Sovereign Defaults: The Price of Haircuts Juan Cruces Univ. - PowerPoint PPT Presentation

Sovereign Defaults: The Price of Haircuts Juan Cruces Univ. Torcuato Di Tella Christoph Trebesch University of Munich and CESIfo Debt Crisis Conference, Reykjavik University, 08.10.2011 1

INTROD ODUCTION ON • Theory predicts exclusion from capital markets (Eaton and Gersovitz 1981, Arellano 2008, Yue 2010) • But 30 years of empirical research finds small effects • Puzzle: Do markets really forgive and forget? • Our contribution 1) Compute “haircut” estimates 1970-2010: 180 restructuring cases in 68 countries 2) We find indication that effects of defaults are increasing in Haircut size. Effects “larger” and last longer 2

HAIRCUTS IN 180 R RESTRUCT CTURINGS 100 YEM BOL TGO NIC ETH SEN GUY CO MOZ STP MRT BIH GUY ZMB IRQ SLE UGA TZA GIN CMR CMR NER HND 80 ALB BOL ARG HND CRI YUG PER CIV 60 POL Haircut in % MEX BGR MDA JOR NIC VNM RUS SDN MDG DOM POL POL ZAR DOM GMB MOZ CUB NIC KEN RUS RUS NER TGO PHL ECU CUB JAM 40 NGA CUB MAR PHL NIC MDG NGA POL ECU MWI VEN CRI ZAR PAN POL NER ZAR MDA MKD ZAR GRD SEN CRI ARG LBR JAM CHL MEX ROM PRY BRA ROM JAM SEN ZAR NGA ZAR RUS ARG BRA SEN MWI URY ZAR PHL POL GIN URY BLZ NIC DZA MAR POL ZAF ARG 20 TUR URY MAR BRA YUG TUR BRA MEX TUR NGA MDG TTO CHL GAB UKR JAM YUG PAN ECU PHL CHL ROM PAN ZAF PAK MDG JAM UKR HRV DOM POL PAK URY DZA ZAF GAB CHL YUG VEN TUR PER ECU MEX ECU DOM JAM SVN JAM BRA CHL MEX URY NGA BOL NGA MEX 0 NGA PER PER BRA YUG 3 1975m1 1980m1 1985m1 1990m1 1995m1 2000m1 2005m1

HAIRC RCUTS 1970 1970-20 2010: 10: D DESCRI RIPTIVES Obs. Mean Std. Dev. Min Max By Type of Estimate Sturzenegger Zettelmeyer Haircut ("preferred") 180 37.04 27.28 -9.80 97.00 Market Haircut 180 40.01 27.02 -9.80 97.00 Write Off (of Face Value) 180 16.77 30.55 0.00 97.00 By Type of Creditor Bank Debt Restructuring 162 37.05 27.90 -9.80 97.00 Bond Debt Restructuring 18 36.97 21.60 4.70 76.80 Rescheduling vs. Debt Reduction Rescheduling Only 123 24.15 16.67 -9.80 73.20 With Reduction in Face Value 57 64.84 24.94 -8.30 97.00 By Era Pre-Brady (1970-1989) 99 25.57 18.83 -9.80 92.70 Brady Era (1990-1997) 48 51.81 28.48 3.30 92.30 Post-Brady (since 1998) 33 49.96 31.30 -8.30 97.00 4

TW TWO HY O HYPOT OTHES ESES ES Higher haircuts are associated with 1) Higher borrowing costs later  Bond Spreads 2) Longer periods of market exclusion  Duration until re-access Restructuring Yiel eld Spread eads ?  Haircut in Negotiations n % Exclusion on Time me? t+1 t+… t-2 t-1 Default 5

1) HAI HAIRCU CUTS AN S AND BOR ORROWING COS COSTS TS • Dependent variable: EMBI / EMBIG stripped yield spread over US Treasury. • Monthly panel regression with time and country fixed effects. • Sample: all 47 countries in the EMBI/EMBIG index. (24 defaulters + 23 non defaulters) 1993-2010. 5000+ obs. • Explanatory variables : Haircuts, Default Dummies, Debt/GDP, Budget Balance, Real Growth, Inflation, Political Risk, Reserves/Imports, High-Yield Spread, Credit Rating Residual. 6

HAIRCUTS AN HAI S AND POS OST-RESTRUCT CTURING S G SPREADS Mean stripped EMBIG spread differential between defaulters and non defaulters after restructuring for groups of countries (by haircut level) 800 700 600 EMBIG Spread 500 400 300 200 100 0 0 1 2 3 4 5 6 7 Years after restructuring 37% is the median H_SZ among defaulters All_defaulters_mean H<37% H>37% 7

