The potential stickiness of pandemic-induced behavior changes in the United States

Human behavior is notoriously difficult to change, but a disruption of the magnitude of the COVID-19 pandemic has the potential to bring about long-term behavioral changes. During the pandemic, people have been forced to experience new ways of interacting, working, learning, shopping, traveling, and eating meals. A critical question going forward is how these experiences have actually changed preferences and habits in ways that might persist after the pandemic ends. Many observers have suggested theories about what the future will bring, but concrete evidence has been lacking. We present evidence on how much US adults expect their own postpandemic choices to differ from their prepandemic lifestyles in the areas of telecommuting, restaurant patronage, air travel, online shopping, transit use, car commuting, uptake of walking and biking, and home location. The analysis is based on a nationally representative survey dataset collected between July and October 2020. Key findings include that the “new normal” will feature a doubling of telecommuting, reduced air travel, and improved quality of life for some.


Introduction
Disruptions in our lives present opportunities to learn and practice new ways of doing things, and to re-evaluate old choices and habits (1). The COVID-19 pandemic has been perhaps the largest disruption event in modern human history. Nearly every human on the planet has been forced to modify their habits to adjust to the pandemic, creating an opportunity for long-term change. Importantly, the pandemic has coincided in time with the widespread availability of technologies such as broadband internet service and videoconferencing, as well as many app-based services available through mobile phones.
To provide insights into the potential stickiness of pandemic-induced behavior changes, we developed an extensive survey and collected 7,613 responses in the United States between July and October 2020 (2). The dataset is weighted to be representative of U.S. adults, and captures pre-pandemic, pandemic-era, and expected future behavior in the areas of telecommuting, restaurant patronage, air travel, online shopping, transit use, car commuting, uptake of walking and biking, and home location.
We compare respondent expectations about their own future choices to their pre-pandemic lifestyles. Although we recognize that stated intentions do not always accurately predict future choices, both the survey's design and the choice context itself alleviate this concern.
The survey instrument prompted respondents to provide reasons when they reported that they expect to behave differently post-pandemic than was their pre-pandemic norm. These questions served both as a check on whether a change was actually expected and provided information that informs whether the change is likely to stick.
Further, respondents understand the choice context well. They experienced one lifestyle prepandemic, their daily lives changed during the pandemic, and our future-looking questions ask them how they plan to mix and match the two ways of life. Respondents have experience with both as well as time to reflect on this question during the pandemic, so their answers are wellinformed. Indeed, more than 70% indicated there were aspects of pandemic life they would like to continue.

Telecommuting and its consequences
The most transformative long-term change identified in our data is a large increase in telecommuting. We asked respondents whether they expect to have the option to telecommute post-pandemic, and if so, how often they expect to do so. Therefore, answers reflect individual preferences tempered by expectations about what their employers will allow. The fraction of workers who expect to telecommute at least a few times each week has doubled from the prepandemic period, increasing from 13% to 26% (Fig 1).
A shift to telecommuting is important for its direct impacts on quality of life, worker productivity, and commuting. Among those new to telecommuting at least a few times a week during the pandemic, two-thirds identified telecommuting and/or commuting less often as key features of pandemic life they would like to continue into the future. More than 70% of those new to regular telecommuting report that their productivity has stayed the same or improved during the pandemic, consistent with pre-pandemic research (3). This is remarkable, since many pandemicera telecommuters are juggling childcare and have suboptimal working environments.
The long-term increase in telecommuting is not equitably distributed across the population; the percentage of people with bachelor's degrees who expect to telecommute at least a few times a week post-pandemic is double that of those without bachelor's degrees. Similarly, people from households earning over $100,000 per year are 1.7 times more likely to telecommute. Thus, these quality of life improvements will flow primarily to high-income, highly-educated individuals. The direct impacts of telecommuting on car commuting are substantial. We estimate that less frequent commuting (Fig 1) will reduce car commute kilometers by approximately 15%. The fraction of commuters who choose the car as their primary commute mode is not expected to change substantially.
Telecommuting also impacts transit demand. Commuting accounted for about half of all transit trips pre-pandemic (4). Our data suggest nearly a 40% decline in transit commute trips postpandemic, relative to pre-pandemic. About half of this decline can be attributed to changes in commuting frequency, while the remainder comes from commute mode shifts. A shift to telecommuting will have indirect effects on many aspects of our economy. There is likely to be reduced demand for office space and downtown parking. Patronage of office-district businesses is likely to decrease. Restaurants will continue to be hard-hit. Our data suggest that the number of people who plan to dine in restaurants at least a few times each week will decrease by more than 20% post-pandemic, compared to the pre-pandemic era. Since the restaurant industry employed 8% of U.S. workers pre-pandemic (5, 6), a decrease in restaurant patronage translates to a significant economic hardship for service workers.

