Sovereign Bonds Since Waterloo

The paper asks a question the sovereign debt literature has mostly approached from the borrower’s side: given how often governments default, why do investors keep buying their bonds? It answers by measuring what creditors actually earned, assembling two new datasets and matching them bond by bond. The first is monthly price quotations for 1,552 foreign-currency sovereign bonds issued and traded in London and New York between 1815 and 2016 – 266,134 observations covering up to 91 countries in an unbalanced panel. The second is an archive of external default and restructuring events built largely from the annual reports of nineteenth- and early-twentieth-century bondholder organisations, yielding haircut estimates for 313 restructuring events in 91 countries and, crucially, the timing and size of missed or partial coupon payments at monthly frequency. The central finding is that the average real ex-post yearly return on a global portfolio of external sovereign bonds was 6.85% – about 4 percentage points above the “risk-free” benchmark of long-term UK and US government bonds, with the excess return running 2% to 4% depending on the era. Two things make that survive the defaults. First, defaults do not wipe creditors out: the average haircut is 44% (39% when weighted by amount restructured), with a standard deviation of about 30% and no visible time trend across 200 years, and outright repudiation is confined to revolutions and imperial break-ups. Second, roughly 70% of the 8.0% average nominal return – 5.6 percentage points – comes from coupons rather than capital gains, so returns keep accruing even while prices are depressed. The risk is real and priced: bonds of the 51 “serial defaulters” earn the highest returns (7.1% real, 4.6% excess) and also the highest volatility; after a default the cumulative return index falls about 15%, and an investor entering two years before default breaks even about four years after it, though the lower quartile of episodes never recovers within six years. Compared with other asset classes over the same two centuries, only US equities and a 16-country advanced-economy equity portfolio returned more, while the external sovereign bond portfolio’s Sharpe ratio is on a par with US equities and above US corporate bonds, UK equities, and domestic sovereign bonds. The authors are explicit about the scope limits: the sample is unbalanced with a near-total gap in the 1970s and 1980s syndicated-bank-loan era, so the 1980s debt crisis is largely absent; and they caution that the unusually good modern performance should not be read as a “new normal.”

Sovereign Defaults: The Price of Haircuts

The paper attacks a long-standing empirical consensus – that sovereign default carries little or no penalty in credit markets – by arguing that the consensus rests on a measurement choice. Earlier work coded credit history with a binary default indicator, “capturing any missed payment,” which throws away the enormous variation in how much creditors actually lose. The authors therefore build the first complete set of present-value haircut estimates for all 180 sovereign debt restructurings with foreign banks and bondholders between 1978 and 2010, covering 68 countries, assembled from nearly 200 sources including IMF archives, offering memoranda, private-sector research and the financial press, and discounted using a new procedure that imputes a deal-specific “exit yield” from low-grade US corporate yields and the sovereign’s rating at the time. The resulting facts are themselves the paper’s first contribution: the average haircut is 37% (about 30% volume-weighted), half the cases lie below 23% or above 53%, haircuts rose by roughly 25 percentage points on average between the 1970s-80s and the 1990s-2000s, deals with outright face-value write-offs average 65% against 24% for pure reschedulings, and restructurings by highly indebted poor countries average 87% – nearly three times the middle-income figure. The sovereign average is far below the 64% the authors cite for US corporate restructurings, which they find “surprising because US corporate debt, in contrast to sovereign debt, can be enforced in courts.” The second contribution is the link to what happens next. Replicating the standard specification with a binary restructuring dummy reproduces the received result – spreads 260 basis points higher in year one, around 150 in year two, and insignificant or marginal thereafter – but substituting the continuous haircut changes the picture: one extra percentage point of haircut goes with EMBI Global spreads about 6.75 basis points higher in year one and still about 3.16 basis points higher in years four and five, so a one-standard-deviation (22 percentage point) increase implies spreads 149 basis points higher in year one and 70 higher in years four to five. In the fully specified model that includes both the haircut and the restructuring dummies, the incremental spread of a restructuring is statistically significant for haircuts above 40% throughout years one to seven, and a one-standard-deviation rise in haircut implies spreads 122 basis points higher in years four and five and 149 higher in years six and seven – against the at most 50 basis points earlier studies attributed to a past default. On market access, across 67 “final” restructurings the average time to partial reaccess is 5.1 years (median three), but 2.3 years for haircuts below 30% against 6.1 years above; a Cox proportional hazard model puts a one-percentage-point higher haircut at 2.37% lower odds of reaccess in a given year, so a one-standard-deviation increase (30 percentage points in that sample) implies “a 51 percent lower likelihood of reaccess in any given year.” The authors are careful about what this does and does not show: they include country and year fixed effects and a large set of fundamentals, but say this “mitigates, but not necessarily completely eliminates” the risk of an omitted confounder, that the findings “should not be interpreted as direct evidence” for either punishment or information revelation, and that what they have is “indicative evidence” of a trade-off rather than an identified channel.

Christoph Trebesch | Macro Paper Warehouse

Sovereign Bonds Since Waterloo

Sovereign Defaults: The Price of Haircuts