Credit rating agency downgrades and the Eurozone sovereign debt crises

Accepted Manuscript Title: Credit Rating Agency Downgrades and the Eurozone Sovereign Debt Crises Author: Christopher F. Baum Dorothea Sch¨afer Andreas Stephan PII: DOI: Reference:

S1572-3089(16)30024-9 http://dx.doi.org/doi:10.1016/j.jfs.2016.05.001 JFS 437

To appear in:

Journal of Financial Stability

Received date: Revised date: Accepted date:

27-6-2015 1-1-2016 4-5-2016

Please cite this article as: Christopher F. Baum, Dorothea Sch¨afer, Andreas Stephan, Credit Rating Agency Downgrades and the Eurozone Sovereign Debt Crises, (2016), http://dx.doi.org/10.1016/j.jfs.2016.05.001 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Credit Rating Agency Downgrades and the Eurozone Sovereign Debt Crises Department of Economics, Boston College, Chestnut Hill, MA 02467 USA b DIW Berlin, Mohrenstraße 58, 10117 Berlin, Germany c Jönköping International Business School, Jönköping University, Sweden d The Ratio Institute Stockholm, Sweden

us

cr

a

ip t

Christopher F Bauma,b,1 , Dorothea Schäferb,c , Andreas Stephanc,b,d

Abstract

Ac ce p

te

d

M

an

This paper studies the reaction of the Euro’s value against major currencies to sovereign rating announcements from Moody’s, S&P and Fitch CRAs during the Eurozone debt crisis in 2010–2012 based on event study methodology combined with GARCH models. We also analyze how the yields of French, Italian, German and Spanish government long-term bonds were affected by CRA announcements. Our results reveal that CRA downgrades, watchlist and outlook announcements had no impact on the value of the Euro currency but increased exchange rate volatility. At the same time, downgrades as well as negative outlook announcements increased the yields of French, Italian, and Spanish bonds and even affected the German bond’s yields. This shows that the monetary union has led to a breakdown of the consequences of the rating shocks between currency value and sovereign bond yields. The reason is that part of the rating shock is absorbed by an internal repricing of sovereign bonds. (JEL Classification Numbers: G24, G01, G12, G14, E42, E43, E44, F31, F42, F65) Key words: Credit Rating Agencies, Euro Crisis, Sovereign Debt, Euro Exchange Rate

Email addresses: [email protected] (Christopher F Baum), [email protected] (Dorothea Schäfer), [email protected] (Andreas Stephan) 1 Corresponding author. Phone 1-617-834-4615, Fax 1-617-552-2308.

Page 1 of 46

1. Introduction Since the bursting of the real estate bubble in the United States in the summer

ip t

of 2007, European economies have been badly damaged by banking and sovereign debt crises. Most European Union member states suffered from rapidly increasing

cr

fiscal deficits. Credit rating agencies (CRAs) responded to these developments with a series of downgrades.2 The Eurozone crisis reached its height during our period

us

of study: the years 2010–2012. Within this period particularly severe downgrades occurred. For example, on January 13, 2012, S&P announced credit rating down-

an

grades for nine EU member states, including France, which lost its AAA rating to AA+, while Portugal and Cyprus were assigned junk-bond ratings. In addition to this, 14 countries within the EU were given negative outlooks, while only Germany’s

M

premium credit rating remained unaffected. The goal of this paper is to examine the impact of CRA rating announcements on the exchange rate movements and bond

d

yields of large Eurozone countries during the worst of the Eurozone crisis.

te

Sovereign credit ratings are employed as an indicator for the quality of a country’s economic fundamentals and its perceived ability to repay its sovereign debt. Many

Ac ce p

investors sell the affected country’s sovereign bonds if their ratings deteriorate. In most parts of the world the ‘flight to quality’ goes hand in hand with the retreat of capital from the country and an ensuing devaluation of the domestic currency.3 However, the situation is different in the recent debt crisis of the Eurozone. The

flight to quality is not necessarily a flight away from the common currency. The 19 member states share a common currency but retain full fiscal sovereignty, subject only to the rather toothless Stability and Growth Pact. Investors can retain the Euro but still shift their investments from downgraded countries to more stable ones. And 2

The CRAs rating grades are defined in Table 1. For example, Chen et al. (2013) study the impact of rating changes on private investment starting from the assumption that sovereign rating downgrades are associated with an increase in capital outflows from the downgraded country. 3

1 Page 2 of 46

Ac ce p C

ip t

us

Definition Prime High grade

Upper medium grade

an

Fitch AAA AA+ AA AAA+ A ABBB+ BBB BBBBB+ BB BBB+ B BCCC

Lower medium grade

M

d

S&P AAA AA+ AA AAA+ A ABBB+ BBB BBBBB+ BB BBB+ B BCCC+ CCC CCCCC C SD

te

Moody’s Aaa Aa1 Aa2 Aa3 A1 A2 A3 Baa1 Baa2 Baa3 Ba1 Ba2 Ba3 B1 B2 B3 Caa1 Caa2 Caa3 Ca

cr

Table 1: CRA Rating Grades and Definitions

Non-investment grade Speculative Highly speculative

Substantial risks

Extremely speculative DDD DD D

In default

2 Page 3 of 46

indeed, there is anecdotal as well as strong empirical evidence that investors behave in this way. Beck et al. (2015) show that the 2013 announcement of the Swedish Central Bank (Riksbank) of having shifted its Euro currency reserves from Spanish

ip t

and Italian bonds to German bonds was not a singular phenomenon.4 They find that portfolio reallocation was common within the Eurozone, both with respect to

cr

external investors and intra-Eurozone investors. In the extreme, if investors in the Eurozone reacted to negative sovereign rating news by selling bonds of some member

us

states and buying bonds of others, the Euro’s value would not be affected at all. The existence of the Euro currency union may decouple the usually highly corre-

an

lated movements of prices of downgraded sovereign bonds and the country’s terms of trade. In the absence of the Euro, the currency of a country with an unfavorable

M

balance of payments would depreciate, its domestic goods would become cheaper to outsiders and its current account would consequently improve. This adjustment

d

occurs when the demand for this country’s export and import goods is relatively elastic. But, just as this country’s current account would improve, those of the im-

te

porting countries in the region would worsen. To maintain competitiveness, these

Ac ce p

other countries would then try to devalue their own currency. These ‘competitive devaluations’ could cause speculative attacks on one currency after the other. Such risk is absent within the Eurozone. Second, and closely related, the Euro protects ailing member states from a devaluation of its domestic goods vis-à-vis foreign goods (goods from the non-Eurozone states), and thus from a deterioration of its relative wealth position. The disadvantage is that with the common currency, single member states cannot engineer a currency devaluation in order to stimulate export-led domestic growth. disadvantage. Our research focuses on the implications of this unique feature of the Eurozone during the period of maximum stress. We pose two research questions: First, how 4

See the Swedish economic newspaper Dagens Industri, June 4, 2013.

3 Page 4 of 46

is the Euro currency’s value affected by negative rating signals, given that intraEurozone reallocation of capital is feasible? Second, what impact do negative rating announcements of any member of the Eurozone have on the price of sovereign bonds

ip t

of large Eurozone members? A particular strand of the credit rating literature points to cross-country contagion effects of rating news (e.g., Arezki et al. 2011,

cr

Böninghausen & Zabel 2015) which affect some countries more than others (e.g., Afonso et al. 2012). The motivation for focusing exclusively on the highly liquid

us

bonds of large countries is thus that within a currency union, cross-country contagion should be most visible in the large countries’ bond yields.

an

Following prior research on the impact of sovereign rating changes, we focus mainly on their short-term effects. Therefore, we employ an event-study method-

M

ology combined with a multivariate GARCH model. The event study allows us to evaluate the effects of ratings changes made on different dates, while the GARCH

d

methodology allows us to estimate their effects on both levels and volatilities, as well as take account of common shocks.

te

With this common methodology, we estimate two separate models: an exchange

Ac ce p

rate model and a CAPM model to explain sovereign bond yields for large Eurozone countries. The CAPM allows us to consider the broader market for all Eurozone borrowers’ sovereign debt, rather than focusing on differentials between each borrower’s yield and that of a single benchmark such as the German bund. While the impact of CRA announcements on sovereign bonds’ yields has been

studied before, relatively little work so far has been undertaken on the association between the value of the currency and CRA rating announcements in the recent European debt crisis. Notable exceptions either ignore the specific implications of a currency union (Alsakka & ap Gwilym 2012, Alsakka & ap Gwilym 2013) or concentrate on longer-term effects of changing economic fundamentals and important policy actions including rating announcements (Ehrmann et al. 2013). There has been no specific research to date on the impact of downgrades and other an4 Page 5 of 46

nouncements within a currency union on both the common currency’s movements and the repricing of the country risk of large Eurozone members in times of crisis. Likewise, no research on the impact of CRA statements on currency fluctuations

ip t

specifically examines the later, most severe phase of the European debt crisis. Our work contributes to fill these gaps.

cr

We seek to identify whether rating announcements of Eurozone member states by the three leading credit rating agencies (Standard & Poor’s, Moody’s, and Fitch)

us

were an active driving force for both the value of the Euro and large Eurozone members’ government bond yields. Fratzscher (2011) has shown that the impact

an

of sovereign ratings on capital outflows changed during the crisis.5 Therefore, the focus in our analysis is the Eurozone debt crisis from 2010–2012, as we expect to find

M

different results for this period compared to the prior period. Furthermore, we take concerns of CRA actions lagging behind the market into account (Ferri & Stiglitz

d

1999 and Reinhart 2002) and scrutinize the results for possible reverse causality effects.

te

The remainder of the paper is organized as follows. The next section reviews

Ac ce p

previous literature related to our research questions. In Section 3, we describe the empirical analysis and Section 4 presents the results. Finally, Section 5 concludes. 2. Literature Review

There is ample research on the impact of credit rating announcements on finan-

cial markets. This paper is particularly related to three strands of this literature. First, it relates to the growing research on the effect of sovereign rating changes on debt markets around economic crisis periods. Using a panel of 16 emerging markets, including East Asian, Eastern European, and Latin American economies in 5

Specifically, Fratzscher finds that in the crisis period between 2007 and 2009 countries with a worse sovereign rating, worse institutions or higher short-term debt faced larger capital outflows than in an earlier period.

