 |
||||||
 |
||||||
 |
 |
 |
 |
 |
 |
 |
|
||||||
|
|
|
||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||
|
Essays on Interest Rates and Credit Risk Martin Vojtek Date of defense: July 24, 2009 Dissertation Committee: Evžen Kočenda (chair) Petr Zemčík Viatcheslav Vinogradov
Ronald Anderson Abstract:
This dissertation addresses inefficiencies and problems in the financial
markets of post-transition countries, which denies the use of standard
estimation techniques. It focuses on interest rate markets and empirically
analyzes the situation in the countries that joined the EU in May 2004. These
countries underwent significant changes over the last two decades and markets
in these countries are often not stable and not developed. In my dissertation I
am conducting research in the areas where the empirical results are very
scarce. A deeper understanding of the specifics in the markets of
post-transition countries can be very helpful for example in designing policy
measures touching these markets. Chapter 1 deals with one the
specifics of post-transition countries, namely non-existent or very small
markets with certain types of financial products, in this case the derivatives
of interest rates. However, due to infrequent or non-regular trading, the
prices do not contain sufficient information (or the prices are not quoted at
all) and the implied volatility approach to calibration cannot be used. The
paradigm used in Chapter 1 is the Brace-Gatarek-Musiela model of interest rates
that models the evolution of LIBOR (London InterBank Offered Rates)-type market
interest rates together with the Orthogonal GARCH. The BGM model is among the
most widely used no-arbitrage type of model and its correct calibration is
crucial in the calculation of the correct prices of financial instruments based
on interest rates. The paper builds on calibrated models for the Visegrad 4 countries
and an analysis of interest rate markets with shorter-end maturities is
performed. It answers the question to what extent are these models reliable for
pricing derivatives in transition markets. Chapter 2 is a part of the research conducted at the Czech National Bank
to measure how market participants perceive the prospects of enlarging the euro
area for the four Visegrad countries that joined the EU in May 2004. The
traditional methods to estimate the probability of EMU enlargement for post-transition
countries cannot be used or there are serious limitations. Therefore a method
based on the state space model is developed and used in this chapter. Lund
(1999) builds on the equilibrium interest rate model of the Vasicek type, where
first the so-called ``true local spreads", i.e. the spreads of the local
(domestic) interest rates to the euro interest rates that would occur in the
case that no anticipated entry of domestic country into EMU is possible, are
estimated. Based on the knowledge of the true local spread and the actual
spread it is possible using the Kalman filter to infer how likely is the entry
of the domestic country into the EMU zone. In the Chapter 3 we developed an optimal specification of the credit
scoring model to analyze data on loans at the Czech retail banking market. We
employed two approaches: parametric (logistic regression) and non-parametric
(Classification and Regression Trees, or CART). Along with analyzing our results
we also aimed to assess the determinants of default behavior. We construct
three different models using logistic regression and one model using CART and
compare these models in terms of efficiency and power in discriminating between
bad and good clients. We were able to detect the most important characteristics
of default behavior. Both methods are robust: they found similar variables as
determinants. Further, we show that socio-demographic variables are important
in the process of granting credit and therefore such variables should not be
excluded from credit scoring model specification.  |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
