No. 12/2 July 2012

Rabobank Working Paper Series

Sovereign credit ratings
An assessment of sovereign ratings provided by Moody’s, S&P and Fitch

Authors: Joep Pennartz, Jan Pieter Snoeij
The views expressed in this paper are their own and not necessarily those of Rabobank.
Contact: Joep Pennartz and Jan Pieter Snoeij wrote this working paper as a part of an internship at Rabobank. For further information please contact their supervisor at
Rabobank’s Economic Research Department: S.A.Kamalodin@rn.rabobank.nl

Editors:
Allard Bruinshoofd, head of International Economic Research
Shahin Kamalodin, economist

1

“There are two superpowers in the world today in my opinion. There’s the United States and there’s
Moody’s Bond Rating Service. The United States can destroy you by dropping bombs, and Moody’s can destroy you by downgrading your bonds. And believe me, it’s not clear sometimes who’s more powerful.” (Friedman, 1996).

Introduction
Credit rating agencies (CRAs) are of major importance in international financial markets. Their prominence is explained by the myriad number of traded fixed income securities; one simply cannot assume that every market participant has the resources to assess the credit risk of each borrower.
That is where CRAs step in; they distil public and private information into a simple credit rating. The informative value of these credit ratings lowers information asymmetries and enhances transparency and liquidity (Katz, Salinas, & Stephanou, 2009; IMF, 2010). Furthermore, credit ratings facilitate certification of credit risk, which is widely utilised by financial market participants and supervisory authorities (Kiff, Novak, & Schumacher, 2012).
Given the extensive influence of CRAs on the functioning of capital markets, an appraisal of their performance is desirable, which is illustrated by the constant scrutiny the major agencies –Fitch,
Moody’s and Standard and Poor’s– face in global media. Previous academic literature has examined rating agency operations from multiple perspectives, assessing predictive power of credit ratings 1, corporate rating accuracy and timing 2 and policy responses to CRA flaws 3. Despite the vast amount of research on CRAs, we have found no study that attempts to examine the differences in sovereign rating quality between the major rating agencies. Hitherto, credit ratings have been perceived as a homogeneous set of appraisals, although quality differential between agencies would impose severe consequences. Therefore, this working paper examines the relative rating quality of sovereign credit ratings issued by the three major agencies.
The assessment considers sovereign credit ratings (SCRs) because there tends to be more disagreement in the assessment of sovereign credit risk compared to corporate credit ratings (BIS,
1999; Brooks, Faff, & Hill, 2010). Rating quality is judged based on three dimensions: rating accuracy, timing and stability. The first two criteria are associated with the informative role of ratings and are evaluated based on rating developments prior to a sovereign default. The latter is of importance to the certification function and is assessed by taking all issued ratings into account.
The results indicate that S&P’s SCRs are most accurate proximate to defaults and score best at rating timeliness, due to the aggressiveness of their rating changes. Moody’s ratings are the most stable and also have the highest predictive power at time horizons longer than 1 year prior to default. Hence, this working paper identifies that there are differences in the informational value of the ratings provided by the different CRAs; user preferences should determine which agencies ratings’ are most appropriate. The paper is structured as follows. Chapter 1 discusses the methodology. Chapter 2 encompasses default data and stylized facts. Chapter 3 entails an overview of the results. Chapter 4 concludes.

1
2
3

For example, see Sy (2003), IMF(2010), Kiff et al.(2012) for a survey
For example, see Bheenink (2004), Güttler and Kramer (2008)
For example, see Katz et al., (2009) House of Lords (2011), Kiff et al. (2012)

2

Methodology
This chapter introduces our definition of rating quality and discusses the matching quality measures that form the basis of empirical analysis in subsequent chapters.
Measuring rating quality
There is no commonly acknowledged definition of rating quality available in academic literature (Duff
& Einig, 2008). Therefore, we have identified three characteristics that are of importance for rating users, which are given below:
1. Accuracy
2. Timing
3. Stability
The rationale behind these ‘quality dimensions’ is relatively straightforward. Rating accuracy concerns the discriminatory power of sovereign credit ratings. The more precise an agency can classify credit risk, the higher the quality of its ratings. Rating timing addresses which agency responds most adequately to changes in credit risk. Both these dimensions are essential to the informative value of ratings, since this value decreases when the timeliness or discriminatory power deteriorates. The stability dimension evaluates whether agencies moderate the volatile movements of credit risk. More fundamentally, whether a rating is based on a point-in-time or to a through-the-cycle approach of credit risk (see box 1 for a description). Stability is of importance because of its impact on the certification role of ratings; rating users may need to adapt each time a rating changes, which causes rating users to value stable ratings.
A rating scorecard is derived based on these quality dimensions. Table 1 displays an overview of the methods conducted to examine rating quality. The remainder of this section elaborates upon these quality measures.
Table 1. SCR scorecard
Quality Dimension
Accuracy
Timing
Stability

Measure
Cumulative Accuracy Profile (CAP-curve)
Accuracy Ratio (AR-ratio)
Downgrade profile
Rating cycle leaders and followers
Frequency of rating changes
Frequency of cliff effects (> 2 notches)
Frequency of rating reversals

3

Box 1.

Achieving rating stability

Before discussing how rating stability is studied, it is important to note how rating agencies account for stability in their ratings. Rating agencies minimize rating volatility by basing SCRs on a ‘through the cycle’ (TTC) methodology, rather than taking a point-in-time (PIT) approach. The former implies that ratings are only adjusted when a change in fundamentals occurs; thus credit ratings are unresponsive to cyclical changes (Kiff et al., 2012). Furthermore, the TTC-approach focusses on a long-term default horizon and a prudent migration policy, which implies that the considered time horizon of ratings should be at least one business cycle (Alman & Rijken, 2004). In practise, the TTC methodology translates to several smoothening rules: ratings will only be adjusted if the anticipated change is expected to be persistent, or if the prescribed rating change exceeds a certain threshold
(Cantor & Mann, 2006).
Rating accuracy
The first dimension of rating quality is accuracy; the extent to which a rating correctly reflects the underlying credit risk. Rating failure –in this case inaccurate ratings– can be classified in two different ways, as indicated in table 2. Type I errors occur when a CRA incorrectly assigns a high rating while the underlying credit quality is low. The associated cost for investors is a greater default probability, which can lead to a loss in interest and principal payments or recovery costs. Type II errors reflect the opposite situation; a CRA assigns a low rating while the underlying credit quality is high. The associated costs are opportunity costs, foregone interest income, payment of origination fees and loss due to premature selling at disadvantageous prices (Keenan, Sobehart, & Stein, 2000).
Table 2. Type I and Type II errors
Actual
Low credit quality
Low credit quality

