Contagion Risk During the Euro Area Sovereign Debt Crisis: Greece, Convertibility Risk, and the ECB as Lender of Last Resort

Abstract

Mario Draghi’s famous “Whatever it takes” speech is widely credited as ending the euro area’s sovereign debt crisis by firmly establishing the ECB as the lender of last resort in the euro area and thereby eliminating what is referred to as “convertibility” risk premia. This chapter uses a dynamic conditional correlation model and finds that (non-fundamentally driven) contagion originating from the Greek debt crisis was to some extent responsible for those risk premia.

Appendices

Appendix 1: Descriptive Statistics of Bond Yield Data

Table 5 Descriptive statistics for 10-year benchmark government bond yields between 01.01.2009 and 15.12.2011 for Greece, Ireland, Italy, Portugal, Spain and Germany

Appendix 2: Principal Component Analysis

The results of the principal components analysis are presented in Table 6. In total, five principal components are calculated. By filtering the first principal component, we adjust the original time series for a factor which explains 90% of the joint variation. This percentage is reasonably high to conclude that the global factor, i.e. the most important joint driver of all original yield series, is approximated and extracted by the first principal component.

Table 6 Principal component analysis

Appendix 3: DCC Coefficients

The estimation results of the DCC model proposed by Engle (2002) and Engle and Sheppard (2001) are provided in Table 7. The modified input variables from Sect. 4.2 are statistically adequate for the DCC model. Based on the single time series, univariate GARCH(1,1) equations are estimated for each country in a first step. Conditional volatilities are assumed to be represented by one lagged news parameter and one lagged decay parameter. In a second step, the GARCH residuals are applied to the multivariate MGARCH(1,1) estimation which models the dynamics of the comovements. The conditional correlations are also modelled with one lagged news parameter and one lagged decay parameter. All parameter estimates of both the univariate and the multivariate estimation are highly significant.

Table 7 DCC parameter estimates: Columns two and three refer to the lagged news parameter (ARCH 1) and the lagged decay parameter (GARCH 1) of the six countries’ univariate GARCH(1,1) estimation and the multivariate MGARCH(1,1) equation

Appendix 4: Country Ratings

Sovereign debt rating downgrades of Ireland, Italy, Portugal or Spain are used as control variables in rating regression (5). An overview of the rating development of those four countries is provided in Fig. 6. No rating development for Germany is displayed as within the observation period there was no downgrade for German sovereign debt which is rated as AAA. The linear rating scale used in Figs. 4 and 6 is described in Table 8.

Fig. 6

Irish (upper left subplot), Italian (upper right subplot), Portuguese (lower left subplot) and Spanish (lower right subplot) government debt ratings by the three leading rating agencies. Sources: Moody’s, Fitch and S&P

Table 8 Scaling of rating Figs. 4 and 6

Appendix 5: Results of Ratings Regression Equation (5) for Control Ratings

Table 9 provides estimates for the control rating dummy from rating regression (5). The respective country’s dummy is used as control variable, that is e.g. in the Irish case the Irish rating dummy serves as control variable in the rating regression of the Greek-Irish correlations. No rating dummy is included in the rating regression for the Greek-German correlations as German debt was not downgraded during the observation time span.

Table 9 Rating downgrade regression estimates: Control rating dummy parameter estimates of the respective other country from Eq. (5) are shown for dynamic correlation series of Irish, Italian, Portuguese and Spanish 10-year benchmark government bond yields vis-à-vis Greek 10-year benchmark government bond yields