Tests for identifying financial contagion : new theoretical approaches and empirical evidence

Abstract

This thesis explores the impact of financial contagion following the outbreak of the recent global financial crisis. It provides a new and unified approach to identify contagion. The first aim is to investigate for financial contagion by accounting for the existence of trends in linkages between markets, due to progressing globalisation, and allows for a description of the progress of contagion across the crisis period. Based on different reactions of domestic markets to financial shocks originating abroad, the occurrence of contagion is categorised into three types: “shock”, “recoupling”, and “kink” contagion. The results for a sample of 25 stock markets show that the impact of the 2007-9 crisis was largely heterogeneous and countries were not uniformly affected: those markets, which experienced contagion, were affected in various ways, and those, which did not suffer from contagion, experienced the crisis episode in various ways, too. The second objective is to examine contagion effect at a sectoral level from the world and domestic financial sectors across 25 countries and the findings show that the impact of the 2007-9 crisis was largely heterogeneous and the real economy was not uniformly affected. At least one sector of all countries in our sample was affected during the crisis, either by global or their local financial sectors. Moreover, there is also evidence of more instances of contagion effects in non-financial sectors of developed economies as compared to emerging ones, and the Basic Material sector was more vulnerable to shocks from both global and domestic financial sector relative to other sectors. The final part of this thesis proposes a new approach to test for financial contagion, which accounts for the existence of day-of-the-week effects in returns. The existence of day-of-the-week effects in contagion form the U.S. for twelve European countries before and during recent financial crisis using synchronised data is examined. The results show that countries did not experience contagion consistently during every day of the week; rather, excess co-movements happened only during certain days of the week. This model has the potential to disclose otherwise unobserved contagious effects, and to offer a more detailed picture of those, which could be identified using a more traditional approach.