ANALYSIS: Industry sets sights on anti-procyclicality regulation

The clearing industry has for years been thinking about procyclicality, the phenomenon where the margin models deployed by clearing houses inadvertently increase volatility at times of market stress.

But the issue was brought into stark relief some four years ago by the sell-off that followed the escalation of the coronavirus epidemic.

Given clearing house margin models respond to volatility among other factors, the volatility that erupted in March 2020 led to spikes in margin calls, forcing trading firms and brokers to liquidate positions to find the margin they needed, effectively adding to volatility. The industry then was briefly stuck in a vicious circle that was not broken until volatility finally abated.

March 2020 was the worst recent example and the industry has apparently learned some lessons from that time, particularly around the predictability and transparency of margin models, which have meant things have not been as bad in more recent periods of volatility, such as the trading that followed the Russian invasion of Ukraine in February 2022.

But the issue has not gone away and regulators are still struggling to come up with a framework that can be rolled out universally.

Speaking at a conference last week, Dr Pedro Gurrola Perez, head of research at the World Federation of Exchanges, said: “Through time, margin models have changed. When I first started working in this space, there were still some very simple models because clearing required simple models for simple futures and options strategies but now more complex models are required for clearing CDS (credit default swaps) or more complex products.”

Dr Perez said central counterparty (CCP) margin models have also responded in recent years to changing regulation, of initial and variation margin procedures for example.

“Choosing a model and understanding how a model behaves and what are the consequences of that choice have implications for the different ways in which we address risk but also issues like procyclicality,” Dr Perez added.

Udesh Jha, managing director, head of quantitative modelling at CME Group, said the industry’s initial portfolio-based margin models have stood the test of time but he agreed with Dr Perez the industry’s practices need to evolve.

Jha said: “We believe modelling for procyclicality covers every aspect of the risk model starting with data preparation. If you don’t have good historical data, you will never be able to manage procyclicality because portfolios can change and you might find you simply don’t have the right data for a particular risk profile.”

He added: “Filtering is the general method that people use for building scenarios, also known as EWMA (exponentially weighted moving average) or Lambda, and choice of Lambda is very important for procyclicality because it affects how you are calculating margin. In our experience, almost everything that goes into the margin model has an impact on procyclicality.”

This was a theme developed at the conference by Dr David Murphy, visiting professor, Department of Law at the London School of Economics.

Dr Murphy, who published in 2014 a measure of procyclicality, said: “Procyclicality can be measured but it is a random variable because it depends on the path of returns, that is, something happens in the market and, as a result, margin changes. Obviously how much the margin changes depends on the sequence of events.

“We know we can’t predict financial times series in the future, rather we think of them as random variables so margin changes must be random variables too but that isn’t to say they don’t have some structure.”

Dr Murphy said his 2014 work was based on a “large call measure” looking at margin spikes in five day periods of extreme market stress.

“That measure depends on the precise path of returns but what is less obvious is that if you compare two different models with that measure, sometimes you get one ordering where one is more procyclical than the other, and sometimes you get the opposite ordering, and you don’t have to make a very large change in the return series for the ordering to flip.”

Dr Murphy added: “We need to get away from this dependence on paths of return so we need to think not about the repeat of a particular situation: a repeat of 2008, the commodities market after the Russian invasion of Ukraine etc.

“Don’t pick a single set of events, we need to think instead more broadly. So what we are going for here is a way of comparing initial margin models that can be used to say genuinely this model is more or less procyclical than that one, that doesn’t depend on a particular set of returns and also, ideally, captures some uncertainty.”

Dr Murphy and Dr Perez presented at the World Federation of Exchanges conference last week a new approach to anti-procyclicality based on impulse response function, commonly used in lens design, audio amplifier design and process control systems.

Murphy continued: “We’re going to take a known time series of volatility so we’re going to look at a step function of volatility so … things are fine then a bad thing happens and volatility increases and stays at that higher level. How does the margin model respond to that? But, because we want to get away from a particular path of returns, we’re going to look not at one path but many paths. So we’re going sample a whole bunch of paths consistent with that time series of volatility and examine the margin that comes out.”

Using this metric, the widely used filtered historical simulation (FHS) models are not fit for purpose. “FHS overreacts, it calls for more margin that it needs on average until the early returns are out of the data series and then it gets to the last level. This has got a decent claim to be that peculiar and elusive beast – excessive procyclicality.”

Dr. Murphy added: “We think this impulse response function approach provides a useful tool for comparing margin models that doesn’t depend on a particular risk factor scenario. It allows us to capture model reactiveness to changes in volatility and the use of many scenarios allows us to estimate dispersion or uncertainty in that reaction.

Murphy concluded: “We certainly believe there is no correct amount of procyclicality so we do not believe regulators should set limits on what the number reaction of the model should be. At the end of the day, procyclicality is all about the burden on the users of the CCP and their access to liquidity.”