Determinants of sovereign bond yield spreads in the EMU: An optimal currency area perspective
Introduction
The sovereign debt crisis, which escalated in the European Monetary Union (EMU) in 2010, has sparked big debates about its causes and possible solutions, both in academia and in policy institutions. Since the start of the EMU and before the financial crisis, spreads on 10-year sovereign bond yields relative to the German benchmark were small.2 With the financial crisis the picture completely changed. By the spring of 2009 the Greek sovereign bond spread had reached almost 300 basis points and by 2010 it had skyrocketed to over 1000 basis points (see Fig. 1). Investors started to question the ability of certain EMU governments of meeting their debt obligations and began requiring higher and higher risk premia.
What are the determinants of sovereign bond yield spreads in the EMU? The empirical literature has identified both a common international time-varying factor – commonly dubbed as international risk aversion – and country specific factors – in particular default and liquidity risk – as potential determinants of sovereign bond yield spreads in the EMU. However, both in academic debates and in the context of policy-making, no clear consensus has arisen. As far as the default risk is concerned, Faini (2006), Hallerberg and Wolff (2008) and Bernoth et al. (2012) find that the budget balance and the stock of government debt have, on average, a significant impact on sovereign bond spreads, whereas Codogno et al. (2003) find that public debt plays a role only for Italy and Spain. As regards the liquidity risk component, Codogno et al. (2003) and Sgherri and Zoli (2009) find that liquidity explains only a small fraction of sovereign spreads, while Gomez-Puig (2006) and Barrios et al. (2009) show that liquidity is more important to explain euro area sovereign spreads. With respect to international risk aversion, Attinasi et al. (2010) show that this factor has substantially contributed to the change in sovereign bond spreads during the financial crisis.
This paper adds to this debate by taking a long-run approach to the analysis of the determinants of sovereign bond yield spreads in nine EMU economies (Austria, Belgium, Finland, France, Greece, Italy, the Netherlands, Portugal and Spain) relative to Germany, by looking at the issue from the viewpoint of the theory of optimal currency areas (OCA).3 In particular, we argue (i) that long-run determinants of sovereign bond spreads are more relevant for policy-makers when they have to decide whether, and to what extent, structural policy interventions are needed to reduce sovereign bond yield differentials and (ii) that investors take OCA issues, and in particular diverging competitiveness among EMU members, seriously into account when they have to assign and price sovereign default risk.
In order to take the first point into consideration – and this is also the first innovation of the paper relative to the existing literature – we base our investigation on recently developed panel cointegration techniques that treat cross-sectional dependence via factor models, allow for potential breaks, and are robust to endogeneity. In fact, as regards cross-sectional dependence, we conjecture – and empirically test – that aspects of country interdependence, such as the economic and financial integration processes, the Maastricht convergence criteria, and the common monetary policy framework, cannot be neglected. In addition, given the evident shift in the level of sovereign bond yield spreads experienced during the financial crisis and the subsequent sovereign debt crisis (reported in Fig. 1), we believe that any analysis dealing with the determinants of sovereign spreads should take potential breaks into account. In this paper, we tackle these issues by testing for panel cointegration with break using the approach of Westerlund and Edgerton (2008).
As far as the second point is concerned – and this is also the second novel feature of the paper – in addition to the standard measures of default and liquidity risk, we include cumulated inflation differentials among our explanatory variables, to capture asymmetric shocks leading to a divergence in competitiveness. In fact, even small differences in inflation rates, if persistent, can lead to sizable changes in relative price levels. As shown in Fig. 2, since the start of the monetary union, cumulated inflation differentials among EMU countries have persistently diverged. As noted by Estrada et al. (2012), in principle, persistent inflation differentials may both be a benign phenomenon explained by a structural convergence process according to a Balassa–Samuelson type of argument, and the source of long-lasting and damaging losses of competitiveness.4 In order the former type of argument to hold, however, inflation rates should be positively correlated with the difference between labor productivity growth in the traded versus non-tradable sectors. While there is some evidence that this effect can justify some inflation differentials in the euro area, a consensus seems to have emerged around the claim that the Balassa–Samuelson hypothesis cannot be the general explanation of the persistent inflation differentials across EMU members (ECB, 2003, Estrada et al., 2012). In particular, Estrada et al. (2012) argue that the heterogeneous inertial components of price and wage-setting rules across the EMU, such as those caused by wage indexation clauses, play a predominant role. Given the EMU fixed exchange rate regime, countries that have experienced persistent positive inflation differentials have been subject to an appreciation of the real exchange rate. As noted by De Grauwe and Ji (2012), a country experiencing a real appreciation is likely to bump into problems of competitiveness which in turn may lead to current account deficits and debt problems.5 Regardless of the source of the imbalances, the appreciation of the real exchange rate for some EMU members has represented a gradual large asymmetric shock.6 As a result, one of the theoretical conditions for an OCA, which requires that a shock in one country should be sufficiently correlated with that in the rest of the union, or that the union has put in place measures to balance out asymmetric shocks, has clearly been violated (see Mundell, 1961).7
As far as our empirical results are concerned, we find evidence for a level break in the cointegrating relationship, which we ascribe to the EMU sovereign debt crisis. This indicates that, after the crisis, the expected higher risk awareness of investors keeps government bond yield spreads at a higher level than in the pre-crisis period.
Moreover, results point at fiscal imbalances – in particular expected government debt-to-GDP differentials – as the main drivers of sovereign spreads, although liquidity risks have a non-negligible weight. Cumulated inflation differentials turn out to be a significant variable, its importance being of the same order of magnitude as liquidity risk. But, perhaps most importantly, their inclusion among the regressors allows us to establish that the conclusions we draw on sovereign bond yield spreads determinants are closely interlinked to whether or not diverging competitiveness significantly affects sovereign bond yield spreads themselves.
