Why do the authors emphasize their results for total hours worked more that their results for employment?

Table 3: Distribution of Lost Jobs in the Lower Quartiles, by 25 percent reduction in employment. Salaried women's Employment Sector hours, however, dropped only 14 percent, and employ- Losses in Distribution, % ment fell by only 5 percent. Whether individuals are unionized also appears to be associated with whether Bottom quartile hours losses are also job losses. Although the aggregate Retail trade 20.0 hours lost by union and non-union employees are similar Educational services 10.2 in magnitude (26 and 29 percent, respectively), their em- Health care and social assistance 8.8 ployment losses are not- non-union workers experience Information, culture and recreation 6.0 a 20 percent decline in employment, whereas unionized Accommodation and food services 28.5 workers experience only a 12 percent decline. The experi- All other sectors 26.6 ence of non-union workers is worsened when combined Second quartile with other characteristics. For example, we found that Construction B.8 women who were non-union and paid hourly experienced Manufacturing, durable goods 10.9 a 44 percent reduction in work hours and a 32 percent Retail trade 13.2 reduction in employment. Health care and social assistance 16.0 We expect that differences across occupations and industries explain a substantial part of the differentials in Accommodation and food services 9.5 COVID-19 impacts across various job characteristics, re- All other sectors 41.7 flecting the shutdown of many non-essential activities. To Note: Percentages total more than 100 because of rounding. examine this further, we estimate industry and occupation Source: Authors' tabulations using the Labour Force Survey. shares in the employed population in February 2020 and the COVID-19 effects within each industry-occupation occupations (of 19 percent) that is not matched by a loss in category and use this information to construct counter- employment (at only 2 percent). We believe this is related factual losses that illustrate how much of the differential to the significantly different experience of those who are across quartiles reflects COVID-19 effects across industry- paid hourly and those who are salaried (also presented occupation groups and how much is within industry in Tables 4 and 5). Among women paid hourly, there was and occupation." The results are presented in Table 6. We a 39 percent reduction in hours due to COVID-19 and a find that 60 percent of the relative decline in employmentfor the bottom quartile (compared with the top quartile) is accounted for by COVID-19 effects at the industry- occupation level (so that 40 percent of the difference between quartiles reflects within-industry-occupation employment losses). Similarly, 58 percent of the relative decline in hours for the bottom quartile is accounted for by COVID-19 effects at the industry-occupation level. Finally, looking at classes of workers, Tables 4 and 5 show similar interesting differences. Those working in the public sector saw the smallest decline in hours worked (at 16 percent) and employment (5 percent). We see a 32 percent loss in hours and a 21 percent loss in employ- ment for those in the private sector. Aggregate hours lost Were greatest among the self-employed (at 51 percent), but this results in a loss of only 3 percent of employment. Examining the self-employed reinforces our view that it is important to examine the losses in both aggregate weekly hours worked and employment.representing those at work or absent.) Between Febru- In Figure 2, we present how the employment and ary and April 2020, nearly 2.5 million individuals aged hours losses were distributed across the weekly earnings between 20 and 64 years lost their jobs, representing a 14 quartile. Here we see that nearly half of job losses were percent decline in employment. However, recognizing attributed to workers with earnings in the bottom quar- that population growth and seasonality will generally tile. Those with earnings in the bottom half of the weekly drive some part of the month-to-month change in employ- earnings distribution (in the bottom or second quartile) ment, we also provide adjusted estimates based on 2018 account for almost 80 percent of job losses. Employment changes. With this adjustment, we see that the COVID-19 losses in the top quartile of weekly earnings represent impact represents a 15 percent decline in employment. only 4 percent of all losses. In Figure 2 we also present In Tables 1 and 2, we further characterize the distri- results for workers who moved from being employed and bution of lost work by considering the loss of hours and at work to being employed but absent with a substantial employment within demographic groups. First, we see reduction in hours. Similar to employment losses, nearly that men and women experienced similar job losses be- half (47 percent) of those affected by the loss of work were tween February and April. Overall, aggregate hours fell in the bottom quartile of weekly earnings. by 34 percent for women and 31 percent for men. This We also summarize the overall impact of COVID-19 by contrasts with the initial impact of COVID-19 on jobs in showing the decline in hours worked by quartile (last set March 2020, whereby women were clearly facing larger of bars in Figure 2). Hours losses are not as concentrated losses than men (see Milligan, Schirle, and Skuterud 2020). in the bottom quartile because top quartile workers had In later sections of this article, we return to the gendered their hours substantially reduced even when they kept nature of jobs affected when considering the longer-run their jobs. Workers in the bottom two quartiles were, impacts of the COVID-19 shutdowns and the occupations nonetheless, most adversely affected. They account for 37 most affected. Second, we see that the largest impact is and 34 percent, respectively, of hours losses, compared on younger workers, aged 20-29 years, whose aggregate with 10 percent for workers in the top quartile. hours fell 40 percent and whose employment fell 25 per- The employment losses in the lowest quartiles, illus- cent as a result of the shutdowns. Hours and employment trated in Figure 2, reflect heavy closures within some losses were smallest among those aged 40-49 years, who industries. We break down the losses for the bottom two experienced a 12 percent drop in employment as a result earnings quartiles by industry in Table 3. For those in the of COVID-19. lowest earnings quartiles, job losses in retail trade and With respect to other characteristics in Tables 1 and 2, accommodation and food services account for nearly half the patterns are similar. For women with children, the loss the job losses (presented in the top panel of Table 3). The in aggregate weekly hours and employment is larger for large loss of jobs in health care and social assistance, as those with children aged younger than 12 years than for well as in education, largely reflects home care providers those with kids aged 13-17 years, potentially signalling and education support. These five industries combined the importance of caregiving responsibilities among moth- represent nearly three-quarters of the losses for workers ers with younger children. The impact of COVID-19 on in the lowest quartile. women without children appears larger, consistent with When we look at job losses in the second quartile (in the women without children at home being concentrated in second panel of Table 3), we see a substantial but smaller relatively low-wage jobs, which were most affected by proportion of job losses attributed to some of the same shutdowns.* industries. However, workers in this part of the earnings With respect to regional differences, we see that distribution were also affected by losses in construction Quebec's loss of hours (37 percent) and employment (18 and durable goods manufacturing, together accounting percent) exceeded that of other provinces. Manitoba's for 20 percent of job losses in this quartile. losses were the smallest, with a 23 percent reduction in A more comprehensive representation of the types of hours and 13 percent loss in employment. However, there work lost as a result of COVID-19 is provided in Tables are no obvious patterns across regions. 4 and 5, where we present the loss in aggregate weekly The job losses we highlight in Table 2, however, are hours and employment (respectively) by job characteris- distributed unequally across the wage distribution. To tics. Given the industries facing large job losses (presented examine impacts across wage groups, we first divide in Table 3), it is not surprising to see large declines in em- workers observed in February 2020 into four equal-sized ployment and hours in the most public-facing occupations. groups, ranked on the basis of weekly earnings. To gauge For example, sales and services occupations experienced the magnitudes, the bottom quartile threshold is $646 per a 45 percent decline in hours and a 27 percent decline in week. This bottom quartile includes individuals who work employment (representing the largest loss in employment full time for minimum wage." The February 2020 thresh- among occupation categories). olds are used when measuring employment in April, and In Tables 4 and 5, we see a smaller but substan- February 2018 thresholds are used for 2018. tial reduction in aggregate hours lost in management