PREDICTION OF FINANCIAL CRISES IN SOUTHERN EUROPE AND IRELAND

Abstract

Com esta dissertação pretendemos analisar, no rescaldo da crise financeira internacional de 2008-2009, os fatores fundamentais que nos países do Sul da Europa e na Irlanda contribuíram para potenciar os efeitos desta crise financeira e da subsequente crise da dívida soberana da Zona Euro. Estudámos a evolução destas economias nos últimos séculos, com enfoque nos períodos de crise financeira, discutindo o contexto da arquitetura da União Económica e Monetária e as diferentes circunstâncias e as distintas políticas económicas implementadas em cada país. Para investigar os principais determinantes das crises financeiras, e em particular dos eventos de crises bancárias, aplicámos abordagens econométricas distintas. No capitulo 2, que dedicámos ao estudo das crises bancárias na Irlanda e Espanha, utilizámos um modelo logit binominal, e em anexo, mostrámos a tentativa de utilização de um modelo KLR. No capítulo 3, estudamos os eventos de crises financeiras, onde incluímos crises da dívida soberana, crises cambiais, crises de inflação e crises bancárias, na Grécia e em Portugal. Neste trabalho voltou a ser aplicado o modelo logit binomial, e adicionalmente utilizamos dois modelos logit multinominais. No capítulo 4, aplicámos uma nova técnica de machine learning chamada Support Vector Machine, com diferentes tipos de funções Kernel. Em ambos os trabalhos usamos um conjunto de dados robustos para 69 países, entre 1960 e 2016, e um conjunto alargado de variáveis macroeconómicas relevantes: taxa de crescimento do Produto Interno Bruto (PIB); termos da troca; taxa de depreciação da taxa de câmbio; taxa de juro real; taxa de inflação; défice; crédito ao sector privado, e PIB per capita. Para estudar outros tipos de crises financeiras, que não só as crises bancárias, foram utilizadas também como variáveis explicativas a dívida pública e saldo da balança de transações correntes. Além da componente retrospetiva da análise, em relação aos fatores que contribuíram para as crises financeiras passadas, discutimos o estado atual destas quatro economias, com atenção especial para os principais impactos da pandemia de Covid-19. Principalmente, tentámos ajudar a responder a algumas questões: Quais são os melhores modelos para a previsão de crises financeiras? Quais são as variáveis macroeconómicas mais relevantes na previsão do despoletar de crises financeiras? É uma nova crise financeira (grave) mais provável ou menos provável de acontecer em breve do que há dez anos, quando a crise financeira internacional abriu caminho para a crise da dívida soberana? Dada a arquitetura da Zona Euro, quais são as opções de política económica disponíveis para as autoridades nacionais que buscam reduzir a probabilidade ou o impacto de uma nova crise? Quais foram os impactos principais da pandemia Covid-19 sobre o risco de uma nova crise financeira? Existem diferenças substanciais entre Irlanda, Espanha, Grécia e Portugal? ix Abstract In this dissertation, we provide an analysis of the fundamental factors that, in the countries of Southern Europe and Ireland, contributed to potentiate the effects of the aftermath of the 2008-2009 international financial crisis, and subsequently led to the sovereign debt crisis in the Eurozone. We examine the evolution of these economies in recent centuries, focusing on periods of financial crisis. We discuss the consequences of the context provided by the architecture of the Economic and Monetary Union, as well as the different circumstances and different economic policies implemented in each country. To investigate the main determinants of financial crises, and in particular the events of financial crisis, we applied several quantitative approaches. In chapter 2, which we dedicated to the study of banking crises in Ireland and Spain, we used a binomial logit model. In the appendix to that chapter, we show the result of applying a KLR model to that issue. In chapter 3, we study the events of financial crisis in a broad sense - where we include sovereign debt crises, currency crises, inflation crises, and banking crises - in Greece and Portugal. In this chapter, again we employed the binomial logit model, and additionally, we used two multinomial logit models. In chapter 4, we applied a machine learning technique called Support Vector Machine, with different types of kernel functions. In all chapters, we used a large data set, covering 69 countries between 1960 and 2016, and containing a broad set of relevant macroeconomic variables: Gross Domestic Product (GDP) growth rate; terms of trade; exchange rate depreciation rate; real interest rate; inflation rate; deficit; credit to the private sector, and GDP per capita. In chapters 2 and 3, we also used as explanatory variables the public debt and the current account balance, both in percentage of GDP. In addition to the retrospective component of the analysis, concerning the factors that contributed to past financial crises, we discuss the current state of these four economies, with special attention to the impacts of the Covid-19 pandemic. Mainly, we tried to help to answer some questions: What are the best models for forecasting financial crises? What are the most relevant macroeconomic variables in predicting the onset of financial crises? Is a new (severe) financial crisis more likely or less likely to happen soon than ten years ago, when the international financial crisis paved the way for the sovereign debt crisis? Given the architecture of the Eurozone, what are the economic policy options x

available to national authorities seeking to reduce the likelihood or impact of a new crisis? What were the main impacts of the Covid-19 pandemic on the risk of a new financial crisis? Are there substantial differences between Ireland, Spain, Greece, and Portugal?

In this dissertation, we provide an analysis of the fundamental factors that, in the countries of Southern Europe and Ireland, contributed to potentiate the effects of the aftermath of the 2008-2009 international financial crisis, and subsequently led to the sovereign debt crisis in the Eurozone. We examine the evolution of these economies in recent centuries, focusing on periods of financial crisis. We discuss the consequences of the context provided by the architecture of the Economic and Monetary Union, as well as the different circumstances and different economic policies implemented in each country. To investigate the main determinants of financial crises, and in particular the events of financial crisis, we applied several quantitative approaches. In chapter 2, which we dedicated to the study of banking crises in Ireland and Spain, we used a binomial logit model. In the appendix to that chapter, we show the result of applying a KLR model to that issue. In chapter 3, we study the events of financial crisis in a broad sense - where we include sovereign debt crises, currency crises, inflation crises, and banking crises - in Greece and Portugal. In this chapter, again we employed the binomial logit model, and additionally, we used two multinomial logit models. In chapter 4, we applied a machine learning technique called Support Vector Machine, with different types of kernel functions. In all chapters, we used a large data set, covering 69 countries between 1960 and 2016, and containing a broad set of relevant macroeconomic variables: Gross Domestic Product (GDP) growth rate; terms of trade; exchange rate depreciation rate; real interest rate; inflation rate; deficit; credit to the private sector, and GDP per capita. In chapters 2 and 3, we also used as explanatory variables the public debt and the current account balance, both in percentage of GDP. In addition to the retrospective component of the analysis, concerning the factors that contributed to past financial crises, we discuss the current state of these four economies, with special attention to the impacts of the Covid-19 pandemic. Mainly, we tried to help to answer some questions: What are the best models for forecasting financial crises? What are the most relevant macroeconomic variables in predicting the onset of financial crises? Is a new (severe) financial crisis more likely or less likely to happen soon than ten years ago, when the international financial crisis paved the way for the sovereign debt crisis? Given the architecture of the Eurozone, what are the economic policy options available to national authorities seeking to reduce the likelihood or impact of a new crisis? What were the main impacts of the Covid-19 pandemic on the risk of a new financial crisis? Are there substantial differences between Ireland, Spain, Greece, and Portugal?