Is there any such thing as macro-prudential risk?
I have been grappling with this question ever since reading The Fundamental Principles of Financial Regulation by Markus Brunnermeier, Andrew Crockett, Charles Goodhart, Avinash D. Persaud and Hyun Shin. All the authors are well respected economists; and Brunnermeier, Persaud and Shin, in particular, have been among those whose writings I have admired a lot during this crisis. Yet, I am not convinced about their concept of macro-prudential risk as opposed to the micro-prudential risk that traditional risk models are allegedly concerned with.
My problem is that at least since Markowitz, risk has always meant portfolio risk. The riskiness of any asset is the contribution that it makes to the portfolio in which it is held. For an equity portfolio, therefore the risk of a stock is its covariance with the portfolio which is the approximately the same as the beta if the portfolio is highly correlated with the market portfolio. In a value at risk (VaR) model for a credit portfolio, the marginal risk of a loan is equal to its expected loss conditional on the total portfolio loss being equal to the VaR level (see for example, Hans Rau-Bredow, 2002). This is true in particular of the Basel II models as well.
Brunnermeier et al are rightly worried about herding behaviour where many banks hold similar portfolios and are thus exposed to the same risks. But in this situation, each bank’s portfolio is highly correlated with the aggregate portfolio of the banking system. Thus Basel II and similar allegedly micro-prudential risk models are in fact capturing the macro-prudential risk of each loan. At this point, (following Brunnermeier et al) I am also ignoring the technical inadequacies of the Basel II risk models. The question being asked is whether conceptually they are addressing the right risk. I think they are.
One of the problems I had with Brunnermeier et al is that they seemed to be focused on the wrong conditional probability. They frequently ask the question: conditional on a bank failing what is the probability that there is a systemic crisis. They argue correctly that this probability is higher for a bank with a AAA portfolio than for a bank with a BBB portfolio. I think the correct question to ask is the reverse conditional probability: conditional on a systemic crisis, what is the probability that the banks in question fail. This probability is higher for the bank with a BBB portfolio and this is consistent with the Basel II risk weights.
At this point, it is also worthwhile to remember that the capital required for a AAA loan is far in excess of its unconditional probability of default. In fact, the unconditional probability of default is the “expected loss” which in Basel II is supposed to be covered by the credit spread and is not supposed to come out of the capital charge at all. The capital charge deals exclusively with the “unexpected loss” and is defined by conditionalizing on a systemic shock.
In short, I think today’s risk models are conceptually addressing macro-prudential risk because in any portfolio risk model these are the only risks that matter. Whether they are measuring these risks right is a totally different question (see my last blog post).
Brunnermeier et al have a long discussion about liquidity risk and maturity mismatches as a macro-prudential risk. The example of Northern Rock permeates this discussion. I think the diagnosis of Northern Rock as a liquidity risk which seemed to make sense in 2007 (I was guilty of this diagnosis myself) has been invalidated by developments in 2008. The silent run that banks like WaMu witnessed have shown that there is no safety for a bank in retail deposits. Nor is there safety in retail term deposits. All banks allow customers to prematurely encash their term deposits with modest penalties. In bank runs, people queue up to do exactly that as for example in the mini runs that we had on ICICI Bank in India. The behavioural maturity of core deposits is a meaningful notion in normal times; in periods of crisis, the behavioural maturity is zero. In wholesale markets, maturity is ill defined because of put and call features combined with step-up coupons. Normal maturity assumptions about these bonds have been invalidated during the current crisis. In short, maturity mismatch is not a well defined term during a systemic crisis.
In this context, I am perplexed by the irrational fear of bank runs among regulators and academics alike. We must not forget that frequent runs and near runs are the principal defence that we have against Ponzi schemes (both Madoff and Stanford were ultimately exposed by runs on them). Solvent institutions with normal access to central bank liquidity support can survive runs. Banks that cannot survive a run despite the central bank liquidity window are fundamentally flawed; they need to fail.
Posted at 9:40 pm IST on Thu, 5 Mar 2009 permanent link
Comments