AdobeStock_275068063.jpeg
VoxEU Column COVID-19 Labour Markets

COVID-19 and labour reallocation: Evidence from the US

One of the most urgent economic impacts of the Covid-19 crisis is on labour markets. Widespread job losses, drastic increases in unemployment benefit claims, and the rise of working from home have dominated the discussion during the pandemic so far. This column presents evidence from the US, arguing that the pandemic itself represents reallocation of labour within the economy. As different sectors and occupations are hit with variable severity, the authors argue that policymakers should be wary of this variation, responding with policies that will hold firm over time.

As of early July 2020, economic policy debates in the US centre on whether to extend or to modify major programmes created by the ‘Coronavirus Aid, Relief and Economic Security Act’. On 04 July 2020, President Trump signed legislation that extends the ‘Paycheck Protection Program’ (PPP) into August, giving small businesses more time to apply for loans that incentivise employee retention (Andrews 2020, Westwood and Mattingly 2020). A more contentious issue is whether to extend the ‘Federal Pandemic Unemployment Compensation’ (FPUC) programme, which supplements state-level unemployment benefits by $600 per week. Two-thirds of eligible ‘job losers’ receive unemployment benefits greater than lost earnings under the FPUC programme (Ganong et al. 2020), discouraging a return to work. As Congress went to recess for the Independence Day holiday, there was little agreement about whether to extend the FPUC programme which is currently set to expire on 31 July 2020 (Duehren and Wise 2020).

In thinking about policies to counteract the negative economic effects of the COVID-19 shock, it is essential to correctly diagnose the nature of the shock. In a recent paper (Barrero et al. 2020a), we develop evidence that the COVID-19 shock has major reallocative consequences. In light of this evidence, we argue that there are large benefits to policies (and policy reforms) that facilitate a speedy reallocation of jobs, workers, and capital to newly productive uses in the wake of the pandemic. In contrast, policies that discourage and delay reallocation are likely to slow the recovery from the pandemic and the economic lockdown.

One key point is that millions of jobs lost during the pandemic recession are gone for good. To quantify this point, we start with data from three sources: an employer survey, a Washington Post/Ipsos survey of individuals, and records for unemployment benefit claimants in California. All three sources suggest that about 23% of layoffs during March-May 2020 were seen as permanent when they happened, with the rest seen as temporary. Historically, a sizeable share of layoffs regarded as temporary when they happen do not result in actual recalls. Adjusting for this historical pattern, we project that 32-42% of pandemic-induced layoffs will be permanent (in the sense that the job loser never returns to his or her old job). Thus, a successful recovery from the pandemic recession will require new jobs for at least 10 million workers.

We also quantify the reallocative aspects of the COVID-19 shock in other ways. First, to assess the near-term reallocative effects, we draw on two special questions in the Atlanta Fed/Chicago Booth/Stanford Survey of Business Uncertainty (SBU).1 One question asks (as of mid-April) about the impact of pandemic-related developments on own-company staffing since 01 March 2020. Another asks about the anticipated impact over the ensuing four weeks. Cumulating responses over firms (and across these two questions), we find that the pandemic caused near-term layoffs equal to 12.9% of 01 March 2020 employment, and gross new hires equal to 3.9% (Figure 1). In other words, the COVID-19 shock caused three new hires in the near term for every ten layoffs. After netting out replacement hires, we find 2.5 new jobs created for every ten lost jobs. Similarly, the Job Openings and Labor Turnover Survey reports more than four hires for every ten layoffs in March and April 2020. ADP statistics in Cajner et al. (2020) also show considerable gross job creation and hiring activity in the near-term wake of the pandemic. In short, all three sources show much hiring activity even as US employment fell drastically.

Figure 1 Survey-Based Estimates of Gross Staffing Changes from March 1, 2020 to the Middle of May 2020 in Response to COVID-Related Developments

Source: Authors’ calculations using data from the Survey of Business Uncertainty.

Note: The survey questions that form the basis for this chart do not ask about quits. Thus, the reported estimates for Net Staffing Changes and Gross Staffing Reductions are exclusive of quits. If we impute quits by applying the quit-layoff ratio in JOLTS data for March-April 2020 to the SBU layoff rate, the imputed quit rate is 2.9 percent. The Net Staffing Change becomes -13.7 percent of March 1 employment inclusive of imputed quits, as compared to the -10.8 percent value shown in the chart.