A paradigm shift in air travel
Air travel demand dropped 95% at the height of the pandemic, and has only rebounded to 38% of its pre-pandemic level as of February 2021 (7). Our data indicate that more than 40% of business travelers expect to travel less frequently post-pandemic (Fig 2). Of those reducing business travel, two-thirds attribute this change to new realizations that are likely to stick, primarily about the utility of videoconferencing. Personal air travelers also expect to fly less (Fig 2), but nearly half of these reductions are caused by pandemic-related concerns that will likely soon fade.

Accelerated growth of online shopping for groceries
The pandemic has accelerated the uptake of online grocery shopping, nearly doubling the fraction of grocery spending done online (8). We analyzed survey responses from those who tried online grocery shopping for the first time during the pandemic. Approximately half expect to continue to grocery shop online at least a few times a month post-pandemic, but nearly 90% of them also expect to shop in-store for groceries at least a few times a month. This suggests that online grocery shopping does not completely replace in-store shopping. Among all U.S. residents, 30% expect to grocery shop online at least a few times a month post-pandemic, up from 21% prepandemic.
Our data show online shopping for durable goods following a pre-existing upward trend (9). 63% expect to shop for durable goods online at least a few times a month post-pandemic, compared to 59% before the pandemic.

Marked increases in walking and bicycling
Biking and walking have increased during the pandemic in many U.S. cities (10), a change that improves both transport sustainability and public health. Post-pandemic, 30% of U.S. residents plan to take walks more frequently than they did before the pandemic, and nearly 15% plan to bike more (Fig 2). More than 20% identify taking more walks as an aspect of pandemic life they enjoy.
Many cities have provided temporary infrastructure for walking and biking during the pandemic (11). To support a long-term shift, cities could make these changes permanent. Since commuting traffic is not expected to fully rebound, there is an opportunity to reallocate underutilized road space to pedestrians and bicyclists.

Urban exodus?
Some observers project a long-term decline of city centers, as urbanites seek more space and no longer need to commute as often (12). Other research indicates the pandemic has not led longtime urbanites to leave cities (13).
We compare reasons for moving between those who moved from dense urban neighborhoods and all other movers during the first seven months of the pandemic. A higher proportion of dense urban movers were motivated by telecommuting opportunities than other movers. Nearly 20% of dense urban movers cite not needing to commute as a reason for their move, as opposed to 7% of other movers. Likewise, 40% of dense urban movers expect to telecommute at least a few times per week post-pandemic, compared to 27% of other movers.
Notably, dense urban movers were not more likely than other movers to be motivated by either pandemic-related public health concerns or by a desire for a more comfortable home.
The COVID Future dataset strongly suggests that society should expect and be planning for a "new normal." Although only time will reveal the true impact of the pandemic, these data reflect our collective expectations of what the future will bring, providing important insights to help plan for what's next.

Materials and Methods
The COVID Future survey dataset that is the basis for this article was collected between July and October 2020. The study protocol was approved by Institutional Review Boards at both Arizona State University and the University of Illinois at Chicago. Online consent was obtained from all survey respondents.
The data are weighted to represent the U.S. population along the dimensions of gender, age, educational attainment, Hispanic status, income, vehicle ownership, and presence of children. All analysis presented here used these weights. A complete description of this dataset is available (2), and both the dataset and the survey questionnaire are available for download (14). The supplementary materials provide details for all calculations.

