Sri Lanka’s Bond Deal Should Not Set a Precedent

Rather than reducing the risk of future debt trouble, Sri Lanka’s macro-linked bonds set up the risk that Sri Lanka will fall back into debt trouble in 2029 or 2030.

Sri Lanka’s macro-linked bonds (MLBs) are, in fact, not really macro-linked bonds.

After 2027, they are standard fixed-income instruments which provide no help in the event of future shocks. The final restructuring terms are just left open until just after Sri Lanka’s IMF program ends, and the creditors and Sri Lanka have agreed that the final terms will largely be a function of Sri Lanka's dollar GDP between 2025 and 2027.

The new bonds thus aren’t true state-contingent instruments.

Payments after 2027 aren’t linked to real GDP growth, tourism inflows, or the price of Sri Lanka’s oil imports. Theorists who extol the advantages of state-contingent instruments shouldn’t be fooled. Bond traders still need to worry about the risk that the payments will reset up based on Sri Lanka’s economic performance between 2025 and 2027—a period when risks are low because Sri Lanka is more or less fully funded by the official sector. Good performance over the next few years thus creates more risk of trouble after 2028, as external debt service ratchets up during a period when Sri Lanka is expected to be back financing in the international bond market with a debt to GDP ratio of around 100%.

The new bonds are thus structured to game the IMF's fiscal targets and increase the return bondholders get out of Sri Lanka’s debt exchange rather than to insulate Sri Lanka against future risks (details of the proposed bonds are here).

The innovation in the macro-linked bonds is that the terms of the exchange aren’t fixed until after Sri Lanka’s current IMF program has expired.

a) If all follows the IMF’s (relatively conservative) macroeconomic and exchange rate forecasts, the bondholders get the base case bond—so $12.5 billion in bonds with an average coupon of around 6.5 percent are swapped into $9 billion of new bonds with an average coupon of around 6 percent, and past-due interest (PDI) is settled with a modest discount. In the base case—with GDP averaging $89 billion over the next year—there is a little bit of true debt reduction, as the coupon on the new bonds is below the coupon on the old bonds.

b) If GDP averages $92 billion over the next three years (as seems likely with dollar GDP at $94 billion in the first quarter of 2024), the stock of the new bonds rises to $10 billion and the average post-2027 coupon increases to 6.6 percent.

c)If GDP averages more than $100 billion between 2025 and 2027, the stock of new bonds increases to $10.65 billion, and the average coupon increases to 8 percent. In that case, there is essentially no long-term debt relief as the higher coupon offsets the lost face value.

In the latest proposal, there is also a bit of downside protection for Sri Lanka. The bonds reset only if the real economy has in fact grown from 2024 to 2027. If GDP falls to $85 billion (well below the IMF’s conservative forecast), the stock of new bonds falls from $9 billion to $7.5 billion.

The risk here should be obvious—a strong Sri Lankan currency could drive nominal dollar GDP higher during the program period (when official inflows and limited debt service provide a stable source of foreign exchange on top of exports) even in the absence of any true increase in Sri Lanka’s external payment capacity. Conceptually, dollar GDP isn’t even the actual source of increased external debt service capacity, as a high calculated dollar GDP can reflect an overvalued exchange rate that makes it harder to generate the actual foreign exchange needed to service external debt, a classic consideration that was left out of the IMF’s “fiscal only” market-access debt sustainability framework.*

It is worth stepping back a bit here to go over the IMF’s debt targets for Sri Lanka.

The targets are set exclusively in fiscal terms, so they define limits on overall government debt rather than external government debt.

The targets also allow a quite high level of debt-to-GDP and foreign currency debt service (the foreign currency debt service limit is a component of the broader limit on Sri Lanka’s forecast gross fiscal financing need): specifically, public debt needs to be under 105 percent of GDP in 2027 and under 95 percent of GDP in 2032, the gross financing need needs to be under 13 percent of GDP on average between 2028 and 2032, and foreign currency debt service needs to be under 4.5 percent of GDP between 2028 and 2032. Given how the targets have been defined, a higher dollar GDP mechanically translates into a higher permitted level of foreign currency debt service (to state the obvious, foreign exchange debt service capacity rises with dollar GDP not with exports or foreign currency reserves).

Broadly speaking, the IMF set extremely generous targets for overall debt to GDP in the program—a low-income country would be expected to get general government debt under 55 percent of GDP by the end of the program, and external debt to GDP down to 30 or 40 percent of GDP. At the end of the program, in 2027, Sri Lanka’s public debt-to-GDP is estimated to be over 105 percent; the binding 95 percent target was set for five years later in 2032.

