Working from home: The polarising workplace

Abigail Adams-Prassl, Teodora Boneva, Marta Golin, Christopher Rauh 02 September 2020



What was considered a luxury before the pandemic, has become the new ‘normal’. During the Covid-19 pandemic, working from home has become an integral part of people’s lives. Bick et al. (2020) document that more than one third of the US workforce worked entirely from home in May 2020; Gottlieb et al. (2020) estimate that for developing countries the capacity to work from home is lower. For many, this trend is likely to persist, even once the pandemic is over. Bartik et al. (2020) find that more than one third of surveyed businesses in the US that switched to remote working will make it a more common feature of their work model once the pandemic ends. Barrero et al. (2020) present three reasons why working from home is here to stay:

  • First, the stigma of working from home has fallen sharply. 
  • Second, huge investments have been made to make working from home function more effectively.
  • Third, it has worked surprisingly well. 

To shed light on workers’ ability to work from home and what this means for workers’ labour market outcomes during this pandemic, in this column we analyse data from three waves of surveys collected in the US and the UK in March, April, and May (N=24,924). To capture differences in the ability to work from home, we ask respondents to report what percentage of their tasks they could do from home (0-100%). This allows us to study how capabilities to work from home might be different even amongst workers in the same occupation and industry. 

Working from home buffered the shock

While many might not consider working from home to be a blessing, our results suggest that the ability to work from home sheltered workers not just from the health risks of the Covid-19 pandemic but also from the economic shock brought about by the pandemic (Adams-Prassl et al. 2020a). Figure 1 shows the average share of tasks workers can do from home (left) and the share of workers who lost their jobs since the onset of the pandemic (right) for industry (horizontal axis) and occupation (vertical axis) pairs. The darker the shade of a cell, the higher the mean share of tasks that can be done from home (left) and the higher the job loss probability (right). It becomes clear that the colours of the two panels are the inverse of each other. This means that workers in occupation-industry pairs in which a higher mean share of tasks can be done from home were considerably less likely to lose their jobs. 

Figure 1 Mean share of tasks that can be done from home (left) and job-loss probability (right) for occupation-industry pairs

Within occupation and industry variation

In previous work, it was typically assumed that the share of tasks workers can do from home is constant within occupations and industries (e.g, Dingel and Neiman 2020). In Adams-Prassl et al. (2020b), we provide evidence that this is not the case. In Figure 2, we plot the distribution of the share of tasks workers can do from home for four different occupations. Some striking patterns emerge. Within each occupation, there is considerable variation in workers’ ability to work from home. Moreover, the patterns are rather different for the different occupations. In “Food Preparation and Serving” (top left), which includes cooks and bartenders, working from home is not a feasible option for many. Conversely, in “Computer and Mathematical” occupations (top right), which includes software developers, many workers can do a high share of their tasks from home. 

Figure 2 Distribution of ability to work from home within occupations

The occupations “Architecture and Engineering” (bottom left) and “Office and Administrative Support” (bottom right) reveal nearly identical means. On average, workers in these occupations can do slightly more than half of their tasks from home. Nonetheless, the two distributions are very different. One the one hand, many workers in “Office and Administrative Support” can either do none or all their tasks from home. Very few workers in this occupation can do an intermediate share of tasks from home. On the other hand, for workers in “Architecture and Engineering” we observe the opposite pattern. Many workers can do about half of their tasks from home. 

Another striking feature of the distributions in Figure 2 is how similar they are on both sides of the Atlantic. Not only do the average shares by occupation resemble each other closely, but the median, the standard deviation, and the share of workers who can do no tasks or all tasks from home also look very similar across our US and UK samples. In Figure 3 each bubble represents an occupation and the size of the bubble is proportional to the number of observations we have in our survey. The horizontal axis exhibits the mean, median, standard deviation, coefficient of variation, share of respondents that can do all tasks from home, and share of tasks that can do no tasks from home for the US while the vertical axis shows the same for the UK. The strong correlations corroborate the idea that the variation we capture in the share of tasks that can be done from home is systematic. 

