Global

From L.A. to Boise: How Migration Has Changed During the COVID-19 Pandemic∗

Peter Haslag†

Daniel Weagley‡

March 18, 2022

Abstract

We examine how the broad changes in work arrangements and lifestyles brought on by the COVID-19 pandemic have affected the location decisions of households. We use proprietary data on over 360,000 residential, interstate moves over the last five years with accompanying survey data. We find more than 12% of moves between April 2020–December 2021 were influenced by COVID-19, with a significant shift in migration towards smaller cities, lower cost of living locations, lower tax locales, and locations with fewer pandemic-related restrictions. Higher income households are moving out of more populous cities at greater rates, and they are moving more for lifestyle reasons and less for work-related reasons, consistent with increased location flexibility due to shifts to remote work arrangements. Low income households reasons for moving are relatively unchanged. Among the pandemic-induced movers, over 15% of households are citing remote work as the reason the pandemic has influenced their moves and this proportion has remained steady through 2021. Our results have implications for many outcomes of interest including the future structure of cities, the persistence of remote work arrangements, municipal financing, and real estate.

Keywords : COVID-19, Pandemic, Relocation, Remote Work, Agglomeration. JEL Classification: R23, J61, J11, R10, O15.

∗We are grateful to UniGroup and, especially, Eily Cummings for providing us the data. We also thank Andra Ghent, Andrii Parkhomenko, Jonathan Dingel, and Christopher Smith for helpful comments.

†Owen Graduate School of Management, Vanderbilt University; email: [email protected]. ‡Georgia Institute of Technology; email: [email protected]

Electronic copy available at: https://ssrn.com/abstract=3808326

1 Introduction

The onset of the COVID-19 pandemic has led to broad shifts in remote work arrangements

and lifestyles for a significant portion of the population. Recent theoretical work predicts

these shifts – especially to remote work – will affect where people choose to live and the

structure of cities (Davis et al., 2021; Delventhal and Parkhomenko, 2020; Brueckner et al.,

2021). Motivated by these theories, we examine how households’ motivations for moving

and their location choices changed with the onset of the pandemic. Our analysis provides

insights into people’s expectations about the future of work, lifestyles, and the structure of

cities, and is important information for policy makers, employers and employees.

Consistent with the recent theoretical literature, we find households are moving more

for non-work related reasons, and that they are moving to lower rent, lower tax, and less

populous areas after the onset of the pandemic. Supporting the notion that remote work

arrangements are giving higher income households greater flexibility in where to live, we

find that high-income households are moving more for lifestyle reasons and out of larger

cities – choosing to relocate in smaller suburbs or towns. The re-location of these higher

income households is likely to have a significant impact on local economies and employment

(Moretti, 2010). Many of the changes we document in the pandemic have persisted through

the end of 2021. The lack of a quick reversal in migration patterns and the costly nature of

the moves suggest a significant and ongoing change in location preferences.

For our main set of analyses, we use proprietary move-level data on more than 360,000

inter-state moves within the United States over the past five years to examine how the

nature of re-location decisions of households has changed since the onset of the pandemic.1

The data has detailed information on each individual move (e.g., origin and destination ZIP

code), as well as survey responses on reasons for moving and demographics for a subset of

1The data is provided by UniGroup and covers all domestic moves preformed by the company between January 2017 and December 2021. United Van Lines and Mayflower Transit are brands within UniGroup, C.A.

1

Electronic copy available at: https://ssrn.com/abstract=3808326

movers. The detailed data allows us to not only examine how move origin and destination

locations have changed during the pandemic, but also how households reasons for moving

and the demographics of movers have changed. We supplement our main data and analyses

by examining more representative data from Current Population Survey (CPS) migration

surveys and USPS change of address data, though these data are more limited in sample

size or information on movers. While the main set of moves we study are not representative

of all movers in the U.S., their higher income and interstate nature makes them especially

important with more severe consequences for the impacted local economies.

We find that, while individuals are conducting fewer inter-state moves during the pandemic

than previous years (continuing a downward trend), more than one in seven movers report

that the pandemic influenced their decision to move. The percentage of moves influenced

by the pandemic has persisted even as vaccines have become widely available and the U.S.

has eyed a return to normalcy. In the last quarter of 2021, almost 9% of moves remain

influenced by the pandemic. This suggests that the impact of the pandemic on migration

was not merely a brief shock, but instead will have a longer lasting impact on households’

location decisions.

We further examine how the pandemic shifted households’ migration decisions using

survey data on migrants’ reasons for moving, allowing us to provide direct evidence on how

the drivers of households’ relocation decisions changed with the onset of the pandemic. In our

survey data, respondents were asked to select their main reasons for moving from categories

such as “Job”, “Family”, “Retirement”, “Lifestyle”, “Health”, and “Cost of Living”. These

reasons can be broadly categorized into work reasons and lifestyle reasons. One might expect

the broad shift to remote work arrangements during the pandemic to give individuals more

flexibility on where they can work from, decreasing the shadow cost associated with work

location proximity and increasing the importance of quality of life reasons, which would lead

to more non-work related reasons for moving. On the other hand, job loss was a common

theme in the pandemic, which may lead to more work-related moves. We find households

2

Electronic copy available at: https://ssrn.com/abstract=3808326

are moving much less for new jobs or company-transfers during the pandemic and relatively

more for family and lifestyle reasons (as compared to the pre-pandemic period). These

results are corroborated when looking at CPS data on reasons for moving. We find a similar

drop in job-related reasons for moving during the pandemic both for interstate and intrastate

moves.2 The results are consistent with there being an increased location flexibility from the

broad shift to remote work.

Considering higher income households are more likely to work in remote work-capable

occupations (Bick et al., 2021), we test for whether higher income households experience

a greater shift in their motivations for moving. We find the higher income households are

moving much less for changes necessitated from work (such as job loss or taking a new job)

and much more for non-work related reasons during the pandemic. In contrast, lower income

households are continuing to move for job-related reasons at a similar rate and are less likely

to move for reasons such as retirement, health or lifestyle. In sum, higher income households

have been motivated to move more for non-financial reasons (health and lifestyle), while

lower income households are moving more due to financial reasons (work and cost-of-living)

during the pandemic. The differences across income groups mimic broader trends on the

impact of the pandemic, where economically disadvantaged households are more negatively

affected. Beyond income, mid-career households and those with more than three individuals

(likely families) are also moving less for work and more for lifestyle. The greater shift in

motivations for higher income and mid-career individuals is both surprising and meaningful,

as these groups have relatively higher human capital and tend to be less geographically

mobile.3 This represents a significant shift in the spatial reallocation of human capital.

We gain greater insight into how COVID-19 affected households’ migration decisions by

examining “free responses” for a subset of survey respondents. Responses such as “COVID-19

2In the CPS data, we calculate the proportion change in reason for moving and find the correlation between interstate and within-state movers to be 0.51.

32017-2018 Census mobility data shows workers making $100,000+ are 1.2pps less likely to move, or a 12% relative reduction in the likelihood of moving, as compared to the unconditional probability of moving. Those 30-64 years old are 1.44pps less likely to move.

3

Electronic copy available at: https://ssrn.com/abstract=3808326

and its subsequent requirement to work remotely, eliminated my need to pay high rent in

a prime area near my workplace.” illustrate the impact of the pandemic on households’

location preferences. We classify individuals free responses on how COVID-19 affected their

move and find the three most common reasons are related to family (typically, the desire to be

closer to family), the ability to work remotely, and job loss. The ability to work remotely has

been an increasing factor since April 2020, growing to around 20% of pandemic-influenced

respondents citing remote work as a main determinant of their move. The persistence of this

pattern through December 2021 suggests that remote work arrangements will continue to

play an important role in migration patterns going forward. Again, we find the motivations

for moving differ across income groups: 70% of respondents who mention the ability to work

remotely in their free response are in households earning over $100,000 per year, despite

representing only 42% of the sample of movers.

Interestingly, the local spread of COVID-19 does not appear to be a major motivator,

with only 2% of respondents mentioning the local spread of COVID-19 as motivation for their

move. Restrictions put in place by state and local governments were more than three times as

likely to influence a households’ move – even though still relatively small at 6.7%. These more

temporary shocks related to the spread of the disease and the government’s response, play a

relatively minor role in relocation decisions, while the more permanent shifts in households’

day-to-day lives play a much more significant role in re-location decisions post-pandemic.

