Testing for testing times

Testing for testing times 1

COVID-19 testing for testing times: Fostering economic recovery and preparing for the second wave

There is light at the end of the COVID Crisis tunnel. We will get a vaccine against this new coronavirus in 12-18 months, the scientists assure us. Then we can fight COVID-19 without wrecking our economy; we can use mass vaccination in lieu of mass lockdowns. 

But for the moment, extreme containment policies are all we in the West have to calm calamity at the hospitals and reduce the necessity of witnessing unnecessary deaths. Things are different in parts of Asia. Having learned lessons from the SARS pandemic, several nations were fast to start testing massively – allowing them to isolate the sick from the healthy rather than everybody from everybody. That is no longer an option in the US and Europe – at least for the current wave of COVID-19. 

In this column, I argue that ramping up testing capacities is a priority that is almost as urgent as shielding our firms, jobs, financial system, and productive networks from persistent harm. The ‘big bazooka’ packages should include a hefty slice for COVID-19 testing. 

My argument requires a bit of background – at least for readers who missed the “epidemiology that all economists should know” part of the introduction to the eBook Beatrice Weder di Mauro and I edited three weeks ago, Economics in the Time of COVID-19 (Baldwin and Weder di Mauro 2020a).  

Epidemics for want-to-be curve flatteners

Epidemiologist have really cool models of how diseases spread. They look a lot like what economists use to model the diffusion of innovations (actually it’s the other way around, but never mind). The most famous epidemiological model is the ‘SIR model’ developed in 1927. 

The SIR model classes people into Susceptible (S), Infectious (I), and Recovered (R), hence SIR. 

  • People in ‘S’ can get it, people in ‘I’ can give it, and people in ‘R’ can do neither. 
  • When susceptible and infected mingle, the probability of transmission per period per interaction is ‘beta’.
  • The rate of recovery per period is ‘r’ for infected people. 

Taking S and I to be the number of people in the categories, the number of possible contacts between sick and susceptible is S times I – this is plotted in Figure 1 as the bell-shaped curve.

Figure 1 Source of the hump in the epidemiological curve.

Testing for testing times 2

Source: Author’s elaboration.

Because the number of possible infections (S times I) is hump-shaped (Figure 1), the simplest epidemiological curve (Figure 2) – which plots the number of new cases per day against time – is also hump-shaped. There are also more complex reasons. For instance, sick people recover and, in an actual epidemic, healthy people cut back on interactions once the epidemic is in its acceleration phase. 

Figure 2 The classic epi curve (the daily number of new cases per day) with phases

Testing for testing times 3

Source: Author’s elaboration of CDC.gov diagrams. 

Figure 1 helps us understand the acceleration phase of the epi curve shown in Figure 2. Acceleration happens when I is small and growing, and S is big. The deceleration phase is when S is small and I is shrinking due to recovery and lack of new ‘victims’. 

This is for a single infection incident. Many epidemics come in waves, and many epidemiologists predict that COVID-19 will be one of them (Ferguson et al .2020). More on this below.

Many analysts today use a natural variant of the epi curve – the cumulative number of cases – which looks like a logistics curve. One well-known presentation is published daily by the Financial Times. This cuts off the early phases and plots only the acceleration, deceleration, and resolution phases (Figure 3). 

Figure 3 The Financial Times’ trajectory curves for various nations, 24 March 2020.

Testing for testing times 4

Source: https://www.ft.com/coronavirus-latest 

The cumulative number of cases tends towards a limit, as shown in Figure 4. This limit is where the number of infections (which is low since so many people are immune) falls to the number of recoveries.  

Figure 4 The trajectory of cumulative cases with phases.

Testing for testing times 5

Source: Author’s elaboration. 

What determines the limit? How many people get the disease in the long run if there is no vaccine? The steady-state stock of never-infected people, i.e. S, is S’, where S’ = exp[(1-R0)S’] and R0 (R-nought) is the number of people who catch it from an average infected person. 

If R0 is two, about 80% of the world gets it in steady state. The current estimate for COVID-19 is around 2, but it has varied in different outbreaks.

