27 Apr 2012, Nick Sawyer, Risk magazine
The architects of the original credit support annex (CSA) may well have congratulated themselves on a job well done. By drawing up a document governing the exchange of collateral between two derivatives counterparties, they might have felt they had made a big contribution to the safety and soundness of the over-the-counter derivatives market. Unknowingly, however, they also introduced an incredible amount of complexity. By building a number of options into the agreement, they unwittingly ensured that virtually every CSA is unique – a fact tvhat has led to valuation discrepancies on even the most plain vanilla interest rate swaps.
This has elicited complaints from end-users about a
lack of transparency and an inability to compare prices between dealers. More problematic, it has meant novations between counterparties have almost ground to a halt, as has the back-loading of trades to central counterparties (CCPs) – contributing to a reduction in market liquidity, participants say.
“You need to be confident you can unwind or novate a transaction easily, without getting into a disagreement with the other counterparty over the price. The capacity to novate quickly has become more difficult, and this has affected liquidity,” says Thibaut de Roux, head of global markets for Europe, the Middle East and Africa at HSBC in Paris.
“There is reduced liquidity, and issues on the relationship side. There is no real consistency, and clients have been getting more upset with it,” adds Juergen Feil, a managing director in rates exotic trading at Deutsche Bank in London.
Help could be at hand, though. Led by the International Swaps and Derivatives Association, dealers have been working for the best part of two years to draw up a new standard CSA that aims to eliminate the sources of valuation disputes and bring industry practice more in line with London-based clearing house LCH.Clearnet. The template is more or less agreed, and the first adopters – most likely a handful of big US and European dealers – could be using the new document within months.
Reasons for concern
Not everyone is won over, however. There have been complaints the document is too complex, particularly in its proposed approach for calculating a net settlement currency. Some banks have also expressed concern about the perceived dollar dominance of the document – although these fears were largely addressed earlier this year (see box, Dollar dominance issue resolved through vote). Another source of unease is the changes to systems and processes that will be required, especially because banks will likely have to run two systems in parallel – one for legacy trades under the existing collateral agreement and one for new business under the standard CSA. Taken together, it could slow the take-up of the new document, at least in the initial phase.
“I think we will take a wait-and-see approach,” says Tong Lee, global head of rates at UniCredit in London. “We will need to look very closely and see what the impact will be.”
There may be disagreements over certain aspects of the standard CSA, but most dealers recognise the current situation has led to big problems. The difficulty stems from the huge amount of optionality embedded in the existing document. Each set of counterparties can agree on a list of eligible collateral they will post to each other, as well as the threshold at which they will start to post, the minimum transfer amount, and any additional triggers or termination events. This means virtually no two CSAs are identical: one CSA might restrict the two parties to posting dollars with a zero threshold, while another might allow the counterparties to choose between dollars, euros, US Treasury bonds and even equities with a $50 million threshold.
This has only become an issue relatively recently. Banks used to discount everything at Libor, even though many realised the overnight indexed swap (OIS) curve should theoretically be used as a discount rate for cash-collateralised trades, as this determines the interest rate paid on cash collateral. During the financial crisis, however, the basis between Libor and OIS blew out massively – making it clear that Libor couldn’t be relied upon as an easy, one-stop-shop discount rate. As a result, the large dealers started to move to OIS discounting, with the relevant rate determined by the currency of the collateral being posted – so a swap collateralised with dollar cash would be discounted using the federal funds rate, regardless of the underlying currency of the trade (Risk March 2010, pages 18–22).
This becomes hugely complex when counterparties can choose from a list of eligible collateral, though – particularly when that list includes securities as well as cash. While OIS discounting is now generally accepted for cash collateral, there is much less consensus on what the correct discount rate should be when bonds are posted. Some argue the repo rate should theoretically be used, as this is the rate at which those assets can be funded. But repo markets are very short term, meaning an alternative has to be found – and many opt for Libor.
