Economics Research

UniCredit Global Themes Series Economics, FI/FX & Commodities Research Credit Research Equity Research Cross Asset Research

“

No. 21 26 March 2014

The Damaging Bias of Sovereign Ratings

”

– An eye-watering USD 50 trillion in outstanding sovereign debt is guided by sovereign credit ratings. And, since these set the benchmark for all other credit ratings, the implications of sovereign ratings extend much further. – Just three firms (Moody’s, S&P and Fitch) dominate the market. They set ratings based on a combination of measurable fundamentals (the objective component) and the judgment of their in-house ratings committees (the subjective component). – In this paper, we decompose the ratings into their objective and subjective components. We find that, whilst the objective component of ratings is a good predictor of future sovereign default, the subjective component of ratings has no predictive power on average for defaults one or more years ahead, but adds large distortions to individual ratings. – The implications extend far and wide. History is littered with countries being over- and under-rated by the ratings agencies, with – at times – dramatic consequences. The biggest casualty was the Eurozone periphery, which was downgraded far too heavily during the 2009-11 sovereign debt crisis as the rating committees repeatedly overruled the signals coming from fundamentals and to an unprecedented extent. – We show that the Eurozone’s periphery is on average rated almost five notches below what their fundamentals signal, while the Fragile Five are on average rated almost two notches above their fundamental (or objective) signals. – In light of our findings, we suggest that credit rating agencies should be stripped of their regulatory powers and these transferred to an international body. Failing that, the ratings agencies should be forced to substantially increase transparency, including publishing a separate breakdown of the objective and subjective components of ratings, the minutes of the rating committees, and the voting records. Authors: Daniel Vernazza, Ph.D. (UniCredit Bank London) Erik F. Nielsen (UniCredit Bank London) Dr. Vasileios Gkionakis (UniCredit Bank London)

Editor: Dr. Andreas Rees (UniCredit Bank)

26 March 2014

Economics & FI/FX Research

Global Themes Series

Contents

UniCredit Research

3

Executive summary

4

I. Introduction

6

II. The convoluted methodology of the CRAs

7

III. Quantifying subjectivity

11

IV. Is subjectivity useful in predicting default?

16

V. A look back at the EMU debt crisis

18

VI. Current misalignments

19

VII. Concluding remarks

20

References

21

APPENDICES

21

A. Data and variable construction

22

B. The precision of our estimates

23

C. The results for S&P

24

D. The results for Fitch

25

E. Current misalignment, all countries

27

Global Themes Series List

page 2

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

Executive summary

UniCredit Research

■

Investors the world over rely on sovereign credit ratings - either because they have to for regulatory reasons, or because they want to. More than USD 50 trillion in outstanding sovereign debt is guided by the credit rating agencies' assessments, of which just three firms, Moody's, S&P and Fitch - the so-called Big Three - have a market share of about 95%. And as sovereign ratings provide the benchmark for all other ratings, the Big Three's assessment of sovereign creditworthiness has implications well beyond the direct impact.

■

The Big Three's sovereign ratings are based on a set of relatively standard macroeconomic and political variables (we call them the "objective" component) as well as their ratings committees' non-quantifiable assessments (we call those the "subjective" component.) While some progress has been made with respect to transparency, rating decisions remain amazingly opaque for agencies with regulatory power.

■

In this paper we reveal the "objective" component of sovereign ratings for more than one hundred counties over more than ten years, which makes it possible to deduct the "subjective" component as the difference between the actual ratings and the objective component. This illustrates the degree to which subjective assessments by the rating committees overrule the fundamentals.

■

We find that on average the "subjective" component (the rating committees' overruling of the macro signals) adds no value to the predictability of default until less than one year before the actual event - but what is on average noise hides a multitude of sins, for it means history is littered with sovereigns having been over- and under-rated by the "subjective" component overruling the "objective" component from the macro fundamentals.

■

Since such misvalued ratings guide trillions of dollars of assets, particularly when changes are made, the distortionary (subjective) tweaking of the "objective" signals typically have significant impacts on capital flows and asset prices.

■

In the extreme, during the 2009-11 period, all three rating agencies downgraded the Eurozone periphery in several steps as their ratings committees repeatedly overruled the "objective" macro signals with severely negative "subjective" assessments - only now to slowly begin to repair the damage, which nearly turned into a self-fulfilling disaster prophecy. Ironically, during the same period, the same ratings committees overruled their "objective" macro signals for the "Fragile Five" emerging markets, but in a positive direction. Only now, as their fundamental weaknesses have become obvious, have the ratings agencies begun to reverse their mistake.

■

In light of our findings, we suggest that credit rating agencies should be stripped of their regulatory powers and for these powers to be transferred to an international body. After all, anything with a regulatory role should be carried out by agents of the state, or in this case all states. Failing that (and it would be a fail) the ratings agencies should be forced to substantially increase transparency, including publishing a separate breakdown of the objective and subjective components of ratings, the minutes of the rating committees, and the voting records.

page 3

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

“Benchmarking rating performance against actual default experience constitutes the strongest possible test of a rating system.” Moody’s (2013, p.29), Rating Methodology: Sovereign Bond Ratings

I. Introduction It is difficult to overstate the importance of sovereign credit ratings, which are meant to indicate the probability that a sovereign state will fail or refuse to repay its debt in full or on time 1 (commonly known as ‘default’). Investors rely on them - either because they want to or they have to, or some combination of both. We estimate that a tad more than USD50 trillion of global savings invested in sovereign bonds may be guided directly or indirectly by the views of the 2 three largest rating agencies. Of course, the implications of sovereign ratings are much wider than this, for they set the benchmark for all other credit ratings (corporates and municipalities). It is, therefore, crucial that the behaviour of Credit Rating Agencies (CRAs) – the firms that set credit ratings – is being scrutinized on a continual basis. The market for credit ratings is far from competitive. Just three CRAs - Moody’s, Standard and Poors, and Fitch - account for roughly ninety-five percent of the market. Together, the so-called “Big Three” have a special 3 status, written into law in both the U.S. and Europe. Regulations require some institutional investors to hold only investment-grade debt (the top-tier rating). Consequently, rating changes across the investment-grade boundary can have a big effect on yields as they 4 change the number of potential buyers. There is also a good reason why investors would want to rely on ratings or other external expertise – irrespective of the fact that the law says they may have to. It is, after all, extremely costly for the average investor to undertake the analysis necessary to rate the debt of sovereign states. Ratings are relative – that is, they are ordinal rankings not cardinal – which means that in order to rate one sovereign nation one must consider the fundamentals of all nations in the world – no small task. There already exists a large literature that critically assesses the behaviour of CRAs. These papers have mostly looked at the informational content of ratings and can broadly be split into two branches. The first branch examines whether a ratings action (ratings changes, reviews, watches and outlooks) has an effect on yields. A number of papers find that spreads actually lead ratings actions (see, for example, Reisen and Von Maltzan (1999), Kiff et al (2012), and Gartner and Griesbach (2012)) – suggesting that the CRAs are merely market followers, which means that they – via their legislated and generally guiding role for fund managers – make 5 matters worse when markets get nervous about a given sovereign credit. The other branch assesses whether readily-available fundamentals can explain the variation in ratings – these papers tend to find that much of the variation in sovereign ratings can be explained by a relatively small number of variables (see, among others, Cantor and Packer (1996) and Ferri et al. (1999)). However, Ferri et al (1999) find that the CRAs aggravated the East Asian Crisis by downgrading the affected countries more than the worsening of their fundamentals warranted.

1

More specifically, a credit (or default) event is the failure of the debtor to abide by the full terms of the contract with the creditor. This occurs when the terms of a contract are re-negotiated to lower the net present value of the obligation, even if creditors do so “voluntarily.” In contrast, a reduction in the real value of the NPV of a contract via higher inflation, even if imposed on the creditor via capital controls, is not considered a credit event by the ratings agencies. 2

The total outstanding government debt is around USD53tn (see The Economist global debt clock: http://www.economist.com/content/global_debt_clock).

3

The regulatory power granted to the “Big Three” CRAs commenced in 1975 when the U.S. Securities and Exchange Commission chose their ratings to determine the minimum capital requirements for financial firms to trade certain debt securities, depending on their riskiness. Since 2001, the capital requirements of banks have also been tied to the credit rating of the securities they hold. Under Basel II, regulators can permit banks to use credit ratings from approved CRAs to calculate their net capital reserve requirements. For example, under Basel II, an Aaa/AAA rated security requires capital of just 0.6%; this rises to 4.8% for Baa2/BBB, 34% for Ba2/BB and 52% for Ba3/BB-. 4

Kiff et al. (2012) find that ratings actions that cross the investment-grade boundary have an economically and statistically significant effect on the CDS spreads of sovereigns.

5

There is evidence to suggest that sovereign rating downgrades have a negative impact on the national stock market (see, for example, Brooks et al. (2001) and Kaminsky and Schmukler (2001)).

UniCredit Research

page 4

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

In response to these criticisms, the CRAs typically appeal to two defenses. First, they argue that their ratings are measures of the probability of default, not the market’s perception of the probability of default – and so they do not really care whether yields lead ratings or not. The opening quote from Moody’s (2013) emphasizes this point. Second, they point out that their ratings are based on a mixture of quantitative analysis and committee judgments, and they often use the latter to push-back on attempts by outsiders that fail to ‘explain’ their ratings. And they give reasons to support the use of judgments, which include, among others, that (1) some factors are impossible to quantify and even those that can be quantified are not available for every country at every point in time – some of which is confidential information; (2) there are idiosyncratic country-specific factors; and (3) judgments incorporate forwardlooking expectations and risks. The purpose of this paper is twofold. First, intrigued by the idea that these (self-appointed) committees in each of the CRAs might be able to improve the signal (in terms of default probability) coming from the hard data, we compute a measure of the subjective (or judgment) component of ratings. We find that, in some cases, the committees overrule the signal from the hard data to a considerable extent. Second, and crucially, we test whether ratings are useful in predicting actual default events. Our results are profound. Whilst the objective component (i.e. the signal from the hard data) of ratings is a good predictor of actual default, the subjective component adds no value at all for default one or more years ahead! Importantly, since this is the average outcome of our analysis of more than one hundred countries over more than ten years, it implies that beneath this average zero-value-added “subjective component” lies multiple sins on both sides. In other words, history is littered with countries being over- or under-rated by the CRAs, with – at times – dramatic consequences. For example, between 2009 and 2012, driven mostly by the “subjective component” of the CRAs, the Eurozone periphery suffered a series of downgrades, which contributed to Northern European banks reducing their exposure to these countries by more than 30%. During the same period, in spite of wider current account deficits, the so-called “Fragile Five” (Brazil, India, Indonesia, South Africa and Turkey) benefited from a series of upgrades, which contributed to the same groups of banks increasing their exposure to these five countries by more than 30%. Meanwhile, in 2013, the US seemed to get to the edge of a default (sometimes called a possible “technical default” because it would have been triggered by a poorly functioning governance structure, curiously implying it would have been less serious 6 than a “non-technical” default), yet with no ratings action. We proceed as follows. Section II briefly describes and critically assesses the methodology that the “Big Three” CRAs claim to follow to rate sovereigns. Section III uses an econometric model to ‘reveal’ the implicit model of the CRAs. We use this to decompose the rating into an ‘objective’ component (based on measurable fundamentals) and a ‘subjective’ (or residual) component. Section IV tests whether ratings have predictive power for actual sovereign default. In Section V we detail the evolution of the objective and subjective components of ratings surrounding the 2011 EMU Debt Crisis, a period when the Eurozone periphery was downgraded abruptly. Section VI presents the current misalignment of ratings for selected countries. Finally section VII concludes.

