What are the qualities required from an interest rate benchmark? What would a great pretender to the benchmark throne look like?
The qualities should include (in no particular order)
This does not need a lot of explanation. If you have to pay for something, you don’t want that a someone else is able to decide how much you have to pay without your say.
By opposition to a random number, the benchmark should be based on real economically meaningful items that can be observed by the general public and not only the benchmark administrator. A guess by a self-selected group of benchmark setter, is not an acceptable approach. The best would be to base the benchmark on public information. But there is only a limited amount of public information available in a timely manner in finance. Nobody want to disclose in real time all its transactions. If the benchmark is based on actual transactions, it will probably be difficult to have it really observable.
It should be relevant, not only as an economical abstraction, but as a reality for the users. This is very difficult, as there are two parties in a contract, but here I will take the point of view of the end user, the retail borrower, the corporate, etc. Why would an interbank borrowing rate be relevant to a retail mortgage? For a very specific bank worker, it may be the case as his bonus is linked to the profitability of the bank which itself is negatively correlated with the interbank credit spread. For him, a reverse LIBOR spread mortgage may make sense. But beyond that far-fetched example, it is difficult to see why a retail mortgage would be linked to LIBOR. Similar remarks could be made for municipalities borrowing at rates linked to LIBOR or ICE swap rates (formerly ISDAFIX). See also the ’term’ section below.
It should be possible to trade the benchmark underlying directly, not only derivatives on the benchmark. If we take today’s LIBOR, you can trade futures on LIBOR 3M, but you can not trade the numbers used in the benchmark calculation. The numbers are abstract items existing only in the head of the rate setters, you cannot trade the neurons of the rate setters. An alternative approach would be to let the futures trade up to 11:00am and use the last traded futures price, not only as a settlement price for futures, but also as the fixing for the LIBOR. The LIBOR so determined would be used in other derivatives like swaps. This method of fixing as certainly many drawbacks, that I don’t want to consider here, I just want to point one advantage: the benchmark is directly tradable. You can hedge the fixing and you can influence it, in the economical sense, by trading the future. I have done a similar proposal a couple of years ago for deliverable swap futures where the last settlement price would be used as a fixing for an index used in CMS-like products. In that manner, you need less additional compliance mechanism, the market is your compliance department.
The benchmark should be based on information that are public, with clear rules that can be checked by external parties, not only the benchmark administrator.
I will include also under the ‘transparent’ qualification an element that I have not seen in other places. The benchmark itself should be public information; the numbers should be made public as soon as known, no license or fee should be required to obtain the number. Any analysis, of current data or historical data can be run by all interested parties, including any party in a trade, academics, regulators, etc. All the data used for the computation of the benchmark should be fully public as soon as known, no delay in the publication is allowed. Allowing a license/patent/fee on benchmarks used by retail (I’m not speaking here of the interbank market, where professionals are on an equal footing and free to chose their weapons), is creating an undue legal protection to block the fair assessment of the situation.
In my definition of transparency above, I see a clash between free enterprise and freedom to inform. LIBOR values should be considered as news, not proprietary corporate secret. Full access to the data should be granted to all. Maybe this requirement could appear in some retail consumer protection requirement: if a retail product is linked to a benchmark, full transparency and access to data by all is required for that benchmark. For systematic benchmarks, data access should be granted to researchers (academic, independent) for public research on them. No patent can be declared on news, truth and formulas. The benchmark values should be considered as news. Academics and regulators — and in general any researcher — should be free to develop analysis, including writing code, of the benchmark and publish their results about it, including the code, the data, their findings and their opinions.
What do we need a benchmark for? For saving time in lending activities and related derivatives. Most of the lending is done for a term, i.e. the borrowing is planned to be reimbursed at a agreed date. Credit lines are different, but do not represent the majority of the borrowing activity. Most of the lending is for a period longer than a single day. The alternative rates that have been presented are all overnight based. In that sense they all fail to be relevant for most mortgage and corporate lending. I would prefer to have one of the new benchmarks related to term rates, like 3-month or 6-month rate, rather than all of them overnight.
Some discussion about the alternative rates mention ‘risk free rates’. It is in particular the case of the Bank of England sponsored Working Group on Sterling Risk-Free Reference Rates. Why do we need a risk free rate when risk-free activities, like risk-free money lending, do not exist? It is useful to have a reference rate. LIBOR was a reference rate. The fact that its design was bad and opened the door to obvious manipulation does not remove the fact that it was a reference rate and it that sense did a good job for 30 years. We want to change it, fine for me. The first question is, what is the target. Certainly a risk-free rate is of no practical use to anybody in practice and miss all the above qualities. Can a practical benchmark be created on a risk-free rate? Another question altogether. To my opinion the answer has to be 'no'; risk-free rate is a theoretical concept, you would need a theory to compute it from practical measurements. Then someone else will need a potentially different theory to compute, from the reference risk-free rate, anything he wants to use in practice. I don't see any interest in there.
I confess I have used risk-free rates in theoretical development. But only as an intermediary step, as a guide to intuition. No final formula or price never depended on knowing it or estimating it.
Do we want a unique benchmark to dominate all of them, like the old king? Or several specific benchmarks? The first quality I mentioned was relevant. This implicitly mean multiple benchmarks. This will also mean that the multi-curve framework is here to stay ;)
In a comment to a previous blog, a reader asked about multiple fixing for same benchmark at different time of the day. I understand that this can be handy for professionals to have a fixing at a time which is convenient, like the time they revalue their books. But from a end user perspective, I’m not sure it make really sense. If my mortgage rate is fixed once a year, I don’t really care if it is done at 11:00am or 3:00pm. Also multiplying the fixing time would multiply the derivative underlyings. Do we really want to create a basis swap market between LIBOR 3M 11:00am and LIBOR 3M 3:00pm? I would prefer not.