Work from home: how COVID-19 sped up the inevitable

At the start of the COVID-19 pandemic in early 2020, a radical shift occurred in how much people worked from home. Almost overnight, the majority of white-collar workers began working from their kitchen tables, garages, and home offices to avoid the risk of getting the virus and to comply with stay-at-home measures. While employers perceive the productivity of working from home (WFH) to be lower than productivity of work at the office in normal times (Bartik et al. 2020, Morikawa 2021), social distancing reduced the productivity of work at the office so drastically that WFH became more productive by comparison.

While initially viewed as a stopgap measure, as the pandemic dragged on both employers and employees grew accustomed to WFH and now expect a permanent increase in the practice after the pandemic ends. When surveyed, US employees report that they anticipate a fourfold increase in the share of hours they will WFH going forward (Barrero et al. 2020). In a recent survey of 133 US executives, employers similarly anticipate a dramatic increase in the share of work done from home going forward (PwC 2021). Scandinavian employers also expect the share of WFH to more than double post-pandemic (Mortensen and Wetterling 2020).

What happened during the pandemic that made this change permanent and how will it affect where we live, work, and our incomes going forward? Rather than happening in 2020, the WFH revolution has been slowly brewing for more than 30 years. Distinct technological advances have contributed to our ability to work effectively at home. In the early 1990s, affordable PCs running Microsoft Word and Excel became widely available. In the mid 1990s, work email became commonplace and the Netscape IPO lead a surge in information posted on the World Wide Web. In the early 2000s, high-speed-internet became widely available and by 2010 mobile phones turned to smartphones. Finally, between 2010 and 2020 video-conferencing technology became useable facilitating remote meetings across the world.

A common theme of each of these innovations is that their impact on the ability to do work depends at least in part, on the prevalence of adoption. There is no point to writing an email if no one reads it, video-conferencing becomes very difficult if the person on the receiving end has slow internet, and so forth. While many home-office technologies have been around for a while, the technologies become much more useful after widespread adoption.

In a new paper (Davis et al. 2021), we postulate that the pandemic accelerated the widespread adoption of technologies that enabled households to produce market work at home which, in turn, permanently raised the relative productivity of working at home. To understand the consequences of this change in home productivity, we specify a model where high-skill workers choose how to allocate their time between working at home or in the office. We assume WFH is only an option for high-skill workers, consistent with the detailed occupational analysis of ability to WFH that Dingel and Neiman (2020) provide. There is no commute when working at home, but the productivity of WFH differs from that at the office. High-skill workers also choose how much physical space to rent at home and (on behalf of firms) in the office. All workers choose where to live, how much to consume, and how much housing to rent.

We use the model and data from before the pandemic to estimate the elasticity of substitution between WFH and work at the office. Our strategy involves measuring the extent to which, in pre-pandemic times, workers from the same industries and occupations but with different commuting times made different choices about how frequently to work at home. The rate at which time allocation changes as commuting cost changes is informative about how substitutable WFH is with office work. We find that WFH is an imperfect substitute with work at the office which has important implications for understanding the future of the post-pandemic workplace. If working at the office and at home are not perfect substitutes, most workers in the future will do some work both at the office and some at home in each week or month rather than choosing a corner solution of all work at the office or all WFH. This may limit the ability of workers and firms to relocate from large cities to distant parts of the country with more favourable tax regimes and lower land costs. Thus, our understanding of how changes to WFH technology will affect outcomes is intimately related to the degree of substitutability between work at the office and WFH.

In addition to our estimates rejecting perfect substitutability, perfect substitutability is inconsistent with the historical evidence on the rise in WFH before the pandemic and employer expectations for what the workplace will look. As Figure 1 shows, the biggest increase in WFH in the years leading up to the pandemic was in the share of workers sometimes working from home.

Figure 1 Share of EU workers working from home sometimes and usually, 1995-2019.

