>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>

RISK & VALUATION A Bloomberg Professional Service Offering

SRSK <GO>
FRAMEWORK, METHODOLOGY & USAGE
CONTENTS
02 BLOOMBERG SOVEREIGN RISK MODEL — METHODOLOGY
05 CDS MODEL
06 SRSK FUNCTION
08 BIBLIOGRAPHY
INTRODUCTION
The Bloomberg sovereign risk function (SRSK) provides
transparent, quantitative estimates of the 1-year default
probability of a country’s sovereign debt and 5-year CDS
spread. The model inputs are displayed on the screen,
providing transparency into the drivers of sovereign risk.
Major inputs are: deficit levels, foreign currency reserves,
non-performing bank loans, GDP growth and a political risk
indicator. Country financials are obtained from the World
Bank, Eurostat and the IMF. These inputs are override-able,
allowing clients to perform scenario and sensitivity analyses
(e.g., assess effect of changes to major inputs). SRSK
currently covers 85 countries. We will continue to expand
coverage as data becomes available.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Countries are divided into two categories: Reserve Using the strong relationship between the factors that drive
Currency (RC) and Non-Reserve Currency (NRC) countries. default probability and market CDS spreads, we also estimate
RC countries use one of the six world reserve currencies: the 5-year CDS spreads, thus providing a valuable reference point
US dollar, euro, Japanese yen, British pound, Swiss franc and for countries without traded CDS (53 out of 85 countries we
Canadian dollar. We also assign the Scandinavian countries to cover do not have traded CDSs). Finally, countries are assigned
the RC category. All other countries are NRCs. a default risk measure as a high-level summary of their 1-year
sovereign risk using an explicit mapping from default probability
to default risk (low, medium or high risk).
BLOOMBERG SOVEREIGN
RISK MODEL — METHODOLOGY
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Definition of Default MULTI-FACTOR MODEL FOR SOVEREIGN RISK
Our initial list of defaulted countries comes from an MIT
press release “Sovereign Defaults and Debt Restructurings: Countries Unique Variables Common Variables
Historical Overviews.” The original source is Lindert and Morton » Reserve » Government » GDP Growth
(1989), Beim and Calomiris (2001) and Standard & Poor’s. Currency Surplus » EIU Political
The list contains defaults and/or debt restructurings (including » Refinancing Ability Risk Score
reschedulings) involving external creditors. Payment delays and » Non-Reserve » Reserve-to- » Non-Performing
other technical defaults that eventually resulted in full repayment Currency External-Debt Bank Loans
are excluded. We have added bailouts collected from the
general media to our list of defaults. Our default coverage Figure 2: Input Variables
includes 85 defaults and bailouts from 1972 until 2012. Unique Variables
Drivers of Default Reserve currency countries issue debt only in their own
currency. Therefore, distance-to-default is driven by government
surplus—the difference between government revenues and
Default Coverage obligations. The surplus, as we define it, represents funds
12
left over for paying interest expense. Government surplus is
Number of Defaults

10
calculated as:
8
Govt Surplus = (Govt Revenues–Govt Expenditure) —
6
(Debt Due in the next 12 Months + 20%M Govt L.T. debt)
4
= Govt Net Revenues — (Debt Due in the next 12 Months +
2
20% Govt L.T. debt)
0
78 80 82 84 86 88 90 92 94 96 98 00 02 04 06 08 10 12 The majority of the debt issued by non-reserve currency
19 19 19 19 19 19 19 19 19 19 19 20 20 20 20 20 20 20 countries that is accessible to international investors is external
Year
debt. External debt is issued in one of the reserve currencies,
Figure 1: Default Coverage mainly the US dollar, the euro, or the British pound. Therefore,
distance-to-default is driven by the difference between the
Countries are divided into two categories: Reserve Currency stock of foreign reserves and external debt obligations. We
(RC) and Non-Reserve Currency (NRC) countries. Each define the ratio of reserves to external debt as:
category has unique as well as common drivers. Figure 2
summarizes the drivers. Reserves Excluding Gold
Reserve Ratio =
S.T. External Debt + 30% L.T. External Debt

Sovereign Defaults and Debt Restructuring: Historical Overview, MIT Press.

