Assessment Methods And Early Warning Models Of Sovereign Default Risk

Debt crisis, triggered by Sovereign default, has a long history. Almost all of the countries had suffered in debt crises. These crises leads to economic downturn and serious defaults are even followed by decades of economic stagnation. The recent European debt crisis represents one of the major uncertainties threatening the current global financial stability. Thus, sovereign default risk evaluation and crisis early-warning are hot debate topics both in academic research and policy analysis.This dissertation studies sovereign default risk evaluation methods and crisis early-warning models, with focus on new model development and econometric method innovation. Three models are intensively discussed in this dissertation:early-warning method based on Probit panel model, on financial time series analysis, and theoretical sovereign default model.Major contributions are as following. Firstly, considering crisis contagion effects, we develop a crisis early-warning system based on spatial Probit panel model, where contagion effects are represented as spatial dependence within continent in the model. As results, we show that our model has larger likelihood value and the parameter representing contagion effect is statistically significant. Implementing this model in practice, we show that this model can predict various debt crises in Argentina, Brazil, Nigeria and Turkey during 1980 to 2009 correctly, by indicating that the crisis probabilities are much higher before sovereign default than in other periods.Secondly, we construct a time-varying MS-GARCH model, and use the changing probability of high volatility regime to predict the coming crisis. The case studies in East Asia countries during 1997 financial crises show that this model achieves good forecasting results. Especially, Singapore experienced a crisis probability jump from 6% to 20% half year before the 1997 financial crisis, and this probability is higher than 50% one month before the crisis.Thirdly, we develop a sovereign default model with asymmetric output shocks, acknowledging the fact that the real world shocks are mostly asymmetric. As results, the sovereign default probabilities and debt levels are significantly higher with asymmetric output shocks, and our model matches the Argentinean economy over period 1993Q1-2001Q4 quite well, with an overall data fitting better than Chaterjee and Eyigungor (2009). We show that our model can match high default frequency, high debt/output ratio and other cyclical features, such as countercyclical interest rate and trade balance in emerging countries. Additionally, we develop some useful tools in evaluating sovereign default risk and predicting debt crisis with the theoretical sovereign default model.We make contributions in all the three above mentioned models, and achieve better forecast performance. As to model comparison, we point out that all of these models have their own merits. It is recommended that various models are used in building the effective crisis early-warning system. Probit panel models suffer from country heterogeneity and crisis dynamics, while the performance of time series analysis depends on whether financial markets incorporate crisis information. Comparing to empirical models, the theoretical models have unique advantages in emphasizing decision making process of economic agents and crisis triggering mechanism in economics. By modeling deeply rooted economic dynamics, the theoretical models are widely considered as better suited for sovereign default risk evaluation.Moreover, we also make contributions in developing new econometric methods for parameter estimation. In chapter 3, we show modified MCMC methods to estimate parameters in spatial Probit panel model and in chapter 4, we develop a Griddy-Gibbs sampling method to estimate time-varying MS-GARCH model, which impove in 3 aspects comparing to Bauwens et. al. (2010). As results, our method shows quicker parameter convergence speed and less time consuming in calculation.As another contribution in econometric methods, we propose a simulation-based approach to approximate transition function of asymmetric output shocks between finite states, which is used in dynamic programming in obtaining the steady state. Comparing to Tauchen's method, our method is more flexible in transferring different shock process, which make it possible to analyze the impact of different shock process to the economic system.