01/11/2012 02:43 pm ET Updated Mar 12, 2012

The Causes of Systemic Disasters

I'm a bit late on this, but Karl Smith kicked off an interesting discussion with large, if elusive, import over the weekend at his Modeled Behavior blog. Smith tackles the subject of causality in epidemics -- and by that he means not only conventional disease epidemics, like the flu or SARS, but also a variety of systemic disasters from economic (and presumably market) breakdowns to "social" epidemics, by which I take he means examples of mass migrations or genocides, or of mass behavioral trends, like crime waves or obesity or drug addiction. (Smith's post may have been provoked by Mother Jones' Kevin Drum arguing for a correlation between crime and the use of lead paint.) Once you open it up so widely, you do get bogged down into definitional problems; the metaphorical tendency comes powerfully to the fore. If it's a proliferating mass phenomenon of any kind, it's an epidemic. One of Smith's commenters, for example, refers to cancer as an epidemic of a single organism, which is a fascinating notion but one that cries out for more rigor. Smith gestures toward a definition of epidemics that focuses on its transmissibility, via communication or transportation lines, though in some cases, the term comes awfully close to existing in the eye of the beholder. Smith's "epidemics" in fact veer closely to the extensive work Yale sociology professor and systems expert Charles Perrow has done on what he calls "normal accidents" and disasters.

Smith tosses up two cases of "epidemics" -- the purported ties between lead paint and crime, and the medical theory that most peptic ulcers are "caused" by a microbe, H. pylori, contrary to previous multicauses notions -- and declares: "I don't know the research behind lead and crime. However, overwhelmingly the presumption should be that epidemics have a single precipitating cause."

And so the debate is off and running. Why should we care? Well, there's the prophylactic question: How do you contain systemic accidents or disasters? This is Perrow's subject, and many of his conclusions speak directly to containment, whether it's a hurricane like Katrina or a nuclear plant meltdown like Fukushima or a market breakdown. Obviously, if you believe in the one-factor theory, containment is a lot simpler. (Perrow, by the way, doesn't seem to.) Second, there's a deeper matter that verges on the political or philosophical. Certainly when you move from the biological to the social and economic you edge into a realm where determinism, particularly driven by single factors, emerges. Is autism caused by vaccines? Perhaps lead paint does have a major effect on crime. But are we really driven, as the popular theory has it, by wayward genes or single environmental factors? Do we have a single gene for obesity, or IQ, or larceny? Where does free agency fit in? Is history really that simple? Did the Roman Empire fall because of lead in its pottery? Or, like nature versus nurture, perhaps the answer is that some epidemics may have one cause, and that others are quite complex.

The gist of Smith's single-factor declaration seems to rest on probabilities. "People are sometimes confused by the fact that complex conditions have a long list of necessary factors," he writes. "However, the odds against more than one necessary factor pushing the phenomenon across the line into epidemic at the exact same time is astronomical." This seems to me far too rigid a way to look at this. Epidemics, if that's a useful word at all (I'd prefer a systemic accident, catastrophe or disaster following Perrow), may differ in the number of necessary factors. Most systemic disasters require a crisis and a pathway of transmission, both of which may have to develop over time; each of those can have component parts. And as one of Smith's commenters suggests, some of these necessary factors may at times provide buffers and at other times generate strong correlations. Thus in the subprime disaster, you clearly had underlying economic factors (lagging incomes, low interest rates) that were correlated and that may well have been linked to larger policy factors like deregulation and perhaps even technological innovation, which drove both Wall Street and globalization. That's why it's so difficult to identify a small number of factors -- real estate alone, say, or Fannie Mae -- or culprits: It was a complex set of probabilities that converged. (The same commenter argues that Smith needs to consult "random matrix theory," which, simply put, I think, suggests that small changes can sometimes lead to big results.)

Another Modeled Behavior commenter takes us back to Smith's language and how we describe these phenomena. He argues that when Smith talks of "precipitating factors," there can, "almost by definition, only be one." He continues: "But to me that seems people usually care about the causative factors: what happened to cause this thing, so we can prevent it next time? In that case, it's a mistake to single out one event when there are many equal contributors. Not just in cancer, but in recessions and crime too."

That makes a great amount of sense. History tells us that the further one studies a major historical event -- World War II, the Great Depression, the fall of the Roman Empire -- the more multicausal these events become; the "event" is less a single cause than a systemic event. (Journalism, it is true, has a weakness for single-cause explanations.) In fact, "precipitating events," whether it's the Crash of '29 or Pearl Harbor, may well be spectacular, but their importance decreases with study. They are, in a sense, nearly random: Another event might well have kicked off a disaster -- the same or one quite different -- because the conditions, the environment, required very little to nudge it over into systemic disaster or epidemic. The same seems to apply to many cancers, which are extremely heterogeneous and difficult to categorize. A few cancers, usually rare or childhood, stem from single genetic changes. But most cancers seem to require multiple changes. The organism or cells have to reach a tipping point (to recklessly use another popular metaphor) where they proliferate out of control. But the "precipitating factor" is not the cause.

What gives this discussion currency, of course, is our own recent bout with financial disaster. The popular response to 2008 and what has followed is to focus on the single cause -- subprime, deregulation, globalization, greed, compensation -- and fail to separate out precipitating factors from less satisfying, more complex and ambiguous causes. But subprime alone was not enough to nearly bring down the world financial system. Much more was going on. Perhaps lead paint does play a major role in crime; perhaps that single bacteria is the major "cause" of peptic ulcers. But I would argue, even if that was the case, that single-factor causes of systemic disasters are as rare as single defective genes as a cause of cancer.

Robert Teitelman is editor in chief of The Deal magazine.