THE ESTIMATES BELOW WERE MADE CIRCA MARCH 20, 2020.
Updated estimates are available here, complete with additional cost categories and references and based on important new data.
----------------------
We are currently fighting a war against the COVID-19 virus. The war presents an obvious and massive tradeoff between “guns” – activities whose primary purpose is war production – and “butter,” which refers to the normal activities of households and businesses. Without any improvement in our techniques for fighting the war, the sacrifices by households and businesses will be staggering and historically unprecedented.
This document enumerates and quantifies the sacrifices using two novel methods. The results suggest that negative 50 percent is an optimistic projection for the annualized growth rate of U.S. GDP in 2020 Q2 if the nonessential businesses were not allowed to operate during that quarter. GDP losses, while massive, nonetheless understate the true costs of the sacrifices that households and businesses are making, which I estimate to total almost $10,000 per household per quarter. This is why better techniques for fighting the war are incredibly valuable.
Government officials around the world have ordered businesses shut and families to stay in their homes except for essential activities. My purpose here is to enumerate and quantify the real economic costs of fighting the war in this way. This document estimates the opportunity costs of lockdown relative to a normally functioning economy, which is the relevant comparison for the purposes of policy decisions such as medical innovation (or, possibly, a statistical sampling effort) that could end the lockdown earlier.
To be clear, the cost-enumeration exercise can be consistent with a conclusion that the war is worth fighting – that depends on quantifying the benefits, which are surely significant given the value that people place on health and longevity. Although this document does not address the question of whether a lockdown is better than no policy response, it offers some of the essential ingredients for such an analysis. No policy response would itself involve economic contraction during the period of time in which no treatment or vaccine is available.
The lockdown method of fighting the war on the virus directly affects how people allocate their time. That includes what they do, where they do it, and with whom. Because normal time allocation includes elements of saving and capital accumulation, such as learning skills, the economic effects of the war are felt into the future as stocks of physical and human capital are reduced. For enumeration purposes, I distinguish market production activities from all other activities, especially because market production is counted in conventionally-measured GDP whereas leisure activities are not. Although I refer to the non-market activities as “leisure,” they include religious gatherings and forms of effort such as student effort in school and effort put toward housework. Both market production and leisure activities are significantly affected by lockdown.
A. Net Costs Associated with Market Production
Lockdown reduces the amount and effectiveness with which people work. The effects of this can be measured on either the production side of the national accounts, as the value of goods and services not produced, or the income side as reductions in total incomes. Either approach yields the same result, up to measurement error.
However, the incidence – the distribution of impact across industries, occupations, and income groups – is different from the production and income perspectives.
Momentarily putting aside costs associated with leisure activities, the lockdown can be analogized with a change in the number of holidays and weekends (“nonwork days”). A well-studied, albeit obscure, element of national income accounting is the adjustment for the fact that the number of nonwork days normally varies from year to year.
A normal year has about 251 working days and about 114 nonworking days.
The national accountants have found that adding a nonwork day to the year reduces the year’s real GDP by about 0.1 percent and have been applying this estimate to both the production and real income accounts.
Adding a nonwork day to a quarter would therefore reduce the quarter’s unadjusted real GDP by about 0.4 percent.
Extrapolating from this finding, removing all of the working days from a quarter is 62 or 63 times this, or 25 percent.
In other words, if seasonally-adjusted GDP 2020-Q2 would have been $5.5 trillion at a quarterly rate (see Table 1), then changing all of that quarter’s working days to the functional equivalent of a weekend or holiday would reduce the quarter’s GDP to $4.2 trillion.
Applying the same approach to 2020-Q1, with a lockdown occurring for one-eighth of the quarter, 2020-Q1 real GDP (in 2020-Q2 prices) would be $5.4 trillion. The quarter-over-quarter growth rate of seasonally-adjusted real GDP would, expressed at annual rates, therefore be
-10 percent in Q1 and
-63 percent in Q2.
The Q2 growth rate would be less negative to the extent that a lockdown was in place for only part of the quarter or for part of the country.
[Table 1 contains POINT ESTIMATES, not worst-case scenarios. If these costs are to be netted against health benefits, then those benefits should be POINT ESTIMATES too. Worst-case scenario health benefits should be compared with worst-case scenario costs, which far exceed what I provided].
Lockdown is not exactly the functional equivalent of changing workdays to weekends or holidays. On one hand, a segment of the workforce will engage in telework during lockdown that they would not perform on a normal weekend or holiday. Other segments or regions will be exempt from shutdown. This by itself suggests that the $4.2 trillion estimate is too pessimistic. On the other hand, much of the normal weekend activity such as restaurants, entertainment, and religious activities is not occurring during lockdown. This by itself suggests that the $4.2 trillion estimate is too optimistic.
A second method uses the production side alone. Labor is reduced by the number of “non-essential” employees, which has been about 30 percent during Federal shutdowns.