Introduction
The results shared in this section of the article are directly calculated from simple tabulations of the survey data, as below.
• "… more than 70% indicated there were aspects of pandemic life they would like to continue." The following tabulation illustrates that the exact percentage of the weighted sample that reported that they would definitely or maybe like to continue some aspect of pandemic life is 73.9%. In the survey, this question was followed by a list of possible pandemic-era lifestyle aspects, and respondents were instructed to select up to three of them. The top choices were "Working from home, at least some of the time", "Taking more walks", and "Spending more time with family". .

Remote work and its consequences
• "The fraction of workers who expect to work remotely at least a few times each week has doubled from the pre-pandemic period, increasing from 13% to 26%." The following tabulations illustrate. Note that these tabulations include only those workers who were employed pre-pandemic and expect to be employed post-pandemic.
. • "Among those new to working remotely at least a few times a week, two-thirds identified remote work and/or commuting less often as key features of pandemic life they would like to continue into the future." The tabulations below indicate the fraction of those new to frequent remote work who value each of these aspects of pandemic life, and also the fraction who value either of them. Because we allowed survey respondents to choose up to three aspects of pandemic life that they value, many who were new to frequent remote work chose both of these. We identify those new to frequent remote work as workers who did not work remotely at least a few times a week pre-pandemic, and who do work remotely at least 2 times per week during the pandemic.
. • "More than 70% of those new to regular remote work report that their productivity has stayed the same or improved during the pandemic." Note in the tabulation below that the "decreased" productivity categories added together sum to 27.7%. There is another category selected by 8.7% of those new to remote work: "in some ways it has increased and in other ways it has decreased". Because there are both effects for these workers, we included them in the "stayed the same" category in our reporting.
. A limitation of the survey in the commuting section is important to mention. Those respondents who were not employed during the pandemic period were not asked what they expect their commute mode to be in the post-pandemic period. In addition, those who commuted for both employment and to school were only asked to provide details on the commute that was their primary activity. In all commuting analyses in this article, therefore, only those survey respondents who were employed in both the pre-pandemic period and the current period, and for whom work was their primary activity, are included.
• "…we estimate that less frequent commuting will reduce car commute kilometers by approximately 15%." The total car commute distance decline is the net result of four effects: 1.
Pre-pandemic car commuters expecting to switch away from cars.

2.
Pre-pandemic non-car commuters expecting to switch to cars.

3.
Pre-pandemic car commuters expecting to increase their remote work frequency without switching modes.

4.
Pre-pandemic car commuters expecting to decrease their remote work frequency without switching modes.
To estimate the total change in car commute kilometers, we estimated car commute distances in both the pre-and post-pandemic periods for those survey respondents who were commuting to work in both periods, and calculated the percent change. To be sure that the change was largely due to less frequent commuting rather than mode switching, we also separately calculated the changes due to items (1), (2), and (3+4) above. The total percent change in car commute distance was 14.5%, which decomposed into a 15% decrease in car commute distance from changes in remote work frequency, and a 0.5% increase in car commute distance from mode switching.
To accomplish this calculation with the COVID Future survey data, however, a number of steps and some assumptions were required.
First, we needed to impute commute distances for the 285 pre-pandemic car commuters who did not provide them, but did provide commute times. To do this, we first calculated the average car commute speed for respondents who reported both car commute distances and times, and used this average speed to estimate commute distances for those car commuters who only reported commute times.
Calculating the average weekly car commute kilometers for private car commuters in the prepandemic period is straightforward: multiply the reported commute distance by the reported number of days per week of commuting.
For the post-pandemic period, the survey data do not include the number of days per week that each person expects to be commuting. The data do include both the pre-pandemic frequency of remote work and the expected post-pandemic frequency of remote work, however, which can be used together with the pre-pandemic number of days commuted per week to estimate postpandemic commute days per week. Multiplying this by the reported commute distance gives us the post-pandemic expected weekly car commute kilometers.
Specifically, we estimate the post-pandemic commute frequency as follows. 1.
If the pre-and post-pandemic frequency of remote work is the same, then we assume the post-pandemic commute frequency is the pre-pandemic number of days per week.

