What the current situation boils down to is this: is economic meltdown a price worth paying to halt or delay what is already amongst us?
— Tom Jefferson and Carl Heneghan, Centre for Evidence-Based Medicine, Oxford University
April is mathematics awareness month, and this essay is about math and probability. Actuarial tables are used by the insurance industry to calculate risk. Actuarial data suggests that for Covid-19 there are significant variations in risk. In all likelihood, the number of deaths from Covid-19 will be comparable to other causes of death in the United States, such as influenza, car crashes, or suicide. And using an actuarial approach to pandemics helps us think about how to use data in balancing the risk of death against the certainty of self-inflicted recession and mass unemployment.
Let’s start by using Italy and China to set upper and lower bounds on risk. It is unlikely the U.S. mortality rate will be higher than Italy’s or lower than China’s reported figures. The effect of Covid-19 on the United States will fall between the two. The U.S. population is five times larger than Italy’s, suggesting that with the same mortality rate, we should expect 87,000 deaths. Italy has the older population, more smokers, and the highest population of antibiotic-resistant individuals in Europe, making it vulnerable in ways that the United States is not. Statistics from China are suspect, since they are tailored to fit the Communist Party’s political needs. China’s claim that it has a fraction of the infections the United States has is ludicrous. China’s population is four times larger than the United States, yet it reports around 83,000 cases and 3,300 deaths. Some reporting suggests that China may conceal two-thirds of its cases. Other estimates suggest China underreports deaths by almost a factor of 20 . Even if adjusted for underreporting, China’s data suggests only 18,000 U.S. deaths. To put the figures from Italy and China in perspective, the Centers for Disease Control (CDC) says United States suffered around 34,000 deaths from the flu in 2019.
Bad Data Means Bad Decisions
Deciding when the cost of not letting people go back to work outweighs the risk of increased mortality requires estimating the likelihood of death. The dilemma here is that we do not know the actual number of infections, and respected researchers say that mortality estimates are off by an order of magnitude. One study in Lancet, a leading medical journal, concludes that “[a]lthough highly transmissible, the CFR [case fatality rate] of COVID-19 appears to be lower than that of SARS (9.5%) and Middle East respiratory syndrome (34.4%), but higher than that of influenza (0.1%).” Another estimate by two Stanford medical researchers concluded that the risk of death may be even lower.
This means we can dismiss predictions of a pandemic like the Black Death or the 1918 influenza epidemic as exaggerated. Some of this exaggeration is driven by discontent with an administration widely perceived as erratic. Unfortunately, some reporting on Covid-19 is equally erratic. Approaching the pandemic as an actuarial problem suggests that too much of the public discussion is irrational. Reports that the United States could face 200,000 deaths are at variance with the experience of every other country and would require a mortality rate three times higher than the Italy’s.
The problem with a maximalist “complete shutdown” is that it maximizes economic harm. We need to weigh the Covid-19 response against the social distress caused by increased unemployment. Some estimates predict massive unemployment and a loss of as much as 24 percent of GDP because of the shutdown. If the 2008-2009 global financial crisis is a precedent, suicide rates will increase dramatically as will addiction and homelessness. We know that recession and unemployment decrease life expectancy and are accompanied by and exacerbate a range of other social problems that can last for years. These social problems will harm a much larger percentage of the U.S. population if we do not change course. There is also a real risk that absent a data-driven approach to risk, recovery will be much slower.
Actuarial data shows that Covid-19 mortality rates vary widely by age, health conditions, and location. Testing data would let us assess whether there are intermediate measures that reduce risk of mortality without massive economic harm. The initial lack of testing data is a significant handicap for determining how to manage risk. Actuarial data suggests that tailored policies could be effective. Countries such as Singapore and South Korea have contained Covid-19 with widespread testing, aggressive tracking of contacts, and selective quarantines. Singapore is much smaller, but similar methods could work in the United States if it were better organized. Countries may have been forced to impose mass shutdowns because they missed the opportunity to use less severe measures.
Data and Democracy
The math behind the Covid-19 response points to several areas for further work. The first, judging from the experience of the countries that were more successful in managing the outbreak, is how the United States can better organize itself to acquire data and use it to guide policy, making it easier to develop tailored responses. This may require some rethinking of data protection rules as well as the development of technologies that maximize anonymization of the personal data needed to track and control infection. A data-driven policy could reduce mortality while avoiding economic collapse. The CDC provides a foundation, and we need to ensure it has the adequate resources to take advantage of the most advanced data analysis and artificial intelligence (AI) tools. In this, Covid-19 points once again to the risks of federal policies driven only by a desire to cut spending.
Covid-19 also highlights a larger problem for democratic policymaking. Elected officials are influenced by public opinion. In turn, public opinion is shaped by media reporting, some of which can be inaccurate or alarmist. The internet and social media increase a natural human proclivity to emphasize risk. The mechanisms of democratic policymaking need to adjust to this new era of rapid and pervasive access to information and opinion presented in ways that make it difficult to distinguish between the two. Technology can remedy this, but the need for better data in democratic governance comes at a time when traditional sources of expertise in academia and the media face skepticism and new competition in the online environment.
In the future, better technology might allow citizens to ask, “Siri, is this claim of 200,000 American Covid deaths accurate?” The response from an improved AI assistant would be, “No, it is wrong, and here’s why.” The algorithms currently used by digital assistants are not sophisticated enough to do more than scrape the web, guaranteeing inaccuracy. What the new mechanisms for public policy in the information age will look like, how they will be shaped by improvements in the automation of knowledge and data analytics, and, above all, what voters will accept as authoritative is far from clear. Technology has brought us a contaminated information space; now the task is to see if technology can clean it up.
We Have Made This Mistake Before
The public discussion of Covid-19 has overemphasized risk. This is important because self-inflicted damage from overreaction to risk does real harm. After the 9/11 attacks, the Intelligence Community held a wargame on the responses to nationwide terrorist attacks. The game found that draconian responses (which included shutdowns) did more harm than terrorists by snarling the economy. We are approaching a similar situation in the response to Covid-19. The best policy would balance the risk of additional deaths against the certainty of economic damage, but much of the discussion has underestimated social and economic harm.
Actuarial data helps us make decisions about risk, and in this case, it is the risk of ending the draconian shutdowns. Shutdowns cost the United States about $100 billion a week. We can predict that mortality will crest in mid-April and then decline. By then, Americans will be both impoverished and fed up with house arrest. While there will be expressions of fear and concern, social need should drive the United States to reopen in May if we are to avoid economic tragedy.
James Andrew Lewis is a senior vice president and director of the Technology Policy Program at the Center for Strategic and International Studies in Washington, D.C.
No comments:
Post a Comment