Business startups are sparked by a creative solution to a problem, but only a rare few will become spectacularly successful, while many others flame out. It turns out terrorist organizations follow a similar pattern.
In a new study published in the Proceedings of the National Academy of Sciences, Brian Uzzi and his team at the Kellogg School of Business at Northwestern University present a method to predict which incipient terror organizations are likely to turn dangerously lethal by conceptualizing them as business startups. Business analysts have developed tools to evaluate startups and try to predict their future returns. Usually, analysts look at a company’s balance sheet, evaluating the startup’s capabilities and resources. Businesses with a lot of resources plan the deployment of those resources carefully. The converse is also true: Businesses that deploy their products in a haphazard manner typically have unstable resources.
Terrorist organizations don’t publicly share balance sheets, so it’s difficult to estimate their resources and capabilities. In fact, these organizations may intentionally send false signals about their access to resources. But it turns out that they deploy attacks much like businesses deploy new products. Uzzi’s team found that the more systematic ones are the most lethal and have access to more resources, whereas the ones that produce haphazard attacks tend to be the least lethal and have fewer resources. If a terrorist group had sophisticated technology or specialized skills (e.g., bomb-making, assassinations, hostage-taking, etc.), but lacked stable resources, the terrorist group still wound up being a relatively non-lethal group, Uzzi says.
“Our model is able to look at the first few attacks by a terror organization and predict with accuracy whether that terror organization is going to become a monster or if it’s going to be a relatively low-lethality organization,” Uzzi said. Using that model, he said, counterterrorism organizations could dedicate their resources to targeting the terrorist organizations that present the highest risk.
To build their model, researchers used data from the Global Terrorism Database, that includes information about terror attacks from thousands of organizations between 1970 and 2015. The researchers restricted their analysis to those groups that carried out at least 10 attacks, limiting the analysis to 342 groups. They then tested out their model’s predictions on another terrorism database with events from 2016 to 2019.
Although the forecasting ability of the model is novel and interesting, there are a few limitations of this approach, researchers say. The first is that the model considers a relatively small subset of the total number of terrorist groups.
“When you’re looking at groups that carry out at least 10 attacks, then you end up truncating the sample quite a bit,” says Kristian Gleditsch, a political scientist at the University of Essex in the UK who studies conflict and cooperation and was not involved in the study. The vast majority of terrorist groups carry out one or two attacks, and then we never really hear anything about them again, he says.
He adds that, although the database contains abundant information about attacks in Afghanistan and Iraq, the frequency of terrorist attacks in Europe has actually decreased since the 1970s; thus the model may apply to terrorist groups in Iraq and Afghanistan, but may not be applicable to other regions of the world.
Victor Asal, a terrorism expert at SUNY Albany, says that the model does not take into account factors that are known to influence the deadliness of terrorist groups, such as the organization’s size, religious ideology, and territorial control.
And there’s the luck factor too: It’s not clear that there is something inherent in the different capacities of terror groups that determines a group’s success in initial attacks, Gleditsch says. He and colleagues, in an earlier analysis, argued that some organizations are just lucky in launching successful initial attacks, and this allows them to establish an advantage going forward.
In spite of these limitations, the model could still be useful to counterterrorism analysts, Uzzi says. The model does not predict where attacks are likely to occur, or what kinds of targets a group is likely to choose.
“With so many things you have to watch, and with an inability to watch everything at the same time, this model can help them [counterterrorism analysts] pinpoint who they need to keep their eyes on,” Uzzi said.
Viviane Callier is a science writer based in San Antonio, Texas.