By Eric Carlson, Benjamin Glasner, Adam Ozimek

Too much research on remote work treats it as one uniform category of working. Although full-remote and hybrid work may seem fairly similar, the differences of their economic impacts are substantial. 

In some ways, hybrid work can be more similar to in-person work, given that coming into the office three days a week still means workers have to live within a commutable distance. What counts as an acceptable commutable distance goes up with hybrid work, but it does not eliminate the question entirely. 

In contrast, full-remote work can sever the connection between individuals and their local labor markets. This distinction doesn’t just matter for individuals but can also be consequential for local economies. The impact is apparent in so-called “Zoomtowns,” where full-remote workers are lured in by strong amenities and cheaper cost of living, despite fewer local job options. Meanwhile, cities with many job options but expensive housing are losing families as full-remote work allows them to move away without giving up their desirable big city jobs.

Using new data from the Bureau of Labor Statistics (BLS), this working paper examines important differences between full-remote and hybrid work arrangements for disabled employment, long-distance migration rates, and housing price growth. 

Key findings include:

  1. Remote work generally improves employment opportunities for disabled individuals. Disabled workers are 22 percent more likely to be full-time remote than otherwise similar workers, while the impact of hybrid remote is half as large. 
  2. Remote work is increasing migration rates, but only full-remote work impacts long-distance migration rates. Hybrid remote, in contrast, impacts shorter distance moves. 
  3. Remote work has different effects on the local housing market, with full-remote work associated with stronger house price growth from 2019-2022 and hybrid remote associated with weaker house price growth. We provide preliminary evidence that these effects of remote work on housing demand are causal.

New Data on Remote Work Intensity

Until recently, the conventional wisdom was that hybrid was the most common type of remote work, while full-remote—where employees rarely or never physically come into the office—was believed to be substantially less common. However, new Bureau of Labor Statistics (BLS) data upends this belief. BLS’ Current Population Survey (CPS) data shows that in September 2023, 10 percent of all workers were full-remote compared to 9.8 percent who were hybrid. These findings contrast with survey data from researchers at WFH Research, who found that only one out of four remote workers are full-remote.

There are three additional reasons why the new BLS data is useful. First, it provides detailed demographic and economic data that enables examination of who exactly is working full-remote and who is working hybrid. Second, the survey has a large sample size and contains geographic information, which allows us to estimate how many individuals work full-remote or hybrid for a sample of metro areas. Third, the survey is conducted in person and over the phone, which avoids the risk of bias that can come from online-only surveys about remote work.

These factors make the new data useful for examining how full-remote versus hybrid matters for three important outcomes: employment of disabled individuals, migration, and housing markets.

Full-Remote Work Draws Workers with Disabilities into the Labor Market

It is no coincidence that the rise of remote work has coincided with the highest employment rates for disabled individuals in 15 years. While research has provided some empirical evidence for the relationship between full-remote work and rising employment among disabled individuals, studies to-date have been hampered by data that fail to distinguish between full-remote and hybrid. Yet, it is easy to see why this distinction is critical for disabled individuals, for many of whom the daily commute and office environment may be insurmountable barriers to work. In these cases, full-remote work can provide a previously unavailable opportunity to work, while hybrid would be of little to no help.

We explore this relationship using the detailed information in the CPS on disability, which suggests that the disabled are substantially more likely to be full-remote. Using regression analysis, we see that this is true even when controlling for education, occupation, industry of work, and a set of other individual characteristics.

A worker who reports any disability is 2.4 percent more likely to be full-remote than an otherwise similar worker. This represents a 22 percent increase in likelihood of being full-remote compared to the population-wide average of 10.7 percent. The likelihood of participating in a hybrid work arrangement among those reporting any disability is 1.1 percent higher, just under half the effect size of full-remote work arrangements.

For some types of disabilities, the distinction between full-remote and hybrid is even larger.[1] The likelihood of full-remote work is slightly higher for those with hearing (1.4 percent) or visual (2.2 percent) impairments, while workers with a disability that limits physical mobility[2] are 4.2 percent more likely to be full-remote.

Those with a physical or mental health condition with difficulties in self-care[3] saw the largest increase in the likelihood of being full-remote, up 5 percent relative to otherwise similar workers[4] —a 47 percent increase relative to the population average.

In contrast, most types of disabled workers were not likely to have hybrid work arrangements, with the exception of those with cognitive difficulties (remembering, concentrating, or making decisions), who had a 2.3 percent increase in likelihood in hybrid work relative to otherwise similar workers.

