Hide/Show Apps

Analyses of factors of market microstructure: price impact, liquidity, and volatility

Karasan, Abdullah
First chapter of this thesis is an attempt to model the price impact by extending themodel proposed by Kyle [54]. It is assumed that the market is not perfectly efficientso that it takes to time to adjust new equilibrium price. Thus, in order to model theprice impact, two new concepts are introduced which are market resiliency and speedof price informativeness. It is showed that market resiliency and price impact tend toraise as speed of price information increases which emphasizes the fact that speed ofinformation matters in financial markets and market resiliency is not a phenomenonthat can be neglected.In the second chapter, it is tried to stress the importance of the liquidity which is con-sidered as the neglected dimension of the financial risk. To do that, a new approachcalled Liquidity Augmented Stochastic Volatility with Jump (LASVJ) model is in-troduced and it is compared with the Stochastic Volatility with Jump (SVJ) model interms of stability and performance. This analysis includes both simulation and cali-bration analysis. The simulation results suggest that LASVJ model outperforms SVJas it has lower bias and Root Mean Square Error. In the calibration part, ten compa-nies listed in Dow-Jones 30 are used and it is found that the estimated probability ofdefault and credit spread with LASVJ model are higher than those with SVJ model.The 2008 Crisis period is even aggravated this result. The findings imply that thevii probability of default and credit spread are underestimated if liquidity dimension ofrisk is neglected and this partly accounts for why 2007-2008 financial crisis and itsfull-scale effect could not be predicted.In the third chapter, it is aimed to improve the volatility prediction which includedin the financial risk management. As a well-performing volatility prediction shedslight on the uncertainty in the financial market, it is an important task to model it. Tothis end, GARCH-type models as well as SVR-GARCH model are used to model thevolatility and the results are compared based on the performance metrics. In part ofempirical analysis thirty stocks listedS&P-500 are included and the period coveredis between 01/01/2010-09/01/2019. The finding indicates that SVR-GARCH outper-forms the traditional models in predicting volatility and also produce more reliableresult in Value-at-Risk estimation based on Proportion of Failures and Basel’s TrafficLight Backtesting approaches.