Author: Peter Marber, March 2021

The changing quality and growing quantity of political risk is directly related to three distinct yet interconnected forces occurring in the last 50 years: the rise of globalization, the explosion of information, and the growing importance of, to use George Soros’s term, “reflexivity.”

Image by Unsplash

As a young trader for a large Swiss bank in early 1990, I built a small position in a heavily discounted US-dollar bond of the National Bank of Yugoslavia at roughly 25% of face value. With a coupon of 5%,  I smugly enjoyed running yields of 20% for much of that year. The smugness didn’t last long. Within twelve months, civil war broke out and there was no longer a functioning Yugoslavia, let alone a functioning central bank. I was left with my bond claims but had no idea which person or what entity to contact about them.

This kind of “political” risk—the risk of investment losses due to civil war, nationalization, or expropriation—is typical of what crossed portfolio managers’ minds back then. There’s good evidence to believe that those types of risks, indeed, are largely remnants of the past. But if we define political risk a little more broadly as eventual financial losses due to policy shifts, then political risk today may be even more acute and pervasive than in the days of the late twentieth century, albeit subtler and harder to detect. 

The changing quality and growing quantity of political risk is directly related to three distinct yet interconnected forces occurring in the last 50 years: the rise of globalization, the explosion of information, and the growing importance of, to use George Soros’s term, “reflexivity.”

The Globalized World

It would be an understatement to say the global economy is different today than it was in 1970; the same can be said for political risk. Back then, world output was less than 20% of what it is in 2020, with only modest trade between countries. The world during the Cold War was mainly a patchwork of inward-focused economies, each of which relied on domestically manufactured and sold goods. Very little cross-border trade in finished products existed, and what did exist occurred between only 20 or 30 “First World” countries.  China and the former Soviet Union’s communist and socialist economic models closed them off from trade. Even in the United States, trade comprised less than 1/7th of economic activity. Cross-border risks were more geopolitical than economic. Analysis was not easy, but the demand for it was limited: only a few “risky” countries even had accessible financial markets, and aside from a couple dozen global banks, few financial investors routinely invested abroad. 

The 21st-century world economy has been transformed into a complex system of interdependent and constantly changing relationships. No longer a patchwork quilt, the global economy today looks more like an intricate tapestry. Just when you think you understand the overall pattern, it quickly unravels and reconnects to form an entirely new tapestry, reflecting the rapid rate of change in supply chains and financial markets that now include more than 100 countries. Today’s cross-border investments entangle trillions of dollars, with financial markets now multiples the size of the world economy. 

The difference between political risk a generation ago and today is like the differences between checkers and chess: The 8×8 board of squares may look the same, but the games could not be more different. While even a child can master the finite universe of checkers moves, chess requires an ability to think many moves ahead and to understand the future interactions and consequences of both your own and your opponent’s moves. 

The Informationalized World 

The analysis of politics and policy change has never been easy, but globalization has made both of them infinitely more complex. The amount of information generated by and about all the players in this new world economy is unprecedented. In fact, from the beginning of recorded time until 2003, humanity created only five exabytes of data—the equivalent of five billion gigabytes. By the time you read this, we might be creating five exabytes every 10 seconds.

Modern networks can now churn out huge volumes of data that flow to supercomputers for analysis. When objects can both sense the environment and communicate, they become tools for understanding complexity, tools with an unprecedented ability to respond swiftly. What’s truly amazing about this revolution is that it’s quietly being deployed already, often working without human intervention. This new megatrend, often called “Big Data,” really is big. Coupled with faster computing, Big Data has the potential to act like global X-rays, allowing us to see things in places that had previously been invisible to the naked eye. 

With a world so complex—and a volume of data so vast—analysis of political risk may seem impossible. But a symbiotic evolution of machine learning has occurred alongside and in conjunction with this information explosion. Humans minds have been ahead of machines for complex analysis for most of history, but that is certainly changing. To return to the board game analogy, it took until 1997 for IBM’s Deep Blue computer to beat grandmaster chess champion Gary Kasparov, and it took another nineteen years for AlphaGo, a program driven by Google’s DeepMind artificial intelligence, to beat Lee Sedol, one of Go’s most dominant players. Today, the speed at which machines can grab, calculate, and process information makes AlphaGo seem quaint. 

Algorithms, no matter how sophisticated, can only analyze the information they are given; the old “garbage in-garbage out” maxim still applies. The masses of information coming from incalculable sources may be accurately gathered and observed, or it may be inaccurate due to mistake, sloppy recording, misinterpretation, fraud, or even malevolent manipulation. No one has yet found a reliable way to filter the “real” from the “fake.” Perhaps former US Secretary of Defense Donald Rumsfeld was onto something when he infamously quipped in 2002 about the war in Iraq:

There are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns—the ones we don’t know we don’t know . . . 

