Is the EMU government bond market a playground for asymmetries?

The Journal of Economic Asymmetries 10 (2013) 21–31

Contents lists available at ScienceDirect

The Journal of Economic Asymmetries www.elsevier.com/locate/jeca

Is the EMU government bond market a playground for asymmetries? ✩ Catalin Dragomirescu-Gaina a,∗ , Dionisis Philippas b a b

Unit of Econometrics and Applied Statistics, European Commission, Via E. Fermi, 2749 Ispra, Italy Department of Finance, ESSCA Grand Ecole de Management, 55 Quai Alphonse Le Gallo, 92513 Boulogne-Billancourt, Paris, France

a r t i c l e

i n f o

Article history: Received 2 July 2013 Received in revised form 5 September 2013 Accepted 6 September 2013 Available online 25 September 2013 Keywords: Government bond yields Economic Monetary Union Asymmetries

a b s t r a c t We investigate the volatility dynamics of some major European Monetary Union sovereign bond markets. We provide an endogenous identification in terms of two Markov switching regimes for market volatility and analyze the impact of capital and trade flows together with policy actions on the persistence of volatility swings. The empirical findings indicate that, with some notable exceptions, capital and trade flows measures were a matter of minor importance for European Monetary Union sovereign bond markets included in our set. On the contrary, central banks’ liquidity provision indicators had important but asymmetrical effects on the persistence of the European Monetary Union’s bond market volatility swings. Although we do not straightly reject the increased market integration hypothesis, these asymmetries suggest that certain domestic factors still weigh heavily in times of stress for market sentiment. © 2013 Elsevier Inc. All rights reserved.

1. Introduction Government bond markets play a major role in the transmission mechanism of monetary policy in the European Economic and Monetary Union (hereafter EMU). The common currency became the vehicle for exercising effective and appropriate policies on many levels and avoiding the consequences of asymmetrical shocks. Over the first years of Eurozone life, the previous literature provided evidence that the common currency has led to government bond yields’ integration due to the elimination of asymmetric shocks and better liquidity conditions, despite deteriorating macroeconomic fundamentals (see for example Abad, Chuliá, & Gómez-Puig, 2010; Christiansen, 2007; Codogno, Favero, & Missale, 2003; Favero, Pagano, & Von Thadden, 2010; Geyer, Kossmeier, & Pichler, 2004; Kim, Moshirian, & Wu, 2006; Pagano & von Thadden, 2004; Schuknecht, Von Hagen, & Wolswijk, 2009). Overall, these studies conclude that the harmonization of EMU macroeconomic policies, together with the liquidity risk and the global risk, plays an important role in affecting bond spread fluctuations. The presence of global risk factors should capture portfolio allocation from an international diversification perspective; domestic factors would become less significant over time and government bond spreads would reflect increased market integration, and much less so macroeconomic or liquidity conditions (see for instance, Longstaff, Jun, Lasse, & Singleton, 2011), as mentioned above. Although we do not straightly reject the increased market integration hypothesis, we provide evidence that would instead reveal an asymmetric market response to the same concerns and same policy actions, especially in times of high uncertainty. We therefore assume that certain domestic factors are responsible for these asymmetries. In fact, because Euroarea members gave up monetary policy and national currency, they had to survive negative macroeconomic and financial

✩ We gratefully acknowledge Professor Costas Siriopoulos (Department of Business Administration, University of Patras, Greece) for his comments on earlier versions of this paper. Corresponding author. E-mail addresses: [email protected] (C. Dragomirescu-Gaina), [email protected] (D. Philippas).

*

1703-4949/$ – see front matter © 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jeca.2013.09.001

22

C. Dragomirescu-Gaina, D. Philippas / The Journal of Economic Asymmetries 10 (2013) 21–31

shocks by relying on other policies. Moreover, retaining national responsibility for fiscal policy and financial regulation did not make things easier. We take an empirical approach and look at volatility swings in some major European sovereign bond markets. We use a Markov switching model suitable for capturing non-linearities in the data and providing an endogenous characterization of two separate regimes observed for market volatility. The identification of the regimes overlaps with the key events that took the center stage in European financial markets, starting with the first Greek bailout in May 2010. We then investigate how market concerns and fears regarding imbalances caused by capital and trade flows and how different policy actions have affected the persistence of volatility swings. Our findings indicate that measures of net foreign assets’ positions were a matter of concern only in the case of Greece, given that the country was relying on external capital inflows to finance its deep-rooted current account deficit. Trade balance developments had a favourable impact on the market volatility only in Germany. However, central banks’ liquidity provision measures affected the persistence of market volatility swings favourably in Spain and Ireland but negatively in Greece and Portugal. As government cash and debt management heavily depends on country-specific regulations and traditional arrangements with the central monetary authorities, we regard this as one of those domestic factors affecting market volatility dynamics. Much as in Lane (2012) and other related studies concerning the European sovereign crisis, we support the idea that higher coordination is needed to mitigate the impact of country-specific factors, regulations or institutional arrangements. We consider that our analysis has strong implications for international investors’ optimal portfolios decisions. Reallocation between individual EMU countries’ assets should be evaluated according to the asymmetries outlined in this study and the probability of lying within one market regime or another. The structure of the paper is as follows. Section 2 discusses the proposed methodology, Section 3 presents the data, while Section 4 shows the empirical analysis and discusses the main findings. Finally, Section 5 summarizes the conclusions. 2. Methodology The Markov switching model was first considered by Hamilton (1990) and Kim (1994). The data generating process (hereafter DGP) of time series may display fat tails or point to multimodal distributions. Structural and persistent shocks and different causality and dependency relations with other indicators could potentially explain these non-linearities. The different dynamics are the “states” or “regimes” and they specify the model equation as:

yt = β( S t )xt + σ ( S t )εt

(2.1)

where εt ∼ N (0, 1) is a normal distributed sequence of innovations (i.i.d.) and the parameters β and S t , an unobservable variable that follows a first-order Markov process:

Pr ob( S t = j / S t −1 = i ) = P i j ,

i , j ∈ [1, k]

σ depend on the state (2.2)

where k is the number of states. The P i j probabilities, which should be interpreted as the probability of switching from state i to state j, are collected in a symmetric matrix P (k×k) = [ P i j ] where each row sums to unity. The properties of the dependent variable are jointly determined by the characteristics of the innovations εt and the state variable S t . In particular, the Markovian state variable yields random and frequent changes in model dynamics, matching the observed non-linearities in the data, while its transition probabilities determine the persistence of each regime. The time-varying transition probabilities are specified as a function of an exogenous variable zt , using the cumulative normal distribution function as illustrated below. For a two state Markov switching model (i.e. k = 2), the expressions for the transition probabilities follow:



P (2×2),t =

P 11,t 1 − P 22,t

P j j ,t = N ( zt θ j j ),

1 − P 11,t P 22,t

j = 1, 2



(2.3) (2.4)

where N is the cumulative normal density function, zt is the exogenous variable and θ j j are parameters that need to be estimated. Given that the derivative of P j j with respect to the exogenous variable z in (2.4) is a function of θ j j , one can assess the direction and the statistical significance of the influence of zt on yt . The time-varying specification could be tested using a simple LR test against the restricted one involving constant transition probabilities as in Hamilton’s original model. The value of the unobservable variable S t in this case must be uncorrelated with the value of zt but this can be achieved by using the lagged values of variable z. The estimated transition probabilities give a much more realistic characterization of the “states” in this case, being conditioned on the current set of information, and not on the entire information set. We use the empirical approach described above to investigate the volatility patterns observed in some major EMU sovereign bond markets, by following a two-step approach. In the first step, we depart from theory and derive an empirical testable specification for the main model equation, using constant transition probabilities as shown in (2.1) and (2.2). In a

C. Dragomirescu-Gaina, D. Philippas / The Journal of Economic Asymmetries 10 (2013) 21–31

23

second step, we let the transition probabilities vary over time in order to explain the persistence of volatility swings, as illustrated in (2.3) and (2.4). As determinants, we use various indicators reflecting institutional sector imbalances generated by capital and trade flows and proxies for policy actions that have the potential to affect market behavior. A standard theoretical framework for modelling interest rates and the term structure is represented by the general equilibrium models of Cox, Ingersol, and Ross (1985) and Longstaff and Schwartz (1992). Cox et al. (1985) explain the term structure as a function of short-term interest rates and then extend the model with an exogenous process for inflation uncertainty in order to provide a role for money and nominal prices. Longstaff and Schwartz (1992), use the short-term interest rate volatility as a second factor when explaining the term structure. With a two-factor model, they illustrate the interplay between the short-term interest rate and its volatility in matching the twists and humps of the term structure in the real data; they derive a specification where the volatility of the term structure depends on the same state variables as the term structure itself. Nowadays, most of the changes in short-term interest rates are under the control of monetary authorities, although some volatility remains given imprecise implementation of monetary policy changes, liquidity regulations or other exogenous factors affecting money markets and interest rates (Manganelli & Wolswijk, 2009). The European Central Bank (hereafter ECB) is keen on its price stability mandate and acts with regard to inflation developments across all euro-area members. In this context, country-specific differences in price trends could arise reflecting either structural or transitory developments. However, under a common monetary policy, high structural inflation in one member state could snowball into a loss of external competitiveness over time, deterioration in public finances and high debt levels. These developments would certainly create additional uncertainty, reflected in the long-term interest rates as investors discount the future value of poor economic management. The empirical literature explaining long-term interest rates and term structures is rich and relies on various determinants. Some studies look at trends in aggregate savings and investment, which can be linked to financial liberalization, capital returns or even demographic developments (see for example, Dean, Durand, Fallon, & Hoeller, 1990; Howe & Pigott, 1991). Other authors focus instead on public deficits and government fiscal policy, highlighting the role of long-term inflation expectations (see for example, Nicoletti, 1988). Given theoretical considerations outlined above and drawing on previous empirical literature, we specify and estimate an equation1 that captures the volatility swings in the sovereign bond market, as follows:

vol(r ) = β0 ( S ) + β1 ( S ) vol(i ) + β2 ( S )infl + β3 ( S ) W + σ ( S )ε

(2.5)

where r represents the government bond yield, i represents a short-term interest rate measure, infl represents the inflation volatility proxy, W captures other factors such as international influences on market sentiment and vol() denotes the realized volatility.2 We estimate the model parameters β ’s and σ together with the coefficients characterizing the Markovian variable S via maximum likelihood and then select the best specification according to standard selection criterions (such as BIC). 3. The data We use a dataset consisting of the monthly realized volatility for the ten-year government bond yields for eight EMU countries, namely Italy, Spain, Greece, Portugal, France, Germany, the Netherlands and Ireland, covering the period from January 1999 to December 2012.3 The sample therefore contains the period since the introduction of the common currency and ends well after the time point when countries like Greece, Portugal and Ireland had to withdraw from the international bond markets. As discussed by Chordia, Sarkar, and Subrahmanyam (2005), and Pastor and Veronesi (2009), uncertainty is the reality under which investors operate, learn and update their evaluations. We aim at capturing the main symptoms of sovereign bond market stress by directly observing the realized volatility as a better measure of the market/investors’ sentiment (Boyer et al., 2006). Modelling volatility instead of mean returns provides a more efficient use of information, less parameters to estimate and a more straight interpretation of the results. 3.1. Constant transition probabilities specification At the empirical specification of Eq. (2.5), we use the three-month Euribor as a proxy for short-term interest rates. This proxy reflects counterparty risk in the inter-bank market, it captures stress features like flight-to-quality, flight-to-liquidity, as well as the price impacts of enhanced adverse selection problems in times of banking stress and it could also provide information on the liquidity conditions in the long-term segment (see Acharya & Skeie, 2011; Beber, Brandt, & Kawajecz, 2009). Table A1 (in Appendix A) reports the summary statistics for realized volatility calculated for government bond yields and Euribor. The statistical analysis shows that the mean and the variation of the bond yields’ volatility both fall within a

1

This is the empirical counterpart of Eq. (2.1), ignoring time subscripts. Realized volatility is calculated as the volatility of the daily average bond yields’ returns or short-term interest rate. Please note that the omitted time subscript would apply to the time series of calculated realized volatility. 3 Source: all the data are extracted from Datastream, ECB, Eurostat or are authors’ calculations. 2