In some of the industries, real capital will continue to be used, albeit by fewer employees. Other industries will not use their capital, although it may be repurposed, such as a hotel being used as a hospital ward. To be conservative, I assume that few industries increase their labor-capital ratio.
The reduction in capital input is therefore somewhere between 0 and 30 percent; I assume 15 percent. History has repeatedly shown that labor is more important in the production process than capital, so that by the second method real GDP is reduced 26 percent.
The estimates above assume no black markets. But, as seen with border patrol and the war on drugs, any government regulation attempting to block valuable gains from trade will result in black market activity. Businesses will also work the gray area, lobbying and distorting their operations to have more activities declared “essential.”
Black-market activity is far less productive than legitimate activity, which is why it does not come close to replacing the “non-essential” sales that were banned. But it still has value, which is why the best welfare effects of shutdown may be less pessimistic than analysis assuming zero black market.
I assume that black markets replace 25 percent of the gains from trade, based on studies of illegal drugs.
However, value generated in black markets is typically not measured as part of GDP. Indeed, black markets compete with legitimate markets for the factors of production and by this channel would reduce measured real GDP even more than would occur without black markets
(Fleming, Roman and Farrell 2000).
Table 1 shows only averages, but the distribution of costs is unequal. Revealed preference -- that fact that the
demand for social insurance increases in these situations -- suggests that the inequality itself is a cost large enough that people are willing to tolerate even further increases in the average costs (i.e.,
further decreases in GDP) in order to mitigate the costs for those disproportionately affected.
Although the national income accounts were designed on the basis of the principles of welfare economics, GDP growth is not exactly a benefit and GDP reduction is not exactly a welfare cost because valuable activities and assets such as home production, elements of human capital accumulation, and environmental quality are not yet recognized in the official national accounts (Hartwick 1990, Nordhaus and Kokkelenberg 1999, Jorgenson 2010). However, as discussed further below, the GDP losses cited above prove to reasonably approximate more comprehensive welfare losses.
B. Net Costs Associated with Nonmarket Activities
The nonmarket/home sector is affected by lockdown through two basic channels, as shown in Table 1. The first channel is discussed above: the nonmarket sector has additional labor that has been forced out of the market sector. The second channel is that the nonmarket sector becomes less productive, both for the nonmarket time that normally exists as well as the additional nonmarket time coming from the market sector, because even in their nonwork activities people are restricted in terms of where they go and how they associate with others. The percentage change in the value created in the nonmarket sector combines the two channels and is approximately the sum of the (positive) percentage change in labor input and the (negative) percentage change of nonmarket productivity.
An important example of the second channel is the time allocation of children and young adults who would normally be enrolled in school and now spend their time at home. Their learning from normal face-to-face interactions with teachers and fellow students is not fully reflected in GDP, but is nonetheless valuable. In other cases, as with religious gatherings, entertainment, and tourism, lockdowns reduce the value of these activities by limiting how people can congregate and the market inputs that can be used as part of the leisure activity.
Because the national accounts are based on the principles of welfare economics, GDP would ideally capture value created or destroyed in both the market and nonmarket sectors. Measurement challenges have so far limited the scope of conventional GDP measures to the market sector. Conventional GDP measures therefore miss the value of additional nonmarket time added by the shutdown (the first channel) as well as the reduced productivity of nonmarket time (the second channel). This section provides estimates of the two, which can be added to the GDP losses from Section I.B to arrive at a welfare loss of shutdown as compared to normal economic activity.
To estimate the nonmarket value of added labor, I use the short run of the neoclassical growth model, which is essentially a labor supply and demand framework. The average nonmarket value of time is below the after-tax real wage that would normally prevail, but above the marginal value of time with a shutdown, which I estimate to be 49 percent of the former.
With a 48 percent marginal tax rate (inclusive of implicit taxes on labor income), the total nonmarket value of the extra time is about $7 billion per day (see Table 1), or about 30 percent of the reduction in real GDP. Simply put, about two-thirds of the $22 billion daily GDP loss is a welfare loss, even without considering any productivity change in the nonmarket sector.
Full-time schooling, where there are normally about 73 million children and young adults enrolled, is the part of the non-market sector’s productivity loss that is easiest to quantify. Their time and efforts, which are known as “foregone earnings” and not counted in conventional GDP measures, are combined with direct schooling costs such as the education industry’s payroll and capital expenses because the students, their parents, or their community value the results of schooling. The direct costs were $370 billion in 2018. Various studies, such as Breton’s
(2013) estimate that foregone earnings are about 102 percent of the direct costs, which would be $377 billion in 2018, or about $4.5 per hour that the average student was in school. Assuming that some schooling will still occur during lockdown, I take the loss of student output attributable to their time and effort to be half, or about $2.25 per hour that they would have been in school.
