Getting the polled rates right: Financial Express 14th July 2012

After the Libor scam, the way Mibor is calculated needs to be changed. A system that uses actual traded numbers is best The recent controversy on the rigging of the London Interbank Offered Rate (Libor) is interesting as it invites debate on the best practices to pursue given the assumption that human frailty gives in to temptation to game the system. The Barclays story is quite straight. The Libor rate is determined by a poll where a fixed set of companies provide their rates, at which they have traded or feel is the right one. There is a trimming of these rates where the outliers are removed and the final number is averaged. Evidently, as the numbers given by any organisation polled is not based on actual trade data but a guess, there are ways of stating prices in conjunction with others or on their own if they have clout to keep the rate at a certain level. The rate may not matter if it is just an academic one, but when trillions of dollars of transactions, especially in derivatives across the globe are linked to this rate, then it is a serious issue. Mibor in India is a money market rate polled by Reuters and NSE, and while it is similar in terms of use, the volumes transacted may not be that large. And considering that the polled institutions are dominated by public sector banks, the risk of gaming could be lower. How is one to view this system of polling rates or determination of any spot price? Ideally, one should be using actual rates at which any product is traded, which is possible if the market is regulated. But the issue is of having these rates reported at the right time. If it is online trading it can be tracked, but OTC transactions would be out of the ambit unless it is made mandatory to report the transactions immediately. For this we need a regulator, and while RBI can impose on money market transactions, there is no such entity for global transactions. Hence, at the global level this would be impossible. Also, if one wanted to game the system, then the volumes could be changed so as to reflect a desired weight at the time of polling, say 9am or 10am, while the rest would be conducted after these hours. In stock markets, in the cash segment, transactions are online and the price is hence a revealed traded price. Price bands ensure that prices cannot be overstated or understated at any time and there are surveillance systems to ensure that the market is orderly. This is possible because it happens on an exchange which is organised and is hence the best platform for trading and settlement. But the same does not hold for the forex or money market as the OTC segment dominates. In the derivative segment, the spot price is already determined in the cash market, and the futures prices move in accordance with the spot price with bands being fixed. To ensure that the prices are not rigged by large players, position limits are placed so that no one can corner the market and drive the prices. Asking players to bring in more margin capital ensures that price movements are orderly. But, if the analogy is carried to the commodity market, there are no unique spot prices given its fragmented nature, and hence these are polled by the exchanges. This brings it back to the integrity of the polled participants. Commodity exchanges like NCDEX outsource this exercise to agencies like CMIE and NCMSL, which conduct these polls by jumbling the participants. So, from a universe of say 100 participants, a different set is polled everyday to ensure there are fewer chances of rigging the prices. This is more acute here because several commodities are not traded in the spot markets and hence do not have a ready price. But with futures trading taking place, the spot price becomes one which is guessed by participants from a wide cross section of the community. Simultaneously, there are checks carried out by the exchange in the cash market to ensure that the prices are in line with the reality. Now, Mibor is polled by Reuters in the morning based on conjectures and again in the evening based on traded numbers as reported by the participants. The two normally tend to converge. NSE does it in the morning and uses a more scientific process of boot strapping which goes beyond plain haircuts of the top and bottom quotes. In simple language, a multitude of samples are drawn and their averages and standard deviations are calculated through an iterative process and the one with the lowest standard deviation becomes the quote for the day. This is also replicated by NCDEX in the commodity market. So clearly, a plain vanilla average removing the outliers is inferior to a bootstrapped method that delivers more accurate results. Second, the samples used should be large and change regularly. If the polled participant is not sure if its quote is being used in the sampling, then the incentive to cheat is lowered. Therefore, ideally the entire universe of banks and PDs should be polled and the bootstrapping method used on a section of them. It could suffer from the exclusion syndrome for if a major bank is excluded which dominates the market, then the number may not be representative of the reality. But this will ensure the integrity of the system. Third, we should try and migrate to a system that uses actual traded numbers, which is the best option. A change definitely is required as the present system for polled rates for any product which is not regulated or traded on an exchange is susceptible to both judgment and fudge.