In Geopolitical Risk Management, a Chance for AI, Alternative Data and Predictive Analytics to Shine
Computer science yields "more data driven, systematic and methodologically rigorous" approaches, one CEO says
Friday, July 19, 2019
By Katherine Heires
In common with other categories of risk, those arising from geopolitical events - trade and tariff disputes, regional military conflicts and terrorism, domestic political uprisings and instability - are increasing in frequency and complexity and challenging conventional risk management assumptions and approaches.
“A key error on the part of any global business would be to ignore and fail to regularly monitor geopolitical risks,” including their impact on operations, employee safety, asset values, currencies, interest rates, regulatory compliance and more, says Marvin Zonis, professor emeritus at the University of Chicago Booth School of Business and co-author of Risk Rules: How Local Politics Threaten the Global Economy.
Just as there are more geopolitical events than ever to track, monitor and analyze, advanced technological tools employing algorithms and artificial intelligence are making the task more manageable for chief risk officers, political risk specialists, board risk committees and others seeking to raise their level of understanding and analysis. These tools are seen as augmenting traditional inputs into geopolitical risk management that are heavily dependent on human labor, personal contacts and consulting teams.
“Traditionally, the management of political risk involves the hiring of subject matter experts who are former secretaries of state, secretaries of defense and ambassadors who provide their considered advice to CEOs, C-suite executives, and board members,” says Gary Shiffman, adjunct professor at the Georgetown University Center for Security Studies and founder and CEO of Arlington, Virginia, software company Giant Oak.
Demand remains healthy for political risk advisory services, such as those of APCO Worldwide, Control Risks Group, Eurasia Group, Oxford Analytica and Teneo, Shiffman points out.
However, with the rise of new technologies - including Giant Oak's behavioral-analytics-based software - alternative methods have emerged for monitoring the risks of an increasingly globalized and interconnected world.
“Phenomenally High” Risk Level
A measure of the challenge is the University of Cambridge Judge Business School's Cambridge Global Risk Index, which for 2019 sees “sustained levels of high risk from geopolitical events and financial crises.” It ranks market crashes and interstate conflicts as the top threats in terms of “GDP@Risk.” Floods, cyber attacks and commodity price shocks also rank relatively high in the Cambridge Centre for Risk Studies' “comprehensive threat analysis for 279 cities that represent 41% of global GDP.”
The index highlights the Iran-Syria conflict, territorial disputes in the East and South China Seas, and the tripling in number of global Islamic terrorist groups in the past decade.
“The level of geopolitical risk is phenomenally high for businesses today,” Zonis says. “The risks were far more constrained in the past, even during the Cold War, but now they are everywhere - in China, Russia, Iran, but also in the Middle East and South America, in particular Venezuela.”
To Shiffman, “It's a very volatile environment right now, but it's important to keep in mind that displacements in the geopolitical environment, while creating disadvantages for some, can create clear advantages for others.” Trade tensions between the U.S. and China could benefit commerce between the U.S. and Vietnam, for example.
“Businesses need to keep in mind that this is not a zero-sum game,” Shiffman says
Introducing Computer Science
Giant Oak, Geoquant and Predata are examples of innovative companies helping their clients navigate changes and uncertainties in the geopolitical landscape. They differ in the ways that they employ technologies and integrate human intelligence, but all aim to deliver relevant and impactful risk alerts in a timely fashion while separating signal from noise.
Geoquant co-founder and CEO Mark Rosenberg says political risk was long considered too idiosyncratic to define, quantify and measure systematically. Geoquant was launched in San Francisco in 2016 to change all that.
A political economist and University of California at Berkeley PhD who formerly worked at Eurasia Group, Rosenberg says that Geoquant “fused political science with computer science” to collect data and then benchmark measures of geopolitical and political risk for risk management and investment purposes.
“We have introduced a new element into the business of managing geopolitical risk, making it more data driven, systematic and methodologically rigorous,” the CEO explains.
According to a Geoquant white paper, the firm employs a proprietary, deep factor model incorporating 22 fundamentals of politics, such as governance practices, policy actions and security factors, derived from political science and economic literature. It is also the basis for more customized risk models targeting asset classes, geographies and other topics
The measure of each political fundamental is based on an assessment of both structured and unstructured data, producing two risk scores that are then folded into the model
The structured data is derived from 250 traditional sources, such as country-level databases maintained by multilateral institutions, NGOs, governments and social scientists. The unstructured data comes from more up-to-date, unstructured text on social media and traditional news sources, generated by customized machine learning and natural language software.
Human Element Is Critical
All of this data is then integrated, with the help of a Tel Aviv-based engineering team, and using a proprietary algorithm. It is then reviewed by Geoquant's in-house and freelance political scientists and regional experts, to produce final risk scores
“The human element is very much a part of the daily scores and is critical to the quality and accuracy of the product,” says Rosenberg.
