The effect of credit rating downgrades on sovereign
bond yield during the Eurozone sovereign debt crisis
Bachelor thesis Joeri Schouten 22-7-2013 Second draft
Introduction
The importance of the assessments of sovereign credit ratings by credit rating agencies has increased dramatically in the pre-Eurzone crisis period (Hooper et al., 2008). The capability of credit rating agencies of assigning the correct risk profile to individual sovereigns has however been heavily criticized. This critique consist of the fact that credit rating agencies failed to predict the Asian crisis and that they usually provide pro cyclically up- and downgrades which might have worsened previous crisis (Hooper et al., 2008). It is however not clear whether or not the change in credit rating
assessment actually effects sovereign bond yields significantly, because when efficient market
hypothesis holds the impact of credit rating changes are not likely to have a significant impact (Cantor and Packer, 1996).
Sovereign credit ratings are an assessment of the relative probability that a borrowing sovereign will default on its obligations. Because nowadays more high- risk governments are borrowing and because many investors prefer credit rated securities over unrated securities, the demand for credit ratings of these specific high-risk governments has also increased significantly (Cantor and Packer, 1996). Not only are sovereign ratings important for issuing (international) capital, they are also important because the sovereign rating assessments usually affect non-sovereign
borrowers of the same nationality (Cantor and Packer, 1996). The three big credit rating agencies, Standard&Poors’, Moody’s and Fitch, provide two types of credit rating assessments. The first is an actual credit rating status assessment and the second is a change in outlook (IMF, 2012).
Because the number of credit rating changes during the Eurozone crisis from 2008 until 2013 increased substantially, the role and influence of credit rating agencies has been subject to intense debate (IMF, 2012). Although a IMF working paper is written about financial markets spillovers due to credit rating changes during the Eurozone crisis, no specific research about the effect of credit rating changes on European sovereigns during the 2008-2013 Eurozone sovereign debt crisis has been done yet. In this thesis the effect of credit rating downgrades on sovereign bond yields during the European sovereigns debt crisis will therefor be assessed.
In order to form a basic framework for the empirical research in this paper a literature review of the most relevant studies on the topics of explaining bond yields (IMF, 2011), efficient market hypothesi (Hooper et al, 2007), the role of credit rating agencies in the economy (Bannier and Hirsh, 2010) and the effect of credit rating changes on the bond yields (Packer, 1996) is done. The first empirical part consists of a case study on the impact of credit rating changes on the bond yields of all Eurzone countries. For every country the number and date of the downgrades between 2010 and 2013 are analyzed. The effect of these downgrades is assessed for every country by looking at how time series data on 10 year bond yields changes during a period of 20 days before and 10 days after the downgrade. The results of these specific time series are then summarized in several graphs. This is done by creating index numbers of the bond yields time series (date of announcement = 100) so that
the bond yields can be compared and then the average of these index number will be computed and plotted. Then a multiple regression model will be made which explains changes in bond yields by relevant economic factors provided by in the paper on Cantor and Packer (1996) and the is credit ratings by the three credit rating agencies. After the empirical analysis the results from both the case study and the regression analysis will be analyzed and possible model misspecifications will be assessed. Then a advice for follow up research is provided and the thesis will end with a conclusion.
Literature review
Although a lot of research has been done about the effect of credit rating announcements a, solid benchmark study which focuses primarily on the effect of credit rating downgrades on sovereign bond yields during the 2010-2012 Eurozone sovereign debt crisis is not available. This is why past research about credit rating announcements and the Eurozone sovereign debt crisis have been selected to construct a framework for the empirical research in this thesis.