Correct prediction

Type II error

High credit quality

Model

High Credit quality

Type I error

Correct prediction

Cumulative accuracy profiles (CAP)
Sovereign defaults are used as a benchmark to assess SCR accuracy because they represent the only event when the underlying credit risk is known – the default probability is 100%. For this reason, they are the only proper standard to test the predictive power of the ratings. The subsequent chapter discusses this topic in more detail.
SCR accuracy is examined with “cumulative accuracy profiles” (CAP), which is the most common method in finance to assess rating accuracy (Irwin & Irwin, 2012). An example of the CAP-curve is shown in figure 1. The x-axis indicates the cumulative proportion of issuers, which increases from zero to one while moving along this axis. The issuers are ranked from riskiest to safest sovereign ratings. Hence, D ratings are given on the left-hand side of the x-axis while AAA rated bonds are on the right-hand side. By doing so, each rating is treated as a threshold to distinguish defaulting from non-defaulting sovereigns (Irwin & Irwin, 2012). To determine the cumulative proportion of sovereigns rated in a certain category, all issued ratings between 1998 and 2011 are taken into account. 4

The y-axis indicates the cumulative proportion of defaulters. Therefore, the curve is determined by the “cumulative share of defaults accounted for by the cumulative share of ratings” (Moody’s, 2011b).
The CAP curve is determined by the percentage of defaults that occurred within a fraction ‘x’ of all sovereigns whose ratings are equal to, or lower than the maximum rating of fraction ‘x’.
The better the predictive power of a SCR, the lower the ratings of sovereigns prior to default and thus the more the curve is situated at the north-west region in the graph. A perfect rating-model with perfect discriminative power would assign D-ratings to all defaults, whilst no defaults would occur in any of the other rating buckets; the corresponding curve is illustrated in figure 1. On the contrary, a random model without discriminative power between defaulters and non-defaulters would have equal proportions of defaults in any fraction of debtors, resulting in a line with a 45-degree slope between coordinates [0,0] and [1,1].
It is important to note that the CAP-curve provides a snapshot of rating accuracy at a given time prior to a default. To this end, we work with different time intervals in our analysis, ranging from 0 – 5 years prior to default, to provide an overview of the predictive power at different time horizons. Note that the method accounts for both Type I and Type II errors; a Type I error would result in a curve shifting down towards the random walk model since the defaults will fall in higher rating thresholds. A
Type II error implies that non-defaulters are given low ratings, shifting the entire curve to the right; again away from the perfect model.
Figure 1. The CAP-curve

cumulative proportion of defaulters

1

perfect model

0,9

Ap

0,8

rating model

0,7

Ar

0,6

random model

0,5
0,4
0,3
0,2
0,1
0
0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9

1

Cumulative proportion of sovereigns

Accuracy ratio (AR)
The accuracy ratio (AR) is the summary statistic of the CAP-curve, which eases the interpretation of the outcome. It is an indication of the accuracy of a rating model and can be described as a measure of discriminative power (Tasche, 2006). The ratio 4 (illustrated in figure 1) is defined as the area between the random model and the rating model (Ar), divided by the area between the perfect model and the random model (Ap). Therefore, the closer to 1, the higher the accuracy of the SCRs.

4

The AR-ratio has a similar relation to the CAP-curve as the Gini-coefficient to the Lorentz curve.

5

An important downside of the AR ratio is its non-exclusiveness. Every CAP curve only has 1 AR ratio, but a certain AR ratio can apply to multiple CAP curves (Moody’s, 2011a). Therefore, AR-ratios do not provide a view on underlying differences between rating systems. Hence, no distinction between Type
I and Type II errors can be made. Another weakness of the AR ratio is the difficulty to interpret. A ratio of 1 describes a perfect rating model and a ratio of 0 represents a random model. Nonetheless, the AR ratio is no indication of the percentage of correct ratings or correctly forecasted defaults.
Rating timing
The second rating quality dimension concerns downgrade timing preceding a sovereign default.
Timing is considered an important feature because an immediate rating response to a change in the creditworthiness of a sovereign is clearly preferred to lagged rating action. To assess the timingquality, downgrade profiles are created, showing SCR developments prior to default. The downgrade profiles are created for all sovereign defaults that were rated by the three agencies, resulting in a total sample of 13 defaults. An example is presented in figure 2. Each graph contains the rating developments of all three agencies in order to compare the timeliness. The investment grade to speculative grade boundary is represented by the upper horizontal line, while the lower line indicates a default rating. The sovereign default date is indicated by the vertical black-dotted line.
In order to structure the comparison of downgrade profiles we identify two measures to assess rating timeliness: • The first to downgrade preceding a default;
• The first to issue a default rating (downgrade to SD rating or lower)
Being the first to downgrade preceding a default indicates which agency was the first to identify a deterioration of credit quality. While being the first to recognize a default 5 is where this ‘race to the bottom’ is all about: who is the first to determine a country heading for default, or in default?
Additionally, we will determine the agency with ‘best timing’ and ‘worst timing’ per case study. These are qualitative assessments of the downgrade profiles, taking into account the initial rating, the height of the rating over the entire period and the timing and magnitude of the downgrades. However, because a downgrade profile is a ‘race to the bottom’ it is likely that an agency that downgrades most aggressively wins. The profiles only show the instances when a default actually occurs, but neglects the instances when the winner is too aggressive. Overly aggressive rating actions are not taken into account here, but the effects may show in the subsequent section as they lead to large rating changes
(i.e. rating instability). Furthermore, data on downgrade cycles is examined to show which agency tends to be the first to move, and which one tends to lag behind. In combination, these indicators provide a view on the timing and downgrade aggressiveness of the rating agencies.

5

The recognition of a sovereign default by a CRA can potentially cause a default. Therefore, the first agency to acknowledge a default can affect the market. This endogenous effect can affect our analysis. However, literature assessing whether CRAs are leading or lagging the market is inconclusive. See for example: Afonso, Furceri, &
Gomes (2011), Gande & Parsely (2005), Brooks, Faff, Hillier, & Hillier (2004) and Faff & Hill (2007). Therefore, our analysis does not take this effect into account.

