Systemic risk is a big topic these days for obvious reasons – the near collapse of the financial system has had profoundly bad effects on people around the world, and its continued instability is a major source of concern. Finding a way to measure and control threats to the stability of the entire system – and by extension, the rest of the world – is an important step toward making sure this doesn’t happen again.
Given that, it’s surprising that there remains no coherent, commonly understood definition of systemic risk, let alone a single way of measuring it. As Lars Peter Hansen notes in a recent NBER paper, this has implications for regulatory policy in the absence of meaningful analysis. Hansen focuses on the dangers of government enacting “discretionary” policies based on political urgency, though the opposite is more likely – if we treat the financial system as too complex to understand, then every problem simply becomes a “normal accident” that couldn’t possibly be mitigated by regulation.
Although Hansen has interesting things to say about modeling, what I find most useful in the paper is the way he groups together the various sources of systemic risk. Hansen uses three primary categories, which I paraphrase as:
Concentration: Some firms are either so large, or play such a central role in the financial system, that their failure can pose a risk to the entire system. The obvious example here is the too big to fail (TBTF) banks.
Interconnectedness: This clumsy word has become the term of art (see only the latest example here) for describing the density of linkages between financial institutions, as well as to the broader economy. Hansen is sharp in observing that the connections can have two damaging effects. First, they can transmit volatility across networks of institutions simply by virtue of their being connected. Worse still, the structure of the network itself can make the spread of risk worse by magnifying its effects, particularly in the case of the failure of an institution that is both densely interconnected and very large.
The big banks are an obvious example here as well, given the daisy chains of liabilities created by swaps, repo agreements and other transactions that link them.
Behavior: If enough investors become worried about solvency, the resulting sell-off can become “a modern-day counterpart to a bank run triggered by liquidity concerns,” as Hansen puts it. A less innocent example is what William Black calls the “epidemic of control fraud” in mortgage lending and securitization, which was so widespread that it arguably threatened the entire system.
In reality, none of these risks exists in isolation, and they are best thought of as the ingredients that recombine to create emergent systemic risks. A bank run, for example, only becomes systemic to the extent that it affects institutions that are linked to other institutions, or are sufficiently large to create problems, or both. Or consider ETFs and the Flash Crash. Trading in index ETFs affected index futures, individual stock prices and the ETFs themselves (interconnectedness), resulting in an electronic stampede (behavior).
This intermingling of sources of risk is one of the primary challenges facing regulators, who are operating (as Hansen notes) without sufficient data or analysis in the face of heavy resistance from industry and Congress. I think Hansen’s three-part framework can also help clarify how regulators are approaching this challenge, which is the subject of the next post.