Debt and financial market contagion

Fry-McKibbin et al. (2019) consider a multiple-channel test of contagion, but the channels correspond to higher-order co-moments only.

To be clear, we consider contagion across equity markets based on debt characteristics rather than contagion across bond markets, such as considered in Gravelle et al. (2006).

Results would be mostly the same to three decimals given only 25,000 simulations.

The equity indices, expressed in US dollars, are collected from Datastream. The mneumonics are: Argentina – Argentina Merval price index; Australia - ASX200 price index; Austria - MSCI Austria price index; Belgium - BEL20 price index; Brazil - MSCI Brazilian price index; Bulgaria - Bulgaria Se Sofix price index; Canada - S&P Composite price index; Chile - Chile Santiago Se General price index; China – Shanghai Se A Share price index; Colombia – Colombia IGBC price index; Denmark - OMX Copenhagen price index; Finland - OMX Helsinki price index; France - CAC40 price index; Germany - MDAX Frankfurt price index; Greece – Athex Composite price index; Hong Kong - Hang Seng price index; Hungary – Budapest price index; India - CNX500 price index; Indonesia - IDX Composite price index; Ireland – Ireland Se Overall price index; Italy - FTSE MIB price index; Japan - NIKKEI225 Stock Average price index; Korea – Korea Se Composite price index; Malaysia - FTSE Bursa Malaysia Klci price index; Mexico – Mexico IPC price index; Netherlands - AEX price index; New Zealand – MSCI New Zealand price index; Norway - MSCI Norway price index; Peru - MSCI Peru price index; Philippines –PSEI price index; Poland – MSCI Poland price index; Portugal - MSCI Portugal price index; Romania – Romania Bet price index; Russia – Russia RTS price index; Singapore – MSCI Singapore price index; South Africa - FTSE All Share price index; Spain - IBEX35 price index; Sweden - OMX Stockholm price index; Switzerland – Swiss Market price index; Taiwan – Taiwan Se Weighed price index; Thailand – Bangkok SET price index; Turkey - MSCI Turkey price index; UK - FTSE100 price index; USA – Dow Jones industrials price index.

The country in the African region is South Africa. The 8 countries in the Americas region are Argentina, Brazil, Canada, Chile, Colombia, Mexico, Peru and the USA. The 13 countries in the Asian region are Australia, China, Hong Kong, India, Indonesia, Japan, Korea, Malaysia, New Zealand, Philippines, Singapore, Taiwan and Thailand. The 22 countries in the European region are Austria, Belgium, Bulgaria, Denmark, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Netherlands, Norway, Poland, Portugal, Romania, Russia, Spain, Sweden, Switzerland, Turkey and the UK.

The subprime mortgage crisis and Great Recession are separated by the severity of the collapse of Lehman Brothers on 15 September of 2008, and the Great Recession and European debt crisis are separated by the Greek bailout in the first quarter of 2010.

In order to avoid arbitrary selection of the pre-crisis and crisis periods, we consider Bai and Perron (2003) structural break tests to identify the non-crisis and crisis dates. These tests support the dates specified in Fry-McKibbin et al. (2014). We also perform the sensitivity analysis by taking the dates as one-month before and after the structural break on July 25, 2007 for robustness. These structural break and robustness test results are provided in Appendix D.

We filter the data in the same way as in Forbes and Rigobon (2002). Before conducting the contagion tests, a vector autoregressive (VAR) model is used to control for market fundamentals (country-specific and cross market relationships that always exist) and address any serial correlation. The VAR model is given bywhere \(R_{t}=[x_{t},y_{t}]^{\prime }\) is a transposed vector of returns across a set of equity markets during the non-crisis (\(x_{t}\)) and crisis (\( y_{t}\)) periods; \(\phi \left( L\right) \) is a vector of lags, and \(u_{t}\) is a vector of the residual terms. In order to deal with equity markets open in different time zones, two-day rolling average returns are used in the VAR model. The residuals \(u_{t}\) are treated as financial shocks and are used in the calculation of the correlation contagion statistic in (8), the co-skewness contagion statistics in (10) and (11), the co-volatility contagion statistic in (14), and the multiple-channel test of contagion statistic in (17). To conduct the contagion tests during the three episodes of financial crises from 2007 to 2013, the VAR model with 5 lags is estimated based on the sample period of 2005 to 2013. As for testing contagion during the COVID pandemic, a VAR model with 5 lags is estimated based on the sample period of 2019 to 2021.

The market cap of selected 44 countries are shown in Appendix E .

As the threshold of high- and low-debt groups is selected arbitrarily, we also consider the 70th and 60th percentiles for the robustness check. The results based on 70th and 60th percentiles are very similar to those of using 80th percentile in high-debt group.

We also consider the 70th and 60th percentiles as the threshold for the high-debt countries and the 30th and 40th percentiles for the low-debt countries for the robustness check.

Developed countries in the Americas region include: Canada, Chile and Mexico. Emerging countries in the Americas region include: Argentina, Brazil, Colombia and Peru. Developed countries in the European region include: Austria, Belgium, Denmark, Finland, France, Germany, Hungary, Ireland, Italy, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, Turkey and the UK. Emerging countries in the European region include: Bulgaria, Poland, Romania and Russia.