5 Page 6 of 46

the period 1990–2000, Kaminsky & Schmukler (2002) explore how sovereign ratings affect bond and equity markets. They find that changes in ratings have an impact on sovereign bond spreads and equity returns. Kräussl (2005) shows that negative

ip t

ratings significantly increase an index of speculative market pressure consisting of daily nominal exchange rate changes, daily short-term interest rate changes and

cr

daily stock market changes. In contrast, ratings upgrades and positive outlooks show a weak or even insignificant impact.

us

Second, our work is linked to the credit rating literature on cross-country contagion effects. Kaminsky & Schmukler (2002) report that rating changes are transmit-

an

ted across countries, with neighbor-country effects being more significant. Recent research on the Eurozone indicates the presence of statistically significant spillover

M

effects of rating news from an event country to markets of non-event countries. Arezki et al. (2011) report that rating downgrades to near-speculative grade for

d

sovereigns are especially contagious across countries and financial markets. For example, a rating downgrade of Greece from A- to BBB+ grade, as announced by

te

Fitch on 8 December 2009, resulted in substantial spillover effects across mem-

Ac ce p

bers of the EMU including 17 and 5 basis point increases in Greek and Irish CDS spreads, respectively. Afonso et al. (2012) find that ratings downgrades of lowerrated countries have strong spillover effects on higher-rated sovereigns in the region, which is consistent with prior studies by Gande & Parsley (2005) and Ismailescu & Kazemi (2010). Interestingly, CRA announcements by different agencies differ in their impact. Rating announcements by Moody’s tend to have the strongest persistence effect, of a 1.5% higher bond yield spread for six months following a rating downgrade. Our work is also complementary to that of Blommestein et al. (2016), who analyze European sovereign credit default swap (CDS) spreads from September 2008–December 2011 using a regime-switching framework. In contrast to our event study methodology, in which we seek to identify the impact of CRA announcements on high-frequency sovereign bond yields, Blommestein et al. (2016) identify tran6 Page 7 of 46

quil and turbulent regimes within the data, as well as international spillover effects across borrowers. The third closely related strand includes research on the association of sovereign

ip t

rating changes and the value of their currency. Alsakka & ap Gwilym (2012) completed the first empirical study on this question. With a dataset of 124 developed

cr

and emerging economies, they find that rating news have an impact on the affected country’s exchange rate and cause cross-country spillovers. Moreover, their results

us

reveal that sovereign watch and outlook signals as well as multiple-notch rating changes have a greater effect than one-notch rating changes.

an

In another recent study, Alsakka & ap Gwilym (2013) analyzed the direct shortterm impact of sovereign credit ratings on foreign exchange markets with data from

M

2000–2010. Their event study covers a broad set of countries in Europe and Central Asia. The evidence reveals that rating news affects a country’s bilateral exchange

d

rate against the US dollar as well as the exchange rates of other countries. The domestic and foreign spillover effect is reported to be stronger during the crisis

te

period from October 2006 to July 2010. Negative news from all three major agencies

Ac ce p

have an impact, but Fitch’s announcements triggered the strongest reactions. Finally, a recent study by Ehrmann et al. (2013) examines the effects of macroe-

conomic fundamentals, policy actions including rating news, and the public debate on the evolution of the Euro exchange rate and its volatility from October 1st, 2009 to November 30, 2011. Specifically, all rating changes, including watchlist and outlook announcements, are transformed into an indicator variable that represents either a rating improvement or deterioration. They find that similar to many of the other potential determinants, the rating signals did not significantly affect the Euro exchange rate during this earlier phase of the debt crisis. Our research complements Afonso et al. (2012), Alsakka & ap Gwilym (2013), Ehrmann et al. (2013) and Blommestein et al. (2016) by focusing on CRA announcements as a determinant of the exchange rate and contagion. We consider only the 7 Page 8 of 46

Eurozone member countries as our main interest is the impact of CRA announcements on the common currency’s value in combination with a possible reallocation of Eurozone sovereign bond portfolios. We primarily focus on short-term effects

ip t

during the period of the acute debt crisis, 2010–2012, unlike Alsakka & ap Gwilym (2013) and Afonso et al. (2012). Finally, Afonso et al. (2012) use the spreads of

cr

Eurozone sovereign bond yields to German bond yields for the EU countries. Our

us

preferred CAPM approach, in contrast, uses excess bond yields. 3. Empirical analysis

an

3.1. Rating announcements and exchange rate movements

The daily spot exchange rate movements (Figure 1) reveal that the Euro had

M

already depreciated significantly against the US dollar before the Eurozone debt crisis started in 2010. In March 2009 Ireland lost its triple A status because of an increased sovereign debt burden after a series of recapitalization measures for

d

domestic banks. The downward pressure eased in mid-spring 2009 as a result of

te

government assistance to financial institutions in the Eurozone. Fiscal assistance in combination with negative growth figures caused national debt ratios to raise further

Ac ce p

(see Figure 2). In late 2009, after the Greek government announced severe upward corrections of the deficit and debt data of 2005–2008, a series of downgrades for Greek government bonds was announced, and devaluation pressure rose again. In June 2010 the Euro reached its lowest point within the period of the acute Eurozone debt crisis 2010–2012: 0.837 Euro per USD according to the closing price on 8 June 2010. This period of extensive Euro depreciation occurred in conjunction with rising public debt levels in Greece, Ireland, Portugal and Spain and the negative sovereign rating announcements made by each credit rating agency within this time frame. The situation stabilized somewhat after Ireland and Greece were provided with a series of bailout packages. The stabilizing effect did not persist, though, as in 2011 more countries of the EMU were becoming involved. Rising fears among investors 8 Page 9 of 46

ip t

01/01/2009

01/01/2010

4.9 4.7 4.8 ln(Japanese yen/Euro spot rate) 4.6 4.5

te

Ac ce p

−.2

d

0

M

.2

an

us

.4

cr

Figure 1: Exchange Rates and CRA Downgrade Announcements, 2009–2012

01/01/2011 Date

01/01/2012

Event downgrade

ln(nominal effective trade weighted Euro rate)

ln(US dollar/Euro spot rate)

ln(British pound/Euro spot rate)

01/01/2013

ln(Japanese yen/Euro spot rate)

Sources : Datastream, Events: S&P, Fitch or Moody’s Downgrade of Eurozone States’ Sovereign Debt

9 Page 10 of 46

of possible spillover effects to other countries within the Eurozone put a substantial strain on the Euro. Figure 1 shows that the bailout packages were, in fact, accompanied by increased downward pressure on the value of the common currency.

ip t

The Euro lost value against the Japanese yen, the Swiss franc and the US dollar. Around the end of 2010 to the beginning of 2011, the common currency experienced

cr

a sudden slump in its value. At that time S&P and Fitch assigned junk bond status to Greek government bonds. In December 2011, a series of negative watch and out-

us

look revisions were announced by S&P and Fitch, followed by massive downgrade announcements for Eurozone countries in January 2012.

an

The Euro began depreciating again in the second quarter of 2011. In July 2011, Eurozone leaders decided in favor of a haircut of Greek bonds held by non-sovereign

M

lenders. In the beginning of 2012, the Euro reached a further minimum. The haircut was executed in February 2012 accompanied by S&P’s SD (in default) rating for

d

Greece. In the following month the Euro recovered to some degree, but downward pressure was renewed in early summer. When Moody’s threatened to downgrade

te

three top-rated European sovereigns in July 2012, the Euro’s value was again de-

Ac ce p

pressed, as the Euro lost about 10 percent of its value between January 2011 and July 2012. However, the overall change of the Euro’s value in 2011 and 2012 was miniscule. For example, against the US dollar, the Euro traded at the beginning of 2011 for 0.749 Euro per dollar. By the end of December 2012, the exchange rate was 0.758 Euro per dollar: a loss in value of only about one percent in two years. In the period from January 2010 through December 2012 (771 trading days), a

total of 176 rating announcements were made by the three CRAs. The most affected economies of the Eurozone debt crisis are Greece, Ireland, and Portugal along with Spain and Italy (the GIPSI countries). Figure 2 illustrates the downgrading activity using the example of S&P. Of these 160 CRA announcements, 132 were made during

10 Page 11 of 46

the most turbulent phase of the Eurozone debt crisis: 2011–2012 (Table 2).6 Table 2: CRA Rating Announcements 2010–2012, 2011–2012

15 22 20 47

32 52 48 132

ip t

44 70 62 176

an

Fitch Moody’s S&P Total

21 28 25 74

cr

Fitch Moody’s S&P Total

Down All

us

Outlook Watch 2010–2012 15 8 27 15 20 17 62 40 2011–2012 11 6 22 8 15 13 48 27

M

3.2. Event study methodology with GARCH models

Table 5 provides a description of the variables that are employed in the analysis. We combine event study methodology with a multivariate GARCH model for the

d

analysis of the short-term effects of the rating agencies’ downgrades on the trajecto-

te

ries of the Euro’s value and those of selected government bond excess yields during

Ac ce p

the Eurozone crisis. The GARCH model has the advantage that it can quantify effects from the events both on the mean and on the volatility of exchange rates and bond yields. The events are defined as binary variables, where event days are coded as 1 and non-event days are coded as 0. We test three types of events: watch, outlook and downgrade. Because the effects from the event might not only occur on the day of the announcement, but also before and after the announcement, we define event windows including days before and after the event, as is common in the event study literature. 6 Credit rating announcements published by Standard and Poor’s, Fitch Ratings, and Moody’s were obtained from the official websites of each credit rating agency. We ignore upgrades, as only Greece was upgraded after the July 2011 haircut. Previous research reports anyway that the effect of upgrades is weak compared to downgrades (e.g., Cantor & Packer 1996, Brooks et al. 2004 and Hooper et al. 2008).