Cantor and Packer (1996) find that individual sovereign credit rating downgrades affect bond yield spreads. The conclusion from the case study was that the announcement of a change in a rating agencies sovereign credit risk assessment is followed by a statistically significant change in the sovereign bond yield. A very important finding in their study is that the impact of rating changes is significantly smaller for investment grade sovereigns then for non-investment grade countries. This paper also conclude that rating announcements which are completely anticipated by the market create a larger impact than credit rating changes which are not. This conclusion was completely the opposite of their expectations given that the efficient market hypothesis predict that when markets already anticipate the credit rating change is not likely to cause any significant change in bond yield spreads. Another important conclusion form Cantor and Packer (1996) comes from their regression analysis which includes the credit rating status of sovereigns explains a higher percentage of the total variation than the model without the sovereign credit rating status. This confirms their conclusion from the case study that credit rating agencies do provide additional information to financial markets. A big
difference from the current (2013) situation and the situation during the research of Cantor and Packer is that at that time only two rating agencies, Moody’s and S&P were active. Currently a third credit
rating agency, Fitch, is providing the market with similar credit rating information. The effect of sovereign rating changes was also researched in a more recent paper by Hooper
et al. (2007) which focuses on the reaction of stock markets to sovereign credit rating changes instead of bond yields. The conclusions of the study by Hooper et al. were mostly in line with the study of Canton and Packer (1996). The first conclusion which is highly important for this thesis is that the impact of credit rating downgrades for countries in a crisis situation is especially significant. This is also the case for high debt countries. They also find that the market impact associated with credit rating changes from as early as five days prior to the announcement date are highly significant.
Another important conclusion from the paper of Hooper et al. (2007) is that sovereign credit downgrades lead to portfolio restructuring by large institutional investors. This implies that large investors reduce their exposure to a country when it’s sovereign credit rating decreases which leads to a decrease in demands for bonds which leads to an increase in sovereign bond yields.
Beetsma et al. (2012) performed a recent study on the effect of news on European sovereign bond yields during the sovereign debt crisis. News in this study are defined as news about economic indicators of a country provided by the Eurointelligence newsflash. The main difference between this study and the research performed is this thesis is that this study focuses on all news available instead of credit rating news only. This study is very relevant for this thesis because it focuses on the effect of news, which a credit rating announcement is too, on sovereign bond yields during a time span and a region which are identical to the research performed in this thesis. An important finding of Beetsma et al. is that when good and bad news are individually analyzed, bad news causes especially upward pressure on bond yields from non-investment grade countries, the so called GIIPS countries (Greece, Italy, Ireland, Portugal and Spain) . They also find substantial spillover effect of negative news between these non-investment grade countries. Spillovers from non-investment to investment grade countries have also been found, but these effect are substantially smaller.
In an ECB working paper from De Santis (2012) the determinants of sovereign bond yields are assessed. This paper is particularly relevant because it investigates spillover effects of sovereign credit ratings during the financial crisis of 2007-2011 in the Eurozone. An important conclusion of this paper estimated spillover effects from a one notch credit rating downgrade of Greece, Ireland and Portugal are significant with respect to other Eurozone countries with weak financial fundamentals. Another important conclusion is that the estimated spillover effect from a credit rating downgrade from Greece leads to severe contagion particularly in the case of Ireland, Portugal, Italy, Spain, Belgium and France. Another interesting part of this paper is that it attempts to explain the reason why a credit rating downgrade has an effect on sovereign bond yields. The first explanation is that most likely market analysts wait for a proper re-elaboration of the available information carried out by the rating agencies before allocating assets. The other reason might be that institutional investors are obliged to hold bonds with a minimum rating; while banks have to meet capital requirements set by regulators and can only use high quality assets as collateral to obtain credit from the central bank, both aspects based on credit ratings systems. Clearly, causation may not only run from ratings to sovereign spreads, but also from sovereign spreads to ratings. In fact, credit ratings respond to developments in sovereign spreads.
Case study
For the case study all individual credit rating downgrades from 2010 until 2012 for Eurozone countries are collected from the websites of Moody’s, S&P and Fitch. Then the 10 year daily bond
yields of all the Eurozone countries except Luxemburg, Estonia and Cyprus, which don’t have debt outstanding which can be compared with ten year bonds (ECB 2010), from 2010 until 2012 are collected. The effect of the credit rating downgrades on the 10 year bond yields is then assessed by placing the individual credit rating downgrades for the individual countries at the day the credit rating downgrade took place. After that a timespan of twenty days before and ten days after this credit rating event is selected in order to assess how the bond yields develop prior and after the credit rating
downgrade. By analyzing the bond yields before the credit rating downgrade announcement the degree to which the market predicts the credit rating downgrade can be determined. Then by analyzing the ten days after de credit rating downgrade the reactions of the market on the actual credit rating downgrade can be determined.
