Power laws -- a mathematical distribution, usually depicted as a curve, that depicts a low number of large extreme events (on the left) tailing off to a large number of mild events (on the right) -- are good at describing lots of things found in nature and society, from the distribution of cities by population to earthquakes by severity. Over 50 years ago, power laws were found to describe the severity of casualties in conventional warfare (Richardson). We've also found that changes to the slope of the power law can depict the shift from conventional warfare to open source warfare (or 4GW) over the past decades (see relatively recent analysis by Johnson, Spagat, et. al.. that shows casualty distributions for modern conflict are converging on the model for non-G7 terrorism).
Another recent effort by Clauset and Young entitled, "Scale Invariance in Global Terrorism," studied terrorism data over a thirty-seven year span and concluded that power laws apply to this data set too. Essentially, that means that despite a marked increase in the frequency in terrorist attacks (this study was focused on "blood and guts" terrorism exclusively), the ratio of large to small attacks, and improvements in weaponry is still the same (the power law curve's exponential is roughly the same). It also shows that 9/11 and other extreme terrorists attacks are merely an expected plot on that curve, and not outliers (although, it was still a black swan). Some other interesting results include:
- There's been a long running and ongoing decrease in the mean time between attacks, with a big fall in 1998. See the blue line in the graph inset above. The sheer volume of terrorist attacks globally is growing at a fairly fast pace.
- While the data in aggregate is able to account for different weapons, there is an indication that the potential for extreme events is much higher with bioweapons (currently a shallow data set). Its exponent is 1.8 vs. the 2.3 average for other techniques (except fire). That means that events are less frequent, but when they occur they are more damaging. Note, there isn't any curve for nuclear terrorism likely due to the fact that these weapons aren't something small groups can readily access (despite hype to the contrary).
- As expected, the exponents for terrorism in developed (1.71, a steep curve) and developing nations (2.5, a shallow curve) are different. This means that while there are fewer attacks in developed countries, they tend to be more violent.
I think that power laws we see in blood and guts terrorism will apply to attacks that focus on systems disruption too. If so, this is worrisome. Here's why:
- The data presented shows that fire (arson) is the closest to using systems disruption techniques for "blood and guts" terrorism and it has the lowest exponent at 1.74. This implies that network effects can radically and potentially irrevocably steepen power law curves such that even badly planned/executed attacks that would have yielded little in "blood and guts" results, can yield exception outcomes via the high ROIs (returns on investment) of system disruption.
- Steeper curves for system disruption might stay steep even if frequency increases. Given the ease at which systems disruption can be accomplished (particularly relative to killing people, measured from complexity of method to level of motivation), we have the potential to see a great many of these attacks. Unfortunately, at the low implied exponential, any increase in systems attack frequency would generate a disproportionate increase in the number and scale of severe results as compared to terrorism over all.
- Finally, bio-terror that focuses on communicable disease (as opposed to biotoxins) would likely be able to access the network effects of systems disruption. If that method become prevalent, the current exponent we see for bioterrorism, already low at 1.8, could steepen even more despite increases in frequency. That means lots of big casualty attacks relative to the total number of attacks. FINAL THOUGHT: the propagation of resilient communities can radically change this assumption, by allowing the scale invariant resilience necessary to break the network-effects of these attacks.