EUROZONE SOVEREIGN BONDS AND RATING ASSESSMENTS : IMPACT ON VOLATILITY

Rating agencies have been very active during the economic crisis and have been blamed for damaging the refinancing possibilities of the eurozone countries. Their decisions concerning sovereign bonds have been widely pointed out as one of the reasons why spreads rose dramatically between 2009 and 2012. Nonetheless, last evolutions of the sovereign spreads in countries such as Spain, Ireland or France show that sovereigns do not respond to rating assessments as extremely as they did before. Therefore, economic actors may wonder whether there has been a recent change in the trend or by contrast those assessments did not influence the volatility of the spreads, which may have been motivated by other variables. In this paper we will intend to determine to what extent S&P announcements were drivers of higher volatility of sovereign bonds’ spreads and how these effects (if any) have evolved over the economic crisis.


INTRODUCTION
The sustainability of public finances has been widely discussed in the framework of the eurozone crisis.Since the intensification of the crisis in September 2008, 10years sovereign bonds yields increased rapidly, affecting mainly those countries whose economic fundamentals were more deteriorated.This impact is clearly observed on the widening of these countries' spreads over German Bunds.Along with the increase of the spreads, we observe a sharp rise of the volatility experienced by eurozone sovereigns, questioning their consideration as safe investment.
In this context, rating agencies played a significant role.They have been very active issuing credit risk announcements, especially in the period 2009-2013.Their announcements were accompanied by drastic jumps of sovereign yield spreads (see There are several studies that have measured the impact of sovereign ratings on sovereign bonds; Cantor and Packer (1996), Larraín et al (1997), Reisen andvon Maltzan (1998 &1999), Ismailescu and Kazemi (2010) and Raimbourg and Ory (2011).In the first three cases they used event study methodologies for a sample in which they mainly included emerging countries.They found evidence that rating opinions affect spreads and market volatility in certain cases.Ismailescu and Kazemi measured the impact of rating news on CDS spreads and they concluded that ratings seem to provide new information to the sovereign CDS market.Raisen and von Maltzan (1998) measured the impact of rating events on financial market volatility.They split volatility into two; bond yield spreads volatility and stock market volatility.Using event study methodologies they found that downgrading, reviews for negative outlooks and upgrading had statistically significant impact on market volatility.
Our paper brings "fresh air" to the literature related to the topic.Although some of the above mentioned papers study the impact of rating assessments on sovereign bonds yields in the context of the eurozone crisis, none of them analyze the issue from the perspective of spreads volatility.Our first objective is to conduct an event study in which we analyze the behavior of sovereign bond volatility, first within an event window of 30 days and later within the 4 days surrounding the rating events in order to determine whether abnormal volatility is observed.This choice has been motivated by the methodology used by Reisen andvon Maltzan (1998), Hand et al (1992) and Bouraoui (2009).Our second objective is to observe how the influence of rating announcements on spreads volatility evolves over the eurozone crisis.
The paper is structured in four sections; second section will be devoted to a literature review as well as a description of the trends followed by sovereign bonds volatility over the period of analysis.In the third section we will perform an event study that will allow us to test both graphically and analytically whether rating announcements trigger excessive volatility in the market.We will distinguish between both types of announcements; actual rating changes and watch/outlook changes implemented by the agency Standard and Poor´s over the period covered by our study.To finalize we will draw some conclusions and policy recommendations.