11 Page 12 of 46

Figure 2: Debt ratios and rating movements of selected Eurozone countries

ip t

200

cr

150

us

100

50 2009

2010 Year

2011

an

2008

(a) Debt to GDP Ratios, 2008–2012 haircut on Greek debt in early 2012 reduced the debt ratio by the end of 2012 to about 158 percent of GDP.

te

d

∗ The

Germany Greece* Spain

M

EU (27 countries) France Ireland

2012

AAA

Ac ce p

AA AA! A

BBB BB B

CCC+ CCC

CCC! CC

C SD 1/1/2009

7/1/2010 Spain

1/1/2012

Portugal

Italy

7/1/2013 Ireland

Greece

(b) GIPSI countries: Rating activities of S&P 12 Page 13 of 46

ip t

13 January 2012

S&P

4 December 2012

Moody’s Moody’s

18 October 2011 13 February 2012

Moody’s

24 July 2012

M

d te

19 Nov 2012

Ac ce p

Moody’s

us

S&P

Event 15 Eurozone member states put on credit watch, France among them 16 Eurozone countries downgraded, France among them threat to downgrade all Eurozone (negative watch)

an

Date 5 December 2011

cr

Table 3: Significant Sovereign Debt Rating Announcements, 2010–2012

Source Standard and Poor’s (S&P)

Fitch’s Ratings

19 December 2011

Fitch’s Ratings

27 January 2012

Fitch’s Ratings

11 May 2012

threat to downgrade France Downgrading of six Eurozone countries (Italy, Malta, Portugal, Slovakia, Slovenia and Spain) Threat to downgrade 3 Aaa-rated economies the next 18 months (Germany, the Netherlands and Luxembourg) due to developments in Spain and Greece Downgrading of France Threat to downgrade France and other Eurozone countries (Belgium, Cyprus, Ireland, Slovenia and Spain) Downgrading of five Eurozone countries (Belgium, Cyprus, Italy, Slovenia and Spain) while Ireland gets a negative outlook. Fitch puts all Eurozone states negative watch if Greece exits.

13 Page 14 of 46

Table 4: Overview of the covered EU actions and events, 2010–2012

22 July 2011 02 September 2011 09 September 2011 27 October 2011

ip t

cr

us

an

Ac ce p

01 November 2011

M

May 2010 May 2010 November 2010 December 2010 February 2011 March 2011 May 2011 June 2011 June 2011 July 2011

d

03 10 29 07 14 24 16 17 20 04

Event Euro area Heads of State agree to offer, together with the IMF, financial support to Greece in the form of coordinated bilateral loans Announcement of an economic adjustment programme for Greece Agreement on the European Stabilisation Mechanism (ESM) Announcement of an economic adjustment programme for Ireland Decision on financial assistance to Ireland Agreement about ESM lending capacity of Euro 500 bn European Council agrees on the Euro Plus pact Official approval of the Euro 78 bn bailout package for Portugal Increase in effective lending capacity of EFSF to Euro 440 bn Finance ministers agree to broaden the EFSF mandate Decision to disburse the fifth tranche of the Greek rescue package (Euro 12 bn) Agreement about Euro 109 bn of new funds for the Greek package and a private sector involvement of 21 % The 5th EU/IMF/ECB Review Mission to Greece has left Athens unexpectedly Jürgen Stark resigns from the ECB’s Executive Board Restructuring of the second rescue package for Greece: increase in financing to Euro 130 bn and in private sector involvement to 50% Greek PM Papandreou announced his intention to hold a referendum over the rescue package, including the 50% private haircut ”Six-pack” approved by the European Council The ”fiscal pact” agreed by the EU in December 2011 is signed ECB allows the Eurosystem national central banks to temporarily accept additional collateral Agreement on Greek debt restructuring Eurozone backs a second Greek bailout of 130bn Euros Eurozone finance ministers announce an expansion of the EFSF and ESM ECB approves measures to increase collateral availability for counterparties Mario Draghi Speech: Whatever it Takes to Save the Euro Eurogroup postphones Greece’s fiscal consolidation targets by two years, and the programme’s control mechanisms are enhanced.

te

Date 25 March 2010

08 November 2011 30 January 2012 9 February 2012 21 13 30 20 26 27

February 2012 March 2012 March 2012 June 2012 June 2012 November 2012

14 Page 15 of 46

ip t cr

Table 5: Variable Descriptions Description

NEER

nominal effective Euro rate for Euro area-17 countries vis-à-vis the EER-19 group of trading partners, trade weighted Euro exchange rate, K ∈ {U SD, GBP, JP Y } France 5 year government bond yield Spain 5 year government bond yield Germany 5 year government bond yield Italy 5 year government bond yield downgrade, watch or outlook announcement (S&P, Fitch or Moody’s) European Union actions, see Table 4

vixeu

an

M

eu event

d

spotK/EU R FRA ESP GER ITA cra event

Source

ECB

us

Variable

ECB Datastream Datastream Datastream Banca d’Italia authors’ collection Ehrmann et al. (2013), and authors’ collection STOXX

Ac ce p

te

VSTOXX (symbol V2TX), volatility conveyed by EURO STOXX50 realtime options prices vixus near-term volatility conveyed by S&P 500 Datastream stock index option prices (CBOE) repo ECB’s repo rate Datastream (iK − iEU R ) K ∈ {U SD, GBP, JP Y } difference of St. Louis Fed 3-month LIBOR rates for currency K and Euro Economic Research cds credit default swaps on government bonds: Datastream ITA, GER, FRA and ESP rf government 5-year bond par yield of ECB triple A–rated Eurozone bonds (changing composition) market yield government 5-year bond par yield of all Eurozone bonds ECB Notes: All continuous variables exhibit a unit root in levels, but are stationary in first differences.

15 Page 16 of 46

We formulate two different models for the purpose of our analysis. In the first model, we study the relationship between the Euro exchange rate and CRA events, considering a number of control variables including important EU actions, listed in

ip t

Table 4. In the second model we formulate a CAPM approach to explain sovereign bond yields for the various countries.

cr

Model I is formulated as follows:

us

∆ ln (Spot)t = β0 + β1 cra eventt + β2 ∆(iK − iEU RIBOR )t + β3 ∆ ln(vixeu )t

an

+β4 ∆ ln(vixus )t + β5 eu eventt + β6 ∆ ln(repo)t + εt

where ∆ denotes a first differenced variable, and Spot ∈ {N EER , USD/Euro,

M

GBP/Euro, JPY/Euro} is the respective Euro exchange spot rate.7 In the first set of estimates, the dependent variable is the nominal effective trade-weighted nominal

d

Euro rate (N EER), which captures the value of the Euro against 19 other major

te

currencies. Higher values of the N EER indicate that the Euro has appreciated against other major currencies.8 To analyze the impact of CRA events via Eurozone

Ac ce p

capital outflows into ‘safe havens’, we also estimate the model for the bilateral rates of the Euro against the USD, GBP and JPY as a robustness check. As our major control variable we include the interest rate differential, (iK −

iEU RIBOR ), K ∈ {U SD, GBP, JP Y }, in the bilateral rate equations. The purpose of this variable is to capture uncovered interest parity effects on the Euro rate. We expect that the interest rate differential to be inversely related to the Euro rate: an increase in the interest rate differential leads to a depreciation of the Euro against 7

Unit root tests using a modified Dickey–Fuller (DF-GLS) test in which the series has been transformed by a generalized least-squares regression (Elliott et al. 1996) indicate that all continuous variables should be specified in first differences. 8 The NEER is calculated by the European Central Bank (ECB) as a weighted average of bilateral Euro exchange rates against 19 trading partners of the Euro area. (Schmitz et al. 2013).

16 Page 17 of 46

the respective currency and in consequence also lowers the effective Euro rate. As we do not have a trade-weighted average of interest rates of the major Eurozone trading partners, in the model for N EER we utilize the difference between three-

ip t

month money market rates (US LIBOR minus the EURIBOR rate) as a proxy, as the USD has a major weight in the N EER. We also include the change in the ECB’s

cr

repo rate to control for actions undertaken by the ECB during the sovereign debt crises. A further control variable is eu event, an indicator variable describing dates

us

of important actions undertaken by the EU during the crisis period 2010–2012, as listed in Table 4.9

an

Other control variables in Model I are the volatility indices of the EuroStoxx 50 (VSTOXX) and the Chicago Board Options Exchange Volatility Index (VIX).

M

The latter measures the implied volatility of the S&P 500 stock index over a 30-day period. In our context the vix variables are used as a proxy for the level of investor

d

uncertainty in European and US equity markets. Figure 3 illustrates movements of the VSTOXX and VIX series as well as the various interest rate differentials.

te

Model II is formulated to measure the effect of CRA announcements on gov-

Ac ce p

ernment bond yields. We hypothesize that in particular CRA downgrades trigger reallocation of capital within the Eurozone. As government bonds are a key asset in many institutional portfolios, geographic rebalancing of portfolios will affect the yields of these bonds. We expect that price changes of possible contagion effects from downgrades will be reflected in the yields of the highly liquid bonds of large Eurozone member countries. Specifically, we investigate the effect of ratings downgrades on the yields of France (FRA), German (GER), Italian (ITA), and Spanish (ESP) sovereign bonds. Figures 4 and 5 show the evolution of excess yields of bonds and of CDS spreads respectively. 9

We are grateful to an anonymous reviewer for recommending this control variable as a robustness check.