Lambasted by the press for this purported gobbledygook, Rumsfeld was basically talking about political risk. If he were speaking today, Rumsfeld would have needed a fourth category: “unknown knowns,” or things we think we know but which are, in fact, untrue. And knowing what is true is getting tougher and tougher in the 21st century. 

The Reflexive World

Today we have countless economic and political actors generating, receiving, and analyzing unfathomable amounts of information that may or may not be “true” or “real.” The way this information influences personal and political decision-making—and vice versa—is what George Soros called “reflexivity” in his 1987 book The Alchemy of Finance, and it is the essence of today’s political risk. Soros has theorized that there are two realities: objective and subjective. Objective realities are true regardless of what participants think about them. For example, if someone says that it’s 40 degrees Fahrenheit outside and, in fact, it is 40 degrees Fahrenheit outside, then that is an objective truth. Someone might claim that it’s 85 degrees, but that would not make it 85; it would still be 40 degrees Fahrenheit. 

Subjective realities, on the other hand, are shaped by what participants think about them. Soros says that financial markets fall into this category because participants can never know all the objective realities driving a complex, 24/7 interconnected financial system. Investors, therefore, use their best judgement as to what assets are worth. Soros believes that collective thinking is what moves markets and produces outcomes. That is, our collective subjective reality affects objective reality, and that objective reality, in turn, affects our subjective reality once again in an endless feedback loop. 

For example, how did Amazon become one of the world’s most valuable companies? For years after its 1997 initial public offering the company made little or no money, but investors pushed its stock price higher, perhaps believing that its Internet model would eventually dominate the global economy and crush brick-and-mortar retailers. According to Soros, the reasons people believe something are unimportant. What is important is that these beliefs shape objective reality.

How did this reflexivity work with Amazon? Subjective beliefs kept its stock price high for years while the company lost money. High stock prices allowed Amazon to curb salaries and use stock options as part of compensation. This helped Amazon attract exceptional talent, which, in turn, led to increased innovation and growth beyond bookselling. Because investors kept believing, Amazon could continue on for many years profitless while grabbing market share from profitable, traditional retailers. If subjective realities were different, Amazon may have died like Booksamillion and thousands of other dot.coms. Some would argue that Elon Musk’s upstart Tesla, in contrast to stodgy incumbents like Ford and General Motors, may be riding (figuratively and literally) such a subjective reality.

Soros’ theory about the reflexivity of financial markets applies similarly to the realm of politics. For example, what if voters believe polls that predict a frontrunner will be a landslide winner over an underdog? The frontrunner’s supporters may be less motivated to vote, while the underdog’s supporters may be more motivated, leading to an underdog win:  subjective beliefs altered the objective outcome.

The Business of Fake News

On June 23, 2016 the world was stunned: British citizens voted to leave the European Union. Less than five months later, on November 9, 2016, another stunning upset occurred when American voters elected businessman Donald J. Trump as the 45th president of the United States. More recently, in May 2019, Australia’s Labor Party was widely favored to win national elections but suffered a surprise defeat.

These events were especially shocking because, for weeks leading up to these national referendums—even hours before the votes were tallied—pundits, policy wonks, pollsters, professional handicappers, and some of the world’s most sophisticated investors confidently predicted the opposite outcomes. Despite enjoying access to torrents of data and the computer capacity to analyze their findings, the experts were proven wrong. 

The risk of unexpected political and governmental developments such as these have always been a critical factor in cross-border investing, but the effects were much more localized in the past. Certainly, in the mid-20th century, a surprise Brexit vote or Trump election would have had some impact on American and, perhaps, certain European financial markets. But in 2016 we saw synchronized falls (and subsequent rises) on dozens of stock exchanges worldwide that demonstrated an unprecedented interconnectedness of global markets. 

Many got Brexit, Trump, and Australia wrong because they failed to account for intentional—and successul—manipulation of subjective realities through the manufacturing of misinformation. Hindsight has revealed the ubiquity of the misinformation industry. For example, many of the fake news websites that sprang up during the 2016 US election campaign have been traced to the small city of Veles in Macedonia, where teenage sock-puppets were found to be manufacturing sensationalist stories to earn cash from advertising (Kirby). In 2016, according to Twitter, an army of Russian and Iranian trolls sent thousands of messages urging Brexit on the day of Britain’s referendum on EU membership, with approximately 3,800 fake accounts tweeting out 1,102 posts with the hashtag #ReasonsToLeaveEU (Field). Fake messaging has gone so rampant that in mid-2019, Facebook announced removing more than 2.2 billion fake accounts—nearly the equivalent of the number of real people who use Facebook (Price).