24

C. Dragomirescu-Gaina, D. Philippas / The Journal of Economic Asymmetries 10 (2013) 21–31

tight range in all the countries in our sample. However, the distributions of all the bond yields’ volatilities are skewed to the right, excess kurtosis indicates fat tails and the hypothesis of normality (J–B test) is rejected for all the cases. In fact, the non-normality property of the volatility distributions was one of the main arguments for using a Markov switching model specification in the first place. This type of model is suitable for describing data generating processes characterized by mixed or multi-modal distributions, fat tails etc. The second determinant in Eq. (2.5) relates to inflation, which has been viewed in the literature as a proxy for the quality of economic management, which could have consequences for sovereign default risk (Mihaljek & Klau, 2008). High inflation hampers economic development, investment decisions and maintains uncertainty regarding future policy actions designed to bring price stability. Whenever the rise in inflation is perceived as having a transitory rather than a structural source, high inflation and low interest rate volatility may coexist. We construct a ratio between the core4 inflation in each country and the eurozone total as we wish to concentrate on country-specific structural inflationary pressures, rather than global or regional trends. Next, we take the idiosyncratic shock of the conditional volatility of an estimated GARCH model, following Christiansen (2007). Table A2 (in Appendix A) provides an initial perspective on the idiosyncratic shock of relative core inflation time-series and reports their summary statistics. As regards other determinants of government yields volatility described in (2.5) by the W variable, we have experimented with different proxies mainly reflecting global influences such as CBOE5 Volatility Index VIX, EURUSD currency pair volatility or European stock index STOXX50 volatility. The effect of global and regional factors on financing costs in EMU countries was analyzed within the context of the literature on the determinants of sovereign debt spreads. Global risk aversion was found to be a significant factor driving the spreads (see Baldacci & Kumar, 2010; Dailami, Masson, & Padou, 2008; Gonzales-Rozada & Levy-Yeyati, 2008) and in periods of financial distress had a larger impact on bond yields. However, in the empirical analysis section we report model estimates including only the first two determinants described above, as they had better statistical properties and lower information criterion (BIC) numbers across various alternative specifications. This does not mean we are downplaying the importance of global factors in explaining volatility dynamics in EMU sovereign bond markets; but, from an empirical perspective, our simple model looks well adapted at capturing the needed dynamics. One possible explanation could be that global influences have already been captured by the inclusion of short-term interest rate volatility, given that the much of the transactions in financial markets are clustered around a short-term maturity. 3.2. Time-varying transition probabilities specification The discussion up to now has followed the specification presented in (2.5), which is the main model equation. However, a Markov switching model could be estimated under constant or time-varying transition probabilities. We next turn and have a look at the potential determinants of transition probabilities, namely variable z in specification (2.4). We intend to describe the dynamics and persistence of volatility swings between the two identified regime states, by relying on indicators expressing both market concerns or fears and policy actions. A priori, one could assume that fears would drive market volatility upwards and policy actions would drive it downwards in times of stress. Much of the discussion about the indicators below has been constrained by the availability of time series needed to capture all relevant information for market sentiment at the monthly frequency. In light of our research objective, by using monthly data for our model specification we had to strike a balance6 between the high frequency of market price discovery process and the low frequency of statistical data availability. In the literature on bond yields’ determinants, openness and trade flows play an important role in explaining the cost of borrowing as the penalty for sovereign default is higher in terms of capital reversion (see, for instance, Berger & Nitsch, 2010). We construct a rough measure of imbalances caused by trade flows (or current account position); more exactly, we take net exports and scale them by exports value, which gives us a trade balance proxy more adequate for addressing financial crisis effects and increased vulnerability (see Lane, 2012; Chen et al., 2013). In a related study, Attinasi, Checherita, and Nickel (2009) take macroeconomic announcement surprises as a determinant for sovereign bond spreads, but use another dataset and reach different conclusions. Table A3 (in Appendix A) provides the summary statistics of the trade balance proxy used here. With few exceptions (Netherlands and France) data show non-stationarity and therefore we prefer to use the first difference of the indicator for all the countries in order to have comparability. Lane and Milesi-Ferretti (1999) have laid the basis for a rigorous analysis of the relation between economic development and capital flows or external indebtedness (measured by the stock of foreign assets). Beber et al. (2009) analyze capital flows and EMU sovereign bond yields finding a strong relation between liquidity and quality. Given the major contribution of banks’ financing in Europe, we highlight the potential role played by the banking sector in influencing the sovereign bond market. 4 We prefer to use core inflation instead of general consumer price inflation, which may include items such as unprocessed food prices or other very volatile components. Our core measure for inflation is the “overall index excluding energy and unprocessed food” from Eurostat database. 5 Chicago Board Options Exchange. 6 For example, as one move towards the high frequency spectrum, there could be daily available data for sovereign bond yields, but there will be no daily proxy to capture imbalances generated by trade or capital flows. At the low frequency spectrum, better macroeconomic indicators become available, but the financial market data would be too sparse to reflect investors’ sentiment.

C. Dragomirescu-Gaina, D. Philippas / The Journal of Economic Asymmetries 10 (2013) 21–31

25

We use the net foreign assets divided by the total foreign assets as a proxy for the external position of the banking sector. A worsening external position leaves a country banking sector more exposed to a change in the global risk appetite. Depending on the inter-linkages between domestic banks and the government sector, this vulnerability could further extend to the public sector and bond markets. Table A4 (in Appendix A) provides a perspective of the summary statistics for the net external position proxy for the eight EMU countries. The statistical analysis reveals that the hypothesis of the unit root (ADF test) is accepted for all the cases, motivating us to use the first difference of the indicator. Government cash flows have been a major cause of short-term volatility in money markets and banks’ liquidity. This has complicated both liquidity management and the central bank’s monetary policy implementation. The need to mop up excess liquidity in the banking system can be costly to the central bank. The implementation of government cash and debt management through a central bank can conflict with its monetary policy goals and operations. This has been the case particularly in relation to interest rate setting and signalling. Central bank lending to the government increases the monetary base, which has potential consequences for inflation. Good indicators for policy actions are especially hard to find and measure based on publicly available information. We prefer using statistical indicators which are objective and measurable in order to capture the policy actions commanded by monetary authorities (see Giannone, Lenza, Pill, & Reichlin, 2012) and governments/central authorities. Some authors have used other proxies such as, for example, based on official statements communicated through media channels (see for instance, Blinder, Ehrmann, Fratzscher, De Haan, & Jansen, 2008 who present a survey on how central banks communicate and the associated empirical findings on market impact and, moreover, see Attinasi et al., 2009 who present estimates of macroeconomic press release surprises in the case of bond spreads). Table A5 (in Appendix A) provides the summary statistics for the share of central bank funding in credit institutions liabilities. Table A6 provides statistics about the ratio between the government deposits and the loans taken from the banks, which we take as a measure of government liquidity (or cash and debt) management. Both indicators are available in the ECB statistical data warehouse. The statistical analysis motivates us to use both indicators in first differences to mitigate any non-stationarities issues and assure comparability. Financial markets are usually exposed to large amounts of information, much of it being simply noise. The current state of technology has increased the speed and the volume of the information transmitted through standard media channels and Internet networks. Some of it would certainly need more time in order to be digested and properly integrated by market participants into their evaluations. Because we want to let the market react to the information and not the noise, we use two types of transformation of the four indicators described above (included in the transition probabilities specification): a quantitative and a qualitative one. Both transformations would amplify the impact of the indicator if over consecutive months the change takes place in the same direction. The quantitative transformation adds the previous month’s change to the current change in the indicator, while the qualitative transformation just counts the number of consecutive changes that take place in the same direction. This means that an improvement in the trade balance proxy on a monthly basis might be disregarded when it first happens, but it would more certainly be noticed if the improvement repeats itself for a second consecutive month or even a third consecutive month. These transformations are more common in the literature addressing central bank foreign exchange interventions and market impact, ours being close with the ones in Égert (2007). The power of these transformations rests in the insights they could provide about the effectiveness of policy intervention; moral suasion or public commitment to a future policy course could be more powerful tools than the effective deployment or the size of policy action. We extend these transformations to the proxy indicators capturing economic fundamentals such as imbalances caused by trade and capital flows. In the next section, we assign the z variable in specification (2.4) with the quantitative and qualitative transformations of our proxy indicators for policy actions and imbalances generated by capital and trade flows. 4. Empirical analysis In this section we provide details of the estimated specifications for the eight countries in our set. We experiment with several model specifications for each country according to Eq. (2.5). In the first step the β coefficients and σ are allowed to switch between two states but the transition probabilities are kept constant over time. We base our selection on BIC information criterion, choosing the most conservative specifications with the lowest BIC figures. Richer alternative specifications also included various proxies for global risk,7 but they were discarded in the final specifications. Under constant transition probabilities, we find that the best specifications entails that only the intercept and the variance in Eq. (2.5) switch between states, namely β0 and the standard deviation σ . The coefficients of the interest rate volatility and the core inflation idiosyncratic shock have remained fixed across both regimes. Table 1 below provides a summary of the estimates for each country. The set of estimated coefficients in the table above, especially the intercept and the standard deviation, clearly differentiates the two estimated regimes. Therefore, we label regime 1 as the calm regime and regime 2 as the volatile one. The exact identification of each regime in each country can be observed in Fig. 1.

