Pro-cyclicality and repo: do regulators understand how haircuts are determined?
Pro-cyclicality is every regulator’s favorite whipping boy right now. Regulators pound the table over ways to offset the tendency to ease financing terms when the market is strong, only to pull back when the markets stumble. Repo haircuts are ground zero for the debate. We think most people have gotten the cause and effect wrong.
First, what is pro-cyclicality? Investopedia defines “procyclic” as “A condition of positive correlation between the value of a good, a service or an economic indicator and the overall state of the economy. In other words, the value of the good, service or indicator tends to move in the same direction as the economy, growing when the economy grows and declining when the economy declines.”
Are haircuts in repo pro-cyclical? Not necessarily. A May 2, 2012 article in the FT (sorry, behind the pay wall) by Richard Comotto described the connection between pro-cyclicality and repo as such: “When confidence is high, asset prices are buoyant and perceived risk is low, haircuts are reduced. Collateral generates extra cash, which fuels asset purchases. Prices rise further and sentiment is reinforced, so haircuts are reduced again, releasing more liquidity. The result is a self-reinforcing upward spiral into excess leverage. But when a shock damages confidence, asset prices fall and perceived risk increases, so haircuts are raised, choking off liquidity. Borrowers are forced to sell assets, depressing prices further and worsening the perception of risk, so haircuts are raised again. The result is a self-reinforcing downward spiral into crisis.” This implies a cause (buoyant/weak markets) and effect (lower/higher haircuts). It makes a nice soundbite, but the reality is more complicated.
Lets remember what haircuts in repo are for. Haircuts absorb the risk of a security falling in price during that period of time between one side of a repo trade defaulting and the collateral being liquidated. Those haircuts are determined based on a VaR-like analysis. How far down will a security fall (or, for that matter, rise) given X% confidence over Y days? The analysis (at least for fixed income) is driven, more often than not, by maturity (lower PV01s=lower haircuts). The shorter the paper, the less prices change for a given yield shock and the lower the haircut necessary to protect the cash lender.
Then what drives a change in a haircut? In the bilateral world, where haircuts can be negotiated (in theory) on a trade-by-trade basis, credit people looks at the underlying collateral’s volatility. If the vol goes up, then the potential distance traveled during the default to liquidation period will go up and haircuts rise accordingly. And visa-versa. Feeling good about the market might be correlated to lower volatility (or visa versa), but it is not the cause of lower haircuts.
Tri-party repo is another story. In the crisis, those haircuts didn’t change. This is not proof that repo broadly (and haircuts in particular) didn’t aid transmission of systemic risk during the crisis. With all due respect to Mr. Comotto who uses this as evidence of little connection between rising haircuts and systemic risk, it is the opposite. Tri-party haircuts are negotiated in master schedules. It is a huge pain to change them. Rather than doing the work to update those schedules, cash lenders would rather turn tri-party repo off or demand only certain types of very high quality collateral from cash borrowers. This unintended consequence of inflexible design created static haircuts. Cash lenders went on strike, which then prompted a run on the tri-party repo system. If those haircuts weren’t so sticky, the market might have adjusted, like it did for bilateral trades.
What to do? Is a “to big to fail haircut” the answer? Regulators have talked about 20%. One size fits all is a bad idea. First, for some paper it is absurdly high. For other securities it might be low. And what was high one day, might be low the next. Think about a short-term corporate bond issued by a borrower than suddenly reports bad news. The bond, which had very small PV01 one day, now looks more like equity in free fall. No one cares about yield anymore. Furthermore, what is mandated as a minimum haircut inevitably ends up being institutionalized as a maximum.
We think back to MF Global and how LCH.Clearnet raised the haircuts on the sovereign Euro debt repos midstream, based on the increase in volatility of the underlying collateral and uncertainty around MFG. Someone once called this “bad boy margin”. You can argue if this is fair or not, but it did protect the CCP members and the market. If you advertise the fact that haircuts can rise, even in a term trade, repo participants will (hopefully) maintain adequate reserves to cover the potential for additional collateral. Make the rules absolutely transparent: if volatility goes up (either security or counterparty) then haircuts will go up. Even tri-party haircuts can be included in the process. Maybe it won’t protect from sudden cataclysmic risk – but short of living in a bunker, not much will. Not even 20% haircuts. Perhaps this is a reason to push all repos through CCPs?
February 22, 2017
October 21, 2015
May 10, 2016