In his latest scientific publication, Professor Michael Tröge and his co-author studied Libor, a widely-used interest rate benchmark. Their key contribution is to demonstrate how a market benchmark that does not rely on market data, but on non-verifiable “cheap-talk” announcements can function well and even improve the functioning of the “over the counter” interbank markets. But not in the advent of a crisis… which helps understand its woes and could contribute to a deeper understanding of other market benchmarks, some of which have also come under scrutiny.
The London Interbank Offered Rate (Libor) is an interest rate benchmark that is supposed to reflect the borrowing conditions in the interbank market. These benchmark rates are used widely as reference rates for financial contracts. For example, interest rates on student loans and mortgages as well as loans to small businesses and large corporations often depend on Libor.
Libor has had its fair share of controversies, including a major scandal of rate rigging following the publication by the Wall Street Journal of a study suggesting that banks might have understated borrowing costs they reported for Libor during the 2008 credit crunch. The ensuing scandal was one of the primary reasons why its administration was shifted from the British Bankers' Association to Intercontinental Exchange in 2013. It will be abandoned in 2021…
Non-verifiable information can and did improve the functioning of interbank markets
But for Prof. Tröge, the surprising fact is not so much that Libor was manipulated, but rather that Libor functioned so well for such a long time. “We studied the Libor benchmark setting mechanism, which like other interbanking rates was until recently based on judgmental estimates of borrowing costs,” he explains. It is not obvious why in this type of situation, banks should truthfully reveal their true borrowing costs, given that lying was difficult to prove and penalise. “We constructed a model of this setting as a search market where banks communicate non-verifiable information about their opportunity cost to potential counterparties, interpreting it as a cheap-talk game.”
Cheap-talk games were developed in game theory to analyse “communication between players that does not directly affect the payoffs of the game.” In these models, one actor has information and can choose strategically what to say and what not to say, the other one has the ability to act upon this information. In cheap-talk games, information can only be transmitted if the sender and receiver of the information have overlapping interests.
In the article they published in The Review of Financial Studies, Michael Tröge and Durham University Business School’s Prof. Ángel Hernando-Veciana show that this is true for the interbanking market under normal market conditions. Non verifiable signals such as the estimates of borrowing costs in the Libor mechanism can then contain true information, and lead to a welfare-maximizing equilibrium where banks truthfully disclose their borrowing cost in order to obtain the optimal loan offers from the lenders.
Unfortunately, in times of financial stress, the interests of borrowers and lenders in the interbanking market differ. In this case, cheap-talk communication breaks down as lying becomes now profitable. If the difference between what borrowers and lenders want to achieve does not become too large, only “coarse” information survive where instead of precise numbers only approximate intervals are indicated. This could be the reason why the Libor mechanism failed during the financial crisis between 2008 and 2011.
A deeper understanding of market benchmarks for better reforms?
Professors Tröge and Hernando-Veciana explain that their model predicts a number of patterns in the precision of the banks' submissions that can be identified in the data and seem difficult to explain otherwise. They argue that submitting rounded numbers is a simple and intuitive way to implement “coarse” cheap-talk equilibria that their model predicts in a crisis. “We tested our model’s empirical implications with data from the Libor benchmark setting process (each bank's individual submissions), showing that it can explain a number of so far undocumented patterns in the precision of the Libor submissions: banks round more frequently if the risk of the bank increases. Rounding is also more frequent for the more liquid short-term rates and certain benchmark maturities,” add the ESCP Business School Professor of Finance and his co-author.
They think that their search model is a reasonably realistic approximation of how Libor worked in the early days. It explains why the surprisingly informal Libor mechanism largely performed well and allowed Libor to become a widely followed benchmark. “We hope that beyond the insight generated about the Libor process, our model will contribute to a deeper understanding of other market benchmarks and benchmark setting mechanisms. Market benchmarks are used in many illiquid OTC (over the counter) markets, and the calculation of benchmarks in these markets is based on a bewildering range of different mechanisms involving past transactions, binding or partially binding quotes and pure cheap-talk signals,” they explain.
Following the Libor investigations, a number of these other market benchmarks have come under the suspicion of manipulation: in addition to interest rate benchmarks such as ISDAfix, RONIA and SONIA foreign exchange benchmarks such as the WM/Reuters FX rates as well as commodity benchmarks such as the Gold/Silver Fixings and energy benchmarks such as the Platts, ICIS and Argus have recently been investigated. Their results should help reforming these benchmarks in a way that preserves their efficiency-enhancing properties…