When French president Francois Hollande recently attacked the decision of former European Commission chief Jose Manuel Barroso to accept a position at Goldman Sachs, he cited the US bank’s role helping Greece fiddle its Maastricht ratios.
Fifteen years after it was signed, and thirteen years after I first wrote about it in 2003, the notorious Goldman-Greece swap retains a larger than life symbolism in the debate about the European Union. A year ago I spoke about this to the New York Times, in a column written by Bill Cohan. As that article shows, there is a continued appetite for information about this deal.
The deal has attained an importance out of proportion to its initial impact on Greece’s balance sheet. As Goldman never tires of pointing out, Greece’s debt-GDP ratio declined by ‘only’ 1.8 per cent as a result of the swap. And Goldman says it can’t be blamed for the subsequent explosion of this ratio by over 70 percentage points and the collapse of Greece’s economy.
This misses the point. As the uproar over Barroso’s move to Goldman shows, the US bank is loathed in continental European capitals for having done the deal. With its prestigious name, Goldman gave Greece vindication that Maastricht rules could, and should, be flouted at will. The Brexit vote merely heightens the feeling that Goldman undermined the EU project right at the beginning, and the bank is not being forgiven for it.
Since I first learned about the swap in a City of London restaurant in May 2003, I have accumulated an archive of unpublished interview notes, including background conversations with senior Goldman executives and my 2012 interviews with former Greek debt agency officials. As an exclusive for Risky Finance subscribers, here are some interesting details.
No bank wants to defend a controversial deal purely on the grounds that it was legal. Goldman Sachs, like other investment banks, has business practice and risk committees whose job is to examine large transactions from all angles and protect the firm from potential problems. Goldman has a good track record but every now and again, such as with Libya, Malaysia, BHS and Greece, it makes a mistake.
Although the bank’s senior management will never admit it, the Greek swap was in reputational terms one of Goldman’s worst ever blunders. How did the firm’s vaunted compliance and review process allow it to happen?
One reason is that there was a precedent. Confirming what had been rumoured for years, Italy’s then-finance minister Giulio Tremonti said in 2010 that his country used off-market swaps to flatter its deficit ratios prior to joining the Eurozone in 1998. Italy is often cited by Goldman executives as justification for the Greek deal.
However, the precedent is misleading. As Tremonti was careful to point out, not only were Italy’s swaps done in lira, before the euro rules were codified. They were also explicitly approved by the European Commission. Neither of these were true in the case of Greece. Perhaps the best thing that can be said about Goldman is that it allowed itself to be hoodwinked.
In 1999, Christoforos Sardelis was hired from his job as chief economist for Bank of America in Athens to run Greece’s public debt management agency. Overseeing the PDMA was a branch of the finance ministry called the General Accounting Office, in particular a shadowy directorate called Delta 23, whose boss was deputy chairman of the PDMA. According to Sardelis, D23 was a ‘back office’ for the PDMA, served as its auditor and also handled relations with Eurostat and Greece’s statistical agency.
D23 has been dogged by allegations of corruption and dishonesty. In 2007, the directorate was discovered to have arranged the issuance of €280 million of structured notes with JP Morgan, using a separate military procurement balance sheet which bypassed the PDMA. The bonds were then missold to Greek pension funds at a huge markup by a corrupt broker. In the ensuing fallout, JP Morgan fired its banker who had negotiated with D23.
In 2010, Eurostat published a report in which it accused D23 of lying about the Goldman swap. Goldman has always insisted the transaction was perfectly legal and compliant. On paper, using Eurostat’s published accounting rules, it was. Under pressure in 2012, Goldman gave me email evidence that it had met Eurostat before the Greek swap deal, and Eurostat had approved the “out-of-money swaps” central to the deal. After seeing the email, Eurostat responded that this discussion was only in general terms and not specific to Greece.
Given the gulf that remains between Goldman and Eurostat, did the bank do enough to ensure compliance? When I first spoke to a senior Goldman banker about the deal in 2003, he told me “I don’t think these sovereigns are violating Eurostat rules. I think they interpret them and they go backwards and forwards with Eurostat and Eurostat accepts these numbers.”
