February 12, 2020 - Folks that know LIBOR is likely to end soon after December 2021 probably also know that SOFR, the likely replacement for USD loans, is a different kind of rate. While LIBOR theoretically includes an element of bank credit risk, SOFR is an overnight risk free rate. Thus, their levels should be different. This difference should be solvable for new loans originated on SOFR, as new loans can have a higher margin to compensate for a lower base rate with different embedded optionality. However, it will matter for loans that “fall back” from LIBOR to SOFR upon LIBOR’s death. Because SOFR is a different rate, there needs to be a “spread adjustment” that minimizes the difference between LIBOR and SOFR for these falling back loans. And, indeed, ARRC “fallback language” specifically includes such spread adjustments.
How will they be calculated? First, ISDA has already announced the spread adjustment methodology for derivatives – historical five-year median difference – and Bloomberg should be publishing the indicative levels soon. It is likely that cash products could use the ISDA spread adjustment. But, in addition, in January, the Alternative Reference Rates Committee (“ARRC”), the body tasked with LIBOR transition, released a consultation on LIBOR-SOFR spread adjustments for cash products like loans, FRNs and securitizations. We summarize key points below.
First, as mentioned above, a spread adjustment is meant to minimize the difference between LIBOR and SOFR when LIBOR ceases. Both ISDA and the ARRC will use “static” spread adjustments; in other words, this spread adjustment would be calculated once at LIBOR cessation. It would not be a dynamic spread adjustment meant to capture differences between LIBOR and SOFR going forward. (The ARRC notes that potential dynamic spread adjustments suffer the same problems as LIBOR itself: i) Limited transactions in normal times that could be used to calculate the spread adjustments, ii) even more limited transactions in periods of stress, and iii) an unstable sample of firms that borrow in unsecured wholesale markets, which introduces borrower-based variability. Moreover, the ARRC shows in Table 1 (p. 9) that the mean absolute difference (MAE) using a static spread or dynamic spread adjustment are very similar. Long Story Short: No dynamic spread adjustment.)
That settled, what would be the methodology for the static spread adjustment? ISDA has consulted repeatedly on spread adjustments for derivatives. Based on respondents’ feedback, ISDA will use the five-year historical median difference between LIBOR and SOFR Compounded in Arrears. There would be different spread adjustment for different tenor pairs, e.g., three-month LIBOR to three-month SOFR, one-month LIBOR to one-month SOFR, and so on. The ISDA methodology is set and, if cash products use a different methodology, this could introduce some basis between derivatives and cash products.
Indeed, the ARRC acknowledges the importance of consistency between asset classes where possible. Tables 2 and 3 show that the five-year historical median difference (the ISDA methodology) has relatively small MAEs for term SOFR, SOFR Compounded in Advance and SOFR Compounded in Arrears. (Note that with the loan market’s interest in potentially using Simple SOFR in Arrears, the ARRC also says it could produce a spread adjustment for Simple SOFR. That said, Simple SOFR in Arrears and Compounded SOFR in Arrears are very similar – the difference has averaged a basis point or less over the last 20 years – so the same spread adjustment likely would work for both.)
ARRC does say that a “transition period” may be one point of departure from ISDA’s methodology. In effect, ISDA would apply the spread adjustment all at once at LIBOR cessation. The ARRC notes that if the LIBOR-SOFR differential is very different from the historical median at LIBOR cessation, this could create a “cliff effect” where a borrower could suddenly be paying significantly more or less when its loan switches to SOFR. Appendix 1 gives an example. At the end of 2014, the difference between LIBOR and SOFR was 35 bps, but the long-run difference was 60 bps. If LIBOR had ceased then, borrowers would have immediately paid 25 bps more when converting to SOFR. Conversely, in mid-2016, the difference between LIBOR and SOFR was higher. If LIBOR had ceased then, the rate paid by the borrower would have dropped by 60 bps immediately upon transition to SOFR.
A one-year transition period, during which the spread adjustment would be interpolated between the long-run difference and the cessation date difference, would basically eliminate such cliff effects. (The interpolated spread differential would only be applied at reset dates, and would not change every day.) The benefit of a transition period is the reduction in the cliff effect; the potential downside is more complexity in LIBOR transition and basis risk between derivatives and cash products. Responses to the consultation are requested by March 6th. The LSTA will help educate on the consultation, but – inasmuch as the ARRC is seeking responses from individual institutions – we are not anticipating producing an industry-wide response.