Thursday, January 16, 2014

Fixing MIBOR: Comments on the Draft Report of the RBI Committee on Financial Benchmarks

The problem

LIBOR is made by asking a few dealers what the price on the market is. The LIBOR scandal involves allegations that dealers in London made up the numbers. The dealers at twenty major global banks chose values that were convenient for the banks. This is considered terribly unfair.

Understanding the LIBOR scandal

The inter-bank money market is an OTC market. It is based on conversations between counterparties and when a position is taken, there is credit risk of the counterparty. At the height of the global crisis, this market broke down. In August 2007, HSBC and Citibank were not willing to take each others' credit risk in London. When HSBC and Citibank were not comfortable doing transactions on each other, they were also not willing to do inter-bank unsecured money market transactions. As a consequence, the underlying inter-bank unsecured money market became dramatically illiquid.

In an illiquid market, and particularly in an illiquid OTC market, the very participants in the market do not know what is the market price. Nobody knows what is going on.

Every morning, quotes had to be supplied by dealers to the BBA. There were two choices:

  1. Admit that the unsecured money market had broken down. This would have generated a legal crisis for myriad contracts which are predicated on LIBOR. A large fraction of those would go into litigation. Many contracts would have deemed to be in auto-default in the absence of LIBOR. This would have amplified the market meltdown.
  2. Or, make up some numbers and proceed. The dealer at HSBC felt that Citibank was bankrupt and was not willing to lend to Citibank. However, for the purpose of the polling, the dealer at HSBC could ask himself: If Citibank were solvent, at what price would I would be willing to lend to him? The rates reported were lower than those at which the fragile entities could borrow, as the credit risk was ignored.

The banks chose option 2. Government agencies were aware of this choice and complicit in it. They supported option 2 as they did not want one more spanner in the works of the global financial system. Many central banks privately talked with banks to help guide LIBOR rates since LIBOR has a big impact on the rate at which governments borrow. As an example, the Bank of England did this with major British banks operating in the LIBOR market.

I believe that the decisions of the banks and the financial agencies, to keep the boat aloft with invented LIBOR rates, were good for the world at large. Under any such messy arrangement, there are bound to be individual episodes of excesses, but at a strategic level, I think what was done was good for the world economy. I believe there is less criminality in the LIBOR scandal than is sometimes claimed.

Trade prices are not a panacea

Can such problems be avoided in the future? Suppose we switch from a reference rate made through polling, to the volume weighted average (VWA) of trades. This assumes that all the trades are observed in some trade reporting or trading or clearing system. This approach has three problems:

1. It does not solve the LIBOR crisis situation : What if there is no trading and the VWA makes no sense? For a contrast, the bid and offer price on an order book in an electronic exchange represent the market price even if there is no turnover.

2. The reference rate must be a rate that you can reliably attain in a trade. Suppose we make the VWA of all the trades of the day. This is an unambiguous number. The trouble is, this rate is not implementable for arbitrageurs. Suppose you are doing spot/futures arbitrage with a cash-settled futures. On the expiration date, the futures contract is worth the same as the spot, so you need to unwind your spot position alongside the termination of the futures contract. This requires a predictable time at which the spot unwinding is done. The VWA of the whole day is not replicable for a trader. Also see.

In contrast, a reference rate at 11:00 AM is replicable for a trader: he just does the unwind at 11:00 AM.

The Nifty reference rate is a very successful VWA, as it has been the foundation of one of the world's great index derivatives contracts. It is made as the average of 30 minutes of trading, from 3:00 PM to 3:30 PM. Why does this work well? The NSE spot market is an extremely liquid market. Hence, there is a feasible implementation algorithm through which an arbitrageur can attain this price: To split up each required trade into 100 trades that are placed once every 18 seconds. The average execution obtained over these 100 trades comes close to the Nifty spot that's computed as the VWA over 1800 seconds. This implementation strategy is hard for a human, but it can be done using algorithmic trading.

If Nifty was a less liquid market, then this uniform placement of orders (equal sized orders every 180 seconds) would not reliably yield VWA-execution so the reference rate would then be unattainable.