6

Fitch did put the U.S. on negative watch on 15 October 2013, but did not actually downgrade the U.S..

UniCredit Research

page 5

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

II. The convoluted methodology of the CRAs The CRAs point out that their ratings are decided based on a combination of measurable (or objective) and judgment (or subjective) factors. In the past there was much opaqueness surrounding their rating methodologies, and staff were reportedly sworn to secrecy. Recently the CRA’s have made some small movements to become more transparent - in response to the weight of criticism parked at their door in the wake of the financial crisis. In the case of Moody’s latest methodology (see Moody’s, 2013), they claim that by following their ‘scorecard’ one can get to within a three-notch range of the actual sovereign debt rating in most cases. We looked at the latest methodologies from the Big 3 CRAs and came to the following conclusions:

■

First, the CRA’s have an unenviable job. They have to predict the probability of sovereign default when there are few past incidents of actual default to go on. This makes it very difficult to identify the drivers of sovereign default, not to mention the weights and functional form of different factors.

■

In Moody’s case, the result is an incredibly convoluted method. They employ 31 variables (so-called sub-factors) to determine the ranking. The value for each sub-factor is assigned a rank on a 15-point scale from ‘Very High plus’ to ‘Very Low minus’. There appears to be little or no empirical foundation for the cut-offs between these ranks. These sub-factors are then aggregated using a set of ad-hoc weights (see Moody’s, 2013). The final rating is based on the quantitative assessment and the judgment of their rating committees.

■

S&P's methodology (see Standard and Poor’s, 2012) appears similar to Moody's although less transparent. More specifically, each sovereign is given a score from one (the strongest) to six on each of the following five criteria: political, economic, external, fiscal and monetary. Although there is a list of variables taken into consideration, many of them are not defined explicitly and there is little or no discussion of the weights used. Indeed, the agency says that, “rather than providing a strictly formulaic assessment, Standard & Poor's factors into its ratings the perceptions and insights of its analysts”.

■

Fitch employs a number of variables in its sovereign ratings model (see Fitch (2012)). These variables are listed under four headings: macroeconomic, public finances, external finances, and structural. The weights are estimated using regression analysis. However, just like Moody’s and S&P, Fitch says that “the actual rating determined by the sovereign rating committee can and does differ from that implied by the rating model". In other words, subjectivity plays an important role.

Whilst the CRAs face a difficult task, this should not stop them from evaluating the forecast performance of their model. To their credit, they do present some evidence of the historical performance of their ratings in predicting actual default. For example, Moody’s (2013, p.29) declares that since 1983 their ratings “accurately rank-order sovereign default risk” – by which they mean that higher-rated sovereigns are less likely to default than lower-rated sovereigns, over both one-year and five-year horizons, hardly a sufficient criteria for success. More importantly, this blurs the predictive power of the ‘objective’ and ‘subjective’ components of ratings. Indeed, later we will show that while the ‘objective’ component of ratings is informative of future default, the ‘subjective’ component of ratings has no predictive power at all for default one or more years ahead and, hence, is distortionary at the individual country level.

UniCredit Research

page 6

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

III. Quantifying subjectivity There is little dispute about which variables should be included in an analysis of a country’s creditworthiness, namely GDP, per capita GDP, growth, public and external debt, among others and, broadly speaking, which are good and bad levels for these variables. This is relatively straightforward, so the value-added by the CRAs would supposedly be the ‘subjective component’, i.e. the assessment by the rating committees that modify the signals from the “objective component”. In this section we back-out this ‘subjective’ component of ratings. We simply estimate the component of ratings that depends on measurable ‘fundamentals’ of creditworthiness – the ‘objective’ component of ratings – and compute the ‘subjective’ component as the residual. To this end, we pooled the data on the end-of-year sovereign ratings of the Big Three CRAs as well as a host of ‘fundamentals’ of creditworthiness. The number of sovereigns rated varies over time; for example, in the case of Moody’s (the oldest CRA) the number of rated sovereigns has gradually expanded from 13 nations over the period 1949-1985 to 121 nations last year. In all, we have data on 1,569 ratings from Moody’s, 1,719 ratings from S&P, and 1,339 ratings from Fitch. Initially we started out with fifteen fundamental variables, but based on goodness-of-fit tests we were able to whittle this down to the following parsimonious set of ten: Nominal GDP, GDP per capita, average real GDP growth, public debt, current account, external debt, an indicator variable for whether the state has a past default, an indicator variable for whether the state is 7 an advanced country, government effectiveness, and rule of law. These are largely selfexplanatory: nominal GDP, GDP per capita and GDP growth affect the ability of the state to finance government borrowing. Of course, the larger the public debt the harder it is to service it, while the external debt measures the indebtedness of the whole nation and current account deficits have to be financed by borrowing from abroad. Finally, credit events are often associated with poor governance, which in our model is captured by the World Bank’s indices of government effectiveness and rule of law. Not only is the rationale for these fundamentals grounded in economic theory, they too are 8 variables the CRAs themselves purport to use, among others, to rate sovereign debt. A detailed description of the variables used, including their definition, units, and source is given in Table A1 in Appendix A. In the interests of space, hereafter we will focus on the results for Moody’s; the estimates for S&P and Fitch can be found in Appendices C and D, respectively. The results for the Big Three are not significantly different, which follows from the fact that their ratings are correlated to an extremely high degree (see Table 1). TABLE 1: PAIRWISE CORRELATIONS FOR THE RATINGS OF THE “BIG 3” Moody’s

S&P

Moody’s

1.00

S&P

0.97

1.00

Fitch

0.97

0.98

Fitch

1.00

Notes: 1,214 observations. The correlation is for the numeric conversion of the alphanumeric ratings given in Table A2. Source: Bloomberg, UniCredit Research 7

The five variables that we dropped are average inflation, inflation volatility, the budget balance, the public debt trend, and the World Bank’s control of corruption index. These were statistically insignificant for explaining the sovereign ratings of the Big Three. 8

Moody’s claims to use some 31 variables in their ‘scorecard’ methodology. It is important to note that our parsimonious subset of these variables will to some extent proxy for some of the ‘omitted’ variables. For example, the correlation between government debt-to-GDP (included) and government debt-to-revenue (omitted) is very high at 0.87, as is the correlation between the World Bank’s rule of law index (included) and the World Economic Forum’s competitiveness index (omitted) at 0.87.

UniCredit Research

page 7

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

Table 2 shows the means of the measurable fundamentals by Moody’s rating category. Casual observation reveals that high ratings tend to be associated with advanced countries with high nominal GDP, high GDP per capita, a low external deficit, with no history of past default, high government effectiveness and a strong rule of law. For example, of those countries rated either prime or high grade, just one percent have defaulted since 1960 and 98 percent are advanced countries. From the table the relationship between ratings and either GDP growth, public debt or external debt is not clear cut. This likely reflects the fact that emerging markets tend to grow faster than advanced countries as they catch-up, while due to their generally lower creditworthiness emerging markets are constrained to hold lower debt than their more developed peers. TABLE 2: AVERAGES BY RATING CATEGORY Prime/High grade (Aaa-Aa3)

Upper medium grade (A1-A3)

Lower medium grade (Baa1-Baa3)

Speculative (Ba1-Ba3)

Highly speculative (B1-B3)

Extremely speculative (Caa1 and below)

Nominal GDP

2.72

0.38

0.19

0.17

0.15

0.16

GDP per capita

36.5

18.2

12.0

7.50

6.83

5.75

GDP growth

2.80

5.03

4.40

4.36

3.91

2.53

Public debt

60.8

40.9

40.2

49.1

63.2

68.0

Current account

1.81

-2.13

-1.44

-2.54

-3.31

-3.30

External debt

916.8

625.8

1,340.9

361.6

176.3

130.9

Past default

0.01

0.15

0.46

0.67

0.63

0.82

Advanced country

0.98

0.47

0.08

0.02

0.01

0.02

Government

1.72

0.80

0.33

-0.14

-0.49

-0.69

Law

1.61 255

0.64 124

0.09 168

-0.33 210

-0.66 193

-0.81 49

No. of observations

Notes: The table displays the means for the ten fundamental variables listed in the first column broken down by Moody’s broad rating category. Past default is an indicator variable that takes the value one if the country has defaulted on its debt in the past (since 1960) and zero otherwise. Advanced country is an indicator variable that takes the value one if the country is deemed to be an advanced country by the IMF and zero otherwise. See Table A1 in Appendix A for the definition and units of all the variables used. The sample is restricted to that of column (3) in Table 3. Source: Bloomberg; IMF; World Bank; BIS; UniCredit Research

To proceed with formal estimation of the reaction functions of the Big Three, we must convert their alphanumeric ratings into numeric ones. We use a linear mapping where the higher the rating the higher the number assigned. More specifically, the top rating (Aaa/AAA) is assigned the value 24, the next rating below (Aa1/AA+) is assigned the value 23, and so on. See Table A2 in Appendix A for the details of the conversion Formal statistical estimation (see Box 1) reveals that Moody’s attaches the following weights to our fundamentals. A USD 10,000 rise in GDP per capita adds 1.5 notches to the credit rating; 5pp higher growth adds 0.5 notches, a 50pp rise in the debt-to-GDP ratio subtracts 2 notches; a rise of 5pp in the current-account-to-GDP ratio subtracts 0.25 notches; a 100pp rise in external debt subtracts 0.15 notches; a past default subtracts 1.75 notches; being deemed an advanced country adds 3.23 notches; and government effectiveness and rule of 9 law are positively associated with ratings. A USD 10tn rise in nominal GDP adds 1.3 notches to the credit rating, although the effect is not statistically significant.

9

These marginal effects are from the estimation in column (3) of Table 3.

