by Adam Ozimek and Eric Carlson
Key stats
- A new estimate of remote work from the Bureau of Labor Statistics (BLS) suggests that remote work is less common than previously thought.
- While the new BLS data reveals hybrid remote work to be substantially lower than other surveys estimated, fully remote work is close to where other surveys show, at around one out of ten workers.
- As a result, fully remote work appears slightly more common than hybrid remote work.
- Another new estimate of remote work by the U.S. Census Bureau shows that the highest work-from-home (WFH) places have seen the largest declines in WFH, with some low WFH places seeing an increase. The 10th percentile is now 6.3 percent and the 90th percentile is 29.8 percent.
- Despite some convergence, Southern states still lag with less than 3 percent of workers remote, while more than 30 percent are remote in parts of the Northeast Corridor, the Pacific Northwest, and Northern California
From increased household formation and housing demand to increased fertility rates, remote work has left its mark on the American economy. But with employers calling for more in-office days, is remote work becoming a thing of the Covid past? New data from two government surveys sheds light on how remote work is evolving both at the national level and in communities across the U.S. The first survey from the BLS shows that remote work is both less common than shown in other surveys and also that a greater share of workers are fully remote. Over the last year, remote-working rates have declined somewhat but remain significantly elevated above pre-pandemic levels. The second survey, from the Census Bureau, shows that remote work remains highly uneven across the U.S., with some places having more than a third of their workforce remote while other places having barely any. However, there is some convergence occurring, with the most remote places seeing declines, and the least remote places seeing some increase.
Revising Remote Work
There are a variety of estimates of the share of the workforce that is remote, and these sources have differences owing to sampling methods, coverage universe, and other measurement issues. Last week, the BLS published new estimates of remote work from the Current Population Survey (CPS), which is the monthly survey used to measure the unemployment rate and other important labor market indicators. Starting in October of 2022, the BLS began collecting data on how many people were working remotely, including a question about pre-pandemic workplace. The CPS is a generally reliable survey, making it a useful complement to other surveys. Importantly, the CPS is not conducted online or by email and is thus free of measurement bias that can come from online panels.
The headline is that remote working has still increased dramatically compared to before the pandemic, but is substantially less common than previous surveys suggested. In August 2023, the most recent data available, 19.5 percent of workers ages 16 and up teleworked or worked at home in a given week. In contrast, data from WFH Research, jointly run by the University of Chicago, ITAM, MIT, and Stanford University, and is among the most commonly cited WFH data, suggests 45.9 percent working remotely in that same month.
Figure 1. Source: CPS telework estimates and Barrero, Jose Maria, Nicholas Bloom, and Steven J. Davis, 2021
Part of the gap may be due to the sample universe. For example, WFH Research includes full-time wage and salary workers who earn more than $10,000 a year, while the headline CPS estimate includes all professionals at work, regardless of average earnings. Yet if we look at the full-time remote work rate in the CPS, the remote rate only increases to 20.7 percent, closing very little of the gap.
Another substantial difference between the CPS and WFH Research data is that fully remote makes up a far larger share of remote workers in the CPS. In the CPS, 53 percent of remote workers are fully remote, while WFH Research puts this at 26 percent. Indeed, as a share of workers, both sources are much closer when it comes to the fully remote share of the workforce, with the CPS at 10.3 percent and WFH Research at 12.1 percent. To the extent we should revise down our estimates of remote workers as a result of this new data, it is almost entirely fewer hybrid workers. If the CPS is correct, fully remote is actually slightly more common.
While the downward revision in the share working hybrid remote suggests a smaller change in the economy than previously thought, it is nevertheless still a massive shift. The CPS question about pre-pandemic remote working suggests 9.4 percent of people worked remotely. This share suggests a doubling has occurred. Again, this is substantially below the six-fold increase found by WFH Research, but nevertheless one out of five workers is now remote, and one in ten is fully remote. Looking at the college educated, it is nearly one in five who are fully remote. Among advanced degree holders, nearly 40 percent are hybrid or fully remote. Among skilled workers, remote working is now a substantial share of the labor force, including fully remote.
Importantly, the data shows remote work is stable over the last year, and if anything, has increased slightly. There is no sign that return-to-office is gaining steam on a national scale.
Future work will need to reconcile the sources of discrepancy between the various measures. But the CPS provides another data point in favor of an estimate of somewhere around one in five workers remote, consistent with work from Scott Winship and Thomas O’Rourke, who estimate a 2023 remote work rate of 19 percent using American Time Use Survey, another BLS product.