But the IMF’s initial GDP forecast was as conservative as debt to GDP target was generous. Sri Lanka wasn’t forecast to return to its pre-crisis level of dollar GDP even by the end of its IMF program. The bondholders argued, unsurprisingly, that the low level of GDP in the IMF baseline unfairly limited the available debt service for foreign creditors. The GDP forecast has subsequently been adjusted up (to around $90 billion over the next three years, on average) without any downward adjustment in the debt to GDP targets—which has the effect of reducing the amount of debt relief that Sri Lanka needs from all its external creditors in the baseline.

There seems to be a strong desire from all parties involved to declare victory, collect some fees (the consent fee for the bonds is $225 million), start collecting past-due interest, and move on. The IMF is arguing that the current debt restructuring process is now starting to work.** The Paris Club, India, and the Export-Import Bank of China (Exim) have done their separate deals (on undisclosed terms, and yes, that is a problem, see my paper with Sean Hagan). Market analysts think the bonds should trade at around 60 cents on the dollar, even at a relatively high 11 percent exit yield.

So, my guess is that there will be a lot of pressure to let this deal sail through.

But there are two reasons for concern.

The first is obvious: the macro-linked bonds structure essentially makes a generous debt-to-GDP target even more generous. The structure of the IMF’s debt targets set completely aside what I consider to be the core issue, namely whether $45 billion of debt is too much for an economy that struggles to generate foreign exchange and has no history of collecting tax revenues. In the upside scenario, the coupon on the $10.5 billion in new bonds is approximately $870 million per year, which is higher than the $770 million contractual coupon on Sri Lanka’s defaulted bonds. Sri Lanka’s difficulties are thus being treated by both the IMF and its creditors as problems of extended illiquidity, not a problem of fundamental solvency—a judgment that I question given the rise in Sri Lanka’s gross (and net) external debt over between 2012 and 2022.***

The IMF’s market access framework through doesn’t look at the absolute size of external debt relative to measures of external payment capacity. It only looks at the size of the gross financing need, or more precisely, the size of the gross financing need relative to the size of the domestic banking system. That is a pretty narrow basis for assessing fiscal solvency as, for example, it leaves out interest payments relative to tax revenues, and Sri Lanka’s Achilles’ heel, as Theo Maret has emphasized, has long been its low levels of tax collection. Government revenue was around 10 percent of GDP before the COVID-19 Pandemic, and it fell to ridiculously low levels thereafter, thanks to a poorly timed tax cut and even with the IMF program adjustments was still just under 10 percent of GDP in 2023. The stock of high coupon bonded debt after 2027—and the need to refinance that debt at high debt-to-GDP levels—thus pose a set of classic fiscal risks.

The second issue is one of symmetry. In the upside scenarios, both the stock of bonds and coupon increase. In the downside scenario, the coupon is constant. As a result, the $570 million coupon in the base case (the one used in the IMF debt sustainability analysis) rises to $870 million in the upside scenario and falls only to $470 million in the downside scenario. Looking at all debt service, the upside in 2028 for the bonds is over $500 million (0.45 percent of estimated GDP) while the maximum downside adjustment is less than $200 million (0.2 percent of estimated GDP).

Moreover, the probability distribution for a payout is skewed toward the upside. The bondholders are correct in arguing that the IMF’s forecast for dollar GDP during the program period are a bit conservative. The problem is what happens after the program ends and Sri Lanka, in theory, has to survive on its own in the market with high levels of public external debt and overall public debt.

Absent disclosure of the terms of the China Exim restructuring and the Paris Club/India deal, it is hard to know whether there will be concerns about comparability. The NPV relief provided by the bonds may be significantly higher than the NPV relief provided by the official creditors in the base case.

But the asymmetry in the bondholder deal could still be an issue—as the bonds almost certainly will get a lot more in both the near and medium term if GDP hits the $100 billion threshold. The bonds are also getting most of the near-term cash flow (over $700 million a year) out of a total external debt service envelope for bonds, the China Development Bank (CDB), and bilateral claims of $900-975 million; bilateral creditors and the CDB have more of the restructuring stock—at least $16 billion relative to the $14.5 billion of bonds including PDI—so there is, as usual, a big tilt toward the bonds in the near-term cash flows.

Put simply, “comparability” under the current formula requires more NPV effort from the bonds given the cash flow tilt toward the bonds, which may or may not be the case. We don’t know the NPV relief in the official deal, or how the official sector will assess the greater upside provided in the private deal.

But the most important issue is a broader one. A rise in GDP—if it isn’t just a function of an overvalued exchange rate—frees up additional resources for a host of uses. Under the IMF’s framework, higher dollar GDP increases Sri Lanka’s calculated capacity to service foreign currency debt in a linear way—$10 billion more in GDP means $450 million more in estimated debt service capacity.*** The proposed macro linked bonds give all of that upside to one group of external creditors, at least between 2029 and 2032. The bondholders get something like 80 percent of the notional upside from higher dollar GDP while being at most one third of the overall external debt stock ($14.5 billion of $43 billion).

