Understanding the varied impact of work from home across society is important for businesses, workers, and policymakers alike. This column examines patterns in the US before the COVID-19 pandemic and since. Before the pandemic, the relationship between potential work from home rates and actual rates was weak. During the pandemic, many industries reached their full work from home potential. Post-pandemic, many industries reverted to lower levels but the relationship between potential work from home rates and actual rates remains stronger than pre-pandemic. Long-term adoption depends on additional factors, such as productivity, collaboration, and business needs.
The ability to work from home (WFH) varies greatly among workers. Some job tasks can be performed remotely, while others require a physical presence (Neiman and Dingel 2020). Some firms have a policy that forbids WFH, while others permit or even encourage it (Bick et al. 2023). Some workers’ homes can accommodate a quiet and dedicated workspace, while others cannot. Variation in WFH ability leads to different WFH rates across demographic groups, occupations, industries, and locations (Mondragon and Wieland 2022, Barrero et al. 2023, Bick et al. 2023). As a result, the economic impact of WFH is unevenly distributed.
Understanding the varied impact of WFH across society is an important area of research. This column examines WFH patterns using six major nationally representative US surveys with information on WFH. We analyse which groups of workers were more likely to WFH before the COVID-19 pandemic, how these patterns have evolved since, and the extent to which different data sources produce consistent findings. 1
Consistency across data sources
In our recent paper (Bick et al. 2024a), we document WFH rates in six major nationally representative surveys with information on WFH in the US. Four of these surveys are run by the US Census: the American Community Survey, the American Time Use Survey, the Current Population Survey, and the Survey of Income and Programme Participation. The other two are online surveys by teams of academic economists: the Real-Time Population Survey by Bick and Blandin (2023), and the Survey of Working Arrangements and Attitudes by Barrero et al. (2021).
Because datasets measure WFH differently and survey different population samples, we first harmonise measures and samples across surveys. Using these comparable measures and samples, we find that our six datasets are mostly consistent with one another regarding differences in WFH across demographics (age, sex, education), geography, and firm size distribution. This holds true for both full-time WFH and hybrid work. Below we highlight some of the key patterns we document in our research paper.
One notable trend is that while women already exhibited higher rates of WFH before the pandemic, this gap expanded post-pandemic. For example, in the Real-Time Population Survey, the share of women WFH every workday pre-pandemic was 0.5 percentage points higher than for men, whereas in 2024, this gap had widened to 2.9 percentage points. Similarly, workers with a four-year degree consistently had higher full-time WFH rates than those without such a degree. In the Real-Time Population Survey, this gap grew from 0.5 to 6.6 percentage points in 2024. While the magnitude of these estimates differs somewhat across the datasets, we find similar patterns in all of them.
WFH patterns also vary considerably by firm size. In three of the four datasets (Survey of Income and Programme Participation, Real-Time Population Survey, Survey of Working Arrangements and Attitudes), full-time WFH follows a U-shaped pattern, with medium-sized firms (10–499 employees) having the lowest WFH rates. In the Current Population Survey alone, full-time WFH is similar across small and medium-sized firms and most prevalent in large firms.
Additionally, data from the Survey of Income and Programme Participation and Real-Time Population Survey enable a comparison of WFH rates by firm size pre- and post-pandemic, revealing that WFH increased most in small and large firms. For instance, in the Real-Time Population Survey, full-time WFH rates for small and large firms increased from about 8.6% pre-pandemic to 11.0% and 14.6%, respectively, in 2024. The rise in medium-sized firms was more modest, from 5.2% to 6.7%.
WFH across industries
WFH also differs across industries. To facilitate cross-dataset comparisons, we focus on full-time WFH, which is available in five of our six datasets. Figures 1 (pre-pandemic) and 2 (post-pandemic) illustrate that different datasets provide a consistent ranking of industries by WFH prevalence.
Figure 1 Pre-pandemic WFH rates by industry
Notes: Figure 1 displays heterogeneity in WFH by industry. Each panel plots the pre-pandemic share of workers WFH every workday (‘WFH-only rate’) by industry in the American Community Survey (ACS) on the horizontal axis against the pre-pandemic WFH-only rate by industry in either the Survey of Income and Programme Participation (SIPP) or Real-Time Population Survey (RPS) on the vertical axis. The pre-pandemic baseline is 2019 in the ACS and SIPP and February 2020 in the RPS. Industries are plotted as bubbles with areas directly proportional to their share of employment in the 2019 ACS release. The (rounded) population-weighted correlation of industry-level WFH-only rate in the ACS with industry-level WFH-only rate in another dataset is plotted in the upper left-hand corner of each panel. The 45-degree line is also plotted.
