The cows were not safe. To be sure, they were mad. But what made them unsafe was that anyone consuming them may well become mad. That is what the UK discovered in the 1990s. It was found that cattle affected by bovine spongiform encephalopathy (or BSE) could cause a variant of Creutzfeldt-Jakob disease in humans. That disease would mentally impair its victims and eventually take their lives. As of 2013, 177 people in the UK had died. Not surprisingly, no one wanted to consume cattle that might have BSE.
The reaction of the United States to cases of BSE is instructive. In 2003, a cow imported to the US from Canada was found to have BSE. Imports were banned. In Canada, cattle prices fell by a half and retail beef prices by 14 percent. Canada’s annual beef export revenues to the US fell by two thirds. At the time, Canadian beef was three quarters of US beef imports so this imposed costs on both countries; with losses estimated in the billions.1 When, in later in 2003, an infected cow was discovered in Washington state in the US, the trade bans fell on the other foot.
As internal bans were neither palatable or practical, the US Department of Agriculture (or USDA) ramped up testing. It favored what was argued to be a less accurate ‘rapid’ immunologic test (with results delivered in hours rather than weeks). The costs of these tests was about $200 million but the positive impact on reviving the US beef export industry was far in excess of this.
This chapter is about the value of testing and how it can improve the functioning of markets when there are infectious diseases. The BSE example indicates that for the beef trade and it has strong lessons in the wake of COVID-19 for how the testing of people can make it safe for people to interact with one another. But before getting to the meat(!) of the issue, there was one more twist in the USDA’s handling of BSE testing. Having successfully demonstrated the economic value of tests, the USDA promptly banned them.
You read that right. The USDA forbade cattle exporters from paying for the tests themselves for their own livestock. A producer of black Angus beef for sale to Japan, Creekstone Farms Premium Beef wanted to use the USDA’s approved rapid test as part of its production and marketing efforts. The reasons were obvious. It was commercially lucrative to provide that information to customers. The USDA claimed that using the test was for “surveillance” purposes and was concerned that if some producers tested their cattle this would imply that the cattle of others was unsafe. Cattle trade associations feared that this would lead to an unravelling necessitating all producers to incur the costs of testing.
Creekstone sued the USDA and, initially, prevailed.2 The USDA’s position wasn’t ludicrous as a matter of economics. Many economists had been concerned that in some markets, particularly higher education, there may be undue costs to signaling and that there may be a social rationale for banning such contests.3 (For instance, students spending enormous effort to get into a slightly higher-ranked college even though the learning outcomes were the same). However, in this situation, the Court realized that there was a customer who was particularly sensitive about certification of quality and that in the absence of a threat to public safety, there was no reason to prevent a business’s right to use tests to assist in their marketing. The only rationale for prohibiting the use of the test is if the test was uninformative. They weren’t. In other words, the tests could not simultaneously be effective in identifying a safety concern and ineffective in certifying product quality. The USDA appealed and the US Court of Appeals reversed the decision and returned to the USDA the power to regulate BSE test kits which it exercised. Private testing was banned.
The interests of economy and public health collide because the most important way to deal with a pandemic in the interests of public health is to slow the rate of infection (that is, R0). A person having a disease is a health problem that requires knowing how to treat that person and then doing so. A person having an infectious disease is a public health problem because, in addition, that person can pass the disease onto others. Being infectious is what turns an isolated health problem into an interdependent one. Because our typical dealings with other people rely on others being safe to interact with, pandemics destroy interactions and with it, the economy.
The BSE infections showed a microcosm of how a lack of safety impacts interactions — in this case, between cattle and humans. But they also showed the importance of knowledge. There is a big difference between knowing someone you interact with is infectious and having to make a guess as to whether that person is infectious. In the former case, you can act and limit the interactions. In the latter case, you have to take a risk. And in evaluating that risk what we care about is not just whether you become infected but also whether you might pass that on to others.
If your goal is to minimize the public health risk, a lack of knowledge dramatically reduces your number of options. If you have no information whatsoever regarding whether any given person is infected, then you have to engage in blanket isolation policies to reduce the rate of infection. You are forced to make judgments regarding what is and what is not essential and draw your isolation boundaries around those lines.
