Sovereign haircuts / SRB

Serbia and Montenegro

6 concluded restructurings, between 1895 and 2004. The earliest default they settle began in 1895. The worst cost creditors 70.9% of the present value of their claim, in 2004. The longest gap between a default and its settlement was 35 years. The file records these episodes under 2 country names (Yugoslavia, Serbia and Montenegro); the page is titled with the most recent.

Restructurings

15.9%

Median present-value haircut

70.9%

Worst present-value haircut

4 of 6

Cut face value by zero

$31.81B

Debt treated (2020 $)

Every restructuring

Two measures of the same deals, never combined. The present-value haircut discounts what creditors got against what they were owed; the face-value reduction counts only principal written off.

Present-value haircutFace-value reduction

Default	Settled	Years to settle	Present-value haircut	Face-value reduction	Debt treated (2020 $)	Source
1895	1895	0	17.3%	0.0%	$1.63B	Meyer, Reinhart and Trebesch (2022)
1932	1967	35	0.0%	73.1%	$1.10B	Meyer, Reinhart and Trebesch (2022)
1983	1984	1	-1.3%	0.0%	$4.78B	Cruces and Trebesch (2013)
1983	1985	2	14.5%	0.0%	$7.47B	Cruces and Trebesch (2013)
1983	1988	5	19.7%	0.0%	$13.23B	Cruces and Trebesch (2013)
1992	2004	12	70.9%	59.3%	$3.60B	Cruces and Trebesch (2013)

Source: Meyer, Reinhart & Trebesch (2022), Sovereign Bonds since Waterloo | Cruces & Trebesch (2013), Sovereign Defaults: The Price of Haircuts | Asonuma & Trebesch (2016) Debt treated is the amount restructured, deflated to constant 2020 US dollars in the source file. A negative haircut means the new instruments were worth more than the old claim. Methodology

Bond prices

1 of 6 Serbia and Montenegro restructuring episodes has Waterloo prices during the restructuring interval

The haircut file dates episodes by year. Waterloo matches them by country code and plots a padded window from one year before default through one year after settlement.

The public Waterloo bond panel in this repo runs from Jan 1, 1822 to Dec 1, 1980. This selector uses price only, never pbid or pask, and shows one line per bond from one year before default to one year after settlement. The author README defines price as “bond price in GBP and USD”; it does not document normalization to face value. Across all 329 haircut episodes, 111 have no Waterloo country-code match and 177 start after Dec 1, 1980. Those two counts overlap. 4 of 6 episodes start after the panel ends. 1 episode overlaps the panel but still has no non-null price observations in that window.

Episode

Selected episode: default 1932, settled 1967, case 10, spell SRB_1932-1967. Window: Jan 1, 1931 to Dec 31, 1968.

USD bond prices

Independent USD y-axis; values are not scaled against another currency.

Bond legend (USD)

7% External Gold Loan of 1922, due 1962 (series B) (USD)

8% External Gold Loan of 1922, due 1962 (series A) (USD)

7% State Mortgage Bank of Yugoslavia of 1927, due 1957 (USD)

1,086 monthly observations across 3 bonds. 1,060 of those observations fall inside the shaded restructuring interval. Earliest point in this window: Jan 1, 1931. Latest: Oct 1, 1968. One line equals one bond; no cross-bond average is computed.

Bond	Currency	ISIN	Obs.	First	Last
7% External Gold Loan of 1922, due 1962 (series B)	USD	not recorded	437	Jan 1, 1931	Oct 1, 1968
8% External Gold Loan of 1922, due 1962 (series A)	USD	not recorded	433	Jan 1, 1931	Jul 1, 1968
7% State Mortgage Bank of Yugoslavia of 1927, due 1957	USD	not recorded	216	Feb 1, 1949	Nov 1, 1967

Each option includes its case id and default-spell id. The chart includes every priced bond for that country in the window, so it does not identify bonds belonging to the selected restructuring. The one-year padding is a display window, one year either side; an editorial choice rather than a source-defined threshold. It uses waterloo_bonds.price only; pbid and pask are not averaged into it. The author README defines price as “bond price in GBP and USD” and does not document face-value normalization. GBP and USD observations use separate labelled panels and y-axes. Of the 329 episodes in the haircut file, 111 have no Waterloo country-code match and 177 start after 1980-12-01; these counts overlap. Later episodes show an absence message rather than an implied zero.