1

Peter H. Lindert & Peter J. Morton (1989), How Sovereign Debt Has Worked, NBER Chapters, in: Developing Country Debt and the World Economy, pp. 225-236 National Bureau of Economic Research, Inc.
2

David O. Beim and Charles W. Calomiris (2001), Emerging Financial Markets, New York: McGraw Hill.
3

BLOOMBERG SOVEREIGN DEFAULT RISK // 02

>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Common Variables
Estimating the ability to meet debt obligations through a surplus After correctly ordering countries based on default risk, the
measure for RC countries and reserve ratios for NRC countries model should also predict the default rate with reasonable
is necessary but not sufficient to assess credit health. Growth is accuracy. To test this, we perform a Goodness-of-Fit test by
also an important factor, since it helps predict whether the gap plotting model-implied default rates against realized rates. We
between assets and liabilities is widening or decreasing. Growth also evaluated our model’s overall ability to provide early warning
affects debt capacity, the cost of debt and, ultimately, default of default through event studies.
probability. We include GDP growth as a driver of default for all
Accuracy Ratio (AR) Test
countries.
The accuracy ratio measures the quality of the default ordering
Another variable common to all countries is the health of the produced by the model. The model is penalized both for
banking sector. During 2012, both Belgium and Ireland had assessing safe countries as too risky and vice versa (type I
similar deficit levels. However, SRSK considered Ireland to and II errors). The AR test is performed as follows:
have a larger probability of default than Belgium because of the
» Rank countries by decreasing default probability.
relative health of their banking sectors—Belgium's was better.
We capture this effect by including, as a default driver, the ratio » Draw countries in order of decreasing rank without
of non-performing bank loans to total loans. replacement. On the y-axis, mark the percentage of defaulted
countries that have been removed and, on the x-axis, mark
All variables so far capture financial and economic risk.
the percentage of the total number of countries that have
In addition, as a proxy for political/social risk, we use the
been removed.
Economist Intelligence Unit (EIU) Political Risk Score. This
evaluates a range of political factors relating to political » After completing the draw, plot the curve (Cumulative
stability and effectiveness that could affect a country’s Accuracy Profile). Compute the area under the curve. This is
ability to meet and commitment to servicing its debt the Accuracy Ratio. For a model that is generating random
obligations—both of which could cause turbulence in the results, the Cumulative Accuracy Profile is a 45 degree line.
foreign-exchange market. This Risk Score is pertinent to For a perfect model, the line would rise almost vertically
sovereign, currency and banking sector risks. The Risk (the exact angle would depend on the overall default rate)
Scores are measured on a scale of 0–100. Higher scores until it intersects the 100% level and then would form a
indicate a higher level of risk. The effect of recent instability horizontal line.
in Egypt on default risk can only be captured by adding soft In-sample accuracy tests indicate that our model has around
factors. All information, with the exception of the EIU Political 89% accuracy ratio (Figure 3).
Risk Index, is derived from the World Bank, IMF and Eurostat.
Comprehensive Statistical Tests
The primary focus of SRSK is to discriminate between defaulting
and non-defaulting countries. To this end, we evaluate its
performance by calculating the Accuracy Ratio, or Cumulative
Accuracy Profile. This test quantifies a model’s ability to identify
defaulting countries as more risky than non-defaulting countries.

BLOOMBERG SOVEREIGN DEFAULT RISK // 03

BLOOMBERG SOVEREIGN
RISK MODEL – METHODOLOGY
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Goodness-of-Fit Test Early Warning Performance Event Study
The goodness of fit was tested by We have complete model inputs for 69 of the 85 defaults
and bailouts in our default database. Defining year 0 as the
» Grouping country-level observations of default by
event year, model estimated default probabilities are plotted
default probability.
starting nine years before default and ending nine years after
» Plotting the calculated default probabilities on the X axis default (Figure 5). Model default probabilities start to rise
and the actual default rates from default data on the Y axis, significantly two-to-three years before default, providing early
e.g., ex-ante versus ex-post default probabilities. warning of deteriorating financial and economic conditions in
the defaulting countries.
The curve generated (Figure 4) lies close to the 45 degree line,
indicating that the model estimation is free of bias.
9%
8%
100%
7%
Model Default Probability
% of Defaulting Countries

6%
80%
5%
4%
60%
3%
2%
40%
1%
0%
20%
-10yr -5yr 0yr 5yr 10yr
0%
Event Window
0% 20% 40% 60% 80% 100%
% of Total Countries Figure 5: Event Study

Figure 3: In-Sample Cumulative Accuracy Profile (Accuracy Ratio Test)