Imagine, for the moment, that instead of no knowledge you had perfect knowledge of whether any particular individual is infected or not. To give you a picture, imagine the virus was such that it inflated people’s noses and made them shine bright red like Rudolph the Red Nosed Reindeer. Imagine also that as those people moved, they left a trail of red that you could see even after a number of hours. Then anyone could easily identify who is safe to interact with and who is not. For those who were unsafe, we could isolate them or only approach them if they or you had suitable protective gear.
The difference between perfect knowledge and no knowledge is what makes an infectious disease have its impact on social and economic interactions. With perfect knowledge, some people get sick, they are isolated and life is essentially unchanged. With no knowledge at all and no interventions to prevent infections, then for COVID-19, at its peak, about 21 million people in the US alone would likely be infectious at one time. With no restrictions on activity, the probability that you interact with one of the infectious people on a given day is 21m/327m or 6.4 percent.4 However, suppose you interact with only ten people a week. In that situation, the probability that you are able to avoid any of those infected people is about fifty-fifty. When you think about the fact you go to public spaces, it is more likely you interact with over a hundred people a week. In that case, your probability of avoiding an infected person becomes close to zero. In other words, perfect knowledge allows you to avoid all infected people. No knowledge makes it near certain that you will encounter one.
The key to making people safe is knowledge. To be sure, one way that can occur is to let the virus run its course without interventions. Of course, that is tantamount to saying that public health will not be prioritized over the economy. In the US, that likely means 15 million hospitalized at the peak and more sick at home. And there are another 3 million likely to die. That is the underpinning of the dark recession scenario. Suffice it to say, if the goal is to make people safe for interactions, making them completely unsafe for a period of time would appear to defeat the point.
How do we gather that knowledge? The answer with respect to COVID-19 was testing.5 There are two types of tests that are relevant. First, there are tests that can indicate the presence of the coronavirus in an individual. Second, there are tests that can indicate the presence of COVID-19 antibodies. One type tests whether you have the virus and are likely to be infectious (e.g., equivalent to the Rudolph thought experiment above) while the other tests whether you have had the virus and are likely to be immune. At the onset of the COVID-19 outbreak, tests for the presence of the virus were available and, depending on the country and the test, there were differences in how quickly they could yield a result. As of the time of writing, antibody tests were being developed but were not verified, let alone widely available.6 The two types of tests, which I will refer to as HAVE and HAD respectively, play different roles in making people safe.7
The first thing to note is that a HAVE testing regime potentially makes a HAD test redundant. With a perfect HAVE testing regime, you would test everyone at a regular interval and the test would, with high confidence, tell you if you HAVE the virus or not. Given this, on the assumption that having the virus would give you immunity, you would not expect to learn much more from a HAD test.
For COVID-19, no country had a perfect HAVE testing regime. In general, as tests were not widely available, different jurisdictions would have different policies regarding the factors that, if present, might require a test. This is somewhat paradoxical because, if you had COVID-19 symptoms (such as a fever, cough or shortness of breath) you were more likely to test positive for the virus. Thus, if a person with symptoms had a positive test, this is actually less information that would be gained if a person without symptoms had a positive test result. Put simply, the stronger are your COVID-19 symptoms, the less valuable a test is.
This is especially the case as many infected people were, in fact, asymptomatic. At the time of writing, it was not clear how infectious asymptomatic people who carried the virus were. However, what was understood was that some symptoms, particularly a cough, would make people more infectious.
Moreover, one value of testing is that it can inform public health officials of the characteristics of the disease including the base epidemiological properties such as R0 and how infectious asymptomatic carriers were. Put simply, if you test symptomatic people and find that, say, 80 percent of them have COVID-19, then without knowing how many asymptomatic people have the virus, you only know that a random person in the population likely has COVID-19 with less than 80 percent probability and not how much less.
