Shutting down the workplace is an effective means of reducing contagion but can induce large economic losses. We harmonize the American Time Use Survey and O*NET data to construct a measure of infection risk (exposure index) and a measure of the ease with which a job can be performed remotely (work-from-home index) across both industries and occupations. The two indexes are negatively correlated but distinct, so the economic costs of containing a pandemic can be minimized by sending home only those workers that are highly exposed to infection risk but that can perform their jobs easily from home. Compared with a lockdown of all non-essential jobs, which includes many jobs not easily performed from home, a more selective policy can attain the same reduction in aggregate infection risk (32 percent) with one-third fewer workers sent home to work (24 percent vs. 36 percent) and only half the aggregate wage loss (15 percent vs. 30 percent). In addition, moving to such a policy reduces the infection risk of low-wage workers the most and the wage losses of high-wage workers the most. Our crosswalk between the American Time Use Survey and O*NET data can be applied to a broader set of topics.
Most countries have implemented lockdowns and social distancing to varying degrees to contain the COVID-19 pandemic. The obvious downside is the economic costs, since most economic activities depend on in-person interaction. Thus, at least in the short run, policymakers face an inherent trade-off between the risk of contagion and economic losses.
To analyze this trade-off, it is important to know, first, the exposure-to-infection risk from performing a given job and, second, the ease with which the job can be performed remotely. The actual trade-off will depend on how jobs are distributed along these two dimensions, which is the focus of this article.
We begin by constructing an index of exposure-to-infection risk across occupations by using O*NET data and an index of how easily a job can be performed remotely (from home) across industries and occupations by using the "time worked from home" variable in the American Time Use Survey (ATUS). Such indexes are not new, but our approach is novel in that we harmonize the two datasets through our own crosswalk to quantify the trade-off between infection risk and the economic losses present in the economy, using the distribution of workers across jobs in the American Community Survey (ACS).
We find that, although jobs not easily performed from home (low work-from-home [WFH] ability) tend to be more exposed to infection risk on average, the negative correlation between the two indexes across occupations and industries is far from tight. First, there are a number of jobs with high exposure that can be easily performed from home (high WFH ability), such as IT sales agent. Second, infection risk varies widely even among jobs with the same WFH ability: For example, neither medical therapists nor experimental physicists can work from home, but the latter's workplaces pose almost no risk of contagion. In addition, even the same occupation can have a very different WFH ability depending on the industry: For example, a registered nurse employed by a hospital has low WFH ability, but one in consulting services has high WFH ability.
In light of these findings, we consider an optimal policy that selectively sends home specific occupations in specific industries to minimize the aggregate wage loss subject to a given reduction in the aggregate exposure-to-infection risk. Intuitively, it is optimal to first send home workers with jobs with high exposure at work and small productivity and wage losses when working from home, the latter of which can be computed from a job's wage and WFH ability. Mathematically, this translates into a linear threshold in the two-dimensional plane of exposure and wage loss.
The aggregate wage loss under the optimal policy is much smaller than under a lockdown of all non-essential jobs as implemented in many U.S. states and European economies. Our version of the real-world lockdown reduces aggregate exposure by 32 percent by sending home 36 percent of all workers, costing 30 percent of aggregate wages. Our optimal policy attains the same reduction in aggregate exposure by sending home only 24 percent of all workers, costing only 15 percent of aggregate wages. That is, the optimal policy achieves the same reduction in aggregate infection risk for half the economic cost. Under a constrained optimal policy in which healthcare-related workers must continue to work normally, the aggregate wage loss is 20 percent, still a third smaller than under a real-world lockdown. These gains are possible because the optimal policy exploits the large variation in WFH ability across occupations and industries for any given level of exposure—the novel fact we establish in this article.
It has become clear that low-wage workers are not only bearing the brunt of the pandemic economically but are also bearing the brunt of the infection risk. Compared with a lockdown of all non-essential jobs, the optimal policy reduces low-wage workers' infection risk the most. On the other hand, a move from such a lockdown to the optimal policy generates the largest wage gains for high-wage workers, although more workers across the entire wage distribution are allowed to work normally and thus earn more than when working from home. In fact, under all policy scenarios we consider, high-wage workers are the least exposed to infection risk and also lose the least economically, pointing to the importance of redistributive policies during a pandemic.
A word of caution: By design, our optimal policy is simple and static, abstracting from the essentiality of certain jobs (other than healthcare-related workers) that need to be performed even in the midst of a pandemic; the complementarity among jobs that need to be performed in person; the economic propagation across jobs and sectors; and the possibility that people switch to jobs with lower exposure or higher WFH ability (to help prevent wage loss). It also assumes that the indexes we construct are constant, ignoring the potential change in exposure or WFH ability of specific jobs due to more subtle non-pharmaceutical interventions such as wearing masks, reorganizing the workplace, and changing individual behaviors as the population adjusts to the pandemic. However, it does capture the direct trade-offs given the work patterns at the onset of the pandemic. Furthermore, our analysis presents simple guidance for policymakers that is easy to implement in practice, while also providing a benchmark for structural economic models that consider some of the dimensions that we abstract from.
It should also be noted that our industry-occupation crosswalk between ATUS and ACS/O*NET is not specific to exposure and WFH ability and can be applied to a broader set of topics.
Read the full article.