The author has published an update to this piece with new data and analysis on September 16, 2020: “COVID outcomes update: Health and employment impacts in the US compared to other countries”
In response to the COVID-19 crisis, the US chose to lock down its economy around mid-March, quite a bit later than many other countries; and then Congress passed (and President Trump signed) the CARES Act and other relief bills in late March and after, with a variety of measures (like payments to individuals, loans to affected businesses, and an expanded safety net) in some ways similar to and in other ways different from other countries.
Given these decisions, how well has the US performed, in both employment and health outcomes, relative to other countries during the crisis? By comparing both employment and health data for the US with those of other OECD countries in the first five months of 2020, we can ascertain the extent to which our federal policy response has either mitigated or exacerbated damage on both dimensions.
In Table 1 below, I present employment data and virus case outcomes for the richest 25 OECD nations (in terms of per-capita income), among those with populations of over 4 million people (and excluding Saudi Arabia, for whom no economic data in 2020 were available). I have drawn the data from two highly credible websites: 1) Trading Economics for the economic data; and 2) the Johns Hopkins University Coronavirus Resource Center for data on virus cases and mortality. The economic data appear in part A of the table, and the virus case data in part B.
Table 1: U.S. v. other OECD countries (Population > 4M): Employment and virus outcomes in 2020
Part A. Unemployment rates in 2020 (%)
|Country||UR 1/20||UR 3/20||UR 4/20|
Part B. Virus cases in U.S. v. other OECD countries
|Country||Cases/1,000||New Cases/1,000||Deaths/1,000||New Deaths/1,000,000|
Note: Unemployment rates are obtained from Trading Economics (www.tradingeconomics.com) on May 28, while COVID Cases were obtained from Johns Hopkins University Coronavirus Resource Center (www.coronavirus.jhu.edu). The sample of countries includes the 25 richest OECD countries (in per capita terms) with at least 4 million people, excluding only Saudi Arabia (for whom no economic data are available in 2020). An * indicates unemployment rates that are measured for the entire first quarter of 2020, rather than March. Of the countries showing quarterly rates, only Spain’s rate has risen substantially from the fourth quarter of 2019 (when it was 13.8 percent). The cases and deaths of each country are defined as of May 28.
The three columns of part A present unemployment rates in January, March and April 2020 respectively. While most OECD countries already report data for the first three months of the year, only about half have already done so for the month of April (as of May 28, 2020). Furthermore, four of the countries report only quarterly rather than monthly rates, and these only for the first quarter of 2020. (These can be compared to increases during the last quarter of 2019, though with the understanding that changes over a 3-month period might dampen effects in those cases that occur primarily at the end of the quarter.)
The columns in part B of the table present four kinds of data on the virus caseloads in the relevant countries: total cases and deaths documented (as of May 28, 2020) per thousand residents, and net new cases and deaths, defined as the changes in numbers on that date relative to the previous one, per thousand or million residents. While the total cases and deaths can be found for almost every country in the sample, we have the new ones only for 10 industrialized countries: the USA and nine countries in the EU.1
What do the data on employment outcomes in Table 1A indicate? The unemployment rates reported for January, March and April 2020 (or for the first quarter of that year) indicate the following:
- Since the US began its shutdown of businesses to combat the virus only in mid-March, the rise in its unemployment rate for that month (relative to January, which represents the pre-virus labor market in all cases) is just .8 percentage points (from 3.6 to 4.4%).
- But even this increase is larger than what we observe in all but 3 of the other 22 OECD countries – Austria, Canada and Spain – that present such data in either a monthly or quarterly fashion.2
- The rise in unemployment in the US from January until April – 11.1 percentage points – is larger than that observed for any other OECD country that we consider here.
The data in Table 1B, for total cases and deaths as well as new cases and deaths per capita, indicate the following:
- Only 2 countries – Singapore and the Czech Republic – out of the 22 others have larger numbers of cases per capita than the US.
- Only 6 countries out of the 22 others indicate larger numbers of deaths per capita than the US.
- Among the nine other industrial countries for which we have new caseload data, none indicates as high a number of new cases per capita as does the US, indicating we have been less successful than any other of these countries in bringing its virus caseload under control and in “flattening the curve.”
- Only 2 countries – Sweden and the United Kingdom – indicate higher numbers of new deaths per capita than the US.
What do these numbers imply about the US experience, and the administration’s role, in handling of the crisis to date?3Because it waited a very long time to begin shutting down the economy and imposing emergency measures, US workers ultimately lost very large numbers of jobs by April (plus more in May that will be measured in the next Bureau of Labor Statistics report). These delays no doubt aggravated the ultimate severity of the crisis, in both economic and health terms. And, by failing to develop a federal strategy for “test and trace” and for careful data-based reopening, the US has experienced nearly the least progress of any industrial nation in combating the spread of the virus.
A number of caveats are in order at this point. First, differences in measured unemployment rates reflect differing measurement rules and timing, as well as some real policy differences. For instance, at least 15 countries on this list – including the US – have relied on subsidizing firms to keep workers on payroll, thus limiting their unemployment increases.4
Most of these countries have relied much more heavily on these subsidies than the US, with 40-50 percent of workers covered in France, New Zealand and Switzerland, and approximately 20 percent in several other countries. And, for quirky measurement reasons, at least 3 countries – Canada, Israel and Ireland – report larger increases in Unemployment Insurance receipt but much smaller increases in unemployment than the US (Rothwell, op. cit.).