Fortunately, there were situations where HAVE testing was conducted without reference to underlying symptoms. One case was the Diamond Princess cruise ship that was quarantined in Japan for a period of time and ended up having many victims. However, a cruise ship does not match the properties in the population in terms of like transmission rates (it is a unique situation) nor in terms of other factors such as mortality (as the demographics were different). A better indication came from a pro-active study of the town of Vò in Italy whose entire population of 3,300 was tested and retested regardless of symptoms. It was discovered there that half of the positive cases were asymptomatic.8
The Vò experience also highlighted the effectiveness of using HAVE tests to identify who should be isolated. The first testing round found three percent of the population with the virus. They were isolated and a second round of test found only 0.3 percent still infected. Importantly, that was not zero and there were still six infected people who also had no symptoms. Identifying them prevented a re-emergence of the infection in the population.
There is one final remark to make regarding testing and symptoms. Symptoms are themselves a type of test albeit one with error. For instance, one cost-effective way of regular testing is taking temperatures. These are done at some border crossings and other places where there might otherwise be larger gatherings of people. The problem with this test is that an elevated temperature may be consistent with other things; for example, the flu. This can be important, as shown in the following table:
Shortness of breath
Notice that, alongside fever, other symptoms are common to both COVID-19 and the flu. The main symptoms that are more clearly common with COVID-19 than the flu are shortness of breath or respiratory issues. Thus, it is these symptoms that give the clearest indication that a person has COVID-19.9 They may, of course, be hard to measure if they are mild as the baseline may differ between individuals.10
There is one method that would assist in targeting asymptomatic people for testing and then isolation: contact tracing. This requires an intensive effort to identify who any person who came in contact with someone who tested positive for COVID-19 (or was otherwise suspected to carry it) over the past week or so. In doing this, those people can be identified and then prioritized for tests (and potentially further contact tracing) even if they do not exhibit symptoms. Again, the goal with testing or gathering information is to be able to isolate people on a more targeted basis than blanket policies that lockdown entire regions.11
To summarize, a HAVE test is useful because it enables an action. That action is to isolate or quarantine any individual with a positive test until such time as they are held (through additional testing or otherwise) to no longer be infectious. The value of this strategy is that it is a potentially more cost-effective (in terms of impact on economic and social life) than using blanket isolation policies to reduce rates of infection.12 In this way, the availability and use of HAVE testing is a potential way in which countries can reduce the depth of the decrease in production possibilities during a pandemic.
While the initial response to the COVID-19 pandemic in many countries (especially in Europe and North America) was varying degrees of blanket isolation, there would come a time when those isolation policies needed to be relaxed. Because HAVE testing was either non-existent for most or otherwise imperfect, the only safe individuals to be removed from isolation would be those who were known to have COVID-19 and recovered. Because COVID-19 can be asymptomatic, even if a large share of the population did have the virus at one point, even they may not be sure they are now immune. More broadly, even if they suspect they are immune, there would no easy way to communicate to others that they were safe.
Perhaps no example better illustrates the desire for certification than what happened during the yellow fever plague that hit New Orleans in the 1800s. In 1853 alone, one in ten died. The only known defence was ‘acclimation’ – to contract the disease and not die from it. You had a 50:50 chance of that last step. Historian Kathryn Olivarius documented that despite this, the city (and its region) managed to grow. She recounts the experience of a German immigrant Vincent Nolte:
Nolte cherished one form of capital above all. In 1806, three months after his arrival in New Orleans, he was bitten by a tiny mosquito and fell sick with yellow fever, the most terrifying disease in the Atlantic World.... Nolte survived his “acclimation.” And now what had made him sick made him strong. He possessed “immunocapital”: socially acknowledged lifelong immunity to a highly lethal virus, providing access to previously inaccessible realms of economic, political, and social power.13
In New Orleans, an acclimation certificate was a key asset that determined whether you could engage in economic activity. Indeed, it was so valuable that many immigrants arriving actively tried to get sick as this would be a ticket to economic prosperity and marriage if they survived.14 Unfortunately, without a test, it turned out that the best way to become certified would be to prove proof that you lived in a yellow fever affected area for more than two years.
