RAND Tries to Model Risks of Terrorist Attacks
Jack Riley, director of RAND's Public Safety and Justice section, explained their new computer models of possible terrorist attacks to a UCLA honors class April 23.
Published: Friday, April 25, 2003Can we predict how likely a terrorist attack on New York's Grand Central Station might be -- compared to a possible attack on the Washington Monument? The RAND Corporation, the venerable Santa Monica, California, think tank, believes it can, with at least some assurance that this is not a pure guess. Jack Riley, director of RAND's Public Safety and Justice section, explained RAND's new mathematical computer model of potential terrorist actions to the April 23 session of Honors Collegium 155 in UCLA's Dodd Hall. This was the fourth installment of an unusual 10-part seminar for undergraduates which is also open to the general public. The class is sponsored by the Ronald W. Burkle Center for International Relations.
RAND has formed a partnership for this study with a San Francisco company called Risk Management Solutions (RMS). RMS is in the business of advising insurance companies on the likely scale of risks from natural disasters such as hurricanes and earthquakes to help the insurance companies know how to price their policies. RMS staffer Gordon Woo, a London-based computer programmer and game theorist, has devised a modeling program that attempts to calculate the potential values of property in a given area correlated to the potential scale of damage for such natural disasters and how likely very high damage events may be.
While people do not regard deliberate destruction in the same way they look on the impersonal destructive forces of nature, the property values are the same in both cases. RAND has been working with RMS to adapt the modeling program to cover terrorist attacks, using a database of some 3,339 terrorist attacks that involved U.S. citizens or property between 1968 and 1998 to provide a rough graph of frequency and scale of damage -- modified by some hypothetical assumptions in the aftermath of the September 11, 2001, attacks in New York and Washington, which were the largest to date.
"This model," Riley told the audience, "really fills a gap in terms of the policy making process and in terms of the issues that are important."
The gap is not that the RAND-RMS model can tell you if you will be blown up or not, but its goal is to help the U.S. government decide how much it is reasonable to spend on deterrence of potential attacks at particular locations in the United States, and to give insurance companies, which paid out some $50 billion after September 11, something like actuarial statistics on how much risk they are taking in selling anti-terrorist insurance in Akron, Ohio (not much), compared to Chicago, Illinois (pretty high) -- and don't even think about such insurance in New York City.
"Terrorism, unlike natural disasters, often operates on a microscale," Riley said. He gave as an example of a natural disaster the Kobe earthquake of January 1995. In Kobe, "significant loss to buildings, damage such that the buildings couldn't be reoccupied" stretched along the seacoast for some 30 kilometers. In contrast, the Twin Towers event was less than 2 kilometers. But "the insured losses from the World Trade Center attack were roughly three times the losses from the Kobe earthquake."
The amount of damage from terrorist attacks has a wider spread than that of natural disasters. While most hurricanes do more damage than most terrorist attacks, hurricanes peak in ferocity at about $200 billion in damage, and only one ten thousand does that much. In contrast, most terrorist attacks do very little damage, but a very small number actually or potentially can exceed the damage of any hurricane, with potential losses of as much as $1 trillion.
An Aid to Government Preparedness Planning
"Imagine that you are Secretary of Homeland Defense Tom Ridge," Jack Riley said, "and that you are responsible for developing formulas and allocating the tens of billions of dollars to municipalities and governments and organizations all across the government to help them prepare against terrorism. What's the right formula? How do you do it? I would submit that you need some kind of modeling framework that helps you assess the tradeoff between, say, sending another dollar of preparedness money to New York City versus sending that dollar to Cincinnati or even here to Los Angeles."
In the model RAND and RMS are working with, the risk of an attack either for a particular city or for a particular building or location within a city is ranked by a combination of the place's economic value, its symbolic importance to the United States, how easy it is to attack, if attacked how likely it is that it could be destroyed, and the extent to which the target would be familiar to an audience in the terrorist's home country who would be impressed by the feat.
Costs and Risks to the Terrorists
The RAND model includes trying to predict the likelihood that the terrorists would use extremely destructive weapons. Their conclusion is that conventional explosives and firearems are far more likely. "It is often lost on people but there are costs to the terrorists of attempting to field new technologies and use new weapons in their attacks," Riley said. "For example, if they want to use biological agents against us there is a significant capital investment in seeking to acquire those weapons, there is the cost of training their people, including the potential cost of losing some of their best scientists to accidents and unintended things while they are trying to figure out how to successfully use the weapon.
"And then there is the problem that if they aren't successful they have probably called attention to themselves and they may really limit their ability to do that kind of thing in the future."
Where Will the Terrorists Strike Next?
The RAND-RMS computer model tries to rank cities by the how likely they will become targets for a terrorist attack (see graph below). New York is overwhelmingly the most probable spot, according to their calculations, with Chicago a distant second and Los Angeles only number six.
"Once you get beyond six cities in terms of relative risk," Riley said of his model, "it is very hard to differentiate beneath those six cities where the risk is really all that much different from any place else." He added that his model would probably have ranked Washington, DC, as more at risk than the graph shows but that the modeling program currently rates economic damage more highly than damage to symbolic sites.
"The bottom line in thinking about this kind of thing in a structured modeling framework," he said, "is that there is a certain distribution across attack modes that exists and we know something about that, and there is a certain distribution of targets that is important to the people who are doing the targeting selection. And the result is that you can generate a relative likelihood for each combination of those two things. This is a relative risk map for Manhattan" (see illustration at the top of this article).
"The areas that are in darkest red are those combination of targets and attack modes that result in the highest simulated losses across insurance lines. The upper section are the hotels around Grand Central Station, the lower red area is the business district downtown. There is a lot of green space in there, and those are areas where the risk is relatively low."
Jack Riley suggested that this kind of modeling could prove useful to government agencies and private companies in weighing how much to spend on security and what locations to prioritize. He noted, for example, that "radiological, chemical, and biological attack, at least at this point, are the relatively low probability events. The conventional attacks are the relatively high probability events. We are writing large checks as taxpayers to first responders who are buying all sorts of equipment -- all-hazard hazmat suits, sophisticated detection equipment for chemical, biological, and other attacks, but it is not clear that that kind of decision making is being done in a structured framework where first responder training and equiping strategies are really being mapped to the kinds of risks that are out there."