6

Figure 2. Downgrade Profile

Rating stability
In accordance with CRAs own stability measures (Cantor & Mann, 2003), the following indicators are reviewed to examine rating stability:
• Frequency of rating changes;
• Frequency of cliff-effects;
• Frequency of rating reversals.
A high frequency of rating changes indicates rating volatility that should be moderated by the TTCapproach. Cliff-effects are large rating changes, which are identified by reviewing rating changes of 3 or more notches. They are undesirable from a stability perspective as they signal large adjustments in credit worthiness that should – from a rating stability point of view – be embedded in the ratings in a series of smaller steps. Rating reversals are downgrades followed by upgrades, and vice versa. They can partly be explained by changes in underlying credit risk. However, they indicate a violation of rating stability as the TTC-methodology aims to eliminate this rating volatility. The smoothening measures that accompany rating stability are somewhat incompatible with rating timeliness, indicating agencies have to position themselves with respect to this trade-off.
To recap, the results of the quality measures are provided in a scorecard. The conducted methods are both of quantitative nature, such as CAP curves and AR-ratios, as of qualitative nature, like the interpretation of downgrade profiles. The goal of these different measures is to provide insight into the characteristics of credit ratings, exposing how rating agencies position their ratings with respect to accuracy, timeliness and stability.
Data description
This chapter introduces sovereign default definitions and introduces the sovereign defaults that serve as data for empirical analysis in the subsequent chapter.

7

Overview of sovereign defaults
CRAs each state their own sovereign default definitions. Surprisingly, rating agencies differ in the recognition of default type and timing. For example, Fitch states that Indonesia defaulted in June
1998, 2000 and 2002, whereas S&P claims they defaulted in March 1999 and April 2000, while
Moody’s does not acknowledge any Indonesian default. Possible explanations for these differences are different default definitions and contradictory opinions on debt restructuring processes, which are ex post assessments of negotiations prior to a sovereign default (Bhatia, 2002).
To be able to compare agency rating performance, we consider an external overview of sovereign defaults as framework for our study. We have decided to utilize Reinhart and Rogoff’s (2011) database, which is a landmark study on the history of financial crises, as benchmark on historical sovereign default data. This choice is motivated by the extensive scope of their dataset. Their definition of sovereign defaults is stated in appendix 2.
In accordance with the default definitions of credit rating agencies, their definition comprises failure to meet the contractually agreed fees, and instances when rescheduled payments are less favourable than originally agreed upon. In addition to foreign currency defaults, domestic defaults include the freezing of bank deposits or the forcing of currency conversions. These scenarios only apply in the case of domestic currency cases because there is a possibility that an entity –usually the central bank– has the legislative backing to enforce such measures, whereas this authority is not enforceable in the case of foreign currency denominated debt. On top of the defaults identified by Reinhart and
Rogoff, we have added data for omitted countries. An overview of the consulted sources and included sovereign defaults is provided in appendix 3.
Datasets
Most of the empirical analysis in the succeeding chapters is based on rating developments prior to sovereign defaults. The included sovereign defaults are shown in table 3, that also indicates which rating agency had issued a sovereign credit rating of the defaulting sovereign at the time of default.
Table 3. General default database
Country
Argentina
Belize
Cameroon
Dominican
Republic
Ecuador

Greece
Grenada
Indonesia

Default
Type

Date

Fitch

S&P

Moody’s

Domestic
Foreign
F
D
F

December 2001

X
X

X
X
X
X
X

X
X
X

D
F
F
D
F
F
F

August 1999

December 2006
July 2004
April 2005

December 2008
March 2012
December 2004
January 1999

X
X

X
X
X
X

X
X
X
X
X

X
X
X
X
X
X
X

8

Jamaica
Moldova
Pakistan
Paraguay
Russia
Seychelles
Ukraine
Uruguay
Venezuela

F
D
F
F
F
D
F
D
F
F
F
D
F
D
F

July 2002
January 2010

X
X
X
X

June 2002
January 1999
July 2003

X
X
X

X
X
X
X
X
X
X
X
X

X
X
X

August 1998
X
July 2008
January 2000
April 2003

X
X

X
X

X
X

X

July 1998
January 2005

X
X
X
X
X

X

The size of the database is shown in the succeeding table:
Table 4. Number of defaults
Foreign defaults
Fitch
S&P
Moody’s

Domestic defaults

Total

11
15
16

5
6
8

16
21
24

Sovereign credit ratings
Before addressing the obtained results, we would like to specify the source of the actual data: historical sovereign credit ratings. We retrieved the data from research by Hollaar (2011). The study entails data on S&P’s sovereign credit ratings from 1976 to May 2011, Fitch’ data from 1995 to July
2011 and Moody’s rating history from 1970 up until early 2011. The datasets have different starting dates because each rating agency issued its first SCR at different times. Unless stated differently, the entire period of sovereign credit ratings is used in the analysis.

9

Results
This chapter presents the results obtained from the rating quality measures. The first three sections cover the dimensions of rating quality discussed above: accuracy, timeliness and stability. The chapter concludes with a scorecard to provide an overview of the derived results.
Accuracy
To

assess

rating

accuracy,

CAP-curves

and

AR-ratios

have

been

constructed.

Figures 3a-3g show the derived CAP-curves for different time intervals: at the time of default, 6 months before, and annually until 5 years prior to default. This structure allows to analyse rating accuracy over different time horizons. The figures also include AR-ratio’s to ease the comparison between CRAs.

Figure 3a -3g

AR-ratio’s
S&P:
Fitch:
Moody’s:

99%
96%
94%

10

AR-ratio’s
S&P:
Fitch:
Moody’s:

91%
90%
86%

AR-ratio’s
S&P:
Fitch:
Moody’s:

71%
68%
74%

11

AR-ratio’s
S&P:
Fitch:
Moody’s:

57%
56%
60%

AR-ratio’s
S&P:
Fitch:
Moody’s:

50%
55%
54%

12

AR-ratio’s
S&P:
Fitch:
Moody’s:

50%
43%
54%

AR-ratio’s
S&P:
Fitch:
Moody’s:

43%
38%
53%

13

The first figure shows CAP-curves based on SCRs at the time of default. Surprisingly, none of the
CRAs show a ‘perfect’ curve. One would expect all agencies to be able to, at the time of default, rate defaulters in the appropriate rating class. However, as indicated in chapter 2, identifying sovereign defaults is an ambiguous practice, which explains the imperfect CAP-curves. The curve indicates that
S&P has the highest rating accuracy. This finding is supported by the AR-ratios, which reveal that S&P has a very high accuracy ratio (99%). Fitch and Moody’s AR ratios are 96% and 94%, respectively.
When the time period increases to six months prior to default, the relative rating accuracy remains the same: S&P still performs best. However, the accuracy deteriorates in absolute terms for all CRAs; the curves shift away from the north-west corner, leading to a drop in the AR-ratios. This reduction in explanatory power is explained by CRA’s lack of perfect foresight; which is understandable. The oneyear CAP-curves signal an important development. For the first time Moody’s becomes more accurate than the other rating agencies. The AR-ratio of Fitch falls to 68%, while S&P attains a score of 71% compared to 74% for Moody’s.
At even longer time horizons, the accuracy differential between S&P and Moody’s increases. Fitch also attains a lower accuracy at the 2-year time interval, then scores high at a 3-year time interval.
However, Fitch performs worst during the 4 and 5-year time intervals. This could be explained by their smaller dataset 6; at the 3-year time interval Fitch’s CAP-curve is based on just 11 defaults; whereas S&P and Moody’s curves are based on 16 and 18 defaults, respectively. The AR-ratios indicate that the accuracy of Moody’s ratings falls by just 1% over the last two years, while the accuracy of Fitch’ ratings deteriorates by 17%. These developments are also shown in the following table. Table 5. AR Ratios
0 year