We further explore how the pandemic shifted people’s location preferences through

revealed preferences. A unique aspect of the UniGroup data is that we can observe the

origin-destination pair at the zip code level. Utilizing differences between the features of

origin and destination locations and identifying systematic differences through time allows

us to shed light on how the relative value of certain area features changed post-onset of the

pandemic. We find migrants are moving relatively more to less dense areas and lower cost of

living (taxes, rent) areas than in the pre-pandemic period. We find households are moving to

less populous CBSAs during the pandemic and that the movement to less populous CBSAs

4

Electronic copy available at: https://ssrn.com/abstract=3808326

is concentrated among origin cities with greater remote work capability. In sum, there was

a relative exodus out of dense, costly, highly populated cities towards smaller, less costly

cities.

Prior to the pandemic, there was a growing concentration of high income individuals

in the largest cities (Gaubert et al. (2021); Moretti (2012)). Theoretical models of cities by

Davis and Dingel (2019) and others, predict that higher-income households are drawn to

larger, more expensive cities because of the agglomeration benefits of idea sharing, which

leads to higher individual productivity. With the onset of the pandemic, there were many

changes to work and social arrangements that potentially altered the direct benefits of

agglomeration, particularly the idea-sharing productivity gains. Thus, we examine if the

trend of higher income individuals concentrating in large cities has at least partially reversed.

We find higher-income households are leaving larger, more expensive cities at a higher

rate and landing in less populated areas during the pandemic. This relative exodus of

higher-income individuals out of more expensive cities has the potential to reduce the idea

sharing agglomeration benefit of living in a major city (e.g. Davis and Dingel (2019)).

Overall, our paper makes three key contributions to the literature and our understanding

of how shifts in migration patterns were brought on by the pandemic. First, we document

that the pandemic has continued to impact migration decisions almost two years after

the onset, and the ability to work from home is an increasingly important and persistent

reason for people to move. Second, the reasons people are moving have shifted during

the pandemic towards non-work related reasons, particularly for higher income, mid-career

individuals. The differences in motivations across demographic groups reinforce remote work

arrangements playing a key role. Third, and related, we show the relative importance of

particular location features have changed with migrants putting greater emphasis on rent,

taxes and density post-onset of the pandemic. Understanding the reasons households are

moving – for more permanent reasons like lifestyle and remote work, and not for temporary

reasons, like the spread of the virus – helps predict migration flows and how human capital

5

Electronic copy available at: https://ssrn.com/abstract=3808326

will be spatially allocated in the future. The implications of which are key inputs for

local government decisions, firm decisions, and real estate development. As human capital

reallocates, these factors will likely continue to influence decisions such as investments in

public goods and municipal debt issuance, real estate investments, and how firms will

structure work arrangements for their employees.

2 Literature Review

There has been broad interest in how the pandemic has changed work arrangements, migration

and housing markets with a number of relatively contemporaneous papers written on the

subject. For instance, several papers have shown residential and commercial real estate

prices changed dramatically through the pandemic.4 A main theme from this literature

is that house prices in the suburbs benefited at the expense of city centers. While real

estate prices are an important outcome to study, movement in house prices does not provide

direct evidence on the motivations behind migrants moves. Studying migration and migrants

motivations, therefore, nicely complements the real estate analysis by relying on direct action

and is less subject to market conditions like low housing supply.

Our data on interstate migration flows and households’ motivations for moves allows

us to contribute to the literature in ways other papers can not. The interstate moves we

study are a costly endeavor for many reasons and, therefore, are more likely to capture

permanent re-locations of individuals and reflect expectations of long-term work and living

arrangements. Another advantage of the UniGroup data is that we can observe origin-destination

pairs, which allows us to examine changes in the relative value of locational features through

time. Furthermore, we can link moves to survey responses, which provides greater insight

into who is moving, why they are moving, and how motivations differ across demographic

4Examples include, but are not limited to: Ling et al. (2020), Liu and Su (2021), Gupta et al. (2021), Guglielminetti et al. (2021), Ramani and Bloom (2021), Rosenthal et al. (2021).

6

Electronic copy available at: https://ssrn.com/abstract=3808326

groups.

There has also been a growing literature that takes an interest in the role of remote

work on migration. Recent theoretical work predicts that large scale adoption of remote

work will shift migration patterns both intracity as well as intercity (Davis et al. (2021),

Delventhal and Parkhomenko (2020), Brueckner et al. (2021)). We find people are moving

in response to the pandemic and the shift to remote work, and choosing to move out of large,

expensive cities. We are one of the first papers to show the direct link between migration

and remote work capabilities on a large scale. Thus, our empirical results complement such

predictions for intercity migration. Survey data from Ozimek (2020) show 14-23 million

Americans are planning on moving in response to remote work capabilities, with more than

half moving more than two hours away. Again, the UniGroup data corroborates this evidence

and gives more clarity to the motivations for such moves. In contemporaneous related work,

Ramani and Bloom (2021) use USPS change of address data to document a “donut effect”

of the pandemic on where people are living. They find a sizeable increase in population in

city suburbs and a decline in population in urban areas mainly in the largest U.S. cities.

While intracity moves are not the focus of our paper, we do find results consistent with

and supporting of Ramani and Bloom (2021) and the theoretical work mentioned above.

Specifically, we find a similar preference for suburban versus urban areas in our interstate

moves data with a greater frequency of moves out of urban cores and into suburban or rural

areas.5,6 Compared to their analysis, we are able to observe where people are moving from

and to, examine reported motivations for moving and examine differences in behavior and

motivations across movers demographics.

Overall, our results support many of the theoretical predictions and corroborate survey

evidence that remote work is having a significant impact on the migration of workers and

5See Figure A2 in the Appendix. 6In Section 4.3.1, we use the USPS change of address data to examine how our measure of state-level

changes in origin and destination flows compares to changes observed in permanent moves in the USPS data. We find a correlation of approximately 0.36 and 0.48 for origins and destinations, respectively.

7

Electronic copy available at: https://ssrn.com/abstract=3808326

is likely to continue to play a role in migration decisions in the future as well. At the same

time, we are able to offer a broader understanding of these shifts by relying on our unique

data.

3 Data and Summary Statistics

3.1 Move and Survey Data

Our main data on moves is sourced from UniGroup, C.A. The data includes all domestic,

non-military moves performed by MayFlower and United Van Lines brands from January

2017 through December 2021, with details including the ZIP codes of origination and destination,

dates, and freight weight. We focus our analysis on interstate moves, which represent 98.6%

of moves performed by UniGroup. Our sample contains just over 360,000 interstate moves,

which is roughly 5% of all interstate moves during this time period.7 Only 3.5% of the moves

are partial moves suggesting the vast majority of the moves we examine are permanent in

nature. We provide the time-series of moves in Figure A1 in the Appendix.

For approximately 25% of moves (or approximately 90,000 moves), movers completed a

survey with questions on the main reason(s) for their move, income bracket, age bracket, and

size of household.8 The survey is mainly used to capture additional details about individuals

who moved and their satisfaction with their experience.

In June 2020, UniGroup began including additional survey questions regarding how

COVID-19 affected the respondent’s willingness to move.9 These data include both form

and free responses citing particular reasons for their move. We categorized the over 3,400

responses from March 2020 to December 2021 using the following procedure: (1) we perform

7Interstate moves represented 14% of all moves in the U.S. in 2019 8Some respondents provide only one piece of demographic information, so total responses differ across

demographic variables. 9UniGroup also reached out to customers who moved prior to June to solicit additional responses. United

Van Lines’ assessment can be found at https://www.unitedvanlines.com/newsroom/covid-moving-trends.

8

Electronic copy available at: https://ssrn.com/abstract=3808326

a preliminary read through of the responses to create 13 categories, (2) assign three research

assistants to separately classify the free responses into one or more of the 13 categories,

(3) if ≥2 of the research assistants agree on a category, then the response is given that

categorization. We allow for responses to have multiple categorizations.

3.2 Additional Migration Data

In addition to the UniGroup migration data, we utilize two additional migration data sets.

We use CPS ASEC (Annual Social and Economic Supplement) migration survey data from

2017-2021. The CPS surveys are collected and weighted to give a representative view of

the U.S. on a host of issues. Within the CPS surveys, ASEC surveys are given to a subset

of respondents. The data contains locations at the state-level and reasons for moving for

both interstate and within-state moves. This allows us to calculate how reasons for moving

changed during the pandemic for both interstate and intrastate moves. The set of reasons

for moving that respondents can select from in the CPS data does not perfectly overlap

with the UniGroup data, but there is enough similarity to make general comparisons across

the two data sets. The data contains demographics, reasons for moving, and state-to-state

migration pairs as well. The most recent survey was conducted in 2021 and will capture

migration during 2020.

