Date of Award

August 2022

Degree Type

Dissertation

Degree Name

Doctor of Philosophy

Department

Economics

First Advisor

Kundan N. Kishor

Committee Members

Rebecca Neumann, Jangsu Yoon, Vivian Lei

Keywords

Credit Gap, Early Warning Indicator, Forecast Combination, Model Average, Property Prices, State-Space Model

Abstract

My dissertation studies credit expansion and its effect on house prices and financial stability. In the first chapter, I examine the idea that house prices and credit to households are jointly determined, affecting each other in the short and long run. I decompose the movements of the two variables of interest into permanent long-run and transitory short-run components using an unobserved components vector autoregressive model. The dynamic model shows findings to support the hypothesis that a short-run positive shock to house prices is associated with an increase in household credit above its long-run trend. Furthermore, by utilizing additional information generated by the unobserved component model, our multivariate model performs better than univariate models in capturing the dynamics of household credit and house prices over the last three decades, especially during the recent financial crisis. I also estimated the predictive ability of cyclical components of a variable on their counterparts from other variables by employing cross-correlation coefficients in the VAR model.

In the second chapter, I propose a new method to measure the credit gap - the deviation of the credit-to-GDP ratio from its long-run trend. Here, I utilize the idea proposed in Nelson (2008) that the deviation of a non-stationary variable from its long-run trend should predict future changes in the variable. Since different trend-cycle decomposition methods of credit-to-GDP ratio provide different credit gap measures, I handle the model uncertainty by assigning weights to these different credit gap measures based on their relative out-of-sample predictive power based on Bates and Granger (1969) forecast combination method. I apply this approach to estimate the UK and the US credit gap using credit-to-GDP ratio data from 1960-2020.

My proposed credit gap measure dominates the alternate credit gaps, including the one provided by the Bank of International Settlements (BIS) regarding its relative out-of-sample predictive power. The proposed gap also has superior features in terms of early detection of turning points and relative insensitivity to the endpoint problem.

The third chapter of my dissertation overcomes model uncertainty in using the credit gap as an early warning indicator (EWI) of systemic financial crises in a binary outcome setting. I propose using model averaging of different credit gap measurements to achieve a better averaged model fit and out-of-sample prediction. I use binary logistic regression in a panel setup consisting of 40 countries. In this paper, I also propose a novel, superior criteria to judge the performance of an EWI than the one currently popularly used in the literature. The empirical results showed that our Bayesian averaged model could synthesize a single credit gap that outperforms other popularly studied credit gap measurements in terms of an early warning indicator.

Included in

Economics Commons

COinS