6 months 1 year

2 year

3 year

4 year

5 year

S&P

99%

91%

71%

57%

50%

50%

43%

Fitch

96%

90%

68%

56%

55%

43%

38%

Moody's

94%

86%

74%

60%

54%

54%

53%

To sum up our findings so far; S&P’s ratings have the highest explanatory power at time horizons at or close to default, while Moody’s performs worst. But over longer time spans, Moody’s outperforms the others. This finding especially holds at higher rating categories, shown by the steepness of
Moody’s cap-curve between an x-value of 0.2 and 0.3 at 3, 4 and 5-year intervals. This suggests that
Moody’s is better at distinguishing non-defaulters and defaulters in the corresponding rating classes
(B1 – Ba3). The divergence in predictive power gets stronger the longer the time horizon. At a 5-year interval, 80% of the defaulters occur among the lowest 30% of Moody’s ratings, while Fitch’ ratings classify them amongst the lowest 55% of their ratings. These results suggest a difference in methodology between the two agencies, whereby Moody’s ratings tend to be more in line with the
TTC-methodology. Box 2 describes a robustness test that re-assesses these results.

6

Unfortunately, the small dataset does not enable the execution of a significance test on the obtained results.
Appendix 1 discusses this matter in more detail.

14

Box 2. robustness test
Besides the default database based on Reinhart and Rogoff’s sovereign default definition, another database has been used to conduct a robustness test. This restricted database comprises the defaults that are recognized by the agencies themselves – which tend to distinguish a limited number of defaults. This translates into a different set of defaults for each agency. The analysis of this dataset is not included in this paper, although the outcomes are briefly discussed. The detailed analysis is available on request.
The CAP-curves derived from the restricted database show similar outcomes. In accordance,
S&P outperforms the other agencies in the short-run, and Moody’s ratings prove to be more accurate when considering longer time periods. Two differences with the general database are the higher concurrence of S&P’s and Moody’s ratings, and the deterioration of Fitch’ ratings explanatory power at shorter time periods. This outcome is unexpected, because normally one would expect rating accuracy to increase when only taking into account acknowledged defaults. This increase in explanatory power is observed in Moody’s and S&P’s ratings. In conclusion, the robustness test has similar outcomes compared to the general analysis.
Timing
This section examines the rating timeliness of the SCRs, by analysing downgrade profiles of 13 internal and external defaults that have been rated by all three CRAs. Appendix 4 shows all downgrade profiles and a short analysis of the rating developments. Table 6 summarizes the results of the qualitative assessment, which appendix 5 shows in detail:
Table 6. Summary of downgrade profile analysis
Best timing
S&P
8

Moody’s
4

Worst timing
Fitch
2

S&P
2

Moody’s
7

Fitch
5

First to downgrade
S&P
6

Moody’s
1

Fitch
4

First to recognize default S&P
8

Moody’s
2

Fitch
2

The table shows that S&P has the best rating timing. They score highest at ‘best timing’ and lowest at
‘worst timing’, are most often the first to downgrade and usually recognize a default the fastest. S&P had the best rating timing during the defaults of Indonesia, Russia, Uruguay, the Dominican Republic and Jamaica, while Moody’s was the best during the Argentine and Greek defaults. Fitch outperformed the others during the defaults of Ecuador and Uruguay.
What are the causes of these results? S&P is usually the first to acknowledge a default, but this could be explained by the fact that they also recognize the most defaults. They tend to be more aggressive in their rating movements, which is illustrated by the downgrades in the Indonesia and Venezuelan profiles. On the other hand, Moody’s recognizes the least defaults and tends to maintain a more stable rating over longer periods, thereby occasionally neglecting imminent defaults in the process.
Table 7 indicates a potential reason why S&P has better timing: S&P most often is the first to take negative rating actions, making them to be the first to move.

15

Table 7. Order of taking negative rating actions
S&P
Percentage of rating actions Moody's

Fitch

First

Second

Third

First

Second

Third

First

Second

Third

59

13

28

13

51

36

28

36

36

Source: IMF 2010.

Table 7 shows that S&P is most often leading rating downgrade cycles, and Fitch and Moody’s are often second and third in their rating actions, which indicates a potential reason for S&P to have the best rating timeliness in our assessment. However, the related downside is that they might be too aggressive. This is not taken into account by the case studies, since they do not take into account
Type-II errors as no cases are considered where downgrade cycles did not end in a sovereign default.
Where S&P tends to have the best rating timeliness due to their aggressive rating adjustments,
Moody’s seem to adjust their ratings less often, reducing timeliness. The associated rigidness is shown in the downgrade profiles of Indonesia, Russia, Uruguay, Venezuela, Dominican Republic and
Jamaica (which are located in appendix 4). The inertia leads to omission in the recognition of some sovereign defaults: of the 13 defaults considered, they only acknowledge 5.
The following table provides more insight in the rating developments of Fitch. On average, they tend to be less timely than S&P but better than Moody’s. Table 8 shows that the performance differential compared to S&P can be explained by Fitch’ initial rating levels. On average they have higher ratings prior to a downgrade cycle. Therefore, they have to downgrade more severely to reach a lower credit rating. Table 8. Height of initial rating prior to downgrade cycle
Moody’s
Highest
Lowest

Fitch

S&P

5
6

7
4

3
8

To conclude, because of the limited amount of case studies there is no strong indication that the timing of initial downgrades differs structurally amongst rating agencies. However, S&P seems to be more timely in the recognition of sovereign defaults because they are more aggressive in downgrading their ratings, while Moody’s tends to be the least timely as they fail to recognize defaults. Fitch is more aggressive in its rating adjustments compared to Moody’s as well, but their initial ratings are the highest and their ratings less timely than S&P’s.

16

Stability
Rating stability is examined to finalize the assessment of rating quality. The examination of stability can potentially explain the results derived in the previous sections: assessment of rating accuracy suggests that Moody’s rating accuracy hardly changed between a 3 and 5 year time horizon, which can be explained by a high rating stability. Furthermore, the downgrade profiles indicated that S&P seems to be more aggressive with its rating adjustments, while Moody’s appears to be more inert. So the question is, are these findings backed by differences in rating stability?
To provide the answer, the following statistics have been calculated:
• Frequency of rating changes
• Frequency of cliff effects
• Frequency of rating reversals
The frequency of rating changes, measured as a share of the total amount of SCRs issued, is stated in table 9. The data shows that over a 10-year period S&P and Fitch change almost 25% of their ratings a year (on average), while Moody’s adjusts less than 19% annually. Over the last 5 years those differences are converging, which might be explained by changes in the methodology or a changing global economic environment. Still, Moody’s has the lowest frequency of rating changes, which is in line with the observations in the previous two sections.