17 Page 18 of 46

ip t

01/01/2009

01/01/2010

4 3.5 ln VIX / VSTOXX 3 2.5

−2

Ac ce p

te

0

d

M

Interest differential (%)

2

an

us

4

cr

Figure 3: Interest Rate Differentials, VSTOXX, CBOE VIX, and CRA Downgrade Announcements, 2009–2012

01/01/2011 Date

01/01/2012

Event downgrade

int diff USD − EUR

int diff GBP − EUR

int diff JPY − EUR

ln CBOE VIX

ln Euro VSTOXX

01/01/2013

Sources : Datastream, Events: S&P, Fitch or Moody’s Downgrade of Eurozone States’ Sovereign Debt

18 Page 19 of 46

ip t

an M d te

−1

Ac ce p

0

1

Gov. bond excess yield (%) 2 3 4 5

6

us

7

cr

Figure 4: Five-year Sovereign Bond Excess Yields and CRA Downgrade Announcements, 2009– 2012

01/01/2009 01/07/2009 01/01/2010 01/07/2010 01/01/2011 01/07/2011 01/01/2012 01/07/2012 01/01/2013 Date Event downgrade

FRA

ESP

GER

ITA

Sources : Datastream, Events: S&P, Fitch or Moody’s Downgrade of Eurozone States’ Sovereign Debt

19 Page 20 of 46

ip t

d te

0

Ac ce p

100

200

M

CDS

300

an

400

us

500

cr

Figure 5: Credit Default Swaps and CRA Downgrade Announcements, 2009–2012

01/01/2009 01/07/2009 01/01/2010 01/07/2010 01/01/2011 01/07/2011 01/01/2012 01/07/2012 01/01/2013 Date Event downgrade

CDS FRA

CDS ESP

CDS GER

CDS ITA

Sources : Datastream, Events: S&P, Fitch or Moody’s Downgrade of Eurozone States’ Sovereign Debt

20 Page 21 of 46

The basis of Model II is a CAPM formulation that explains bond excess yields with the market excess yield. The strength and direction of this correlation is captured by the coefficient βbond . The higher is βbond , the higher the market risk

ip t

of the bond, and the higher should be its expected excess yield. This approach differs from that of Afonso et al. (2012), who base their event study on the observed

cr

bond yield spreads between country-specific bonds and German bonds. However,

Accordingly, the equation is

us

we prefer the CAPM approach because it also captures the market risk of the bond.

∆ (yieldt − rtf ) = α0 + α1 cra eventt + βbond ∆ (mktyieldt − rtf ) +α2 ∆ ln(vixeu )t {z

}

|

{z

an

|

bond excess yield

market excess yield

}

+α3 ∆ ln(vixus )t + α4 eu eventt + α5 ∆ ln(repo)t + νt ,

M

where rf denotes the risk-free yield, defined as the yield curve spot rate of a changing composition of government bonds of all Euro area issuers with a triple A rating.10

d

We expect that α0 = 0, as bond yields should be only related to their systematic

te

risk. The market yield, mktyieldt , is defined as the average yield of all government bonds of the Eurozone with a tenor of five years. The respective data are provided

Ac ce p

by the ECB. If βbond = 1, the respective bond has the same systematic risk as the entire bond market.

Furthermore, the variables vixeu and vixus are included as a measure for per-

ceived uncertainty in the European and US stock markets. Given that Eurozone government bonds constitute alternative investment opportunities, we hypothesize that increased uncertainty in equity markets should increase the demand for Eurozone sovereign bonds. As a consequence excess yields of government bonds should fall. However, there also might be adverse spillover effects from equity markets on bond yields if overall uncertainty in financial markets increases due to the crisis. 10

By the end of 2012 four Eurozone countries had triple A-rated bonds according to Standard & Poor’s: Finland, Luxembourg, the Netherlands and Germany.

21 Page 22 of 46

ip t

te

0

Ac ce p

1

d

2

M

3

an

us

4

cr

Figure 6: Eurozone government bond yields (all ratings and triple-A), ECB repo rate, and CRA Downgrade Announcements, 2009–2012

01/01/2009 01/07/2009 01/01/2010 01/07/2010 01/01/2011 01/07/2011 01/01/2012 01/07/2012 01/01/2013 Date Event downgrade

ECB repo rate (%)

Bond yield, all ratings (%)

Bond yield, only triple−A rated bonds (rf)

Sources : Datastream, ECB, Events: S&P, Fitch or Moody’s Downgrade of Eurozone States’ Sovereign Debt

22 Page 23 of 46

The CRA event may not only affect the mean equation, but also the variance of the dependent variable (Bollerslev 1986). For instance, one could expect that during event times we observe a higher variance of the excess yield compared to non-event

ip t

times. Diagnostic LM tests indicate the presence of GARCH effects after OLS estimation. The simplest generalized autoregressive conditional heteroskedasticity

cr

model is GARCH(1,1). One of the primary benefits of the GARCH specification is its parsimony in identifying the conditional variance. For instance, a specifica-

us

tion of multiplicative heteroskedasticity allows us to specify a linear combination of variables that are hypothesized to affect the conditional variance. The resulting





2 γ0 + γ1 σt−1 + λ1 ε2t−1 × [θ1 cra eventt + θ2 ∆ ln(vixeu )t

M

σt2 =

an

GARCH(1,1) conditional heteroscedasticity equation is specified as

+θ3 ∆ ln(vixus )t + θ4 eu eventt + θ5 ∆ ln(repo)t + θ6 ∆ ln cdst ] ,

d

see Judge et al. (1985, p. 843). In the variance equation we include the variable cds,

te

which is an index of credit default swaps for government bonds of the respective country. We use it as a proxy variable for the perceived investor risk for the govern-

Ac ce p

ment bonds from a specific country. We expect a positive relationship between cds and bond yield variance, since again higher systematic default risk should give rise to higher bond yield volatility.11 In principle we could have utilized more complicated models such as EGARCH

(Nelson 1991) but these models often give rise to numerical estimation problems and non-convergence. Therefore we have decided to estimate simple GARCH model specifications.12 We also noted that in some models, the residuals appear to be distributed with fatter tails than those of a normal distribution. To allow for this 11

A recent study by Klose & Weigert (2014) used CDS premia to evaluate the potential effects of a Eurozone breakup on sovereign bond yields over a subset of our 2011–2012 sample period. 12 We also found that higher order GARCH models tend to provide implausible estimates or suffer from convergence problems.

23 Page 24 of 46

flexibility, a generalized error distribution (GED) model is specified (Bollerslev et al. 1994).13 In contrast to the Euro exchange rate models, it is possible for Model II to allow

ip t

for cross-equation correlations of residuals. If residuals are correlated this can be interpreted as evidence that there are unobserved factors, such as common shocks,

cr

that affect the various markets. In the extreme case, it could indicate contagion effects across financial markets (e.g., Missio & Watzka 2011).

us

In our case we specify a varying conditional correlation multivariate GARCH model (Tse & Tsui 2002, Engle 2002) which is more flexible than the constant

an

correlation model. The specification tests based on the adjustment parameter show that the varying conditional correlation MGARCH is the preferred specification.

M

4. Results

Tables 6 and 7 provide summary statistics of the variables for models I and

d

II. Depending on the definition of event type and event window, one can see how

te

frequently the various events occurred. For instance, from Table 6 one can infer that

Ac ce p

downgrades according to the event window definition occurred during around 37% of all days, while watch announcements occurred during 19% of all trading days. As previous research—e.g., Alsakka & ap Gwilym (2012)—has done, we consider

not only downgrades but also weaker rating actions, such as watchlist and outlook events. The estimation results in Tables 8, 9 and 10 show the effect of downgrade, watch and outlook announcements on the effective Euro exchange rate and the bilateral rates of the USD, GBP and JPY vs. the Euro over the period 2010–2012. The event window for the impact of the respective event type on the Euro currency includes one day before the event and three days after it. 13

The GED distribution includes the normal distribution as a special case when its estimated shape parameter is 2.0. Note that in the estimation results Tables ln(shape) is reported.

24 Page 25 of 46

Table 6: Summary Statistics Model I, 2010–2012

Max. 1 1 1 1 0.0111 0.0184 0.0293 0.0379 0.0997 0.0923 0.1017 0.2500 0.2981 0.4055

us

cr

ip t

Mean Std. Dev. Skew. Min. 0.3664 0.4821 0.5547 0 0.1851 0.3887 1.6213 0 0.3155 0.4649 0.7940 0 0.1538 0.3610 1.9188 0 -0.0002 0.0035 -0.2075 -0.0138 -0.0001 0.0065 -0.2289 -0.0271 -0.0001 0.0049 0.0198 -0.0208 -0.0002 0.0082 0.0540 -0.0301 0.0008 0.0083 2.6261 -0.0700 0.0006 0.0077 2.5703 -0.0675 0.0006 0.0078 3.0959 -0.0700 -0.0003 0.0202 -2.4296 -0.2500 -0.0004 0.0627 0.6288 -0.2491 -0.0000 0.0722 0.7664 -0.3506

an

Variable [window] cra down [-1,+3] cra watch [-1,+3] cra outlook [-1,+3] eu event [-1,+3] ∆ ln (NEER) ∆ ln (USD/e) ∆ ln (GBP/e) ∆ ln (JPY/e) ∆(iU SD − iEU R ) ∆(iGBP − iEU R ) ∆(iJP Y − iEU R ) ∆(repo) ∆ ln vixeu ∆ ln vixus

te

d

M

Notes: daily frequency, period from 1.1.2010 to 31.12.2012 (767 obs), NEER= nominal effective exchange rate, GBP=British pound, JPY=Japanese yen, USD= US dollar, ∆ denotes first difference, x(t) − x(t − 1)

Table 7: Summary Statistics Model II, 2010–2012

Mean Std. Dev. 0.1695 0.3754 0.1447 0.3520 0.1851 0.3887 0.1538 0.3610 0.004 0.151 0.003 0.139 0.0002 0.067 -0.0002 0.032 0.002 0.047 0.0011 0.0560 0.0011 0.0570 0.0005 0.0564 -0.0004 0.0612

Ac ce p

Variable [window] cra down [0,+1] cra outlook [0,+1] cra watch [-1,+3] eu event [-1,+3] ∆(ESP-rf ) ∆(ITA -rf ) ∆(FRA - rf ) ∆(GER - rf ) ∆(market yield -rf ) ∆ln(CDS ESP) ∆ln(CDS ITA) ∆ln(CDS FRA) ∆ln(CDS GER)

Skew. Min. Max. 1.7618 0 1 2.0197 0 1 1.6213 0 1 1.9188 0 1 -0.636 -0.899 0.717 -0.477 -0.933 0.633 0.132 -0.925 0.915 0.336 -0.155 0.188 -0.138 -0.225 0.178 -0.4110 -0.4175 0.2826 -0.4991 -0.4520 0.2220 -0.1552 -0.2339 0.1918 -0.1656 -0.3714 0.3187

Notes: daily frequency, sample period from 1.1.2010 to 31.12.2012 (767 obs), event window=[0,+1], ∆ denotes first difference, x(t) − x(t − 1) (ESP-rf ), (ITA -rf ), (FRA - rf ), (GER - rf ) are government bond excess yields.

