Warning! — long post. But with investors still grappling with the terms of the Greek debt restructuring, it’s worth reading on…
Avid readers will remember Option 4 of the proposed Greece financing offer.
Published by the Institute for International Finance last week, but like all the options, it was drawn up by banks — not the IIF itself (we’ll get to the distinction in a bit, but just note that it’s there):
It’s a tricky one. Maybe the pinnacle of complexity in the quest to secure “private-sector involvement” on Greek debt. Discounts to bonds’ face value, partial collateralisation of the remainder, escrow arrangements for this collateral and so on. Tricky, we thought, but we’re dumb journalists. It can’t be forbiddingly complex. Let alone apparently impossible for some analysts to reconcile with the offer’s overall proposed haircut to Greek debt, which should be the same across each option: 21 per cent of net present value, assuming a 9 per cent discount rate. Can’t be, but… there are rising doubts, of a technical but worrying nature. They encompass the actual size of Greece’s haircut, the actual basis for adding up the discount rate — and not least, the meaning of Option 4.
_____________
Option 4
It’s probably best to start with the typo query.
On Monday, Pavan Wadhwa of JPMorgan published a European Rates Strategy note on the intrinsic value of Greek government bonds, following the offer’s terms. We’ll get to his conclusion in a moment, but it’s the footnote that stood out first:
Note that the IIF term sheet detailing the fourth option is not clear. Specifically, it says that “The principal is partially collateralized with 80% of losses being covered up to a maximum of 40% of the notional value of the new instrument”. We think this is a typo since 40% collateralization results in a 30% haircut using 9% assumed yield. Assuming 80% collateralization results in a 21% haircut which is similar to the haircut on the other three options. We therefore assume that 80% (instead of 40%) of the €80 face value will be collateralized…
Now, there are analysts elsewhere who don’t think it’s a typo. Also, they calculate the sequence of protection in Option 4 in a way that does deliver 21 per cent (using a 9 per cent discount, etc).
To be sure we asked the IIF directly whether this was indeed a typo. They asked the bank that did the option (again, the IIF didn’t write the options), and got back to us with the following answer:
The pricing of Option 4 is complex, but that is not a typo.
To value Option 4 on an approximate basis one would split the cash flows into parts and value them separately: the 40% up front collateral (of the 80pct discounted bond), the remaining 60% “uninsured” principal discounted at the assumed rate of 9%, and the coupons also discounted at 9%. That should sum to provide roughly a 21% NPV reduction.
To give a precise valuation you would need a stochastic model, as well as assumptions on recovery rates.
OK – first, the context here, which is relevant to the pricing of all the other options as well. It makes sense to discount principal differently depending on whether it is ultimately repaid by Greece itself or from AAA collateralised sources.
But Wadhwa is cautious as to whether even this approach creates a 21 per cent haircut. As he told FT Alphaville:
“the coupons also discounted at 9%”. Coupons are 80% of 5.9% = 4.72%. This works out to a PV of 38.05. No ambiguity here.
“the remaining 60% “uninsured” principal discounted at the assumed rate of 9%”. PV = 60% * 80 / (1.09)^15 = 13.18. No ambiguity here either.
“the 40% up front collateral (of the 80pct discounted bond)”. This could mean one of two things.
I. 40% of 80 is the PRESENT VALUE of the guarantee. This would make it 32. The total PV would therefore be 38.05 + 13.18 + 32 = 83.22, or a haircut of around 17%. This doesn’t work out to a 21% haircut.
II. 40% of 80 is the FUTURE VALUE of the guarantee. This would make it 40% * 80 / 1.038^15 = 18.29. The total PV would therefore by 38.05 + 13.18 + 18.29 = 69.51, or a haircut of around 30%. Still doesn’t work out to a haircut of 21%.
He adds that it shouldn’t be necessary to use a stochastic model, as payments under the bonds wouldn’t be random. Even if Greece did default on paying coupons at a random time, collateralised principal is accelerated in that scenario. Notably it might not be necessary to use stochastic modelling for the other three options to derive 21 per cent haircuts.
However, Tim Brunne, a credit strategist at Unicredit, does accept the explanation given via the IIF. Nevertheless even in doing so he picks out another assumption — only in Option 4 does it seem that principal is accelerated in the event of default. In the other options investors are guaranteed their principal payback on maturity of the bond (for example 30 years for Option 1). Brunne says that Option 4′s rolling payments into an escrow account account for this crucial difference:
In the 4th option, a default of the new 15Y GGB seems to accelerate payments i.e., the principal redemption. Therefore the NPV of the guaranteed fraction of the principal redemption is alway 40% of the bond notional amount (and 32% of the intial investment notional amount) during the whole lifetime of the bond. Strictly speaking, this reasoning is conditional on a 50% loss assumption, but this simplification is helpful to understand the different NPV of the guaranteed payments in the 4th option compared to the first three options.
However, what happens if GGBs need to be restructured in the future once again, is not explicitly discussed somewhere, according to my knowledge…
Which is interesting. If Greece does default in the near future, it’s likely that there will be enough in the escrow account to help cover losses, whereas owners of the 30-year par bonds (for instance) are left running for cover. Naturally the prospect of a second default is going to weigh on investors as they decide whether (and what part of) the offer is worth accepting, or whether there is an incentive to free-ride after all.