To understand how data correlates with a broader movement of individuals, we use

data provided by the USPS on change of addresses from January 2017 to December 2021

(originally obtained by Ramani and Bloom (2021) through a FOIA request). Specifically,

we use the permanent number of moves to a particular state and from a particular state in

the pre- and post-pandemic periods. The data encapsulates both interstate and intrastate

migration, but does not distinguish between them.

9

Electronic copy available at: https://ssrn.com/abstract=3808326

3.3 Discussion of the Migration Datasets

Our main data set covers interstate moves conducted through a large moving company,

UniGroup. There are a few limitations with our main sample of movers. First, our sample

does not have intra-city moves such as those from the center of a city to less dense, lower

price-per-square-foot suburban neighborhoods. While there is considerable debate about

the geographical-level at which to examine migration, our focus on interstate moves is

not uncommon (Molloy et al., 2011). The second limitation is that the moves are not

representative of all migrants. In Section 3.5, we compare the UniGroup population of

movers to the CPS set of movers and find the UniGroup sample is tilted towards higher

earning, older households.

By supplementing our main analysis with analysis of the CPS and USPS data, we are

able to address both of these issues. Because the CPS data has both interstate and intrastate

moves, we can examine whether there were similar changes in reasons for within-state moves

as across-state moves during the pandemic. In addition, we can examine how reasons for

moving have changed for a more representative set of migrants ensuring our results are not

contained to our particular sample of movers.

While there are some clear benefits to the CPS data, there are two key limitations

compared to our main data. The CPS sample is much smaller with some states only having a

few respondents each year, and (2) the data only extends through 2020, unlike our UniGroup

data that goes through December 2021. The smaller number of responses may provide

noisier estimates of changes in reasons for moving and makes examining one-year changes in

migration patterns difficult (especially at the state level).

We use the USPS change of address data to examine broader, more representative

migration patterns. The data includes all movers who utilize USPS change of address services

giving it a broader coverage of migration patterns. We compare our interstate moves to this

broader set of overall migration. The USPS data is very limited in that it does does not

10

Electronic copy available at: https://ssrn.com/abstract=3808326

include origin-destination pairs information, demographics, or stated reasons for moving, so

we cannot use it as a comparison set for most of our analyses.

Although the UniGroup data is not representative of all movers, the subset of movers

we study are an especially important subset to study. These individuals are more likely to

be switching local labor markets and tax jurisdictions. Moreover, due to their above average

income level, these re-locations will have more severe consequences for local consumption,

which can lead to new employment opportunities (Moretti, 2010), and collected taxes.

Despite higher than average wages, we are able to exploit heterogeneity in income within

our sample to conduct cross-sectional analysis across income groups.

Finally, one may be concerned that there may be idiosyncratic, time-varying demand

for the services of the particular moving company providing our data. Any variation of this

kind is unlikely to materially affect the interpretation of our results. The bulk of our analysis

is looking in the cross-section or accounts for variation in the number of moves over time.

In this way, we remove any common variation in demand for our moving company’s services

over time. We have no reason to suspect that there are any local-level demand shocks for

the services of the moving company or that any changes in demand would likely arise from

systematic factors that would impact our interpretation of the results.

3.4 Data on Locations

In order to understand how location characteristics are related to the location decisions

of moving households, we collect location-specific characteristics data for the origin and

destination locations in our sample.10 We utilize ZIP-code level telework proportions using

data from Su (2020).11 Density and population data is collected from the 2010 Census at

the ZCTA-level. We collect median rent at the ZCTA-level from Manson et al. (2020). State

10As a base we utilize Anthony D’Agostino’s improvements on a zip-county crosswalk - (https://anthonylouisdagostino.com/a-better-zip5-county-crosswalk/).

11We thank Yichen Su for sharing this data. For some tests we aggregate the data to the CBSA-level. This measure is highly correlated with the measure in Dingel and Neiman (2020).

11

Electronic copy available at: https://ssrn.com/abstract=3808326

tax data is collected from NBER’s tax rate database (https://users.nber.org/~taxsim/

state-tax-rates/real.html). We use 2018 data for the marginal income tax rate for an

income of $100,000. For some tests, we aggregate these finer geographic measures up to the

CBSA. Population and work from home employment are naturally aggregated by summing

variables, while rent is taken as the simple average across the CBSA.

We collect data related to the spread of COVID-19 and local government responses

to measure the severity of the COVID-19 outbreak across locations and the stringency of

the government response. Data on daily COVID-19 cases and deaths by county is from

the New York Times.12 The per capita calculation is scaled by 10,000. Data on state-level

COVID-19-related restrictions are collected from Hale et al. (2020). In particular, we use the

Stringency Index which counts the number of categorical restrictions based on containment or

closure policies. We use the COVID-19 cases per capita and Stringency Index as of July 1st,

2020 to capture cross-sectional variation in COVID-19 spread and government stringency.

We chose July 1st, 2020 as most locations had experienced the first wave by this point.

Indices as of July 1st, 2020 are highly correlated with the cross-sectional sorting through

time, so the point-in-time measure proxies well for cross-sectional ranks over time.

3.5 Summary Statistics

Panel (A) of Table 1 summarizes the data. The majority of these moves are of significant

distance. The 25th percentile of move distance (Actual Miles) is 603 miles and the average

distance is 1,191 miles, which is approximately the distance from Los Angeles to Seattle.

Given the interstate nature of the moves, these moves are not individuals moving from

within city centers out to the suburbs. We provide separate summary statistics for origin

and destination locations for the percentage of occupations with remote work capabilities

(Remote Work), state-level Stringency Index (July ‘20 Stringency Index), density (Density),

12https://github.com/nytimes/covid-19-data

12

Electronic copy available at: https://ssrn.com/abstract=3808326

population (Population), median rent (Rent), tax rates (Tax Rate), and COVID-19 cases

per capita (July ‘20 Cases Per Cap.).

Comparing the characteristics of the origin and destination locations and changes in

the mean values of these characteristics between pre-pandemic and during the pandemic

highlights many of the key results from the paper. The difference between the mean values

for moves pre-pandemic and mean values for moves post-onset of the pandemic (all moves

on or after April 1, 2020 are in the post-period), along with the statistical significance, are

in the last two columns.13 We find movers are leaving locations with slightly higher remote

work ability during the pandemic and that they are leaving zip-codes that are more dense

and heading to destinations that are less dense. The economic magnitude of the change in

destination density is staggering, as it represents a 13% decrease as compared to the overall

mean. People are also leaving more populated CBSAs and moving to less populated CBSAs

on average during the pandemic.14 This result echoes anecdotal evidence on intra-city moves

from the urban core to the suburbs and is consistent with related research by Ramani and

Bloom (2021); Su (2020); Gupta et al. (2021); Guglielminetti et al. (2021). In a similar

vein, we find households are moving away from more expensive areas and relocating to less

expensive areas. In terms of state tax rates, there is no statistical difference in the average

origin tax rate, but we do find that households are tending to relocate to lower tax states

during the pandemic.

For COVID-19-related stringency measures and COVID-19 cases per capita, comparing

the pre-period and post-period is not informative since there were zero COVID-19 cases and,

therefore, no COVID-19 response pre-pandemic. However, we can compare the means during

the post period between origin and destination (i.e., column “Post Mean”). We see that on

13To test statistical differences, we double cluster standard errors by year-month and origin or destination state, depending on the outcome of interest. We use the delivery date as the relevant date. Our results are virtually unchanged if we use registration dates.

14Appendix Figure A2 corroborates this interpretation. The figure shows a steep drop in moves to urban destination zip codes across all origin types and an increase in moves to suburban and rural zip codes. Urban destinations have seen a slight uptick in 2021 but still remain below pre-pandemic levels.

13

Electronic copy available at: https://ssrn.com/abstract=3808326

average households are leaving locations that have stricter government mandates (as of July

2020) during the pandemic and moving to areas that have fewer mandates. Examining

COVID-19 cases per capita, we see households are moving to areas that have relatively

lower COVID-19 incidence rates compared to the areas they are leaving. Finally, to get a

more dynamic understanding of how infection rates and the government response are related

to migration patterns, we calculate moving averages over the 30-days prior to the move for

stringency and cases per capita. We find a similar pattern in the moving averages for the

Stringency Index as when we use the July 2020 value. The difference between origin and

destination COVID-19 cases per capita dissipates when we use the moving average though.