Sources: American Community Survey, Survey of Income and Programme Participation, and Real-Time Population Survey
Figure 2 Post-pandemic WFH by industry
Notes: Figure 2 displays heterogeneity in WFH by industry. Each panel plots the pre-pandemic share of workers WFH every workday (WFH-only rate) by industry in the American Community Survey (ACS) on the horizontal axis against the 2022 WFH-only rate by industry in either the Current Population Survey (CPS), Survey of Income and Programme Participation (SIPP), Real-Time Population Survey (RPS), or Survey of Working Arrangements and Attitudes (SWAA) on the vertical axis. Industries are plotted as bubbles with areas directly proportional to their share of employment in the 2019 ACS release. The (rounded) population-weighted correlation of industry-level WFH-only rate in the ACS with industry-level WFH-only rate in another dataset is plotted in the upper left-hand corner of each panel. The 45-degree line is also plotted.
Sources: American Community Survey, Current Population Survey, Survey of Income and Programme Participation, Real-Time Population Survey, and Survey of Working Arrangements and Attitudes
These figures also reveal that variation in WFH adoption increased sharply after COVID-19. Prior to the pandemic, WFH adoption varied little across industries, with only a few exceeding 15% WFH adoption. By contrast, data from 2022 (the most recent year with available data for all five datasets) show substantial divergence across industries, with some surpassing 40% WFH adoption. Notably, all datasets with pre-pandemic data indicate that the financial activities, professional/business services, and information industries experienced the most significant increases in WFH.
WFH potential versus actual adoption
One natural explanation for industry-level variation in WFH adoption is that WFH is more feasible in some industries than others. To examine this, we use the concept of WFH potential – the share of jobs in an industry that can feasibly be performed entirely from home. Neiman and Dingel (2020) estimated WFH potential across occupations and scaled their findings to industries, identifying the highest potential in education, financial activities, professional/business services, and information.
Given the substantial WFH growth documented earlier, it is natural to ask whether high WFH potential accounts for this trend. For our analysis, we use the Real-Time Population Survey as it is the only dataset that includes pre-pandemic information, high-frequency data during the pandemic, and already available data for 2024.
The pre-pandemic relationship between WFH potential and actual WFH rates was weak. In February 2020, the correlation between these measures was only 0.35, and each percentage point increase in WFH potential corresponded to just a 0.05-point rise in actual WFH (Figure 3a). This changed dramatically during the pandemic: by May 2020, many industries reached their full WFH potential, driving the correlation to 0.94 (Figure 3b). This underscores how industries where WFH was feasible adapted to crisis conditions, using WFH to maintain business continuity and worker safety.
Figure 3 WFH potential and actual WFH by industry in the Real-Time Population Survey
Notes: Figure 3 plots the measure of WFH potential by industry from Neiman and Dingel (2020) on the horizontal axis against the actual share of workers WFH every workday (WFH-only rate) in the Real-Time Population Survey by industry on the vertical axis. Industries are categorised according to NAICS classifications and are plotted as bubbles with areas directly proportional to their share of employment in the 2019 American Community Survey release. The solid line is the line of best fit using these industry weightings, while the dashed line is the 45-degree line. The slope of the line of best fit and correlation coefficient are reported in the upper left-hand corner. Each panel corresponds to one of three distinct time periods: February 2020, May 2020, and 2024 (April and June).
By mid-2024, many industries reverted to lower levels of WFH. However, the relationship between WFH potential and actual WFH had not fully returned to its pre-pandemic state. While the correlation declined from its peak of 0.94 to 0.54 (Figure 3c), it remained substantially above the pre-pandemic value of 0.35. Moreover, each percentage point increase in WFH potential was associated with a 0.12-point increase in actual WFH – below the pandemic peak, but still more than double the pre-pandemic value.
Why have some industries maintained high WFH levels?
Sustained high WFH adoption in financial activities, professional/business services, and information suggests that firms in these industries found ways to integrate remote work more permanently. In contrast, other industries with high WFH potential, such as education, have largely returned to in-person work.
The case of education is particularly striking: despite having one of the highest WFH-potential estimates, the sector has returned to in-person work, likely due to concerns over the effectiveness of remote instruction (McKee et al. 2020). Survey evidence from Japan (Morikawa 2022) further reinforces this view, indicating that WFH is only 60–70% as productive as office work.
The policy and economic implications of persistent WFH
The variation in WFH adoption across industries carries important implications for policy and economic planning. Persistently high WFH rates in certain sectors are reshaping urban labour markets, commercial real estate demand, and public transportation usage (Glaeser 2022). Early adopters of WFH have made significant investments in remote collaboration tools and flexible work arrangements, while others continue to prioritise in-office work.
A key takeaway is that WFH potential sets an upper bound on actual WFH rates, but long-term adoption depends on additional factors, such as the productivity of WFH relative to in-person work. Even in high WFH-potential industries, prominent firms like Amazon and JP Morgan have pushed for a return to full-time office work (Anand and Binnie 2025, Bindley and Rana 2025).
Conclusion
Work from home has become a defining feature of the post-pandemic labour market, but its adoption varies widely across industries. While some industries have permanently integrated WFH, others have returned to traditional models despite its feasibility. This suggests that WFH potential alone does not determine long-run remote work adoption – productivity, collaboration, and business needs remain critical factors. As the labour market continues to evolve, understanding these patterns will be critical for businesses, workers, and policymakers alike.
Source : VOXeu