If R0 is two, about 80% of the world gets it in steady state. The current estimate for COVID-19 is around 2, but it has varied in different outbreaks (Cereda 2020). Dr Syra Madad, who runs preparedness efforts for NYC Health and Hospitals, said: “I think that it is certainly plausible that 40% to 70% of the world’s population could become infected with coronavirus disease.”1 

Two last critical points: 

  • The estimate of R0 is critical to economic thinking about the COVID-19 Crisis since it explains how fast the number of cases blows up during the acceleration phase.
  • A vaccine for COVID-19 will be a game-changer since then people won’t have to get the disease to become immune to it. 

The steady-state calculations above assume no vaccine.

Why Europe and the US are trying to flatten the epi curve

COVID-19 is lethal (at least ten times deadlier than the regular flu), but the good news is that it is not that lethal by pandemic standards (SARS was ten times worse).  The virus, by contrast, is very contagious. It’s R0 is high enough to make its progression explosive in the acceleration phase. That makes this pandemic quite unlike other 21st century epidemics (see Annex 1). It all comes down to the R0.

Small differences in R0 matter hugely. After ten transmission rounds, for example, one infected person will have infected R0 raised to the power of ten. If R0 is 2 then after ten rounds of transmission, each infected person gives it to 1,024 others; when R0 is 3, the figure is 59,049. Of course, each newly infected sets off their own path, so explosive is not an understatement when the basic reproduction rate, R0, is large. 

Small differences in R0 matter hugely … If R0 is 2, then after ten rounds of transmission, each infected person gives it to 1,024 others; when R0 is 3, the figure is 59,049.

The only epidemiological study to date on Western outbreaks, Cereda et al. (2020), estimates the basic reproduction number, R0, to be 3.1, and the length of each transmission round as 7.5 days (technically called the ‘serial interval’, i.e. the time between successive cases in a chain of transmission). The actual transmissibility of the disease – the estimated net reproduction number R(t) – varies in the field by cluster and across time. Figure 5 shows the estimates for one of Italy’s worst outbreaks, in Bergamo. 

Figure 5 Estimated infectiousness (net reproduction number Rt)

Testing for testing times 6

Source: Figure S10, Cereda et al. (2020), “The early phase of the COVID-19 outbreak in Lombardy, Italy”.

While only some COVID-19 patients require intensive care treatment in a hospital, the explosive growth of the disease can overwhelm even very well-prepared medical systems. That happened in Italy, where 18% of the detected cases required intensive care (Cereda et al. 2020). Hospitals in Spain are starting to see similar eruptions in admissions. 

Overwhelmed hospitals are a human calamity. Hospitals that are unable to provide the care that people need to survive will see more people dying than would be the case with a slower rate of infection. This is why epi curve flattening (slowing the rate of infection) is a life-and-death imperative.

A recent paper by economists using epidemiological models (and some bold assumptions) simulates deaths with and without curve-flattening policies (Alvarez et al. 2020). This is to be taken as an illustration, but it brings home, and forcefully, what is meant by “life-and-death imperative” when it comes to flattening COVID-19’s epi curve. The difference with and without curve-flatten is more than 2% of the population. For the US, that is over 6 million avoidable deaths – to say nothing of the sort of civil unrest and violence that calamity at hospitals on this scale could provoke (Baldwin 2020c).

Figure 6 Hypothetical simulation of unnecessary death without containment policies

Testing for testing times 7

Source: Figure 1, Alvarez et al (2020), simulated impact of a partially effective lockdown.

Epi curve-flattening possibilities

There are no 21st century tools against COVID-19 – no vaccine and no cure. The only way to slow the rate of infection (flatten the epi curve) is to separate the infected and uninfected. There are, in turn, two ways of doing that. 

  • The Singapore model: separate the sick (test, track, and trust). 

The 2003 SARS epidemic traumatised East Asia. While Singapore was not particularly hard hit, SARS woke up the Singaporean government and they put in place plans for the next one. Those plans involved rapid response, extensive testing, and isolation of infected people and people they might have given it to. The key to all of that was testing. Singapore was able to determine who was infected. Tracking (of the infected and their contacts), transparency (by the government about what measures were being taken), and trust (by the people of the government and each other to obey the rules) were also essential (Straits Times 2020, Singapore government 2020).  