Source: Meyer, Reinhart & Trebesch, Sovereign Bonds since Waterloo (bond-price panel) | Harvard Dataverse, Sovereign Bonds since Waterloo Methodology

Default spells

Spell	From	To	Years	Restructurings
SRB_1895-1895	1895	1895	1	1
SRB_1932-1967	1932	1967	36	1
SRB_1983-1988	1983	1988	6	3
SRB_1992-2004	1992	2004	13	1

Source: Meyer, Reinhart & Trebesch (2022), Sovereign Bonds since Waterloo | Cruces & Trebesch (2013), Sovereign Defaults: The Price of Haircuts | Asonuma & Trebesch (2016) Duration counts both endpoints. The spells table also carries a cumulative-haircut column; it is not published here because it does not reconcile with the per-episode haircuts, see the methodology. Methodology

The Cruces-Trebesch record: a third measure and the discount rate

The Cruces-Trebesch file carries 5 restructurings here, and adds the market haircut (the new instruments against the face value of the old claim, undiscounted) and the exit yield used to do the discounting. The deal label is the file’s own and is finer than the country: it names the instrument class the deal covered. The two files were assembled separately, so a deal here need not line up one-to-one with a row above. 4 of them are filed in that file under iso3 YUG rather than SRB, and are shown here because that code has no page of its own.

Deal	Date	Present value	Market	Face value	Exit yield	Structure	Data quality
Yugoslavia	1983-09-01	6.5%	6.5%	0.0%	14.7%	-	2 / 5
Yugoslavia	1984-05-01	-7.5%	-7.5%	0.0%	14.7%	-	2 / 5
Yugoslavia	1985-12-01	14.5%	18.5%	0.0%	14.5%	-	3 / 5
Yugoslavia	1988-09-01	19.7%	22.1%	0.0%	14.3%	-	2 / 5
Serbia and Montenegro	2004-07-01	70.9%	73.2%	59.3%	9.7%	-	3 / 5

The two files disagree here. For 1984, the long-run file records a present-value haircut of -1.3% and the Cruces-Trebesch file records -7.5%. Both are shown above. They are not averaged, and no third number is constructed from them.

Source: Cruces & Trebesch, haircut dataset (2014 update) Data quality is the file's own 1-to-5 index of how well the deal terms are documented. Methodology

Debt still in default

$240.0M across 1 creditor class in 2024. A haircut can only be measured once a restructuring concludes, so any of this that is still being negotiated is by construction absent from the tables above.

Creditor class	In default, 2024
Domestic (fiscal) arrears	$240.0M

Source: Bank of Canada-Bank of England Sovereign Default Database (CRAG) Bank of Canada terms (attribution). Includes domestic arrears, so it is not comparable with the debt-treated column above. Methodology

Restructurings with Chinese creditors

2 restructurings with Chinese creditors are recorded here. No haircut is attached to any of them, so they cannot be placed on the scale above; the file records only whether face value or the interest rate was cut.

Year	Type	Creditor	Terms
2003	Face Value Reduction	Sinochem	In 2003, Sinochem International and Naftna Industrija Srbije (both SOEs) signed a debt relief agreement. The agreement cancelled 70 mn USD in outstanding payment obligations and rescheduled the remaining 196 mn USD in debt to be repayable in 16 semi-annual installments of approximately $14.2 million over an eight year period from June 30, 2004 to December 31, 2011. The interest rate on these outstanding obligations were reduced from 4.4% to LIBOR plus a 0.7% margin.
2009	Face Value Reduction	Multiple	China cancels and reschedules several outstanding debts of Serbia, incl. claims of SAFE on the National Bank of Serbia, claims by China Ex-Im Bank on state-owned Serbian banks (with sovereign guarantees) and claims by China Ex-Im Bank on the central government. All of these debts date back to 1999, when Chinese entities provided large financial support to Serbia.

Source: Horn, Reinhart & Trebesch, China's overseas lending and debt restructurings Methodology

Methodology, the measures, and what this data cannot tell you