The Crucible of Correlation and Coordination: On the Nature of Systemic Risk

Among the basic tools of risk management — avoidance, diversification, hedging, and insurance — risk transfer to specialized institutions that pool risks too treacherous for others would seem the likeliest vector of systemic risk.  But insurance poses no such threat.  Insurance reveals a paradox demanding a more systematic definition of systemic risk.

Correlation among certain losses, such as floods, is so strong that private insurers ordinarily refuse to underwrite these risks.  The consideration of correlation by insurers mirrors the treatment of R-squared and beta in portfolio management.  Indeed, beta is often treated as the simplest measure of “systemic” risk that cannot be managed by mere diversification.  The P branch of mathematical finance (portfolio theory) thus unites investment and insurance.

In response to private market failure, governments intervene by underwriting policies and/or serving as reinsurer of last resort.  Nevertheless, insurance may be one of the least worrisome branches of the financial industry.  Few if any insurers are too big to fail, or otherwise pose risks of such magnitude that their failure would doom insurance markets, the financial industry, or the broader economy.  By contrast, other financial industries labor under menacing amounts of systemic risk.

This presentation takes first steps toward a more systematic definition of systemic risk.  It will identify the basic mathematical tools for measuring systemic risk.  It will also recognize limitations on those tools.

First, I will distinguish between static and dynamic models of systemic risk.  Whereas tests gauging whether institutions are “too big to fail” rely on static measures of size and dominance such as the Herfindahl-Hirschman Index, more sophisticated tests predict whether destructive coordination makes institutions “too interconnected to fail.”

Second, I will identify behavioral constraints on the regulation of systemic risk.  Correlation is so counterintuitive that it deserves its own branch of behavioral finance: correlation neglect.  Given how small actors acting in coordinated fashion may increase fragility, diversification may perversely amplify rather than cushion financial shocks.  Regulatory arbitrage through “shadow banking” may inject a uniquely worrisome dose of systemic risk.

Finally, I note the underappreciated role of liquidity.  At sufficient scale, even a single institution’s unwinding of risky positions can result in an independent and calamitous form of systemic risk.  Liquidity risk therefore deserves a place of its own in theories of systemic risk.