Discounting big bang: CME auction process

CME has proposed a couple of webinar related to the collateral and discounting big bang planned for October 2020. The general process is based on cash and discounting risk compensation. The discounting risk compensation involves the booking of EFFR-SOFR basis swaps in the accounts of all participants (members and clients). They clients can sell those basis swaps through an auction process (at their risk and cost). In this post, I look at some elements related to the auction as presented by CME in its recent webinars.

Some elements of the discounting risk compensation

• Cash and discounting risk compensation based on the position and market levels as estimated by CME on Friday 16 October (EOD)
• Basis swaps obtained using DV01 of the portfolio for 6 tenors (2Y, 5Y, 10Y, 15Y, 20Y, 30Y), 6 trades are booked in all accounts with clearing date Monday 19 October
• Basis swap restore approximate pre-transition DV01. There are 3 sensitivity types (EFFR, SOFR, LIBOR). At most one can be restored. To my understanding, it is the SOFR sensitivity that will be restored to its pre-transition level.

The auction is based on the following principles:

• Timing: the auction will take place on the Monday 19 October (9:00-10:00 am NY time), based on computation done on Friday 16 October
• Optional, total and bindings: participants (clients) have to declare their intention to participate (legal agreement required). The clearing members will collect intentions and pass them to CME. Once they participate, the result of the auction will be mandatory for them, what ever the result is. The participation of a client is on his full account or on nothing, no partial participation is possible.
• No guarantee: there is no guarantee that the auction will be successful. Participants may still have the basis swaps in their account after the auction.
• Maximum loss: a maximum "loss" will be set for the auction to take place. The loss is computed between the Friday level estimated by CME and the Monday auction level. The maxim loss level is decided by CME and is the same for all participants. It will depend on the market liquidity.
• Dutch-style auction: The netted total position is offered for auction. The position is divided in similar slices. Each participant bid for as many slices as he likes at any price he likes. The final price for all slices is the price of the last slice.
• The total cost is split between client participating in the auction on a pro-rata of gross DV01.
• Bidding in the auction is by invitation only. CME will decide of the invitation, it plans to invite members and potentially some non-members.
• Bidders will be invited to bid on two portfolios: the actual portfolio and its mirror image (in term of payer/receiver). The actual portfolio is not disclosed by CME to the bidders.

Some comments:

• Netting (good): if there are many offsets between clients. They don't have to pay each the bid/offer.
• Netting (bad): it is possible in some cases to manipulate the result in favour of some clients. Suppose that there are 2 clients participating with almost perfectly offsetting portfolios and one is your "friend". You skew your bids to make his portfolio cheap (and the rest of the market expensive). Your cost is very small (due to offsets for you) and his benefit is large (absence of offset for him). This type of conflict of interest will probably exists as some auction invitees will be link in some sense (same share holders, clients, etc.) to institutions wanting/requiring to get out of the basis swaps.
• The idea of "blind auction" is based on absence of information leakage. Large clearer, which collect requirements from their clearing clients, will have more information than outsiders. Even if the information is not total, they will have more information than auction participants with less or no clearing clients. They may have a good sense of which side of the auctioned portfolio is the real one. It leave the door open to skewed biding. How is the Chinese wall between those functions (clearing member and auction participant) guaranteed?
• The auction is probably good for clients on the "bad side" of the market, i.e. client for which the side (payer/receiver) is the same as the net one of the market. The don't pay the full bid-offer. But it means it will probably be bad for clients on the other side, those bringing offsets in this illiquid market. They are not encouraged. If the netted side is somehow known or guessed by the market, this could encourage "netters" to not participate in the auction and eliminate their basis bilaterally.
• Rates have decreased significantly over the last years. This means that clients on the same side of the payer/receiver market will probably be on the same side of the basis. Pension funds tend to receive fixed long term, they are globally in-the-money, with a large discounting impact (long cash flows, negative OIS EFFR exposure). The auction may create "industry competitions": pension funds vs asset managers.
• The market underlying the auction is largely illiquid. Potentially there will be crosses between bid and offers. The auction participants will not be able to take advantage of those. In some cases, it would be better to go directly to the market. The auction is no guarantee to be the best price for each client. In some sense, the auction guarantees some profit for the members (bid/offer) in exchange for the clients to have a not extremely bad price.
• The maximum "loss" is between the CME decided level based on EFFR discounting on Friday and the auction level on Monday based on SOFR discounting. The Friday level is in an illiquid market. If the CME decided level is largely incorrect or the market is moving significantly over the week-end, the auction will fail. The week-end transition is good from an operational perspective, but bad from a market movement perspective.
• Will the results of the auction be published? Not each participant name, but the bid/offer for each size and the size of the netted DV01 (not the actual side, just the gross sizes).
• Still no mention of convexity adjustment in the collateral transition. More forthcoming on my side on this in the coming weeks.

Note added: I have not found similar details about LCH auction process. I will comment on them as soon as I find them. Don't hesitate to indicate links to the LCH description in the comments if you know about one.

Note added 2020-06-14: I have contacted CME about the auction and the price transparency. They indicated that "We are currently engaged with market participants and regulators on this same topic. Our intention is to request real time SDR reporting relief on the auction to enable bidders to provide more aggressive prices. Subsequently, we would be open to making a subset of the auction results public provided that the winning bidder had adequate time to manage their risk." This is a first step, but not real transparency. A relief in reporting is not really helping transparency. Also the openness for data is about the "auction results", not the auction actual bids. The auction results are anyway public to the extend that each client participating will know the prices at which they have been served (which is the same for all). What would be really interesting, beyond the actual result, which is simply two numbers (size, rate) for each tenor, is the full market depth. That would allow to estimate the cost/benefit for the clients (netting is a global benefit, using a unique price is a cost for some participants) and the depth/liquidity of the market (amounts really offered, price range, etc.)