Related literature: impact of rating announcements on spreads
The effects of sovereign rating news on bond risk premium have been widely studied by the economic literature.Cantor and Packer (1996) analysed the macroeconomic determinants of sovereign ratings.They found that per capita income, GDP growth, inflation, fiscal balance, external debt level and default history are the main factors behind the assessments.Additionally, they used an event study to analyse the impact of rating news on sovereign spreads finding evidence that rating news affect independently market spreads, especially in the case of non-investment grade ratings; spreads increase with negative news and viceversa.
Reisen and von Maltzan (1998) associated rating events with changes in sovereign spread and financial market volatility.First, they analysed the granger-causality of rating news and changes in sovereign spreads.Then, using the same event study methodology as Cantor and Packer, they found significant impact of rating decisions on sovereign yields which was interpreted as a signal for the market about the credit worthiness of the country.In addition, they tested this impact on the volatility of sovereign bond spreads3 and stock markets.In both cases they found significant impact of negative news on volatility and no impact of positive news.
A similar working paper was published by the OECD in which Reisen and von Maltzan (1999) mainly focused on the link between rating agencies decisions and emerging countries bond spreads.They concluded that Credit Rating Agencies' opinions tend to foster boom-bust cycles in sovereign bond markets.Larraín et al (1997) performed an event study to examine the market's assessment of the rating announcements by Moody´s and S&P in the context of the "tequila crisis".They concluded that credit rating opinions tended to lag the market instead of leading it because their assessment evidenced that rating events were anticipated by the market, especially regarding rating outlook reviews.
More recently Ismailescu and Kazemi (2010) tested through an event study methodology the reaction of CDS spreads to Credit Rating Agencies' news.They determined that rating announcements seem to provide new information to the market.Comparing to the previous literature they found that positive news have a more significant impact on sovereign CDS spreads than negative news.They based their argument on the fact that negative events have a higher probability of being predicted by the CDS market, therefore their impact is lower.In the same line, Kiff et al (2012) developed an event study to test whether rating agencies provide new information to the sovereign debt market and whether certification services matter.They determined that there is a change in sovereign CDS spreads in the expected direction that led to a significant widening of the risk premia by about 37 basis points on the event day and immediately after the event.
Afonso, Furceri and Gomes (2011) tested the impact of rating announcements on CDS spreads and yield spreads for 24 European Union countries.They concluded that EMU4 and non-EMU CDS spreads responded similarly to rating news.Positive rating news triggered weak and non significant changes in CDS spread, whereas negative news increase spreads considerably.Their event study technique only found statistically significant market movements on eurozone countries upon negative S&P announcements.
Raimbourg and Ory (2011) developed an empirical investigation of the effect of downgrades and negative outlooks on European sovereign bond spreads.They used an alternative methodology to measure the significant impact, Perron structural break test.They came to the conclusion that the opinions issued by the three leading rating agencies are not "harmful" for the market.
Gande and Pasley (2005) tested the impact of rating news on sovereign bond yield spreads from a different point of view.They checked the "announcement spillovers" that rating agencies news may trigger over the period 1991 -2000.They include in their sample various worldwide countries that issued public debt denominated in dollars.They found significant impact of rating events on both affected country spreads and other countries' spreads.One notch downgrade was associated with an average increase of 12 basis points of the risk premia of non-downgraded countries.
Regarding news spillovers as well, Arezki et al ( 2011) analysed this phenomenon in the context of the European sovereign debt crisis.They found contagion across different countries.CDS spreads from one country may respond to rating events from another State.This addressed important regulatory implications about the ability of rating agencies to boost financial instability.
Metiu (2011a:7) investigated a question similar to the one of the two previous cited documents in the context of the EMU crisis.He showed that "sovereign debt crisis in small open economies may become systemically important due to the high degree of integration between government bond markets".He found a significant impact of rating news on sovereign yield premia volatility.He captures the asymmetric causality effect of rating news on volatility through an ARMA -GARCH model, given by the coefficient and referred to an estimation of the time varying volatility.
He determined that an unexpected decrease of spreads raises the volatility of the sovereign premia less than an expected increase of spreads.
More recently, a working paper published by the ECB analyse the impact of sovereign bonds assessments performed by rating agencies in the context of the eurozone sovereign crisis.De Santis (2012) deals with the impact of rating events on the eurozone countries that presented more deteriorated public finances (so-called PIIGS5 ) in the period 2008-2011.He found that country-specific credit rating is one of the factors that can explain the developments in sovereign spreads mainly for those countries presenting more deteriorated economic fundamentals, although they are also influenced by the existence of spillovers from other eurozone countries (i.e.Greece).He also finds evidence that spreads for Austria, Finland and the Netherlands depend on the higher demand on German bonds during the crisis (flight to quality) and not on rating events or lack of liquidity.