UniCredit Research

page 8

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

BOX 1: Estimating the ‘objective’ and ‘subjective’ components of ratings We estimate the following equation:

10

rating it = β' x it + α i + τ t + ε it

i = 1,..., N ; t = 1,..., Ti

(1)

where rating it is a numeric value for the alphanumeric long-term foreign currency rating of country i at the end of year t; x it is the vector of measurable fundamentals; β is a vector of coefficients (or weights) to be estimated; α i is a countryspecific fixed effect; τ t is a macro time effect; and ε it is an error term. We convert each CRA’s alphanumeric rating into a numeric rating using a linear scale – a higher rating receives a higher value (see Table A2 in Appendix A for the details of 11 the conversion). Table 3 presents the coefficient estimates from equation (1) for Moody’s ratings. The specification in column (1) controls for nominal GDP, GDP per capita, GDP growth, public debt, whether the state has a past default and whether the state is an advanced country. All variables have the expected sign and are highly statistically significant, with the exception of nominal GDP. The specification in column (2) additionally controls for the current-account-to-GDP ratio and the external-debt-to-GDP ratio. Both are statistically significant but the magnitude of their coefficients suggests they are somewhat less important as drivers of Moody’s ratings than the variables in column (1). The estimates in column (3) also control for the World Bank indices of government effectiveness and rule of law. Ratings are strongly increasing in government effectiveness and the rule of law. The final column of Table 3 displays the estimates when we control for country-specific fixed effects. This identifies the model of the CRAs using only the variation of ratings within-state and across-time. The ratings agencies tell us that these 12 country-specific effects are important. Most of the parameter estimates are not significantly altered by controlling for country fixed effects, although nominal GDP becomes statistically significant at the 10% level while government effectiveness is rendered insignificant, presumably because of little time variation within the latter. TABLE 3: ESTIMATES OF MOODY’S REACTION FUNCTION

Nominal GDP GDP per capita GDP growth Public debt Current account External debt Past default Advanced country Government Law Country fixed effects No. of observations No. of countries R-sq: within R-sq: between R-sq: overall

DEPENDENT VARIABLE: MOODY’S SOVEREIGN CREDIT RATING (NUMERIC) (2) (3) (4) Coef. S.E. Coef. S.E. Coef. S.E. 0.19 0.13 0.13 0.09 0.30* 0.17 0.16*** 0.04 0.15*** 0.04 0.14*** 0.05 0.12*** 0.05 0.10** 0.05 0.09** 0.05 -0.04*** 0.01 -0.04*** 0.01 -0.04*** 0.01 -0.06*** 0.01 -0.05*** 0.01 -0.06*** 0.01 -1.6E-4*** 2.3E-5 -1.5E-4*** 2.6E-5 -1.6E-4*** 3.6E-5 -2.36*** 0.61 -1.99*** 0.54 -1.75*** 0.51 -2.41*** 0.70 6.21*** 0.73 4.94*** 1.16 3.23*** 1.09 0.64*** 0.41 0.12 0.49 0.48** 0.45 0.54** 0.51 NO NO NO YES 1,569 1,016 999 999 111 94 94 94 0.39 0.42 0.40 0.41 0.75 0.77 0.82 0.65 0.75 0.74 0.79 0.66 (1) Coef. S.E. 0.07 0.08 0.09*** 0.02 0.17*** 0.05 -0.04*** 0.01

Notes: Standard errors are clustered at the country level. Significance levels: * 10%; ** 5%; *** 1%.

Source: UniCredit Research

10

The inquisitive reader may ask why we do not follow the methodology of the CRAs to a tee, for example, by applying Moody’s ‘scorecard’ and then compute the subjective residual as the difference between the actual rating and the rating implied by the scorecard. There are several reasons for this. First, following Moody’s scorecard is not an easy task because it is so convoluted. Second, and more importantly, it would also not be appropriate. Our purpose is to estimate the part of the actual rating that cannot be explained by fundamentals – that is, the subjective component of the rating. Since Moody’s ad-hoc weights are themselves part of the subjective component, we would not want to follow their methodology. Finally, our regression approach has the desirable property that each measurable fundamental determinant of the sovereign rating is given the same weight for every country and year. It is right that our derived ‘objective’ component of ratings uses these consistent weights. In contrast, idiosyncratic factors necessarily belong to the residual or ‘subjective’ component.

11

In order to deal with potential non-linearities we also ran an ordered probit regression for ratings. The results were qualitatively similar to those of the linear model and are available on request.

12

Moody’s says that “The use of supplementary adjustment factors is an attempt to capture idiosyncratic country-specific factors which may not be universally available or relevant (Moody’s, 2013)”.

UniCredit Research

page 9

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

The “objective” component of the sovereign rating is then simply the weighted sum of the fundamentals (i.e. the fitted value). The “subjective” component is computed as the difference between the actual rating and the “objective” component (i.e. the residual). As illustrated in Table 3, the overall fit of the model is fairly good. The percentage of the crosscountry variation of ratings that is explained by the “objective” component is 75–82 percent, 13 leaving the remaining 18–25 percent to be explained by the “subjective” component. 14

Chart 1 displays the distribution of the “objective” component across all rated countries. Recall that the top credit rating Aaa/AAA corresponds to 24 rating notches, while noninvestment grade ratings have values equal to 14 notches and below. The distribution is fairly evenly spread with tentative evidence of it being bi-modal with peaks at prime grade and speculative grade. Chart 2 shows the distribution of the “subjective” component. There are two things of note here. First the subjective component is not far off normally distributed, which suggests our measurable fundamentals do a good job of explaining ratings. Second, most of the subjective components are small in magnitude. CHART 1: DISTRIBUTION OF OBJECTIVE COMPONENT

12

CHART 2: DISTRIBUTION OF SUBJECTIVE COMPONENT

percent of countries rated Speculative grade

50

Investment grade

10

percent of countries rated

40

8 30 6 20

4

10

2 0

0 5

7

9

11

13

15

17

19

21

23

25

-9

number of rating notches

-8

-7

-6

-5

-4

-3

-2

-1

0

1

2

3

4

5

number of ratings notches of subjective overruling

Source: UniCredit Research

13

The “objective” component does less well at explaining the variation in ratings within country and across time – only between 39 percent (column (1) of Table 3) and 42 percent of the time variation can be explained by movements in the “objective” component.

14

The “objective” components here are the fitted values from the regression in column (1) of Table 3. The reason why the estimated “objective” component can fall outside the top rating of 24 (=Aaa/AAA), and indeed some of the fitted values are equal to 25, is because the regression assumes that ratings (the dependent variable) are a continuous variable and not “limited” to just 24 categories.

UniCredit Research

page 10

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

IV. Is subjectivity useful in predicting default? Ultimately the usefulness of credit ratings depends on their ability to predict actual default events. Table 4 presents the means and standard deviations for our measurable fundamentals of creditworthiness in our sample, where we have split the sample into those countries that actually defaulted and those that did not. Unsurprisingly, default events tend to be associated with low GDP per capita, low GDP growth, high public debt, a history of past default, poor government effectiveness and a weak rule of law. This suggests the “objective” component of ratings will be informative of future default, but what about the “subjective” component? TABLE 4: FUNDAMENTALS IN TIMES OF DEFAULT AND NO DEFAULT Mean

No default

Std. dev.

Mean

Default

Std. dev.

Nominal GDP

0.86

3.43

0.16

0.27

GDP per capita

17.0

14.4

6.84

5.15

GDP growth

3.92

2.60

1.72

3.57

Public debt

52.8

33.8

75.2

35.8

Current account

-1.39

7.84

-0.50

6.90

External debt

663.1

2,574.3

166.9

250.0

Past default

0.39

0.49

1

0

Advanced country

0.33

0.47

0.05

0.23

Government

0.46

1.00

-0.49

0.49

Law

0.29

1.06

-0.69

0.61

No. of observations

980

19

Notes: The table presents the means and standard deviations for our ten fundamentals of creditworthiness conditional on default. The sample is restricted to that of column (3) in Table 7. Source: UniCredit Research

Table 5 lists the 26 default events in our sample since 1997, along with the “objective” and “subjective” components of ratings one and two years prior to default. Actually, to be precise, since we use end-of-year ratings whereas the default can take place any time within the year, the one-year prior to default actually refers to a horizon of up to one year, and similarly the twoyears prior to default refers to a horizon of up to two years. There are two things to note here. First, simple adding up reveals that the subjective component up to two years prior to default was positive in exactly half the cases – that is, the rating committees got it right half the time – and wrong half the time, so they performed as well (or bad) as a coin toss. Second, in 17 cases (65%) the objective component fell as the sovereign approached default – and this from an 15 already low level; whereas the subjective component fell in only 15 cases (58%).

15

The picture does not change much when one takes account of the fact that the “objective” and “subjective” components are measured with error. Chart B1 in Appendix B plots the standard errors of the objective components. The average standard error is around 0.5, which is fairly small. More specifically, a standard error of 0.5 means that 95% of the time we can be sure the “objective” component lies within 1 notch of the central estimates in Table 5.

UniCredit Research

page 11

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

TABLE 5: SOVEREIGN DEFAULT EVENTS IN OUR SAMPLE Objective component (number of rating notches) Country

Year of default

Subjective component (number of rating notches of subjective overruling)

1Y prior to default

2Y prior to default

1Y prior to default

2Y prior to default 0.1

Russia

1998

12.4

12.9

0.6

Ecuador

1999

6.5

7.2

2.5

3.8

Indonesia

1999

13.1

14.8

-3.1

-0.8

Pakistan

1999

11.1

11.4

-3.1

-0.4

Indonesia

2000

9.8

13.1

0.2

-3.1

Morocco*

2000

13.2

13.1

-0.2

-0.1

Ukraine

2000

9.3

9.4

-0.3

-0.4

Argentina

2001

10.0

10.5

2.0

1.5

Suriname*

2001

9.2

9.8

-0.2

-0.8

Suriname*

2002

9.0

9.2

0.0

-0.2

Paraguay*

2003

9.1

9.7

-0.1

0.3

Moldova

2002

7.6

6.9

0.4

2.1

Uruguay

2003

10.7

13.0

0.3

2.0

Grenada*

2004

10.9

10.6

1.1

1.4

Dominican Republic

2005

12.3

11.9

-3.3

-1.9

Venezuela

2005

10.0

9.4

0.0

-1.4

Belize

2006

10.6

10.5

-4.6

1.5

Ecuador

2008

7.9

7.9

-0.9

0.1

Seychelles*

2008

9.7

9.5

0.3

0.5

Nicaragua

2008

11.1

10.0

-3.1

-2.0

Jamaica

2010

10.4

10.8

-2.4

0.2

Belize

2012

8.3

8.6

0.7

0.4

Greece

2012

14.3

15.9

-3.3

-1.9

Grenada*

2012

8.2

8.4

-0.2

-0.4

Grenada*

2013

7.8

8.2

0.2

-0.2

Jamaica

2013

7.3

7.6

1.7

1.4

Notes: The table presents the objective and subjective components of Moody’s ratings within one and two years of the listed sovereign defaults. * indicates that the sovereign was not rated by Moody’s but was rated by S&P and the values for the objective and subjective components refer to those from the model using S&P ratings (see Table B1 in the Appendix). Source: Bloomberg; IMF; World Bank; BIS; UniCredit Research

Summing up only those “subjective” components that are large enough in magnitude to fall outside the margin of error, up to two years prior to default the subjective component was positive in six cases and negative in five cases – that is, the rating committees performed worse 16 than a coin toss. As default becomes imminent (i.e. within a year), the rating committees do much better, with the subjective component being positive in just three cases and negative in eight cases. The same point is illustrated in Chart 4, which plots the frequency distribution of the subjective component conditional on whether the country defaults one-year and two-years ahead. The distribution of the subjective component is approximately symmetrical about zero even though the state is about to default within two years. This improves slightly as the default falls to within a year (as can be seen by the subtle shift leftwards of the distribution). But still, in almost half the cases of actual default occurring in less than one year (namely 9 out of 19 instances), the credit committees either did nothing to downgrade the country relative to what the objective signals suggested, or even modified the objective signal towards a higher rating.

16

Table B1 in Appendix B reproduces Table 5 for the subjective component and adds the 95% confidence interval.