But what made him so confident that the EU’s accounting watchdog had accepted Greece’s numbers? It was not the PDMA’s job to tell them. Did D23 lie to Goldman about approval from Eurostat or the Commission, in the same way it subsequently lied to Eurostat itself?
Clearly, Goldman didn’t do enough. At the very least, it should have obtained a letter from Eurostat to D23 that specifically approved Greece shrinking its debt ratio using off-market swaps. And it should have made more enquiries in Athens about the probity of the finance ministry. That mistake still haunts Goldman today.
Starting in the 1990s, European countries that committed themselves to Maastricht limits looked for ways to reduce debt and deficit ratios. Short of outright privatisation – always politically risky – they found ways of gaining the same cash benefit using securitisation. The trick was to keep assets on balance sheet while funding them off-balance sheet through a special purpose vehicle.
After joining the Eurozone, Greece looked for cash flows to securitise – things like air traffic control revenues, lottery tickets and interest payments on municipal loans. My favourite one was borrowing against EU community support funds intended to improve Greek infrastructure. A bit like a poor person being given a tuition stipend and then pledging the payments for instant cash in a pawn shop.
Between 2000 and 2001, Greece raised €2.7 billion in this way. D23 supervised the deals according to Sardelis, and they were publicly marketed to investors by arrangers such as Morgan Stanley and BNP Paribas. Although they exploited off-balance sheet accounting, their terms were no secret, with public prospectuses. And as a percentage of the face value of bonds sold, the issuer fees and commissions were typically below two per cent.
Goldman’s fixed income sales team had other, more lucrative ideas. Consider foreign currency debt issued by Greece before joining the euro. Back then, the worry had been about Drachma devaluation, so under Spyros Papanicolaou, the Greek central bank used cross-currency swaps to lock in the value of this debt using market exchange rates.
Addy Loudiadis handled this account for Goldman and got kudos for warning Papanicolaou about dodgy products offered by competitors. When she finally pitched the deal in 2000 it was not a ‘product’ for the PDMA but financial alchemy aimed at D23:
“Why don’t we quietly lend you some upfront cash by using a fictitious exchange rate in your cross currency swaps and then you repay us with a separate interest rate derivative that has a positive value for us?”
I don’t claim that Loudiadis said these words, but people from her firm have said them. This is similar to borrowing against future cash flows by securitising them, Goldman argued. How so? I asked when I first heard this argument. After all, the securitisations involved real cash flows like lottery tickets that obviously have a financial value that serves as collateral. How can you compare that to creating a loan out of nothing by using a fictitious rate in a currency swap?
Goldman bankers respond to such bewilderment with infinite patience. Stop getting hung up on what is ‘real’ and what isn’t, they say. As soon as you invoke the fictitious exchange rate you have a debt. To repay it requires a stream of future cash flows as real as an itch on your arm.
As they pointed out to Greece, the end effect is the same. In a normal securitisation, a country gets upfront cash from off-balance sheet lenders, which it can use to buy back bonds in the market, and reduce public debt ratios, at the cost of future adverse impact on deficits.
Using swaps is even better, coaxed Goldman, because you get the Maastricht benefit of a debt buyback – via Eurostat rules – without the driving up your bond prices in the market. And we’ll construct the repayment swap as a hedge against increasing interest rates for your national balance sheet. All in complete secrecy.
Does this sound familiar? On a far bolder scale (this is Goldman after all) it’s the characteristic pitch of all derivatives salesman who argue that their ‘solution’ offers ‘benefits’ not found in something more transparent and far less profitable to the bank proposing the deal.
Much of this conversation took place within D23, but since the proposal involved managing Greece’s debt portfolio, the Goldman bankers went to see Sardelis. According to Papanicolaou, Goldman would only negotiate on the basis that Sardelis keep the deal secret. Sardelis became sceptical about the idea, and tried to turn it down – only to be overruled by his paymasters in D23 and the finance ministry, for whom secrecy was attractive.
On 30th June 2001 the swap was signed, creating a secret debt of €2.8 billion, more than all of Greece’s public securitisations added together. According to Sardelis, there was no dinner to celebrate. The Goldman bankers begged off – they had to return to London to supervise the hedging. In a minute we’ll find out why.