25 Page 26 of 46

Table 8: Estimation Results Model I for Euro Exchange Rates (2010–2012), Event Downgrade

eu event constant Variance equation cra event down ∆(repo) ∆ ln(vixeu )

0.0004 (0.0005) -0.0901*** (0.0277) 0.0300* (0.0156) -0.0501*** (0.0054) 0.0000 (0.0051) -0.0001 (0.0007) -0.0001 (0.0003)

0.2617 (0.1817) 7.3121 (6.6159) -0.7111 (2.6963) 5.3475** (2.4718) 0.1472 (0.2684) -12.6351*** (0.4349) 0.0710 (0.0456) 0.5693*** (0.1642) 0.4679*** (0.0743) 104.7(6) 0.0000 3305.1 767

0.3101* (0.1736) 5.0205 (7.9930) -0.9719 (2.4858) 4.8193** (2.3408) 0.0879 (0.2399) -11.3511*** (0.4641) 0.0332 (0.0429) 0.5669*** (0.1892) 0.5099*** (0.0743) 166.7(6) 0.0000 2851.6 767

eu event

constant

ARCH(1)

GARCH(1)

GED lnshape

Chi2(df) p-value Log-Likelihood Obs

cr

ip t

-0.0001 (0.0003) -0.0564*** (0.0193) 0.0173* (0.0093) -0.0147*** (0.0036) 0.0033 (0.0030) -0.0003 (0.0005) 0.0001 (0.0002)

Ac ce p

∆ ln(vixus )

0.0001 (0.0004) -0.1032*** (0.0233) 0.0221** (0.0102) -0.0368*** (0.0046) -0.0031 (0.0043) -0.0010* (0.0006) 0.0002 (0.0003)

us

∆ ln(vixus )

-0.0001 (0.0002) -0.0487*** (0.0125) 0.0160** (0.0064) -0.0160*** (0.0025) -0.0002 (0.0024) -0.0003 (0.0003) -0.0001 (0.0001)

an

∆ ln(vixeu )

∆ ln (JPY/e) (4)

M

∆(repo)

∆ ln (GBP/e) (3)

d

∆(interest)

∆ ln (USD/e) (2)

te

Mean equation cra event down

∆ ln (NEER) (1)

0.7508* (0.4400) 5.0988 (40.0128) 10.5460** (4.2377) 0.8998 (4.3560) 0.1053 (0.5325) -13.8596*** (0.6111) 0.0759*** (0.0257) 0.8406*** (0.0453) 0.5348*** (0.0809) 35.0(6) 0.0000 3040.9 767

0.0410 (0.1726) 7.7650 (5.6836) -0.4577 (3.0844) 7.1638*** (2.3269) -0.0408 (0.2588) -10.8881*** (0.2712) 0.0714** (0.0360) 0.5360*** (0.1103) 0.4260*** (0.0651) 157.7(6) 0.0000 2697.2 767

Notes: Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1 dependent variables: spot rates, model (5): NEER (nominal effective exchange rate, cra event window= [-1,+3], eu event window= [-1,+3].

26 Page 27 of 46

Table 9: Estimation Results Model I for Euro Exchange Rates (2010–2012), Event Watch

eu event constant Variance equation cra event watch ∆(repo) ∆ ln(vixeu )

-0.0003 (0.0006) -0.0933*** (0.02769) 0.0301* (0.0157) -0.0502*** (0.0054) 0.0001 (0.0051) 0.0001 (0.0007) 0.0001 (0.0003)

0.4808** (0.2100) 7.8858 (6.1478) -1.5576 (2.5024) 5.5729** (2.3146) 0.0835 (0.2607) -12.5412*** (0.3821) 0.0636 (0.0449) 0.5451*** (0.1591) 0.4819*** (0.0775) 105.7(6) 0.0000 3306.8 767

0.4230** (0.2089) 5.4468 (7.2194) -1.7810 (2.6635) 5.3864** (2.3973) 0.0288 (0.2530) -11.4162*** (0.4597) 0.0214 (0.0390) 0.6162*** (0.1698) 0.5223*** (0.0767) 154.6(6) 0.0000 2852.0 767

1.9579* (1.0109) 10.2133 (37.2242) 5.1310 (5.0130) 7.0345 (4.7509) 0.4585 (0.8840) -15.9547*** (0.9355) 0.0254* (0.0145) 0.9527*** (0.0194) 0.5596*** (0.0744) 37.7(6) 0.0000 3043.6 767

-0.0372 (0.2284) 7.7681 (5.8201) -0.4719 (3.1517) 7.1411*** (2.3726) -0.0147 (0.2704) -10.8846*** (0.2762) 0.0739** (0.0363) 0.5398*** (0.1113) 0.4201*** (0.0652) 155.3(6) 0.0000 2697.2 767

eu event

constant

ARCH(1)

GARCH(1)

GED lnshape

Chi2(df) p-value Log-Likelihood Obs

cr

ip t

-0.0006 (0.0004) -0.0517** (0.0187) 0.0131 (0.0097) -0.0148*** (0.0034) 0.0034 (0.0030) -0.0003 (0.0005) 0.0001 (0.0002)

Ac ce p

∆ ln(vixus )

-0.0002 (0.0006) -0.0996*** (0.0239) 0.0213* (0.0114) -0.0375*** (0.0046) -0.0027 (0.0043) -0.0010* (0.0006) 0.0002 (0.0002)

us

∆ ln(vixus )

-0.0003 (0.0003) -0.0503*** (0.0122) 0.0157** (0.0069) -0.0163*** (0.0025) -0.0002 (0.0024) -0.0003 (0.0003) 0.0001 (0.0001)

an

∆ ln(vixeu )

∆ ln (JPY/e) (4)

M

∆(repo)

∆ ln (GBP/e) (3)

d

∆(interest)

∆ ln (USD/e) (2)

te

Mean equation cra event watch

∆ ln (NEER) (1)

Notes: Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1 dependent variables: spot rates, model (5): NEER (nominal effective exchange rate, cra event window= [-1,+3] and [-1,+2] for GBP, eu event window= [-1,+3].

27 Page 28 of 46

Table 10: Estimation Results Model I for Euro Exchange Rates (2010–2012), Event Outlook

eu event constant Variance equation cra event outlook ∆(repo) ∆ ln(vixeu )

0.0002 (0.0005) -0.0930*** (0.0274) 0.0300* (0.0158) -0.0502*** (0.0054) 0.0001 (0.0051) -0.0001 (0.0007) -0.0001 (0.0003)

0.2221 (0.3581) 13.0517*** (3.8420) 5.5366 (7.5961) 2.3787 (7.1597) 0.5179 (0.4688) -14.2662*** (0.6389) 0.0303 (0.0199) 0.8890*** (0.0518) 0.4666*** (0.0790) 91.9(6) 0.0000 3305.3 767

0.1465 (0.1962) 5.1031 (8.4683) -1.0591 (3.3353) 5.3083* (2.7905) 0.1957 (0.2609) -11.6987*** (0.5886) 0.0279 (0.0356) 0.7003*** (0.1630) 0.5015*** (0.0777) 161.6(6) 0.0000 2850.8 767

eu event

constant

ARCH(1)

GARCH(1)

GED lnshape

Chi2(df) p-value Log-Likelihood Obs

cr

ip t

-0.0002 (0.0003) -0.0564*** (0.0190) 0.0172* (0.0092) -0.0147*** (0.0036) 0.0032 (0.0030) -0.0003 (0.0005) 0.0001 (0.0002)

Ac ce p

∆ ln(vixus )

0.0004 (0.0005) -0.1014*** (0.0236) 0.0210* (0.0109) -0.0369*** (0.0046) -0.0028 (0.0043) -0.0011* (0.0006) 0.0001 (0.0003)

us

∆ ln(vixus )

0.0001 (0.0002) -0.0453*** (0.0127) 0.0158** (0.0070) -0.0161*** (0.0026) 0.0005 (0.0023) -0.0004 (0.0003) -0.0001 (0.0001)

an

∆ ln(vixeu )

∆ ln (JPY/e) (4)

M

∆(repo)

∆ ln (GBP/e) (3)

d

∆(interest)

∆ ln (USD/e) (2)

te

Mean equation cra event outlook

∆ ln (NEER) (1)

0.7628 (0.5271) 3.6302 (42.9235) 11.7875*** (4.2651) 0.5094 (4.4506) 0.2498 (0.5653) -14.1060*** (0.6882) 0.0723*** (0.0249) 0.8590*** (0.0424) 0.5338*** (0.0813) 36.3(6) 0.0000 3040.4 767

-0.0827 (0.1883) 8.0083 (5.7617) -0.5183 (3.1062) 7.1330*** (2.3384) -0.0271 (0.2595) -10.8475*** (0.2722) 0.0732** (0.0354) 0.5344*** (0.1112) 0.4256*** (0.0661) 157.7(6) 0.0000 2697.2 767

Notes: Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1 dependent variables: spot rates, model (5): NEER (nominal effective exchange rate, cra event window= [-1,+3], eu event window= [-1,+3].