Panel (B) of Table 1 documents summary statistics of demographic variables recorded

from the survey data. The survey response rate hovers around 25% and about 95% of survey

respondents are willing to provide some demographic information. The key areas of interest

are household size, income, and respondent age. The movers in our sample tend to be

older, with a median age between 55 to 64; higher income, with a median income of at least

$100,000; and have smaller families, with a median household size of 2.

To examine how the UniGroup data compares to a representative dataset of all movers,

we compare the demographic distribution of UniGroup movers to the demographic distribution

of interstate movers in the CPS data set during the March 2017 - March 2021 time period.

Results are presented in Panel A of Table A1 in the Appendix. We find the UniGroup sample

is tilted toward older and higher income individuals as compared to the general population

of movers. For instance, 42% of CPS respondents are under the age of 35, while only 14% of

our moves are of the same age group. For those who respond, 63% of the UniGroup sample

are in households making more than $100,000, while only 33% of the CPS sample falls into

that category. The survey respondents in our sample are typically from smaller households

with 75% of households consisting of less than 3 individuals as compared to 53% in the CPS.

In sum, the UniGroup sample is tilted towards longer, inter-state moves and smaller,

14

Electronic copy available at: https://ssrn.com/abstract=3808326

higher-income households compared to all movers in the U.S. While the UniGroup population

is not necessarily representative of the entire population of movers, the tilt towards higher

income, remote work-capable individuals allows us to focus our main analysis on the subset

of movers that are likely to have the greatest impact on the local economy. They are arguably

the moves which are most likely to disrupt housing, taxes, and political structures. Moves

by these types of households may affect the costs and benefits of certain locations for lower

income households over time as well. Even though our main sample is less representative of

lower income households, the patterns we observe for this demographic within our sample

should be similar to what is happening for do-it-yourself cross-state moves although we are

hesitant to over-extrapolate.

4 Results

4.1 How have motivations for moving changed during the pandemic?

We begin by investigating how migrants motivations for moving have changed following

the onset of the pandemic. In Panel A of Table 2, we document the proportion of survey

respondents that selected each reason for moving.15,16 We contrast the proportional response

rate during the pandemic (starting April 1, 2020) with the rate in the period prior to the

pandemic.

The most significant increase was for “Family” reasons with a 6.6 percentage points

increase (or 27% increase from 24.7% to 31.3%). The increased proportion of family-motivated

moves could partially be due to the desire to create social “bubbles” with family during the

pandemic (e.g., for grandparents to help with childcare). Also, being closer to family became

a more viable option for people in jobs that shifted to remote work. A similar reasoning

could be driving the increase in “Lifestyle” (13.7% to 16.1%) reasons. We see a modest

15Cost of Living was added to the survey in June 2020. 16Respondents can select one or more primary reasons for moving.

15

Electronic copy available at: https://ssrn.com/abstract=3808326

increase in “Retirement” (19.7% to 21.2%) reasons likely due to the uncertainty, job loss,

and shift in work arrangements brought on by the pandemic. “Health” shows only a minor

increase during the pandemic. This provides some of the first evidence that the spread of

COVID-19 did not play a major role in households’ relocation decisions. The most drastic

change was the 13.5 percentage point decrease for reasons related to new jobs or company

transfers (“Job”).

The observed changes in reasons are perhaps the result of decreased labor mobility

during the pandemic, the shift to remote work decoupling jobs and locations freeing people

to locate based on non-work considerations, or a more general shift in moving to fulfill

non-work preferences as opposed to an emphasis on labor or income. Many of the factors

affecting reasons to move during the heart of the pandemic are likely to persist, especially the

shift towards remote work. Considering the frequency of partial moves decreased during the

pandemic to only 3.5%, many of these moves are likely to be permanent moves and reflect

permanent shifts in households’ location preferences.

In Panel A of Appendix Table A2, we further examine how the reasons for interstate

migration have changed using the more representative CPS data. The set of reasons does

not perfectly map between CPS and UniGroup and in the CPS survey the respondents

are only able to select one reason. That said, we see broadly similar patterns in the CPS

sample as in the UniGroup sample. There is a large decrease in work-related reasons for

moving with an 8.8 percentage point drop in “New Job or Job Transfer”, a 1.3 percentage

point drop for “Other Job-related Reason” and a small 0.2 percentage point increase in “To

Look for Work or Lost Job”. The reason with the second largest decrease is “For Easier

Commute,” this is consistent with the shift to remote work decreasing the importance of

commute distance for some households. We find small increases in “Other Family Reason,”

“Change in Marital Status,” and “Relationship with Unmarried Partner.” These reasons are

likely captured by “Family” reasons in the UniGroup data, suggesting there are a number

of potentially interesting shifts in family-related moves during the pandemic. Finally, we

16

Electronic copy available at: https://ssrn.com/abstract=3808326

find many housing-related reasons experienced an increase in relative frequency during the

pandemic. Reasons such as “For Cheaper Housing,” “Wanted Better Neighborhood,” and

“Want to Own Home, not Rent” all increased by at least 1.2%. These shifts are consistent

with the observed increase in lifestyle reasons for moving in the UniGroup sample.

One limitation of the UniGroup data is that it is isolated to interstate moves and,

therefore, may not capture broader migration trends in the reasons for moving, limiting our

interpretation and the broader external validity. To address this concern, in the Appendix, we

perform the same exercise of comparing changes in reasons for moving across the post-pandemic

and pre-pandemic periods, focusing only on intrastate moves using the CPS data.17 We

find similar, though more mooted changes for intrastate moves. There is a relatively high

correlation between the interstate and intrastate changes in reasons of 0.51. Overall, these

results suggest there are broadly similar shifts in motivations for moving during the pandemic

for within-state moves as for across-state moves.

Returning to the UniGroup data, we examine the proportion of moves that explicitly

state being influenced by the pandemic. Panel B of Table 2 shows the proportion of

respondents who indicate that COVID-19 influenced their move by month. The proportion

remained under 6% through July 2020 then spiked and stayed between 11-19.3% from August

2020 through August 2021.18 Even in the last quarter of 2021 – almost two years after the

onset of the pandemic and several months of increased vaccination rates – a significant

proportion of moves were motivated by the pandemic (8.7%).19 The cumulative effect of

these likely permanent moves suggests a significant reallocation of individuals in response

to the pandemic. To the extent that we can extrapolate from this single company’s moving

17Results presented in Panel B of Appendix Table A2. 18UniGroup began surveying movers on the influence of COVID-19 in August 2020. For months prior to

August, the company attempted to re-survey previous respondents. This can potentially account for the large jump in pandemic-related moves in August 2020.

19As of this draft, UniGroup has only collected a fraction of the expected response rate for moves that end in December 2021. The proportion of COVID-19 influenced moves tends to increase as more surveys are collected. For instance, in a previous iteration of the paper with partial data for November 2020, we reported 11% of moves were influenced by the pandemic during the month of November 2020. The proportion jumped to 19.3% in this version.

17

Electronic copy available at: https://ssrn.com/abstract=3808326

data, and given 4 million individuals move across state lines in a typical year, a conservative

estimate implies that more than seven hundred thousand individuals will have moved across

state lines since the summer of 2020 in response to the pandemic.20 The persistence of

COVID-19-influenced moves suggests many more households will choose to relocate due to

the ongoing societal shifts brought on by the pandemic.

We next examine movers’ free responses, which provide a significant amount of additional

insight into how the pandemic has influenced peoples moves. These responses are provided

by movers citing the pandemic as a reason for their move. This group was asked to provide

free responses detailing how COVID-19 has impacted their decision.

Panel C of Table 2 documents the proportion of free-response answers that are classified

into each category. The most common COVID-19-related reason for moving is “Family,” at

31.8%, which mimics the broader survey. Strikingly, the direct risk of the disease was not a

main driver of moves as evidenced by only 2% of respondents mentioning the local infection

rate. This could be due to several reasons: the wide-spread and somewhat unpredictable

nature of the spread of the virus; individuals not expecting COVID-19 infection rates to

persist longer-term; or individuals not weighing the risk heavily in their decision.