In particular, we argue that a statistical significance attached to cumulated inflation differentials is an indication that the economies included in the sample of countries do not belong to an OCA. In fact, if shocks were sufficiently correlated or if the monetary union were able to absorb and balance out asymmetric shocks, then cumulated inflation differentials would be small and unimportant for sovereign bond yield spread determination.8 We iteratively run this test by excluding one country at a time from the full sample of countries, starting from that with the highest cumulated inflation differential relative to Germany, and going forward until such a variable becomes statistically insignificant. This process leads (i) to a grouping of countries into two categories corresponding to EMU core (Austria, Finland, France, Germany and the Netherlands) and EMU periphery (Belgium, Greece, Italy, Portugal and Spain), and (ii) to the finding that cointegrated panel regression results are clearly driven by the inclusion of the observations belonging to the peripheral EMU economies considered. In fact, when such observations are excluded, debt-to-GDP differentials turn out to be the least important determinant of the sovereign bond yield spread, while expected budget balance differentials and the liquidity risk carry the highest weights.
It is noteworthy that while in the sample of peripheral countries a one-percent-point rise in the expected public-debt-to-GDP ratio differential leads, on average, to an 8.63 basis points increase in the sovereign bond yield spread; in the restricted sample pooling only core EMU economies, the same increase in the expected public-debt-to-GDP ratio differential leads, on average, to an increase in the sovereign spread of only 0.46 basis points. These results clearly unveil the fact that international investors heavily punish the deterioration of expected debt positions of those countries that face competitiveness gaps and hence are not being perceived as OCA members.
The remainder of the paper is structured as follows. Section 2 describes the data employed in the estimation. Section 3 outlines the econometric methodology. Section 4 reports and discusses the results. Finally, Section 5 concludes and highlights policy implications. Technical details are appended to the paper.
Section snippets
Data
Our panel dataset contains monthly data of nine euro-area countries over the period 2001:1–2011:12. The countries, selected on the basis of data availability, are Austria, Belgium, Finland, France, Greece, Italy, the Netherlands, Portugal and Spain. In the remainder of the paper subscript i refers to the cross-sectional dimension (country) and subscript t refers to the time dimension (month).
The variable to be explained is the ten-year government bond yield spread over Germany (rit). Data are
Econometric methodology
The discussion on the variables of interest in Section 2 leads to the following specification for an empirical model on long-run determinants of sovereign bond yield spreads:where are coefficients to be estimated, the notation on the regressand and the regressors is that described in Section 2 and εit is an error term.
To estimate Eq. (1) we have to take three important econometric issues into account. First, we cannot ignore
Results
Three tests, the results of which are reported in Table 1, point at a clear-cut evidence of cross-sectional dependence for all variables. This reinforces our prior conjecture based on economic considerations. In particular we run the LM test by Breusch and Pagan (1980), the CD test by Pesaran (2004), and the standardized Spacing Variance Ratio (SPR) by Ng (2006).12
Conclusions and policy implications
This paper provides useful information for policy makers facing the difficult task of tackling high sovereign bond spreads with the aim of fostering greater public finance stability and ultimately guaranteeing EMU survival. Results primarily point at expected fiscal imbalances (namely expected government debt-to-GDP differentials) and liquidity risks as the main determinants of sovereign bond yield spreads in the long run. We find evidence for a level break in the relationship, occurring during
Acknowledgments
Comments by Yunus Aksoy, Panagiotis Konstantinou, Ron Smith, Paola Paiardini, Joe Pearlman, Lucio Sarno, Stefania Villa and two anonymous referees are gratefully acknowledged. The usual disclaimer applies.
References (46)
- et al.
Do asymmetries matter for European monetary policy?
Eur. Econ. Rev.
(2002) - et al.
The EMU sovereign-debt crisisfundamentals, expectations and contagion
J. Int. Financ. Mark., Inst. Money
(2012) - et al.
Sovereign risk premiums in the European government bond market
J. Int. Money Financ.
(2012) Unit root tests for panel data
J. Int. Money Financ.
(2001)- et al.
Capital mobility and global factor shocks
Econ. Lett.
(2013) - et al.
The Greek financial crisisgrowing imbalances and sovereign spreads
J. Int. Money Financ.
(2012) Size matters for liquidityevidence from EMU sovereign yield spreads
Econ. Lett.
(2006)Some tests for unit roots in autoregressive-integrated-moving-average models with deterministic trends
Biometrica
(1993)- et al.
An LM test for a unit root in the presence of a structural break
Econom. Theory
(1995) - et al.
What explains the surge in euro area sovereign spreads during the Financial Crisis 2007–2009?
Public Financ. Manag.
(2010)
Structural changes, common stochastic trends, and unit roots in panel data
Rev. Econ. Stud.
On the estimation and inference of a panel cointegration model with cross-sectional dependence
A panic attack on unit roots and cointegration
Econometrica
Estimating and testing linear models with multiple structural changes
Econometrica
Determinants of intra-euro area government bond spreads during the financial crisis
Eur. Eco. Econ. Pap.
The LM test and its application to model specification in econometrics
Rev. Econ. Stud.
EMU and trade revisitedlong-run evidence using gravity equations
World Econ.
Panel data stochastic convergence analysis of the Mexican regions
Empir. Econ.
Yield spreads on EMU government bonds
Econ. Policy
Cited by (0)
- 1
The author completed part of this work while he was visiting Cass Business School, City University London.