28 Page 29 of 46

We find that none of the announcement types has a significant impact on the Euro rate. However, in some cases the volatility of the Euro exchange rate increases in the event window, as shown in the variance equation of the GARCH models

ip t

reported in Tables 8 and 9. Downgrades increase the volatility of the GBP–Euro and the USD–Euro rate, while watch announcements affect the volatility of the

cr

NEER–Euro, the USD–Euro and the GBP–Euro rate positively. Finally, negative outlook announcements have neither an impact on the mean nor on the volatility of

us

exchange rates (see Table 10).14

Regarding our control variables, we find that in all models the interest rate dif-

an

ferential of the US–Euro money market has the expected negative sign: a positive change leads to a depreciation of the Euro. The VSTOXX EuroStoxx 50 index

M

has the expected negative impact on the Euro exchange rate for all three types of announcements. In times of higher uncertainty in the European stock market, the

d

demand for investments in the Eurozone decreases and the Euro depreciates. The results also reveal that higher uncertainty on the European or the US stock market,

te

measured by implied option market volatility but now included in the variance equa-

Ac ce p

tion, typically triggers higher volatility of the Euro rate against major currencies. As expected, positive (negative) changes of the ECB repo rate are associated with an appreciation (depreciation) of the Euro against other currencies for all three announcement types. Tables 11, 12 and 13 display the estimation results for sovereign bond yields based on the CAPM formulation described above as Model II. The event window in the case of government bond yields is defined as the event day and the day after the event for downgrades and negative outlook announcements, while it is defined as one day before and two days after the event for watch announcements.15 14

As further confirmation of the absence of effects of CRA announcements on the Euro exchange rate, we also estimated model I for the Swiss franc–Euro rate, as the Swiss franc in particular might be subject to safe-haven capital inflows in times of high uncertainty. Again, we find no significant effects from CRA events on the Swiss franc–Euro exchange rate. 15 The reason for using a different event window definition for watch announcements is the oc-

29 Page 30 of 46

Table 11 shows effects from downgrades on the yields of the Spanish, Italian and French sovereign bond yields, but no effect on the German bonds.

ip t

Table 11: Estimation Results Model II for Sovereign Bond Yields (2010–2012), Event Downgrade

∆(mktexcessyield) ∆(repo) ∆ ln(vixeu )

eu event

∆ ln(vixus )

∆(repo)

∆ ln(cds)

eu event

constant

ARCH(1) GARCH(1

9.434* (5.244) 26.813** (11.985) -4.537 (7.706) 33.752*** (11.603) -40.46 (24.72) 4.739** (2.083) -21.54*** (8.186) 0.061*** (0.012) 0.937***

te

Ac ce p

∆ ln(vixeu )

d

constant Variance equation cra event down

0.0080** (0.0031) 0.1810*** (0.0260) -0.0195 (0.0651) -0.0074 (0.0227) 0.0049 (0.0215) -0.0050 (0.0038) -0.0015 (0.0010)

0.0031 (0.0023) -0.0554** (0.0219) -0.0646 (0.0530) -0.0176 (0.0198) -0.0560*** (0.0166) 0.0005 (0.0026) -0.0005 (0.0009)

1.925 (1.200) 12.523 (7.641) -2.194 (5.101) -19.494*** (6.204) 15.94*** (5.815) 3.042*** (1.122) -12.38*** (2.092) 0.079*** (0.014) 0.906***

1.776*** (0.383) -16.763*** (3.701) 25.710*** (1.918) 7.348* (3.866) 8.786** (3.528) 0.436 (0.378) -10.33*** (0.388) 0.290*** (0.045) 0.652***

0.510 (0.435) 14.818*** (3.293) 2.011 (3.351) -12.525*** (4.508) 7.971*** (2.967) 0.988** (0.445) -11.83*** (0.499) 0.031** (0.014) 0.928***

M

∆ ln(vixus )

0.0098* (0.0056) 1.4250*** (0.0617) -0.0288 (0.1550) 0.2129*** (0.0484) 0.0246 (0.0403) -0.0032 (0.0072) -0.0010 (0.0022)

us

0.0248*** (0.0086) 1.6505*** (0.0877) 0.1683 (0.2360) 0.1971*** (0.0749) 0.0005 (0.0633) 0.0038 (0.0100) -0.0043 (0.0037)

an

Mean equation cra event down

cr

Varying conditional correlation MGARCH model ∆ESP ∆ITA ∆FRA ∆GER

currence of convergence problems for the [0,1] window.

30 Page 31 of 46

. . . continued Varying conditional correlation MGARCH model ∆ESP ∆ITA ∆FRA ∆GER (0.010) (0.013) (0.039) (0.018) 767 767 767 767 Coef. Std. Err. z 0.6433 (0.0801) 8.03*** -0.0867 (0.1062) -0.82 0.0649 (0.1093) 0.59 -0.1843 (0.1058) -1.74* -0.0107 (0.1057) -0.10 -0.1064 (0.1244) -0.86 905.7(24) 0.0000 4796.7

an

us

cr

ip t

Observations Cross-equation correlations corr(∆ESP,∆ITA) corr(∆ESP,∆GER) corr(∆ESP,∆FRA) corr(∆ITA,∆GER) corr(∆ITA,∆FRA) corr(∆GER,∆FRA) Chi2(df) p-value Log-Likelihood

Ac ce p

te

d

M

Notes: Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1, event windows= [0,+1], distributions: Gaussian, vcc adjustment parameters: λ1 =0.017*** (0.004), λ2 =0.974*** (0.006)

31 Page 32 of 46

Table 12: Estimation Results Model II for Sovereign Bond Yields (2010–2012), Event Watch

∆(repo) ∆ ln(vixeu ) ∆ ln(vixus ) eu event

Ac ce p

∆(repo)

te

∆ ln(vixus )

∆ ln(cds)

eu event

constant

ARCH(1)

GARCH(1) Observations

-4.863 (3.181) 27.41*** (6.972) -20.27*** (5.325) 33.56*** (7.737) -17.67** (7.473) 3.739*** (1.289) -11.31*** (1.492) 0.065*** (0.012) 0.927*** (0.012) 767

d

Variance equation cra event watch ∆ ln(vixeu )

-0.435 (0.539) 12.67** (5.391) -3.744 (4.615) -20.16*** (5.057) 12.60*** (4.824) 2.511*** (0.595) -10.67*** (0.584) 0.081*** (0.015) 0.898***a (0.014) 767

-2.712** (1.356) -9.31** (4.081) 27.40*** (2.428) -5.452 (8.260) -3.869 (3.378) 0.878 (0.713) -10.12*** (0.403) 0.330*** (0.047) 0.661*** (0.041) 767

M

constant

-0.0041* (0.0022) 0.1850*** (0.0259) -0.0324 (0.0439) -0.0163 (0.0214) 0.0042 (0.0198) -0.0064** (0.0031) -0.0003 (0.0010)

-0.0035 (0.0023) -0.0555** (0.0223) -0.0682 (0.0536) -0.0209 (0.0200) -0.0604*** (0.0168) 0.0012 (0.0026) 0.0003 (0.0009)

cr

∆(mktexcessyield)

-0.0094* (0.0055) 1.4245*** (0.0615) -0.0136 (0.1565) 0.2174*** (0.0467) 0.0137 (0.0396) -0.0003 (0.0074) 0.0019 (0.0022)

us

-0.0041 (0.0082) 1.6585*** (0.0918) 0.1870 (0.2302) 0.1944*** (0.0744) -0.0067 (0.0647) 0.0040 (0.0101) 0.0005 (0.0036)

an

Mean equation cra event watch

ip t

Varying conditional correlation MGARCH model ∆ESP ∆ITA ∆FRA ∆GER

-0.429 (0.411) 14.72*** (3.019) 2.744 (3.220) -13.54*** (3.601) 7.063** (2.832) 1.028** (0.427) -11.54*** (0.409) 0.029** (0.014) 0.926*** (0.017) 767

32 Page 33 of 46

. . . continued Varying conditional correlation MGARCH model ∆ESP ∆ITA ∆FRA ∆GER Coef. Std. Err. z 0.6359 (0.0770) 8.26*** -0.0747 (0.1004) -0.74 0.0177 (0.1079) 0.17 -0.1643 (0.0994) -1.65 -0.0322 (0.1009) -0.32 -0.0755 (0.1224) -0.62 893.5(24) 0.0000 4788.5

us

cr

ip t

Cross-equation correlations corr(∆ESP,∆ITA) corr(∆ESP,∆GER) corr(∆ESP,∆FRA) corr(∆ITA,∆GER) corr(∆ITA,∆FRA) corr(∆GER,∆FRA) Chi2(df) p-value Log-Likelihood

Ac ce p

te

d

M

an

Notes: Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1, event windows= [-1,+3], distributions: Gaussian, vcc adjustment parameters: λ1 =0.016*** (0.004), a GARCH(2) instead of GARCH(1) included. λ2 =0.976*** (0.006)

33 Page 34 of 46

Table 13: Estimation Results Model II for Sovereign Bond Yields (2010–2012), Event Outlook

∆(repo) ∆ ln(vixeu ) ∆ ln(vixus ) eu event

Ac ce p

∆(repo)

te

∆ ln(vixus )

∆ ln(cds)

eu event

constant

ARCH(1)

GARCH(1) Observations

2.102** (0.930) 23.94*** (8.167) -17.59** (8.267) 29.92*** (6.793) -8.250** (3.240) 3.256*** (1.045) -12.05*** (1.506) 0.062*** (0.012) 0.931*** (0.011) 767

d

Variance equation cra event outlook ∆ ln(vixeu )