Table 9. Annual frequency of rating changes as measured to total ratings issued.
S&P

Fitch

Moody's

2000-2010

24.76%

24.99%

18.60%

2000-2005

26.43%

28.39%

17.60%

2005-2010

21.99%

20.09%

19.01%

Table 10. Rating reversals and large rating movements
30 days
Large Rating
Upgrades (%)
>2 notches
Large Rating
Downgrades
(%)
>2 notches
Rating
Reversals (%)

60 days

180 days 1 year

S&P

0.16%

0.16%

0.29%

0.62%

Fitch

0.22%

0.22%

0.26%

0.48%

Moody's

0.39%

0.42%

0.48%

0.70%

S&P

0.86%

0.72%

1.63%

1.82%

Fitch

1.41%

1.30%

1.56%

1.89%

Moody's

0.79%

0.84%

1.12%

1.12%

S&P

0.00%

0.05%

0.16%

0.54%

Fitch

0.07%

0.15%

0.26%

0.41%

Moody's

0.00%

0.03%

0.06%

0.14%

17

Table 10 includes data on rating reversals and large rating movements. The data are calculated as a percentage of the total amount of ratings over the entire time period. This explains the small percentages. The first row shows that Fitch has the lowest amount of large upgrades (of 3 or more notches) relative to the total amount of Fitch’ ratings at periods longer than 180 days, while Moody’s has the largest share of large upgrades. Considering the opposite rating action -rating downgradesMoody’s has the lowest percentage of large adjustments while Fitch has the most. This indicates that
Fitch’ ratings are more prone to negative cliff-effects, while Moody’s is more likely to carry out large positive upward adjustments. However, the chance of a negative cliff-effect is more than twice as high. In total, Moody’s large rating changes are lower than those of S&P and Fitch, indicating that its rating is more stable.
The data on rating reversals shows that S&P and Fitch are far more likely to have rating reversals, especially when longer time horizons are considered. Although rating reversals at a 1 year period can be affected by actual changes in underlying credit risk, the analysis shows clear differences between the three agencies, also at shorter intervals. The results indicates that once Moody’s has decided to execute a specific rating action, they are less likely to execute an opposite adjustment. Again, this directs to a higher rating stability by Moody’s; it seems that they are more conservative when changing their rating, but once a rating adjustment has been executed they are less likely to revise their action.
Against this backdrop, we can argue that Moody’s ratings appear to be most stable as they change their ratings less frequently, are the least subject to large rating downgrades and have the smallest chance of being reversed once a rating action has been executed. The difference between S&P and
Fitch is less clearcut

Conclusion
This working paper assesses which rating agency attains the highest sovereign credit rating quality by assessing their ability to predict sovereign defaults. Rating quality is examined with respect to three dimensions: accuracy, timing and stability. Rating accuracy assesses the discriminatory power of credit ratings, while rating timing addresses which agency responds the quickest to changes in credit risk; both are vital for the informational value of ratings. The stability dimension evaluates whether agencies find the optimal balance between adapting to changes in underlying credit risk and rating user preferences, which is essential for the successful execution of their certification role. Table 11 displays the main findings.
Table 11. SCR scorecard
Quality Dimension

Best scoring Agency

Accuracy

Short-term: S&P
Long-term: Moody’s
S&P
Moody’s

Timing
Stability

As summarized in table 11, the examination of rating accuracy indicated that S&P’s ratings had the highest explanatory power close to a default, while over longer time horizons Moody’s rating accuracy outpaced the other agencies. Another noteworthy trend is that Moody’s managed to maintain a relatively stable accuracy over longer time-horizons, whereas the precision of S&P’s and especially

18

Fitch’ ratings worsened. Our results suggest that Moody’s focusses more on long-term indicators, thereby ruling out business-cycle induced changes in creditworthiness.
Analysis of downgrade profiles and rating stability supports this outcome. Case studies of 13 defaults rated by all agencies (provided in appendix 4) indicate that S&P is most aggressive, and most timely with its ratings, whereas Moody’s ratings tends to be the most stable. To illustrate how static Moody’s ratings are, consider that during 8 out of the 13 defaults Moody’s does not downgrade its sovereign rating to default status; explaining the inaccuracy of their ratings close to a default. In contrast, S&P’s rating actions are more volatile, which improves the timeliness of their ratings but may go at the expense of the explanatory power in the long run. Analysis of initiators and followers during downgrade cycles supports that S&P is most aggressive in its rating actions. The assessment of rating stability fits that conclusion: Moody’s has the most stable ratings, followed by S&P and Fitch.
Eventually, it is not feasible to address one agency with the highest sovereign credit rating quality.
Depending on user preferences, S&P performs best when timing or accuracy proximate to default are most important, while Moody’s outperforms its competitors at stability and long-term accuracy; Fitch seems to strike a different balance between timeliness and stability, never outperforming the other two on the identified aspects of rating quality.