_____________
Discount rates
Ultimately we are going to leave this portion of the problem — on pricing these bits of Option 4 — as an exercise for the reader. But back to Wadhwa’s main conclusion, as promised. We’re now going beyond Option 4 a bit but the issues are related. They involve the cash curves used to price the discount rate for the offer’s haircut.
Wadhwa explains that it does make sense to price the new bonds’ collateralised principal using a Libor curve, but that he’s used an annuity curve to calculate the cashflow arising from the coupon payments, rather than the zero coupon curve that might be used. Again this reflects the balance of collateralisation and risk between principal and coupon. Seems technical but it’s a big difference. As Wadhwa argues:
The exhibit [above] shows that the Greek annuity curve flattens out around 12%. This annuity yield is significantly higher than the 9% yield assumed by IIF to estimate the PV of coupons on the new Greek bonds. Additionally, the IIF uses a 3.80% yield to discount 30Y AAA-rated bonds that will be used to collateralize the principal. We believe this yield is a tad low since 30Y EFSF zeros will likely be issued around L+50bp; as of Friday’s close, that would mean a zero coupon yield around 4.10%.
Since both IIF yield assumptions are too low, the exchange price estimated by IIF is too high (€79), and therefore the estimated €21 haircut (off of par) is too low. We find that the correct haircut is around €34 based on Friday’s closing prices…
It’s debatable whether the EFSF itself would issue the AAA-rated zero coupon bonds, of course, but the point stands for similar AAA issuers who might be turned to. Anyway, the main thing here is a bigger haircut, of 34 per cent.
Getting a bit more technical here, but Wadhwa reaches the same conclusion another way, using the zero coupon curve itself, noting that looking across the entirety of this curve, the 10-year point is situated around the 12-13 per cent yield level.
In short, it’s clear that key components of the bond offer are still causing considerable confusion a few weeks before the offer opens. There are also bigger themes at work, notably the risks of further haircuts down the line. What’s also clear is that this would affect the deal’s collateralisation in tricky ways, i.e. affecting one of the more innovative parts of the offer and a key element of the argument that the eurozone is edging toward fiscal union for Greece. It’s a bit like Ecuador’s 1999 debt restructuring, the first to restructure Brady bonds.
_____________
The IIF
Time then to say a quick word on the IIF’s role in this complicated tale. The IIF’s recent mistake over the details of banks supporting the offer (which is not necessarily the same as the banks who would tender Greek bonds anyway) highlights that the banking body has had to move quickly to engage bondholders before the August date when the exchange is scheduled to start. The complexity of detail within the offer compounds this, not least because the IIF wasn’t drafting the options themselves last week.
Very important bit of context this, because let’s compare the IIF’s role to the way similar debt swaps were conducted in Argentina or Uruguay. The latter case is a good one in that it also featured high expectations for take-up (which were met) but did this by making clear that coercion wasn’t too far away. Indeed investors were told that the central bank would not accept old bonds as collateral, and bondholders voted to bind hold-outs at the same time as accepting the new bonds’ terms. Impossible for Greece in 2011 when the ECB will not accept a CDS trigger or any threat to collateral stability. At the same time, Wadhwa observed in his Monday note that Greek bond prices beyond two-year issues are considerably priced in for just this kind of follow-up involuntary restructuring, featuring bigger haircuts. If anything, that’s going to make CDS triggers even trickier, particularly if there are problems with delivering bonds and generating recoveries.
It’s also worth noting that the IMF advised the Uruguayans in detail. There’s little political appetite for the IMF to play a similar role in Greece it seems (though the eurozone is clearly quite happy for the Fund to keep its loans coming). How times change, incidentally.
Intensive Fund or government involvement in corralling investors would be normal in most emerging-market sovereign defaults of course, but it’s remarkable how a banking association is executing a rapid, vast sovereign debt restructuring with relatively little official input (but plenty of official limitations). And it is vast. The financing offer foresees €135bn (€196bn) of gross involvement in the deal, from 2011 out to 2020 as various rollover provisions play out.
However, that’s gross. It’s rather different when you look at net contributions. For example in this official follow-up document, net private-sector involvement squeezes down to €37bn once you discount the credit enhancement provided to investors, and the delays that will come from bonds being rolled over after 2014. To give some context, Argentina swapped $30bn and then $10bn in two exchanges in 2001 (and eventually defaulted on $82bn of debt).
Very striking, we think, that this Greece offer has to secure €135bn of commitment right away, but the initial impact is still €37bn, more similar to the Argentine situation. You can see why emerging market board members at the IMF think this isn’t actually enough private sector involvement.
And yet, paradoxically, the point still stands:
The IIF is being asked to help corral, by September, a greater amount of bonds than were defaulted on by Argentina and Russia combined – still the biggest sovereign defaults in history, for now. It’s hardly a wonder that the overall bailout figure trumpeted in the halls of Brussels is melting away, pending the IIF’s painstaking work on bondholder participation.
All to play for, really.
Anyway, certainly shows why tricky stuff like Option 4 matters, no?
Related links:
The shaky basis of the Greek rescue – FT
Banks jump for cash before Greek swap – Reuters
Understanding the Greece bailout: the missing €7bn? – FT Brussels Blog