Why testing a representative sample of the population must be done now

Andrea Galeotti, Paolo Surico 08 April 2020

a

A

In order to win a war, you need to know your enemy. We are fighting the war against Covid-19 with very little information, and the information we have is compromised by measurement errors. This makes it impossible for governments to formulate health and economic policies. Without such information, any action we take is like flipping a coin. 

To be more precise: how can governments formulate health policies to contain new and further spreading of the virus if social distancing measures are relaxed without knowing even basic information, such as the prevalence of infection in the population and, consequently, the immunity the population has built up? Italy is now approaching this second phase with, as far as we are aware, no plan for how to relax social distancing measures. Spain is no different from Italy in this regard, and we doubt the UK will be ready for the second phase unless action is taken now.

Similar considerations will apply as governments think about designing economic policies to allow the much needed re-start of economic activities. Understanding the extent to which the virus has already spread across different demographics will inform what sectors can be easily re-opened without further large disruption to the economy (Baldwin 2020). 

What we hear from the media and amongst medical practitioners across different countries is the same: test, test, test! In reality, many governments are struggling to test even key workers in the health system and related industries. At this moment, governments are operating under steep constraints; making proposals that ignore this fact is not helpful.

However, as we have repeated over the last four weeks (as a consequence of our analysis presented in these slides), in order to obtain reliable information about how widely the virus has spread, and how this spread differs across individual and household characteristics, we do not need to test everyone. Governments should now formulate a strategy based on selecting and testing a representative sample of the population (see also Dewatripont et al. 2020 and Grosso 2020). The cost of this strategy is minor, relative to the simplistic proposal to test everyone, but its value would be enormous, allowing us to estimate, with standard statistical methods, how much immunity a country has built up and how such immunity is distributed across individuals with different characteristics.

It is perhaps worth pointing out that the estimates on 30 March from the Imperial College Covid-19 Response Team on the prevalence of the virus are highly uninformative, since their range is enormous. They estimate that the percentage of the Italian population that has been infected is between 3.2% and 26%. The Imperial College Team is providing a remarkable service, but without good data to feed the model, any quantitative assessment is mere speculation. 

Our feasible proposal, articulated briefly in the following steps, is this: 

1. Determine a representative sample of the population. (In Italy, for example, one could start with the regions of Lombardy, Emilia-Romagna, and Veneto. In the UK, one could start with London.) Different statistical governmental institutions, together with research centres, will already have constructed such representative samples, as this is the first step of every statistical investigation. 

2. Test such a representative sample for Covid-19; for each individual in the sample, record at the same time their socio, economical, demographic, and locational characteristics at the household level.

3. Use standard statistical methods to infer prevalence of the contagion in the whole population and detect the household characteristics most likely to predict whether or not someone within the whole population is infected. For example, what is the age bracket that has been exposed more to the virus and so developed larger immunity? Are those people working in specific sectors of the economy? 

Why is testing a representative sample useful? Most likely, the existing social distancing measures will be relaxed, to start with, for only a subset of the population. By testing a representative sample, we will know, for example, the percentage of individuals between ages 20-40 that have acquired immunity. We can then decide whether it is worth focusing, in the second phase of testing, on this group of individuals, before relaxing social distancing measures for this group. Or, given the test results, whether it is better to focus the second phase of testing on another category of individuals. Testing an individual has an opportunity cost. Before we commence mass testing, we should understand better what those costs are. This is what testing a representative sample of the population now will accomplish.

We conclude by reiterating that this strategy is not only cost effective, it can be implemented rapidly. But time is running out. 

References

Baldwin, R (2020), “COVID-19 testing for testing times: Fostering economic recovery and preparing for the second wave”, VoxEU.org, 26 March.

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. 

Flaxman et al. (2020), “Estimating the number of infections and the impact of non-pharmaceutical interventions on Covid-19 in 11 European countries”, Imperial College Covid-19 Response Team, 30 March.

Gross, D (2020), “Creating an EU ‘Corona Panel’: Standardised European sample tests to uncover the true spread of the coronavirus”, Voxeu.org, 28 March.

Galeotti, A and P Surico (2020), “A user-guide to Covid-19”, Voxeu.org, 27 March.

a

A

Topics:  Covid-19

Tags:  COVID-19, coronavirus

Professor of Economics, London Business School

Professor, Economics Department, London Business School; Research Affiliate, CEPR

Events

CEPR Policy Research