Rating announcements as drivers of volatility
In this section we analyse the evolution of sovereign spreads volatility and the factors that may link historical volatility with credit rating announcements.Source: Compiled data from Financial Times.
We observe in figure 2 that up to mid 2007, volatility of eurozone sovereign bonds was fairly low.Such a low levels of volatility and the fact that spreads were just few basis points above German bund made all eurozone sovereigns be perceived as roughly equivalent.However, spreads started widening in mid-2008 and the situation worsened as of the beginning of the financial crisis.
Sovereign crisis distorted the perception of investors about credit risk, which was reinforced by the increasing trend followed by volatility.In circumstances of low volatility, higher yields may compensate for higher risk, and bonds of high-spread countries could still be attractive 7 .But in such a context of high volatility, these greater yields are not sufficient to compensate for more risk (IMF, 2011).Since the beginning of 2010, volatility of high spread eurozone sovereign bonds was well above the level of those countries with low spreads (see figure 3) what had an obvious negative impact on their attractiveness; increasing volatility implies that daily movements in bond yields (or prices) are less predictable what means greater risk from holding these bonds8 (Sosvilla and Morales, 2011).
From a macroeconomic point of view, the increase of uncertainty has a negative impact on the economy.First, government debt represents a relevant percentage among the financial assets and higher volatility of sovereign spreads can translate into more uncertainty in general expectations.In addition, it also undermines the value of the securities issued by local banks and companies due to the fact treasury bonds have commonly been considered as a benchmark risk free asset (Masciandaro, 2011).Moreover, sovereign bonds are held by banks, companies and private investors, mainly national but international as well, so higher bond spread volatility has important systemic consequences.As credit ratings are a measure of the credit capacity (or willingness to pay), we can affirm that volatility of yield spreads is potentially influenced by rating announcements (Metiu, 2011).A deterioration of the sovereign credit quality causes a reduction of the number of potential buyers for these bonds, triggering spread movements and difficulties to roll-over debt issuances (Whelan, 2011).Financial literature has shown that credit ratings appear to influence sovereign yields in addition to their correlation with other publicly available information (Cantor and Packer, 1996:349 ; Reisen von Maltzan, 1999 andMetiu, 2011).Most of the announcements made by the Credit Rating Agencies have been concentrated in the period 2009-2013.Of course, rating news have not been the only driver of volatility since the deterioration of public finances and other news have also played an important role.As it was empirically established by Borio and McCuley (1996:75) fiscal disequilibria enhance yield volatility, but it is not the only factor; news related to sovereign bonds also explain variations.
Certain elements attached to rating assessments may contribute to raise volatility upon their announcements.The main channels that we have found in the literature are: a) Attachment of ratings to regulation Ratings are embedded in regulatory capital and thresholds.Under prudential regulation rules, banks are allowed to hold less capital or own reserves against highly rated securities within their balance sheet10 having to increase its capital in case of higher credit risk (lower rating) is assigned to its holdings.
For instance, Basel II rules allowed for two approaches to measure credit risk; the standard one that relies on rating assessments by the rating agencies and the internal rating based system (IRB).Although banks must have their own methodologies to assess credit risk from counterparties or securities, there is a high degree of reliance on external assessments (BIS, 2010:52) (Sy, 2009:10).
Other example of attachment of ratings to regulation is the minimum collateral requirement of the ECB.The ECB determines through a rating grade scale which assets can be used as collateral for money market operations11 (ECB, 2012).
Sometimes, ratings are linked to financial contracts.The application of rating-based models is sometimes due to the fact that valuing securities may be tied to ratings such as swap protections against downgrading (Duffie and Kenneth, 2003:139).This may lead to further collateral calls or force selling by fund managers.In other cases the investment mandate of the institution (life insurer, pension funds, etc.) prevents managers from investing in securities that are under the eligibility criteria that is often set through a rating issued by one of the leading Credit Rating Agencies (Deb et al, 2011).This "hardwiring" of ratings into regulation grants more systemic importance to Credit Rating Agencies creating a sort of "license effect" (Masciandro, 2011a).The relevance of the rating announcement plays an important role and makes sovereign debt yields and prices be sensitive to rating events.In the context of sovereign bonds this fact mainly affects to those that are downgrading towards the speculative grade.Downgrading may cause sell offs because investors will perceive these bonds as distressed.Buy and hold investors will try to rebalance their portfolios in order to reduce their exposure and meet the mentioned requirements; price movements will also be translated into more capital requirements.This so-called "cliff effect"12 after the rating announcement will trigger large price movements and greater yield volatility meaning higher market uncertainty.This higher volatility of returns will make sovereign bonds less attractive and consequently market risk premia will tend to increase (IMF, 2012).This uncertainty led to a "flight for quality" that took place in the eurozone during the economic crisis.Foreign banks started reducing their exposure to sovereign debt stocks from countries that had been heavily downgrade (Italy, Spain, Belgium -figure 4) during 2010and 2011(IMF, 2012)).