At the same time, some of the differences in measured unemployment increases are no doubt real. Reliance on payroll subsidies to keep workers on payrolls likely leaves these workers less uncertain about their future employment prospects, and reduces the share of workers who will experience permanent and very costly job loss over time.5
Indeed, our official unemployment rates understate job and earnings losses in the US by a large amount, even as of April; and such numbers in May (and for real GDP in the second quarter) will no doubt be much larger.67 We do not know whether these downward biases in officially reported rates are smaller or larger in other OECD countries, in April and beyond.
And there seems less doubt that the differences in reported virus cases and deaths are real. Had the Trump administration acknowledged the crisis earlier, and implemented the shutdown plus economic relief earlier as well, how much difference might it have made? A report from Columbia University last month estimates that, had the administration implemented emergency measures just one week earlier, the numbers of people infected and their mortality rates might have been dramatically lowered. While we have no comparable study on economic impacts, one way to roughly estimate the impacts of an alternative policy scenario is simply to average the employment and health outcomes of the other OECD countries and compare them to the US:
Table 2: Changes in employment and virus outcomes: U.S. v. other OECD countries (Population > 4 million)
|Unemployment Rates (%) Jan.-Mar. 2020||0.32||0.8|
|Unemployment Rates (%) Jan.-April 2020||1.44||11.1|
|Viral Cases and Deaths|
I report all of these numbers in Table 2, and some in Figures 1 and 2. They indicate the following:
- The unemployment rate in the US grew by about a half percentage point more than in the other OECD countries between January and March, and by nearly 10 percentage points more between January and April.
- Total virus cases per capita have been approximately two-thirds higher in the US, with over 2 cases per thousand more, while deaths have been about 50 percent higher, at about .1 more per thousand; and
- New cases in the US have been about 3 times as high as in the other OECD countries (69 v. 21.5 per million), while new deaths are almost twice as high (2 v. 3.7 per million).
To get a sense of how large these differences are between the experiences of the US and other OECD countries, we can apply them to the overall levels of each measure in the US. Taking the gaps in measured unemployment rates in April at face value – and remembering that they are driven partly by substantive differences across countries in reliance on payroll subsidy – the higher unemployment rate in the US translates into approximately 15 million more jobs that Americans would still have had our unemployment rate remained at the level of the other OECD countries.
And, had our virus caseloads and deaths been similar to the OECD average, the US would have had nearly 700,000 fewer cases of the virus, and a death toll of over 30,000 less.
Furthermore, as the new cases portend a much larger second wave of viruses in the second half of 2020 in the US than elsewhere, these numbers will threaten to substantially worsen our unemployment (and real GDP) numbers later in 2020, relative to other OECD countries. By opening the economy in many states very early, and before their new case curves have substantially flattened, these states risk large new waves that will either generate tragically large numbers of fatalities, a second wave of economic shutdown, or possibly both.
Again, some caution n is in order as we review these results. Differences in unemployment rates reflect measurement as well as policy differences, including various downward biases in such measures that I note above. Whether the measured gaps in jobs across countries would be smaller or larger if job losses not currently counted were included in unemployment rates here and elsewhere, and when rates for May are announced, remains unclear. And differences in unemployment as well as virus caseloads and deaths likely reflect a wide range of differences – in policy but also country demographics, health outcomes and other factors – for which I cannot control.
Still, the differences across countries in outcomes strongly suggest that certain features of the policy response to the virus in the US – such as its delayed implementation and the lack of a clear national strategy on testing and tracing, as well as some features of the relief program (like its lack of emphasis on payroll protection) – have generated much higher job losses than were necessary, as well as tens or hundreds of thousands of more lives lost. And much larger losses of output, income, jobs and lives are still to come in May and beyond.
A previous version of this publication had in error in the GDP figures, which have been removed.
Pei, Sen et al. 2020. “Differential Effects of Intervention Timing on COVID-19 Spread in the United States.” Department of Environmental Health Sciences, Mailman School of Public Health, Columbia University.
Rothwell, Jonathan. 2020. The Effects of COVID-19 on International Labor Markets: An Update. Brief, Economic Studies, Brookings Institution.
- While a 5- or 7-day moving average is preferable for new cases and deaths data than a one-day response, the Johns Hopkins website made calculation of such an average somewhat more cumbersome. But I did calculate new case and death rates on a few other earlier days, and the broad patterns were the same qualitatively.
- Spain reported unemployment rates of 13.8 and 14.4 percent for 2019 Q4 and 2020 Q1, implying somewhat larger rates for the moths of February and March only.
- The CARES Act likely had little effect on outcomes observed through the first quarter (or end of March), and only modestly affected the April unemployment rate, measured during the second week of that month.
- Rothwell (2020) reports 13 countries that are using payroll subsidies, and Japan and South Korea are not included in his list. The US program is the Payroll Protection Plan that has been part of the loans for small businesses that are being dispersed through the Small Business Administration.
- On the other hand, a large degree of payroll subsidy might well be costlier fiscally, and create less flexibility for employers to adjust their employment levels to new production technology or market forces.
- As of April, about 8 million US workers had left the labor force relative to its level in February, while another 8 million were forced to work fewer hours. These two groups would each add another 5 percentage points to the official US unemployment rate in April. Since new Unemployment Insurance claims have risen by several million workers in the four weeks since the April surveys were administered, cumulative measured unemployment and job loss in May will be much larger than in April.
- It appears as though real GDP will follow a similar pattern to unemployment. Real GDP in the US fell by 5% in the first quarter on an annualized basis, and 1.25% without annualizing. In the other OECD countries, the decline in GDP averaged 2.4 percent. Still, the decline observed in the US was comparable in magnitude to that observed in some other OECD countries such as Norway (-1.5%), South Korea (-1.4%) and Japan (-0.9%). But our declines for the second quarter will likely be much higher than our first quarter estimate, consistent with the employment numbers.