12%

10%
Emprical Default Probability

8%

6%

4%

2%

0%
0% 2% 4% 6% 8% 10% 12%
Model Default Probability

Figure 4: Realized versus Model Default Probability

BLOOMBERG SOVEREIGN DEFAULT RISK // 04

CDS
MODEL
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
The SRSK model provides implied CDS spreads for all 85 » Political/social risks are high in the country. These risks
covered countries, while only 53 of these countries have might not be fully captured by the EIU risk score.
market-traded CDS. The SRSK drivers of default probability
» Countries with unique situations such as Argentina
are the inputs for the econometric model of CDS spreads.
where continuing litigation by bondholders keeps CDSs
The SRSK output is “real-world,” “physical” or “P-measure,”
at elevated levels.
whereas CDS-implied default probabilities are “risk-neutral”
or “Q measure.” The risk premium inherent in the conversion » Relationship between debt capacity and indebtedness is
between the two measures is implicit in our regression different from the average trend. Two countries—the U.S.
framework, allowing the market data to specify the and Japan—have superior access to debt capital compared
appropriate conversion. with other countries with similar or even lower levels of debt-to-
GDP. We attribute this to the special safe haven status of their
For non-reserve currency countries, the model predicts market
currencies and their quantitative easing policies.
CDS spreads using the same inputs (reserve-to-external-debt
ratio, non-performing bank loans, EIU Political Risk Score and We wish to re-emphasize that, in all these cases, we leave
GDP growth) as the DP model. Similarly, for reserve currency the model output untouched—our aim is not to match the
countries, market CDS spreads are explained by default market but to provide an independent, quantitative method
probability drivers (government surplus, non-performing of assessing country default risk—even if that leads to
loans, GDP growth and refinancing ability). significant model-to-market CDS mismatch (and, implicitly,
default probability).
For some countries, the SRSK default probability and
model CDS are not consistent with market belief. Next Correlation Between Model and Market CDS
to these country names, we have placed information icons. We test the performance of our model by doing correlation
Clicking on the icon provides an explanation for the unique tests with market spreads. For reserve currency countries,
situation of that country. The information is critical to the implied CDS spread has a 98% Pearson correlation
understanding model output. and a 95% rank correlation with actual spreads. For
non-reserve currency countries, implied CDS spread has a
There are several possible reasons for the differences
75% Pearson correlation and an 81% rank correlation with
between model and market CDS:
actual spreads. Overall, these numbers suggest a strong
» The market is mispricing the level of sovereign risk. relationship between SRSK model-implied CDS spreads
and market spreads. Once again, we are not trying to match
» SRSK is mispricing the level of sovereign risk — latest financial
market CDS, but, rather, we wish to estimate an intrinsic
data reported by the country (and used by SRSK) is outdated
CDS given the level of default probability for a country.
(e.g., non-performing loan levels being reported in the media
and being priced into the CDS are significantly different from
the last available World Bank data for the country).

BLOOMBERG SOVEREIGN DEFAULT RISK // 05

SRSK FUNCTION
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
The Bloomberg Default Probability Model and the CDS
Spread Model can be accessed through SRSK <GO>.
This section describes the function and its capabilities.
Main Page
The main page is divided into three sections: Country list,
Default Risk section and CDS section. The drop-down
menu allows the user to select the category of countries.
The country categories are:
» All
» Americas
» Asia/Pacific
» Central and Eastern Europe
Under the Default Risk (%) section, there are five columns.
» Middle East and Africa The first column is titled—“Current”—and shows current
» Western Europe 12-month default probability. Default probabilities are assigned
three different colors: red, yellow and green—which correspond
» Developed to high-, medium- and low-risk categories. Click on the detailed
» Less-Developed legend at the bottom left to see definitions of these risk
categories. For reserve currency countries, the high-risk
» Emerging category corresponds to a larger-than-2% default probability
» Eurozone (DP); medium-risk corresponds to a DP between 1% and
2%; and the low-risk category coincides to a DP of less than
» Non-Eurozone 1%. For non-reserve currency countries, these cut-off points
» Highest 1-Yr DP (%) are 0.5% and 1.5%. The cut-off points are determined
empirically. We observe that 90% of historical non-reserve
Each category is divided into Reserve Currency and currency defaults occured when DP > 1.5% and the other
Non-Reserve Currency countries. The first column shows 10% when DP is between 0.5% and 1.5%. For reserve
the list of countries. To the left of the country name, is a currency countries, there have been only five historical
news heat icon. Clicking on this icon takes you to the defaults, but they all occurred when calculated default
specific country’s news page, where the most relevant probabilities were higher than 2%. The difference in cut-off
economic news is shown. points could be explained economically as well. Non-reserve
currency countries rely on foreign reserves to meet their debt
obligations, which makes them vulnerable to currency crises
and to crises in developed countries. Reserve currency
countries, in contrast, can either print money to pay their debt or
are guaranteed or protected by countries that have the power
to print money to meet debt obligations. For example, Greece
is being supported by other Eurozone governments, which
have the power to print money and buy Greek debt. These
differences allow reserve currency countries to have higher
debt capacity than non-reserve currency countries.