Learning does not stop at graduation. Post-graduation workers learn on the job, which shifts the composition of their compensation toward skill acquisition and away from the cash and other fringe benefits that are part of conventionally measured GDP (Rosen 1972). Although the market sector may be the physical location of this learning, I count the foregone earnings as “nonmarket” because it is usually unmeasured. I estimate the value of foregone earnings using the cross-section age-earnings profile and the average of two estimates of the age-training profile (Mulligan 1998). During a shutdown, this learning does not occur for 30 percent of the workforce, although (as with market production) I assume that about one-third of its value is replaced with nonmarket activity. The net opportunity cost associated with on-the-job training (OJT) is therefore about $107 billion at an annual rate, as shown in Table 1.
The normal population has even more adults not in the labor force than full-time students, not to mention all of the time that workers normally spend outside of work. If a shutdown also reduced the hourly value of their time by $2.25 for 2000 hours per year for those out of the labor force and for 500 hours per year for those who work, that would be a loss of $767 billion at an annual rate.
C. The Incidence of the Net Costs
The massive costs of shutting down “non-essential” activities are not shared equally among Americans. Some workers are still able to draw a normal salary even while their industry is inactive. Others work in industries such as parts of healthcare that are booming as a result of the pandemic. Because the aggregate reduction in the value of what is produced must equal the aggregate reduction in total income, the costs of lockdown will fall disproportionately on the remainder of the population that are not in these circumstances.
Public programs are being created and expanded with the intention of helping some of those who are disproportionately bearing the costs. Redistribution policy may help distribute the aggregate costs more fairly, but in no way can it reduce the aggregate cost. Even while these policies assist those who are not working because of the pandemic, they do not replace the work and production that the workers would have been doing. Instead, redistribution itself has its own aggregate costs, for example, by reducing incentives to work and incentives of workers to shift into industries that need them most (Mulligan 2012).
D. Bibliography
Barro, Robert J., José F. Ursúa, and Joanna Weng. "The Coronavirus and the Great Influenza Pandemic: Lessons from the “Spanish Flu” for the Coronavirus’s Potential Effects on Mortality and Economic Activity." NBER working paper, no. 26866 (March 2020).
Breton, Theodore R. "Were Mankiw, Romer, and Weil right? A reconciliation of the micro and macro effects of schooling on income." Macroeconomic Dynamics 17 (2013): 1023–1054.
Christian, Michael S., and others. Human capital accounting in the United States: 1994 to 2006. BEA, 2010.
Council of Economic Advisers. Mitigating the Impact of Pandemic Influenza through Vaccine Innovation. Executive Office of the President, September 2019.
DiMasi, Joseph A., Henry G. Grabowski, and Ronald W. Hansen. "Innovation in the pharmaceutical industry: new estimates of R&D costs." Journal of health economics 47 (2016): 20–33.
Eichenbaum, Martin S., Sergio Rebelo, and Mathias Trabandt. "The Macroeconomics of Epidemics." googlegroups.com. March 23, 2020. https://fb8280a8-a-62cb3a1a-s-sites.googlegroups.com/site/mathiastrabandt/home/downloads/EichenbaumRebeloTrabandt_EpidemicsMacro.pdf?attachauth=ANoY7coRODwA_z1gJoSLHsTEXF0IQcWOQbFP5bfX9cHSypuO2NuwJPyq7T2A4p2wdppOl0jWVgmAzN4-16-irT7vZqkwvBs8_8PNb3hs0ZJmZtr.
European Central Bank. "The impact of the number of working days on euro area GDP in 2004." Monthly Bulletin, June 2004: 51-63.
Fleming, Matthew H., John Roman, and Graham Farrell. "The shadow economy." Journal of International Affairs, 2000: 387–409.
Hartwick, John M. "Natural resources, national accounting and economic depreciation." Journal of Public Economics 43 (1990): 291–304.
Jaffe, Sonia, Robert Minton, Casey B. Mulligan, and Kevin M. Murphy. Chicago Price Theory. Princeton University Press (ChicagoPriceTheory.com), 2019.
Jorgenson, Dale W. "A new architecture for the US national accounts." Review of Income and Wealth 55 (2009): 1–42.
Jorgenson, Dale W. "Human capital and the national accounts." Survey of Current Business 90 (2010): 54–56.
Katz, Eric. "Agencies Paid Federal Employees $3.7 Billion Not to Work During Recent Shutdowns." govexec.com. September 17, 2019. https://www.govexec.com/pay-benefits/2019/09/agencies-paid-federal-employees-37-billion-not-work-during-recent-shutdowns/159936/.
Mulligan, Casey B. The Redistribution Recession. New York: Oxford University Press (redistributionrecession.com), 2012.
Nordhaus, William D., and Edward C. Kokkelenberg. Nature’s Numbers. National Academy Press Washington, DC, 1999.
