2014-09-26

Change of benchmark overnight index is a difficult task

In the book, I indicated why I dislike the term “OIS discounting” (OIS = Overnight Indexed Swap) as it hides the complexity of the valuation process and the numerous hypothesis and results required to achieve the valuation formula. The correctness of the approach depends on a number of market conventions and practice, unstable to my opinion. Unstable in the physical sense, that a small perturbation to the situation can destroy the whole process, not only move it slightly.

This blog entry is written in reaction of the Risk article titled OIS rate change easy to absorb, says ISDAs O'Connor (subscription required). As you can guess from the introduction, I disagree with the “easy” part of the title and I believe the term “OIS” used in the title is as misleading as the one used in “OIS discounting”.

If I understand correctly the article, there is a General Collateral (GC) overnight index and some suggestion, in particular by the US Federal Reserve, to change the standard index from Federal Fund
Effective rate to the GC index as a reference for the OIS swaps, as a rate paid on cash collateral under bilateral Credit Support Annex (CSA) agreements and as a rate for Price Alignment Interest (PAI) by central counterparties (CCPs). The ISDA’s chairman seems to think that this is an “relatively easy” change.

The index used would be the GC index published by DTCC: http://www.dtcc.com/charts/dtcc-gcf-repo-index.aspx

Let’s start by a positive comment. The cash collateral referred to in CSA applies, by definition, to a collateralized situation. The derivatives are collateralized by the cash and the cash is collateralized by the derivatives. It seems appropriate to use a collateral rate to remunerate the cash; the GC index is an appropriate index.

The idea of using another index is not a bad one in a stable situation. But the proposal does not only concern a stable situation, it is supposed to apply to a change to the new index. A change is a very different proposal. The “OIS discounting” by clearing houses and for bilateral contracts works only because
  1. The rate paid on the cash collateral (and the PAI) is linked to the same overnight index for its entire life.
  2. The rate reference in the OIS is the same index.
  3. The OIS themselves are collateralized with the same rule as the other instruments for the entire life of the OIS.
Suppose we decide to change the rate paid in PAI in central clearing to the GC index. To create the stable situation that allows a clean and standard valuation, the following steps are required.
  1. Create a liquid OIS market linked to GC.
  2. Make sure that bilateral traded new OIS refers to a (new) CSA related to GC collateral interest
  3. All dealer have to sign new CSA agreements for the new trades
  4. CCPs create a new category of trades with GC collateral; it is not possible to net the existing trades and variation margin as the trades have different reference collateral rates and no full netting is possible. Cross margin for Initial Margin (IM) is probably possible, but not at the Variation Margin (VM) level. Or more exactly a payment netting is possible but not a netting of new and old instruments in term of positions.
  5. Run, up to the maturity of the longest trade existing in the world linked to fed funds, two parallel markets, one for the Fed Fund trades and one for the GC related trades. This includes computing convexity adjustment between the different trades (good for me, more work for quants, see Theorem 8.11). The dual collateral rules require each user to run a double set of multi-curve framework with collateral, one with each collateral.
  6. As an alternative to 4/5, users with existing position could decide to cancel the existing trades. That would require each individual user to agree on the valuation of each position and agree to cancel it for that price. The cancellation could be replaced by an alternative move from one collateral to the other, but in the same way all the parties have to agree on the valuation (as an up-front fee or a fee over the life of the trade) impact of the transfer.
Changing the overnight benchmark index has a greater impact that changing the LIBOR benchmark index. The overnight index is used as reference for OIS but also, and this is the trickiest part, as a reference index for standard CSA collateral and CCPs. Not only all the trades that refer directly to the index are impacted, but also all the trades that refers to it indirectly, through collateral rate or PAI. This is roughly all the OTC interbank market.

My conclusion: OIS rate change very difficult to absorb (due to foundations and implementation issues, to parody the subtitle of the book).




P.S.: I understand regulators wanting to improve the situation. There is an important item that they seems to ignore: transparency. Regulator are there to protect the public, not private interest. My request to regulators is that if they push for a new benchmark, the first point they have to ensure is that the benchmark is public information. All the historical data of the benchmark should be available publicly, without conditions, without fees and without registration. The data should be public as soon as available, no “insider information” is allowed, even for one minute. This does not seems to be the case for the moment.

If the regulators are ready to enforce the transparency and if they are willing to work with knowledgeable people (from the industry and academics) to analyze the direct and indirect impact of changes, I’m ready to support them. My contact details are available in the blog!

The graphs below represent the fixing of the two indexes over the last year and the a posteriori discount factors using the fixing (i.e. how much was required at those dates to get 1 USD today by investing overnight at the fixing).

Comparison between the Federal Fund Effective index and the General Collateral index.

A posteriori discount factor from the indicated dates to today (25 September 2014): Fed Fund vs GC.