19

References
Afonso, A., Furceri, D., & Gomes, P. (2011). Sovereign credit ratings and financial markets linkages application to European data. [European Central Bank Working Paper Series, 1347]. Retrieved from http://www.ecb.europa.eu/pub/pdf/scpwps/ecbwp1347.pdf
Altman, E.I., Rijken, H.A. (2004). How rating agencies achieve rating stability. Journal of Banking &
Finance. 28(11). pp. 2679-2714. Retrieved from http://ac.elscdn.com.proxy.library.uu.nl/S0378426604001050/1-s2.0-S0378426604001050main.pdf?_tid=6ca636dffad54aee4a5b314775251a3a&acdnat=1338386276_2e4c59d32abc77de f8e094e2b1c99d92 [subscription required].
Andrews, L. J., & Antoine-Clyne, P. (2007). Government of Granada prospectus for $100.0 million 91day treasury bills. Retrieved from http://www.gov.gd/egov/docs/reports/Grenada_Prospectus.pdf Bhatia, A.V. (2002). Sovereign credit rating methodology: An evaluation. [IMF working paper,
02/170]. Retrieved from http://www.uoit.ca/sas/International%20Finance/SovereignCredit.pdf
BIS. (1999). Supervisory lessons to be drawn from the Asian crisis. [Basel Committee on Banking
Supervision Working Papers] Retrieved from http://www.bis.org/publ/bcbs_wp2.pdf
Bissoondoyal-Bheenick, E. (2004). Rating timing differences between the two leading agencies:
Standard and Poor’s and Moody’s. Emerging Markets Review 5. pp 361 – 378. doi:10.1016/j.ememar.2004.03.006 Brooks, R., Faff, R.W., Hillier, D.,& Hillier, J. (2004). The National Market Impact of Sovereign Rating
Changes. Journal of Banking & Finance. 28(1), pp. 233-250. Retrieved from http://ac.elscdn.com.proxy.library.uu.nl/S0378426602004065/1-s2.0-S0378426602004065main.pdf?_tid=c5c65129b195efb05b2f6a9a33f1a68f&acdnat=1338385428_375dc68e9126ad62 cf71b51575e3acaa [subscription required].
Brooks, R., Faff, R., & Hill, P. (2010). Variations in sovereign credit quality assessments across rating agencies. Journal of Banking & Finance. 34(6). pp. 1327 – 1343. Retrieved from http://ac.elscdn.com.proxy.library.uu.nl/S0378426609003252/1-s2.0-S0378426609003252main.pdf?_tid=d7ddb9f28715469e7cc75e5418efa0ff&acdnat=1338387759_c2cf3a7a8ee60753d
2503745bed6258f [subscription required].
Cantor, R., Mann, C. (2003). Measuring The Performance Of Corporate Bond Ratings. [Moody’s special comment]. Retrieved from http://ssrn.com/abstract=996025
Cantor, R., & Mann, C. (2006). Analyzing the Tradeoff Between Ratings Accuracy and Stability.
[Moody’s Investors Service]. Retrieved from http://www.moodys.com/sites/products/DefaultResearch/tradeoff.pdf Damodaran, A. (2010). Into the Abyss: What if nothing is risk free? Retrieved from http://ssrn.com/abstract=1648164 Duff, A., & Einig, S. (2008). Credit ratings quality the perceptions of market participants and other interested parties. The British Accounting Review. 41(3). pp 141–153. Retrieved from http://ac.els-cdn.com.proxy.library.uu.nl/S0890838908001224/1-s2.0-S0890838908001224main.pdf?_tid=2bdac1c4121d64771852b640d48051eb&acdnat=1338388570_d99494be52716e 961aa769c86f1e83e1 [subscription required].

20

Engelmann, B., Hayden, E., & Tasche, D. (2002). Measuring the discriminative power of rating systems. [Deutche Bundesbank, discussion paper]. Retrieved from http://www.econstor.eu/dspace/bitstream/10419/19726/1/200301dkp_b.pdf Engelmann, B., Hayden, E., & Tasche, D. (2003). Testing rating accuracy. Retrieved from http://www.netegrate.com/index_files/Research%20Library/Catalogue/Quantitative%20Analysis /Model%20Validation/Testing%20Rating%20Accuracy%20with%20ROC%20and%20CAP%20(En gelmann).pdf Enderlein, H., Trebesch C., & Von Daniels, L. (2012). Sovereign disputes: A database on government coerciveness during debt crises. Journal of International Money and Finance. 31(2). pp 250 –
266. DOI: http://dx.doi.org/10.1016/j.bbr.2011.03.031
Evans-Pritchard, A. (2008). Ecuador default: Fears grow that others will follow. Retrieved April 27,
2012 from http://www.telegraph.co.uk/finance/globalbusiness/3759173/Ecuador-default-Fearsgrow-that-others-will-follow.html
Faff, R., Hill, P. (2007). Do credit watch procedures affect the information content of sovereign credit rating changes. Retrieved from http://www.fma.org/Prague/Papers/FMAEuro_2008_Submission_HillnFaff.pdf
Fitch. (2011). Fitch Ratings Sovereign 2011 Transition and Default Study. Retrieved from: http://www.fitchratings.com/creditdesk/reports/report_frame.cfm?rpt_id=670071 [subscription required]. Friedman, T.L. (1996, February 13). The news Hour with Jim Lehrer: Interview with Thomas L.
Friedman. [Television broadcast transcript]. Retrieved from http://www.pbs.org/newshour/gergen/friedman.html Gande, A., & Parsley, D. (2005). News Spillovers in the Sovereign Debt Market. Journal of Financial
Economics. 75(3), pp. 691-734. Retrieved from http://ac.elscdn.com.proxy.library.uu.nl/S0304405X04001576/1-s2.0-S0304405X04001576main.pdf?_tid=03372582898fe21a5c33546da468f9f4&acdnat=1338384954_d4b12088709f67e5
4b85ed1c9ddc84b4 [subscription required].
Güttler, A., & Krämer, W. (2008). On comparing the accuracy of default predictions in the rating industry. Empirical Economics. 34(2). pp 343-356. Retrieved from http://www.springerlink.com.proxy.library.uu.nl/content/m94v805j7r772373/fulltext.pdf Hollaar, A.D. (2011). Effects of changes in sovereign credit ratings on investors’ behavior.
[Unpublished manuscript].
House of Lords. (2011). Sovereign Credit Ratings: Shooting the Messenger? Oral evidence and associated written evidence. Retrieved from http://www.parliament.uk/documents/lordscommittees/eu-sub-com-a/Creditrating/CRA%20Corrected%20and%20Evidence%20Final.pdf
IMF. (2006). Cross-Country Expericence with Restructuring of Sovereign Debt and Restoring Debt
Sustainability. Retrieved from http://www.imf.org/external/np/pp/eng/2006/082906.pdf
IMF. (2010). Global Financial Stability Report - Sovereigns, Funding and Systemic Liquidity. [World economic and financial surveys]. Retrieved from http://www.imf.org/external/pubs/ft/gfsr/2010/02/pdf/text.pdf 21