7 Global Influences included CBOE Volatility Index VIX, EURUSD currency pair volatility and European stock index STOXX50 volatility. See the discussion from the previous section.

26

C. Dragomirescu-Gaina, D. Philippas / The Journal of Economic Asymmetries 10 (2013) 21–31

Table 1 Markov switching model with constant transition probabilities.

β1 β2 Regime 1

β0 Std. dev. Regime 2

β0 Std. dev. BIC P 11 P 22

GR

NL

−0.01 −0.29

−0.08 0.14

FR

DE 0.03 0.42∗

−0.17∗∗ −0.35

IE 0.08 0.23

IT

PT

−0.03 0.14

−0.11∗ 0.23

ES 0.09 0.09

0.65∗∗∗ 0.09∗∗∗

0.77∗∗∗ 0.11∗∗∗

0.66∗∗∗ 0.08∗∗∗

0.84∗∗∗ 0.13∗∗∗

0.64∗∗∗ 0.11∗∗∗

0.71∗∗∗ 0.07∗∗∗

0.71∗∗∗ 0.08∗∗∗

0.75∗∗∗ 0.16∗∗∗

1.92∗∗∗ 1.40∗∗∗

2.13∗∗∗ 0.56∗∗∗

1.55∗∗∗ 0.45∗∗∗

2.63∗∗∗ 0.95∗∗∗

1.73∗∗∗ 0.85∗∗∗

1.77∗∗∗ 0.79∗∗∗

2.08∗∗∗ 1.16∗∗∗

2.07∗∗∗ 1.64∗∗∗

283.028 0.97∗∗∗ 0.95∗∗∗

226.342 0.97∗∗∗ 0.88∗∗∗

205.686 0.96∗∗∗ 0.88∗∗∗

269.118 0.97∗∗∗ 0.89∗∗∗

252.059 0.97∗∗∗ 0.89∗∗∗

214.966 0.96∗∗∗ 0.88∗∗∗

258.407 0.97∗∗∗ 0.95∗∗∗

268.418 0.99∗∗∗ 0.98∗∗∗

The table presents the empirical model estimated in line with the specification described in (2.5). According to the BIC criterion, the specification that best describes the data allows only the intercept and the standard deviation to switch between two states. One, two and three stars (∗ , ∗∗ and ∗∗∗ ) denote statistically significant coefficients at 10%, 5% and 1%, respectively.

Fig. 1. Volatile regime identification under constant transition probabilities.

Just for illustration, we chose to group the countries into two categories: the first group contains Greece, Portugal, Spain and Ireland, whereas the second group contains Italy, Germany, France and the Netherlands. Our model allows us to identify8 exactly the start of the volatile regime in Greece as of January 2010, in Portugal as of February 2010 and in Ireland as of March/April 2010. For Italy, France, the Netherlands and Germany there were only a few volatile months between May 2010 and January 2011. The first months of 2011 brought a change in market volatility in Spain’s sovereign bond markets, and then, beginning with July 2011, an upwards spike occurred for all the countries in our sample. Lately, the volatility came down by the end of 2011 only for Ireland. The persistency of the identified regimes is quite high, with higher persistency for the calm or stable regime rather than the volatile one. We should highlight that the persistency of the volatile regime is higher in Greece, Portugal and Spain. It is also interesting to note that the calm regime has more or less the same coefficients across all the countries, highlighting a possible convergence or market integration scenario. However, all this changed in the high-volatility regime with volatility spikes that were more erratic and sudden in Greece, Portugal and Spain, as illustrated by the standard deviation estimates. Although the β1 and β2 coefficients in Table 1 are not statistically significant in overall, this model specification was the most parsimonious one, capturing determinants as laid down in the second section where we discuss the methodology. In fact, our intention is to control for the impact of major theoretical determinants and then explore the remaining residual volatility by using our proxies capturing market concerns and policy actions in the second step. Next, we allow the transition probabilities to be time varying, determined according to (2.3) and (2.4). One by one, we introduce proxies reflecting developments in: (a) the trade balance, (b) the net foreign assets’ position (NFA) or external borrowing of the banking sector, (c) the central bank funding or liquidity providing measures to domestic banks and (d) government cash and debt management. The first two proxy indicators (a) and (b) reflect macroeconomic and banking sector imbalances or disequilibria in two most relevant institutional sectors as a consequence of trade and capital flows. The last two proxy indicators (c) and (d)

8

The smoothed probability of being in the volatile regime is above 90% starting with these dates.

