The procyclicality and margin debate is one of those topics that reads like a damned if you do, damned if you don’t. Earlier this month the Bank of England published an interesting paper “An investigation into the procyclicality of risk-based initial margin models” (Financial Stability Paper No. 29 – May 2014) by David Murphy, Michalis Vasios and Nick Vause. We have some thoughts.
Initial margins (and their first cousin haircuts) are often lowered when markets are more stable and presumably less risky (and raised when market risk goes up). Statistically based risk models helped along by competitive pressures are driving the train on this behavior. But lower haircuts have been criticized as facilitating a buildup of leverage that, in turn, enables crises.
Hiking haircuts or initial margin is seen as equally problematic, especially when it forces the collateral poster over a cliff, taking the financial system along with it. Just think about AIG – which the paper cited — and the need to post collateral once their ratings were cut.
From the paper:
“…If initial margin increases substantially, the requirement to post margin on a timely basis may pose a substantial liquidity burden on the poster, often just at the time when they are least able to bear it…”
The solutions are somewhat unsatisfying. For example, regulators have suggested raising minimum haircuts to subdue excess leverage. Increases in haircuts means investors have to put more skin in the game to buy a security, reducing leverage. But regulators haven’t asked for haircuts to go up too much, since that could trigger the cliff problem. We saw this from the FSB where one proposal they suggested in their paper “Strengthening Oversight and Regulation of Shadow Banking, A Policy Framework for Addressing Shadow Banking Risks in Securities Lending and Repos” was to raise minimum haircut levels. So far, this has gone nowhere, but that could change with all the talk from regulators on haircuts. We wrote about this in a post “Some thoughts on the FSB mandating repo haircuts” back on November, 12, 2012.
How to deal with the other side of the equation? Advocating that haircuts not be increased when volatility would otherwise suggest they should be is securities financing heresy. From the paper:
“…When markets become more volatile, risk is higher, and hence margin requirements should be higher. But it is undesirable for margin models to overreact to changing conditions…”
Regulators know this and have attacked the problem in other, somewhat indirect, ways. By increasing capital, banks can better absorb either their own margin deficits or a client failure based on (inability to satisfy) increased margin demands. This includes not only initial margin or haircuts but – importantly –variation margin problems as well. Recent focus on systemically important asset managers could extend regulator’s control on leverage to these important end clients. The TBTF debate (and we suppose the solutions should they ever appear) address the interconnectedness issue that was at the core of AIG’s bailout.
The paper examined several different volatility-driven initial margin risk engines. We won’t go into the model specifics beyond the simplification that the base-line is using historical data assuming constant volatility (99th percentile, normal distribution) while other models the authors examined layered on top variables driven by most recent changes in volatility (either smoothing out data — making it less reactive — or doing the opposite), impose floors on volatility to create minimum a margin amount or a VaR approach that would establish margin based on neutralizing the Xth worst loss (in this case the 5th). It makes for good, if slightly wonky, reading – especially for credit officers looking for different ways to think about margin and exposure.
But how can you create a margin model (and policies) that will at once cover risk when market volatility demands it, but at the same time not be oversensitive, risking a push to leveraged players (and their bankers) over the cliff? The two goals seem mutually exclusive. From the paper:
“…we want [models] to ignore ‘temporary’ changes in volatility caused by noise or estimation error, and we want them to react quickly to permanent or structural changes in volatility. These two goals are in opposition to each other…”
In the end there is no golden model and no perfect solution. Pushing on one side makes the other worse. The article does suggest several different measurements that could be used – peak-to-trough and n-day pro-cyclicality measures – which would “…help clearing members and their clients to anticipate potential increases in margin requirements, and to prepare accordingly…” Peak-to-trough and n-day both are basically measurements of how much margin might be required in a scenario. It is hard to argue with advocating every client understanding how much margin they might have to pony up on a given trade or portfolio.