b) Communication effect
The communication policy of the rating announcements can be another driver of bond spreads volatility.Although empirical evidence has not been found, the importance of communication methods in the release of opinions by Credit Rating Agencies has been highlighted by some authors and institutional reports13 .
An example of the impact of communication strategies on market behaviour can be found in different events such as central bank governor speeches, monthly unemployment figures or monthly inflation.The same type of analysis could be done for rating agencies news.As stated Masciandro (2011b) these are the main features of the communication of rating events: • First, the system of symbols (AAA, AA, AA-, etc.) is easy to understand by market participants regardless of their level of financial literacy.This fact does not require them to go into further analysis when they receive the announcement and market actions may be immediate.• Rating news are often communicated when markets are open.This may trigger instant reactions in the trading venues translated into massive sell-offs which may lead to higher market volatility.• Despite most of the rating changes are preceded by outlooks or credit watches reviews, rating actions are not always foreseen.Since there is not a fixed schedule of rating announcements, sovereign rating announcements are not always anticipated.Even if they are expected they may trigger greater volatility; especially when they are very severe (i.e.downgrades by more than 1 notch14 ).

Sovereign ratings
Rating notations are assessments of the ability of the sovereign bond issuer to pay back capital and interests.The different notches represent a rating grade, with the Our data related to credit rating assessments come from the rating agency Standard and Poor's (S&P).We have focused on S&P's assessments due to three reasons: First, S&P can be considered as the market leader (table 2 and figure 5) as it is the agency that issues the highest number of sovereign ratings and is often the first credit agency to take action.Secondly, this agency has a wide availability of rating series.Thirdly, it has been widely stated by other authors (Kiff et al, 2012;Reisen andvon Maltzan, 1999 andGande andPasley (2005)) that S&P ratings announcements are less often anticipated by the market.
Data related to sovereign rating assessments and outlook changes were provided by the agency S&P (appendix 2).We focus on those ratings denominated in local currency because debt placements issued by eurozone members are mostly done via competitive auctions of long maturity, fixed coupon and domestic denominated currency (De Broeck and Guscina, 2011).

Data set
We cover the following countries in the analysis: Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, Netherlands, Portugal and Spain.
The reasons why we have chosen these countries are, first the wide availability of data related to these countries from the sources we had access to; secondly, all these countries have been members of the eurozone during the whole period of study; third, S&P has issued assessments linked to each of these countries during the period, at least once (either rating changes or outlook reviews).Moreover all of them are considered developed economies and their bonds were rated as investment grade at the beginning of the period of study which allows us to set a common starting point for our analysis.
The daily data set covers the period between 1 January 2001 and 31 October 2013.Since 2001, there have been 80 rating announcements related to these countries issued by S&P (40 downgrades, 12 upgrades, 2 reviews for possible upgrade and 26 reviews for possible downgrade).
The daily data for sovereign 10 year Government bond yields (end of day) comes from Financial Times database.Thus, we have worked with 3,346 observations for each country which leads to 36,806 observations.Some transformations have been performed in order to make the data workable as explained in the following section.We use 10 years government bonds spreads computed as the yield of each country's bond over the German Bund.