Contrary to the options available in the nineteenth century, in order to make the labor market safe again at some point, most countries will likely need to deploy HAD tests widely. Those who are found to have the requisite anti-bodies can then be certified safe. Obviously, this will require careful recording and verification of HAVE test results as well. Then some method of identifying the safe individuals will need to be devised. All this is within the realm of our current institutions and technology but setting up the apparatus will likely be costly and require some time.15 Indeed, one could imagine innovative ways of rationing access to such tests when they are scarce – say, by testing in conjunction with blood donations, thereby, encouraging that activity as people try to establish their immunocapital.
The question that will arise is what to do with people who do not test positive for HAVE or HAD. One option is for them to remain isolated but the difficulty here is that there is no obvious end date for that policy. Another policy would be to have guidelines and other preventative measures imposed on those people that limit their interactions with other people who have negative HAVE tests because if one of those people does end up contracting the disease, they would be able to transmit it to other people who have not yet had it. Overall, the right policy will depend on the proportion of people who test negative. The fewer who test negative, the safer are those people as they return to normal economic life because their chances of interacting with other non-immune people are reduced. Moreover, the tests can assist in certifying people for interactions with high-risk others (such as older people) or in health care.
Nonetheless, the downside and potentially unavoidable consequence of moving to a testing economy in this way is that it will reduce social cohesion. Just as the beef producers who worried that having some producers become certified as BSE safe would cause producers who were not certified to be seen as unsafe, we should be concerned that not being certified safe might become stigmatized with all of the costs that entails.
The discussion here thus far has glossed over an important issue with any kind of test — that it is imperfect. In particular, a test conducted on a person who has COVID-19 can return negative — this is a false negative — while a test conducted on a person who doesn’t have the virus can be returned positive — a false positive. This impacts on policies regarding what to do contingent on a test result.
Recall that with a HAVE test, what we want to do is isolate those who test positive and release those who test negative. We are doing this to prevent having to isolate everyone. Thus, if a person has a false positive, relative to the fact that our plan was to isolate that person anyhow, the fact that we choose to isolate them impacts them but not by much relative to the alternative. By contrast, if a person has a false negative, our plan is to release that person from isolation. In that case, however, we are putting someone we wanted to isolate into the population. Suffice it to say, that is costly.
But is it so costly that we should not use a ‘test then release’ strategy? Typically, there is a trade-off between false positive and false negative rates with one rising while the other falls. Often this is because a test is not just a test for one factor but for the presence of multiple factors. So if your test involves looking for the presence of, say, three factors, then you might choose to conclude that the test is positive only if all three factors are present. That means that, given this approach, you are less likely to have a false positive test but more likely to have a false negative test. This is why for many COVID-19 tests there was a reported false negative rate of between 10 and 15 percent (in line with other viruses) but a false positive rate of only 1 percent.16
The reason many tests appear to err on the side of minimizing false positives is because anti-viral treatments might be harmful, and you do not want to use them on people who do not have a particular virus. By contrast, a false negative test can be followed up with a future test for that patient that may reverse the finding. In other words, you want to be confident that you are treating the right person and if you have the option to continue observation and test, so you are comfortable perhaps initially missing a treatable person.
This weight of characteristics changes when you are dealing with a different decision — whether to release a potentially infectious person. In that case, you would want to err on the side of minimizing false negatives. If you want to release someone who has tested negative, you may not have an option to re-test them before they do more harm. By contrast, if someone tests positive falsely, you can keep them isolated and then re-test them later. This same logic applies to both HAVE and HAD tests but is stronger for HAD tests as the goal is not to re-test using that regime. By contrast, a HAVE regime would involve continued testing of people who returned negative results in the past.
This suggests that our medical practices will need to be informed by the decisions that have to be made — treatment versus release — to an extent that we haven’t done to this date. Of course, it goes without saying that tests that can reduce both false positives and false negatives will be more valuable as well. Interestingly, however, our tolerance for tests with errors may be greater than would be apparent at first. For instance, Nobel laureate economist, Paul Romer, conducted simulations of the movement of infectious diseases like COVID-19 through the population and compared the use of a blanket isolation strategy versus a test and release strategy even when tests had high degrees of false negatives.17 His analysis suggested that even tests with a false negative rate of 20 percent or more, could lead to two or three times fewer people eventually infected than a no isolation approach but also involves fewer people required to be in isolation when even imperfect tests are used.