Emerging market countries are also victims and perpetrators of fake news in the political arena. In Brazil, for example, television ads traditionally dominated presidential politics (as they had in the US). But in the last decade nearly 70% of the electorate opened an active social media account, making the Internet the dominant “news” medium. Research suggests that at least 86% of Brazilian voters—and 98% of supporters of 2018 upstart presidential candidate Jair Bolsonaro—were delivered fake news (Harden)

. Similar to the Brexit and Trump votes, Bolsonaro won despite predictions to the contrary.

One could argue that these information manipulation tactics, which to some extent seem like modern iterations of an old KGB Cold War playbook, epitomize the real political risks of our time: discrediting and destabilizing political systems 24/7, planting stories to erode faith in governments and leaders, undermining trust in established media, and stoking conspiracy theories. As deepfakes become ubiquitous, citizens around the world will likely struggle to believe what their eyes and ears are telling them—even when the information is real. In our 21st century of hyper-competition, such tactics can divide long-stable nations into competitive tribes, preying on old and new grievances and inequities and leading to civil dysfunction, greater authoritarianism, and perhaps even social collapse. 

The increased reliance on social media for information, combined with high—but often misplaced—levels of trust in online content, has made both industrialized and emerging market populations vulnerable to falsehoods – with reflexive consequences. This trend is not limited to either commerce or electoral politics but rather is a new form of political interaction with the power to change both individual and group values and behavior, enhance subcultures (for better or worse), and spread misinformation with unprecedented speed. Propaganda has been with humans for millennia, but in the modern context, fake news and misinformation can reach willing and open minds as well as manipulate and amplify opinions that already exist in ways we are just beginning to understand. In the political realm, fake news propagated from subcultural movements—often written by bots—can strategically leverage racist, sexist, or religious predispositions. Its fallout alters political outcomes and policy trends. In extreme cases, such information can lead to protest, violence, or even jihad. This new form of political interaction has truly created new forms of political risk. 

At a minimum, fake news leads to confusion over what information is true or false. These beliefs can also lead to an increase in government mistrust as well as disbelief over hard news sources. During the 2020 COVID-19 pandemic, for example, fake news and skepticism may have accelerated and amplified the spread of the virus in some countries. Believing online accounts that COVID-19 was a hoax led many to resist health authority measures to prevent the virus’ spread. There were also countless recommendations of unproven drugs and remedies (including injecting detergent bleach)  that, in some cases, led to hospitalizations and deaths. 

The Future: From Alternative Facts to Alternative Data

Political risk today encompasses constantly shape-shifting interactions between innumerable political, corporate, and individual actors; incomprehensible amounts of data; and technologies for manipulation that have yet to be successfully challenged. Can the business of political risk analysis adapt to this new world? A look at the evolution of the business may provide some insight.

Data about political risks have driven Wall Street decisions for decades, but information was once much scarcer than today. For example, in the early ‘90s the Mexican government had borrowed heavily in US dollars and dollar-linked instruments while trying to maintain a managed peg currency.  While investors could follow coverage of Chiapas uprisings or the leading presidential candidate’s assassination, they had to wait weeks or even months for official economic data. More specific information about Mexico’s dollar assets and liabilities was almost impossible to obtain. Therefore, it was a big surprise when Mexico abandoned its peg in December 1994 which created a contagion effect that hit many emerging market asset prices. 

Back then, “political risk” firms were small boutique consultancies that would publish newsletters (via mail or fax)  and hold conference calls. Some like that still exist today. Many of them, based in New York or Washington, were staffed with multilingual former diplomats, government officials, or global Fortune 500 executives, each with some version of a “Man in Havana,” a local foreign contact who had an ear to the ground. Without Internet data and computer power to crunch numbers, these firms provided largely what can be called political risk “chat therapy” —consultations that made clients worry less. True, some of them did offer valuable insights into specific places and events, but many were not much more than well-groomed and well-educated soothsayers. That’s not to say that such services were totally worthless. Sometimes chat therapy, even without unique insights, can calm anxious investors and allow for more rational decision-making, particularly in the absence of timely, credible data. 

But by the end of the assorted foreign crises of the 1990s, the Internet spread. Central banks created websites, and data became more timely, granular, affordable, and plentiful. By the time Argentina abandoned its peg and defaulted in late 2001 there were few contagion effects: the country’s central bank had been publishing daily US dollar reserve positions for years. Argentina’s collapse surprised few;  it was a slow-motion train wreck watched by the world for much of 2001. Investors had plenty of time to devise strategies to deal with a potential meltdown.

This is how things are different. Yes, it is still a caveat emptor world, but there are much more data available today than in the pre-Internet world to help investors avoid many of political risks feared in the past. Some of this data includes the same official government statistics, with less time lags and better accuracy than in the past. But this data is only the beginning of a larger disruption in the political risk business.