In order to compare the outcomes of this data analysis the bond yield time series have been converted into index numbers with the day of the downgrade being 100. Then the average of all these index numbers have been computed in order to determine a general trend for the Eurozone countries. Recent studies from Beetsma et al. (2010) have shown however that there is a significant difference between the way investment grade and non-investment grade countries react to credit rating
downgrades. In order to test whether these credit rating downgrades actually cause a structural break in the average bond yield index trend Chow’s test on structural breaks is performed. The data on which these tests are based are five days prior and five days after the credit rating downgrade invent in order to isolate the break due to the downgrade event from other event that might have happened during the same time span.
Not all of the credit rating downgrade events have been used for the case study analysis. This is because a large part of the credit rating downgrades from different credit rating agencies were clustered in the same time span, especially in the case of non-investment grade countries. In this case is it harder to isolate the effects of these different downgrades from each other. Removing all the clustered credit rating downgrade events would however have serious consequences for the number of observations that could be used for the analysis. For this case study analysis the decision was made that when a clustering of credit rating downgrade events occurred then only the first would be used for computing the average bond yield index. The total number of credit rating downgrade events for all the countries of interest between 2010 and 2012 was 68. After further analyses 52 credit rating downgrades were found to be found sufficiently isolated to use for the computation of the average bond yield index. In the case of some of the investment grade countries no credit rating downgrade has taken place. These countries are therefor also excluded from the analysis. The investment grade countries are Austria, France and Slovakia which have been downgrade 15 times in total. The non-investment grade countries are Greece, Portugal, Spain, Ireland, Italy and Slovenia which have been downgraded 37 times all together.
The three different case study show very different results. For the case study based on the aggregated data a steep increase in the bond yield index takes place after twenty days of almost continuously increasing bond yields. This shows that markets either anticipate on the credit rating downgrade or that markets already incorporate the information on which the credit rating downgrade
at t=0 is based. The last option would be in line with the efficient market hypothesis. The increase in bond yield index after the credit rating downgrade however only continues for five days until it starts decreasing up till a level of about 99. The total increase between the beginning and the end of the time span is 2 percent point.
The case study for investment grade countries show an entirely different trend. A sharp increase 10 days prior and a sharp decrease 4 days prior to the credit rating downgrade event takes place after which the bond yield index more or less stabilizes at 99. In contrast to the case study of the aggregated data in this case the total change in bond yield index during the time span of 30 days is minus 2.
The case study for the non-investment grade countries share some grade similarities with the case study of the aggregated data which is mainly because the number of credit rating downgrades for non-investment grade countries dominates this number for investment great countries. What
distinguishes the non-investment grade country bond yield index however is that the increase prior to the credit rating event is much more gradual and after the downgrade it does not decrease as much as in the case of the index of the aggregated countries. The total change in bond yield index is for the non-investment grade countries is 5 percent point.
In order to test whether the credit rating downgrade event actually causes a statistically significant break in the bond yield index trend Chow’s test on structural breaks is performed on all three of the case studies. The results from these test show that in all of the three cases the a significant structural break has taken place after the credit rating downgrade event.
Regression method
In order to estimate the effect of credit rating downgrades on sovereign bond yields a panel data regression is performed. To perform this panel data regression analysis it is important to
determine the explanatory variables. A regular regression analysis has been performed by Cantor and Packer (1996) and they have determined seven highly relevant variables which are also used by the credit rating agencies themselves in order to assess sovereign credit ratings.