UniCredit Research

page 12

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

CHART 3: OBJECTIVE COMPONENT BEFORE DEFAULT conditional on default 1Y ahead

CHART 4: SUBJECTIVE COMPONENT BEFORE DEFAULT conditional on default 1Y ahead

conditional on default 2Y ahead

conditional on default 2Y ahead

frequency of countries rated

frequency of countries rated

7

5

6 4

5

3

4 3

2

2 1

1 0

0 7

8

9

-5

16

15

14

10 11 12 13 number of rating notches

-4

-3

-2

-1

0

1

2

3

4

number of rating notches of subjective overruling

Source: UniCredit Research

Of course, when assessing whether the “objective” and “subjective” components are useful or not in predicting default, one should use all the variation in the data – that is, cases where a country defaults and does not default. This is what Charts 5 and 6 illustrate, they plot the distribution of the objective and subjective components, respectively, conditional on whether the state defaults or not within two years. Chart 5 shows that the “objective” component of those countries that are about to default tends to be much lower than those that are not about to default. In stark contrast, Chart 6 shows there is very little difference between the “subjective” components assigned to those that are about to default and those that are not. The implication is that the “objective” component is useful in predicting default but the “subjective” component is not. CHART 5: OBJECTIVE COMPONENT IF NO DEFAULT/ DEFAULT conditional on default 2Y ahead 20

CHART 6: SUBJECTIVE COMPONENT IF NO DEFAULT/ DEFAULT

conditional on no default 2Y ahead

conditional on default 2Y ahead

percent of countries rated 50

conditional on no default 2Y ahead

percent of countries rated

40

15

30 10 20 5

10 0

0 5

7

9

11

13

15

17

19

21

23

-9

25

-8

-7

-6

-5

-4

-3

-2

-1

0

1

2

3

4

5

number of rating notches of subjective overruling

number of rating notches

Source: UniCredit Research

This is confirmed by the formal estimation results detailed in Box 2. We find that the “objective” component of ratings is a good predictor of default one to five years ahead. In stark contrast, although the “subjective” component is helpful in predicting imminent default (i.e. within a year), it provides no value at all two years out and, worse still, three years out it systematically overrules the signal from the objective component in the wrong direction! This is important for two reasons. First, typically the game is already up when default is imminent – so it is to be expected that the subjective component has some predictive power then. Indeed, the subjective component embodies ‘news’ that the objective component does not because the latter uses only historical data. Second, long-term credit ratings are meant to be

UniCredit Research

page 13

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

just that – long term. They are marketed as being informative of default three years ahead. And yet we have shown that the “subjective” component of ratings systematically and wrongly overrules the signal from fundamentals for the probability of default within three years. Therefore, history is littered with cases where sovereigns have been under- and over-rated because of this misguided subjectivity.

BOX 2: Estimating the probability of future default We estimate the following model for the probability of default for country i in year t:

defaultit* = β i + β1 (objectivei ,t −τ ) + β 2 log(subjectivei ,t −τ ) + uit

defaultit = 1[defaultit* > 0]; i = 1,..., N ; t = 1,..., Ti ; τ ∈ {1,2,3,4,5}

(2)

*

where default it is the latent propensity to default for country i in year t; objectivei ,t −τ is the objective component of the rating for country i in year t-τ, subjectivei ,t −τ is the subjective component for country i in year t-τ; default it is the observed binary default decision that takes the value one if country i defaults in year t and zero otherwise; u it is an error term, and 1[.] is the indicator function. The objective and subjective components of ratings are lagged τ periods because we are interested in the ability of these to forecast the probability of actual default τ-years ahead. Note that, by construction, the subjective component of ratings is independent (or orthogonal) to the objective component of ratings, and hence there is no danger of one variable proxying for the other. We proceed by estimating equation (2) for various lag lengths of the objective and subjective components. Table 6 presents the results for the random-effects probit model for the probability of default within one year. The estimates in each of the four columns correspond to the models in Table 3, which is where the objective and subjective components are derived from. The one-year lagged objective and subjective components are both highly statistically significant with the correct sign – they are good predictors of imminent default (i.e. default within a year). TABLE 6: THE PREDICTIVE POWER OF MOODYS RATINGS WITHIN ONE YEAR DEPENDENT VARIABLE: DEFAULT DUMMY

Objective component lagged one year Subjective component lagged one year No. of observations No. of countries Pseudo R-sq

(1) Coef. S.E. -0.20*** 0.04 -0.20*** 0.07 1,719 114 0.25

(2) Coef. S.E. -0.19*** 0.05 -0.20*** 0.07 1,144 104 0.22

(3) Coef. S.E. -0.20*** 0.05 -0.18** 0.07 1,108 103 0.22

Notes: Random-effects probit estimation. Significance levels: * 10%; ** 5%; *** 1%.

(4) Coef. S.E. -0.19*** 0.04 -0.21*** 0.07 1,108 103 0.22 Source: UniCredit Research

It is usually pretty clear that the game is up when default is imminent. Therefore, it is to be expected that the “subjective” component is useful for predicting imminent default, because while the rating committees can react to ‘news’ the objective component only uses past information. Table 7 presents the estimates when the objective and subjective components are lagged two years. The objective component remains overwhelmingly statistically significant and has the expected negative coefficient – that is, strong measurable fundamentals imply that it is less likely that the country will default within the next two years. In stark contrast, the subjective component of ratings now adds no predictive value at all – the coefficient is not statistically different from zero. While on average the subjective component has no predictive power, this disguises individual sins. TABLE 7: THE PREDICTIVE POWER OF MOODYS RATINGS WITHIN TWO YEARS DEPENDENT VARIABLE: DEFAULT DUMMY

Objective component lagged two years Subjective component lagged two years No. of observations No. of countries Pseudo R-sq

(1) Coef. S.E. -0.15*** 0.03 -0.02 0.08 1,569 111 0.18

(2) Coef. S.E. -0.14*** 0.04 1.3E-3 0.08 1,016 94 0.15

Notes: Random-effects probit estimation. Significance levels: * 10%; ** 5%; *** 1%.

UniCredit Research

page 14

(3) Coef. S.E. -0.14*** 0.04 1.8E-3 0.08 999 94 0.15

(4) Coef. S.E. -0.13*** 0.03 -0.03 0.08 999 94 0.14 Source: UniCredit Research

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

Table 8 displays the estimates for when the objective and subjective components are lagged three years. The results are striking, whilst the objective component is strongly statistically significant with the expected negative sign, the subjective component is now weakly statistically significant but with the wrong sign. This means that the subjective component is no longer noise, it is actually systematically overruling the signal from macro fundamentals in the wrong direction! This is important because long-term credit ratings are meant to be informative of the probability of default three years ahead. TABLE 8: THE PREDICTIVE POWER OF MOODYS RATINGS WITHIN THREE YEARS DEPENDENT VARIABLE: DEFAULT DUMMY

Objective component lagged three years Subjective component lagged three years No. of observations No. of countries Pseudo R-sq

(1) Coef. S.E. -0.12*** 0.03 0.16* 0.09 1,569 111 0.16

(2) Coef. S.E. -0.11*** 0.03 0.19* 0.10 1,016 94 0.14

Notes: Random-effects probit estimation. Significance levels: * 10%; ** 5%; *** 1%.

(3) Coef. S.E. -0.11*** 0.03 0.19* 0.10 999 94 0.14

(4) Coef. S.E. -0.10*** 0.03 0.14 0.10 999 94 0.12 Source: UniCredit Research

Looking at longer horizons, the objective component remains negative and highly significant four and even five years ahead. In contrast, the subjective component is either insignificant or significant but with the wrong sign.

UniCredit Research

page 15

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

V. A look back at the EMU debt crisis Of particular interest (because the consequences nearly led to a disaster of a potential meltdown of the financial system) is what role the CRAs’ subjectivity played during the 2009-2011 Eurozone debt crisis, a period when the Eurozone periphery was sharply downgraded. Chart 7 displays the subjective component of ratings for Italy, Spain, Portugal, Ireland and Greece from 1989 to 2013. For these countries the subjective component was, on average, slightly positive in the first decade of the Euro. But then between 2010 and 2011 the subjective component turned sharply negative, on average subtracting from the objective signal three notches by end-2011, and then six notches by end-2012. This deeply negative “subjective” component is much bigger in absolute value than almost anything else seen in our sample of more than one hundred countries over more than ten years (see Chart 2). An important exception where a large negative component now appears to have been justified is for Greece. Greece is a special case because it lied about its accounts. Excluding Greece, on average the periphery was still rated five-notches below its objective rating at end-2012. The negative sentiment of the rating agencies towards the Eurozone periphery stands in stark contrast to their positive view of the so-called “Fragile Five” (Brazil, India, Indonesia, South Africa and Turkey). On the onset of the Eurozone debt crisis, the subjective component of the ratings of the “Fragile Five” became more positive (see Chart 8). Over the same period, the core Eurozone countries were on average rated roughly in line with fundamentals, while the UK and the US were rated slightly higher (Chart 9). The rating committees’ overruling of the objective macro signals contributed to causing distortions to capital flows on a magnitude that became systemically dangerous. To illustrate, following the downgrades of the Eurozone periphery – directly or indirectly (dictated by bank supervisors) – banks reporting to the BIS reduced their exposure to the Eurozone periphery by a whopping USD 1.4 trillion (or 43%) between the end of 2009 and the third quarter of 17 2013. European banks accounted for almost all (USD 1.3 trillion) of this withdrawal of funds (down 45%). Apart from clearly contributing to the depth of the recession in the periphery during these years, and hence untold social hardship, the downgrades and withdrawal of capital appear to have become a virtually self-fulfilling spiral that was only broken by the 18 ECB's introduction of the OMT. As a virtual mirror image, and as a bizarre irony of the flawed system, while the CRAs rating committees were busy overruling the objective macro signals in a negative direction causing unjustified downgrades in the Eurozone periphery, they overruled the objective macro signals from the so-called Fragile Five, delivering ratings upgrades not justified by their own macro fundamentals for this group of countries. And hardly as a coincidence, during this same period, banks reporting to the BIS increased their exposure to the "Fragile Five" by USD 282bn (or 32%), of which European banks increased their exposure by USD 144bn (up 23%).

17

See BIS international banking statistics, Table 9D: Consolidated foreign claims of reporting banks – ultimate risk basis, January 2014. Gartner and Griesbach (2012) find that the downgrades of the Eurozone periphery were a central cause of the European debt crisis. More specifically, the authors find that every spread widening by 140-150bp triggers a downgrade, which can push up yields to a level that triggers a further downgrade. This self-fulfilling prophecy is aggravated by the non-linear reaction of yield spreads to a rating action. 18

UniCredit Research

page 16

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

CHART 7: A STRONG DISLIKE FOR THE EMU PERIPHERY greece 4

italy

ireland

portugal

CHART 8: FAVOURING THE FRAGILE 5

spain

turkey

south africa

subjective component, in rating notches

4

2

3

0

2

brazil

india

indonesia

subjective component, in rating notches

1

-2

0

-4

-1

-6

-2

-8

-3 -4

-10 1989

1993

1997

2001

2009

2005

2013

1994

1998

2002

2006

2010

Source: UniCredit Research

CHART 9: SUBJECTIVITY DURING THE EMU DEBT CRISIS EMUperiphery 3

EMU periphery excl. Greece

Core EMU

US

UK

Fragile 5

subjective component, in rating notches

2 1 0 -1 -2 -3 -4 -5 -6 -7 2007

2008

2009

2010

2011

2012

2013

Source: Bloomberg, Haver Analytics, UniCredit Research

UniCredit Research

page 17

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

VI. Current misalignments Eventually ratings should catch-up with those suggested by fundamentals as the rating committees realize their errors. Table 9 displays the current misalignment (at the end of 2013) between the actual Moody’s credit rating and the objective rating for selected countries. The table is quite staggering. Spain, Portugal, and Ireland are all five notches below where they should be, while Italy is underrated by four notches. For Portugal this is the difference between its current junk status and investment grade. At the other end of the scale, Brazil, Indonesia and Turkey are overrated. The core of the Eurozone is rated about right. The current misalignments for 108 countries are given in Appendix E. TABLE 9: IMPLIED CURRENT RATING MISALIGNMENT Moody’s rating

Objective rating

Overvaluation (in notches)

Error band (in notches)

Brazil

Baa2

Ba2

+3

± 0.9

Indonesia

Baa3

Ba3

+3

± 1.0

Turkey

Baa3

Ba2

+2

± 0.8

Country

Finland

Aaa

Aa1

+1

± 0.9

Baa3

Ba1

+1

± 1.2

Netherlands

Aaa

Aa1

+1

± 1.1

South Africa

Baa1

Baa2

+1

± 0.9

UK

Aa1

Aa2

+1

± 1.1

US

Aaa

Aa1

+1

± 2.3

Austria

Aaa

Aaa

0

± 4.9

France

Aa1

Aa1

0

± 1.1

Germany

Aaa

Aaa

0

± 1.0

Luxembourg

Aaa

Aaa

0

± 2.1

Belgium

Aa3

Aa1

-2

± 1.2

Italy

Baa2

A1

-4

± 1.4

Greece

Caa3

B1

-5

± 2.6

Ireland

Ba1

A2

-5

± 1.5

Portugal

Ba3

Baa1

-5

± 1.4

Baa3

A1

-5

± 1.1

India

Spain

Notes: The “error band” is the 95% confidence interval.