We do find a small, but not insignificant, number of moves were driven by the government

response to the pandemic, which is consistent with households “voting with their feet”

(Tiebout, 1956). We see moves in response to both governments being too lax (only 1.4%)

and governments being too strict (6.7%). Of the households moving because the government

response was too strict, 94% left democratic states.21,22

“Work from Home” and “Job Loss” were the second and third most common reasons,

representing 17.2% and 14% of respondents, respectively. While both of these reasons are

20Using a conservative average for the COVID-19-influenced move rate from March 2020-December 2021 (10%) and 1.75 year’s worth of moves (4.7 million in 2018-2019), we arrive at 700,000.

21We classify states based on their voting in the 2020 presidential election outcome. 22Other corollary effects such as Social Unrest and Politics reinforce the fact that the pandemic has become

a political issue for a portion of the population.

18

Electronic copy available at: https://ssrn.com/abstract=3808326

“job-related”, the underlying motivations are very different. Considering the costs associated

with moving hundreds of miles and across state lines, the moves motivated by “Work from

Home”-related reasons are likely due to households expecting a permanent change in work

structure, not just a temporary shift during the height of the pandemic. These effects

are likely to have an impact on the geographic structure of labor markets (local versus

national labor markets) and the geographic dispersion in incomes. In particular, Azar et al.

(2020) and Marinescu and Rathelot (2018) show that labor markets are largely local and

fairly concentrated. Our results suggest that de-coupling work from location in response to

COVID-19 could result in broader labor market opportunities for households.

To provide more direct evidence on the importance of remote work in driving moves,

we investigate the time series of the “Work from Home” response rate. Figure 1 presents

a quarterly time series of the pandemic-induced moves citing work from home. There is a

quick reaction to the pandemic with more than 5% of responses citing remote work in the

second quarter of 2020. The prevalence of remote work gradually increased over the fist

six quarters of the pandemic before leveling off. In the second quarter of 2021 almost 1

in 5 pandemic-influenced moves was related to this newfound geographic flexibility. This

effect has remained above 17% through the end of the sample in 2021, indicating that this

trend was not short-lived. The persistence of migration influenced by remote work suggests

many households expect a permanence in their remote work arrangement going forward

and that more and more households are learning of the permanence of their arrangement.

Moreover, it is likely that other individuals outside this survey are moving as a result of the

increased work-from-home flexibility, but may not have tied themselves to the response of

being pandemic-induced and, therefore, were not asked for their free response.

19

Electronic copy available at: https://ssrn.com/abstract=3808326

4.2 Changes in Reasons for Moving Across Demographic Groups

We next examine heterogeneity in the reasons for moving during the pandemic across income,

age and household size groups. We expect there to be a differential impact of the pandemic on

households reasons for moving as the COVID-induced shifts in work and social arrangements,

and the impact of mandates are unlikely to fall on all households equally.

Table 3 presents the results. We examine the proportion of moves induced by COVID,

the proportion of moves related to cost of living during the pandemic, and the change in

the proportion of moves related to Job, Family, Retirement, Lifestyle, and Health reasons

during the pandemic as compared to pre-pandemic (defined as January 2019-March 2020).23

These within-demographic changes will remove any time-invariant differences across groups.

Thus, comparing across different groupings serves as a difference-in-difference for changes in

proportions. We cannot calculate a change in the Cost of Living reason because this was

not an available option pre-pandemic.

In Panel A of Table 3, we examine changes across the income distribution. In the first

column, we investigate the proportion of pandemic-era moves induced by the pandemic.

We find higher-income households were more likely to cite COVID-19 as affecting their

move. Recall our sample is tilted towards higher income individuals, so the absolute number

of higher income households moving in response to the pandemic is quite high. We then

investigate any changes in the reasons for moving during the pandemic. The most significant

differences across the income distribution are related to “Job” and “Lifestyle”. As household

income increases, the change in the proportion of “Job”-related moves is monotonically

decreasing. High income households (> $100,000) are citing “Job” 13.8 percentage points

less during the pandemic, while lower income households (≤ $49, 999) are citing “Job”

only 2.81 percentage points less during the pandemic. High income households are citing

“Lifestyle” reasons more during the pandemic (4.02 pps), while lower income households are

23The results are not sensitive to accounting for seasonality or using the entire pre-pandemic period.

20

Electronic copy available at: https://ssrn.com/abstract=3808326

citing “Lifestyle” much less (-4.24 pps). This difference across high and low income groups

represents a 57% swing as compared to the unconditional “Lifestyle” proportion of moves.

For the other categories, we see similar but slightly smaller differences across groups: changes

in “Family” and “Retirement”-related reasons for moving are increasing in income, while

“Health”-related reasons are only differentiated by the lower-income cohort. Examining the

the pandemic-era proportion of responses for “Cost of Living”, we find cost of living is cited

about 51% (2.98 pps) more often by those households making less than $50,000 in income

than by households making more than $100,000 (8.77% compared to 5.79%).

Taken together, these results suggest higher income households are moving more because

of COVID, less for job-related reasons, and more for non-job, lifestyle-related reasons during

the pandemic. Lower income households, on the other hand, are moving less for lifestyle

reasons and relatively more for financial reasons (“Job” and “Cost of Living”). These results

highlight how the disparate impact of COVID-19 on lower versus higher-income households

has affected households migration decisions. These results add to the results in Chetty et al.

(2020), who show a wide gap in the net effects of the pandemic based on income or wealth.

We repeat this process to look across both respondent age in Panel B of Table 3.24

Examining age brackets, we see the pandemic had a hump-shaped effect with the strongest

impact on those households where the respondent was 35-54 years old. Both younger and

older households seem to be slightly less affected. Middle-age households experienced the

largest decline in job-related moves with a drop of almost 17 percentage points and the

greatest increase in lifestyle-related moves with an increase of over 6 percentage points.

Examining variation across household size in Panel C of Table 3, we see those households

with three or more individuals (likely families) were more likely to move due to the pandemic.

Again, those households most likely to move because of the pandemic were also those

households who experienced a more significant drop in job-related motivations and a significant

increase in lifestyle-related reasons. Specifically, households with at least 3 people experienced

24We also perform this analysis by state. It can be found in Figure A5 of the Appendix.

21

Electronic copy available at: https://ssrn.com/abstract=3808326

a 17 percentage points drop in job-related reasons for moving. These groups also experienced

an increase in “Lifestyle”-related moves of 5.5 percentage points. This once again suggests

that geography and labor mobility are less intertwined during the pandemic.

In the Appendix, we investigate the time series aspects of these analyses. In Appendix

Figure A3, we find the differences in COVID-induced moves across demographic groups is

persistent over time. Figure A4 in the Appendix shows the proportion of movers selecting

each response over time for each income group. The differences in reasons for moving we

see during the pandemic are not merely a continuation of longer-term trends. For instance,

between 2017 and 2019, higher income households moving for “Job” related reasons dropped

slowly from around 63% to 59%. Yet in 2020, only 48% of moves were for “Job”-related

reasons for the highest income group. This one-year 11% decline is 2.75 times larger than

the previous 3 years combined. One of the more striking figures shows how lower-income

households used to cite “Lifestyle” significantly more than any other group. However, in

2020-21 there is a complete reversal where higher-income households become 33% more

likely to cite migration for “Lifestyle” reasons.

To examine if these patterns are present in a more representative sample, we examine

responses across demographics in the CPS survey. We focus on interstate moves and

calculate changes across income, household size, and age brackets. Results can be found in

columns (2)-(7) of Panel A of Appendix Table A2. We find similar patterns as documented

above. For example, we find high-income households and larger households move less for

work-related reasons and more for “Health,” “Change of Climate,” and “Other Family

Reasons.” These groups also move for more housing and neighborhood reasons. The patterns

are not quite as stark when comparing younger to older respondents. This may be because

survey respondents for CPS surveys include everyone within a household while UniGroup

respondents tend to be head of household. Nevertheless, we find similar changes in reasons

for moving across demographics for the CPS and UniGroup respondents.

22

Electronic copy available at: https://ssrn.com/abstract=3808326

Overall, we find higher income households, middle-aged households, and families of 3

or more people (likely parents with children) are moving less for job reasons and more for

lifestyle reasons. These are also the types of movers most likely to state that the pandemic

influenced their move. The higher income, middle aged family was relatively more exposed

to a number of pandemic-related factors that could motivate moves: the shift to remote

work, virtual schooling for children, and older parents with health risks. The higher rates

of moves for this group is interesting considering they are mid-career and might previously

be more tied to their location via their job, and they are more likely to have school age

children. These factors make them ex-ante less likely to move and ex-post more likely to

stay put following the move. The costly, interstate nature of the moves suggests that the

migrants believe their remote work arrangements are likely to persist in the aftermath of the

pandemic. Lower income households, on the other hand, are less likely to be in occupations

that experienced this broad shift towards remote work and, therefore, do not have the same

flexibility to relocate for lifestyle reasons.