Once the correct discount curves have been determined for each of the eligible collateral types, dealers need to work out which one should be used to calculate the net present value (NPV) of each trade under that CSA. Most dealers agree the correct discount curve should be based on the cheapest-to-deliver collateral, on the assumption that counterparties will always look to post whatever asset is cheapest for them. However, what is cheapest now may not be cheapest in the future, meaning dealers need to look at the discount curves for each of the eligible collateral types, swap them into a single currency for comparison and consider what the cheapest collateral is at each point in time. The final theoretical step is to consider how this could change as market conditions alter.
Banks are taking a variety of approaches to this – some are looking at what is cheapest now and using a single discount rate for the entire duration of the trade, while others are constructing blended discount curves that try to reflect the cheapest-to-deliver collateral at any point in time (Risk March 2011, pages 18–23). As a result, each counterparty can end up with different valuations for the same transaction, which can contribute to disputes.
Crucially, any change in collateral terms will also have an economic implication for each counterparty – the main reason for novations slowing, say dealers. To take a simple example, a dollar-denominated swap backed by a CSA that allows the counterparties to deliver euro cash only should be discounted using the euro overnight index average (Eonia). If another dealer steps in via a novation, and has an existing CSA in place with the remaining counterparty that allows the posting of dollar cash only, that transaction would now be discounted at the federal funds rate – a change that can significantly alter the NPV of the trade. If the impact is large enough, the party that stands to lose out is likely to decline the request for novation. The same dynamics exist with the back-loading of trades to CCPs.
The standard CSA should eliminate many of the problems by removing the optionality. Instead, every trade will be allocated to one of 17 silos, based on the currency of the underlying trade. Counterparties will only be allowed to post cash collateral in that currency, and the trades in each silo will be discounted using the relevant OIS rate or – if a liquid OIS market doesn’t exist – an agreed alternative. So, for example, dollar-denominated trades would be allocated to the dollar bucket, the counterparties would be required to post dollar cash collateral, and the federal funds rate would be used to discount those trades.
The 17-silo proposal
This is a significant change from earlier proposals, which included just five silos, corresponding to those currencies with liquid OIS curves: dollar, euro, yen, sterling and Swiss franc. Expanding the number of silos means alternative discount rates will need to be agreed for silos six to 17, at least until liquid OIS curves develop in those markets.
That means it may not be as theoretically ‘pure’ as the five-silo proposal – but this approach is already employed by LCH.Clearnet for its SwapClear interest rate swap clearing service, and dealers are keen to ensure the methodologies for cleared and uncleared trades mirror each other, eliminating any mismatch that might occur between a bilateral transaction under the standard CSA and a cleared hedge.
“One of the overarching goals of this process is to allow for an ease of transfer and harmonisation between the cleared and uncleared market, so this is an example of us being aware of what is happening in the cleared market and moving towards that,” says Andrew Kayiira, assistant director, market infrastructure, at Isda in New York.
Any mismatch could have been problematic, say participants. Under earlier proposals, a Norwegian krone trade might have been allocated to the euro silo under the standard CSA, requiring collateral to be posted in euro cash and Eonia to be used to discount the transaction. However, it is likely a hedge conducted with an interdealer counterparty would be cleared through LCH.Clearnet, requiring Norwegian krone cash to be posted as collateral, with a local Norwegian krone Libor-style curve used to discount the trade.
“Aligning the core variation margin structure across cleared and bilateral hemispheres of the market makes risk management sense. If you were not using the Group of 17 (G-17) currency silos in the bilateral context, your Norwegian krone swap with the client would be collateralised in, say, euro, but a hedge cleared through LCH.Clearnet would require Norwegian krone collateral to be delivered, so there’s cross-gamma risk there. Having greater alignment also makes it easier to transfer risk to clearing houses,” says Michael Clarke, a managing director at Goldman Sachs, and co-chair of the Isda collateral steering committee.