It has a reputation for political patronage, poor scrutiny and a liking for using derivative-based products to flatter its financial position. No, not Greece but the London Borough of Newham. In 2015 I uncovered for Channel 4 Dispatches how £150 million, or a third of the council’s borrowing, was in a particularly risky product called an inverse constant maturity swap (CMS) floater, that contributed to the council recently paying more than 7 per cent interest on its loans.
It turns out that Greece also got into inverse floaters, via the repayment leg of the swap with Goldman. As with Newham, the justification was “hedging” against higher interest rates. Similar to Newham, there was a teaser period of three years in which Greece paid no coupons. But it was the formula for the coupons that began after that which caused Greece problems, on a staggering scale.
According to Sardelis, this complex formula involved Greece paying a cap rate minus the 10 year euro swap (CMS) rate multiplied by a leverage factor. The notional for the swap was at least €15 billion, and as much as €30 billion according to other sources, with a maturity date in 2019. If we assume conservatively that the notional was €25 billion and the leverage factor was two, then the delta of this swap was €50 billion, and possibly significantly more than that.
As Newham council learned, inverse floaters are extremely sensitive to shifts in market expectations of interest rates, or the forward curve. A modest decline in long-term yields can result in all the expected inverse floater coupons increasing at the same time, resulting in a surging mark-to-market value. This hurt Newham in 2014 and similarly Greece in late 2001, after the 9/11 attacks caused euro bond yields to plunge.
Using the parameters above with some reasonable assumptions I priced the Greece inverse CMS floater on a Bloomberg terminal (see chart). The swap starts out in July 2001 with a value of €2.6 billion for Goldman (reflecting its purpose to repay off-market currency swaps), and initially moves in Greece’s favour. Then after the 9/11 attacks, the euro yield curve fell, and the value of the swap for Goldman balloons to above €6.5 billion.
Unlike Newham, Greece never even got to the stage of actually paying high coupons on its inverse floater. The mark-to-market increase alone was enough to spook Sardelis who demanded a different formula for the repayment leg from Goldman.
When I first talked to a senior Goldman executive about the trade in 2003, he told me it was a “principal transaction”, implying that Goldman was taking proprietary risk on its balance sheet. It was not supposed to be credit risk – the exec confirmed what my sources told me about Goldman using a credit default swap with Dublin-based Depfa Bank. He also explained that the interest rate and foreign exchange risk was hedged in the market. Until Sardelis and Papanicolaou talked me through the deal in 2012, I wasn’t sure how to reconcile these statements.
So what risk was Goldman taking?
When trading opened on 1 July 2001, Goldman faced Greece in swaps whose inverse floater component had a delta of €50 billion or more. It was effectively going long a trade similar to the ones that felled California’s Orange County in 1994. No wonder the bankers were required to fly home from Athens so quickly.
It probably took days to hedge that kind of delta, perhaps starting with Bund futures and progressing to swaps. My background conversations indicate that Goldman didn’t execute the swaps at ‘mid’, and built a spread into what Greece paid in order to cushion itself from the risk it was taking. We’ll discuss how this impacted profits below.
But there was a component that sources close to Goldman say wasn’t hedgeable at all: the convexity or nonlinear exposure to interest rates in the CMS trade. This convexity is there because the value of liquid fixed rate hedging instruments like vanilla swaps or bonds is affected differently by discounting of future coupons compared with CMS products which use a fixed rate as a floating index.
Given the leverage of the inverse floater, this convexity was enormous, equivalent to an interest rate option notional in the billions. Because of the caps and floors embedded in the trade, the bank was both long and short. Goldman quants wrote memos on how the convexity might be replicated using swaptions, but traders knew the market would be unable to handle the volume, so they left the exposure in place. This was the proprietary bet that Goldman made.
The trade’s volatility also meant that Goldman took additional Greek credit risk, despite what the firm’s execs told me at the time. Think again about the credit default swap hedge with Depfa. This was always described to me as being a static hedge, as if Goldman had made a fixed loan to Greece with a bullet repayment date.