As employers consider remote work as a means to reach a wider range of potential employees and increase workforce diversity, it is important to understand the significant difference between full-remote and hybrid remote. A labor market that includes a greater number of full-remote jobs will open the door for far more otherwise qualified workers who have been unable to participate due to difficulties beyond their control. In contrast, hybrid remote would provide much less improvement in opportunity.

Only Full-Remote Work Increases Long-Distance Migration

Early evidence suggested that remote work was increasing individuals’ desire to move as early as November 2020, about nine months after shutdowns from Covid began. While subsequent research shows that remote work has influenced migration trends, our analysis directly examines whether the distinction between full-remote and hybrid work had an impact on relocation decisions.

Using the 2023 Annual Social and Economic Supplement for the CPS that includes questions about migration, we examined the difference between hybrid and full-remote’s effects on migration. Using a similar regression model as above, we examine whether workers with full-remote or hybrid jobs are more likely to have moved over the past year than otherwise similar workers. Because migration can be a household decision rather than an individual one, we conduct the analysis at the household level using the share of working adults in the household who are full-remote and the hybrid share as independent variables.

Our results show that full-remote workers are slightly more likely than hybrid workers to have moved over the past year; however, columns two through four in Table 1 show that the effect varies significantly by type of move. We find a statistically significant increase (1.7 percent) in the likelihood that a person moved out of their county if they were working full-remote—a 35.3 percent increase relative to the population average rate of moving out of one’s county (4.8 percent). Though we do not find a statistically significant impact of full-remote work on within-county moves, we do see an increase in likelihood (1.1 percent) of within-county moves linked to hybrid work arrangements, although just shy of significance with a 95 percent confidence level.

Overall, hybrid work appears to have some effect on short-distance migration. In contrast, full-remote makes one substantially more likely to make a longer-distance move. Again, we observe that the effect of remote work varies significantly depending on the nature of the remote work arrangement. The long-term impacts of remote work on migration are yet to be determined, but will surely vary depending on which kind of remote work grows more in the future.

Table 1: Effect of Remote Work on Household Moves

Remote Work Distinctions Reveal Significant, Divergent Impacts on House Prices

The fact that full-remote work has a different impact on migration than hybrid remote suggests that impacts on the housing market may differ as well. Indeed, even with less granular data on remote work, past research has shown that the impact varies by type of area. In particular, more dense and expensive cities tended to see weaker housing demand as a result of remote work while less dense, less expensive places often saw stronger housing demand. Thanks to more data on full- and hybrid remote arrangements, we examined the various impacts on the housing market stemming from different types of remote work, finding a direct impact on prices.

Using the large sample size and geographic detail of the CPS, we calculated the percent who work from home by type of remote work for 280 metro areas. For the housing market, we measured the change in average house price from 2019 to 2022 using the Census Bureau’s American Community Survey.

On average, the simple percent who work from home has an economically small and statistically insignificant impact on house price growth. However, if we measure full-remote and hybrid remote shares separately, we find that full-remote has a positive and statistically significant impact on house price growth while hybrid has a negative impact. A one percentage point increase in the share of full-remote workers is associated with about a 2 percentage point increase in the growth rate of home values between 2019 and 2022. This implies, for example, that a metro with a 10 percent full-remote share had their home values grow by 10 percentage points more than an otherwise comparable metro with a 5 percent full-remote share. In contrast, a one percentage point increase in the share of hybrid remote workers is associated with about a 2 percentage point decrease in the growth rate of home values between 2019 and 2022.

This change may be due in part to reverse causality, reflecting the movement of full-remote workers away from expensive metros rather than a higher share of local workers going full-remote who then drove up home values. Full-remote workers have more mobility as noted above; whether a place is able to attract or retain its full-remote workers can be a sign of strong demand to live there. In contrast, the results show hybrid remote workers are less likely to migrate.

To help disentangle causality, we utilize the industry, occupation, and demographic mix of a metro’s pre-pandemic residents to predict the share of each metro’s workers that would be full- and hybrid remote. Some industries and occupations have relatively higher shares of full-remote while others have relatively higher shares of hybrid remote. Figure 1 shows that while the share of full-remote and hybrid remote workers in an occupation is correlated, there is substantial variation. For example, writers and database administrators have relatively high levels of full-remote work but low levels of hybrid work, while researchers like astronomers, physicists, and natural scientists have the opposite.

