Four years after the onset of the financial crisis, regulators remain finely attuned to the dangers of systemic risk in the financial sector. Yet financial institutions continue to operate in regulatory silos where rules are focused on and applied to individual firms rather than the sector as a whole. That’s of concern, Professor Ciamac Moallemi says, because systemic risk can’t be effectively measured or managed at the individual level.
“The many individual units that comprise a system — in the case of finance, the units are banks and the system is the financial sector — are, to varying degrees, interdependent,” Moallemi says. In today’s economy, the largest financial institutions are likely to have similar investments across a wide swath of sectors, so significant failure in one may be linked to the health of another.
Yet the central feature of financial sector regulation — and virtually all proposed systemic risk measures — addresses only individual firms. Typically, regulators seek to limit risk by placing limits on how much a firm can lose, with no provisions for whether other firms hold similar positions or how closely other firms might be connected to the same sectors.
Managing systemic risk does, in part, involve making sure firms don’t lose too much money. But an effective measure of systemic risk must look at risk in terms of joint outcomes for all firms and how those losses are split up. “Regulators should not focus so much on individual firms as on the cross-sectional profile of gains and losses,” Moallemi says.
This is more complex than simply noting outcomes for each firm in the economy and summing them all up. “It’s not that some firms make money, some lose money, and if it all nets out all is well. We argue that from the perspective of a systemic regulator you can't allow gains to net with losses because you can’t take money from firms that have made money to subsidize those going bankrupt,” Moallemi says. “Similarly, if you compare two scenarios, one where a large number of firms each lose a small amount of money and another where a single firm loses that same amount of money, it is clear that the distribution of losses matters. A regulator must express a preference for one over the other.”
Moallemi worked with Garud Iyengar and graduate student Chen Chen of Columbia University’s Fu Foundation School of Engineering and Applied Science to create a framework for defining systemic risk as a function of the joint collection of possible outcomes in a system. The framework allows regulators to define cross-sectional profiles of acceptable limits of systemic risk. For instance, rather than only monitoring individual firms’ overnight losses in isolation from others, regulators could monitor the collective overnight losses from a designated set of firms and assess how much loss each firm contributed. The framework also allows regulators to identify how much risk each firm contributes to overall risk in the system.
The ability to attribute risk to specific firms could also be used as a basis for a taxation mechanism. “If banks take positions that are highly correlated and create a systemic risk to society, we might consider taxing those banks a small amount to take into account the risk they are creating and to offset potential damage,” Moallemi says.
This approach to systemic risk is not limited to the financial sector: most industries and firms can be broken down along system lines, where a collection of agents is overseen by a system manager concerned with the overall outcome generated by the system. A telecom carrier could use the framework to assess how disruptions in individual links of a network impact the overall service, and attribute the risk of the overall network going down to specific links. Similarly, a utility regulator might use the framework to gauge the contribution to the overall risk of an electrical grid of any given component failing, and thus assess the robustness of utility companies’ investments in upgrading electrical grid infrastructure.
“This isn’t just about measuring money or purely financial outcomes,” Moallemi says. “It works for any system in which we are interested in the risk of the collective behavior of a set of agents.”
Ciamac Moallemi is associate professor of decision, risk, and operations at Columbia Business School.