You can think about the off-market exposure that Goldman had on the day of the trade as being like a fixed rate loan, but the analogy quickly unravels. Suppose that Goldman had done €25 billion of vanilla currency swaps with Greece at the market price and fully hedged them. Then Goldman’s exposure at the trade date would have been zero.
However, the value of the trade would have soon become either positive or negative as bond yields and exchange rates moved around. Without a collateral agreement in place, Goldman would be exposed to Greece defaulting while still having to pay out to hedge counterparties.
Hedging that credit exposure requires dynamic, not static CDS trading. Similarly with the actual Greek trade: Goldman’s exposure quickly diverged from the initial €2.8 billion, accelerated by the tremendous volatility of the inverse CMS floater component. As the CMS trade went against Greece and the value climbed from €2.8 billion to above €4 billion, Goldman was taking increasing credit risk, something that it was unlikely to have been comfortable with.
There is an old quip about the Greek breakfast that consists of a pack of cigarettes. Greeks are the fourth most voracious smokers in the EU according to OECD statistics, with 38 per cent of the population lighting up at least once per day. Given that national habit, one might expect the cost of tobacco to feature prominently in Greece’s domestic inflation figures.
When Sardelis talked to Goldman about restructuring the repayment leg of the swap in 2002, they found an index for him that fortuitously excluded tobacco costs: the European harmonised index of consumer prices (HICP). That wasn’t why Sardelis liked it. He was happy because it was less volatile than a leveraged inverse CMS product, which wasn’t very difficult.
The €50 billion CMS hedge was unwound, and the mark to market loss that Greece had already incurred was added to the debt owed to Goldman. Recognising the profits on that trade would have been premature, because the bank had a new proprietary position to manage.
While the HICP was already used as a calculation basis for index-linked French government bonds (OATIs), doing a €25 billion notional HICP swap trade with Greece put Goldman on a new level. For comparison, between 1998 and 2004, France issued a total of €48 billion index-linked bonds. By receiving cash flows on the synthetic equivalent of €25 billion of HICP-linked debt, Goldman would become the biggest player in the nascent euro inflation swap market overnight.
It took a good year for Goldman traders to take the HICP risk warehoused from Greece and find buyers for it in the market. In September 2003, Goldman was voted top inflation swap provider in a Risk magazine poll. The same month, I published an article in which Italian originators of structured inflation-linked products extolled Goldman as the biggest inflation swap counterparty in the market. Somewhat testily, the bank did acknowledge publicly that it had a position on with Greece (acknowledging a size of €500 million, far smaller than the real figure).
In that article I quoted a Goldman trader, Driss Ben-Brahim, who in 2004 was featured in news reports for having been paid a £30 million bonus (denied by Goldman): “A strong proprietary bent is necessary to run a successful inflation business”, he said.
A theme that emerges in this story is how Greek officials tinkered with the portfolio over time. First there were some vanilla cross-currency swaps that Loudiadis rejigged into a giant hidden loan. Then Sardelis lost patience with the inverse CMS floater and asked Goldman to “change the formula” to one involving Eurozone inflation. Next was the change of government in 2004 and Sardelis’s replacement by Papanicolaou. That led to the swap being restructured again, transferred away from Goldman and onto the balance sheet of National Bank of Greece, leading to its securitisation in 2009.
At each stage, former Greek officials accuse each other of making things worse by tweaking the existing deal. Take each step in turn. Start with the leveraged inverse CMS floater, which was supposed to be a hedge against rising interest rates on Greece’s broader debt portfolio. Former Greek officials criticise Sardelis for panicking over the mark to market value, which they say had no material consequence for Greece whatsoever (under Eurostat accounting rules).
What if Sardelis had left this trade in place? Greece was not obliged to pay any coupons to Goldman until mid-2004. Using my assumptions for the trade, and market inputs over time, Greece would have had to pay Goldman about 6 per cent a year thereafter, or €1.5 billion per year.