2.143*** (0.699) 14.70* (7.740) -5.340 (6.389) -21.70*** (7.409) 14.87*** (4.672) 3.249*** (0.745) -12.28*** (1.225) 0.075*** (0.014) 0.908*** (0.012) 767

1.938*** (0.380) -17.02*** (3.778) 25.49*** (1.898) 7.941** (3.769) 10.18*** (3.284) 0.396 (0.409) -10.40*** (0.405) 0.276*** (0.046) 0.663*** (0.041) 767

M

constant

0.0085** (0.0033) 0.1796*** (0.0256) -0.0225 (0.0670) -0.0024 (0.0224) -0.0001 (0.0211) -0.0036 (0.0037) -0.0016 (0.0010)

0.0053** (0.0024) -0.0569** (0.0221) -0.0649 (0.0527) -0.0177 (0.0197) -0.0557*** (0.0165) 0.0007 (0.0026) -0.0008 (0.0009)

cr

∆(mktexcessyield)

0.0127** (0.0059) 1.4307*** (0.0613) -0.0256 (0.1458) 0.2150*** (0.0462) 0.0189 (0.0396) -0.0023 (0.0073) -0.0013 (0.0022)

us

0.0276*** (0.0092) 1.6831*** (0.0885) 0.1878 (0.2313) 0.1998*** (0.0742) -0.0187 (0.0626) 0.0059 (0.0100) -0.0042 (0.0037)

an

Mean equation cra event outlook

ip t

Varying conditional correlation MGARCH model ∆ESP ∆ITA ∆FRA ∆GER

0.779* (0.421) 15.10*** (3.298) 1.615 (3.310) -12.94*** (4.473) 8.200*** (2.989) 1.050** (0.437) -11.93*** (0.505) 0.033** (0.014) 0.927*** (0.018) 767

34 Page 35 of 46

. . . continued Varying conditional correlation MGARCH model ∆ESP ∆ITA ∆FRA ∆GER Coef. Std. Err. z 0.6609 (0.0896) 7.38*** -0.0953 (0.1150) -0.83 0.0925 (0.1202) 0.77 -0.1900 (0.1143) -1.66* 0.0101 (0.1145) 0.09 -0.1408 (0.1414) -1.00 919.5(24) 0.0000 4803.4

us

cr

ip t

Cross-equation correlations corr(∆ESP,∆ITA) corr(∆ESP,∆GER) corr(∆ESP,∆FRA) corr(∆ITA,∆GER) corr(∆ITA,∆FRA) corr(∆GER,∆FRA) Chi2(df) p-value Log-Likelihood

an

Notes: Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1, event windows= [0,+1], distributions: Gaussian, vcc adjustment parameters: λ1 =0.017*** (0.004), λ2 =0.976*** (0.006)

M

The βbond coefficients on ∆(mktexcessyield) are significant and have the expected magnitude. Spain and Italy have βbond coefficients above unity for all three

d

announcement types and thus exhibit a heightened sensitivity to the market risk of Eurozone bonds. On the other hand, France and Germany have much lower βbond

te

values, with the German coefficient even being negative for downgrade and watch

Ac ce p

events, implying that German expected excess yields fall if the market excess yield increases. For Spain, Italy and France, CRA announcements have a unambiguously positive impact on government bond yields, which means that during the event window, the prices for these bonds fell. With regard to the impact of downgrades on the risk in the bond markets,

shown in the variance equation of the GARCH model in Table 11, we find that the volatility for Spanish and French bonds increases but not for the other two issuers. Uncertainty in the equity markets is related to bond yield volatility for all three announcement types, with the VIX, VSTOXX or both having meaningful effects, often differing in sign. Changes in CDS spreads also have important (and generally positive) effects on volatility.

35 Page 36 of 46

Table 12 displays the effects from watch announcements on bond yields,16 showing that, overall, the evidence of an impact from watch announcements is weak. Finally, Table 13 presents the effects from outlook events on the government bond

ip t

yields. These announcements have the strongest effects, both on the mean of bond yields as well as on their volatility. Though very small in magnitude, even the

cr

German bond yield increase is statistically significant, at around 0.005 percentage points, when outlook announcements occurred.

us

In the lower panel of Tables 11–13, correlations between the error terms across equations are displayed. A strong and significant correlation is found between Italian

an

and Spanish bond yields, and a negative significant correlation between Italian and German bond yields. As these correlations accounts for unobserved factors, or shocks

M

that affect both markets, we take this as evidence of spillover effects in bond markets. The spillovers work in both directions, as bond markets respond either in the same

d

way (ESP and ITA) to rating announcements or in opposite ways (GER and ITA). Figure 7 shows the evolution of the correlations over time. In particular at the end

te

of 2011 strong correlations exist, but of opposite signs. However, towards the end

Ac ce p

of 2012 we can observe a type of convergence to a much lower level of correlation across the issuers’ errors. Note also that the positive correlation between German and French bonds reverts to a negative correlation at the end of 2011. Table 14: Pairwise correlations between CRA announcements and EU actions 2010–2012 (T=767)

down watch outlook eu

down 1 0.3181 0.7878 0.0202

watch

outlook

eu

1 0.1478 0.1058

1 -0.0254

1

16

We also tested models where all event types are considered simultaneously but because of the relatively high correlation of events, as shown in Table 14, these models typically do not estimate significant effects from CRA announcements.

36 Page 37 of 46

ip t

te

−.4

Ac ce p

−.3

−.2

−.1

d

0

.1

M

.2

.3

.4

an

.5

.6

us

.7

cr

Figure 7: Cross-equation Correlations from Table 13, CRAs’ Negative Outlook Announcements, 2009–2012

01/01/2009

01/01/2010

01/01/2011 Date

01/01/2012

Event outlook

Corr(ESP,ITA)

Corr(ESP,GER)

Corr(ESP,GER)

Corr(ITA,GER)

Corr(ITA,FRA)

01/01/2013

Corr(FRA,GER)

Sources : Datastream, Events: S&P, Fitch or Moody’s Negative Outlook of Eurozone States’ Sovereign Debt

37 Page 38 of 46

ip t

02/01/2009

02/09/2009

03/05/2010

.002 .0005 .001 .0015 lpoly smooth: Variance of MV Bond Portfolio 0

0

Ac ce p

.1

te

.2

d

.3

M

.4

.5

an

.6

us

.7

cr

Figure 8: Variance of the Minimum Variance Bond Portfolio (MVP) and the Weights of ESP, ITA, FRA and GER Governmental Bonds in MVP, based on Covariances from Table 13, 2009–2012, Kernel-weighted Local Polynomial Smoothing

01/01/2011 Date

01/09/2011

01/05/2012

Event outlook

lpoly smooth: portfolio weight ESP

lpoly smooth: portfolio weight ITA

lpoly smooth: portfolio weight FRA

lpoly smooth: portfolio weight GER

lpoly smooth: Variance of MV Bond Portfolio

30/12/2012

Sources : Datastream, Events: S&P, Fitch or Moody’s Negative Outlook of Eurozone States’ Sovereign Debt

38 Page 39 of 46

Based on the covariances from the model reported in Table 13, we computed the variance of the minimum variance portfolio (MVP) containing the four bonds as well as the corresponding weights of the different bonds in this MV portfolio. Figure 8

ip t

shows both Var(MVP) as well as the evolution of the weights17 of the various bonds in this portfolio.18 One can infer from this graph that the share of German bonds in

cr

the MVP in the second half of year 2009 is around 40 percent, while that of Italian bonds is about 20 percent and that of Spanish bonds below 5 percent. However, in

us

2012 the share of German bonds in the MVP increases to slightly above 60 percent, whereas the share of the Italian bonds decline to 5 percent and the Spanish bonds

an

have a weight close to zero in the MVP during the second half of 2012. One can also see from Figure 8 that the variance of the MVP peaked during fall 2011 when several

M

significant CRA announcements were made, but decreased sharply afterwards. Overall, these results are consistent with Beck et al. (2015) and also confirm

d

anecdotal evidence that CRA announcements triggered considerable portfolio rebalancing across member countries, from troubled states’ sovereign debt into more

te

stable sovereign issues. The evidence also indicates that during the depth of the debt

Ac ce p

crisis, the existence of the common currency served as a kind of insurance mechanism against a severe currency crisis. However, the protection against a severe currency crisis comes at a price. While CRA announcements have little influence on the Euro exchange rate, they tend to accentuate fluctuations in the price of sovereign bonds. Finally, as a robustness check, we experimented with different definitions of the

event window. It turns out that the insignificance of CRA events for the Euro exchange rates holds irrespective of the length of the event window. In the case of sovereign bond yields, we find stronger effects from CRA announcements for shorter 17 18

Kernel-weighted local polynomial smoothing has been used for the displayed time series. Note that the portfolio weights of bonds sum to one, and short-selling is not possible.

39 Page 40 of 46

event windows. Two days after the event, the impact on bond yields is negligible.19 5. Conclusions

ip t

This paper evaluates the effects of CRA announcements on the value of the Euro and the sovereign bond yields of major Eurozone countries during the most

cr

stressful period of the financial crisis. We combine event study methodology with multivariate GARCH models to evaluate the announcements’ effect on both the level

us

and the volatility of our two main analysis variables. For sovereign bond yields, we formulate a CAPM–based model that links bond excess yields to market excess

an

yields. We find no significant effects of various event types on the value of the Euro over the period 2010–2012, but the volatility of the Euro exchange rate increases in

M

some cases. In contrast, we find stronger impacts from CRA events on sovereign bond yields. Specifically, our estimates reveal effects from downgrades on both the yields and volatilities of Spanish, Italian and French bonds, but the strongest effect

d

is found for negative outlook announcements that also increased the yields of the

te

German bonds. Furthermore, the estimates show significant positive and negative

Ac ce p

correlations of residuals across equations, showing that shocks affected bonds in the same (Spanish and Italian bonds) or in the opposite direction (German and Italian bonds).