Put differently, if Sri Lanka’s (dollar) GDP increases by 10 percent (versus the baseline), debt service on external bonded debt rises by 50 percent, not by a more reasonable linear 10 percent.

That raises one final point:

Sri Lanka has a relatively high level of external debt: $43 billion, which will rise to at least $45 billion in the IMF program period, as future Fund disbursements, $2.5 billion in multilateral development bank (MDB) fiscal financing, and the recognition of past-due interest offset the reduction in the bond’s face value. Sri Lanka also has an exceptionally low level of reserves. A $100 billion economy should have around $15 billion of reserves—Sri Lanka now has $5 billion, $3 billion of which are borrowed. Sri Lanka actually defaulted back in 2022 because it ran out of reserves, a fact that is glossed over in the IMF’s fiscally-focused debt sustainability framework.****

Net external debt—a concept that the IMF’s market access framework doesn’t incorporate—is thus high. Sri Lanka has certain inelastic import needs (fuel) together with a modest export base (goods exports are only $10 billion or so; total foreign exchange proceeds are around 20 percent of GDP).*****

Sri Lanka is thus an economy that cannot—in my view—easily support a lot of expensive external debt.

So, I do worry that the IMF assumes a return to external market access in 2027 with a debt-to-GDP level that will be over 100 percent in the IMF’s base case and could be over 90 percent even in an optimistic case. The potential spike in debt service in 2029 also stands out as an obvious source of risk. The macro-linked bonds could easily become high-rate, relatively rapidly-amortizing bonds just when Sri Lanka loses the lifeline from its IMF program.

State-contingent instruments were originally viewed as a mechanism to reduce risk by aligning payments to payment capacity. Lots of academics still find them appealing for that reason. But in practice, they have generally become instruments that add to, rather than reduce, the risk of future default.  

To borrow a phrase from Michael Pettis, the macro-linked bonds create a “volatility machine” just as Sri Lanka (in theory) is set to exit from its IMF program.

That is a problem.

* IMF staff have resisted suggestions that they should review their current debt framework and do a better job of of incorporating external debt variables into the medium-term module of the new market access debt sustainability framework.  External variables actually enter into the confidential short-term assessment of risk, but the medium-term assessment defines parameters for a debt restructuring—the long-term module appears to have little substantive value. I also worry, more generally, that the IMF staff has lost interest in balance of payments issues, in part because the balance of payments has dropped out of a lot of PhD programs.

** Much of the debate on the debt restructuring architecture has focused on the G-20’s “Common Framework” which only applies to low-income/IDA eligible countries. Sri Lanka was just a bit too wealthy to qualify for the Common Framework.  This has two concrete implications:

a) China’s official creditors were not required to participate in a single official creditors’ committee with the traditional bilateral creditors (i.e., the Paris Club; in Sri Lanka’s case, the biggest traditional creditor is Japan).  The Export-Import Bank of China negotiated separately in this case, and the China Development Bank is part of the commercial rather than the official restructuring.

b) Countries outside the Common Framework aren’t assessed using the IMF’s low-income country debt sustainability framework. The “retro” low-income country framework focuses on external debt and generally is more constraining; the “modern” market access country framework doesn’t have the cliff effects of the low-income country framework but also entirely drops balance of payments-based measures of debt.

*** The general government’s external debt table includes the central bank’s swap liabilities, which is commendable.

**** Ecuador, interestingly, has a very similar external debt profile to Sri Lanka despite a very different overall public debt profile. Both Sri Lanka and Ecuador have relatively large stocks of external debt and very few reserves, and a reasonably high share of each countries’ debt is relatively expensive commercial debt. Both countries are examples of why looking more holistically at debt-payment capacity matters. The issue for Sri Lanka should be whether its relatively high level of domestic debt makes it too risky to also carry a relatively high level of external debt; the issue for Ecuador is whether its lack of a domestic buffer (thanks to dollarization) makes it too risky to carry its current stock of foreign external debt even with a low level of domestic public debt.  I am not convinced that the IMF's current framework draws out these questions in an effective way, as it (intentionally) blurs the line between domestic and external debt.

***** The macro-linked bonds have a high risk of being viewed as a sophisticated version of the initial bank flow rescheduling that marked the first phase of the 1980s emerging market debt crisis. That rescheduling was defined by a belief on the part of creditors that the country only faced a liquidity problem, and thus an unwillingness to give up on principal (the banks also needed time to rebuild their own capital). The actual resolution of the crisis required a truly fixed payment schedule that the country could meet, which created a realistic upside for both creditors and the country⍐.

Blog Post by Brad W. Setser July 16, 2024 1:10 pm (EST)