C. Dragomirescu-Gaina, D. Philippas / The Journal of Economic Asymmetries 10 (2013) 21–31

27

mainly reflect policy actions or available instruments through which policy makers can influence market volatility dynamics and persistence. For each of our four proxy indicators we experiment using the quantitative and qualitative transformations suggested at the end of the previous section in order to disentangle information from noise. We start by introducing the first lag of each proxy indicator into (2.4), therefore the qualitative and quantitative transformations above refer to adding a second or a third lag to the newly constructed indicator. We present in Table 2 only the specifications for which we obtain statistically significant estimates for the time-varying transition probabilities of at least one regime. Depending on the sign of the estimated θ coefficients, we can evaluate whether a certain change in our proxies would keep the sovereign bond market volatility in one regime or push it out of that state. Specifically, we concentrate on the probabilities of remaining in the same regime, i.e., P 11 and P 22 , as shown in (2.3). Table 2 also reports the LR tests comparing the restricted specification with constant transition probabilities, as shown in (2.2), against the unrestricted specification, which includes the exogenous proxies in the transition probabilities, as in (2.3) and (2.4). Observing the estimated coefficients for β0 and the standard deviation for regime 1 across countries, we see a very tight range, again suggesting a possible convergence or market integration scenario. However, in the volatile regime number 2, things change, as indicated by our estimated intercept and standard deviation, which now vary widely across countries. We perform LR tests and accept (at mostly 10% level of significance) that the specifications illustrated in Table 2 are generally superior to the constant transition probabilities’ specifications presented in Table 1. According to our findings, imbalances generated by trade flows were a decisive factor only for the persistence of market volatility in Germany, keeping the market in the calm state. Imbalances generated by capital flows were a matter of concern but only for Greece, pushing the market out of the calm state. An increase in the net foreign assets’ position (equivalent to an increase in the foreign assets and/or a decrease in the foreign liabilities) has negative consequences because it means a reduction in foreign capital inflows, potentially accelerating the economic downturn. It is interesting to note that markets need to perceive quite a strong signal coming from fundamentals (i.e., two or three months of consecutive changes in the same direction) in order to digest the information and react to it. Increased reliance on central bank financing had positive consequences for Ireland and Spain, keeping the market volatility in a low range. In the case of Greece and Portugal, the same thing has caused uncertainty pushing the market out of the calm regime. This asymmetry means that international investors perceive and evaluate countries on different grounds. Spain and Ireland were treated in more favorable terms by the market than Greece and Portugal as a reaction to central bank liquidity measures. A possible explanation of this asymmetry considering the timing of the official bailout request or announcement9 for each of these countries would not entirely solve the puzzle, because Greece and Ireland request for official financial assistance came in 2010, followed by Portugal in 2011 and Spain in 2012. Looking at the regimes identification according to our Markov switching model, we see no major time gap between Greece, Portugal and Ireland coming under market pressure. This leaves us only with an explanation that rests on other domestic factors or country-specific characteristics. In the case of Germany, the coefficient in the calm regime is also statistically significant and negative, but the LR test did not reject the constant transition probability model as a better representation of the data. As an additional observation, the impact of central bank liquidity on market volatility is more rapid, even within the following month, compared to the impact of imbalances generated by capital and trade flows. An improved government liquidity position keeps the market volatility subdued in Greece and Germany but has an unfavorable impact in France, Ireland, Spain and Portugal. There is no particular distinction observable between these two groups of countries when looking at the descriptive statistics in Table A6 that could help us identify a possible explanation here. Back to the discussion above regarding country-specific factors, we could consider this as a factor in that list given that institutional set-up and regulations associated with government liquidity management and its relation with monetary authority are very much country-specific. In fact, government operations represent one very important autonomous source of variation for money market liquidity in general. 5. Conclusions In this paper, we adopt a Markov switching model with both a constant and time-varying specification of the transition probabilities in order to better characterize volatility dynamics in some major EMU sovereign bond markets. Our endogenous identification of the market volatility regimes corresponds to major events that have shaped the European financial crisis in the last years. We investigate several determinants of market volatility dynamics to which we apply two types of transformation to better segregate information from data. We focus on determinants that capture institutional sector imbalances generated by capital and trade flows and proxies for policy actions that have the potential to mitigate crisis effects. The findings support the proposition that net foreign assets’ positions and trade balance developments are a matter of minor importance for volatility dynamics, liquidity provision measures of Eurozone central banks have important but asymmetrical effects, which could be possibly explained by relying on country-specific factors such as, for example, gov-

9 The exact dates are as follow: Greece had the first official bailout program approved on 2 May 2010; Ireland official request came on 21 November 2010; Portugal request came on 7 April 2011; Spain request came on 25 June 2012. Source: European Commission, DG Economic and Financial Affairs.

28

GR

GR

GR†

GR†

FR

FR

FR

FR

DE

DE

DE

DE

DE

IE

IE†

ES

ES

ES

ES

PT

PT†

PT†

NFA

Gov’t

Central bank

Central bank

Gov’t

Gov’t

Gov’t

Gov’t

Trade balance

Gov’t

Gov’t

Central bank

Central bank

Gov’t

Central bank

Gov’t

Central bank

Central bank

Central bank

Gov’t

Central bank

Central bank

Type of C transformation

Q

C

C

C

C

C

Q

C

C

C

C

C

C

C

C

C

C

C

C

C

C

#no consecutive #2 months

#3

#1

#3

#1

#2

#3

#3

#3

#1

#2

#1

#2

#3

#2

#1

#1

#2

#3

#3

#2

#3

−0.10 0.21

−0.09 0.22

Proxy

β1 β2 Regime 1:

β0

Std. dev.

θ11 Regime 2:

β0

Std. dev.