In summary, while CRA announcements had very little or no impact on the value

of the Euro, events significantly affected sovereign bond yields. Thus, our findings are consistent with the observation that investors rebalanced their sovereign bond portfolios within the Euro member countries to reduce their exposure to riskier borrowers. These results have important policy implications. The single currency has proven to be an effective stabilizer against massive downward pressure triggered and fueled 19

Results for different event windows are available from the authors upon request.

40 Page 41 of 46

by repeated downgrading from rating agencies. CRA announcements did not accelerate the movement of foreign funds out of the Eurozone. Thus, the existence of the single currency deterred foreign investors to sharply reduce investments in the Euro-

ip t

zone, and prevented a sudden and strong deteriorations of ailing countries’ terms of trade. The presence of stronger states within the Eurozone offered investors an effi-

cr

cient way to insulate their portfolios from the allegedly higher sovereign risk signaled by rating agencies’ announcements. They could simply reallocate their funds from

us

weaker to stronger states without changing the currency. However, the dampening of the adverse impact of downgrades on the ailing countries’ terms of trade comes

an

at a clear price. The reallocation of investors’ funds has distributional effects, as it dramatically reduced Germany’s cost of debt service (indicated by a negative Ger-

M

man bond beta) while downgraded countries have had to offer significantly higher interest rates on their sovereign issues after downgrade events, and therefore bear

d

considerably higher costs when issuing new debt. In addition, export-led growth fueled by devaluation of the own domestic currency was absent. In light of our find-

te

ings, the continuing measures taken to lower financial tensions and to preserve the

Ac ce p

Eurozone, as well as to support ailing member countries, seem highly appropriate. 6. Acknowledgements

We thank seminar participants at the National Bank of Poland, and in particular

Michał Konopczak, Dobromił Serwa and Zbigniew Polański for helpful comments and suggestions. We are also grateful to seminar participants at Tufts University, in particular Michael Klein and Marcelo Bianconi. The paper was presented at the International Conference on Economic Modeling (EcoMod2014) in Bali, Indonesia and the School of Economics and Finance, Queensland University of Technology. Two anonymous referees have also provided crucial feedback for which we are very thankful. The usual disclaimer applies.

41 Page 42 of 46

References

ip t

Afonso, A., Furceri, D. & Gomes, P. (2012), ‘Sovereign credit ratings and financial markets linkages: Application to european data’, Journal of International Money and Finance 31(3), 606–638. URL: http://ideas.repec.org/a/eee/jimfin/v31y2012i3p606-638.html

us

cr

Alsakka, R. & ap Gwilym, O. (2012), ‘Foreign exchange market reactions to sovereign credit news’, Journal of International Money and Finance 31(4), 845– 864. URL: http://ideas.repec.org/a/eee/jimfin/v31y2012i4p845-864.html

an

Alsakka, R. & ap Gwilym, O. (2013), ‘Rating agencies’ signals during the european sovereign debt crisis: Market impact and spillovers’, Journal of Economic Behavior & Organization 85(C), 144–162. URL: http://ideas.repec.org/a/eee/jeborg/v85y2013icp144-162.html

M

Arezki, R., Candelon, B. & Sy, A. (2011), Sovereign rating news and financial markets spillovers: Evidence from the European debt crisis, IMF Working Paper 11/68, International Monetary Fund. URL: http://www.imf.org/external/pubs/ft/wp/2011/wp1168.pdf

te

d

Beck, R., Georgiadis, G. & Gräb, J. (2015), The geography of the great rebalancing in euro area bond markets during the sovereign debt crisis, Working Paper Series 1839, European Central Bank. URL: https://ideas.repec.org/p/ecb/ecbwps/20151839.html

Ac ce p

Blommestein, H., Eijffinger, S. & Qian, Z. (2016), ‘Regime-dependent determinants of Euro area sovereign CDS spreads’, Journal of Financial Stability 22, 10–21. Bollerslev, T. (1986), ‘Generalized autoregressive conditional heteroskedasticity’, Journal of econometrics 31, 307–327. URL: http://www.sciencedirect.com/science/article/pii/0304407686900631 Bollerslev, T., Engle, R. F. & Nelson, D. B. (1994), Chapter 49: Arch models, in R. F. Engle & D. L. McFadden, eds, ‘Handbook of Econometrics’, Vol. 4 of Handbook of Econometrics, Elsevier, pp. 2959 – 3038. URL: http://www.sciencedirect.com/science/article/pii/S1573441205800182 Böninghausen, B. & Zabel, M. (2015), ‘Credit ratings and cross-border bond market spillovers’, Journal of International Money and Finance 53, 115–136. Brooks, R., Faff, R. W., Hillier, D. & Hillier, J. (2004), ‘The national market impact of sovereign rating changes’, Journal of Banking and Finance 28(1), 233–250. URL: http://linkinghub.elsevier.com/retrieve/pii/S0378426602004065

42 Page 43 of 46

Cantor, R. & Packer, F. (1996), ‘Determinants and Impact of Sovereign Credit Ratings’, The Journal of Fixed Income 6(3), 76–91. URL: http://www.iijournals.com/doi/abs/10.3905/jfi.1996.408185

cr

ip t

Chen, S.-S., Chen, H.-Y., Chang, C.-C. & Yang, S.-L. (2013), ‘How do sovereign credit rating changes affect private investment?’, Journal of Banking & Finance 37(12), 4820–4833. URL: http://EconPapers.repec.org/RePEc:eee:jbfina:v:37:y:2013:i:12:p:48204833

us

Ehrmann, M., Osbat, C., Stráský, J. & Uusküla, L. (2013), The euro exchange rate during the european sovereign debt crisis - dancing to its own tune?, Working Paper Series 1532, European Central Bank.

an

Elliott, G., Rothenberg, T. J. & Stock, J. H. (1996), ‘Efficient tests for an autoregressive unit root’, Econometrica 64(4), 813–836.

M

Engle, R. (2002), ‘Dynamic conditional correlation: A simple class of multivariate generalized autoregressive conditional heteroskedasticity models’, Journal of Business & Economic Statistics 20(3), 339–350.

d

Ferri, L.-G. & Stiglitz, J. E. (1999), ‘The Procyclical Role of Rating Agencies: Evidence from the East Asian Crisis’, Economic Notes 28(3), 335–355. URL: http://doi.wiley.com/10.1111/1468-0300.00016

Ac ce p

te

Fratzscher, M. (2011), Capital flows, push versus pull factors and the global financial crisis, Working Paper Series 1364, European Central Bank. URL: http://ideas.repec.org/p/ecb/ecbwps/20111364.html Gande, A. & Parsley, D. C. (2005), ‘News spillovers in the sovereign debt market’, Journal of Financial Economics 75(3), 691–734. URL: http://linkinghub.elsevier.com/retrieve/pii/S0304405X04001576 Hooper, V., Hume, T. & Kim, S.-J. (2008), ‘Sovereign rating changes – do they provide new information for stock markets?’, Economic Systems 32(2), 142–166. URL: http://linkinghub.elsevier.com/retrieve/pii/S0939362508000034 Ismailescu, I. & Kazemi, H. (2010), ‘The reaction of emerging market credit default swap spreads to sovereign credit rating changes’, Journal of Banking & Finance 34(12), 2861–2873. URL: http://linkinghub.elsevier.com/retrieve/pii/S0378426610001913 Judge, G. G., Griffiths, W. E., Hill, R. C., Lütkepohl, H. & Lee, T.-C. (1985), The Theory and Practice of Econometrics, 2nd edn, Wiley.

43 Page 44 of 46

Kaminsky, G. & Schmukler, S. L. (2002), ‘Emerging Market Instability: Do Sovereign Ratings Affect Country Risk and Stock Returns?’, The World Bank Economic Review 16(2), 171–195. URL: http://wber.oupjournals.org/cgi/doi/10.1093/wber/16.2.171

cr

ip t

Klose, J. & Weigert, B. (2014), ‘Sovereign Yield Spreads During the Euro Crisis: Fundamental Factors Versus Redenomination Risk’, International Finance 17(1), 25–50.

us

Kräussl, R. (2005), ‘Do credit rating agencies add to the dynamics of emerging market crises?’, Journal of Financial Stability 1(3), 355–385. URL: http://linkinghub.elsevier.com/retrieve/pii/S1572308905000069

an

Missio, S. & Watzka, S. (2011), Financial Contagion and the European Debt Crisis, CESifo Working Paper Series 3554, CESifo Group Munich. URL: http://ideas.repec.org/p/ces/ceswps/_3554.html

M

Nelson, D. B. (1991), ‘Conditional Heteroskedasticity in Asset Returns: A New Approach’, Econometrica 59(2), 347–370. URL: http://www.jstor.org/stable/2938260

d

Reinhart, C. M. (2002), ‘Default, currency crises, and sovereign credit ratings’, World Bank Economic Review 16(2), 151–170. URL: http://ideas.repec.org/a/oup/wbecrv/v16y2002i2p151-170.html

Ac ce p

te

Schmitz, M., De Clercq, M., Fidora, M., Lauro, B. & Pinheiro, C. (2013), ‘Revisiting the effective exchange rates of the euro’, Journal of economic and social measurement 38(2), 127–158. Tse, Y. & Tsui, A. (2002), ‘A multivariate generalized autoregressive conditional heteroscedasticity model with time-varying correlations’, Journal of Business & Economic Statistics 20(3), 351–362.

44 Page 45 of 46

*Highlights (for review)

Credit Rating Agency Downgrades and the Eurozone Sovereign Debt Crises Highlights

Ac ce p

te

d

M

an

us

cr

ip t

• We examine the effects of credit rating agency announcements in the Eurozone. • We consider their effects on the value of the Euro exchange rate. • We consider their effects on the yields of major borrowers’ Eurozone sovereign bonds. • The Euro’s value was not substantially affected, but its volatility increased. • Sovereign bond yields reacted to announcements depending on the issuer’s credit rating.

Page 46 of 46