2.
If the pre-pandemic frequency of remote work is "never" or "few times/year" and the postpandemic frequency of remote work is "once/week" or "few times/month", then we assume the post-pandemic commute frequency is the pre-pandemic number of days per week minus one.

3.
If the pre-pandemic frequency of remote work is "never" or "few times/year" and the postpandemic frequency of remote work is "few times/week", then we assume the postpandemic commute frequency is the pre-pandemic number of days per week divided by two.

4.
If the pre-pandemic frequency of remote work is "once/week" or "few times/month" and the post-pandemic frequency of remote work is "few times/week", then we assume the post-pandemic commute frequency is the pre-pandemic number of days per week plus one, divided by two.

5.
If the pre-pandemic frequency of remote work is "once/week" or "few times/month" and the post-pandemic frequency of remote work is "never" or "few times/year" then we assume the post-pandemic commute frequency is the pre-pandemic number of days per week plus one. 6.
If the pre-pandemic frequency of remote work is "few times/week" and the post-pandemic frequency of remote work is "never" or "few times/year" then we assume the postpandemic commute frequency is the pre-pandemic number of days per week multiplied by two. 7.
If the pre-pandemic frequency of remote work is "few times/week" and the post-pandemic frequency of remote work is "once/week" or "few times/month", then we assume the postpandemic commute frequency is the pre-pandemic number of days per week minus one, multiplied by two. 8.
If the post-pandemic frequency of remote work is "every day", then we assume the postpandemic commute frequency is zero.

9.
If the reported post-pandemic primary mode choice is "I expect to work only from home and not commute", then we assume the post-pandemic commute frequency is zero. 10. If the pre-pandemic reported frequency was zero because the person worked exclusively from home, but they expect to commute post-pandemic, then we add 5 commute days per week for those who expect not to work remotely, 4 commute days per week for those who expect to work remotely "once/week" or "few times/month", and 2.5 commute days per week for those who expect to work remotely "few times/week". 11. Finally, we adjust so that any resulting post-pandemic commute frequency that has been estimated to be less than zero is reset to zero, and any that has been estimated to be greater than seven is reset to seven. There are a small number of observations in each category.
The resulting average weekly car commute kilometers for the pre-and post-pandemic periods are illustrated below.
First, we calculate the total change in car commute distance for all of those who said that they commuted by car in either pre-or post-pandemic period AND who had non-missing numbers of commute days in both periods AND that had less than 500 car commute miles per week in both periods.
. . scalar carmiles_change_wfh=pre_carmiles_week3-exp_carmiles_week3 . scalar pct_carmiles_change_wfh=carmiles_change_wfh/carmiles_change_total . display "Fraction of change in car commute distance per week due to changes in > remote work = " pct_carmiles_change_wfh Fraction of change in car commute distance per week due to changes in remote w or > k = 1.0302229 Importantly, in both this and the analysis of transit commute impacts of COVID, we assume that if people are working remotely on a given day, then they are not commuting on that day. Because the COVID Future survey is focused on impacts of the COVID-19 pandemic, we believe that this assumption is valid. A major impact of the pandemic has been that many workers have switched from working at a workplace to working from their homes. Therefore, the pandemic context clearly suggests that respondents to the COVID Future survey likely would interpret questions about remote work frequency to mean remote work instead of commuting to a workplace, rather than remote work in addition to commuting to a workplace.
In both analyses, we further assume that commuters use only their reported primary commute mode on the days when they travel to their workplace.
• "The fraction of commuters who choose the car as their primary commute mode is not expected to change substantially." The following tabulations illustrate that 86% of U.S. workers were car commuters pre-pandemic, and this figure is expected to be 85% post-pandemic.
. For comparison purposes, we determined the percentage of transit trips that were for commute purposes pre-COVID using the 2017 National Household Travel Survey. We subsetted the trips file to only transit trips (TRPTRANS codes 11 public/commuter bus, 12 paratransit, 15 Amtrak/commuter rail, and 16 Subway/elevated/light rail). We did not include boat trips as they may include both ferries and privately-operated boats, and excluded trips where the trip purpose was not included. We computed the weighted proportion of these trips which were for commuting to work or school purposes (either to work or from work).
• "Our data suggest nearly a 40% decline in transit commute trips post-pandemic, relative to pre-pandemic." In order to arrive at this result, we use the assumptions outlined above to estimate the postpandemic commute frequency per week. We then calculate the total number of commute trips per week by transit across all commuters for the pre-and post-pandemic periods, and compute the percent change.
. • "About half of this decline can be attributed to changes in the frequency of remote work, while the remainder comes from commute mode shifts." The analysis behind this statement is exactly analogous to that which we conducted for car commute distance. For transit, we do not calculate the change in distance commuted by transit but instead focus only on the change in the number of transit trips.
The total transit commute trip decline is the net result of four effects: 1.
Pre-pandemic transit commuters expecting to switch away from transit.