4.3 Analysis of Migration Patterns

4.3.1 Graphical Analysis of Migration Patterns

In this section, we leverage the data on move origins and destinations to examine how the

places people are moving from and to has changed during the pandemic. Before running tests,

we graphically illustrate the shifts in migration patterns across states during the pandemic.

Figure 2 shows the proportional change in moves by origin (Panel A) and destination (Panel

B) since the onset of the pandemic. Specifically, for move origins (destinations), we compute

the proportion of moves from (to) a particular state in the pandemic period then subtract

off the proportion of moves from (to) that state during the pre-pandemic period. Comparing

moves as a proportion of all moves allows us to abstract away from the ebbs and flows of

the total number of moves and strictly focus on the cross-sectional shock.

23

Electronic copy available at: https://ssrn.com/abstract=3808326

Delventhal and Parkhomenko (2020) provide a theoretical model for how migration

responds to an increase in remote work arrangements. One of their key predictions is

a move away from the coasts and into middle America. This prediction appears to be

manifesting with coastal states representing a significantly larger proportion of move origins

in the pandemic relative to the pre-pandemic period. The mountain west states of Utah,

Idaho, and Montana experienced significant increases in inflows, as well as some southern

states including Texas, Florida, Tennessee, and the Carolinas. The magnitudes of these shifts

are non-trivial. Colorado experienced the largest increase in outflows at 0.61pps and Florida

saw the largest decrease of -0.59pps. Florida saw the biggest proportional increase in inflows

of 1.27pps, while California saw the largest drop in inflows with a -0.95pps proportional

drop. The states people are moving into at a greater rate tend to have lower tax burdens

and lower overall cost of living. In Section 4.3.2, we provide more regression-based evidence

that tax rates and cost of living are related to the pandemic-era shifts in migration patterns.

We utilize USPS change of address data to see how the UniGroup changes in state-level

migration compare to changes for a broader set of moves. One complicating factor with

comparing UniGroup move data to the USPS data is that the USPS data does not delineate

between within-state and out-of-state moves, so aggregating at the state level may duplicate

movement from within. Nevertheless, we find that the correlation in the change of permanent

address flows for states has an approximate 0.36 and 0.48 correlation for the origin and

destination states, respectively. Again, the intra-state migration patterns may diminish the

correlation given interstate moves represent 14% of all moves. Given this limitation, it is

reassuring to find that the UniGroup changes in state migration rates have a reasonably high

correlation with similar measures constructed from the broader USPS sample.

24

Electronic copy available at: https://ssrn.com/abstract=3808326

4.3.2 Regression Analysis of Migration Patterns

We formally examine changes in flows of migrants between CBSA pairs. The goal of this

analysis is to uncover changes in the importance of locational features with the onset

of the pandemic via revealed preference.25 For this analysis, we examine flows at the

origin-destination-year level. We examine origin-destination pairs at the CBSA-level because

CBSAs are large enough to create meaningful variation in the the number of moves, but still

small enough to capture local characteristics.

We run a Poisson regression with the number of moves between CBSA-pairs each

year as the dependent variable. Cohn et al. (2021) advocate for the use of a Poisson

regression model because of the unbiased nature and efficiency of the estimation.26 The

data is well-balanced with all CBSA-pairs represented each year, but CBSA-pairs which

experience no moves between them over the five years will be omitted due to the fixed

effects. We include origin-destination pair fixed effects to absorb any time-invariant flows

between CBSA-pairs. We also include year fixed effects to address broad time trends (e.g.,

reduction in overall migration or changes in demand for moving services).27 Thus, we are

comparing the differential impact in the cross-section of CBSA characteristics on migration

flows.

The independent variables of interest are differences in location characteristics between

destination CBSA and origin CBSA (e.g., MedianRentDestination − MedianRentOrigin). The

differences capture the comparison of a characteristic between origin and destination. Essentially,

we are testing whether these differences promote different movement patterns after 2019, as

compared to before. The variables are standardized to give an easier interpretation of the

25Alternatively, we could analyze origin CBSAs and destination CBSAs, separately. This analysis (unreported) gives similar insights as comparing changes in the means from pre-pandemic to during the pandemic as provided in the summary statistics. Additionally, in Table A3 in the Appendix, we provide results examining changes in the proportion of inflows and outflows for each state pair. These results reinforce the interpretation of the CBSA-pair results.

26We use the PPMLHDFE package from Correia et al. (2019) in our analysis. 27Figure A7 contains the raw averages of Origins and Destination characteristics. These figures mimic

many of the results presented here.

25

Electronic copy available at: https://ssrn.com/abstract=3808326

data and standard errors are clustered at the origin-CBSA level.28

We begin by examining how migration is related to the stringency of the state government

response to the pandemic, COVID-19 cases per capita, the population density, median rent,

and state income tax rates. Results are presented in columns 1–6 of Table 4. Examining the

stringency of COVID-19 response (column 1), we find households are more likely to move

to relatively less stringent CBSAs during the pandemic. A one standard deviation increase

in relative July 2020 stringency leads to a 2.7% decline in the number of moves between

CBSAs. In the free responses, we observe a number of individuals moving out of places

with an opposing political party in power and to places they agree with more politically and

with a COVID-19 response they agree with. A larger proportion were upset with a strict

government response in the place they were leaving than the government being too lax. A

typical example would be a migrant leaving New York because of the stringent response

to COVID-19 and moving to Florida where there was a less stringent response. The free

responses combined with the regression result suggests that government response may have

been a contributing factor in households’ location decisions.

Using the comparative July 2020 infection rate, we find destinations with relatively

lower infection rates, as compared to the origin CBSA, experience a relative increase in the

number of moves. People may move to avoid infection or this result could be due to urban

areas being hit harder earlier on and households leaving larger cities and more urban areas.

The results in column 3 show individuals are more likely to relocate to less dense areas

following the onset of the crisis. The time series of the average destination density suggests

the bulk of this effect is coming from the destination choice as opposed to the origin (see

Figure A7 in the Appendix). In column 4, we find a greater frequency of moves to lower

cost of living areas, as proxied by rent, during the pandemic. This result has the largest

economic magnitude, where a one standard deviation decrease in relative rent leads to a

28We do not cluster along the year dimension since clustering with a small number of clusters (5) risks biasing the standard errors (Thompson (2011)).

26

Electronic copy available at: https://ssrn.com/abstract=3808326

6.4% increase in the number of moves between the pairs.29 These results suggest moving

increased across CBSAs to more affordable places and not merely within a given CBSA.

The results in column 5 further support the role of living costs. A one standard deviation

increase in the relative state marginal income tax rate is associated with approximately 4.1%

fewer moves.

In column 6, we present results for a horse race regression in an attempt to determine

whether one or two particular dimensions are more important than the others. We find that

rent differential is economically the most important and statistically the most significant

coefficient, and the coefficient on the differential in income tax rates remains statistically

significant as well. Both continue to exhibit a negative relative relationship to migration rates

during the pandemic as compared to years prior. These results indicate that characteristics

broadly related to cost of living in an area are more informative about shifts in migration rates

during the pandemic than the other characteristics. The more temporary characteristics,

COVID-19 case rate and government stringency, have little explanatory power suggesting

moves are motivated more by the characteristics that are expected to be more persistent.

That said, all independent variables are inherently correlated and the statistical significance

of one variable does not necessarily negate the importance of the other features.

There are at least three potential explanations for the negative relationship between

relative cost of living and number of moves during the pandemic. One potential reason is

the decline in the value of the amenities in higher cost of living areas during the pandemic.

Although it is not clear that a temporary decline in access to amenities should drive out-migration

from high cost of living cities since individuals should be maximizing lifetime utility. Either

households believe the shock to amenities is not temporary, have high discount rates, or

believe the quality of these amenities is likely to decline. A second potential explanation is

the labor market and idea sharing benefits of high cost of living locations may have fallen

as well. A third potential explanation is high cost of living cities have more individuals in

29= (1 − exp(−0.066)) × 100

27

Electronic copy available at: https://ssrn.com/abstract=3808326

occupations that are able to work from home. These individuals may have transitioned to

full-time work from home, which would allow them to live anywhere. Relatedly, remote work

typically requires home office space that many people living in high cost of living areas may

not have. By moving to lower cost of living areas, they can afford more space on a similar

budget.

