Managing the economic and health costs of the COVID-19 pandemic will require inputs from different disciplines. Epidemiology models in the tradition of Kermack and McKendrick (1927) will certainly play an important role in the process. But those models won’t suffice because they generally don’t allow for interactions between peoples’ economic decisions and rates of infection. The absence of these interactions limits their usefulness for forecasting and policy analysis. In three papers (Eichenbaum et al. 2020a, 2020b, 2020c), we propose a framework for combining economics and epidemiology in a way that allows for these interactions.
In this framework, an epidemic gives rise to negative shifts in aggregate supply and aggregate demand. The supply effect arises because an epidemic exposes people who are working to the virus; people react to that risk by reducing their labour supply. The demand effect arises because consumption activities exposes people to the virus; people react to that risk by reducing their consumption. Working in tandem, these supply and demand effects generate a large, persistent recession. This recession arises even if governments didn’t institute any containment policies.1 But our framework implies that the government should undertake containment policies which improve health outcomes even though those policies worsen economic outcomes during the epidemic. The challenge is designing policies which improve the trade-off between health and economic outcomes.
Our framework builds on the classic epidemiology model proposed by Kermack and McKendrick (1927). The population is composed of four groups who differ by their health status: (i) susceptible (people who haven’t been exposed to the disease), (ii) infected (people who have contracted the disease), (iii) recovered (people who have survived the disease and have acquired immunity), and (iv) deceased (people who have died from the disease). In Kermack and McKendrick (1927), transition probabilities between states are exogenous.
In contrast, transition probabilities in our three papers reflect, in part, people’s choices regarding market activities. New infections arise from consumption-based activities. The number of these infections is proportional to the total amount of time that infected and susceptible people spend on market-based consumption activities. New infections can also arise from working. The number of these infections is proportional to the total amount of time that infected and susceptible people spend in work-related activities. Finally, new infections also arise from susceptible people interacting with infected people in non-economic based activities. Other things equal, the probability of a person getting infected in any of the ways just discussed is an increasing function of the fraction of the population that is infected.
People understand the health risks associated with market activity so they reduce their consumption and hours worked during an epidemic. In making their individual decisions, people take the aggregate number of infections as given. For any given person, this assumption is rational. But in the aggregate, it means that there is an externality associated with a virus, analogous to the one associated with pollution. The classic solution to this type of externality is to impose a Pigouvain tax on polluting activities. So, too, in our model it is socially optimal to use containment measures to reduce market activities during an epidemic. The case for externalities is even stronger when the health system can become overwhelmed during an epidemic. In Eichenbaum et al. (2020a), we capture this effect by assuming that the death rate from the virus is a convex, increasing function of how many people are infected. This ‘overcrowding’ problem magnifies the externalities associated with infections.
Figure 1 displays the impact of an epidemic in the model economy of Eichenbaum et al. (2020a). Here we adopt the calibration used in Eichenbaum et al. (2020b) to ease comparability across the two papers. The solid blue line depicts the aggregate dynamics absent any government intervention. Note that an epidemic generates a large number of deaths (totalling roughly 0.16% of the initial population) and a large recession, with an average peak-to-trough decline of about 7% of real GDP.2
Figure 1 Epidemic dynamics
Source: Eichenbaum et al. (2020a) based on calibration in Eichenbaum et al. (2020b)
What policies should a welfare-maximising government pursue to deal with the infection externality? Suppose that the government can’t treat people differently depending on their health status. In Eichenbaum et al. (2020a), we investigate the nature of the optimal containment rate on everyone’s consumption. The dotted line in Figure 1 depicts the solution to what we call the optimal simple-containment policy.3 The initial jump in the containment rate reflects the interaction between the dangers of overwhelming the medical system with the possibility that vaccines and treatments will, at some point in the future, become available. After the initial jump, the containment rate mirrors the infection itself because the size of the externality reflects the fraction of the population that is infected.4
The optimal simple-containment policy makes the recession worse than the no-intervention equilibrium. But the policy improves welfare because it saves an enormous number of lives. The political pressure to abandon policies like simple containment because of their large economic costs have, in many cases, proven unbearable. Numerous countries, including the US, prematurely abandoned initial containment measures. We analyse the results of doing so. Consistent with the evidence, abandonment leads to a short-lived economic revival followed by a surge in infections, epidemic-related deaths and a subsequent second recession.