  • The Italian model: separate everybody (lockdown, school closures, etc.)

Given the lack of testing in Europe and the US, governments one after another have turned to the only option they had to calm calamity in the hospitals (actual or looming). Such policies are like using the ‘off’ switch to shut down a nuclear reaction, but it’s the economy, not a reactor. The result is a massive fallout – economic fallout. Given the lack of testing facilities, the recession in the West is really a public health measure (Baldwin 2020b). Since hospitals are projected to be overloaded despite the lockdowns, the result may be a COVID-19 upheaval scenario, as I called it in my 15 March column (Baldwin 2020c).

Figure 7 The two-curve problem: Flattening the epi curve worsens the recession curve

Testing for testing times 8

Source: Author’s elaboration based on Gourinchas (2020).

The epi flattening produces a recession, or what I’ve called the ‘two-curve problem’ (Baldwin 2020d); see Figure 7. The dilemma – flattening the epi curve worsens the recession curve – is solved with economic policies that ‘shield’ jobs, firms, banks and networks. The goal is to ensure that the economy that workers will return to once the medical crisis passes has been shielded from unnecessary damage. The policy solutions to the two-curve problem are symmetric. We flatten the epi curve to avoid unnecessary deaths while waiting for a vaccine; we shield the economy to avoid unnecessary economic scarring while waiting for a vaccine. 

The policy solutions to the two-curve problem are symmetric. We flatten the epi curve to avoid unnecessary deaths while waiting for a vaccine; we shield the economy to avoid unnecessary economic scarring while waiting for a vaccine.

The consensus among the authors in the second eBook that Beatrice Weder di Mauro and I edited last week is that economic policymakers need to: “Act fast and do whatever it takes” (Baldwin and Weder di Mauro 2020b). Governments should deploy policies that ‘flatten the recession curve’ while avoiding long-lasting damage to our economies. Governments should do whatever it takes to ‘keep the lights on’ until the recession is over, and especially shield the most vulnerable from dire hardship. 

But it’s not that simple. COVID-19 and the containment policies have reduced the flow of labour to businesses and this has massively reduced output, i.e. the production side of GDP. The ‘shield packages’ are attempting to maintain the expenditure side of GDP (Baldwin 2020d). 

That brings us to testing and economic recovery.

The two-test problem: Getting workers back to work without getting them sick

As Dewatripont et al. (2020) put it: “restarting economic activity as quickly as possible is crucial but … requires the reliable identification of individuals who will not contract the virus or transmit it to others.” And even now some production must go on. 

Since the production and delivery of goods requires social clustering (in factories, logistic chains, etc.), we must either provide medical protection to the workers involved (rearranging work such that transmission is less likely, providing protective kit, etc.) or figure out which workers are immune. 

This latter option – identifying which workers are immune and not infectious – requires two tests (Dewatripont et al. 2020): 

Serology (blood tests that look for antibodies) can be used to find evidence that the person is not in the S class. If the anti-bodies are found, then the person is either infectious or recovered. Technically, these are known as ELISA tests for SARS-Cov-2 specific antibodies.

This test, technically known as a RT-PCR4 test, looks for the virus’s RNA. If you don’t have the RNA in your blood, you are either an S (susceptible) or an R (recovered). 

If you do have the anti-bodies but don’t have the RNA, then you are ‘recovered’ and thus can neither give it nor get it. These are the workers we will find really useful in keeping the essential goods flowing. 

If you do have the anti-bodies but don’t have the RNA, then you are ‘recovered’ and thus can neither give it nor get it. These are the workers we will find really useful in keeping the essential goods flowing.

This could be called the two-test problem, or the ‘two-test test’. 

Testing to avoid having to use recession as a public health measure in the 2nd wave

If a second wave of COVID-19 comes, say in the autumn of 2020, then we will have to induce a second recession, or a double-dip recession, unless we have the Singapore option to hand. That, as mentioned, requires massive and agile testing capacities, a government that is willing to face up to reality quickly, and a population that trusts that isolated sick people will actually remain isolated until they are no longer infectious. 