Given these characteristics have a great deal of common variation – many of these

variables are strongly related to city size – we use the size of the city as a proxy for city

characteristics. In column 7, we find a one standard deviation increase in CBSA population

differential leads to 1.8% fewer moves across the CBSA-pair. Columns 8 and 9 repeat this

regression, but subset on above and below median levels of origin remote work capability.30

We focus on origin remote work capability instead of destination as this is likely the more

influential proxy of one’s ability to work remotely and choosing to move to a location where

remote work is more prominent is secondary. Interestingly, the effect only exists among the

areas with above average ability to work remote. This is consistent with increased movement

from high population areas to lower populations areas being related to the ability to remote

work, though we caution over-interpreting this result as remote work ability is correlated

with CBSA population.

In the Appendix, we examine whether the patterns documented differ across income

groups or have changed as the pandemic has progressed. To do so, we repeat the regression

analysis for subsets of movers based on income and subsets of the COVID-period (2020

versus 2021). Results are presented in Table A4. In Panel A, we separate the sample into

above and below median origin zip-code income per capita. We use a zip-code level income

measure to allow us to include movers who do not fill out the survey. Broadly speaking,

we find the movement patterns of higher income households are similar to the full sample,

while lower income households are only sensitive to cost of living. These results mimic our

30Differences in observation counts arise from areas with lower remote work having more origin-destination pairs with zero moves across them. Thus, more observations are absorbed by the origin-destination pair fixed effects.

28

Electronic copy available at: https://ssrn.com/abstract=3808326

previous cross-sectional results on reasons for moving, where high income households are

moving to satisfy a change in preferences and lower income households are moving more out

of necessity.

In Panel B, we repeat the analysis but only include either 2020 or 2021 in the post period.

We find the coefficient on differences in local restrictions (Stringency) is significant for the

2020 period, but insignificant in 2021 as many states opened back up. All other variables are

significant for both 2020 and 2021. It is perhaps unsurprising that the more persistent factors

(rent and tax differentials) remain key predictors of migration as the pandemic continues.

Overall, we find that households are moving to smaller cities during the pandemic and

this relationship is only present among CBSAs with higher levels of remote work capability

in the population. In the next section, we further analyze the flight of households from larger

cities, with a focus on the higher income households who are more likely to be able to work

from home.

4.4 High Income Exodus from Large Cities

Evidence in the previous section suggests that the types of places people are moving from

and to has changed significantly in response to the onset of the pandemic. In this section,

we focus on how the increased frequency of moves out of large cities varied across income

groups.

Panel A of Figure 3 shows the average population of the origin CBSA by income bracket

over time. There is a marked increase in the average origin CBSA population for high income

earners during the pandemic, while the time series for low income households is relatively

flat. Both lower and medium income households experienced a slight uptick in the proclivity

to leave more populous CBSAs, but have remained comparable to their pre-pandemic trends.

We further examine the flows of higher income individuals in Panel B of Figure 3. We

29

Electronic copy available at: https://ssrn.com/abstract=3808326

calculate the difference between destination and origin populations for the highest income

group ($100,000+) and present the distribution of this difference for the pre-pandemic period

and pandemic period, separately. We bucket the variable in 5 million person increments. It

is evident that high income individuals are not only leaving origins with higher populations,

but choosing to move to dramatically less populated areas. There is an absolute decrease in

households moving up to more populous areas and an increase in households moving to less

populous areas.

These results are broadly consistent with the theoretical predictions of Davis et al. (2021)

and Delventhal and Parkhomenko (2020) who predict a migratory response to the enormous

uptake in remote work during the pandemic. Both models predict a permanent component of

remote work moving forward. In response, high income and remote work-capable individuals

find locations that better suit their preferences while these moves produce productivity and

amenity externalities for those who stay. The long-run benefit of this shift hinges on the

productivity of workers in the remote work environment moving forward.

5 Discussion

Implications of our results:

Overall, our results show that a significant number of interstate migrations are being

motivated by expected changes to everyday life that were brought about by the pandemic,

while exposure to the coronavirus is a small or non-existent factor in move decisions. People

are choosing to move away from larger cities and more dense, higher cost areas within cities.

Remote work situations have enabled this location flexibility, especially for individuals in

higher income, white-collar occupations. Barrero et al. (2020) find the fraction of workers

who will work remotely at least 5 days a week is expected to increase from 3.4% to 10.3%

and Bick et al. (2021) find that the percentage of high income households that expect to

30

Electronic copy available at: https://ssrn.com/abstract=3808326

work from home full time in 2022 is 14.8% compared to only 8.9% pre-pandemic. The nature

of the moves for high income households is consistent with this expectation of more remote

work in the long-run.

The shift in migration patterns we document are likely to have a number of additional

implications for the future structure of cities and urban economics. For instance, the

migratory response we document is likely to attenuate the recent upwards trend in the

dispersion of income across locations for higher earning households Gaubert et al. (2021),

as high income households are leaving major cities at a relatively higher rate. Ramani and

Bloom (2021) have shown that people are moving more from urban centers to the suburbs

during the pandemic. We show this is not only an intracity effect, but that a portion of

this is driven by intercity moves. This suggests that even if remote work is not localized,

that individuals are still valuing the additional space and amenities that are present in the

suburbs.

Our results also suggest that the remote nature of work has the potential to alter

information sharing and the value of physical proximity in that process. Theoretical models

on the value of idea sharing and agglomeration (e.g. Davis and Dingel (2019)) suggest

in-person interaction is an important component of individuals’ location decisions, especially

for higher skilled individuals. Larger cities are attractive because of the increased interaction

with other high-skilled individuals. During the pandemic, the value of in-person interaction

dropped significantly, especially for high-skilled labor, which would suggest higher skilled

individuals should be less attracted to large cities after the onset of the pandemic.31 Empirically,

we examine income and location population jointly and find higher-income households are

leaving larger, more expensive cities at a higher rate and landing in less populated areas

during the pandemic. As higher-income individuals move out of more expensive cities this

31If new forms of virtual idea sharing are a reasonable substitute for in-person interaction, the move to remote work and virtual communication will further deteriorate the value of living in major cities. On the other hand, if virtual interaction is a complement to in-person interaction Gaspar and Glaeser (1998) we would not expect this shift to virtual communication to lead people to leave major cities.

31

Electronic copy available at: https://ssrn.com/abstract=3808326

has the potential to further reduce the idea sharing agglomeration benefit of living in a major

city in the long-run.

The out-migration of high income individuals from cities also has the potential to lead

to more permanent changes and less opportunity for lower income workers. In equilibrium,

low income individuals are willing to accept a higher cost of living in larger cities because

their wages are higher (Davis and Dingel (2019)). With the out-migration of high income

individuals from cities, this could lead to a decreased wage for low income individuals. An

extreme version of this occurred during the pandemic, Chetty et al. (2020) document that

the significant drop in consumption in affluent neighborhoods during the pandemic led to

job losses for lower income households. Large cities may not be worth the high cost of living

for some lower income households (Althoff et al. (2022)) in the future or living costs may

need to adjust downwards.

Persistence of these effects:

The reader may be concerned that the pandemic moves we analyze are simply a temporal

shift in relocation demand. While we cannot predict the future, the results suggest that these

patterns are likely to continue to some degree into the future. The incidence of COVID-19

related moves peaked in late 2020 but there is still a significant number of households moving

in response to the pandemic at the end of 2021. The percentage of these moves related to

remote work has also trended higher over time. These trends continued and show no signs of

a reversal despite the rollout of the COVID-19 vaccine and the hope of a return to normalcy.

Permanent shifts in where people desire to locate seems likely considering individuals are not

moving because of the virus itself, but because the structure of everyday life including work

arrangements has changed. Some of these changes, such as the ability to work remotely,

are likely to persist to some degree into the future. Nevertheless, the cumulative number of

moves paired with their significant cost, suggests an economically meaningful change in the

geographic landscape of resources and human capital going forward.