The last set of results makes clear the need to find policies that improve the trade-off between health and economic outcomes. A natural class of such policies involves testing people for their health status. In Eichenbaum et al. (2020b), we modify the framework in Eichenbaum et al. (2020a) to take into account that neither people nor the government know the health status of any given individual.
In our framework, testing without quarantines actually worsens the economic and health consequences of an epidemic. The reason is as follows. People who know they are infected, reduce their economic activity by less than people who don’t know their health status. There’s just less to lose from consuming and working once you know you are infected compared to someone who is susceptible to the virus. Testing people reveals to them whether they are infected. With more ‘known’ infected people engaging in economic activity, social interactions become riskier for people who don’t know their health status. The latter respond by cutting back on their economic activity. The net result in our model is a deeper recession and more deaths compared to a no-testing scenario.
Suppose now that test results are used to implement the following simple-quarantine policy: infected people aren’t allowed to work and receive consumption from the government but they are allowed to engage in non-economic social interactions. We refer to these policies as ‘smart containment’. A strict-quarantine policy goes one step farther and restricts infected people from non-economic social interactions. We refer to these policies as ‘strict containment’.
Smart- and strict-containment policies would be costly to implement. But they dramatically improve the trade-off between economic activity and health outcomes. Smart- and strict-containment improve health outcomes because they greatly reduce the interactions between infected and susceptible people. But these policies also lead to much smaller recessions than a no-intervention policy. Smart- and strict-containment policies isolate infected people from social interactions related to consuming and working. The resulting reduction in the risk of being infected leads to higher consumption and work by everyone at risk of becoming infected compared to the non-intervention policy. So, testing and quarantine policies are a win-win from an economic and a health perspective. For that reason, they are far more likely to be politically sustainable.
Standard epidemiology models abstract from the way in which people’s economic behaviour changes in response to policies aimed at curbing infections. There are at least two reasons to worry about this shortcoming. First, it limits the usefulness of these models for predicting the results of changes in health policy. This point is a classic application of the Lucas (1976) critique. The statistical ‘fit’ of an epidemiology model to data generated under one set of public health policies is not a reliable guide to how that model will predict outcomes under a different set of health policies. Second, policies designed to deal with the health consequences of the COVID-19 epidemic inevitably involve some curtailment of economic activity. These considerations mean that it is important to integrate economic and epidemiology models. In Eichenbaum et al. (2020a, 2020b and 2020c), we propose a framework for doing so.
We use this model to address the challenge of designing and implementing policies that improve the trade-off between economic and health outcomes during an epidemic. The results suggest that testing and quarantine policies should play a central role in minimising the social costs of the COVID-19 crisis.
Chetty, R, J N Friedman, N Hendren, and M Stepner, (2020), “How Did COVID-19 and Stabilization Policies Affect Spending and Employment? A New Real-Time Economic Tracker Based on Private Sector Data”, NBER Working Paper No. 27431.
Eichenbaum, M, S Rebelo, and M Trabandt, (2020a), “The Macroeconomics of Epidemics,” NBER Working Paper No. 26882.
Eichenbaum, M, S Rebelo, and M Trabandt, (2020b), “The Macroeconomics of Testing and Quarantining,” NBER Working Paper No. 27104,
Eichenbaum, M, S Rebelo and M Trabandt, (2020c), “Epidemics in the Neoclassical and New Keynesian Models”, NBER Working Paper No. w27430.
Goolsbee, A and C Syverson (2020), “Fear, Lockdown, and Diversion: Comparing Drivers of Pandemic Economic Decline 2020”, NBER Working Paper 27432.
Lucas, R E (1976), “Econometric Policy Evaluation: A Critique,” Carnegie-Rochester Conference Series on Public Policy 1, 19-46.
1 See Chetty et al. (2020) and Goolsbee and Syverson (2020) for empirical evidence in favor of this property of our framework.
2 In Eichenbaum et al. (2020c), we show that these qualitative conclusions are robust to allowing capital accumulation and nominal rigidities.
3 The objective function in this problem is the weighted average of present value utility of people alive at the beginning of epidemic.
4 In Eichenbaum et al (2020a), we show that very similar results obtain if the government can simply dictate consumption and employment levels that don’t differ by health status.