Figure 3 Correlation of working from home measures across US and UK

In Adams-Prassl et al. (2020a) we document that the ability to work from home predicts job loss over and above what can be explained by occupation and industry fixed effects, highlighting the importance of taking differences across workers within occupations and industries into account. This also becomes clear in Figure 4, where we show the same four occupations as in Figure 2 but we now plot the share of tasks that can be done from home for those who kept their job (blue) versus those who lost their job (transparent). The share of tasks for those who did not lose their job is shifted to the right within each occupation, i.e. those workers tend to be able to do more tasks from home.

Figure 4 Distribution of ability to work from home within occupations for those that kept and lost job 

The differences in the distributions are also important when considering the furlough policy introduced by the UK government. Under this policy, furloughed employees are not allowed to do any work for their employer and they receive 80% of their salary up to a cap of £2,500. While this policy might make sense for occupations such as “Office and Administrative Support”, where many workers can do either none or all of their tasks from home, it confronts a firm employing architects or engineers with a dilemma. Should a firm furlough an architect who is then forbidden from doing half their tasks from home or let them work while paying a full salary for only half the product? This example lays bare how a short-time work scheme could have been better suited for many occupations in which intermediate shares of tasks could be done from home.

Large and growing inequalities

The ability to work from home does not only vary systematically across and within occupations and industries. We also find that men and workers with a college degree can do a substantially higher share of their tasks from home. Workers on low incomes report being able to do a smaller share. 

However, the unequal abilities to work from home are becoming more unequal over time. Barrero et al. (2020) mention that since the onset of the pandemic, firms have implemented changes to their processes to facilitate the transition to the home office. Further, as the pandemic progressed, workers themselves have become more accustomed to working from home. In line with these findings, our survey data show that the average share of tasks respondents report being able to do from home increased between March and May. In Figure 5 we show on the horizontal axis the share of respondents in the UK reporting to be able to do all tasks from home in March and April. The vertical axis displays the increase in this share in May. The increase has been largest amongst occupations in which workers were already able to carry out a large share of tasks from home. This trend hints towards further polarisation in terms of whether or not a job can be done from home. 

Figure 5 Increase in share of tasks that can be done from home in occupations in the UK


The Covid-19 economic crisis has had large and unequal impacts on workers across the globe. In this column, we document that an important determinant of this inequality is workers’ ability to carry out their tasks from home. We also provide evidence of systematic variation in individuals’ ability to work from the home office, both across and within occupations and industries. We argue that policymakers need to take this variation into account when designing policy responses in order to make full use of the potential of the workforce. 


Adams-Prassl, A, T Boneva, M Golin and C Rauh (2020a), “Inequality in the impact of the Coronavirus shock: Evidence from real time surveys”, CEPR Discussion Paper 14665.

Adams-Prassl, A, T Boneva, M Golin and C Rauh (2020b). “Work that can be done from home: Evidence on variation within and across occupations and industries”, CEPR Discussion Paper 14901.

Bartik, A W, Z B Cullen, E L Glaeser, M Luca and C T Stanton (2020), “What Jobs are Being Done at Home During the COVID-19 Crisis? Evidence from Firm-Level Surveys”, NBER Working Paper 27422.

Bick, A, A Blandin and K Mertens (2020), “Work from home after the COVID-19 Outbreak”, mimeo.

Barrero, J M, N Bloom and S J Davis (2020), “COVID-19 and labour reallocation: Evidence from the US”,, 14 July.

Dingel, J and B Neiman (2020), “How many jobs can be done at home?”, NBER Working Paper 26948.

Gottlieb, C, J Grobovsek, M Poschke and F Saltiel (2020), “Lockdown Accounting”, IZA Discussion Paper 13397.



Topics:  Covid-19 Labour markets

Tags:  COVID-19, remote working, Work from home, teleworking, lockdown

Associate Professor in Economics, University of Oxford; Research Fellow, Institute for Fiscal Studies

Assistant Professor, Department of Economics, University of Zurich

PhD student in Economics, University of Oxford

Lecturer at University of Cambridge, Research Affiliate at CEPR


CEPR Policy Research