2.
Pre-pandemic non-transit commuters expecting to switch to transit.

3.
Pre-pandemic transit commuters expecting to increase their remote work frequency without switching modes.

4.
Pre-pandemic transit commuters expecting to decrease their remote work frequency without switching modes.
Here, we estimate the portion of the total transit commute trip decline that is due to the last two of these. For completeness, the following Stata code calculates the transit commute trip changes from expected mode shifts as well as that from changes in the frequency of remote work. Together these changes equal the total change in transit demand. . scalar trans_trips_change_wfh=pre_trans_trips_week3-exp_trans_trips_week3 . scalar pct_trans_change_wfh=trans_trips_change_wfh/trans_change_total . display "Fraction of change in transit commute days per week due to changes in > remote work = " pct_trans_change_wfh Fraction of change in transit commute days per week due to changes in remote w or > k = .48711312 • "Our data suggest that the number of people who plan to dine in restaurants at least a few times each week will decrease by more than 20% post-pandemic, compared to the pre-pandemic era." This result is derived from the data in the following tabulations. This figure is calculated by dividing the employment in the "food services and drinking places" industry (12.308 million) by total non-farm employment in the U.S. (152.523 million) for February 2020.

Figure 1
Figure 1 illustrates how the pandemic is expected to change demand for remote work and car commuting. The data behind this chart come from the following tabulations. Car commuting frequency here is calculated more simply than described above. Specifically, those who "Always" car commute report a private vehicle as their primary commute mode and that they work remotely "Never" or only "Few times/year". Those who car commute "Most days" are car commuters who work remotely "Few times/month" or "Once/week". Those who car commute "Some days" are car commuters who work remotely "Few times/week". Finally, those who car commute "Never" either work remotely "Every day" or commute using another transport mode.
As explained above, the commuting analysis is limited to those who answered the survey questions about commute mode for both the pre-pandemic and post-pandemic periods.   about the frequency of engaging in these activities pre-pandemic, how that frequency will change post-pandemic, and, for changes in expected air travel and bicycling frequency, we asked why. Specifically, the survey questions were: • "How much do you expect your airplane travel for leisure/personal (business) purposes to change once COVID-19 is no longer a threat, compared to your level of travel before the COVID-19 pandemic?" • "Why do you anticipate an increase/decrease in your long-distance travel for leisure/personal (business) purposes after COVID-19 is no longer a threat? Select all that apply." (separate questions for increase and decrease, depending on the person's actual response to the previous question) • "After COVID-19 is no longer a threat, how do you expect your use of the following means of transport to change, relative to before the COVID-19 pandemic?" (The prompt also included the text "Please include any walks or bike rides for exercise or enjoyment.") • "Why do you expect to increase your use of bicycles? Please select all that apply." The air travel portion of the Figure focuses only on those who had traveled by airplane at least once per year pre-pandemic for leisure/personal and, separately, for business purposes (based on a separate question about pre-pandemic air travel). Those who expect to increase/decrease air travel for each purpose were then asked to select the reasons why, which we separated into "Pandemic-related", "New realization", and "Other" categories, as follows: If a respondent selected both a "New realization" reason and an "Other" reason, their response was categorized as a "New realization" response. If a respondent selected both an "Other" reason and a "Pandemic-related" reason, their response was categorized as an "Other" response. Therefore, those categorized as "Pandemic-related" are those who only selected a "Pandemicrelated" reason.
With these definitions, below are the tabulations of the proportion of the sample in each category for personal and business air travelers that are illustrated in Figure 2.
. The walking portion of Figure 2 is straightforward, since there was not a survey question that asked about reasons for expected increases or decreases in walking frequency post-pandemic.
Below is the tabulation of the proportion of the sample in each category for walking. Figure 2 puts the "Somewhat" and "Much" categories together for both "More" and "Less" walking.
. The biking portion of Figure 2 includes reasons for expected increases in biking, but not reasons for expected decreases. The reasons for expected increases are categorized into "New realization" and "Other" reasons, as follows: .