The first variable is per capita income. The relevance of the per capita income has mainly to do with the size of the tax base of the government (Cantor and Packer, 1996). The greater the per capita income the greater the government tax base and the higher the probability that a government can repay its debt. So the estimator of this per capita income is expected to be negative. The second variable is the annual growth of GDP. A high annual growth in GDP suggest that a country’s existing level of debt is will be easier to repay over time and vice versa (Cantor and Packer, 1996). The coefficient for GDP growth is also expected to be negative. The third variable is the current account deficit. A large current account deficit measures how heavy the public and private sector depend on foreign funding (Cantor and Packer, 1996). An increase in the current account deficit may become unsustainable over time and therefore the coefficient of this variable is expected to be positive. The fourth variable is government budget. A government budget deficit absorbs domestic savings and implies that a government is not able to raise taxes in order to cover its current spending or to repay its debt (Cantor and Packer, 1996) so the estimator for this variable is expected to be negative. The sixth variable is the percentage of external debt divided by GDP. It tells something about a country’s vulnerability towards currency risk (Cantor and Packer, 1996) but because the Eurozone is as a whole is a relative closed economy it is probably not a substantial risk. The most important implication of high external debt is that it corresponds to a higher risk of default. The seventh variable is economic development. That is, once countries reach a certain income or level of development, they may be less likely to default.
Most of the variables defined by Cantor and Packer (1996) are also used for the regression analysis in this paper. In this regression economic development, as measured by the IMF
industrialization index, is excluded. The reason for this is that this index is simply constant for all of the countries in the index. This was not the case in the research performed by Cantor and Packer
(1996) because the varieties in fundaments between the countries that were examined in their research was much larger than the varieties in the fundamentals of the Eurozone countries. Another variable which has been used in the model or Cantor and Packer 1996) but is omitted from the regression model is inflation. Cantor and Packer state that a high rate of inflation may imply structural problems in the government’s finances. When a government appears unable or unwilling to pay for current budgetary expenses through taxes or debt issuance, it must resort to inflationary money finance. Public dissatisfaction with inflation may in turn lead to political instability. This explanation is however not valid with respect for the Eurozone because individual countries cannot use monetary financing in order to finance government expenditure. Another difference is that a variable which measures total debt to GDP is not included in the regression of Cantor and Packer (1996). Because this variable has proofed to be highly significant in the study of for explaining bond yields Beetsma et al. (2010) it is added to this regression.
The main difference in methodology between this regression and the regression of Cantor and Packer (1996) is however that in this regression dummy variables for the different countries are added in order to control for country specific effects. This is done because it is not likely that the coefficients are constant for the different countries in the data set because de fundamentals of the countries differ strongly. Given the short time span of the data set, it is however quite likely that the estimated coefficients are constant over time. The fact that in this regression the country fixed effects are included decreases the change of the occurrence of substantial omitted variable bias and thus makes the estimators of interest less biased. Because the only functions for the estimators of the country fixed effects is to control and prevent omitted variable bias to occurre they and are not of specific interest, they will not be analyzed.
In order to test the effect of credit rating downgrades on sovereign bond yields four different regressions are performed. The first analysis includes all the relevant variables excluding the credit rating opinions of the three credit rating agencies and is based on the aggregated data. This regression
is mainly performed in order to get a feeling to what extend the model fits the data. The second regression model does include the three credit rating variables afterwards a F test on the possible increase of the models’ explanatory power is performed. One of the main conclusions from the literature review is that the effect of credit rating downgrades differs amongst investment- and investment grade countries. The effect of a credit rating downgrade on sovereign bond yields of non-investment grade countries would be more substantial than for non-non-investment grade countries. This is why the model including the three credit rating variables is applied to the investment and
non-investment grade data individually in order to assess whether or not the outcomes differ substantially from each other.
𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 = 𝛼𝛼 + 𝛽𝛽1 𝑃𝑃𝑃𝑃𝑃𝑃 + 𝛽𝛽2 𝑃𝑃𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 + 𝛽𝛽3 𝐼𝐼𝐼𝐼𝐼𝐼 + 𝛽𝛽4 𝑃𝑃𝐵𝐵𝐺𝐺𝐺𝐺 + 𝛽𝛽5 𝐸𝐸𝐸𝐸𝐺𝐺𝐺𝐺 + 𝛽𝛽6 𝑃𝑃𝐺𝐺 + 𝛽𝛽7 𝐼𝐼𝐺𝐺 + 𝛽𝛽8 𝐼𝐼𝐺𝐺 + 𝛽𝛽9 𝑃𝑃𝐺𝐺𝐺𝐺 + 𝛽𝛽10 𝑆𝑆𝑆𝑆𝐺𝐺𝑆𝑆 + 𝛽𝛽11 𝑆𝑆𝑆𝑆𝐺𝐺𝐺𝐺 + 𝛽𝛽12 𝐴𝐴𝐺𝐺 + 𝛽𝛽13 𝐵𝐵𝐸𝐸 + 𝛽𝛽14𝐼𝐼 𝐺𝐺 + 𝛽𝛽15 𝑆𝑆𝑃𝑃 + 𝛽𝛽16 𝑀𝑀𝐵𝐵𝐵𝐵𝐵𝐵𝑦𝑦′𝑠𝑠 + 𝛽𝛽17 𝑆𝑆&𝑃𝑃 + 𝛽𝛽18 𝐼𝐼𝐼𝐼𝐺𝐺𝑃𝑃𝐺𝐺 + 𝜀𝜀
Results from regression analysis
The first regression analysis was based on the aggregated data of the investment- and non-investment grade countries for all the variables except the credit rating opinions from the three credit rating agencies. The results from this regression analysis show that the model fits the data quite well since, except from the variable current account, all variables are significant at α = 0,01 and the percentage variance explained is 75,60%. There are however also some drawbacks to the quality of the model since the estimated relation between bond yields and the variables GDP per capita and Current Account contradict with rational economic expectations as shown in table 2.
In order to assess whether or not the quality of the model improves by adding the three credit rating variables a second regression analysis on the aggregated data is performed which includes these variables. The results from this analysis are shown in table 4 and indicate that now all
economic variables are significant at α = 0,01 except Current Account, which is significant at α = 0,05. By adding the three credit rating variables the R⇒ has increased by 0,046 to 0,802 which is a
statistically significant improvement (F = 21). A much more important finding is that all three of the credit rating variables are significant at α = 0,05 and the two variables Moody’s and S&P are even significant at α = 0,01. This may indicate that credit rating opinions are actually taken into account by investors when assessing sovereign credit risk.
An important conclusion from the literature review is that there might be a difference in significance between credit rating downgrades of investment grade and non-investment grade
countries. The third regression analysis is therefor based on data from non-investment grade countries only. The results from this analysis are quite in line with the results from the model based on the aggregated data. All credit rating variables turn out to be significant at α = 0,01 except from Fitch which is significant at α = 0,05.
Table 4: Results from regression analysis including the credit rating variables
The last regression analysis is based on data from investment grade countries only. The results from this analysis show some differences with the results from previous regression analysis. The first difference is that except from Moodys’ all variables are significant at α = 0,01. The second and most important difference is that for investment grade countries Moodys’ credit rating opinion does not seem to be significant at all.
Analysis of case study and regression analysis results
In order to compare the results from the case study analysis and the regression analysis it is important to first understand the differences in methodology. The first important difference between the methodology used for the case study and the methodology used for the regression analysis is that for the case study only the bond yield data 20 days prior until 10 days after the credit rating
announcement have been assessed. This was however not the case for the panel data regression analysis which has been performed with data covering the entire three years (2010-2012). Another difference is that in order to create an unbiased case study analysis the credit rating downgrade events from the different credit rating agencies have been largely isolated from each other. This means that the cases in which credit rating downgrade announcements from multiple credit rating agencies were clustered, meaning that they took place during the same time span of less than 20 days prior or less than 10 days after each other, were excluded from the analysis. This is however not done in the regression analysis which could cause the credit rating status opinion of the three credit rating agencies
to be correlated and there for biased. From the case study analysis can be concluded that for non-investment grade countries the
credit rating downgrade is much more anticipated on for investment grade countries. This might be because markets investigate non-investment grade countries better and information which might eventually lead to a credit rating downgrade is better incorporated in the market price of bonds.