EMPIRICAL ANALYSIS
This section is devoted to the analysis of the impact of rating announcements on historical volatility of sovereign spreads.This analysis will cover a number of aspects such as: 1.Whether rating announcements lead to "abnormal" spread volatilities; To what extent these announcements are anticipated by the market; Whether abnormal volatility is caused either by positive or negative announcements; 2. Whether the impact of rating assessments on historical volatility varies over the crisis.

Event Study
We use event study method in order to test whether rating agencies announcements trigger important shifts in historical volatility of sovereign bond yields.We expect an inverse relation between rating changes and increase of the volatility (see appendix 6).This model is based on the idea that markets react immediately to announcements about the credit quality of the bonds (Mac Kinlay, 1997;pp.13).Serra (2002) identifies the following stages for the implementation of this methodology: • Firstly we define our event and the period during which we study the event.For this purpose we will test the existence of "abnormal volatility" during the days surrounding the event; 30 days before and after the rating announcements.• Secondly, we will base our study in the 30 days historical volatility of relative bond yield spreads.The method, as used by other authors, usually focuses on abnormal excess returns for bonds.Here, we are going to use historical volatility.As bonds may display higher risk premia after negative rating news we are going to test how this is reflected into volatility.Volatility cannot be directly observed, so it has to be calculated.We have measured volatility by using the approach developed by Reisen and Von Maltzan (1998) in which they computed historical volatility of the relative yield spreads by using a 30-days moving average.

3
Where X is the relative yield spread.We have computed this yield spread taking the German Bund as a Benchmark 16 : Following the recommendation by Reisen and von Maltzan (1998:13) we compute the relative spread as above.The reason why we use relative spreads is that they are more stable and fluctuate less with the general level of interest.
Third, we determine our estimation and event window (L2).The period of estimation ( L1) mirrors what it should be considered as "normal" volatility and starts 150 days before the event as it is shown in the diagram.During the event window we will determine whether there is presence of "abnormal" volatility.
We have also ensured that both windows do not overlap each other in order to avoid a deficient estimation.
This model will allows us to test the existence of "abnormal volatility" 17 the day of the rating announcement and the surrounding days.The existence of a significant excess volatility will be considered as a credit event.We will consider abnormal volatility as a positive difference between observed volatility on day t and the normal volatility.
Where: = Abnormal volatility on day t; = Volatility observed on day t; = Normal volatility estimated. 16German bund is considered as the less risky Sovereign bond within the EMU. 17The concept of excess volatility or "abnormal" volatility has been used in the financial literature in various articles such as Reisen von Maltzan (1998), Bouraoui (2009).
-3 0 -1 5 0 0 3 0 The statistical analysis shows that the estimation of the cumulative abnormal level of volatility is significant different from zero between 20 and 30 days prior to the announcement.This means that information from rating agencies is anticipated by market players.We also observe that there is a significant level of abnormal volatility in the aftermath of the event, meaning that the announcement further contributes to market instability.
We observe that the reviews for possible negative ratings may contribute to slightly increase the volatility of the bond yield spread, although this abnormal increase tends to be anticipated as well but only few days before the event, as it is shown by the statistical analysis.Volatility goes up close to 3 % above the normal level of volatility the days prior to the announcement of the negative outlook review; nevertheless, it goes down again after few trading days.
In order to fully respect the assumption of not overlapping between event windows Mackinlay (1997:27), we have performed a new analysis with a shorter event window (4 days) that should lead to more accurate results.We have dropped from the analysis all these events of the same nature whose event windows overlapped (see appendix 5).Moreover, as stated above, events related to Greek bonds announced in 2012 have been removed as well.We have ended up with 34 rating downgrades,

Table 4 : Short term impact of rating announcements on spread volatility
Wrana, Javier y Martín, Jose María.Eurozone sovereign bonds and rating assessments: impact on volatility.