Between 2001 and 2005, the average interest coupon on Greece’s debt portfolio declined from 7 per cent to 5 per cent, saving the country about €1.5 billion a year over this period. The Goldman swap wiped out this gain. Now you might argue that you can’t blame a hedge if the event it protects against – rising rates – doesn’t happen. But it appears that the only impact of Sardelis’s action was to accelerate the cost that Greece would have paid anyway.
There is a slightly better argument for leaving the inflation swap in place, because HICP declined steadily over time (albeit with considerable volatility). It is hard to say more because we don’t know the precise terms of the deal, which Papanicolaou complained was structured in a way that was bad for Greece.
That takes us to the big decision by the conservative Greek government in 2005 to restructure the swap when its value stood at €5.1 billion. Former officials are bitter about this decision because it locked in a €2.3 billion loss for Greece, even though the pain of this increased debt was softened with a greatly increased maturity of 32 years. The outcome was the best possible one for Goldman: the firm got to release all its risk reserves and recognise the profits made on the deal.
The really fascinating counterfactual is if Greece had kept Goldman in the trade all the way up to 2009 and beyond when Greek creditworthiness tumbled. Faced with a potential writedown, Goldman might have blinked and offered Greece far better exit terms. But like so much of this story, this is purely speculation.
And finally there is Goldman’s last act in the drama: the securitisation of the swap via the Titlos special purpose vehicle in 2009. The purpose of this deal arranged by Goldman was not to sell the swap to outside investors: the sole buyer was National Bank of Greece. The point was rather to convert the swap into a ‘marketable asset’, which could be pledged to the European Central Bank as collateral. The plan didn’t work – a downgrade of Greece soon made Titlos ineligible under ECB criteria and it sits today on NBG’s balance sheet.
Once again, there appears to have been some tone deafness on Goldman’s part. A transaction that helped undermine the Eurozone, now transferred onto the ECB’s balance sheet? What were they thinking of?
Christoforos Sardelis described Greece and Goldman as a pair of “willing sinners”, a phrase that sums up the reputational damage done to both sides once the deal was exposed. But why were the sinners so willing? Greece wanted to subvert Maastricht rules that it was flouting. For Goldman, it was all about money.
When I first heard about the Greek swap at lunch with a source in 2003, one of the numbers I scribbled down was the profit Goldman had [ostensibly] made on the trade – $500 million. I then asked Goldman to comment on my figures.
I realised that I was on to something because Goldman took the story extremely seriously. First came a lawyer’s letter. Then came a friendlier overture from Goldman’s PR department offering access to a very senior executive who was particularly exercised about the $500m profit number.
Let’s think for a moment on what the profit number actually means. We know that Goldman constructed a partial offsetting hedging position for the directional interest rate risk or delta, while taking proprietary risk on the volatility. It also bought a credit default swap to cover some of the counterparty risk it was taking (implying that Greece was not required to post collateral).
The value of what Greece pledged to pay Goldman minus the present value of all the hedges and internal prop trades was the gross profit. However, this gross profit is not the same as the net profit booked by the bank on the day of the trade. Building an offsetting hedge involves risk-taking and modelling decisions, which might turn out to be wrong over time (this was abundantly true in 2008). A bank allows for this uncertainty by keeping some of the gross profit in reserve, depending on risk management, auditing and regulatory constraints. That reduces the gross profit to a smaller net figure.
In his 2003 conversation with me, the senior Goldman executive steered me towards thinking of its profit in net terms, and under pressure from Goldman I reduced the figure of $500m in my story to $200m. When the story came out, my source criticised me for downplaying the $500m number. The source was right to be critical.
Think about things from the Greek perspective. Assuming that in 2003 the PDMA didn’t have the ability to do its own valuation, Greek officials could’ve found a bank willing to quote a price to buy the derivative from them – either the bank that sold it to them in the first place or a third party. If they obtained this buyback valuation for the day the transaction took place, the difference between this and the value Sardelis bought it for indicates the gross profit made by the Goldman on the trade date. I’m pretty sure this is where my $500m number came from.
Eventually the Greeks did their own valuation as well. In 2012, Papanicolaou told me that when he succeeded Sardelis in 2005, he had brought in a former Deutsche bank PhD-qualified trader as a deputy, who worked out that if the trade had been unwound the day after the original transaction on 30 June 2001, Greece would have had to pay €3.4 billion.