Another very interesting result is that during the total time span of 30 days in which the bond yield index is observed bond yields increase by 5 percent point for non-investment grade countries and decease by 2 percent point for investment grade countries. This indicates that the impact of a credit rating downgrade on a non-investment grade sovereigns’ bond yield is much more important than for an investment grade sovereign. This conclusion is in line with the conclusion by Cantor and Packer (1996).
The first regression analysis which only includes seven macro-economic indicators already seems to fit the data quite well. When adding the three credit rating variable to the regression model the explanatory power of the model increases significantly and all three of the credit rating variables turn out to be significant at α = 0,05. Another positive aspect of the results from all the models is that the relation between bond yields and a change in credit rating opinion is estimated as being positive, which is completely is line with expectations. A very interesting difference when comparing the results from the two regression analysis from investment and non-investment data is that for the first case the credit rating opinion turns out to be not significant although all the other variables are significant at α = 0,01. It is however very hard to explain why this would be the case. A reason might be that in the cases when credit rating downgrades are clustered Moodys’ is most of the time the latest credit rating agency to announce a downgrade. Because the announcements of Fitch and S&P is then already largely incorporated in market bond yields and therefore Moodys’ rating downgrade
announcement have a significant effect anymore a sovereigns’ bond yields. This explanation is however not very plausible because in the case of investment grade countries clustered rating announcements can be observed only a very view times. Another interesting conclusion is that the model including the credit rating variables seems to fit the data of the investment grade data better (R⇒ = 0,88) than the non-investment grade data (R⇒ = 0,76). This indicates that bond yields from investment grade countries are easier to predict than bond yield from non-investment grade countries. This however does not imply that credit rating downgrades from investment grade countries are also easier to predict. This is given the results from the case study analysis probably also not the case.
Because a lot of relevant variables, including the country specific effects, have been
incorporated in this model the change of substantial omitted variable bias has been reduced. There are however two technical aspect which might affect the interpretation of the results from the regression analysis substantially and therefor needs to be reconsidered. Credit rating agencies use all the
macroeconomic fundamentals which are also used in this regression analysis in order to determine the credit rating status of a sovereign. Therefore it is very likely that not only are the three credit rating variables highly correlated amongst themselves but are also highly correlated with respect to the macroeconomic fundamental variables used in the model. In order to assess whether or not this is actually the case two test are performed. This first is to establish a correlation matrix and determine which variables within the model are specifically very strongly correlated. The outcome of the correlation matrix shows that the correlation amongst the three credit rating variables ranges between
0,86 and 0,91 which affirms the expectations that the three credit rating variables are highly correlated. The correlation between the macroeconomic fundamentals and the three credit rating variables
however turns out to be not very high. The highest correlation is between debt to GDP (0,64) and the three credit rating variables and the growth rate and the three credit rating variables (0,61). The
problems that come with the occurrence of this multicollinearity is that the estimators might have large standards errors which affect the t-statistics. The second, more severe problem, is that due the
variability of the coefficients the same coefficient may be far from the actual population parameter which can also cause the estimators to have opposite signs.
Another reason to suspect that the estimators might be biased it that the simultaneous causality might apply to this model. The question that needs to be answered in order to determine whether this is actually the case is if the increase of bond yields from an individual sovereign leads to a credit rating downgrade or that the increase in bond yield lead to a credit rating downgrade. One of the conclusions from the case study is that in case of investment grade countries bond markets more or less anticipate to the credit rating downgrade announcement so that the credit rating downgrade itself did not cause a significant change in the timeserie trend. This conclusion implicates that increasing bond yields are not caused by the credit rating downgrade. In the case of non-investment grade countries however there is a significant increase of the sovereign bond yield index after the credit rating downgrade but do not seems to anticipate on this increase in the twenty days prior to the downgrade. This may implicate that sovereign credit rating downgrades actually negatively influence the bond yield of non-investment grade countries. Based on the case study results simultaneous causality can however not be ruled out.