Irwin, J. R., & Irwin, C. T. (2012) Appraising Credit Ratings: Does the CAP Fit Better than the ROC?.
[IMF Working paper 12/122]. Retrieved from http://www.imf.org/external/pubs/ft/wp/2012/wp12122.pdf Katz, J., Salinas, E., & Stephanou, C. (2009). Credit rating agencies - No easy regulatory solutions.
World Bank note. Retrieved from http://rru.worldbank.org/PublicPolicyJournal
Keenan, S.C., Sobehart, J.R., & Stein, R.M. (2000). Rating methodology, Benchmarking quantitative default risk models: a validation methodology. Retrieved from Moody’s Investors Service.
[subscription required].
Kiff, J., Nowak, S., & Schumacher, L. (2012). Are rating agencies powerful? An investigation into the impact and accuracy of sovereign ratings. [IMF Working Paper, 12/23]. Retrieved from http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1997736 Laeven, L. & Valencia, F. (2012). Systemic Banking Crises Database: An Update. Retrieved from: http://ssrn.com/abstract=2096234 Mann, H., and D. Whitney. (1947). On a Test of Whether One of Two Random Variables is
Stochastically Larger than the Other. Annals of Mathematical Statistics 18. pp 50–60. Retrieved from http://webspace.ship.edu/pgmarr/Geo441/Readings/Mann%20and%20Whitney%201947%20%20On%20a%20Test%20of%20Whether%20one%20of%20Two%20Random%20Variables%20 is%20Stochastically%20Larger%20than%20the%20Other.pdf Moody’s. (2008). Sovereign Defaults and Interference: Perspectives on Government Risks. Retrieved from: http://www.moodys.com/sites/products/DefaultResearch/2007100000522782.pdf
Moody’s. (2011a). Sovereign default and recovery rates, 1983-2010. Retrieved from http://www.naic.org/documents/committees_e_capad_vos_c1_factor_review_sg_related_docs_ moodys_sovereign_default.pdf
Moody’s. (2011b). Glossary of Moody’s Ratings Performance Metrics. Retrieved from Moody’s investor service [subscription required].
Moody’s Investor Service. (2012). Credit opinion: Greece, Government of. Retrieved from http://www.moodys.com/research/Greece-Government-of-Credit-Opinion--COP_348330 [subscription required].
Reinhart, C.M., & Rogoff., K.S. (2009). This Time Is Different: Eight Centuries of Financial Folly.
Princeton: Princeton University Press.
Reinhart, C.M., & Rogoff., K.S. (2011). From Financial Crash to Debt Crisis. American Economic
Review, 101(5). pp 1676-1706.
Salmon, F. (2009, May 29). Lessons from Ecuador’s bond default. Retrieved April 27, 2012 from http://blogs.reuters.com/felix-salmon/2009/05/29/lessons-from-ecuadors-bond-default/ S&P. (2011). Sovereign defaults and rating transition data, 2010 update. Retrieved from https://www.globalcreditportal.com/ratingsdirect/RatingsDirect_Commentary_961289_05_30_2 012_07_03_48.articlePDF?rand=emPZbEikXV&id=961289&sourceId=&type=&outputType=&fro m=SR&prvReq=&pager.offset=&SIMPLE_SEARCH_TYPE=&CONID=&entl=&requestFrom=getPD F&articleType=enCommentary [subscription required].

22

Sy, A.N.R. (2003). Rating the rating agencies: Anticipating currency crisis or debt crisis? [IMF
Working Paper, 03/122]. Retrieved from http://www.imf.org/external/pubs/ft/wp/2003/wp03122.pdf Tasche, D. (2006). Validation of internal rating systems and PD estimates. Deutsche Bundesbank.
Retrieved from http://arxiv.org/abs/physics/0606071v1
Theunissen, G. (2009). Seychelles Offers New Debt for $311 Million Defaults (Update 1). Retrieved from http://www.bloomberg.com/apps/news?pid=newsarchive&sid=aMMQx5wLyRpY

23

Appendix 1

Discussion

In this appendix we would like to reflect on the research methods and implications of the study. First of all, we would like to stress that the sample size – the limited amount of sovereign defaults – is a serious limitation to the empirical assessment of rating quality, which decreases the explanatory power of the accuracy and timing assessment. The assessment of timing is based on subjective observations. Moreover, we feel the third measure –‘the first to downgrade’- adds little to the understanding of rating timeliness, because it is hard to determine from which moment prior to default the first downgrade counts. In addition, no significance test has been conducted on the outcomes of the empirical assessments. Mann & Whitney (1947) offer a mathematical approach to assess significance and Engelmann et al. (2002) employ that method to assess the discriminatory power of rating systems. However, Engelmann, Hayden & Tasche (2003) find that this significance test might not be well applicable for small sample sizes (fewer than 50 defaults) and should be interpreted with care. Therefore, given that no significance test is conjured, the outcomes of this study should be interpreted as indications rather than endorsed results.
Second, the assessment of rating accuracy is based on CAP-curves and AR-ratios. These methods were selected because they are the common approach in finance to assess discriminatory power of rating systems. However, other procedures might add to the understanding of rating accuracy. Irwin and Irwin (2012) prefer the “ROC-approach” over CAP-curves because it may reflect more of the underlying rating process. Moreover, CAP-curves weight Type I and Type II errors equally in their valuation of rating accuracy, while the implications of Type I errors are likely to be more austere.
Therefore, adjusting the relative weighting of Type I and Type II errors could improve the examination of rating quality.
Lastly, the current analysis solely focuses on sovereign credit ratings. Credit rating agencies also work with rating announcements to express their views on credit risk. Therefore, a desirable expansion of the research would be to include rating watches and outlooks, which could increase the understanding of sovereign credit rating quality.

24

Appendix 2

Rating timeliness

Sovereign default definitions governing the Reinhart and Rogoff database

Foreign currency default
Definition: ”A sovereign default is defined as the failure to meet a principal or interest payment on the due date (or within the specified grace period). The episodes also include instances where rescheduled debt is ultimately extinguished in terms less favorable than original obligation.”
Comment: “While the time of default is accurately classified as a crisis year there are a large number of cases where the final resolution with the creditors (if it ever did take place) seems interminable.”
Domestic currency default
Definition: “The definition given above for external debt applies. In addition, domestic debt crisis have involved the freezing of bank deposits and or forcible conversions of such deposits from dollars to local currency.
Comment: “There is at best some partial documentation of recent defaults on domestic debt provides by Standard and Poors. Historically, it is very difficult to date these episodes and in many cases (like banking crisis) it is impossible to ascertain the date of the final resolution.”

Source: Reinhart & Rogoff (2009).

25

Appendix 3

Additional sources to Reinhart & Rogoff database

This table was constructed because the database by Reinhart and Rogoff states annual data, while our research demands more specific default data. Therefore, we consulted multiple sources to obtain monthly data; an subjective process because proper data is lacking. Therefore, we only included defaults which we consider were adequately documented. Still, we have been conservative with the
(monthly) interpretation defaults when considering data from rating agencies.