θ22 BIC LR test (p-val)‡

−0.00 −0.23 0.63∗∗∗ 0.08∗∗∗ −16.57∗

−0.01 −0.36

0.06 −0.22

0.05 −0.21

0.03 0.42∗

0.03 0.42∗

0.03 0.43∗

0.03 −0.15 0.50∗∗ −0.43

−0.16∗∗ −0.16∗∗ −0.17∗∗ −0.16∗∗∗ −0.39 −0.39 −0.39 −0.36

0.09 0.23

0.65∗∗∗ 0.65∗∗∗ 0.65∗∗∗ 0.66∗∗∗ 0.66∗∗∗ 0.66∗∗∗ 0.67∗∗∗ 0.84∗∗∗ 0.83∗∗∗ 0.84∗∗∗ 0.85∗∗∗ 0.85∗∗∗ 0.62∗∗∗ 0.09∗∗∗ 0.09∗∗∗ 0.08∗∗∗ 0.08∗∗∗ 0.08∗∗∗ 0.08∗∗∗ 0.08∗∗∗ 0.14∗∗∗ 0.13∗∗∗ 0.13∗∗∗ 0.13∗∗∗ 0.13∗∗∗ 0.11∗∗∗ −1.71∗ −1.36∗ −9.39∗ 0.33 −1.10∗∗ −0.69∗∗ −12.5∗ −11.9∗∗ −11.4∗∗ −1.92∗∗∗ 58.4∗ 52.9∗∗ 35.9∗

1.84∗∗∗ 1.89∗∗∗ 1.93∗∗∗ 1.35∗∗∗ 1.40∗∗∗ 1.40∗∗∗ 1.64 −1.69∗ −0.08

1.92∗∗∗ 1.39∗∗∗ 0.05

1.55∗∗∗ 0.45∗∗∗ 5.03

1.55∗∗∗ 0.45∗∗∗ 4.92

1.56∗∗∗ 0.45∗∗∗ 8.1

1.57∗∗∗ 2.46∗∗∗ 2.61∗∗∗ 2.63∗∗∗ 2.70∗∗∗ 2.69∗∗∗ 0.45∗∗∗ 1.10∗∗∗ 0.94∗∗∗ 0.94∗∗∗ 0.90∗∗∗ 0.91∗∗∗ 0.49 −24.8 −132.9∗ −105.2 −1.51 −0.58

0.44∗∗∗ −0.15 0.18 0.10 0.61∗∗∗ 0.79∗∗∗ 0.18∗∗∗ 0.11∗∗∗ −5.35∗ 0.96∗

1.70∗∗∗ 1.95∗∗∗ 1.98∗∗∗ 0.84∗∗∗ 0.99∗∗∗ 1.31∗∗∗ 0.09 173 −2384

−0.10 0.13 0.79∗∗∗ 0.12∗∗∗ 1.67∗

0.08 0.06 0.76∗∗∗ 0.16∗∗∗ 2.80∗∗

−0.22∗∗∗ −0.10 0.03 0.21 0.85∗∗∗ 0.14∗∗∗ 1.78∗∗∗

2.11∗∗∗ 2.39∗∗∗ 2.36∗∗∗ 1.44∗∗∗ 1.73∗∗∗ 1.52∗∗∗ 1.74 14.14 106

0.71∗∗∗ 0.70∗∗∗ 0.70∗∗∗ 0.09∗∗∗ 0.08∗∗∗ 0.08∗∗∗ 0.85 −1.59∗∗ −1.46∗∗ 2.13∗∗∗ 1.14∗∗∗ 2.92∗

2.09∗∗∗ 1.16∗∗∗ 0.25

2.09∗∗∗ 1.16∗∗∗ 0.16

287.96

287.70

260.71

260.14

212.68

210.49

208.32

199.01

275.18

268.05

268.55

276.09

276.57

257.90

239.88

260.99

265.13

266.44

262.15

264.25

252.32

251.72

0.07

0.06

0.05

0.04

0.20

0.07

0.02

0.00

0.08

0.00

0.00

0.20

0.25

0.11

0.03

0.00

0.00

0.00

0.00

0.11

0.00

0.00

The table presents the estimated empirical models described in Eq. (2.5) by country. According to the BIC information criterion, the specification that best describes the data requires that only the intercept and the standard deviation to switch between two states. The second row refers to the proxy indicating the institutional sector imbalance or the policy action considered in the time-varying transition probabilities specification (2.3) and (2.4). The third row of the table specifies the transformation of the proxy and the number of consecutive months required to observe statistically significant changes in the same direction on the transition probabilities. According to our notation, the first argument C denotes a quantitative transformation and Q means a qualitative transformation of the proxy indicator. The second argument simply specifies the number of months, which could be between one and three. As an example, Q#3 means that the proxy underwent a qualitative transformation and three months were necessary to observe the change in the same direction of the proxy indicator in order to have a statistically significant impact on the transition probabilities, either P 11 (remaining in regime 1) or P 22 (remaining in regime 2). The last row gives the LR test p-value of the restricted model with constant transition probabilities against the unrestricted model with time-varying transition probabilities. †

Shorter sample due to data availability. The data regarding central bank funding start in January 2001 for Greece, April 2001 for Ireland and December 1999 for Portugal. LR tests the restricted model with constant transition probabilities (see Table 1) against the unrestricted time-varying transition probability model. According to our notations in the table above: [trade balance] is the trade balance proxy summarizing imbalances caused by trade flows, [NFA] is the net foreign assets proxy summarizing imbalances caused by capital flows, [central bank] is the central bank share of funding, [Gov’t] is the government liquidity management. ‡

C. Dragomirescu-Gaina, D. Philippas / The Journal of Economic Asymmetries 10 (2013) 21–31

Table 2 Markov switching models with time-varying transition probabilities.

C. Dragomirescu-Gaina, D. Philippas / The Journal of Economic Asymmetries 10 (2013) 21–31

29

ernment liquidity management. In fact, having a common monetary policy, but national fiscal policies and country-specific regulations for the financial sector has not been a very good shield in face of negative shocks. Appendix A This appendix contains Tables A1–A6. Table A1 Descriptive statistics of government bond yields’ realized volatility.

Mean Std. dev. Skew. Kurt. J–B ADF

GR

NL

FR

DE

IE

IT

PT

ES

EUR3M

1.02 0.88 3.03 16.93 161.8 −4.36

0.96 0.65 1.95 8.31 304.9 −3.9∗

0.90 0.56 1.86 8.37 299.8 −7.90

1.07 0.84 2.26 9.39 430.0 −4.91

0.90 0.67 2.62 15.47 127.3 −4.27

0.94 0.65 2.64 13.18 921.6 −4.37

1.04 0.86 2.23 8.81 376.3 −6.01

0.97 0.76 3.17 17.3 171.3 −9.03

0.04 0.05 3.10 14.54 120.4 −6.66

The table presents the summary statistics of the realized volatility of bond yields derived from eight EMU countries and the realized volatility for the three-month Euribor rate. It shows the mean (%), standard deviation, skewness, kurtosis, Jarque–Bera (J–B) test and ADF unit root test (intercept and linear trend). A plus sign (+) denotes that the null hypothesis of normality (J–B test) of the series is accepted at the 1% level of significance. The null hypothesis for ADF is the existence of a unit root, so the time series are tested for stationarity at the 1% and 5% levels of significance (t cr,1% = −4.01, t cr,5% = −3.43). One and two stars (∗ and ∗∗ ) denote that the hypothesis of the unit root existence is accepted at the 5% and 1% level of significance, respectively. Table A2 Descriptive statistics of core inflation idiosyncratic shocks.

