Jorion.01.07.
“The fact that the total of derivatives notional amounts is much greater than cash markets should be a major reason for concern. This could cause systemic risk.”
Discuss this argument. Are notional amounts the appropriate measure of risk?
The conventional answer is, no, notional amounts are not a good measures of systemic risk. However, I think this question is more debatable than it may seem. Please note we need to distinguish between asset (or instrument) risk and systemic risk.
In particular, an instrument like a credit default swap (CDS) may be net riskless: in a zero-sum game, it transfers default risk (and maybe deterioration risk) from the protection seller to the buyer. In terms of fundamental factors (as opposed to technical factors), we can say the CDS represents pure credit risk transfer. However, too many have engaged in a fallacy of composition to assume that the CDS market therefore creates no systemic risk. The system (the market) may introduce “emergent” risks. The issue of whether the CDS market itself creates or reduces systematic risk is a question about a complex system with a high degree of difficulty and a non-trivial answer (where the first challenge is: how exactly do we even define the question; e.g., does abuse of CDS count? does improper modelling? inadequate collateral? in particular, how do we parse people, as they use the instruments, from the instruments per se); we cannot infer the market necessarily creates no new risk merely because a single instrument creates no new risk.
But to Jorion’s question, the traditional answer is along the lines of:
* Notional is clearly an inappropriate measure of risk for an instrument. Jorion says, “For the same notional amount, some bonds have extreme risks and other no risk. The risk manager needs to know how the instruments respond to risk factors, as well as the range of potential movements in risk factors.” Please note how this really leads to risk mapping.
* At the systemic (market) level, the conventional response is also that notionals are not systemic risk measures. For several reason, the amount of actual capital at risk is a small fraction of the gross notional. First, most derivatives to not require counterparties to exchange the notional; e.g., an interest rate swap. Second, counterparties typically net their positive and exposures (“bilateral contract netting provisions”) which substantially reduces exposure. The net counterparty risk is often less than 5% of the notional! To give another example, when Lehman failed, according to the DTCC the notional referenced by CDS was $72 billion but the actual cash exchanged was only about $5 billion.
However, notional volumes are a proxy for market activity. In general, higher notional implies more trades, more counterparties and/or larger reference amounts (by definition). It seems conceivable that notional volume could have at least some bearing on systemic risk, if for no other reason than higher notionals imply larger markets vis-a-vis the system. To illustrate, AIGs exposure as a credit protection seller was some function of the notional on written CDS contracts; the size of that notional likely correlated with AIG’s prominence vis-a-vis the market. If the notional CDS were tiny, then counterparty dependence would be small, and it’s unclear how AIG could have been “too big too fail.” In this respect, it seems plausible to make an argument that notional volume is at least a proxy for one variable that bears on systemic risk.