Comment on "COVID-19 in segmented societies" by Constantino Hevia, Manuel Macera, and Pablo Andrés Neumeyer.

This comment briefly summarizes Hevia et al. (2022), who utilize a unique dataset to shed light on the inequality of the epidemiological burden for low- and high-income individuals in Bogotá. Some suggestions and ideas for future research are provided.

Introduction

The COVID-19 pandemic has had a major impact on health and economic outcomes in the aggregate—claiming more than 5 million lives globally to date and causing one of the largest economic contractions ever—with these impacts heterogeneous across individuals. The paper by Hevia et al. (2022) utilizes a unique dataset to shed light on the inequality of the epidemiological burden for low- and high-income individuals. Using CoVida—a large epidemiological survey conducted in Bogotá— Hevia et al. (2022) document disparities in infections across socioeconomic status: 58 percent of low-income individuals are estimated to have been infected (as of February 2021), compared with 29 percent of high-income individuals. They also find that, within each age group, the case fatality rate is higher in the low-income group relative to the high-income group, though unconditionally it is higher in the high-income group because of its higher average age. Thus, not only are low-income individuals more likely to get infected but they are also more likely to die from the disease, conditional on infection.

Motivated by these empirical findings, they develop a two-agent macro-SIR model that is used to study the distributional consequences of COVID-19 and mitigation/transfer policies. High-income workers have access to financial markets, higher human capital, and lower infection risk, ceteris paribus, compared to their low-income counterparts, who are assumed to live hand-to-mouth. Both agents derive utility from consumption (), leisure (), and social interaction (). The probability of infection depends on their own consumption , labor , and social interactions , and those of infected low-income and high-income individuals. Firms use labor and capital as inputs of production in a standard way. Shutdowns are modeled as a quantity restriction policy, . This is a small open economy environment where the world interest rate is taken as given, but importantly, the wage is determined in equilibrium.

The model is calibrated to match relevant features such as the dynamics of output and infections in Bogotà. Using the calibrated model, the paper studies the effects of shutdowns and redistributive transfer policies. The first main finding is that a shutdown reduces the welfare of both agents and has little impact on long-run health outcomes. Because the shutdown is modeled as a quantity restriction on output, it leads to a large decline in labor demand and in the real wage, increasing the welfare cost of both agents. The second main finding is that redistributive transfers (from high-income to low-income workers) improve the welfare of both agents relative to a shutdown only. This is because low-income workers benefit from the transfer—by allowing higher levels of consumption and leisure—and high-income workers benefit from higher wages that result from the reduced labor supply of low-income workers and inelastic labor demand due to a binding output constraint. Importantly though, they do not find any Pareto-improving policies relative to the Laissez-Faire baseline.

Comments regarding the calibration and model fit

Hevia et al. (2022) use the richness of their dataset to discipline many of the parameters in the model. There are, however, some parameter and modeling assumptions that may not be inconsequential. For example, how important is the assumption of symmetry in transmission? That is, to what extent does transmission occur within income groups relative to across income groups. Is it possible that high-income individuals shop at different shops, work at different workplaces, and socially interact with other high-income individuals? In a related paper that studies the unequal effects of the pandemic in slums and in non-slums in Brazil, Brotherhood et al. (2020) assume that roughly one-third of interactions is exclusively within-group (with the remainder being random interactions). Understanding better this within and across-group transmission would be an interesting avenue for future research. As another example, how important is the hand-to-mouth assumption for the low-income group? Presumably, the effect of the redistributive transfers on welfare and labor supply (and price dynamics) would be smaller if low-income agents have the ability to save.

The calibrated model generally replicates well the dynamics of aggregate economic and epidemiological variables of Bogotá. The model, however, underperforms with regard to the real wage. In the model, labor demand falls by more than labor supply, leading to a counterfactually large decline in real wages (more than 30 percent at the trough). Since the key results depend on these price dynamics, getting this right is important. I suspect that these dynamics are generated, in part, from the assumption that working less reduces transmission within the workplace but increases the transmission from consumption and social interaction. This is somewhat peculiar, as the probability of getting infected likely depends on, for example, the number of parties I attend, how much time I spend at them, and how many people are at the party, not directly on how much time I have overall. In the absence of this assumption, labor supply would decline more, relieving some of the downward pressure on the real wage. This would also bring labor supply closer to the data, as the decline in labor is too small currently, relative to data.

Final thoughts

This paper uses a large epidemiological surveillance study from Bogotá—which contains COVID-19 test results and indicators for socio-economic status—to document the heterogeneous transmission and case fatality rates across income groups. Disciplined by the empirical evidence, the authors build a two-agent economic-epidemiological model to study the aggregate and distributional consequences of the pandemic and government policies undertaken in Bogotá. To the extent that many countries exhibit similar stark inequalities, this paper’s policy implications are applicable to many other countries.

An interesting avenue for future research would involve using the model to study other types of policy instruments. While Hevia et al. (2022) mainly focus on shutdowns and redistributionary lumpsum transfers, policies that more directly address externalities may lead to better outcomes (for example, Pigouvian taxes as in Kaplan et al. (2020) and Hur (2021)).