Mean Std. dev. Skew. Kurt. J–B ADF

GR

NL

FR

DE

IE

IT

PT

ES

−0.30 0.13 −0.44 6.43 87.33 −11.82

0.20 0.14 −0.82 7.89 184.35 −12.8

−0.01 0.11 −0.41 6.09 71.18 −15.2

0.40 0.09 −0.14 6.24 73.22 −10.77

−0.10 0.14 −0.46 6.39 85.81 −13.08

−0.14 0.09 0.37 5.84 59.84 −11.32

0.10 0.14 −0.55 6.66 101.3 −13.01

0.08 0.11 0.14 4.23 11.09 −14.97

The table presents the summary statistics of the idiosyncratic shock (residual series of AR-GARCH(1, 1)) of the ration between country core inflation and euroarea total. It shows the mean (%), standard deviation, skewness, kurtosis, Jarque–Bera (J–B) test and ADF unit root tests (intercept and linear trend). A plus sign (+) denotes that the null hypothesis of normality (J–B test) of the series is accepted at the 1% level of significance. The null hypothesis for ADF is the existence of a unit root, so the time series are tested for stationarity at the 1% and 5% levels of significance (t cr,1% = −4.01, t cr,5% = −3.43). One and two stars (∗ and ∗∗ ) denote that the hypothesis of the unit root existence is accepted at the 5% and 1% level of significance, respectively. Table A3 Descriptive statistics of the trade balance.

Mean Std. dev. Skew. Kurt. J–B ADF

GR

NL

FR

DE

IE

IT

PT

ES

−1.96 0.54 0.92 3.12 23.88 −0.41∗∗

0.09 0.02 −0.84 3.37 20.82 −7.76

−0.08 0.07 −0.02 1.78 10.30 −4.46

0.17 0.03 −0.43 2.68 5.96+ −2.86∗∗

0.39 0.06 −0.08 2.22 4.43+ −1.94∗∗

−0.01 0.04 −0.09 2.44 2.40+ −1.58∗∗

−0.55 0.12 1.16 4.32 50.30 −1.19∗∗

−0.36 0.12 −0.06 2.34 3.07+ −0.46∗∗

The table presents the summary statistics of the (seasonally adjusted) ratio between trade balance and exports, derived for the eight EMU countries. It shows the mean, standard deviation, skewness, kurtosis, Jarque–Bera (J–B) test and ADF unit root tests (intercept and linear trend). A plus sign (+) denotes that the null hypothesis of normality (J–B test) of the series is accepted at the 1% level of significance. The null hypothesis for ADF is the existence of a unit root, so the time series are tested for stationarity at the 1% and 5% levels of significance (t cr,1% = −4.01, t cr,5% = −3.43). One and two stars (∗ and ∗∗ ) denote that the hypothesis of the unit root existence is accepted at the 5% and 1% level of significance, respectively. Table A4 Descriptive statistics of the net external assets ratio.

Mean Std. dev. Skew. Kurt. J–B ADF

GR

NL

FR

DE

IE

IT

PT

ES

0.08 0.21 −0.19 1.78 11.46 −1.93∗∗

−0.18 0.12 0.30 2.08 8.50 −1.52∗∗

0.01 0.10 −0.06 2.10 5.68+ −0.86∗∗

0.22 0.16 −0.59 2.06 16.18 −0.98∗∗

−0.04 0.11 0.65 2.23 16.18 −2.60∗∗

−0.72 0.28 −0.09 2.87 0.36+ −1.63∗∗

−1.09 0.48 0.43 2.11 10.72 −0.63∗∗

−0.55 0.40 0.02 1.75 10.79 −2.59∗∗

The table presents the summary statistics of the (seasonally adjusted) net external assets to total external assets, derived for the eight EMU countries. It shows the mean, standard deviation, skewness, kurtosis, Jarque–Bera (J–B) test and ADF unit root tests (intercept and linear trend). A plus sign (+) denotes that the null hypothesis of normality (J–B test) of the series is accepted at the 1% level of significance. The null hypothesis for ADF is the existence of a unit root, so the time series are tested for stationarity at the 1% and 5% levels of significance (t cr,1% = −4.01, t cr,5% = −3.43). One and two stars (∗ and ∗∗ ) denote that the hypothesis of the unit root existence is accepted at the 5% and 1% level of significance, respectively.

30

C. Dragomirescu-Gaina, D. Philippas / The Journal of Economic Asymmetries 10 (2013) 21–31

Table A5 Descriptive statistics of central bank funding.

Mean Std. dev. Skew. Kurt. J–B ADF

GR

NL

FR

DE

IE

IT

PT

ES

8.51 11.0 1.28 3.21 39.81 −0.92∗∗

1.13 0.48 1.03 5.25 64.8 −4.57

1.11 0.94 1.60 4.96 99.04 −1.74∗∗

2.60 0.96 0.17 2.32 4.09+ −1.5∗∗

6.83 5.83 1.10 2.59 29.62 −1.83∗∗

1.68 1.88 2.72 9.41 495.6 0.25∗∗

2.85 3.54 1.41 3.38 53.16 −0.91∗∗

2.61 2.72 3.13 12.15 860.51 −0.57∗∗

The table presents the summary statistics of the central bank share of funding in the total banks’ liabilities (seasonally adjusted), derived for the eight EMU countries. It shows the mean, standard deviation, skewness, kurtosis, Jarque–Bera (J–B) test and ADF unit root tests (intercept and linear trend). A plus sign (+) denotes that the null hypothesis of normality (J–B test) of the series is accepted at the 1% level of significance. The null hypothesis for ADF is the existence of a unit root, so the time series are tested for stationarity at the 1% and 5% levels of significance (t cr,1% = −4.01, t cr,5% = −3.43). One and two stars (∗ and ∗∗ ) denote that the hypothesis of the unit root existence is accepted at the 5% and 1% level of significance, respectively.