HAI HAIRCUTS AN TS AND POS OST-RESTRUCT CTURING S G SPREADS Lag between DEFAULT DUMMY HAIRCUTS DEFAULT DUMMY AND HAIRCUTS spread and (and controlling for all other determinants of time of (1) (2) spreads) restructuring (7) Coefficient Effect of Coefficient Effect of Effect of Coeff. & SE Coeff. & SE Effect of Effect of & Std.Err. default (bp) & Std.Err. Mean_H SD_H (bp) on Default on Haircut Mean_H SD_H (bp) Dummy 263** 263 6.75*** 100 1.39 1 year lag 270 149 155 (99.99) (2.15) (227.14) (3.93) (10% signif.) 151** 151 4.73*** 107 -1.06 2 year lag 189 104 (72.25) (1.79) (159.82) (3.22) 104 3.89** -124 3.22 3 year lag 155 85 (82.07) (1.87) (104.58) (2.68) 52 3.16** -184** 5.03*** 4&5 year lag 127 70 111 (63.68) (1.38) (73.49) (1.27) -56 0.80 -283** 7.42*** 6&7 year lag 163 (58.88) (1.41) (75.06) (1.68) N 5,369 5,369 4,269 Adj. R2 0.42 0.42 0.51 Other Year & country effects. Year & country effects. US low-grade Year and country effects, US low-grade corp. yield, controls: US low-grade corp. corp. Yield. Debt/GDP, Budget Balance, Real Growth, Inflation, Political Risk, Reserves/Imports, High- Std. Errors: Robust, clustered on country in all specifications. Yield Spread, Credit Rating Residual 8 Dependent variable: spread in basis points. Sample stats: Mean(H)=40%, St.Dev.(H)=22%

2) HAI 2) HAIRCUTS AN TS AND DURATION OF OF EXCL CLUSI SION • Exclusion = Number of years between the end of a restructuring and the first year of market access. • Primary market issuance (bond/loan on international markets from Dealogic) and/or positive credit flows to the public sector (GDF data). • Robustness check “full access” (net inflows > 1% of GDP). • Sample: 60 “final” restructurings, 56 countries, 1980- 2006. • Semi parametric Cox proportional hazard model. 9

HAI HAIRCUTS AN TS AND REACC ACCES: ES: E EXCLU CLUSION D DURAT ATION Kaplan-Meier Survival Functions for Duration of Reaccess 1.25 All Defaulters Proportion of countries still excluded Haircut > 50% 1 Haircut < 50% 0.75 0.5 0.25 0 0 2 4 6 8 10 12 14 Years of Exclusion 10

HAI HAIRCUTS AN TS AND REACC ACCES: ES: E EXCLU CLUSION D DURAT ATION With With External Country Population Higher H Plain Sovereign Political Financing Funda- Full Model Effect of a 1 and GDP lowers Rating Risk Conditions mentals percentage probab. of (1) (2) (3) (4) (5) (6) (7) point rise in reaccessing Haircut? -0.037*** -0.034*** -0.031*** -0.027*** -0.032*** -0.034*** -0.024*** Haircut (SZ, in %) (0.008) (0.008) (0.007) (0.008) (0.007) (0.008) (0.008) e -.024 -1 = -2.4% 0.068*** Credit Rating (Residual) (0.024) Political Risk 0.037 Std.Dev.( H ) = (ICRG) (0.028) 30% Effect = GDP per capita 0.774*** 0.826*** (log) (0.206) (0.281) e -.024 x 30 - 1 = Population 0.414*** 0.159 - 51% (log) (0.102) (0.189) High-yield -0.132* bond spread (0.080) US Treasury 0.136 50% lower 10-year Bond Yield (0.143) chance of -0.094** -0.071* Primary Balance tapping markets (0.044) (0.038) (in % to GDP) in any given Public Debt -0.031*** -0.021* year (in % to GDP) (0.010) (0.012) All estimations have time and regional effects. Sample size: 322 – 237 dep. on spec. 11

CONCL CLUSION 1. A one standard deviation higher Haircut is associated with a. 150 basis points higher spread in year 1, b. 70 basis points higher in years 4 and 5, c. 50% lower likelihood of re-accessing in any year. 2. New support for reputational theories of sovereign debt (Eaton and Gersovitz,1981). NO NO F FRE REE LUNCH! UNCH! NO NO FRE REE E HA HAIRCUT! 12