Dutch bank takes carrot-and-stick approach on interest rate swap for oil and gas equipment firm
Local subsidiaries of EU and US banks may be forced to adopt models before their parents
Changes to rates margin model come as bourse tweaks pricing for Eurodollar futures
COMMENTARY: The weight of history
This time last year, we were approaching an important anniversary: 10 years since the September 2008 collapse of Lehman Brothers. This wasn’t just a cue for a wave of retrospective articles washing over virtually every part of the financial media (not excluding Risk.net, though we did take more of a have-we-learned-anything tack). It also marked the point at which the direst moments of the global financial crisis would start to drop out of a 10-year loss data window.
We looked at the time at how three central counterparties (CCPs) were handling the issue with their interest rate margin models – Eurex had already ‘forgotten’ Lehman due to its shorter three-year historical window, which it supplemented with a floor based on several stress scenarios, including the crash. LCH did something similar, allowing Lehman to drop out of its 10-year window but introducing a stress floor; CME at the time announced it would anchor its window at September 1, 2008, a ‘static look-back’, retaining the memory of the Lehman collapse at the cost of an ever-extending historical window.
A year later, CME has now decided to follow suit in the face of the growing cost of maintaining and using this amount of loss history; CME will revert to a rolling 10-year window, augmented with a stressed VAR calculation based on historical stress periods including the crisis.
The question of appropriate VAR lookback window size came up last week as well, with research suggesting that setting a fixed window – any fixed window – would always be inferior to using mathematical techniques to determine the longest a window could be before including data that was too old to be relevant.
But there’s more to the roll-off issue than the details of margin calculations, important though those are. Observers are fond of pointing out that financial crises themselves tend to occur at roughly 10-year intervals, pointing at the succession of 1987 Wall Street crash, 1997 Asian currency crisis, 2008 global financial crisis, and speculating that this is the time necessary to replace the cautious traders burned by the last crisis with a new generation of gung-ho twenty-somethings who can’t see the point in risk aversion. This is a tempting theory, but not one that stands up; where is the collapse of the Japanese property bubble in 1992 in this 10-year cycle? The Swedish banking crisis of around the same time? The bursting of the dotcom bubble in the early 2000s and the eurozone financial crisis, which peaked in the early 2010s?
That grim list (and it’s not even exhaustive) suggests a different view: that the modern global financial system is not a smooth-running machine for transferring liquidity and risk which occasionally generates an unexpected crisis. It is a system of many closely-linked and volatile nodes: in network theory terms, it is a machine for generating crises. (It’s notable that the far more heavily regulated and simpler pre-1970s financial system produced far fewer systemic disruptions.) The stressed VAR calculations that all three CCPs now have in place are the ones that really matter: for all the best efforts of governments and international bodies to prevent them, crises will continue to be frequent, unforeseeable, inevitable – and severe.
STAT OF THE WEEK
The Bank of Montreal has issued C$13 billion ($9.8 billion) of bail-in debt year-to-date in a bid to meet its total loss-absorbing capacity (TLAC) requirement ahead of a 2021 deadline. Total TLAC, made up of eligible regulatory capital and unsecured debt that can be written off in a solvency crisis, hit C$60.7 billion at the end of July, up 33.5% from the end of January. Bail-in bond sales catapult BMO towards TLAC target
QUOTE OF THE WEEK
“This is an applicable piece of law today that may not be circumvented by ignoring the requirement or substituting the envisaged identifier with something else. There is no such thing as a dummy legal entity identifier. If the code is a dummy, it is not an LEI” – Esma spokesman