Another reason for the expansion is to avert potential problems down the line as new silos are added. For example, if Australian dollar trades were allocated to the US dollar silo initially, but liquidity in the local OIS curve quickly improved and an Australian dollar silo was added after one year, it is unclear what would happen to the US dollar-siloed trades. Moving them to the new Australian dollar silo would have an economic impact, because the discount rate would change from federal funds to Australian dollar OIS. However, keeping them in the existing silo would create operational complexity: it would leave banks having to simultaneously manage Australian dollar trades under the old CSA, new CSA trades allocated to the US dollar bucket, and new CSA trades allocated to the Australian dollar silo.
“People were concerned about adding silos later on. If you start with five, what happens when you need nine or 10? It gets quite messy. So the view was, why don’t we just go for 17 straight away and simplify that part of the equation? The downside is you have to use proxy OIS rates for the G-6 to G-17 currencies, but we already do that in LCH.Clearnet,” says Clarke. It is likely the same proxy rates would be used to discount both cleared and uncleared trades – although this has not been confirmed. Ultimately, it is hoped starting with the extra silos will spur development of OIS curves in those markets.
While the 17 LCH.Clearnet currencies would be allocated to their own silos, those falling outside that universe would be allocated either to the US dollar or euro buckets, according to a mapping table based on analysis of the historical correlations of currencies. In the event a new silo is created, deleted or combined, counterparties will need to adhere to a silo restructuring protocol, which will outline a standard treatment for all affected trades.
Essentially, when a new silo is added, all pre-existing trades will remain in their former silo unless each set of parties bilaterally agrees to switch them over – therefore avoiding the economic impact that could occur from a mandatory change in collateral terms. All new trades would be allocated to the new silo. If a currency is merged with another – for instance, if Denmark chooses to join the euro – then all the trades in the deleted silo would be switched over. Where a currency ceases to exist, the trades would be re-allocated to the silo that had the closest correlation to the deleted currency. “These are the kinds of things you need to clarify – and you want to clarify them now,” says Feil at Deutsche Bank.
While the approaches for single currency trades have altered significantly from earlier proposals, the treatment of cross-currency swaps remains unchanged – despite protests from some banks in Asia-Pacific at the end of last year. The Isda working group had suggested multiple currency transactions be allocated to the US dollar silo, but Japanese banks in particular had complained that approach might create problems for banks that don’t fund in dollars (Risk January 2012, pages 74-78). This prompted a rethink by Isda – with counterproposals including a cascade structure, where a trade is allocated to a particular silo based on the composition of the currency legs, and a free choice. However, Isda decided to stick with the original proposal after conducting a vote earlier this year, in which 63% of respondents favoured the US dollar bucket as a first choice.
Despite these differences in opinion, the silo approach should eliminate many of the problems that currently exist by removing the optionality embedded in the CSA and aligning practices with LCH.Clearnet. But it creates new complexity – 17 separate silos means 17 different collateral flows could, in theory, end up going back and forth between each set of counterparties each day, creating cross-currency settlement risk, or Herstatt risk. That doesn’t exist under the existing document: if two dealers had traded a dollar and a euro swap and those positions were completely offsetting, collateral would only be exchanged on the net exposure, which would be zero. Under the new framework, a dollar collateral flow would pass in one direction, and a euro payment would be made separately in the other.
The Isda working group has devised two solutions to resolve this problem. One is a payment-versus-payment (PVP) settlement system, similar to the service CLS Bank provides to the foreign exchange market. However, this will take time and money to build – while Isda intends to put out a request for proposal later this year, a quick launch is unlikely.
Consequently, an alternative approach will be rolled out for the first phase of implementation. Dubbed the implied swap adjustment (ISA), it essentially allows counterparties to net the various collateral flows into a single payment in one of the G-7 currencies: Australian dollar, Canadian dollar, euro, sterling, Swiss franc, US dollar and yen. The adjustment will be based on foreign exchange spot rates, tomorrow/next day (tom/next) forex swap rates and OIS rates, with Isda publishing a daily fixing of the relevant data. Each counterparty will be able to choose which of the G-7 ‘transport’ currencies to settle in, and can switch up to two times each calendar month by giving two business days’ notice.