Second, to get at medium-term reallocative activity, we draw on firm-level forecasts (at a one-year horizon in the SBU) to construct novel, forward-looking reallocation measures for jobs and sales. These statistics quantify the cross-firm job and sales reallocation expected to occur over the next year in excess of the amount required to accommodate aggregate net changes.2 Figure 2 shows the expected rates of excess job and sales reallocation from September 2016 to June 2020. The expected rates of excess reallocation rose sharply after the pandemic struck. Compared to the pre-COVID period, the rate is twice as high from April to June 2020 for jobs, and more than five times as high for sales. That is, businesses’ expectations at a one-year forward horizon imply much more anticipated reallocation activity after the pandemic struck.

Figure 2 Expected Rates of Excess Job and Sales Reallocation at a One-Year Forecast Horizon, September 2016 to June 2020

Source: Authors’ calculations using data from the Survey of Business Uncertainty.

Third, we provide survey evidence on the pandemic-induced shift to working from home (WFH) in both the near term and over the longer term. To assess the early impact, we fielded a May 2020 survey of American adults who earned at least $20,000 in labour income in 2019. As shown in Figure 3, 42% of respondents worked from home in May 2020, much more than the 26% who worked at business premises. Adjusting for those who are not working, we estimate that working from home accounts for 62% of all labour services supplied in May 2020 (67% when weighting by earnings). These results provide evidence of the extraordinary scale of the WFH phenomenon triggered by the pandemic.3

Figure 3 Working from Home Accounts for More Than 60 Percent of U.S. Labor Services Supplied in May 2020

Notes: This chart summarizes responses to the following question: “Currently (this week) what is your work status?” Response options are “Working on my business premises,” “Working from home,” “Still employed and paid, but not working,” “Unemployed, but expect to be recalled to my previous job,” “Unemployed, and do not expect to be recalled to my previous job,” and “Not working, and not looking for work.”

The data are from a survey of 2,500 U.S. residents aged 20 to 64, earning more than $20,000 per year in 2019 fielded from 21-29 May by QuestionPro on behalf of Stanford University. We re-weight the sample to match the share of individuals at the level of cells defined by the cross product of earnings interval, state and industry (using the current or most recent job) in CPS data from 2010 to 2019. Adjusting for those not working, the results displayed in the bar chart say that (41.9/(100 – 25.9) = 62 percent of labor services were supplied from home as of late May (67 percent on an earnings-weighted basis).

Of course, much of the WFH shift will reverse after the pandemic recedes. To get at the persistent part of the WFH shift, we draw on two special questions in the May 2020 SBU. The first asks firms about the share of full workdays performed at home by their full-time employees in 2019. The second asks them about the share of full workdays they expect their full-time employees will perform at home after the pandemic is over. Comparing responses to the ‘before’ and ‘after’ questions, we find that full workdays performed at home will triple in the post-pandemic economy, rising from 5.5% of all workdays to 16.6%.4 This tripling will involve shifting one-tenth of all full workdays from business premises to residences (one-fifth for office workers). Since the scope for working from home rises with worker earnings, the shift in worker spending power from business districts to locations nearer residences is even greater.

As we discuss in Bloom et al. (2020b), multiple factors lie behind a persistent post-pandemic increase in working from home. First, workers and firms report that the stigma of working from home has fallen sharply. Second, workers and firms have collectively invested huge amounts of time and resources to make working from home function more effectively. Third, working from home has worked better than most firms anticipated, according to our survey evidence. So, even if medical advances or natural forces bring an early end to the health crisis, pandemic-induced shifts in working arrangements will persist to a considerable extent. Similar points apply to pandemic-induced shifts in consumer spending patterns (e.g. toward online shopping) and business practices (e.g. away from air travel and to virtual meetings).

The COVID-19 pandemic is a major reallocation shock with persistent aspects. But what does this mean for policy? Historically, job and business creation responses to major reallocation shocks lag the destruction response by a year or more. Thus, a speedy recovery from the current crisis hinges (at least partially) on policies that enable the economy to reallocate resources effectively, rather than policies that discourage or impede reallocation. For example, extending the FPUC in a manner that makes unemployment more remunerative than work will disincentivise job search, discourage a return to work, and slow the recovery. We prefer income-support programmes (including less generous unemployment benefits) that do not destroy the monetary rewards to working.