UniCredit Research

page 18

Source: UniCredit Research

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

VII. Concluding remarks A priori it is not obvious whether credit ratings should be purely based on a quantitative framework or, rather, be supplemented by the judgment of rating committees. It is, ultimately, an empirical question. In this paper we find that in some cases the rating committees overrule the signal coming from fundamentals of creditworthiness to a large extent, and none more so than during the Eurozone debt crisis. Crucially, however, we showed that these value judgments have no predictive power at all for the probability of default one or more years ahead. Sovereign default is almost always a choice of government and our findings suggest this is typically based on measurable fundamentals. The implications are far and wide. History is littered with cases of sovereigns being over- and under-rated. But none more so than now, with the Eurozone periphery currently suffering from ratings far below that suggested by fundamentals. We conclude that the self-appointed credit committees have caused more damage than good as their over rulings of macro fundamentals have been somewhat random in both directions, and with serious distortions in recent years. We suggest that the CRAs be stripped of their regulatory powers for sovereign ratings. Anything with a regulatory role should be carried out by agents of the state, or in this case all states. Therefore, to replace them in this role, we suggest that an international body be mandated with the development of a set of macro indicators (along the lines of what we have shown here to be the CRAs' objective signal), and that these indicators replace the role in regulatory and other policy matters presently played by the CRAs. In cases where these indicators suggest significant changes in a short span of time, the country's finance minister could be invited to make his or her case to, for example, the board of the IMF after which a pre-assigned body (the IMF Board?) could decide a temporary adjustment to the "rating" of that country. Failing that (and it would be a fail) the ratings agencies should be forced to substantially increase transparency, including publishing a separate breakdown of the objective and subjective components of ratings, the minutes of the rating committees, and the voting records.

UniCredit Research

page 19

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

References

UniCredit Research

■

Brooks, R., R. Faff, D. Hillier and J. Hillier (2004) "The National Market Impact of Sovereign Rating Changes", Journal of Banking & Finance, 28 (1), pp. 233–250.

■

Cantor, Richard and Frank Packer (1996) “Determinants and Impact of Sovereign Credit Ratings”, FRBNY Economic Policy Review, October 1996.

■

Ferri, G., L.-G. Liu and J. E. Stiglitz (1999) “The Procyclical Role of Rating Agencies: Evidence from the East Asian Crisis”, Economic Notes by Banca Monte dei Paschi di Siena SpA, Vol. 28, No. 3, pp. 335-355.

■

Fitch (2012) “Sovereign Rating Criteria: Master Criteria”, 13 August 2012.

■

Gartner, M. and B. Griesbach (2012), "Rating agencies, self-fulfilling prophecy and multiple equilibria? An empirical model of the European sovereign debt crisis 2009-2011", Discussion Paper, University of St. Gallen.

■

Kaminsky, G. and S. Schmukler (2002), "Emerging Markets Instability: Do Sovereign Ratings Affect Country Risk and Stock Returns?," World Bank Economic Review, World Bank, Vol. 16 (2), pp. 171–195, August.

■

Kiff, John, Sylwia Nowak and Liliana Schumacher (2012) “Are Rating Agencies Powerful? An Investigation into the Impact and Accuracy of Sovereign Ratings”, IMF Working Paper 12/23.

■

Kliger, Doron and Oded Sarig (2000) “The Informational Value of Bond Holdings”, Journal of Finance, 55 (6), pp. 2879–2902.

■

Moody’s (2013) “Rating Methodology: Sovereign Bond Ratings”, September 12, 2013.

■

Reisen, Helmut and Julia von Maltzan (1999) “Boom and Bust and Sovereign Ratings”, International Finance, Vol. 2, pp. 273–293.

■

Standard & Poor’s (2012) “How We Rate Sovereigns”, March 13, 2012.

■

The Economist (2013) “Free Speech or Knowing Misrepresentation?” 5 February.

page 20

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

APPENDICES A. Data and variable construction This Appendix defines the variables (see Table A1) and the numeric conversion of ratings (see Table A2) employed in the empirical analysis of sections III and IV. TABLE A1: DATA DEFINITIONS AND SOURCES Variable name

Definition

Units

Determinants of sovereign ratings Nominal GDP GDP in current prices and exchange rates

Data source

USD tn

IMF

GDP per capita

Nominal GDP per person, PPP-adjusted

USD thous.

GDP growth

Average annual real GDP growth t-9 to t

Percent

Public Debt

General government gross debt, end-of-year

Percent of GDP

Current account

Annual current account balance

Percent of GDP

IMF

External debt

Gross external debt

Percent of GDP

BIS; IMF; UniCredit Research

Past default

Indicator variable that takes the value one in all years following a default event since 1960, and zero otherwise

Binary

S&P; UniCredit Research

Advanced country

An indicator variable that takes the value one if the country is deemed advanced by the IMF, and zero otherwise

Binary

IMF; UniCredit Research

Government

World Bank Government Effectiveness index

Index

World Bank

Law Explained variables Moody’s, S&P or Fitch ratings Default

World Bank Rule of Law index

Index

World Bank

Long-term foreign-currency sovereign ratings assigned by Moody’s, S&P and Fitch An indicator variable for sovereign default that takes the value one if the country defaults in the year and zero otherwise

IMF IMF; UniCredit Research IMF

Alphanumeric rating, see Table A2 Binary

Note: IMF=International Monetary Fund; BIS=Bank for International Settlements; S&P=Standard and Poors.

Bloomberg S&P; UniCredit Research

Source: UniCredit Research

TABLE A2: NUMERIC CONVERSION OF ALPHANUMERIC RATINGS

Description INVESTMENT-GRADE Prime High grade

Upper medium grade

Lower medium grade

NON-INVESTMENT GRADE Speculative

Highly speculative

Substantial risks Extremely speculative Imminent default risk

In default

Moody’s Numeric Rating conversion

Rating

S&P Numeric conversion

Fitch Numeric Rating conversion

Aaa Aa1 Aa2 Aa3 A1 A2 A3 Baa1 Baa2 Baa3

24 23 22 21 20 19 18 17 16 15

AAA AA+ AA AAA+ A ABBB+ BBB BBB-

24 23 22 21 20 19 18 17 16 15

AAA AA+ AA AAA+ A ABBB+ BBB BBB-

24 23 22 21 20 19 18 17 16 15

Ba1 Ba2 Ba3 B1 B2 B3 Caa1 Caa2 Caa3

14 13 12 11 10 9 8 7 6

14 13 12 11 10 9

CCC

6

4.5

14 13 12 11 10 9 8 7 6 5 4

BB+ BB BBB+ B B-

Ca

BB+ BB BBB+ B BCCC+ CCC CCCCC C

C -

3 D

2

DDD DD D

3 2 1

Source: Moody’s, S&P, Fitch, UniCredit Research

UniCredit Research

page 21

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

B. The precision of our estimates This Appendix details the precision of our estimates of the “objective” and “subjective” components of Moody’s ratings. Chart B1 plots the standard error of the “objective” component (more specifically, the standard error of the fitted values from column (1) of Table 3). Table B2 provides the 95% confidence intervals for the “subjective” component prior to our 26 default events.

standard error of objective component

CHART B1: STANDARD ERRORS FOR THE OBJECTIVE COMPONENT 3 2.5 2 1.5 1 0.5 0 0

5

10

15

20

25

30

objective component (in notches) Source: UniCredit Research

TABLE B1: CONFIDENCE INTERVALS FOR THE SUBJECTIVE COMPONENT PRIOR TO DEFAULT Country Russia Ecuador Indonesia Pakistan Indonesia Morocco* Ukraine Argentina Suriname* Suriname* Paraguay* Moldova Uruguay Grenada* Dominican Republic Venezuela Belize Ecuador Seychelles* Nicaragua Jamaica Belize Greece Grenada* Grenada* Jamaica

Year of default 1998 1999 1999 1999 2000 2000 2000 2001 2001 2002 2003 2002 2003 2004 2005 2005 2006 2008 2008 2008 2010 2012 2012 2012 2013 2013

Subjective component: 95% confidence intervals 1Y prior to default 2Y prior to default 0.6 ± 1.2 0.1 ± 1.2 2.5 ± 1.0 3.8 ± 0.9 -3.1 ± 0.9 -0.8 ± 0.9 -3.1 ± 0.9 -0.4 ± 0.9 0.2 ± 1.0 -3.1 ± 0.9 -0.2 ± 0.8 -0.1 ± 0.9 -0.3 ± 1.4 -0.4 ± 1.5 2.0 ± 0.8 1.5 ± 0.8 -0.2 ± 0.9 -0.8 ± 0.9 0.0 ± 0.8 -0.2 ± 0.9 -0.1 ± 0.7 0.3 ± 0.7 0.4 ± 1.0 2.1 ± 1.1 0.3 ± 1.1 2.0 ± 0.8 1.1 ± 0.9 1.4 ± 0.8 -3.3 ± 1.1 -1.9 ± 1.0 0.0 ± 0.8 -1.4 ± 0.9 -4.6 ± 1.0 1.5 ± 1.0 -0.9 ± 0.8 0.1 ± 0.8 0.3 ± 1.3 0.5 ± 1.3 -3.1 ± 1.0 -2.0 ± 1.0 -2.4 ± 1.3 0.2 ± 1.1 0.7 ± 1.0 0.4 ± 1.0 -3.3 ± 1.9 -1.9 ± 1.6 -0.2 ± 1.0 -0.4 ± 1.0 0.2 ± 1.1 -0.2 ± 1.0 1.7 ± 1.7 1.4 ± 1.6

Notes: The table presents the 95% confidence intervals for the subjective component of Moody’s ratings within one and two years of the listed sovereign defaults. * indicates that the sovereign was not rated by Moody’s but was rated by S&P and the values for the subjective components refer to those from the model using S&P ratings (see Table C1 in Appendix C). Source: UniCredit Research

UniCredit Research

page 22

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

C. The results for S&P This Appendix contains the regression results for S&P. Table C1 contains the estimates of S&P’s reaction function (it corresponds to Table 3 for Moody’s in the main text). Tables C2 and C3 display the coefficient estimates on the objective and subjective components of S&P’s ratings for the probability of default one and two years ahead, respectively (they correspond to Tables 6 and 7 for Moody’s in the main text). TABLE C1: ESTIMATES OF S&P’S REACTION FUNCTION DEPENDENT VARIABLE: S&P’S SOVEREIGN CREDIT RATING (NUMERIC) (1) Coef. S.E. Nominal GDP GDP per capita

(2) Coef. S.E.

