By Paolo Manasse and Giulio Trigilia
What to do about Greece? So far, much of the debate has split into two camps:
- bail-in-ers, the advocates of coercive (but soft) restructuring of Greek debt, and
- bail-out-ers, those who favour the procrastination of EU-IMF lending plus “voluntary” rollover.
Little of this debate hinges on whether Greece is solvent or not. Everyone agrees that, barring a miraculous rebound in growth, Greece is broke.
It’s the contagion we are struggling against
The argument among in-ers and out-ers hinges on the fear of contagion. Should investors overreact and flee from Spanish and Italian debt, things could get very ugly, very fast. We are talking about widespread European sovereign debt and banking crises. This would force a choice between
- A full-blown monetisation;
- The break-up of the Eurozone.
The bail-out-ers’ argument is well exemplified by Lorenzo Bini Smaghi (2011) of the ECB Board :
…The third feature that generally sets sovereign and corporate debt workouts apart is the externalities they may generate … Imposing haircuts on private investors can seriously disrupt the financial and real economy of both the debtor and creditor countries.. This is particularly the case in a region like the Eurozone…,where [a default could] destabilise the Eurozone financial markets by creating incentives for short-term speculative behaviour.
Could a Greek default be contained?
While integration of the European financial and goods market is clearly key in the possible propagation of the crisis, the “Lehman” argument relies on the fear that markets, as has happened in many crises of the past, would cease to discriminate between borrowers. They would just sell off whatever smells “European”, possibly anticipating/precipitating a sharp depreciation of the euro (if the euro survives). Drawing from research in progress (Manasse and Trigilia 2011), this column argues that the fear of contagion may be exaggerated and that the available empirical evidence points to a decoupling of Greece from other problematic countries, such as Italy.
The first piece of evidence comes from the daily series of credit-default-swaps 5-year spreads of Eurozone sovereigns (Austria, Belgium, Denmark, France, Germany, Greece, Ireland, Italy, Netherlands, Portugal, Slovakia, Slovenia and Spain) from 4 September 2005 to 14 June 2011. CDS spreads are interesting precisely because their (high-frequency) cross-correlations are unlikely to reflect (low-frequency) changes in vulnerabilities, those stemming from market fundamentals such as trade and banks’ balance sheets. Therefore, they may reasonably represent market perceptions on default spillover risks across European sovereigns.
Figure 1 answers the following question: What percentage of the variance of Eurozone CDS spreads can be attributed, over time, to a “European- wide” dimension? It depicts the evolution of the first principal component of the Eurozone (standardised) spreads.1 Initially, this “common dimension” of EU sovereign risk is clearly imported from the US subprime crisis. In August 2008 Lehman reports huge losses on the sub-prime mortgages, and since then the Euro-wide component in sovereign volatility climbs from 30% to 90 % of total volatility. In September 2008 Lehman files for bankruptcy: and again the Euro-wide CDS component climbs back above 90% within a few months. Since April 2010, however, the Euro dimension of sovereign risk appears to be driven mainly by developments in Greece. In April Greek spreads rise sharply, and EU ministers patch up a first agreement on the bailout terms (11/4), the IMF later steps in (23/4), and S&P downgrades Greek bonds to junk (27/4). Since May, however, when the €110 billion EU-IMF package is finally approved, the European wide component has been consistently falling, pointing to a lower degree of “EU bundling” of sovereign risks.
A similar conclusion can be drawn from looking at the weight (the so-called factor loading) of the 5-year Greek CDS spread in the Euro-wide component.
Figure 2 shows that this weight, which had been slowly decreasing since 2007, takes a plunge in October 2009, at the time when Greek CDS spreads take off from the pack. At this time, effectively Greece decouples from the rest of the Eurozone, albeit temporarily. From April, 2011, however, Greece doesn’t affect significantly the Euro-wide spread.
What about the contagion to Italy, admittedly the mother of all defaults?
Figure 3 describes the rolling correlation among Greek and Italian standardised spreads. The two moved very closely together from the crisis outset in November 2007, but their correlation has been steadily declining since November 2010. The correlation is now in negative territory.
Finally, a similar analysis has been done for 10-year government bond yields (details available upon request). Result are similar to those reported with one exception: from November 2010 the share of the EU-wide component of bond yields’ total variance, which had fallen to about 60%, has climbed back substantially, suggesting more interdependence in this market.
We know all too well the recent trends may not necessarily extend to the future. Yet the evidence suggests that contagion has become less likely today relative to the past couple of years. After a relatively long period when markets bundled EU sovereign risks together, financial markets have started to discriminate more.
This means two things. First, an orderly restructuring of Greek sovereign debt is less likely to produce a disruptive “rush out of Europe” than in the past, and second, the fate of other problematic countries, such as Italy, rests, now more than ever, in their own hands.
Bini Smaghi, Lorenzo (2011), “Private sector involvement: From (good) theory to (bad) practice”, ECB.
Joliffe, IT (2002), Principal Components Analysis, Springer.
Kaminsky, Graciela and Carmen M Reinhart (2000), “The Center and the Periphery: Tales of Financial Turmoil”, GWU mimeo.
Manasse, Paolo and Giulio Trigilia (2011), “Contagion in Europe”, mimeo, University of Bologna and University of Warwick.
Rigobon, Roberto (2001), “Contagion: How to Measure it”, NBER Working Paper 8118.
1 Technically, we have divided the sample into 200-observations rolling windows, standardised each spread within each single window, and calculated the first principal component within each window (1053 in total). This variable is plotted in correspondence to the last observation of each rolling sub-sample. (For an exhaustive treatment of Principal Components see.Joliffe 2002, while for an early application to the literature on contagion see Kaminsky and Reinhart 2000. Standardisation is necessary since periods of high volatility would otherwise be associated to higher explained variance, see for example Rigobon (2001).