He adds, “There is a lot of hype around artificial intelligence and machine learning. It is a very useful tool in automating large chunks of an analytical process, but it does not stand alone and in political science, and I'm not sure it ever will.”
Rosenberg is quick to point out that Geoquant's risk scores can't predict the future, but the firm's data generation system combined with political scientists' assessments can be highly effective. The firm's prediction rate, which he says achieves 76% accuracy, “is far greater than the existing industry rate,” providing an early warning system, or what Rosenberg calls “a seismic monitor for global political risk.”
Discerning Patterns and Volatility
Predata, as the New York- and Washington, D.C.-based venture's name implies, is an alternative data provider with an emphasis on predictive analytics.
“It's a complex and dangerous world we live in,” says Hazem Dawani, a former electronic trading entrepreneur who has been Predata's CEO since 2018. “Whether you are a Fortune 500 company, large energy concern, or manager of a large supply chain, you have to manage risk globally and with timely tools.”
Founded in 2015, Predata employs a patented, machine learning method to analyze Internet metadata (data about the data) for insight into group behavior - what people are focused on, researching, and repeatedly viewing. From these insights, risk signals are derived that anticipate political and economic developments.
This was and continues to be a major interest of Predata founder and chairman Jim Shinn, a PhD and lecturer in global affairs at Princeton University who previously served in the State Department, the CIA and as assistant Secretary of Defense.
“We don't employ natural language processing or key word analysis,” Dawani explains. “We purely look at sources curated by our expert analysts on serious topics and employ machine learning technologies to identify patterns of behavior.”
The software performs in any language - it is language-agnostic; can customize queries, posing a broad range of questions; and covers 180 countries.
Predata's behavioral focus allows the software to surface patterns that reflect social unrest, changes in trade relations, concerns about market bubbles or collapses, and the likelihood of shifts in government regulations or policy decisions. “A lot of policy decisions are made after a great deal of research is conducted online,” Dawani notes.
Also unique about Predata, he continues, is that “our signals are not biased by any analyst opinion. They are purely quantitative, and unlike other firms, our system covers a broad range of topics,” including a Political Volatility Index (PVIX)
According to Gary Shiffman, Giant Oak, founded in 2012, represents the shift from broad, human-driven analysis to a more precise form of measurement.
Initially developed to help national security and law enforcement officials identify such activities as terrorism, money laundering and human trafficking, Giant Oak's proprietary search platform - known as GOST - now also helps companies monitor the activities of political actors, customers and business partners. Giant Oak combines behavioral science with artificial intelligence and machine learning algorithms that constantly refine their searches, producing better results over time
In contrast to others that focus on social media or news accounts, “we focus on human activity” on both the open Web - what standard search engines index - as well as the Deep Web, which is not as easily accessible. “You can look at political or industry leaders and ask questions about their patterns of behavior and risk levels,” says Shiffman, a former managing director of Chertoff Group, senior vice president and general manager at systems integrator L-3, and chief of staff of U.S. Customs and Border Protection.
“What is unique about our approach is that it is based on the belief that humans make decisions - not states or governments,” he says. “If I want to understand what is happening somewhere in the world, I look at publicly available patterns of behavior, and that is the signal that is going to give indications of what is happening or could potentially happen.”
He emphasizes, however, that GOST is not a rating system for geopolitical risk, but rather can flag a significant change in behavior.
“GOST is not saying that someone is good or bad, but what it can do is identify a baseline of behavior and then send you an alert when something changes or when there is a dimension of activity worth noting,” Shiffman says.
Given the power of computers and machine learning, GOST works best at scale. “It's always better when you ask it to study the behavior of 100 people over time rather than one person; that's when you get a real return on your investment,” he says.
Processes and People
As sophisticated as these machine-intelligence advances may be, they still operate within existing frameworks that are designed to systematically identify and define geopolitical risks and prepare organizations to respond to them.
The book Political Risk: How Businesses and Organizations Can Anticipate Global Insecurity, by former Secretary of State Condoleezza Rice and Amy B. Zegart of Stanford University, advises that management of geopolitical risks is more challenging than ever as “political actions that impact businesses are happening everywhere,” and they can be set off by anyone with a cell phone, Twitter or Facebook account.
What's more, “Geopolitical issues that used to be the domain of governments alone are now very much corporate concerns as well,” says Frank J. Cilluffo, a longtime cybersecurity adviser and Department of Homeland Security Advisory Council member who is director of Auburn University's McCrary Institute for Cyber and Critical Infrastructure Security. “As a result, the need to integrate that factor into corporate planning and risk management efforts is crucial while companies also need individuals on board who can regularly aggregate and assess the various geopolitical risks and threats.”
Katherine Heires is a freelance business and technology journalist and founder of MediaKat llc.