3. Market abuse. With a polling based MIBOR with 10 dealers (say), you'd need four of them to collude to have a significant impact upon the rate. For more on robust estimation with polled prices, see link and link. In an illiquid market, one person can do circular trading against himself, and dominate the VWA! The polling based MIBOR is vulnerable to market abuse but so is the VWA, particularly in an illiquid market. There is ample experience in India of bad people who `paint the tape' doing circular trading through which apparent turnover is driven up, closing prices are faked, and so on. Our capabilities on enforcement against such market abuse are weak.

These three arguments show that while the VWA is a good technology under certain circumstances, it is not always superior. The VWA and polled rates are not the only choices for dealing with this question. There is an important third choice: Call auctions. Call auctions can be particularly useful when the market is relatively illiquid. Even with Nifty, I believe that we will do better on measurement of the Nifty close by using a call auction (as we do with the Nifty open).

In short, going from a polled reference rate to a VWA is not a free lunch. There are three alternative technologies: polled rates, VWA and call auctions. There is the need for judgement and experimentation in figuring out what works. For the VWA to work well, five conditions should be satisfied:

  1. All trades on the market of interest should be observed in a central trade reporting system.
  2. The window of time over which the VWA is computed must be small (e.g. Nifty uses 30 minutes; ideally it should be even shorter, more like 15 minutes).
  3. The market must be very liquid so that one reliably gets execution within that short window of time, and market abuse through circular trading becomes difficult.
  4. It should be easy to setup algorithmic trading to implement (say) 100 small trades over 15 minutes or 30 minutes so that it becomes possible to achieve execution at the reference rate.
  5. There should be a regulatory capability that detects and enforces against market abuse aimed at falsifying the VWA.

Similarly, conditions can be articulated under which polled rates are optimal, and when call auctions are optimal. These are subtle questions of design of information series and financial products. Financial capitalism solves these problems through a process of crossing the river by feeling the stones. No one technology dominates under all circumstances.

Analysing a recent RBI committee report

RBI has a recent report on benchmarks. They say:
  1. The ownership and control of MIBOR must shift from NSE to FIMMDA.
  2. MIBOR must become a VWA from 9 AM to 10 AM.
  3. The report suggests many other things such as: "Construction of the G-sec yield curve may use volume weighted average rate of the trades executed over longer time window in place of last traded yields" and "FEDAI may stop publishing spot fixings, if it is not used for any meaningful purpose by corporates and other clients". 

I have a few concerns about this.