Figure 1

To that end, we develop a random forest model to predict whether someone is full-remote, hybrid remote, or neither. The model is trained using 2022-2023 CPS data, which includes the actual remote status among workers, allowing us to generate predictions using individual worker characteristics from 2019 data.

A random forest classifier is a method for sorting people into categories (e.g., remote vs on-site). Like a game of “Twenty Questions,” the algorithm proceeds in a series of steps that each introduces increasingly fine amounts of detail into the classification. By the end of the process, we have a collection of digital flowcharts (each chart is a “tree” and the collection of trees gives us our “forest”) that can sort any new person in our data set into one category or the other. The advantage of a random forest over more common methods is that the algorithm can incorporate complicated interactions between variables without interaction from the researchers to specify them.[5] In this case, the random forest classifier proved a better predictor than other methods like linear probability models. For example, our random forest only has an eight percent error rate when predicting whether workers are full-remote, in contrast to the 12 percent error rate of a comparable linear probability model.[6]

We aggregate these predictions for every metro level and compare them to actual remote work percentages, generating a “predicted full-remote share” and a “predicted hybrid remote share” that is exogenous to the actual work-from-home choices of the people who live in that metro.

We also subtracted these predicted shares from the actual share of remote workers to create residual full-remote hybrid remote shares. This can be interpreted as the extent of each type of remote work above and beyond what would be predicted by the characteristics of the population. In short, we have the following remote variables:

Using this approach, we effectively split the full-remote share into two variables: predicted full-remote and residual full-remote (with the same process for the hybrid share). The models in Table 2 below show that both the predicted full-remote share and residual full-remote share are related to faster house price growth. The predicted and residual hybrid remote shares have consistently negative impacts on house price growth.

Table 2: Effect of Full-Time and Hybrid Remote Work on Home Values

The results suggest that full-remote work is more positive for housing demand while hybrid remote is more negative for housing demand. Even when we split remote work into predicted values based solely on pre-pandemic economic and demographic characteristics of the population—as well as when we use the variation in remote work that is explicitly not explained by those factors—we find the same result.

The data also allows us to examine why full-remote is more popular in some metros than in others. The share of remote work that is full-remote exhibits a strong and statistically significant relationship with pre-pandemic house prices: more expensive metros are more likely to have less full-remote work and more hybrid remote work, controlling for the overall level of remote work. One obvious explanation is that full-remote workers who lived in expensive places tended to move away, leaving those expensive places relatively absent of full-remote workers.

The housing market data illustrates the value of distinction between types of remote work for conducting analysis. While the simple measure of the share of remote workers is statistically insignificant, breaking remote into full versus hybrid remote yields two statistically significant variables that explain the 28 percent of the variation in house price growth in a population-weighted regression, and demonstrates that the effects have different directions.


Remote work’s  significance is only likely to grow over time. While it is tempting to look at full- and hybrid remote as subtle variations on a uniform way of working, the differences between the two can be larger than the difference between hybrid and non-remote work. For individuals with mobility challenges or comparable disabilities, requiring in-office work three times a week is just as insurmountable as working fully in-person. At the place level, hybrid work increases the commute shed of a metro area while keeping individuals effectively tied to that metro area, while full-remote work often enables a complete severance between the physical locations of workers and their employers. These differences have profound implications for housing demand and economic geography writ large.

This analysis provides empirical evidence that the effects of remote work are contingent on the distinction between full- and hybrid remote. Whenever possible, future research should attempt to distinguish between the two or provide an explanation for how heterogeneous impacts are otherwise considered.


  1. It is important to note that the CPS tends to have a lower estimate of the prevalence of disability than other large representative datasets.[]
  2. We use the variable “DIFFMOB” to measure physical mobility, which indicates if a “respondent has any physical, mental, or emotional condition that makes it difficult or impossible to perform basic activities outside the home alone.”[]
  3. We use the variable “DIFFCARE” here, which indicates if a “respondents have any physical or mental health condition that makes it difficult for them to take care of their own personal needs, such as bathing, dressing, or getting around inside the home.”[]
  4. We utilize regression analysis to statistically control for all other worker specific factors that affect the propensity to work remotely, allowing the results to show the impact of the variable of interest compared to an otherwise similar worker.[]
  5. Our model includes information about education, sex, race, age, disability status, industry, and occupation. We use the ranger package in R. Data and replication code are available upon request.[]
  6. To evaluate the models’ predictions, we reserve 80 percent of the data set for estimation (our “training set”) and fit our models to the remaining 20 percent (our “testing set”). The reported error rates are based on predictions using this smaller testing set.[]

Remote Work

Related Posts