In other words, the difference between the trade date value (€2.8bn) and the buyback price (€3.4bn) was an estimate of the gross profit earned by Goldman on the trade, or €600 million. This figure appeared in the Bloomberg story I wrote with Elisa Martinuzzi in March 2012.
Which figure is correct? My original $500m figure, Goldman’s preferred $200m number, or the €600m provided by Papanicolaou? A point to bear in mind is that all three are historical snapshots. The day one gross profit assumes that Goldman was able to hedge the entire trade in the market immediately, which we know didn’t happen. The net profit figure is a forward-looking number based on Goldman reserving for risks it anticipated to take over the entire life of the trade until 2019.
None of these numbers incorporates what we know subsequently happened (from Eurostat’s 2010 report and my Papanicolaou interview). In August 2005, Goldman exited the trade, selling the position to National Bank of Greece and restructuring it with a much longer maturity date in 2037. At this point, Goldman was able to release the reserves and recognise its initial profit, net of any trading result during the interim.
And we know something else: by the time of Goldman’s exit, the mark-to-market value of the swaps had increased to €5.1 billion, amounting to a €2.3bn loss for Greece. Could there have been a corresponding €2.3bn gain for Goldman? Sources close to the bank insist that this is a ridiculous claim to make, because Goldman hedged the majority of the risks – particularly the delta exposure to interest rate swings. These swings were substantial, with long-term euro rates falling 200 basis points between June 2001 and August 2005.
There is also the cost of the credit default swap to think about. In July 2001, Greece was considered to be virtually risk-free. Five year Greek CDS spreads were 15 basis points, while Goldman sources have suggested that the spread paid to Depfa was 30bps, as a result of the long maturity and illiquid nature of the trade. On a notional of €2.8 billion, this would amount to €8.4 million per year. If Goldman entered into this trade for the full 18 year period of the initial swap, and then unwound the CDS in 2005 with a large bid-offer spread, we can reasonably state that Greece paid about €100 million for credit hedging costs.
Even allowing for this, the big upfront gross profit suggests that Greece paid handsomely for Goldman’s hedging risks. We also know that Goldman took significant proprietary positions in order to do the trade, both in volatility and inflation, and that during this period, Goldman reported strong trading results, with FICC revenues doubling to $8.5 billion between 2001 and 2005.
Maybe the bank lost money on trading the Greek swaps and achieved the growth elsewhere. But the size of the trade and other circumstantial evidence suggests otherwise. That’s why my interpretation is that Goldman made at least $500 million from Greece, and probably more – a number somewhere below €2.3 billion.
Why is this important? Notwithstanding the deal’s legality, Goldman risked its reputation on the trade. The profit seems to have played a role in convincing the firm’s senior management that the risk was worth taking. They may have been right. Although the trade went far beyond what Goldman should have done, or been allowed to do, the firm is still gambling that people will eventually forgive and forget.
Currently a board member at Ethniki Insurance. Head of the Greek Public Debt Management Agency (PDMA) from 2000-2004. Executed the initial stages of the swap with Goldman, under supervision of the Ministry of Finance.
Now retired. Worked at the foreign currency department of the central Bank of Greece, and was head of the PDMA from 2004-2009 where he worked with Dusi on the swap restructuring and novation.
Goldman’s current chief executive was head of the firm’s Fixed Income, Currencies & Commodites (FICC) division in 2001. He would have been the key decision maker in terms of approving or blocking the Greek swaps from a risk perspective.
Currently vice chairman of Goldman Sachs, Sherwood was head of FICC Europe in 2001 and played a key role in troubleshooting the trading and compliance aspects of the deal.
Now head of Goldman’s pension buyout spin-off Rothesay Life, Loudiadis was head of sales for FICC Europe in 2001. Loudiadis began working with the Greek government as a client in the late 1990s and both Sardelis and Papanicolaou identified her as the mastermind of the deal.
Goldman Sachs partner and managing director. According to Papanikolaou, after Loudiadis left FICC, Dusi took over the account and was responsible for the 2005 restructuring. He also managed the 2009 securitisation.