It is not hard to see why targeting the isolation based on test results reduces the total number of people in isolation. What matters for controlling the infection is how many infectious people it isolates. If people are isolated at random, you have to isolate a lot more to get the same number of people who are infectious.18
The good news here is that, while we may want to calibrate test efficacy for the decision made, there is substantial room for error to still have a substantive impact. In other words, a more perfect test is better but not that much better than an imperfect one.19
Everything in this book thus far, as well as policy discussions regarding COVID-19, has been based on a very important assumption: once you have got the virus and recovered, you are immune. It is for this reason that epidemiologists focus on a sufficient share of the population obtaining immunity from COVID-19 either by past infection or as a result of a future vaccine. If you do this, then even with normal physical interactions, the virus eventually dies out (as R0 becomes less than 1). In particular, this is why we can talk about HAVE and HAD testing as making people safe again. So, while the crisis is awful, the promise of immunity gives us hope.
What if that hope is unfounded? What if you are not immune even if you have contracted the virus? What if a vaccine is not possible for the same reason? In this case, epidemiologists no longer use the SIR (susceptible-infected-recovered) model as there are no recovered people who are not able to infect others. Instead, we must use the SIS (susceptible-infected-susceptible) model. In that situation, when R0 exceeds one, the virus never goes away and a share of the population is always infected.20 The only way to get rid of the virus is by extreme measures – for instance, socially distancing until there are no more infected people or by coming up with treatments such that we don’t care if people are infected or not.21
Is this outcome possible for COVID-19? Because the virus is relatively new, at the time of writing, it is hard to be sure. Scientists were optimistic that, because recovering from COVID-19 required antibodies that such antibodies would give immunity for some period of time. However, in April 2020, South Korea reported 111 of coronavirus patients testing positive again after they had recovered (and tested negative twice in a twenty-four hour period).22 One possibility is that the negative tests were false negatives. Another is that the virus has reactivated. This is a virus that is latent for a time and but remains inside the cells of the host. This happens with chickenpox which can decades later reactivate in adults as shingles. Finally, there could be reinfection. This is why the flu is persistent. The anti-bodies only provide immunity for a time and not against alternative strains of the virus. Coronaviruses are a relatively recent phenomenon, so a lack of immunity remains a possibility.
Let’s take that worst-case scenario and presume that infected patients are not permanently immune. One implication is that HAD tests are of little value. Similarly, a vaccine will not be our savior. Nonetheless, HAVE testing could be of value. While the intensity of testing would have to be even higher than what might otherwise be envisaged, the procedure of isolating those who test positive will reduce the ability of infected people to spread the virus around. In this situation, so long as this results in the rate at which those are being infected falling below the rate at which people are recovering for an instance of the infection, the pandemic can be contained, and the virus will eventually be wiped out.
This highlights another reason to invest in the testing economy. When it comes to HAVE tests, these are valuable whether the virus leads to immunity or not, or something in between. As a policy, they are a hedge against this uncertainty.
Some countries moved to a testing economy very early in the COVID-19 pandemic. For instance, Taiwan started testing travelers from Wuhan for symptoms on the 31st December 2019 and soon after integrated travel histories with national data sets and made them available to hospitals.23 But Taiwan has special characteristics that make their response somewhat atypical (e.g., they are an island with a tighter relationship with China). More instructive in terms of seeing what a testing economy can achieve is to compare the Lombardy and Veneto regions of Italy.
Both regions applied social distancing and locked down retail areas. But only Veneto put in place a testing regime: testing both symptomatic and asymptomatic cases, testing contacts if someone tested positive, having testing carried out in-home and general measures to protect health care professionals.24 The result was that, as of March 26, Veneto (with a population of 5 million) had 7,000 cases and 287 deaths while Lombardy (with a population of 10 million) had five times the number of cases and 5,000 deaths.
The testing economy is what emerges when you have the virus under control, but you do not have widespread immunity either via past infections or a vaccine. This means that tests, like 9-11 security measures, will likely be a part of our daily lives for many years to come lest we end up more like Lombardy than Veneto.