So, if a US dealer and Japanese bank have outstanding swaps in euro, yen and dollars, the trades would be split between the three currency silos, and the relevant discount rate would be used to value the trades in each bucket. However, the Japanese bank might decide to use the ISA to settle anything it owes in yen only, while the US dealer might opt to post collateral in its own domestic currency.
This method appeals to a number of banks – including those in Japan, which had been worried about being forced to settle in US dollars. “The best solution for us is the price adjustment. We want to settle in yen, even if we have euro or dollar swaps. The most practical way is to have a choice in settlement currency,” says Yutaka Nakajima, senior managing director, head of trading, fixed income Japan, at Nomura Securities, and an Isda board member.
In fact, some question whether the PVP solution is needed at all and suggest market participants may ultimately decide to stick with the ISA. “Consensus is that if the ISA methodology works, then the push for PVP will be on the backburner,” says Feil at Deutsche Bank.
Not everyone agrees, though. Some dealers argue the ISA methodology is complex, and introduces an additional manual process that increases the chances of error. Problems could also emerge in the unbundling of the transport currency back into the individual cash positions needed to fund the book, they claim. Banks would want to make that conversion at the fixing rate – but the level at which they actually trade may not exactly match the official tom/next foreign exchange swap fixings published once a day by Isda.
“We can all look at this 11:00am fix, then at 11:01am I am going to be busy finding someone to trade with at a rate as close to the fix as possible. Potentially we are creating a scramble each day to convert currencies when it might be more efficient to exchange the underlying currencies,” says Joshua Luks, fixed income treasurer at BNP Paribas in London.
As a result, several banks say they will likely wait for the PVP solution to emerge before adopting the standard CSA. “We much prefer the CLS-type settlement. That way, we can systematically distribute the right amounts to our business units. We voted for the standard CSA, but we decided to wait for the CLS-type solution,” says Guillaume Amblard, global head of fixed-income trading at BNP Paribas, and an Isda board member. At least three other big European banks have similar opinions.
Those close to the working group concede the ISA calculation would add an extra layer of complexity to the process, but point out that banks are already doing something similar today whenever a net collateral payment is made in a particular asset chosen from a list of eligible collateral. Even putting that aside, the conversion back out into the underlying cash positions will not necessarily occur on a transaction-by-transaction basis, they say.
“The desire to execute at close to the fixing rate would be relevant if banks were taking their transport currency and swapping it into the 17 currencies every day, but they may not actually do that. They may have other activities through the corporate treasury function, which means they have different currencies flowing in and out every day anyway. Foreign exchange transactions may be done as part of that overall treasury function, but those would not necessarily be done at the rate implied by a single transaction – it would be more of a portfolio funding issue,” says one dealer.
Either way, the PVP solution will not be available until next year at the very earliest – possibly later. That means only a core group of dealers will likely implement the standard CSA in the first phase. A few buy-side firms have been involved in the discussions, but are not expected to be among the first adopters – in fact, asset manager BlackRock has decided to renegotiate existing CSAs in order to cut back the list of eligible collateral to a single asset for each agreement.
“Based on our discussions with firms, we estimate about half a dozen firms will adopt this in the first phase. There are buy-side firms that have shown interest, but I don’t think they will be among the first six volunteers,” says Kayiira at Isda.
The document will be voluntary – and Isda acknowledges it may not be appropriate for certain client types to implement at all. Some banks may decide to hold off until the PVP solution is up and running; others might decide to stay on the existing document. Several of the largest dealers, however, are keen to get going.
“We will be in the first phase, and we have put everything in place internally. The operational side is always the issue, and every bank has to keep an eye on budgets. To implement, you need to have the budget, the people and the time,” says Feil at Deutsche Bank.