(3) Coef. S.E.

(4) Coef. S.E.

0.09 0.10

0.25 0.19

0.17 0.12

0.37 0.26

0.06*** 0.02

0.13*** 0.04

0.14*** 0.04

0.12** 0.05

GDP growth

0.27*** 0.06

0.25*** 0.06

0.23*** 0.06

0.23*** 0.07

Public Debt

-0.04*** 0.01

-0.04*** 0.01

-0.04*** 0.01

-0.04*** 0.01

Current account External debt Past default Advanced country

-0.02 0.01

-0.02 0.01

-0.02* 0.01

-1.3E-4 8.9E-5

-1.3E-4* 7.0E-5

-1.8E-4 1.4E-4 -0.39 0.45

-0.80* 0.41

-0.62 0.41

-0.27 0.33

7.86*** 0.61

6.25*** 1.10

3.98*** 0.98

Government Law

1.01*** 0.32

0.66* 0.37

0.27 0.34

0.23 0.37

Country fixed effects

NO

NO

NO

YES

No. of observations

1,719

1,144

1,108

1,108

No. of countries

114

104

103

103

R-sq: within

0.39

0.44

0.45

0.46

R-sq: between

0.75

0.78

0.85

0.62

R-sq: overall

0.77

0.75

0.82

0.62

Notes: Standard errors are clustered at the country level. Significance levels: * 10%; ** 5%; *** 1%.

Source: UniCredit Research

TABLE C2: THE PREDICTIVE POWER OF S&P RATINGS WITHIN ONE YEAR DEPENDENT VARIABLE: DEFAULT DUMMY (1) Coef. S.E.

(2) Coef. S.E.

(3) Coef. S.E.

(4) Coef. S.E.

Objective component lagged one year

-0.33*** 0.07

-0.32*** 0.07

-0.31*** 0.08

-0.31*** 0.08

Subjective component lagged one year

-0.30*** 0.08

-0.28*** 0.09

-0.32*** 0.09

-0.32*** 0.09

No. of observations

1,719

1,144

1,108

No. of countries

114

104

103

103

Pseudo R-sq

0.28

0.25

0.26

0.26

Notes: Random-effects probit estimation. Significance levels: * 10%; ** 5%; *** 1%.

1,108

Source: UniCredit Research

TABLE C3: THE PREDICTIVE POWER OF S&P RATINGS WITHIN TWO YEARS DEPENDENT VARIABLE: DEFAULT DUMMY

Objective component lagged two years Subjective component lagged two years No. of observations

(1) Coef. S.E.

(2) Coef. S.E.

(3) Coef. S.E.

(4) Coef. S.E.

-0.20*** 0.05

-0.19*** 0.05

-0.17*** 0.05

-0.17*** 0.05

-0.03 0.09

0.03 0.10

0.05 0.10

0.04 0.10

1,719

1,144

1,108

No. of countries

114

104

103

103

Pseudo R-sq

0.17

0.14

0.14

0.14

Notes: Random-effects probit estimation. Significance levels: * 10%; ** 5%; *** 1%. D. The results for Fitch

UniCredit Research

page 23

1,108

Source: UniCredit Research

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

D. The results for Fitch This Appendix contains the regression results for Fitch. Table D1 contains the estimates of Fitch’s reaction function (it corresponds to Table 3 for Moody’s in the main text). Tables D2 and D3 display the coefficient estimates on the objective and subjective components of Fitch’s ratings for the probability of default one and two years ahead, respectively (they correspond to Tables 6 and 7 for Moody’s in the main text). TABLE D1: ESTIMATES OF FITCH’S REACTION FUNCTION DEPENDENT VARIABLE: FITCH’S SOVEREIGN CREDIT RATING (NUMERIC) (1) Coef. S.E. Nominal GDP GDP per capita

(2) Coef. S.E.

(3) Coef. S.E.

(4) Coef. S.E.

0.06 0.06

0.15* 0.09

0.13* 0.07

0.18 0.11

0.09*** 0.03

0.14*** 0.04

0.14*** 0.04

0.13*** 0.05

GDP growth

0.16*** 0.04

0.13*** 0.04

0.11*** 0.04

0.13*** 0.04

Public Debt

-0.04*** 0.01

-0.04*** 0.01

-0.03*** 0.01

-0.04*** 0.01

Current account

-0.02* 0.01

External debt Past default Advanced country

-0.02* 0.01

-0.02** 0.01

-1.0E-4*** 1.6E-5

-8.5E-5*** 1.9E-5

-9.5E-5*** 2.0E-5

-2.36*** 0.77

-2.38*** 0.75

-2.05*** 0.67

-3.23*** 0.92

6.06*** 0.94

4.82*** 1.25

2.95** 1.18

Government

1.11*** 0.34

0.81** 0.37

Law

7.6E-4 0.33

-0.07 0.35

Country fixed effects

NO

NO

NO

YES

No. of observations

1,339

972

971

971

No. of countries

102

94

94

94

R-sq: within

0.35

0.39

0.39

0.4

R-sq: between

0.74

0.76

0.82

0.68

R-sq: overall

0.74

0.73

0.79

0.66

Notes: Standard errors are clustered at the country level. Significance levels: * 10%; ** 5%; *** 1%.

Source: UniCredit Research

TABLE D2: THE PREDICTIVE POWER OF FITCH RATINGS WITHIN ONE YEAR DEPENDENT VARIABLE: DEFAULT DUMMY (1) Coef. S.E.

(2) Coef. S.E.

(3) Coef. S.E.

(4) Coef. S.E.

Objective component lagged one year

-0.17*** 0.05

-0.18*** 0.06

-0.19*** 0.06

-0.18*** 0.06

Subjective component lagged one year

-0.29*** 0.09

-0.27*** 0.09

-0.23** 0.09

-0.28*** 0.09

No. of observations

1,339

972

971

No. of countries

102

94

94

94

Pseudo R-sq

0.23

0.21

0.20

0.21

Notes: Random-effects probit estimation. Significance levels: * 10%; ** 5%; *** 1%.

971

Source: UniCredit Research

TABLE D3: THE PREDICTIVE POWER OF FITCH RATINGS WITHIN TWO YEARS DEPENDENT VARIABLE: DEFAULT DUMMY

Objective component lagged two years Subjective component lagged two years No. of observations No. of countries Pseudo R-sq

(1) Coef. S.E. -0.11*** 0.04

(2) Coef. S.E. -0.11*** 0.04

(3) Coef. S.E. -0.11*** 0.04

(4) Coef. S.E. -0.10*** 0.04

-0.02 0.11

0.04 0.12

0.04 0.12

-0.06 0.11

1,339 102 0.10

972 94 0.09

Notes: Random-effects probit estimation. Significance levels: * 10%; ** 5%; *** 1%.

UniCredit Research

page 24

971 94 0.09

971 94 0.08 Source: UniCredit Research

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

E. Current misalignment, all countries TABLE E1: IMPLIED CURRENT RATING MISALIGNMENT Country Brazil Indonesia South Korea Chile China Dominican Republic Kuwait Mexico Oman Peru Poland Romania Trinidad & Tobago Turkey Armenia Australia Azerbaijan Bahamas Bolivia Bulgaria Colombia Costa Rica Croatia Czech Republic Estonia Finland Guatemala India Lithuania Malaysia Mauritius Netherlands New Zealand Philippines Russia Saudi Arabia South Africa Suriname Thailand United Kingdom United States Uruguay Albania Angola Austria Bahrain Bosnia & Herzegovina Cambodia Canada Denmark Ecuador Fiji France Georgia Germany Ghana Honduras Hong Kong

UniCredit Research

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

page 25

Objective rating Ba2 Ba3 A3 A2 A2 B3 A1 Baa3 A3 Ba1 Baa1 Ba3 Baa3 Ba2 Ba3 Aa1 Ba1 Baa2 B1 Baa3 Ba1 Ba1 Ba2 A2 A2 Aa1 Ba2 Ba1 A3 Baa1 Baa2 Aa1 Aa1 Ba2 Baa2 A1 Baa2 B1 Baa2 Aa2 Aa1 Ba1 B1 Ba3 Aaa Baa2 B3 B2 Aaa Aaa Caa1 B1 Aa1 Ba3 Aaa B1 B2 Aa1

Overvaluation (in notches) 3 3 3 2 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Error band (in notches) ± 0.9 ± 1.0 ± 0.9 ± 0.9 ± 2.0 ± 0.9 ± 1.5 ± 0.9 ± 1.2 ± 0.9 ± 0.9 ± 1.0 ± 0.9 ± 0.8 ± 1.1 ± 0.9 ± 1.4 ± 1.2 ± 1.0 ± 0.9 ± 1.0 ± 0.8 ± 1.0 ± 0.9 ± 1.1 ± 0.9 ± 1.0 ± 1.2 ± 1.0 ± 0.9 ± 0.9 ± 1.1 ± 0.9 ± 1.0 ± 0.9 ± 1.3 ± 0.9 ± 0.9 ± 1.0 ± 1.1 ± 2.3 ± 0.9 ± 0.9 ± 1.2 ± 4.9 ± 1.1 ± 1.0 ± 1.3 ± 1.1 ± 0.9 ± 0.9 ± 1.1 ± 1.1 ± 1.2 ± 1.0 ± 1.1 ± 1.0 ± 1.1

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

TABLE E1: IMPLIED CURRENT RATING MISALIGNMENT (CONTINUED) Country Israel Japan Kazakhstan Kenya Lebanon Luxembourg Moldova Montenegro Morocco Mozambique Namibia Nicaragua Nigeria Norway Panama Paraguay Senegal Singapore Slovakia St Vincent & The Grenadines Sweden Switzerland Taiwan Belarus Belize El Salvador Jamaica Jordan Latvia Malta Pakistan Qatar Serbia Uganda Ukraine Vietnam Argentina Barbados Belgium Hungary Venezuela Egypt Tunisia Iceland Italy Greece Ireland Portugal Slovenia Spain

Moody’s rating A1 Aa3 Baa2 B1 B1 Aaa B3 Ba3 Ba1 B1 Baa3 B3 Ba3 Aaa Baa2 Ba3 B1 Aaa A2 B2 Aaa Aaa Aa3 B2 Caa2 Ba3 Caa3 B1 Baa2 A3 Caa1 Aa2 B1 B1 Caa1 B2 B3 Ba3 Aa3 Ba1 Caa1 B3 Ba3 Baa3 Baa2 Caa3 Ba1 Ba3 Ba1 Baa3

Notes: The ratings refer to end-2013. The “error band” is the 95% confidence interval.