The trust part is hard to change quickly, and the governments of today will almost surely still be in power in the autumn, but we can remedy the testing limitations the US and Europe faced this spring. We should do that, and urgently. 

The harsh reality of this disease is that we have to flatten the epi curve to save what could be an unthinkable number of lives. The only way to flatten the epi curve is to reduce the mingling of the sick and the healthy. If we don’t know who the sick are, we must shut down the economy to diminish the intermingling. Neither option is pleasant, but distancing everyone from anyone causes a much larger recession than quarantining the infectious. 

This medical shock will pass. The recession will end. Acquiring rapid testing capacity will help hold down the cumulative death toll and the cumulative economic damage. 

Annex 1: Recent history of pandemics (from Baldwin and Weder di Mauro 2020a)

The 20th century witnessed two pandemics since the historic ‘Spanish Influenza’ of 1918, namely, the ‘Asian flu’ of 1957 and the ‘Hong Kong flu’ of 1968. The 21st century has seen four pandemic outbreaks: N1H1 in 2009 (the ‘bird flu’), Severe Acute Respiratory Syndrome (SARS) in 2002, Middle East Respiratory Syndrome (MERS) in 2012, Ebola which peaked in 2013-14. This box reviews the timelines and mortality of these epidemics.

Asian flu (H2N2): The Asian influenza originated in the Chinese province of Yunnan at the beginning of 1957. The disease reached Singapore in February 1957 and spread to Hong Kong in April 1957. It then spread in the Southern Hemisphere, reaching India, Australia and Indonesia in May, before arriving in Pakistan, Europe, North America and the Middle East in June. South Africa and South America, New Zealand and the Pacific Islands were affected from July, while Central, West and East Africa, Eastern Europe and the Caribbean were reached in August.  This first wave peaked towards the end of 1957 and affected mostly school children, young adults and pregnant women. A second wave arrived in 1958, hitting several regions including in Europe, North America and Japan, with this one tending towards affecting the elderly. 

The estimated number of deaths is not precise, but the consensus figure is around 1.1 million deaths worldwide.  Estimates for the mortality rate (deaths as a share of cases) are likewise imprecise but range between 1 in 4,000 and less than 0.2%. National death estimates are not widely available, but in the US, it was between 80,000 and 110,000; in England and Wales, estimates put it around 6,000.  

Hong Kong flu (H3N2): The Hong Kong influenza was recorded for the first time in Hong Kong on 13 July 1968; 500,000 Hong Kong residents were infected in the first six months (15% of the population).  By the end of July, the outbreak reached Vietnam, Singapore, and started spreading globally, reaching India, the Philippines, Australia, and Europe by September 1968. It entered California via troops returning from the Vietnam War. It ultimately led to about 33,800 American deaths.  The disease reached Japan, Africa and South America by 1969 (Starling 2006). According to the CDC, H3H2 kill about a one million people worldwide, most of them over 65 years old. According to the U.S. Department of Health and Human Services, the virus peaked worldwide in December 1968. 

2009 Avian flu (N1H1): In 2009, a new pandemic flu emerged – the first in 40 years. The first case was detected in California in April 2009; it was declared over by the World Health Organization (WHO) in July 2010. A detailed timeline is provide by the European Centre for Disease Prevention and Control (ECDC).  After the first case was detected in California, it was recognized in Mexico only a few days later. Two days after that, it reached Europe with the first reported cases in Spain and Britain. The WHO Director General announced a world pandemic state on the 11 June 2009 about two months after the first case. 

The CDC estimates that between 151,700 and 575,400 people worldwide (0.001-0.007% of the world population). The total number of cases in 2009 was the highest in the US, Mexico, Canada, and the UK. The number of deaths was the highest in Mexico and the US. 

Severe Acute Respiratory Syndrome (SARS): The SARS is a viral disease originated by the SARS-coronavirus at the end of 2002 in China; the WHO was informed about the outbreak in February 2003. By the end of March 2003, 210 suspect and probable cases of SARS are reported around the world, starting from Toronto.  Between November 2002 and July 2003 8,096 cases were reported with 774 of these leading to death. SARS had a high mortality rate of 9.6%, but it was far less contagious than previous pandemics. Most cases were in China (5,327) and Hong Kong (1,755) where the fatality rates were respectively 7% and 17%; Taiwan and Canada were the next hardest hit with 346 and 251 case and mortality rates of 11% and 17% respectively. 