Similarly, PWC (2021) reports that most employers anticipate a hybrid office model going forward in which employees work 1-4 days per week in the office rather than one in which employees can work entirely remote or only show up at the office a couple times a month. Our results are also consistent with Ramani and Bloom (2021), who find that COVID-19 induced spatial reallocation primarily within rather than across metro areas.

After parameterising the model, we simulate the model to understand the impact of the pandemic on WFH technology and its implications. We first study a before period – call it 2019 – where college-educated workers work at home 20% of the time. Given the model structure, this pins down the level of WFH productivity prior to the onset of the pandemic. We then study an after period – call it 2022 – where college-educated workers double their time working at home, which is the lower bound on estimates of the post-COVID increase in WFH. This expected doubling in hours worked at home allows us to size the gain in WFH productivity that occurred during the pandemic. Finally, we then study the pandemic period itself – a period in which we assume office productivity fell by 50%, reflecting the impact of social distancing on productivity at the office.

One of our key findings is that the model implies that the widespread adoption of WFH technology increased the productivity of working at home relative to the productivity of working in the office by 34% between the onset and the end of the pandemic. The model predicts the higher productivity of WFH and subsequent doubling of hours worked at home will lead to approximately a 20% decline in office rents in the central business district (CBD) in the short run and long run if the supply of office space cannot be reduced relative to pre-pandemic levels. The model suggests residential rents will rise in the short run, especially in the outer suburbs, due to increased demand for home office space. In the long run, once the supply of space in residential areas has a chance to adjust, hours worked at home will increase even more. Since only college-educated workers can work at home, the model predicts large gains to the technology from working at home will increase income inequality. Finally, the decline in working the office will lead to a small decline in productivity at the office due to a decrease in agglomeration economies.

We also simulate what would have happened if the COVID pandemic had occurred in 1990, prior to the existence of many work-at-home technologies. In this simulation we assume that, in 1989, relative home productivity is one-third of its 2019 value and that it does not change after the onset of the pandemic in 1990. As with the 2020 pandemic, we characterise the 1990 pandemic by a 50% drop in relative productivity in working at the office. During this hypothetical 1990 pandemic, people continue to work at the office at the same rate and do not substitute into working at home. Incomes and prices fall, but there is no increased demand to work at home in the suburbs. According to our model, in 1990 working at home is not a practical alternative to working in the office. This implies that the pandemic would have had more dire consequences for household income and mortality had it occurred in 1990 than it did 2019.

As this 1990 counterfactual simulation indicates, the long-term effects of the COVID depend critically on WFH technology being available but not yet fully adopted. Overall, our model suggests the COVID pandemic will lead to higher lifetime income for the working population because it forced many households to work at home which, through learning and adoption effects, boosted WFH productivity. While the measured gains to productivity we report of working at home likely would have happened eventually, the pandemic accelerated this process.

References

Barrero, J M, N Bloom, and S J Davis (2020), “Why Working from Home Will Stick”, ITAM Working Paper.

Bartik, A W, Z Cullen, E L Glaeser, M Luca, and C Stanton (2020), “What Jobs are Being Done at Home During the COVID-19 Crisis? Evidence from Firm-Level Surveys”, Harvard Business School Working Paper 20-138.

Davis, M A, A C Ghent, and J M Gregory (2021), “The Work-at-Home Technology Boon and Its Consequences”, Working Paper, Rutgers University.

Dingel, J I and B Neiman (2020), “How many jobs can be done at home?”, Journal of Public Economics 189: 1-8 (see the Vox column here).

Morikawa, M (2021), “The productivity of working from home: Evidence from Japan”, VoxEU.org, 12 March.

Mortensen, S and N Wetterling (2020), Nordic Real Estate: Remote Work to Permanently Double, Technical Report, DNB Markets.

Ramani, A and N Bloom (2021), “The doughnut effect of COVID-19 on cities”, VoxEU.org, 28 January.

PwC (2021), It’s Time to Reimagine Where and How Work Will Get Done: PwC’s US Remote Work Survey.