  1. If it ain't broke, don't fix it. There was no MIBOR scandal. The report has not demonstrated that there is a problem with MIBOR. (The statistical methods in MIBOR are a bit better than those in LIBOR).
  2. Confusion of terminology. A VWA-based rate is different from MIBOR. It would reduce confusion all around if the new product did not use the name MIBOR. The name MIBOR is taken, and it means : a polling based reference rate. In fact, for contracts that are in flight (and MIBOR has dominant market share today), it would generate legal risk to make an in-flight change over from the polling-based MIBOR to a VWA-based MIBOR, a change that cannot be papered over by holding the name constant. The VWA-based rate can have some other name like VIBOR. For an analogy, a new method for measuring the stock market index, launched by NSE in 1996, was called Nifty and not Sensex.
  3. Why choose one when you can have both. I don't see why we have to close down a polling based MIBOR. Why not do both? All that's required is to have a VWA-based rate competing with MIBOR, and the market can figure out which is better. The CCIL MIBOD/MIBOR is a VWA, has been around, and is competing for influence.
  4. Expropriation alert. I do not see how the event of the LIBOR scandal leads to the recommendation that the coercive power of the State should be used to take MIBOR away from its present owner and give it to FIMMDA. Is this mere turf extension by RBI? Do the laws authorise RBI to expropriate, and even if they do, is such expropriation wise? For an analogy, the reformers of the equity market never proposed that BSE should be expropriated in its ownership of the BSE Sensex. Far from expropriation, there was no State coercion of any fashion in the emergence of Nifty.
  5. The proposed solution has problems. There are five requirements to make a VWA work well (listed above), and the Indian money market does not satisfy them. The proposed window of 1 hour is too long. The spot market is not too liquid, and it is not an active electronic market where we can break up a trading problem into 200 trades, one each 18 seconds, implemented through algorithmic trading. The threats of market abuse will not go away. Shifting to a VWA will impose costs, in return for no visible gain. If the VWA-based CCIL MIBID/MIBOR were compelling, they would have passed the market test.
  6. Cost-benefit analysis. Under the Handbook, page 39, RBI will need to do a cost-benefit analysis before undertaking interventions. It is not clear to me that the cost-benefit analysis of these interventions (expropriation, forcing VWA instead of a robust reference rate) is a win.
  7. Strengthening the polling-based MIBOR. MIBOR construction should use data from a large number of dealers, and these dealers should be as diverse as possible. The problem of the monoculture of the Indian bond market, with domination by banks, implies that by default, we may end up with too many bank participants in the panel that's polled. We should work on obtaining a large and diverse panel, and on giving adequate incentive to dealers to bother to give sound information. A lot can be done on these lines, but I saw none of this in the report.
  8. What about the third strategy? Call auctions may be better than the proposed VWA.
  9. Central planning alert. A financial agency should not be a super-CEO of the financial system, determining who makes what information series and how. Financial agencies must not do industrial policy, picking winners among three rival technologies. Financial agencies should identify and solve market failures. There are difficulties with all three technologies, and the coercive power of the State should be devoted to solving market failures, with neutrality about technology. Policy interventions should be undertaken to make all three technologies function properly, so that private persons who develop markets and products have the full choice between VWA vs. polled rates vs. call auctions. It is not for a financial agency to (say) decide how statistical estimation of the yield curve should be done. The RBI report seeks to replace a self-organising system by the decisions of bureaucrats. Bureaucrats are likely to fare poorly in imagining and innovating on information measures. Private persons in the market economy cross the river by feeling the stones, and reverse themselves from numerous mistakes along the way; this method of nonlinear maximisation gives better outcomes when compared with intelligent design.
  10. Reduced incentives for R&D. Information systems in India were developing on their own, based on the initiative of diverse persons on the market (e.g. Reuters, NSE, FEDAI, CCIL, etc). No bureaucrat would have thought of the innovations that went into information series like Nifty or MIBOR or PPI. When pioneers who create new measures are expropriated, there will be less incentive to think and invest in research and innovation in the future.
  11. Rule of law. The report indulges in central planning without making clear the legal foundations of all the interventions that RBI will use to do these things. The report acts as if RBI has unlimited powers to give orders to a variety of organisations. Can the report be implemented through regulations that are made through the process spelled out in the Handbook and that would withstand judicial review? I doubt it. Many times, I fear, what may be envisaged is verbal instructions or intimidation by RBI. If this happens, it is inconsistent with the rule of law.
  12. MIBOR is likely to matter greatly in the future. Looking into the future, a key objective for India is to build a Bond-Currency-Derivatives Nexus. This requires futures on the short interest rate, which has thus far been banned. A natural candidate for the underlying is MIBOR. There is a natural possibility of doing something like the CME Eurodollar Futures contract in India. Building towards something like the CME Eurodollar Futures contract should be high on our radar, in the project of constructing the Bond-Currency-Derivatives Nexus. We should be very careful before damaging MIBOR, which is a critical ingredient for this possibility.
  13. FIMMDA as a custodian of MIBOR? Who has the best incentives to build towards something like a Eurodollar futures contract? A club of OTC dealers (mostly banks) has an incentive to pursue the rents that come from opacity and barriers to access to markets. From this point of view, an exchange is a better custodian of critical components of the Bond-Currency-Derivatives Nexus rather than FIMMDA. If anything, there is a greater possibility of extreme cooperation, that led to the LIBOR scandal, in a monoculture of bank dealers.
  14. Think it possible that we may be mistaken. In the future, we may decide that a polling based rate was better after all. In this case, we would have lost a chunk of the MIBOR time-series when we closed down the polling-based MIBOR, and this gap can never be overcome. Future estimation of risk models for a polling-based MIBOR will forever be weaker with a gap in the time-series. Let's not do things that we may regret in the future.

1 comment:

  1. a remarkable piece upholding the precious traditions of the rule of law as much as it is an argument in favour of facilitating market innovations that cannot be allowed to be snuffed out. In the longer run it is the market that will create benefits as thought through by the actual players


Please note: Comments are moderated. Only civilised conversation is permitted on this blog. Criticising me is perfectly okay; uncivilised language is not. I delete any comment which is spam, has personal attacks against anyone, or uses foul language. I delete any comment which does not contribute to the intellectual discussion about the blog article in question.

Please note: LaTeX mathematics works. This means that if you want to say $10 you have to say \$10.