On the study of contagion in the context of the subprime crisis: A dynamic conditional correlation–multivariate GARCH approach

Highlights

• We investigate contagion hypothesis for ten developed and emerging markets.

• We apply the DCC-Multivariate GARCH.

• We did not reject contagion for developed countries.

• Market comovements increased after the subprime crisis.

Abstract

This paper investigates the contagion hypothesis for ten developed and emerging stock markets (France, Italy, UK, Japan, China, Argentina, Mexico, Tunisia, Morocco and Egypt) with respect to the US market in the context of the subprime crisis. In order to capture further time-variation in contagion effects and dynamic linkages between stock markets, our paper makes use of a DCC–MGARH model. Our findings make two interesting contributions to the field. First, we highlighted an increase in dynamic correlations following the subprime crisis for most markets under consideration with regard to the U.S. market. Second, with respect to the contagion test by Forbes and Rigobon (2002) that associates “pure” contagion with a significantly higher correlation between markets during the crisis, our results showed that the “pure” contagion hypothesis is not rejected for France, Italy, the UK or Mexico at the level of 1%, and for Argentina at 10%. Simple interdependence with the US market characterizes the other markets. Our findings thus offer interesting insights in terms of investment and diversification, and offer further insights into systemic risk in these regions.

Introduction

Stock markets experienced were extremely bullish from the 1990s onwards. This trend was facilitated by increased access to equity markets, the significant development of financial assets and capital mobility, excess demand by investors, etc. Accordingly, stock prices reached record levels in the US and Europe. Moreover, emerging stock markets underwent several financial reforms and liberalization measures, and have experienced further significant increases in the two last decades. However, such equity market development came to an abrupt end, with stock markets suffering major downturns and severe corrections during the periods of financial crises and crashes. For example, the Dow Jones lost 22.6% in October 1987; the French CAC40 index lost 15% after the Asian crisis in 1997, 21.97% in 2001 and 33.75% in 2002; and the Nasdaq lost 75% in 2000 due to the Internet Bubble. More recently, the subprime crisis also led to large losses for several equity markets: Dow-Jones (6%), CAC40 (12%), Nikkei225 (16.15%), etc.

Interestingly, the major stock markets often seem to evolve in line with one another during periods of both increase and decrease, providing further evidence of strong co-movements and interdependency. These co-movements have been analyzed in different ways by scholars. Indeed, while some authors consider them as a form of financial integration,¹ others have associated them with contagion effects. In the latter case, at least four possible definitions of the concept have been posited (Forbes and Rigobon, 2000). First, contagion reflects excessive shock transmission with regard to fundamental variations. Second, contagion is the result of shock transmission induced by investor behavior. Third, it results from market co-evolution due to external factors. Finally, shift contagion denotes an increase in market linkages due to shock effects that are greater in crisis period than in normal times.² In this paper, we focus on shift contagion in relation to the subprime crisis.

Several authors have studied contagion in its different forms using different econometric tests: e.g., linear cointegration tests, nonlinear error–correction models, and dynamic correlations (Baur, 2003, Chiang et al., 2007, Corsetti et al., 2005, Forbes and Rigobon, 2002, Kaminsky and Reinhart, 2000, Yang and Bessler, 2006). However, findings to date are relatively inconclusive, and often vary per country, per sample and per period. Indeed, Calvo and Reinhart (1996) tested contagion during the Mexican peso crisis and found evidence of contagion with Latin American markets. Baur (2003) did not reject the contagion hypothesis during the Asian crisis, while Serwan and Bohl (2005) found less evidence of contagion for European stock markets over the period 1997–2002. Using dynamic correlations, Chiang et al. (2007) found further evidence of asymmetrical contagion effects for nine Asian markets over the period 1990–2003. Iwatsubo and Inagaki (2007) also pointed to further evidence of contagion between the US market and Asian markets. More recently, Idi Cheffou et al. (2014) did not reject contagion between the US and three major European stock markets (Germany, the UK and France) following the investigation of volatility spillover. Jawadi et al. (2014) extended this study in two ways. On the one hand, they investigated contagion between the same international equity markets using jumps that were identified through realized volatility and high-frequency data. On the other hand, the authors tested contagion effects during overlapping and non-overlapping trading hours, again using high-frequency data. Accordingly, the authors pointed to a leader effect for the US market during overlapping hours, while regional (European) contagion effects seem to be more significant during non-overlapping trading hours.

Following the subprime crisis (2007), the global financial crisis (2008–2009) and the great economic depression (2009), several market losses and bank bankruptcies were seen to carry considerable systemic risk, severe sovereign risk and major debt and confidence crises for several developed and emerging economies. In order to limit the effects of such crises, save the banking system and financial markets, improve risk management, and control credit and the financial markets, several new financial regulation measures and reforms were introduced by Basel III and the European commission. This recent new framework makes contagion a hot topic and a key challenge. Accordingly, a growing body of literature has investigated the issue to help understand its origins and characteristics and to try to prevent it from occurring in the future. For example, Rim and Setaputra, 2007, Rim and Setaputra, 2008 applied a GARCH model to analyze the stock market linkages in South East and East Asia before and after the Asian financial crisis. The authors pointed to further evidence of contagion. Horta et al. (2008) analyzed the contagion effect of the US subprime mortgage crisis on Japanese, Canadian, Italian, Portuguese, UK, German and French stock exchange markets using copula models. The authors confirmed a significant contagion effect for Italy, Japan, Canada, the UK, and France, but not for Germany. Hwang et al. (2010) studied the contagion effects of the US real estate downturn on a couple of global stock markets using a DCC–GARCH model. Their findings showed a significant contagion effect for both emerging and developed markets.

Overall, authors have identified an increase in stock market correlations since the subprime crisis, suggesting further evidence of the contagion effect both during and after the subprime crisis. This can be justified in at least two ways. First, given the circumstances of the great depression, investors have tended toward mimetic behavior and have been more influenced by market opinion than by market fundamentals (Orléan, 1990). Second, as several policymakers, economists and monetary institutions have begun sharing new joint projects that aim to reshape finance and the control of the banking system and financial institutions (Aglietta, 2009, Shiller, 2008), this could be another source of common behavior for investors, and therefore for stock markets.

However, while the crisis effects pointed a priori to the presence of time-varying linkages, several authors have tested contagion through static correlation or linear cointegration tests. The nonlinear cointegration approach might appear more appropriate, but the presence of short historical data after the subprime crisis (data spanning only 7 years, from 2007 to 2014) makes these tests less suitable. To fill this gap, our paper applies a Dynamic Correlation Multivariate GARCH (DCC–MGARCH) model to test contagion in the context of the subprime crisis. This approach has at least two advantages. On the one hand, it helps to test contagion during stable and turbulent (crisis) periods, and to reproduce dynamic cross market correlations. It also enables contagion to be asymmetrically transmitted from one market to another, and allows for time-varying contagion effects. In addition, previous studies focused either on developed markets and/or on major emerging markets, while our study focuses on developed markets as well as on second-level emerging markets. Finally, the application of a contagion test using Forbes and Rigbon's (2001) dynamic conditional correlations is particularly original as it enables us to distinguish contagion from market interdependency.

The paper is organized into four sections. The second section presents the data and the econometric methodology. The empirical results are discussed in the third section. The last section concludes.