A paradigm shift in air travel
All of the specific numbers in this section of the article can be derived from the air travel-related information that is represented in Figure 2. Please see Figure 2's explanation for details.

Accelerated growth of online shopping for groceries
We identified those who were new to online grocery shopping as people who reported that they "Never" shopped online for grocery delivery or pickup at the store pre-pandemic, and that they did one or both of these activities within the seven-day period before taking the survey during the pandemic. There are undoubtedly others in this sample that also tried online grocery shopping during the pandemic, but did not happen to do so during the week before taking this survey. That said, the subsample that we have identified here as new to online grocery shopping is 780 people -large enough to draw conclusions from.
The remaining results reported in this section of the article are directly calculated from simple tabulations of the survey data, as below.
• "We analyzed survey responses from those who tried online grocery shopping for the first time during the pandemic. Approximately half expect to continue to grocery shop online at least a few times a month post-pandemic..." .

Marked increases in walking and bicycling
The results reported in this section of the article are directly calculated from simple tabulations of the survey data, as below.
• "Post-pandemic, 30% of U.S. residents plan to take walks more frequently post-COVID than they did before the pandemic, and nearly 15% plan to bike more" The data for this statement is documented in the Figure 2 data description above.
• "More than 20% identify taking more walks as an aspect of pandemic life they enjoy." Those survey respondents who responded that there were at least some aspects of pandemic life that they enjoyed were asked to specify up to three, choosing from a list that was generated based on an earlier survey. This tabulation indicates the fraction of the full sample (i.e. including those who do not enjoy any aspects of pandemic life, so representative of the U.S. adult population) that selected "Taking more walks".
. For the analysis of urban vs. non-urban movers, zip codes are classified as urban if they have a housing unit density of at least 2000 units per square mile. The results reported below of the article are directly calculated from simple tabulations of the survey data, comparing those movers who previously lived in urban neighborhoods to those who previously lived in lowerdensity neighborhoods.
• "Nearly 20% of dense urban movers cite not needing to commute as a reason for their move, as opposed to only 7% of all other movers." The tabulation below presents the percent of movers coming from urban areas who reported that a reason for their move was "I do not need to commute", compared to movers coming from lowerdensity neighborhoods.
. • "… dense urban movers were not more likely than other movers to be motivated by either pandemic-related public health concerns or by a desire for a more comfortable home." Parallel to the tabulation above, the following tabulations present the percent of movers coming from urban areas and from lower-density neighborhoods who reported that a reason for their move was "I did not feel safe sharing the house with others", "I did not feel safe in my building or neighborhood due to the virus", and "Moved to a more comfortable home." .