Table A6 Descriptive statistics of government cash and debt management. GR Mean Std. dev. Skew. Kurt. J–B ADF

0.21 0.48 1.39 6.28 129.3 −3.28∗∗

NL

−0.60 0.20 3.56 16.37 1606.9 −3.45∗

FR

DE

IE

IT

PT

ES

−0.80 0.08 −0.01 2.05 6.23 −1.89∗∗

−0.68 0.09 0.75 2.46 17.87 0.06∗∗

−0.73 0.11 1.01 4.38 41.9 −4.23

−0.65 0.14 0.05 1.70 11.83 −1.18∗∗

0.88 1.03 1.43 3.88 62.7 −2.16∗∗

0.12 0.37 0.05 2.27 3.82+ −1.54∗∗

The table presents the summary statistics of the ratio between government deposits (seasonally adjusted) at the banks and loans taken from banks, derived for the eight EMU countries. It shows the mean, standard deviation, skewness, kurtosis, Jarque–Bera (J–B) test and ADF unit root tests (intercept and linear trend). A plus sign (+) denotes that the null hypothesis of normality (J–B test) of the series is accepted at the 1% level of significance. The null hypothesis for ADF is the existence of a unit root, so the time series are tested for stationarity at the 1% and 5% levels of significance (t cr,1% = −4.01, t cr,5% = −3.43). One and two stars (∗ and ∗∗ ) denote that the hypothesis of the unit root existence is accepted at the 5% and 1% level of significance, respectively.

References Abad, P., Chuliá, H., & Gómez-Puig, M. (2010). EMU and European government bond market integration. Journal of Banking and Finance, 34(12), 2851–2860. Acharya, V. V., & Skeie, D. (2011). A model of liquidity hoarding and term premia in inter-bank markets. CEPR, Discussion Papers, No. 8705. Attinasi, M. G., Checherita, C. D., & Nickel, C. (2009). What explains the surge in euro area sovereign spreads during the financial crisis of 2007–09? ECB Working Paper No. 1131. Baldacci, E., & Kumar, M. (2010). Fiscal deficits, public debt, and sovereign bond yields. IMF, Working Papers, 10/184. Beber, A., Brandt, M. W., & Kavajecz, K. A. (2009). Flight-to-quality or flight-to-liquidity? Evidence from the euro-area bond market. Review of Financial Studies, 22(3), 925–957. Berger, H., & Nitsch, V. (2010). The Euro’s effect on trade imbalances. IMF, Working Papers, 10/226. Blinder, A. S., Ehrmann, M., Fratzscher, M., De Haan, J., & Jansen, D. J. (2008). Central bank communication and monetary policy: A survey of theory and evidence. Working Papers Series No. 13932. National Bureau of Economic Research. Boyer, B. H., Kumagai, T., & Yuan, K. (2006). How do crises spread? Evidence from accessible and inaccessible stock indices. Journal of Finance, 61(2), 957–1003. Chen, R., Milesi-Ferretti, G. M., & Tressel, T. (2013). External imbalances in the eurozone. Economic Policy, 28, 101–142. Chordia, T., Sarkar, A., & Subrahmanyam, A. (2005). An empirical analysis of stock and bond market liquidity. Review of Financial Studies, 18(1), 85–129. Christiansen, C. (2007). Volatility-spillover effects in European bond markets. European Financial Management, 13(5), 923–948. Codogno, L., Favero, C., & Missale, A. (2003). Yield spreads on EMU government bonds. Economic Policy, 18(37), 503–532. Cox, J. C., Ingersoll, J. E., & Ross, S. A. (1985). A theory of the term structure of interest rates. Econometrica, 53, 385–407. Dailami, M., Masson, P. R., & Padou, J. J. (2008). Global monetary conditions versus country-specific factors in the determination of emerging market debt. Journal of International Money and Finance, 27(8), 1325–1336. Dean, A., Durand, M., Fallon, J., & Hoeller, P. (1990). Saving trends and behaviour in OECD countries. OECD Economic Studies, 14. Égert, B. (2007). Central bank interventions, communication and interest rate policy in emerging European economies. Journal of Comparative Economics, 35(2), 387–413. Favero, C., Pagano, M., & Von Thadden, E. L. (2010). How does liquidity affect government bond yields?. Journal of Financial and Quantitative Analysis, 45(1), 107–134. Geyer, A., Kossmeier, S., & Pichler, S. (2004). Measuring systematic risk in EMU government yield spreads. Review of Finance, 8(2), 171–197. Giannone, D., Lenza, M., Pill, H., & Reichlin, L. (2012). The ECB and the interbank market. The Economic Journal, 122, 467–486. Gonzalez-Rozada, M., & Levy-Yeyati, E. (2008). Global factors and emerging market spreads. The Economic Journal, 118(12), 1917–1936. Hamilton, J. D. (1990). Analysis of time series subject to change in regime. Journal of Econometrics, 45, 39–70. Howe, H., & Pigott, C. (1991). Determinants of long-term interest rates: An empirical study of several industrial countries. Federal Reserve Bank of New York Quarterly Review, 16(4), 12–28. Kim, C. J. (1994). Dynamic linear models with Markov switching. Journal of Econometrics, 60(1–2), 1–22. Kim, S. J., Moshirian, F., & Wu, E. (2006). Evolution of international stock and bond market integration: Influence of the European Monetary Union. Journal of Banking and Finance, 30(5), 1507–1534. Lane, P. (2012). The European sovereign debt crisis. Journal of Economic Perspectives, 26(3), 49–67. Lane, P., & Milesi-Ferretti, G. M. (1999). The external wealth of nations: Measures of foreign assets and liabilities for industrial and developing countries. IMF, Working Papers, No. 115.

C. Dragomirescu-Gaina, D. Philippas / The Journal of Economic Asymmetries 10 (2013) 21–31

31

Longstaff, F. A., & Schwartz, E. S. (1992). Interest rate volatility and the term structure: a two-factor general equilibrium model. Journal of Finance, 47(4), 1259–1282. Longstaff, F. A., Jun, P., Lasse, H. P., & Singleton, K. J. (2011). How sovereign is sovereign credit risk? American Economic Journal: Macroeconomics, 3(2), 75–103. Manganelli, S., & Wolswijk, G. (2009). What drives spreads in the euro area government bond market?. Economic Policy, 24(58), 191–240. Mihaljek, D., & Klau, M. (2008). Catching-up and inflation in transition economies: The Balassa–Samuelson effect revisited. BIS, Working Papers, No. 270. Nicoletti, G. (1988). A cross-country analysis of private consumption, inflation and the debt neutrality hypothesis. OECD Economic Studies, 11. Pagano, M., & von Thadden, E. L. (2004). The European bond markets under EMU. Oxford Review of Economic Policy, 20(4), 531–554. Pastor, L., & Veronesi, P. (2009). Learning in financial markets. NBER, No. 14646. Schuknecht, L., Von Hagen, J., & Wolswijk, G. (2009). Government risk premiums in the bond market: EMU and Canada. European Journal of Political Economy, 25(3), 371–384.