Terrorism, Creative Uncertainty and Rational Prediction

I recently finished "Beyond Fear" by security expert Bruce Schneier. In it, he features a chart showing the number of annual deaths from various causes (e.g., different cancers, automobile accidents, heart disease, stroke, suicide, etc.) Alongside it, he lists casualty figures resulting from specific terrorist incidents over the past century or so.

At the top of the terrorism list, of course is 9-11, accounting for 3,029 deaths. It's followed by several airplane bombings in the 1980's that each cost ~250-300 lives (e.g., Pan Am over Lockerbie, Scotland). Schneier correctly points out that in the context of preventable deaths from other causes, even counting 9-11, terrorism has simply not been that big a deal to date. [Stay with me before rushing to comment...]

The direct and indirect costs of countermeasures the U.S. and other Western countries have taken in response (Schneier argues) are far out of proportion to actual damage and death in the past or the risk of further damage and death in the future. The second part of that statement is utterly flawed.

The  unpredictability of terrorism renders any backwards-looking, purely quantitative, actuarial mode of analysis inappropriate and ineffective. That is, future deaths due to terrorism are something that neither Schneier nor anyone else can possibly predict with any degree of confidence.

Until 2001, the biggest single terrorist incident had caused around 300 deaths. Then in the space of a few hours, that number went up by an order of magnitude. There was no consensus (or even a significant plurality) of expert opinion predicting that that would happen - much less when, where and how.

Yes there were a few isolated voices and even a paper-trail that seem amazingly prescient in hindsight. They don't count. Somebody, somewhere is almost always predicting the future with great precision. Discovering in advance which of those billions upon billions of conflicting and poorly substantiated forecasts is correct is the hard part!

The rub? Nobody can say that the peak casualty figure from a single terrorist incident won't go up again by another order of magnitude (or two, or three) tomorrow morning... or in ten years... or not at all (begging the recursive question of our ability to prevent such a jump by investing prudently in a wide range of broadly geopolitical to highly targeted countermeasures.)

As of the 10th of September, 2001 any rational calculation of the risk and cost of terrorism based on historical data would use numbers in the low 100's for possible deaths and develop countermeasures that were commensurate. After 9-11, the same rational, backwards-looking calculations might use 1000's. Both would be subject to so much uncertainty as to be laughable - because they rely on the same flawed approach to assessing the magnitude of such a risk. (A similar if largely forgotten leap in awareness of the risk occurred in the early '70s)

Terrorism is fundamentally not a forecastable thing. That's especially true during a period of innovation and expansion in that sad, sick "industry". The fact that the peak death number changed so suddenly makes a conservative rational calculation based on past history just as tenuous as any radical emotional guess based on fear. Given that the trend is clearly up however, and that the last jump was by 10X, it is only prudent that we err to the side of assuming high and being wrong than assuming low and waving bye-bye to New York or Los Angeles.

The larger point? What's tough about predicting terrorism is also tough about predicting discontinuous change in business. Applying the same forecasting methodology to discontinuous possible 'left-field' problems as to well-understood, clearly bounded problems with deep actuarial data-sets is like trying to eat soup with a knife. A tool that's extremely precise and powerful for some jobs is utterly misguided for that one.

As we noted last month with regards to "creative uncertainty":

...what was the 'probability' circa 1970 that snowboards (and the as-yet nonexistent entrepreneurial providers who would invent them) would cut into the ski industry? ...Or the 'probability' circa 1980 that digital photography would into silver halide (wet process) film sales?

It is only with a broadly creative process (e.g., scenarios) closely wedded to a structured forecasting methodology (e.g., prediction markets, delphi, etc.) that more reasonable - if still subjective - assessments can be made. In the realm of creative uncertainty, it is the anticipation of entire classes of unconventional threats and opportunities and their rough magnitude, not detailed numerical forecasts that lend strategic power to organizations investing for the long term.