Middle East Respiratory Syndrome (MERS): The MERS is a viral respiratory disease caused by a coronavirus (MERS‐CoV) which has been found in dromedary camels in several countries.  The first outbreak was identified in Saudi Arabia in 2012 and subsequently spread to 27 countries: Algeria, Austria, Bahrain, China, Egypt, France, Germany, Greece, Iran, Italy, Jordan, Kuwait, Lebanon, Malaysia, the Netherlands, Oman, Philippines, Qatar, Korea, Thailand, Tunisia, Turkey, United Arab Emirates, UK, the US, and Yemen. However, it was highly concentrated in Saudi Arab (more than 80% of the cases). All cases identified outside the Middle East were are people who were infected in the Middle East. The disease is highly lethal, with the WHO estimating that about 35% of reported patients died. 

Ebola Virus Disease (EVD): The EVD is a fatal illness in human, whose average fatality rate is around 50%, ranging from 25% to 90% according to the waves of outbreak (see WHO.int for details). The first outbreak was identified in 1976 in in the Democratic Republic of Congo and in Sudan, where the mortality rate was respectively of 88% and 53%, with approximately 300 cases in both states. The second wave was in 2014-2016, starting in West Africa, and it was the largest one since its discovery in 1976, both in terms of cases and deaths. This outbreak spread across states starting in Guinea with 3,811 cases and a mortality rate of 67%, then moving to Sierra Leone, with 14,124 cases and a mortality rate of 28%, and Liberia, with 10,675 cases and a mortality rate of 45%. The most recent outbreak of 2018-2019 started in the eastern Democratic Republic of Congo, and as of now there are 54 cases with a mortality rate of 61%. 


Alvarez, F, D Argente, and F Lippi (2020). “A Simple Planning Problem for COVID-19 Lockdown,” manuscript, 24 March 2020.

Atkeson, A (2020). “What will be the economic impact of covid-19 in the us? Rough estimates of disease scenarios,” NBER WP 26867, March 2020

Baldwin, R (2020b), “Keeping the lights on: Economic medicine for a medical shock”, VoxEU.org, 13 March. 

Baldwin, R (2020c), “The COVID-19 upheaval scenario: Inequality and pandemic make an explosive mix”, VoxEU.org, 15 March.

Baldwin, R (2020d). “The supply side matters: Guns versus butter, COVID-style,” 22 March 2020.

Baldwin, R and B Weder di Mauro (2020a), Economics in the Time of COVID-19, a VoxEU.org eBook, CEPR Press.

Baldwin, R and B Weder di Mauro (2020b), Mitigating the COVID economic crisis: Act fast and do whatever it takes, a VoxEU.org eBook, CEPR Press.

Cereda D et al. (2020). “The early phase of the COVID-19 outbreak in Lombardy, Italy” https://arxiv.org/abs/2003.09320, pre-print on ArXic.org. 

Dewatripont, M, M Goldman, E Muraille and J-P Platteau (2020). “Rapidly identifying workers who are immune to COVID-19 and virus-free is a priority for restarting the economy,” VoxEU.org, 23 March 2020. 

Ferguson, N M, D Laydon, G Nefjati-Gelani et al. (2020), “Impact of non-pharmaceutical interventions (NPIs) to reduce COVID-19 mortality and healthcare demand”, Imperial College COVID-19 Response Team.

Gourinchas, Pierre-Olivier (2020). “Flattening the pandemic and recession curves,” Chapter 2 in R Baldwin and B Weder di Mauro (eds), Mitigating the COVID economic crisis: Act fast and do whatever it takes, a VoxEU.org eBook, CEPR Press.

Singapore government (2020). “Advisories on COVID-19,”

Straits Times (2020). “Singapore’s strategy in fighting Covid-19,” 24 March 2020.

Leave a Reply

Your email address will not be published.