Hours after the results of a risk model reached the public, U.K. Prime Minister Boris Johnson and U.S. President Donald Trump asked nearly 400 million people to upend their lives.
A group of epidemiologists at Imperial College London published the findings of their model on March 16, and the forecasts were grim. Letting the new coronavirus run unchecked through the population could result in 500,000 deaths in the U.K. and 2.2 million deaths in the U.S. The estimates, based on data showing how quickly the virus was spreading among people, made headlines around the world.
That same day, Johnson urged all citizens to stay at home and Trump laid out voluntary social distancing measures. Both leaders made U-turns, after previously downplaying the need for severe restrictions to control the spread of the virus.
Even if you don’t live in the U.K. or the U.S., risk models like the one used by the Imperial team are now governing your life. More than 2.5 billion people are in countries that are on lockdown or face severe restrictions on social gatherings—with the number growing. In many countries with only a handful of Covid-19 cases, officials are using modeling exercises to decide their next steps.
Every person infected with the new coronavirus is estimated to infect nearly three others, who will go on to infect nine, who will infect 27 and so on.
Risk models are mathematical representations of complex systems that help show what can happen under specific conditions, and it didn’t take a pandemic to introduce them into our lives. Some widely used models have low uncertainties, such as those determining the price of your auto insurance premiums based on demographic data produced from millions of insured drivers. Models projecting how the economy will perform if inflation suddenly rises carry far more uncertainty.
No model always gets it right. “Even I find it tough to judge how much I should believe what comes out of a modeling exercise,” said David Spiegelhalter, professor of public understanding of risk at the University of Cambridge.
Public awareness of big problems often starts with modeling. Climate change first entered public consciousness with the help of models that used well-known physics and, initially, little observational data. Scientists are still refining models that have been in use for four decades, with a remarkable record of accuracy. This can give sudden shifts to models an ominous prospect. Recent updates to these climate models shows that the world might warm much more than previously thought, and scientists remain puzzled why that might be the case. In the case of Arctic sea ice, on the other hand, models failed to predict the steep decline observed in the last two decades.
Models are only as good as the assumptions they are based on. Climate scientists are still not sure how much of a role clouds, which can trap or reflect sun’s heat based on their type and their altitude, play in affecting the planet’s warming. And more than three months after Covid-19 emerged in China, scientists still don’t know how many people can infect others without suffering from symptoms themselves. Because of the uncertainties, in both cases, current models produce large differences between the lower and upper bounds of forecasts.
Scores of scientists each spend thousands of hours working on honing the models used to understand infectious diseases and climate change, and the process of improvement won’t be able to iron out all uncertainties. Even when things are not moving as quickly as in a pandemic, the number of variables involved in a model increases the degree of uncertainty. Which is not to suggest models are just figments of computer simulation. They are informed and greatly enhanced by observations and empirical data. It’s just that there will always be uncertainties attached to any model trying to predict the future—no matter how much computing power or human brains you put to the task.
“That uncertainty does not mean we don’t know what to do,” said Spiegelhalter. “It actually is a guide to being more cautious.”
That’s why drastic steps are sometimes taken after reading risk models. In the case of Covid-19, billions of people remain on lockdown until the growth in the number of infections comes under control. For climate change, where policy responses have never been anywhere near as draconian, models have been used to make the case for painful measures such adopting high carbon taxes now rather than gradually increasing taxes as things get worse.
Blindness to Uncertainty
The highest use of models is to save us from our own habitual blindness to uncertainty and risks, especially when confronted with events that don’t always make intuitive sense. “Modeling plays an essential role,” said Baruch Fischoff, professor at Carnegie Mellon University and an expert on risk. That’s because both pandemics and climate change are “non-linear” events, which human beings can’t really wrap their heads around.
The non-linear math of the Covid-19 pandemic works like this: Every person infected with the new coronavirus is estimated to infect nearly three others, who will go on to infect nine, who will infect 27 and so on. With climate change, non-linear math means every one-tenth degree Celsius of warming will have a more severe impact on the climate than the previous tenth degree.
After the pandemic is brought under control, there’s hope that the experience might lead to more “risk thinking” among leaders in governments and businesses, said Nigel Brook, a partner at the law firm Clyde & Co. who runs their practice on climate risk. Having experienced such an event, Brook believes it is likely more leaders will consider “tail risks,” events that had a low chance of occurring. This is especially true if the impact of such an event could be large, and few problems carry greater tail risks than climate change.
There are clear moments in the past when the experience of unlikely tail risks prompted bold action. Back in 1953, for instance, the Netherlands experienced severe flooding when a combination of a storm, high tide and low pressure caused water levels to rise as much as 18 feet (5.5 meters) above sea level. The natural disaster destroyed thousands of homes and killed nearly 2,000 people. Within a few years, the Dutch government launched Delta Works, which became one of the world’s largest infrastructure projects. The series of dams, sluices, locks, levees, dykes, and storm-surge barriers took 40 years to complete and now protects two-thirds of the country’s land that faces the risk of flooding.
Global Not Local
Global risks like pandemics and warming temperatures are qualitatively different from localized catastrophes, such as small epidemics, wars between two countries, or natural disasters. They can’t be solved by one country and necessarily need global solutions.
“[Global problems] change the basis on which we can trust in science and trust in institutions, which are charged with managing these risks,” said Jamie Wardman, professor of risk management at the University of Nottingham. That’s why “risk has become central to our understanding of how the world has been transforming in the past century.”
Will Covid-19 make people more aware of global risks? While each country is mostly dealing with the new coronavirus on its own, institutions like the World Health Organization have proven to be crucial in providing actionable advice on how to manage the spread of the disease and sharing key data between countries. It’s this data that ends up fueling the risk models and the grim forecasts.
With climate change, the data is already shared. So are the models. That’s why the aftermath of the pandemic will create a clear test for global institutions to turn a new awareness of risk into real action.
–With assistance from Eric Roston.
Copyright 2020 Bloomberg.