ARRC Vendor Workshop June 28, 2019

The Alternative Reference Rates Committee (ARRC) hosted a vendor workshop recently at the Federal Reserve Bank of New York, which I attended and in this article I cover some of the key points from the workshop.

• Required SOFR interest rate characteristics
• Required trade economics and processing for SOFR-based derivatives and cash instruments
• Compound in arrears processing which will become widespread
• Tradeoffs between exiting LIBOR portfolios prior to withdrawal and handling under the ISDA protocol and RFR fallbacks
• Inappropriate use of term RFR rates

Despite an evident consensus on the big rate and trade characteristics, the Q&A session and subsequent networking hinted at a mammoth effort across labyrinthine organizational chains and product flows.  

What I heard below with my interleaved observations and notes.

Summary of the session

1. Tom Wipf book-ended the session with opening remarks and conclusion.  
2. David Bowman elaborated SOFR rate and trade characteristics and led the Q&A session.  The presentation was largely based on the SOFR Users Guide.  This document has been well received.
3. Ellen Hefferan spoke on operational and accounting issues in the loans market.
4. There was a little post meeting networking though many slipped away after the speeches.
5. There’s a lot more information in the announcements and publications sections of the ARRC website.  

Tom Wipf’s Speech

The ARRC Chairman and Vice Chairman of Institutional Securities at Morgan Stanley laid the foundation and summed up the meeting.  The headlines were:

• We are deep in the implementation phase of an “unprecedented risk management undertaking” the like of which he would prefer no one sees again.
• LIBOR risk is 95% derivatives vs. 5% cash (commercial loans, debt issuance, mortgages).
• Nonetheless, cash instruments must be diligently addressed in their own right in terms of product-specific variants and unique issues.
• Term rate fixings won’t be usable until close to LIBOR withdrawal in 2021.  Before then, though they are already available indicatively (including from CLARUS01), term RFRs will not reach the robust, transaction-based state required by IOSCO standards.  Even after they are robust, enormous operational risk will be associated with their use under the ISDA protocol.
• Much better is to exit LIBOR well before end 2021 by making early decisions and prompt arrangements to transition LIBOR portfolios to SOFR non-term rate basis well before end 2021. 

Observation: Such decisions demand early and imperfect market risk decisions in 2019 / 2020 despite trader instincts for better liquidity or tighter pricing.   This will entail well-established and empowered LIBOR initiative steering arrangements and early availability of tools which will enable the forward impact of the transition to be assessed.

David Bowman on SOFR Rates

• SOFR is a good choice to replace USD LIBOR as it is published by the FRB for the public good not private profit, is transparent and transaction-based and is based on an active and very deep market which is difficult to manipulate and one of the few markets to weather the 2008 GFC well.
• SOFR is published as of yesterday, unlike LIBOR which is published as of today
• Today’s published rate is for yesterday as of date – unlike LIBOR where today’s rate is as of today.   Among RFRs, SOFR/SONIA/TONIA/ESTER are all like this.  (SARON is the exception being 6pm same day).
• SOFR is published at 8am EST and becomes final at 2:30pm EST.  This allows adjustments due to corrections from sources or technical glitches.
• Adjustments are only made if the adjustment is more than 1bp.   On May 31, there was a glitch in the FICC sourcing which resulted in resorting to the defined back-up source (the repo primary dealer survey) to calculate the rate published at 8am.  When the glitch was rectified, no update was made because the rate properly calculated was less than 1bp from the value calculated from the back-up.

Observation: care is required to carefully calibrate curves / reset processes to handle the date correctly. The suggestion is to take the rate at say 3pm EST so you never have to handle changes.

• SOFR has a solid time series of historical rates
• SOFR has been formally published since April 2018 and the pre-production version is available back to Aug 2014. 
• The FRBNY dealer volumes and rates survey goes back to 1998 and is a very good proxy for SOFR.
• Historic rates show that SOFR and EFFR (Fed Funds) were very close on every day in history bar some exceptions during the worst days of the GFC in 2008.
• The Fed may arrange for central calculation and publication of SOFR average rates to save widespread self-calculation / control issues (maybe 1M/3M/6M both simple and compound).
• Averages serve to smooth out 1-day spikes (e.g. Dec 31 2018) or exceptional divergences between EFFR and SOFR.
• CME SOFR futures already use such averages
• The Fed may sponsor creation of a daily compound average SOFR index to simplify calculation of the compound average rate between any two dates. Chris’s recent blog supported and explored this idea.
• The index starts at 1 and gets calculated daily by multiplying yesterday’s index value by (1 + r_i.n_i/N)
• Conveniently, this allows the compound average rate rate (r) between any start date (s) and end date (e) since i_e/i_s = (1 + r.n/N) means r = (i_e/i_s – 1).N/n

Observation: while compounding in arrears handling is new and causes work to support, if all the above conveniences are created, rate observation / calculation should be simple.

David Bowman on SOFR Trades

All four interest conventions below are needed and should all be engineered into systems solutions as required per product:

• Compound average set in arrears – despite being the most complex to implement, this is viewed as the preferred option for the bulk of both cash and derivatives markets.
• Simple average set in arrears – this is next most complex to implement, though mostly about reuse of existing processing which is narrowly used is expected
• Simple/compound average set in advance – though easy to implement, this will only get narrow use such as for SOFR ARMs
• Forward looking term SOFR set in advance – though easy to implement, this is not likely to be a robust source until late in 2021

On fallback trades, the spread should not be compounded but instead just added onto the compounded daily rate.