A credit rating downgrade announcement itself is new information to the market. Whether or not this announcement actually provides additional information to the market is not relevant for this discussion. This is because the news of the occurrence of a credit rating downgrade might not be the only news which is relevant for sovereign bond markets. Beetsma et al. (2010) already showed that also other news on sovereigns in the Eurozone, like statements about a country from opinion leading institutions like the IMF, ECB or large investment banks have a significant impact on sovereign bond yields. This conclusion might also cause the estimators of the three credit rating variables to be biased because they might include the effect of other news events and would there for be “overestimated”.
Suggestions for follow up research
Although the regression models applied to the 2010-2012 data on economic fundamental and credit rating opinions of Eurozone countries extends the model of Cantor and Packer (1996) there are also some drawbacks to the models used in this thesis which can be solved in future research. One of the main problems of the models used in this thesis is the correlation between the macroeconomic fundamental and the three credit rating variables. Also the simultaneous causality arising between
bond yields and the three credit rating variables is a reason to assume that the estimators in this model are biased. This problem could be avoided in future research by using the instrumental variable method for the three credit rating variables.
Another reason why the estimators in this model might be biased is the fact that the absence of other news related variables might cause the estimators more significant than they actually are. In future research it is recommended to control for these other news related variables in order make the estimators for credit rating announcements more accurate.
Conclusion
The importance of the assessments of sovereign credit ratings by credit rating agencies has increased dramatically in the recent years (Hooper et al., 2008). Cantor and Packer (1996) find that individual sovereign credit rating downgrades affect bond yield spreads. Hooper et al. (2008) conclude that the impact of credit rating downgrades is especially significant for countries in a crisis situation. Beetsma et al. also conclude that the effect of other news, which a credit rating announcement is too, on sovereign bond yields during a time span and a region which are identical to the research performed in this research. An important conclusion of the paper of De Santis (2012) is that a one-notch
downgrade of sovereign bonds by credit rating agencies increases the bond yield of the individual sovereign.
The results from the case study show that the effect of a credit rating downgrade differs strongly between investment and non-investment grade countries. In the case of investment grade countries a credit rating downgrade leads to decrease in the bond yield index. In the case of non-investment grade countries the credit rating downgrade seems to be well anticipated upon and the bond yield index keeps increasing after the downgrade.
The different regression models based on aggregated data, data of investment grade countries and data on non-investment grade countries show very similar results. In case of the regression analysis based on the aggregated and the non-investment grade data all credit rating variables turn out to be significant at α = 0,05. This is in line with the results from the case study. In the case of the analysis based on the investment grade data the credit rating variable Moodys’ turns out to be not significant. There is however not a good explanation for why both Fitch and S&P are significant and Moodys’ is not.
There are however also some model misspecifications. These misspecifications arise do to omitted variable bias which comes from not including other news related variables in the model. But also multicollinearity and simultaneous causality are likely to have biased the estimators in the models. In future research, these problems could be solved by added other news related variables and using instrumental variable regression in order to reduce the multicollinearity and simultaneous causality.
References
Arezki, Candelon and Sy. 2011. “Sovereign Rating News and Financial Markets Spillovers: Evidence from the European Debt Crisis”. IMF Woking Paper.
Beetsma et al. 2012. “Spread the News: the Impact of News on the European Sovereign Bond Markets during the Crisis”, Economic Policy, pp. 640-68
Cantor, Packer. 1996. “Determinants and Impact of Sovereign Credit Ratings”. Economic Policy Review, 22: 37-54.
De Santis, R.A., 2012. “The Euro Area Sovereign Debt Crisis: Safe Haven, Credit Rating Agencies and the Spread of the Fever from Greece, Ireland and Portugal”, ECB Working Paper
Hooper, Hume, Kim. 2008. “Soverign rating changes – Do they provide new information for stock markets?”, Economic Systems, 32: 142-166.
Tabels
Average volatility of index
Greece 20,1 Portugal 9,1 Ireland 7,4 Slovenia 6,2 Slovakia 5,9 Austria 5,0 Spain 4,5 France 4,2 Italy 4,0 Belgium 2,6 Finland 0,0 Germany 0,0 Luxemburg 0,0 Netherlands 0,0
Table 7: Number of downgrades used for case study analysis
