting 'Self- Exciting Threshold CAViaR' (Conditional Autoregressive Value-at- Risk) model is able to detect trigger thresholds for explosive market risk as well as the scale of such a possible expansion in risk. The second part of the thesis is dedicated to the 'Conditional Autoregressive Liquidity' (CARL) model, a multiplicative time series approach to the empirical modelling of market liquidity. The newly con-ceptualised model is capable of picking up self-reinforcing dynamics, i.e. autoregressive patterns in liquidity, which is in accordance with theoretical research on the topic. Moreover, by incorporating a multidimensional liquidity proxy, the model CARL is explicitly designed to take into account the fact that liquidity is a concept with many facets, unlike other empirical treatments that often view liquidity only along a single dimension (e.g. the bid-ask spread, volume, trade duration). In this thesis, we demonstrate the empirical versatility of the model using both fixed interval data (daily and weekly) as well as tick-by-tick intraday data, for which we propose a filtering technique in order to be able to use the model in such a data environment. We note that the model is able to pick up autocorrelation structures in liquidity rather well and find the forecast performance very encouraging for practical use.

Item Type:	Thesis (PhD)
Uncontrolled Keywords:	Economics, Finance
Sets:	Collections > ProQuest Etheses Departments > Finance
URI:	http://etheses.lse.ac.uk/id/eprint/2728

Actions (login required)

Record administration - authorised staff only