• Compounding of (RFR+spread) is logical but leads to compounding calculations per trade with little material difference in resulting rate in practice.

Coupon payment in arrears will require a rate lookback, a payment delay or a rate freeze.  Convergence of market practice on a standard approach / number of days per product variant seems important.

• Plain arrears handling creates a same day interest settlement requirement which no one will want for operational reasons
• Rate lookback means agreeing the trade rate for today is based on the RFR fixing say 2 days prior.  Now payment can be calculated 2 days prior to coupon date.
• Payment delay means calculating the interest payment on the last day of the coupon period but settling payment, say, 2 days later than natural coupon payment date.
• Rate freeze where the rate is frozen for the last 5 days of the period which will render trades inaccurate as hedges and is harder to implement.  For example, Ellen Hefferan’s Operational Issues in the Loan Market presentation included her view that loans will use a 5-day rate freeze but settle payment on unadjusted coupon due date .

Weekend / holiday accruals for lookback trades are tricky and require upfront specific agreement on handling.   A Friday normally has a 3-day accrual, but if the Friday accrual is based on a fixing on Wednesday due to lookback, how many days should be accrued?  It turns out that the answers 1 and 3 both have some support and market conventions require to be nailed down.

Q&A Session

1. How will the spread change over time?   David expects the spreads will become more firmly set at the time of LIBOR withdrawal and will converge over a period of about one more year after that to a permanent fixed long-term spread.  As spreads only apply to legacy LIBOR trades using the fallback, they should only apply to a rump of legacy trades after LIBOR withdrawal.
2. Should the spread remain dynamic?   David’s view is no as this would require a bank survey which is precisely what LIBOR withdrawal is trying to avoid.  
3. How are portfolios expected to transition over time?   David opined that whether it be derivatives, debt issuance or loans, the bulk of the transition should be about exiting LIBOR before LIBOR withdrawal.
4. How will FX forward rate curves be constructed in absence of LIBOR use for interest rate parity?   David suggested the expectation is to use an compound average SOFR rates and that this doesn’t require a term reference rate. 
5. When will SOFR IR volatilities be available to enable SOFR swaption trading?   David’s view that the market may need to make special arrangements to enable transition of options markets before withdrawal
6. A nasty conversion example: the AB Ports FRNs.  An audience member raised the challenges caused by the conversion of AB Ports FRNs from LIBOR to SONIA.   Even though conversion was done on a coupon date, to avoid splitting a coupon, the coupon set on that date was fixed on LIBOR in advance whilst the future periods were SONIA in arrears.   This led to the need for two strips of delta risk on one trade one for SOFR and one for LIBOR.   This caused a lot of work to handle discounting of both LIBOR and SONIA cash flows with a SONIA curve and to correctly accrue the last LIBOR coupon.

Observation: there was little discussion on the myriad issues of how to exit LIBOR before withdrawal.   No doubt other ARRC sub-groups will tackle this.  We have also blogged on this topic compression auctions and portfolio conversion.

Note: I picked up after the session that some draft proposals around: I was passed A Note on Building Proxy Volatility Cubes by Fabio Mercurio.  I also heard some say that the volatilities may be immaterially different despite the basis so that even the belt and braces approach of using LIBOR vols to price SOFR swaptions may work quite well.

Observation:  trading out and replacement by a new trade / issuance / loan will avoid such dual index problems.  However, there were cases noted in cash markets that cannot easily do that.

Post meeting networking

Residual steam on point 3 was vented in the post meeting networking. There was agreement that portfolio transition from LIBOR to RFR is MUCH bigger than set up of SOFR trading / discounting infrastructure.   The AB Ports example illustrates the type of thing which can happen.   There seems not to be enough time and mental capacity to debate each variant in theory and define a perfect solution.    An element of trial and error at small volume for each variant is thus inevitable.

Further new points opined upon included:

1. Hedge inaccuracies or hedge accounting failures may arise if a trade and its associated hedge are no longer on the same index due to one being transitioned to SOFR.  
2. Legacy OIS and FRAs may disappear once SOFR swap liquidity is establishede.g. Fed Funds OIS and USD LIBOR FRAs can be replaced by SOFR swaps (single-period in the case of FRAs). 
3. MM futures will likely stay with us e.g. Eurodollar futures will transition to SOFR futures (though, even here, SOFR swaps could become liquid enough to hedge SOFR swap portfolios).

Summary of Key Messages

• The majority of LIBOR portfolios need to be actively transitioned to be SOFR based well before the 2021 LIBOR withdrawal to avoid a monumental crunch of challenges around the end of 2021.
• This proactive LIBOR exit is the most difficult part of the whole market transition, given cross-organizational trade off decisions made between traders, support CXOs and IBOR program steering committees.
• Despite its scale, proactive LIBOR exit is better than trying to run everything on fallbacks with the associated plethora of legal and operational risks.
• Term RFR rates only become robust late in the process and are not a way to avoid most of the above hard work.
• There remains a not yet fully enumerated set issues which need to reach market practice consensus.  Aside from points 1-6 above, we have also covered some others in the blog – here and here.

Relevant Clarus Products

Clarus offers four services for IBOR to RFR transitions:

• SOFR volumes research – the link is the June 2019 blog post
• SDRView to view SOFR trade activity daily with Researcher or real-time with Professional access
• CLARUS01 is a term SOFR rate (1M, 2M, 3M, 6M)
• RFR Transform for insight into valuation and risk changes leading up to the transition.

For more information or a demo, please contact us and please mention that this blog prompted your enquiry.