Country

Default Type

Date

Source

Argentina
Belize
Cameroon
Dominican
Republic
Ecuador

Domestic & Foreign
Foreign
Domestic
Foreign

December 2001
December 2006
July 2004
April 2005

1,4,5
2,5,6
5
1,2,5

& Foreign

August 1999
December 2008
March 2012
December 2004
January 1999
July 2002
January 2010
June 2002
January 1999
July 2003
August 1998
July 2008
January 2000
April 2003
July 1998
January 2005

6,9,10

Greece
Grenada
Indonesia

Domestic
Foreign
Domestic
Foreign
Foreign
Foreign
Domestic
Foreign
Foreign
Domestic
Domestic
Foreign
Foreign
Domestic
Domestic
Foreign

Jamaica
Moldova
Pakistan
Paraguay
Russia
Seychelles
Ukraine
Uruguay
Venezuela

& Foreign

& Foreign

& Foreign
& Foreign

& Foreign

11, 14
2,5,7,12
1
6,7,8
4,5,6,7,8
4,5
1,4
1,5
5,7
2,5
1,4,5,6,7,8
1,6,7

Sources
1] Reinhart & Rogoff (2011)

7] Standard &Poor‘s (2011)

2] Einderlein, Trebesch, & Von Daniels
(2012)

8]Fitch (2011)

3] Damadoran (2010)

9] Evans-Pritchard (2008)
10] Salmon (2009)

4] IMF (2006)

11] Moody’s Investor Service (2012)

5] Moody’s (2008)

12] Andrews, & Antoine-Clyne (2007)

6] Moody’s (2011)

13] Theunissen (2009)
14] Laeven and Valencia (2012)

26

Appendix 4

Rating timeliness

The following section shows the downgrade profiles of thirteen domestic and foreign currency defaults. These are the only defaults that were rated by all three agencies at time of default. The downgrade profiles show rating developments prior to default, providing an indication of rating timeliness. 27

Indonesia
Indonesia experienced two foreign currency defaults; the first in early 1999 and the second in mid2002. The downgrade profile shows that Fitch issued its first rating just over a year before the default. In terms of rating performance, S&P stated the highest rating prior to default, leading to the largest cliff-effect early in 98’. However, S&P also is the only rating agency to acknowledge the two sovereign defaults, while Fitch and Moody’s maintain their rating at B-/B3. In terms of rating timing, all three agencies start downgrading heavily at the same time, so no real distinction can be made.

27

28

Russia
Russia defaulted on its local and foreign currency bonds in 1998 and 1999. However, no domestic currency ratings were available at that time, so their downgrade profile only consists of foreign ratings. Fitch had the highest rating prior to default, and was last to downgrade. Moody’s was early with its downgrade, but had the highest rating at the time of default at B3; indicating that in their opinion it was a “high-risk obligation”, but nowhere near default. Overall, S&P had the best rating timeliness as it gave the lowest initial credit rating and was the only to identify the default.

28

29

29

Argentina
Argentina’s notorious default is, after Greece in 2012, the second largest in sovereign history. As expected, the external default is recognized by all rating agencies. Moody’s has the lowest rating prior to the downgrade cycle, while S&P is the first to start downgrading and to reach default class. In general, Moody’s performs best as it states the lowest credit rating through almost the entire period.
The domestic currency rating shows an entirely different image. Fitch and Moody’s recognize the default; while S&P does not, and actually upgrades Argentina a year before the default occurred.
Another important observation is that Fitch’ downgrades are of relatively large scale compared to
Moody’s rating actions, caused by Fitch’ higher initial rating. Overall, Moody’s performs best at both the foreign and domestic currency default because it gives the lowest initial rating and has the most smooth downgrade profile.

30

30

31
Uruguay
Moody’s clearly missed Uruguay’s foreign currency default in 2003, maintaining its foreign rating at
B3. Fitch and S&P follow similar downgrade paths and identify the default around the same time. The default on domestic currency denoted debt is not recognized by either of the agencies, although Fitch and S&P do downgrade their rating significantly around the time of default. Overall, Moody’s performed worst by not identifying the defaults, while little distinction can be made between S&P and
Fitch, as the timing and size of their rating actions is quite similar, although Fitch has a large cliffeffect around the time of default.

31

32
Venezuela
The downgrade profile of Venezuela’s sovereign default shows some anomalies. The agencies started upgrading the rating a few years prior to default, and S&P is the only agency to downgrade its rating once the default occurred – giving rise to a large cliff-effect. Because of its abnormality it is hard to draw any timing related conclusions from this downgrade profile, besides that S&P does better at recognizing the default once it occurred.

32

33

Dominican Republic
The Dominican Republic defaulted in April 2005. The downgrade profile clearly indicates that S&P outperforms Moody’s and Fitch: They were first to downgrade and were the first to recognize the default. Moody’s performs worst, as they do not acknowledge the default. This downgrade profile also highlights a difference between Moody’s and the other two agencies: their ratings tend to be much more stable. During the 5-year time period they only adjust their ratings three times, compared to 6 adjustments by S&P and Fitch.

33

34

Ecuador
The Ecuadorian default profile shows a cliff-effect of all ratings at the time of default (i.e. all were caught by surprise). However, Fitch seems to perform best as they do not upgrade their rating within a year prior to default, in contrast with S&P and Moody’s. There is no indication of divergence in rating timeliness as all agencies downgrade heavily at the same time.

34

35

35

Jamaica
The downgrade profiles of Jamaica’s foreign and domestic currency defaults are very similar.
Both indicate that S&P performed best, as they have the lowest initial rating and remain the lowest rating over the entire time period.
However, Fitch is first to downgrade. Initially,
Fitch appears to move together with S&P; although its rating remains higher for most of the period. However, the inserted figure on the right shows a rating error by Fitch: they upgrade their credit rating from C to CCC the day before they recognize the default and set their rating to
RD. Again, Moody’s has a more stable rating but does not recognize the default.

36

36

37

Greece
Similar to Jamaica’s downgrade profiles, the Greek Foreign and Domestic profiles look very similar.
S&P is the first to downgrade and reaches the investment/speculative grade threshold first, but
Moody’s leads the downgrade cycle later on and is the first the rating to default. Fitch performs worst as they tend to lag behind and have the highest rating just before the default occurred.

37

Appendix 5

Rating timeliness II

Detailed results from qualitative assessment rating downgrades

Best timeliness

-7

S&P

Fitch, Moody’s

Moody’s

S&P

Fitch

S&P

S&P

Argentina (Lcl 9) Moody’s

S&P

Fitch

Fitch

Uruguay (Frn) S&P, Fitch

Moody’s

S&P

S&P, Fitch

Uruguay (Lcl) -

-

Fitch

-

Venezuela -

-

-

S&P

Dominican S&P
Republic
Ecuador Fitch

Moody’s

S&P

S&P

Moody’s, S&P

S&P

S&P

Jamaica (Frn) S&P

Moody’s

Fitch

S&P

Jamaica (Lcl) S&P

Moody’s

Fitch

S&P

Greece (Frn) S&P, Moody’s

Fitch

S&P

Moody’s

Greece (Lcl) S&P, Moody’s

9

Fitch, Moody’s

Argentina (Frn 8) Moody’s

8

First to recognize the default Russia S&P

7

First to downgrade Indonesia S&P

38

Worst timeliness

Fitch

S&P

Moody’s

Empty cells indicate that there was no difference in performance amongst CRAs
Frn: foreign currency default
Lcl: local currency default

38