UniCredit Research

page 26

Objective rating A1 Aa3 Baa2 B1 B1 Aaa B3 Ba3 Ba1 B1 Baa3 B3 Ba3 Aaa Baa2 Ba3 B1 Aaa A2 B2 Aaa Aaa Aa3 B1 Caa1 Ba2 Caa2 Ba3 Baa1 A2 B3 Aa1 Ba3 Ba3 B3 B1 B1 Ba1 Aa1 Baa2 B2 Ba3 Baa3 A2 A1 B1 A2 Baa1 A2 A1

Overvaluation (in notches) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -2 -2 -2 -2 -2 -3 -3 -4 -4 -5 -5 -5 -5 -5

Error band (in notches) ± 1.0 ± 2.9 ± 1.1 ± 1.0 ± 1.4 ± 2.1 ± 1.0 ± 0.9 ± 1.0 ± 1.1 ± 1.1 ± 1.0 ± 1.1 ± 1.3 ± 0.9 ± 1.0 ± 1.2 ± 1.8 ± 1.0 ± 1.0 ± 0.9 ± 1.0 ± 0.9 ± 1.0 ± 1.1 ± 0.9 ± 1.6 ± 1.1 ± 0.9 ± 1.0 ± 1.1 ± 4.0 ± 0.9 ± 1.3 ± 0.9 ± 1.0 ± 0.9 ± 1.1 ± 1.2 ± 1.0 ± 0.9 ± 1.1 ± 1.0 ± 1.2 ± 1.4 ± 2.6 ± 1.5 ± 1.4 ± 1.0 ± 1.1

Source: Bloomberg; UniCredit Research

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

Global Themes Series List No

Author(s)

Title

20

24 Oct 2013

Date

Dr. Andreas Rees

Introducing the Global Leading Indicator by UniCredit

19

12 Sep 2013

Michael Rottmann

The return of the macro-yield correlation & its implication for active duration management

18

3 Sep 2013

Vasileios Gkionakis, Daniel Vernazza

Introducing BEER by UniCredit; Our new framework for modeling equilibrium exchange rates

17

5 Jul 2013

Gillian Edgeworth, Dan Bucşa, Carlos Ortiz

CEE: Stress testing external financing shortfalls

16

19 Jun 2013

15

6 Jun 2013

14

21 May 2013

13

7 May 2013

12

Roberto Mialich

Too big to fall soon! Why the USD still remains the world's reserve currency

Gillian Edgeworth

CEE: The 'normalisation' challenge

Marco Valli

Inflating away the debt overhang? Not an option

Harm Bandholz, Tullia Bucco, Loredana Federico, Alexander Koch

The quest for competitiveness in the eurozone

4 Jan 2013

Luca Cazzulani, Elia Lattuga

Short and long-term impact of the introduction of CACs in the EMU

11

2 Oct 2012

Harm Bandholz

US Fiscal Policies at a Crossroad: consolidation through the fiscal cliff?

10

18 Sep 2012

Marco Valli

The eurozone five years into the crisis: lessons from the past and the way forward

9

30 Jul 2012

Erik F. Nielsen

Europe in the second half of 2012: Moving closer together or further apart?

8

18 Jul 2012

Harm Bandholz, Andreas Rees

Reach out for the medal(s)

7

16 Jul 2012

Luca Cazzulani, Chiara Cremonesi

EMU bond correlation & portfolio decisions

6

4 Jun 2012

Andreas Rees

Money scoring goals. Forecasting the European Football Championship 2012

5

23 Apr 2012

Harm Bandholz

How the Great Recession changed the Fed

4

16 Apr 2012

Erik F. Nielsen

Safeguarding the common eurozone capital market

3

10 Apr 2012

Andreas Rees

The hidden issue of long-term fiscal sustainability in the eurozone

2

23 Mar 2012

Alexander Koch

European housing: fundamentals and policy implications

1

12 Mar 2012

Gillian Edgeworth, Vladimir Zlacký, Dmitry Veselov

The EU: Managing capital flows in reverse

UniCredit Research

page 27

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

Notes

UniCredit Research

page 28

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

Notes

UniCredit Research

page 29

See last pages for disclaimer.

26 March 2014

Economics & FI/FX Research

Global Themes Series

Disclaimer Our recommendations are based on information obtained from, or are based upon public information sources that we consider to be reliable but for the completeness and accuracy of which we assume no liability. All estimates and opinions included in the report represent the independent judgment of the analysts as of the date of the issue. We reserve the right to modify the views expressed herein at any time without notice. Moreover, we reserve the right not to update this information or to discontinue it altogether without notice. This analysis is for information purposes only and (i) does not constitute or form part of any offer for sale or subscription of or solicitation of any offer to buy or subscribe for any financial, money market or investment instrument or any security, (ii) is neither intended as such an offer for sale or subscription of or solicitation of an offer to buy or subscribe for any financial, money market or investment instrument or any security nor (iii) as an advertisement thereof. The investment possibilities discussed in this report may not be suitable for certain investors depending on their specific investment objectives and time horizon or in the context of their overall financial situation. The investments discussed may fluctuate in price or value. Investors may get back less than they invested. Changes in rates of exchange may have an adverse effect on the value of investments. Furthermore, past performance is not necessarily indicative of future results. In particular, the risks associated with an investment in the financial, money market or investment instrument or security under discussion are not explained in their entirety. This information is given without any warranty on an "as is" basis and should not be regarded as a substitute for obtaining individual advice. Investors must make their own determination of the appropriateness of an investment in any instruments referred to herein based on the merits and risks involved, their own investment strategy and their legal, fiscal and financial position. As this document does not qualify as an investment recommendation or as a direct investment recommendation, neither this document nor any part of it shall form the basis of, or be relied on in connection with or act as an inducement to enter into, any contract or commitment whatsoever. Investors are urged to contact their bank's investment advisor for individual explanations and advice. Neither UniCredit Bank, UniCredit Bank London, UniCredit Bank Milan, UniCredit Bulbank, Zagrebačka banka, UniCredit Bank Czechia, Bank Pekao, UniCredit Russia, UniCredit Slovakia, UniCredit Tiriac nor any of their respective directors, officers or employees nor any other person accepts any liability whatsoever (in negligence or otherwise) for any loss howsoever arising from any use of this document or its contents or otherwise arising in connection therewith. This analysis is being distributed by electronic and ordinary mail to professional investors, who are expected to make their own investment decisions without undue reliance on this publication, and may not be redistributed, reproduced or published in whole or in part for any purpose. Responsibility for the content of this publication lies with: a) UniCredit Bank AG (UniCredit Bank), Am Tucherpark 16, 80538 Munich, Germany, (also responsible for the distribution pursuant to §34b WpHG). The company belongs to UniCredit Group. Regulatory authority: “BaFin“ – Bundesanstalt für Finanzdienstleistungsaufsicht, Lurgiallee 12, 60439 Frankfurt, Germany. b) UniCredit Bank AG London Branch (UniCredit Bank London), Moor House, 120 London Wall, London EC2Y 5ET, United Kingdom. Regulatory authority: “BaFin“ – Bundesanstalt für Finanzdienstleistungsaufsicht, Lurgiallee 12, 60439 Frankfurt, Germany and subject to limited regulation by the Financial Conduct Authority, 25 The North Colonnade, Canary Wharf, London E14 5HS, United Kingdom and Prudential Regulation Authority 20 Moorgate, London, EC2R 6DA, United Kingdom. Further details regarding our regulatory status are available on request. c) UniCredit Bank AG Milan Branch (UniCredit Bank Milan), Piazza Gae Aulenti, 4 - Torre C, 20154 Milan, Italy, duly authorized by the Bank of Italy to provide investment services. Regulatory authority: “Bank of Italy”, Via Nazionale 91, 00184 Roma, Italy and Bundesanstalt für Finanzdienstleistungsaufsicht, Lurgiallee 12, 60439 Frankfurt, Germany. d) UniCredit Bulbank, Sveta Nedelya Sq. 7, BG-1000 Sofia, Bulgaria Regulatory authority: Financial Supervision Commission (FSC), 33 Shar Planina str.,1303 Sofia, Bulgaria e) Zagrebačka banka d.d., Trg bana Jelačića 10, HR-10000 Zagreb, Croatia Regulatory authority: Croatian Agency for Supervision of Financial Services, Miramarska 24B, 10000 Zagreb, Croatia f) UniCredit Bank Czech Republic (UniCredit Bank Czechia), Na Príkope 858/20, CZ-11121 Prague, Czech Republic Regulatory authority: CNB Czech National Bank, Na Příkopě 28, 115 03 Praha 1, Czech Republic g) Bank Pekao, ul. Grzybowska 53/57, PL-00-950 Warsaw, Poland Regulatory authority: Polish Financial Supervision Authority, Plac Powstańców Warszawy 1, 00-950 Warsaw, Poland h) ZAO UniCredit Bank Russia (UniCredit Russia), Prechistenskaya emb. 9, RF-19034 Moscow, Russia Regulatory authority: Federal Service on Financial Markets, 9 Leninsky prospekt, Moscow 119991, Russia i) UniCredit Bank Slovakia a.s. (UniCredit Slovakia), Šancova 1/A, SK-813 33 Bratislava, Slovakia Regulatory authority: National Bank of Slovakia, Imricha Karvaša 1, 813 25 Bratislava, Slovakia j) UniCredit Tiriac Bank (UniCredit Tiriac), Bucharest 1F Expozitiei Boulevard, RO-012101 Bucharest 1, Romania Regulatory authority: National Bank of Romania, 25 Lipscani Street, RO-030031, 3rd District, Bucharest, Romania POTENTIAL CONFLICTS OF INTEREST UniCredit Bank AG acts as a Specialist or Primary Dealer in government bonds issued by the Italian, Portuguese and Greek Treasury. Main tasks of the Specialist are to participate with continuity and efficiency to the governments' securities auctions, to contribute to the efficiency of the secondary market through market making activity and quoting requirements and to contribute to the management of public debt and to the debt issuance policy choices, also through advisory and research activities. ANALYST DECLARATION The author’s remuneration has not been, and will not be, geared to the recommendations or views expressed in this study, neither directly nor indirectly. ORGANIZATIONAL AND ADMINISTRATIVE ARRANGEMENTS TO AVOID AND PREVENT CONFLICTS OF INTEREST To prevent or remedy conflicts of interest, UniCredit Bank, UniCredit Bank London, UniCredit Bank Milan, UniCredit Bulbank, Zagrebačka banka, UniCredit Bank Czechia, Bank Pekao, UniCredit Russia, UniCredit Slovakia, and UniCredit Tiriac have established the organizational arrangements required from a legal and supervisory aspect, adherence to which is monitored by its compliance department. Conflicts of interest arising are managed by legal and physical and non-physical barriers (collectively referred to as “Chinese Walls”) designed to restrict the flow of information between one area/department of UniCredit Bank, UniCredit Bank London, UniCredit Bank Milan, UniCredit Bulbank, Zagrebačka banka, UniCredit Bank Czechia, Bank Pekao, UniCredit Russia, UniCredit Slovakia, UniCredit Tiriac, and another. In particular, Investment Banking units, including corporate finance, capital market activities, financial advisory and other capital raising activities, are segregated by physical and non-physical boundaries from Markets Units, as well as the research department. In the case of equities execution by UniCredit Bank AG Milan Branch, other than as a matter of client facilitation or delta hedging of OTC and listed derivative positions, there is no proprietary trading. Disclosure of publicly available conflicts of interest and other material interests is made in the research. Analysts are supervised and managed on a day-to-day basis by line managers who do not have responsibility for Investment Banking activities, including corporate finance activities, or other activities other than the sale of securities to clients. ADDITIONAL REQUIRED DISCLOSURES UNDER THE LAWS AND REGULATIONS OF JURISDICTIONS INDICATED Notice to Australian investors This publication is intended for wholesale clients in Australia subject to the following: UniCredit Bank AG and its branches do not hold an Australian Financial Services licence but are exempt from the requirement to hold a licence under the Act in respect of the financial services to wholesale clients. UniCredit Bank AG and its branches are regulated by BaFin under German laws, which differ from Australian laws. This document is only for distribution to wholesale clients as defined in Section 761G of the Corporations Act. UniCredit Bank AG and its branches are not Authorised Deposit Taking Institutions under the Banking Act 1959 and are not authorised to conduct a banking business in Australia. Notice to Austrian investors This document does not constitute or form part of any offer for sale or subscription of or solicitation of any offer to buy or subscribe for any securities and neither this document nor any part of it shall form the basis of, or be relied on in connection with or act as an inducement to enter into, any contract or commitment whatsoever. This document is confidential and is being supplied to you solely for your information and may not be reproduced, redistributed or passed on to any other person or published, in whole or part, for any purpose. Notice to Czech investors This report is intended for clients of UniCredit Bank, UniCredit Bank London, UniCredit Bank Milan, UniCredit Bulbank, Zagrebačka banka, UniCredit Bank Czechia, Bank Pekao, UniCredit Russia, UniCredit Slovakia, UniCredit Tiriac in the Czech Republic and may not be used or relied upon by any other person for any purpose.

