Economic Crises Are Invariably Failures of the Imagination

A fundamental issue in all risk management is oversight vs. seeing.  You can be doing plenty of oversight, analysis, regulation and compliance, with much diligence and having checkers check on checkers, but is the whole process able to envision the deep and surprising risks that are the true fault lines under your feet?  Or are you only analyzing, regulating, writing up and color coding dozens of factors which while important, are not the big risk which is going to take you and perhaps the system of which you are a part, down?  For example, in the midst of your risk management oversight efforts, whether as a company or as the government, could you see in 2005 or 2006, or at the latest by 2007, that U.S. average house prices across the whole country, were likely to drop like a stone?  And see what would happen then?

Most people, including the most intelligent, experienced and informed, could not.

Former Treasury Secretary Timothy Geithner, in his memoir of the 2007-2009 financial crisis, Stress Test (2014), draws this essential conclusion: “Our crisis, after all, was largely a failure of imagination.  Every crisis is.”  If you can’t imagine it, if you consider that the deep risk event is unimaginable or impossible, your oversight will not see the risk.  “For all our concern about ‘froth’,” Geithner continues, “we didn’t foresee how a nationwide decline in home prices could induce panic in the financial system.”  

This is a profoundly important insight.  Geithner expands on it: “Our failures of foresight were primarily failures of imagination, like the failure to foresee terrorists flying planes into buildings before September 11.  But severe financial crises had happened for centuries in multiple countries, in many shapes and forms, always with pretty bad outcomes.  For all my talk about tail risk, negative extremes, and stress scenarios, our visions of darkness still weren’t dark enough.” 

That was not for lack of effort, but for lack of seeing.  “The actual main failing was over-reliance on formal econometric models,” banking scholar Charles Goodhart suggests in his acute essay, “Central bank evolution: lessons learnt from the sub-prime crisis” (2016).  He points out that as the housing bubble was inflating, there was copious housing finance data which could be and was analyzed:

“There were excellent monthly data on virtually all aspects of mortgage finance in the USA starting from the 1950s.  By the 2000s such data provided over 50 years of all aspects of US mortgage finance.  During this period, there had only been a very few months in which the value of houses, and the mortgages related to them, of a regionally diversified portfolio of housing assets over the US as a whole had faced a loss, and then only a very small one.

“While there had been sharp declines in housing valuations in certain specific regions, i.e. the North East in 1991-2, the oil producing states in the mid-1980s, etc., a regionally diversified portfolio virtually never showed a loss, and then only a minor one, over these 50 years.”  The conclusion seemed clear enough at the time: house prices did not, so would not, fall on a national average basis.  A portfolio of mortgages diversified across regions would be protected.  “Virtually everyone was sucked into the general conventional wisdom that housing prices”—on average—“were almost sure to continue trending generally upwards.”

This clear, though in retrospect completely wrong, conclusion could be professionally quantified: “Put those data into a regression analysis, and then what you will get out is an estimate that any loss of value in a regionally diversified portfolio of greater than about three or four percent would be…highly improbable.”  But as the bubble got maximally inflated, its shriveling became highly probable instead of improbable.  As we know, average U.S, house prices went down by 27% and fell not for a few months, but for six years, in spite of all kinds of government interventions.  The housing market went down for longer than a great many people could stay solvent.

“Of course,” Goodhart reminds us, “econometric regressions are based on the implicit assumption that the future will be like the past.” The less of the past we know, the worse an assumption this is. In this case, fifty years and one country, even a very big country, were not enough. 

To expand how much of the past we have studied, both in terms of more time and more places, is perhaps one way to improve our ability to see risks, imagine otherwise unimaginable outcomes, and thus improve our risk oversight.  Perhaps.  There are no guarantees of success.