BLOOMBERG SOVEREIGN DEFAULT RISK // 06

>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
The second column shows the lowest financial default The Model Inputs section is different for reserve vs. non-reserve
probability that our model estimates for the particular currency countries. For reserve currency countries, four main
country during the selected period of coverage. The next inputs are used: Government Surplus/Deficit (Revenues-
column plots the historical range with current and average Expenditure-Debt due in 12 months—20% of Long-Term Debt)
default probabilities displayed. Since historical data availability as a percentage of GDP, non-performing bank loans (NPL) as
varies from one country to the next, we provide a drop- a % of total loans, Refinancing Ability, GDP growth and the EIU
down menu that allows the user to choose the historical Political Risk Score. The refinancing ability is an estimate of the
range. For reserve currency countries, the choices are: country’s ability to roll over the debt coming due in the next 12
the last five years, the last ten years and all available years. months. For the U.S. and Japan, SRSK assumes that this ability
For non-reserve currency countries, we add the last 20 years is at 100% (all their short-term debts can be rolled over easily).
and the last 30 years. For all other countries, the refinancing ability is set at 0%. The
user can override this percentage to estimate the effect on credit
The fourth column shows the highest financial default
health of the partial or complete loss of this refinancing ability.
probability during the period covered. The fifth column
For non-reserve currency countries, model inputs are: Reserve
shows DP’s change from the beginning to the end of the
Ratio (External Debt Due in next 12 months/(Short-Term Debt
range period.
+30% of Long-Term Debt), non-performing bank loans (NPL) as
The third section displays five-year market CDS spread, a % of total loans, GDP growth and the Economist Intelligence
SRSK’s five-year CDS spread and the difference between Unit (EIU) Political Risk Score. The user can do a sensitivity
these two measures. All such measures are shown in basis analysis by overriding any model inputs and observing the
points. This measure can be used as a starting point for model outputs.
deeper dives to investigate investment ideas.
The Market and Model CDS Spreads section shows the time
Country Page series plots of the five-year market CDS versus Model CDS
From the opening screen, click on a particular country Spreads. “1-Yr Default Probability Vs Model Input” provides
name and you will be taken to that specific country’s page. time series plots of model default probability versus each
The page is divided into five parts: model input. Clicking on the circle to the left of the input
variable under the Model Input sections allows the user to
» Model Outputs
select which input to include in the plot. Finally, the bottom
» Model Inputs section shows related economic news for the particular
» Market and Model CDS Spreads country. Clicking on the Related News title takes the user
to an expanded news page.
» Default Probability and “Model Input”
» Related News
In the model outputs section, model CDS spreads, market
CDS spreads and model default probability (DP) are
displayed. In addition, the country’s risk category (low,
medium or high) is displayed in a risk scale.

BLOOMBERG SOVEREIGN DEFAULT RISK // 07

BIBLIOGRAPHY
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Beim, David O., and Charles W. Calomiris. Emerging Financial
Markets. New York: McGraw Hill, 2001.
Lindert, Peter H., and Peter J. Morton. “How Sovereign Debt
Has Worked.” In Developing Country Debt and the World
Economy, 225–236. Cambridge, MA: National Bureau of
Economic Research, Inc., 1989.
Sovereign Defaults and Debt Restructurings:Historical
Overview. Cambridge, MA: MIT Press.

BLOOMBERG SOVEREIGN DEFAULT RISK // 08

>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
BEIJING FRANKFURT LONDON NEW YORK SÃO PAULO SYDNEY
+86 10 6649 7500 +49 69 9204 1210 +44 20 7330 7500 +1 212 318 2000 +55 11 2395 9000 +61 2 9777 8600

DUBAI HONG KONG MUMBAI SAN FRANCISCO SINGAPORE TOKYO

+971 4 364 1000 +852 2977 6000 +91 22 6120 3600 +1 415 912 2960 +65 6212 1000 +81 3 3201 8900

[Link] The data included in these materials are for illustrative purposes only. ©2015 Bloomberg L.P. All rights reserved. S601325776 DIG 0615