As another example, policies that subsidize employee retention irrespective of the employer’s long-term outlook will prevent (or delay) the reallocation of jobs, workers, and capital to more productive uses. The mid-April US Treasury deal with major airlines is a case in point. The deal provides $25 billion in subsidies to airlines in return for barring layoffs and furloughs before October 2020. But it is extremely unlikely that the demand for air travel will recover to pre-pandemic levels by October (or any time in the near future). In circumstances like these, employment-retention subsidies delay the re-deployment of workers and other productive inputs to more efficient uses during the crisis and afterwards. Using taxpayer funds to preserve zombie jobs is not a path to a strong recovery.

To be sure, there is a sound case for providing liquidity support to viable, cash-strapped businesses during the crisis. Delinking liquidity support from employee retention would largely eliminate the incentives to inefficiently retain labour. Low-interest loans without forgiveness provisions would discourage firms with poor economic outlooks from applying for assistance.

References

Altig, D, J M Barrero, N Bloom, S J Davis, B Meyer and N Parker (2020), “Surveying Business Uncertainty”, NBER Working Paper 25956.  

Andrews, N (2020), “House Passes Extension of Paycheck Protection Program”, Wall Street Journal, 1 July.

Barrero, J M, N Bloom and S J Davis (2020a), “COVID-19 Is Also a Reallocation Shock”, Brookings Papers on Economic Activity.

Barrero, J M, N Bloom and S J Davis (2020b), “The future of working from home,” forthcoming.

Berg, J, F Bonnet and S Soares (2020), “Working from Home: Estimating the Worldwide Potential”, VoxEU.org, 11 May. 

Cajner, T, L D Crane, R A Decker, J Grigsby, A Hamins-Puertolas, E Hurst, C Kurz and A Yildirmaz (2020), “The U.S. Labor Market during the Beginning of the Pandemic Recession”, Brookings Papers on Economic Activity.

Davis, S J and J Haltiwanger (1992), “Gross Job Creation, Gross Job Destruction, and Employment Reallocation”, Quarterly Journal of Economics 107(3): 819-863.

Dingel, J and B Neiman (2020), “How Many Jobs Can Be Done at Home?”, VoxEU.org, 07 April. 

Duehren, A and L Wise (2020), “Jobs Data, Coronavirus Cases Cloud Stimulus Plans for Senate Republicans”, Wall Street Journal, 2 July.

Dunne, Timothy, Mark Roberts, and Larry Samuelson, 1989. “Plant Turnover and Gross Employment Flows in the U. S. Manufacturing Sector,” Journal of Labor Economics, 7, 48-71.

Ganong, P, P Noel and J Vavra (2020), “Unemployment Insurance Replacement Rates During the Pandemic”, working paper.

Pagano, M, C Wagner and J Zechner (2020), “COVID-19, Asset Prices, and the Great Reallocation”, VoxEU.org, 11 June.

Papanikolaou, D and L D W Schmidt (2020), “Working Remotely and the Supply-Side Impact of COVID-19”, NBER Working Paper No. 27330.

Westwood, S and P Mattingly (2020), “Trump Signs Paycheck Protection Program Extension”, CNN Politics, 4 July.

Endnotes

1 The SBU is a monthly panel survey of American firms. See Altig et al. (2020) for more information.

2 Our statistics are forward-looking analogs to the backward-looking measures of excess job reallocation examined in Dunne, Roberts and Samuelson (1989), Davis and Haltiwanger (1992), and many later studies. For references to other studies that examine backward-looking reallocation measures in labor, product and financial markets, see Barrero, Bloom and Davis (2020a).

3 Dingel and Neiman (2020) and Berg et al. (2020) estimate the share of work that can be performed at home, which can be seen as an upper bound for the share actually performed at home.

4 The distribution of firm-level equity returns in reaction to the COVID-19 pandemic also suggests that a large part of the shift to working from home will persist. See Pagano et al. (2020) and Papanikolaou and Schmidt (2020).

5,774 Reads