UniCredit Research

page 30

26 March 2014

Economics & FI/FX Research

Global Themes Series

Notice to Italian investors This document is not for distribution to retail clients as defined in article 26, paragraph 1(e) of Regulation n. 16190 approved by CONSOB on October 29, 2007. In the case of a short note, we invite the investors to read the related company report that can be found on UniCredit Research website www.research.unicreditgroup.eu. Notice to Japanese investors This document does not constitute or form part of any offer for sale or subscription for or solicitation of any offer to buy or subscribe for any securities and neither this document nor any part of it shall form the basis of, or be relied on in connection with or act as an inducement to enter into, any contract or commitment whatsoever. Notice to Polish investors This document is intended solely for professional clients as defined in Art. 3 39b of the Trading in Financial Instruments Act of 29 July 2005.The publisher and distributor of the recommendation certifies that it has acted with due care and diligence in preparing the recommendation, however, assumes no liability for its completeness and accuracy. Notice to Russian investors As far as we are aware, not all of the financial instruments referred to in this analysis have been registered under the federal law of the Russian Federation "On the Securities Market" dated 22 April 1996, as amended (the "Law"), and are not being offered, sold, delivered or advertised in the Russian Federation. This analysis is intended for qualified investors, as defined by the Law, and shall not be distributed or disseminated to a general public and to any person, who is not a qualified investor. Notice to Turkish investors Investment information, comments and recommendations stated herein are not within the scope of investment advisory activities. Investment advisory services are provided in accordance with a contract of engagement on investment advisory services concluded with brokerage houses, portfolio management companies, non-deposit banks and the clients. Comments and recommendations stated herein rely on the individual opinions of the ones providing these comments and recommendations. These opinions may not suit your financial status, risk and return preferences. For this reason, to make an investment decision by relying solely on the information stated here may not result in consequences that meet your expectations. Notice to UK investors This communication is directed only at clients of UniCredit Bank, UniCredit Bank London, UniCredit Bank Milan, UniCredit Bulbank, Zagrebačka banka, UniCredit Bank Czechia, Bank Pekao, UniCredit Russia, UniCredit Slovakia, or UniCredit Tiriac who (i) have professional experience in matters relating to investments or (ii) are persons falling within Article 49(2)(a) to (d) (“high net worth companies, unincorporated associations, etc.”) of the United Kingdom Financial Services and Markets Act 2000 (Financial Promotion) Order 2005 or (iii) to whom it may otherwise lawfully be communicated (all such persons together being referred to as “relevant persons”). This communication must not be acted on or relied on by persons who are not relevant persons. Any investment or investment activity to which this communication relates is available only to relevant persons and will be engaged in only with relevant persons. Notice to U.S. investors This report is being furnished to U.S. recipients in reliance on Rule 15a-6 ("Rule 15a-6") under the U.S. Securities Exchange Act of 1934, as amended. Each U.S. recipient of this report represents and agrees, by virtue of its acceptance thereof, that it is such a "major U.S. institutional investor" (as such term is defined in Rule 15a-6) and that it understands the risks involved in executing transactions in such securities. Any U.S. recipient of this report that wishes to discuss or receive additional information regarding any security or issuer mentioned herein, or engage in any transaction to purchase or sell or solicit or offer the purchase or sale of such securities, should contact a registered representative of UniCredit Capital Markets, LLC. Any transaction by U.S. persons (other than a registered U.S. broker-dealer or bank acting in a broker-dealer capacity) must be effected with or through UniCredit Capital Markets. The securities referred to in this report may not be registered under the U.S. Securities Act of 1933, as amended, and the issuer of such securities may not be subject to U.S. reporting and/or other requirements. Available information regarding the issuers of such securities may be limited, and such issuers may not be subject to the same auditing and reporting standards as U.S. issuers. The information contained in this report is intended solely for certain "major U.S. institutional investors" and may not be used or relied upon by any other person for any purpose. Such information is provided for informational purposes only and does not constitute a solicitation to buy or an offer to sell any securities under the Securities Act of 1933, as amended, or under any other U.S. federal or state securities laws, rules or regulations. The investment opportunities discussed in this report may be unsuitable for certain investors depending on their specific investment objectives, risk tolerance and financial position. In jurisdictions where UniCredit Capital Markets is not registered or licensed to trade in securities, commodities or other financial products, transactions may be executed only in accordance with applicable law and legislation, which may vary from jurisdiction to jurisdiction and which may require that a transaction be made in accordance with applicable exemptions from registration or licensing requirements. The information in this publication is based on carefully selected sources believed to be reliable, but UniCredit Capital Markets does not make any representation with respect to its completeness or accuracy. All opinions expressed herein reflect the author’s judgment at the original time of publication, without regard to the date on which you may receive such information, and are subject to change without notice. UniCredit Capital Markets may have issued other reports that are inconsistent with, and reach different conclusions from, the information presented in this report. These publications reflect the different assumptions, views and analytical methods of the analysts who prepared them. Past performance should not be taken as an indication or guarantee of future performance, and no representation or warranty, express or implied, is provided in relation to future performance. UniCredit Capital Markets and any company affiliated with it may, with respect to any securities discussed herein: (a) take a long or short position and buy or sell such securities; (b) act as investment and/or commercial bankers for issuers of such securities; (c) act as market makers for such securities; (d) serve on the board of any issuer of such securities; and (e) act as paid consultant or advisor to any issuer. The information contained herein may include forward-looking statements within the meaning of U.S. federal securities laws that are subject to risks and uncertainties. Factors that could cause a company’s actual results and financial condition to differ from expectations include, without limitation: political uncertainty, changes in general economic conditions that adversely affect the level of demand for the company’s products or services, changes in foreign exchange markets, changes in international and domestic financial markets and in the competitive environment, and other factors relating to the foregoing. All forward-looking statements contained in this report are qualified in their entirety by this cautionary statement This document may not be distributed in Canada. EFI e 2

UniCredit Research

page 31

26 March 2014

Economics & FI/FX Research

Global Themes Series

UniCredit Research* Michael Baptista Global Head of CIB Research +44 207 826-1328 [email protected]

Dr. Ingo Heimig Head of Research Operations +49 89 378-13952 [email protected]

Economics & FI/FX Research Erik F. Nielsen, Global Chief Economist +44 207 826 1765 [email protected]

Economics & Commodity Research

EEMEA Economics & FI/FX Strategy

Global FI Strategy

European Economics

Gillian Edgeworth, Chief EEMEA Economist +44 207 826-1772 [email protected]

Michael Rottmann, Head, FI Strategy +49 89 378-15121 [email protected]

Artem Arkhipov, Head, Macroeconomic Analysis and Research, Russia +7 495 258-7258 [email protected]

Dr. Luca Cazzulani, Deputy Head, FI Strategy +39 02 8862-0640 [email protected]

Marco Valli, Chief Eurozone Economist +39 02 8862-0537 [email protected] Dr. Andreas Rees, Chief German Economist +49 69 2717-2074 [email protected] Stefan Bruckbauer, Chief Austrian Economist +43 50505-41951 [email protected] Tullia Bucco, Economist +39 02 8862-0532 [email protected] Chiara Corsa, Economist +39 02 8862-0533 [email protected] Dr. Loredana Federico, Economist +39 02 8862-0534 [email protected] Chiara Silvestre, Economist [email protected] Daniel Vernazza, Ph.D., Economist +44 207 826-7805 [email protected] US Economics Dr. Harm Bandholz, CFA, Chief US Economist +1 212 672-5957 [email protected] China Economics Nikolaus Keis, Economist +49 89 378-12560 [email protected] Commodity Research Kathrin Goretzki, Economist +49 89 378-15368 [email protected] Jochen Hitzfeld, Economist +49 89 378-18709 [email protected]

Anca Maria Aron, Economist, Romania +40 21 200-1377 [email protected] Anna Bogdyukevich, CFA, Russia +7 495 258-7258 ext. 11-7562 [email protected] Dan Bucşa, Economist +44 207 826-7954 [email protected] Hrvoje Dolenec, Chief Economist, Croatia +385 1 6006 678 [email protected] Ľubomír Koršňák, Chief Economist, Slovakia +421 2 4950 2427 [email protected] Catalina Molnar, Chief Economist, Romania +40 21 200-1376 [email protected] Marcin Mrowiec, Chief Economist, Poland +48 22 524-5914 [email protected] Carlos Ortiz, Economist, EEMEA +44 207 826-1228 [email protected]

Chiara Cremonesi, FI Strategy +44 207 826-1771 [email protected] Elia Lattuga, FI Strategy +39 02 8862-0538 [email protected] Kornelius Purps, FI Strategy +49 89 378-12753 [email protected] Herbert Stocker, Technical Analysis +49 89 378-14305 [email protected]

Global FX Strategy Dr. Vasileios Gkionakis, Global Head, FX Strategy +44 207 826-7951 [email protected] Armin Mekelburg, FX Strategy +49 89 378-14307 [email protected] Roberto Mialich, FX Strategy +39 02 8862-0658 [email protected]

Mihai Patrulescu, Senior Economist, Romania +40 21 200-1378 [email protected] Kristofor Pavlov, Chief Economist, Bulgaria +359 2 9269-390 [email protected] Pavel Sobisek, Chief Economist, Czech Republic +420 955 960-716 [email protected] Dmitry Veselov, Ph.D., Economist, EEMEA +44 207 826-1808 [email protected]

Publication Address UniCredit Research Corporate & Investment Banking UniCredit Bank AG Arabellastrasse 12 D-81925 Munich Tel. +49 89 378-18927

Bloomberg UCCR Internet www.research.unicreditgroup.eu

*UniCredit Research is the joint research department of UniCredit Bank AG (UniCredit Bank), UniCredit Bank AG London Branch (UniCredit Bank London), UniCredit Bank AG Milan Branch (UniCredit Bank Milan), UniCredit Bulbank, Zagrebačka banka d.d., UniCredit Bank Czech Republic (UniCredit Bank Czechia), Bank Pekao, ZAO UniCredit Bank Russia (UniCredit Russia), UniCredit Bank Slovakia a.s. (UniCredit Slovakia), UniCredit Tiriac Bank (UniCredit Tiriac).

EFI 3

UniCredit Research

page 32