The shifting geography of remote work
Data from the 2022 release of the American Community Survey (ACS) also suggests remote work remains well above pre-pandemic rates, but did decline in 2022 in most places. However, because the ACS remote work question is somewhat vague, it is somewhat limited for national remote work trends, but the large sample size makes it the best survey for examining geographic patterns of remote work. The data in 2022 shows that the most remote work-heavy places saw the biggest declines, while remote work increased in the 10 percent of places with the least remote working.
It is worth digging a bit more into the limitations of ACS data for better understanding of remote work trends. The biggest shortcoming of this data is that it uses a binary measure of remote working, based on someone’s primary “mode of transportation to work.” However, the biggest advantage of the survey is that the question has been asked consistently over time and that it contains a very fine level of geographic information. The ACS provides the share of workers who work from home at the Public Use Microdata Area (PUMA) level. PUMAs are non-overlapping geographic regions defined by the Census Bureau. Each PUMA covers part of a state and contains at least 100,000 people. By virtue of its granularity, PUMA level data offer a detailed picture of the geography of remote work.
Nationally, the share of workers who work remotely remains well above pre-pandemic levels despite a 3 percentage point decrease from 2021. As shown in Figure 2, the rate of working from home dropped from 18 percent in 2021 to just over 15 percent in 2022, much higher than the 6 percent of workers working from home in 2019. Measurement issues lead to different estimates of the level of remote working at any given point in time. Yet, both higher and lower estimates agree with the broad trend that remote work has declined from the pandemic peak but remains substantially above pre-pandemic levels. This broad trend masks the wide geographic disparity in the prevalence of remote work across the county.
Figure 2. Source: Author’s calculations of ACS data
As shown in Figure 3, 2021 displayed a wide geographic disparity in work from home rates, ranging from 5.5 percent at the 10th percentile to 38.5 percent at the 90th percentile. However, the data also show evidence of convergence in remote work between regions. PUMAs with initially high shares of working from home experienced significant declines in remote work share while PUMAs with the lowest initial shares of working from home experienced increases in remote work. In 2022, 6.3 percent of workers in PUMAs in the 10th percentile worked remotely while 29.8 percent of workers in PUMAs in the 90th percentile worked from home.
Figure 3. Source: Author’s calculations of ACS data
For example, in San Francisco, CA, 53 percent of workers worked from home in 2021 while only 33 percent of workers worked from home in 2022. Similarly, 55 percent of workers in Washington, D.C. worked remotely in 2021 compared to 34 percent in 2022. In contrast, Tarrant County, just outside of Dallas, TX, went from 10 percent remote in 2021 up to 19 percent remote in 2022.
As shown in Figure 4, remote work remains strong in the Northeast Corridor, the Pacific Northwest, and in Northern California but remains low in many Southern states like Mississippi and Tennessee. The five PUMAs with the greatest share of workers working remotely in 2022 were the Northwestern part of Alameda County, CA; the Western portion of Washington, D.C.; the Northwestern part of Wake County, NC; Downtown Seattle, WA; and the Northern part of Arlington County, VA. The five PUMAs with the lowest share of workers working remotely in 2022 were the East Central region of Mississippi; the Northeastern region of Mississippi; the North Central region of Mississippi; South Central Los Angeles, CA; and Ector County, TX.
Figure 4. Source: Author’s calculations of ACS data
Some of the regions that experienced increases in remote work were areas just outside of regions that had high remote shares in 2021 (Figure 5). For example, parts of Texas like Hays County outside of Austin showed increases in remote work, while the city of Austin experienced a decline in its remote share. This change could reflect relocation on the part of remote workers as part of remote work’s “donut effect” on urban areas. However, in general, declines in working from home tended to be large while increases in remote work tended to be modest.
Figure 5. Source: Author’s calculations of ACS data
The reasons for these changes are unclear but can be explored as more data becomes available. One possible explanation is an occupational composition effect. For example, in 2021 people whose jobs could be done remotely worked from home while many with in-person jobs left the workforce. However, as COVID restrictions waned, people with in-person jobs could work again in 2022. Another possible explanation is a relocation effect. Some remote workers may have left dense urban areas for regions with more space. As a result, we see the declining rate of working from home in big cities and the increasing rate of working from home in less urban settings.
Remote work has changed the landscape of the American economy. While its prevalence has declined in some regions, it remains a significant part of how we work today. It is unlikely that the American economy has settled into a new equilibrium. Policymakers will need to pay attention to these shifts as they have implications for housing, transportation, and urban development.
