Essays on sovereign default

Shi, Liang ORCID: 0000-0001-9680-9047 (2022). Essays on sovereign default. University of Birmingham. Ph.D.

Preview

Shi2022PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (5MB) | Preview

Abstract

This thesis consists of three chapters on sovereign default. The first chapter investigates the influences of fiscal austerity on sovereign debt yield spreads, which has caused heated debates since the Eurozone debt crisis in 2010s. The analysis in this chapter is based on an endogenous sovereign default model with private capital accumulation and fiscal rule. The model provides several contributions: First, it rationalizes the empirical evidence of state-dependent relationships between fiscal austerity and debt spreads: When the economy is under high financial stress characterised by high levels of outstanding debt and low productivity, government spending cuts increase spreads. By contrast, under low financial stress, fiscal austerity reduces spreads. Second, as in Greek data, the model predicts that the pre-default spread surge will be accompanied by fiscal consolidation. Third, it reveals a non-negligible role played by the wealth effect: if expected to be long-lived, austerity harms investment, damages production and eventually raises spreads. In conclusion, even though fiscal austerity could reduce sovereign spreads and debt-to-GDP ratios in the long term, its short-term self-defeating probabilities could be non-trivial.

The second chapter discusses efficiently deriving numerical solutions to macroeconomic models on Matlab with the GPU (Graphic Processing Unit) Parallel Computing toolkit. For many non-linear models such as the endogenous sovereign default model in chapter one, we resort to the discretized value function iteration (DVFI) approacht to obtain robust solutions. Unfortunately, this method is typically slow and the speed problem worsens when a high number of grid points for state variables is needed. This paper shows that the GPU toolkit on the commonly used Matlab platform provides an up to tenfold speed boost compared with using the conventional CPU method. Using the appropriate algorithm is important to achieve this and Matlab favours a combination of vectorization and serial execution, i.e. the Looping Over Exogenous Shocks (LOES) approach. With LOES, records show the solving time spent on Matlab CPU and GPU is significantly shorter than its Julia counterparts. Moreover, implementing GPU computation is easy on Matlab.

The third chapter studies the impacts of shifting long-run growth expectation on sovereign default risk. We show the new evidence of negative non-linear relationships between potential GDP growth forecasts and government debt spreads during the recent Eurozone sovereign debt crisis (2009-2016). Existing equilibrium default models assuming full information rational expectation (FIRE) on trend growth struggle to explain the new evidence because correlation between trend growth and simulated spreads is not statistically significant. In this paper, we build a new sovereign default model where the knowledge of trend growth is assumed to be imperfect and hence agents have to learn about it to make optimal decisions. The simulation results show that embedding such a learning mechanism in sovereign default model provides an easy solution to rationalize our new evidence.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):