Comment on a Proposed Publication of SOFR Averages and a SOFR Index

The Federal Reserve Bank of New York has requested comments on its Proposed Publication of SOFR Averages and a SOFR Index. The proposal and the request for comments is available on their website at https://www.newyorkfed.org/markets/opolicy/operating_policy_191104.

My comments are reproduced below.

SOFR Averages


In general, I consider that the publication of official numbers related to payments to be made by end users in financial markets is to be encouraged. In the context of the transition of some financial products to more usage of overnight rates, proposing a ready made composition / average would be beneficial in certain circumstances.

Unfortunately the proposed 30, 90, 180-day periods do not correspond to any existing or planned financial product (except in unlikely cases when 1, 3 or 6 month exactly cover that round number of days). The introduction of averages published by an official institution but that do not cover exactly and in all circumstances the requirements of the users can only create unnecessary confusion and frustration. It would delay the acceptance of the compounded SOFR as a natural mechanism; many disagreements on the amount to be paid will be born from the proposed NY Fed SOFR Averages publication. End users may rely on the published numbers while their contract specifies something different. In the format proposed, the publication of those SOFR Averages appears a useless and even unnecessarily confusing mechanism.

As described above, the publication of averages not corresponding to market periods by itself appears superfluous, moreover the methodology seems arbitrary and not corresponding to any market standard or economic principle. In particular, the proposal states "Simple interest would apply to any day that is not a business day, at a rate of interest equal to the SOFR value for the last available business day." To my understanding, that means that if a 30, 90 or 180-day period start on a non-good business day, a simple interest would be used for that specific day, even if the market rate is for a different (longer) period. Even the idea of an interest rate starting on non-good business days does not incite to take the proposed number seriously. On the rate convention side, the simple rate used in financial markets is a quotation mechanism, not an economic principle. Most of the processes around the overnight transition is to use the natural and economically justified notion of accumulation of interest by composition. The notion that part of the interest would be accumulated on a non-compounding mechanism (simple interest) appears to negate the foundation of the composition as the foundation of the SOFR Average proposed.

The request for comments asks about other convention than the one currently proposed by the NY Fed. To my opinion, the only average numbers that would be useful are the one coming from standard periods (1, 3 and 6 month) using standard market conventions (modified following) and taking into account the different mechanisms that lead to those periods - a stand alone x-month period is not necessarily the same as a x-month period in a periodic schedule like a swap. To be useful, the average should be published on the standard periods and for all of them. This would potentially include, for one tenor, several periods with the same end date and several periods with the same start date.

The published text on the NY Fed website does not contains a single example where the published number could be used. On my side, I have not found any situation where the proposed SOFR Average could be of interest for any end user.

SOFR Index


The SOFR Index is a more interesting proposal than the SOFR Average in its currently proposed form. From the theoretical index, one could recompute the rate on any period with two divisions while the computation of the rate on a 3-month period typically require around 2x60 multiplications.

One issue will be the precision (number of decimals). As noted on the document, the rounding may lead to a difference in computed interest in some cases. The monetary value of such discrepancy is small but still exists. This would mean that when doing the reconciliation with a counterpaty, one may find a discrepancy because one party uses the Index and the other one uses the actual precise composition mechanism. This discrepancy will probably lead to investigation and a reconciliation procedure that itself will be more time consuming and costly that any advantage one would have in using the Index. Also it would mean that the institution using the index would need to store two time series with essentially the same information: the original SOFR fixing and the derived SOFR Index. This in turn could lead to further technical issues if the two are not fully aligned all the time.

In a financial institution, I would advice for not using the NY Fed published SOFR Index for actual payments in order to make sure that the small discrepancies are not creating issues with clients or counterparties. Such an index may be used internally to speed up some computation, but if this low level improvements is required by a financial institution, it is expected that it can create the index itself and probably improve it by removing the arbitrary rounding at the 8th decimal place by using the precision of their data base/computer language.

For the proposed SOFR Index, it would be interesting to have some examples of application by financial institution or end users.

Comments

Popular posts from this blog

Rigged: part 1 - Will there be a part 2?

A personal statement on the IOSCO Statement on Alternatives to USD Libor

X dominates the world