Finance - Theses

Permanent URI for this collection

http://hdl.handle.net/11343/409

Filters

2014 2
1 1
Reset filters

Settings
Statistics
Citations
End date: 2014 × Start date: 2014 ×

Search Results

Now showing 1 - 2 of 2
  • Item
    Thumbnail Image
    Do bank managers respond to debt-market discipline?
    Tan, Xin Yi ( 2014)
    The 2007/08 financial crisis reignited debates on what is a socially optimal capital structure for banks. One school of thought suggests that debt capital could be beneficial for banks because it provides a disciplinary mechanism over the decisions of bank managers. However, inconclusive empirical evidence of the existence of debt holder discipline creates a gap between the market discipline hypothesis and reality. Given the systemic importance of banks to the financial system and the high amount of debt in banks’ capital structure; it is important to determine if debt can make banks safer through disciplining the decisions of bank managers. The thesis examines the response of bank managers to the Federal Reserve Board’s 2008 clarification of the types of bank debt covenants that are legally enforceable. Unlike the prior empirical literature, which measures market discipline by changes in the price or quantity of debt, I follow the theoretical banking literature and measure market discipline as the ratio of bank debt with legally enforceable covenants to total liabilities. The market discipline that arises from debt covenants comes from the conditionally demandable threat to bank managers and the incentivised monitoring of the debt holders. By examining a sample of bank holding companies in United States between 2005 and 2011, I find evidence of bank managers reducing wealth transfer to equity holders and reducing bank fragility in response to increases in market discipline from the 2008 clarification. I also find that opacity of a bank’s activities reduces the response of bank managers, and that market discipline influences bank managerial actions up to four calendar quarters into the future.
  • Item
    Thumbnail Image
    Disaggregating the predictive information in the S&P 500 option-implied risk neutral distribution
    Low, Yi Ling Michelle-Joy ( 2014)
    This dissertation studies the information content of the S&P 500 index option-implied risk neutral distribution (RND). Our premise is that the long- and short-range persistence in the RND, related respectively to slow structural changes and transient risk factors, contain predictive information. Early literature on RND extraction spans a large number of estimation methods, but we propose extracting the RND using the highly flexible and implementable Generalised Lambda distribution (GLD). This extraction method is motivated by the need for a tractable method which is both straightforward to implement and yields economically interpretable RND moment estimates for financial applications. In a simulation study, we show that extraction by the GLD yields accurate estimates of the RND moments overall. This is true particularly of risk neutral skewness and kurtosis where scarce, discrete tail options price data may affect the estimation accuracy of the popular Bakshi, Kapadia and Madan (2003) model-free method. We test significantly for the presence of structural breaks in the RND moments and model these phenomena with long memory methods. Long memory methods serve as a convenient mathematical construct for capturing the effects of structural change in the RND moment dynamics. Analysing the RND moments in a fractionally integrated vector autoregression framework (FIVARX), we find evidence, consistent with previous literature, of significant co-moment relations. Furthermore, long-term RND persistence is related to macroeconomic influences, whereas short-term persistence is linked to short-term financial market volatility. The FIVARX modelling approach produces superior out-of-sample RND moment forecasts compared to alternative vector models. Finally, we apply Hidden Markov Models (HMMs) to assess the degree of predictive information contained in the RND for the future underlying asset price. HMMs are conditioned on RND-based measures of information derived from the long- and short-term RND moment persistence, as well as the RND quantiles. HMMs specified with long-term RND persistence and tail quantiles significantly outperform the benchmark HMM model in out-of-sample forecasts. Notably, our results suggest that RND-based information is predictive of the future market state governing the observed underlying asset price.