Over the last decade the information revolution has given rise to “alternative data”—new knowns that were previously unknowns. Alternative data is a loose term to describe digitized information, including satellite images, social media feeds, search trends, video footage, digital payments, mobile geo-location data, and sentiment data. Indeed, the digitization of everything makes data out of everything. This is the new world of “data science,” and data science is and should be more important for political analysis than it was in the past. It is based on information that relies on machines to help interpret, process, and analyze. 

When it comes to political risk, most investors and businesspeople are like tourists checking the weather. They’re interested in short-term forecasts—for tomorrow, the next few days, or maybe weeks—to be able to buy, sell, or hedge economic exposure. Interestingly, before the telegraph and the telephone, the only way of predicting the weather was for humans to use immediate experience. Given the weather on a particular day, what kind of weather usually follows during the next day or two? As you can imagine, the success of such forecasting was not much better than a random guess or coin toss. In some ways, old school political risk forecasting hasn’t been that different. 

This data revolution has and will be harnessed by many new political risk players. In addition to more granular and timely economic, financial, and commercial data in digitized form, increasingly political risk firms are capturing data and devising metrics on a myriad of phenomena that directly or indirectly reflects political risk including air quality, child labor, civil and political rights, bureaucratic strength, civil disorder, climate change, corruption, crime, deforestation, education levels, election patterns, foreign pressures and vulnerabilities, gender issues, government stability, health statistics (which became highlighted during the COVID-19 pandemic), law and order, religious tensions, terrorism, water and food stress, and working conditions. Such data—some objective, some subjective—can be regressed against a variety of financial and economic data. Some firms are also creating and analyzing data and “metadata” drawn from traditional media, social media, and search sources. Using keywords and machine-learning software, humans and computing programs can organize data into scores that, over time, create baselines that can be used for political risk decision-making. The advantage of this approach is higher frequency snapshots for investors—something credit rating agencies currently don’t deliver. 

Can these newcomers forecast as well or better than the old tea leaf readers? It’s probably too early to tell. And in many cases, the type of forecasting needs to be refined. What are we trying to forecast? Coups or civil wars? Those “lose everything” situations are rare and becoming rarer. And because of more information, those situations are less likely to be a surprise. Most would argue that we need to better understand shorter-terms political risks that affect markets and cause financial losses. 

In this respect, new and broader political risk data can and should be used like any other statistical data in business. For example, if, on the eve of a presidential election, an investor identifies social media trends that suggest a close election, a hedging strategy could be employed—this is, after all, essentially what investors do before a company reports its earnings. Indeed, Bloomberg, for years a gold-standard platform for professional investors, now tabulates Tweets and social media posts for its clients to analyze linkages to asset prices. 

Like chess, political risk analysis will evolve. When Deep Blue beat Kasparov in 1997, many thought the game of chess would die. On the contrary, “freestyle” chess is thriving and often described as “centaur chess”: half-machine, half-human. Tyler Cowen has written extensively about how this new form of the game amplifies human performance, combining human intuition and creativity with the machine’s power to calculate the endless number of possible chess moves and outcomes. Indeed, there is a growing online community of chess players and their computers playing other teams and their computers. Centaur chess may provide clues as to the optimal interplay between humans and machines and how political risk analysis is unfolding. Teaming the two may lead to better analysis than either humans or computers could manage on their own.

In our brave new world of “alternative data” and “alternative facts”, technical expertise—rather than human sleuthing and networking—will hold the key to future success to managing political risk. Instead of relying exclusively on ex-government officials or a woman in Havana, the typical forecaster will likely rely on a crack squad and their supercomputers in Silicon Valley, Bangalore, or Tel Aviv. Even with such new capabilities, one should always remember superforecasting scholar Philip Tetlock’s wise observation that better predictive judgement is based on a humble acknowledgement that nothing is 100% certain and that anything is possible. With such a mindset and new data tools, future political risk outcomes from close elections or responses to crises like pandemics, will be just that—outcomes—and not surprises.

Peter Marber, PhD. Is a Wall Street veteran and teaches at Harvard University. This article is excepted from his recent book, Quid Periculum? Measuring and Hanging Risk in the Age of Uncertainty, with Christopher McKee (available at

Works Cited

Field, M. and Wright, M. “Russian trolls sent thousands of pro-Leave messages on day of Brexit referendum, Twitter data reveals.” The Telegraph 2018.

Harden, C. Woodrow Wilson International Center for Scholars. 2019. Available at 

Kirby, E.J. “The city getting rich from fake news.” BBC 2016.

Price, R. “Facebook banned 2.2 billion fake accounts in the first 3 months of this year. That’s almost equal to the number of real people who use it.” Business Insider 2019.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s