32

Electronic copy available at: https://ssrn.com/abstract=3808326

6 Conclusion

The pandemic altered many aspects of life, yet it remains unclear what post-pandemic

work arrangements and lifestyles will look like. We document important shifts in migration

patterns after the onset of the pandemic and place the results into a broader conversation

about the transitory and permanent features of the pandemic. Our results suggest that the

reason for moving and the types of locations people are moving to and from have changed

dramatically since the start of the pandemic. Since April 2020, more than 10% of all moves

performed by UniGroup were directly motivated by the COVID-19 pandemic and there is

no indication that this pattern will cease. The prevalence of COVID-19-related moves and

the percentage of these moves driven by remote work arrangements continued and in some

cases strengthened through the end of our sample in December 2021, even as a majority of

the population became vaccinated and there was hope for a return to normalcy.

Examining motivations for moving, we find that migrants are moving less for job-related

reasons and more for family, retirement, and lifestyle during the pandemic. We find that

high-income households’ are moving more for non-work reasons during the pandemic, while

low income households move for work-related reasons at a similar frequency as pre-pandemic

and are moving less for non-work reasons. We corroborate these patterns using more

representative data from the CPS on both interstate and intrastate moves. Higher income

households are moving out of higher population areas at a relatively faster rate. We also

document that households have significantly altered where they are moving from and to

during the pandemic: people are moving to CBSAs with lower housing costs, taxes, and less

dense areas. Households are moving away from CBSAs with higher costs and more dense

areas at a greater rate during the pandemic. Thus, the reasons and locations for moving are

indicative of a more permanent shift in recent migration trends.

Overall, we document a strong and sustained shift in migration patterns, especially

for higher income households, after the onset of the pandemic and the broad shift towards

33

Electronic copy available at: https://ssrn.com/abstract=3808326

remote work. The evidence we provide has important implications for real estate, companies,

and governments, as well as the future structure of cities and their agglomeration benefits. .

34

Electronic copy available at: https://ssrn.com/abstract=3808326

References

Althoff, Lukas, Fabian Eckert, Sharat Ganapati, and Conor Walsh, 2022, The geography of remote work, Regional Science and Urban Economics 103770.

Azar, José, Ioana Marinescu, Marshall Steinbaum, and Bledi Taska, 2020, Concentration in us labor markets: Evidence from online vacancy data, Labour Economics 66, 101886.

Barrero, Jose Maria, Nicholas Bloom, and Steven J Davis, 2020, Covid-19 is also a reallocation shock, Technical report, National Bureau of Economic Research.

Bick, Alexander, Adam Blandin, and Karel Mertens, 2021, Work from home before and after the covid-19 outbreak, Available at SSRN 3786142 .

Brueckner, Jan, Matthew E Kahn, and Gary C Lin, 2021, A new spatial hedonic equilibrium in the emerging work-from-home economy?, Technical report, National Bureau of Economic Research.

Chetty, Raj, J Friedman, Nathaniel Hendren, and Michael Stepner, 2020, The economic impacts of covid-19: Evidence from a new public database built from private sector data, Opportunity Insights .

Cohn, Jonathan, Zack Liu, and Malcom Wardlaw, 2021, Count data in finance, Available at SSRN 3800339 .

Correia, S, P Guimarães, and T Zylkin, 2019, ppmlhdfe: Fast poisson estimation with high-dimensional fixed effects, arxiv. org .

Davis, Donald R, and Jonathan I Dingel, 2019, A spatial knowledge economy, American Economic Review 109, 153–70.

Davis, Morris A, Andra C Ghent, and Jesse Gregory, 2021, The work-at-home technology boon and its consequences, NBER Working paper 28461 .

Delventhal, Matt, and Andrii Parkhomenko, 2020, Spatial implications of telecommuting, Available at SSRN 3746555 .

Dingel, Jonathan I, and Brent Neiman, 2020, How many jobs can be done at home?, Journal of Public Economics 189, 104235.

Gaspar, Jess, and Edward L Glaeser, 1998, Information technology and the future of cities, Journal of Urban Economics 43, 136–156.

Gaubert, Cecile, Patrick Kline, Damián Vergara, and Danny Yagan, 2021, Trends in u.s. spatial inequality: Concentrating affluence and a democratization of poverty, NBER Working paper 28358 .

Guglielminetti, Elisa, Michele Loberto, Giordano Zevi, and Roberta Zizza, 2021, Living on my own: the impact of the covid-19 pandemic on housing preferences, Bank of Italy Occasional Paper .

35

Electronic copy available at: https://ssrn.com/abstract=3808326

Gupta, Arpit, Vrinda Mittal, Jonas Peeters, and Stijn Van Nieuwerburgh, 2021, Flattening the curve: pandemic-induced revaluation of urban real estate, Journal of Financial Economics .

Hale, Thomas, Sam Webster, Anna Petherick, Toby Phillips, and Beatriz Kira, 2020, Oxford covid-19 government response tracker, Database, Blavatnik School of Government.

Ling, David C, Chongyu Wang, and Tingyu Zhou, 2020, A first look at the impact of covid-19 on commercial real estate prices: Asset-level evidence, The Review of Asset Pricing Studies 10, 669–704.

Liu, Sitian, and Yichen Su, 2021, The impact of the covid-19 pandemic on the demand for density: Evidence from the u.s. housing market, Available at SSRN 3661052 .

Manson, Steven, Jonathan Schroeder, David Van Riper, Tracy Kugler, and Steven Ruggles, 2020, Ipums national historical geographic information system: Version 15.0, Database, Minneapolis, MN: IPUMS. 2020. http://doi.org/10.18128/D050.V15.0.

Marinescu, Ioana, and Roland Rathelot, 2018, Mismatch unemployment and the geography of job search, American Economic Journal: Macroeconomics 10, 42–70.

Molloy, Raven, Christopher L Smith, and Abigail Wozniak, 2011, Internal migration in the united states, Journal of Economic perspectives 25, 173–96.

Moretti, Enrico, 2010, Local multipliers, American Economic Review 100, 373–77.

Moretti, Enrico, 2012, The new geography of jobs (Houghton Mifflin Harcourt).

Ozimek, Adam, 2020, Remote workers on the move, Available at SSRN 3790004 .

Ramani, Arjun, and Nicholas Bloom, 2021, The donut effect of covid-19 on cities, Technical report, National Bureau of Economic Research.

Rosenthal, Stuart S, William C Strange, and Joaquin A Urrego, 2021, Jue insight: Are city centers losing their appeal? commercial real estate, urban spatial structure, and covid-19, Journal of Urban Economics 103381.

Su, Yichen, 2020, Working from home during a pandemic: It’s not for everyone, Economic Analysis and Insights from the Federal Reserve Bank of Dallas .

Thompson, Samuel B, 2011, Simple formulas for standard errors that cluster by both firm and time, Journal of Financial Economics 99, 1–10.

Tiebout, Charles M, 1956, A pure theory of local expenditures, Journal of political economy 64, 416–424.

36

Electronic copy available at: https://ssrn.com/abstract=3808326

Table 1: Summary Statistics The table presents summary statistics for our sample, constructed using zip-code level moving and survey data from UniGroup from January 2017-December 2021. Remote work data is at the zip-level from Su (2020). Annual zip-code level Median rent is sourced from Manson et al. (2020). Density and population is calculated from the 2010 Census at the ZCTA-level. The Stringency Index measures the number of state-level restrictions on mobility/closures and are calculated using data from Hale et al. (2020). Cases per capita is calculated using county-level data aggregated by the New York Times and scaled to be cases per 10,000 residents. Cases per capita and Stringency Index are measured as of July 1, 2020 to approximate the first wave of the pandemic in our data. Thus, the pre-pandemic means for these variables are simply capturing the weighted average for origins and destinations as of July 1, 2020. Some measures are additionally aggregated to the CBSA level. State tax rates are sourced from NBER’s marginal tax rate for households making more than $100,000. Column labels indicate the statistic presented. In the last four columns we provide the pre-pandemic mean, the post-pandemic mean, the difference between the post-pandemic mean and the pre-pandemic mean, and the p-value of this difference, respectively. Panel B displays survey demographic counts and rates for all moves, representing approximately 25% of all moves.

Panel A: Summary Statistics

Pre Post Pre

N Mean SD Min. 25% Median 75% Max. Mean Mean -Post p-value

Actual Miles 363,249 1190.6 757.5 0.0 603.0 1028.0 1668.0 8295.0 1183.4 1205.9 22.4 0.074

Weight of Move 339,496 8113.0 5819.2 0.0 3594.0 6680.0 11085.0 137004.0 7994.6 8388.7 394.1 0.000

Orig. CBSA Remote Work 358,401 32.9% 3.7% 19.3% 30.6% 33.6% 35.1% 41.2% 32.9% 33.0% 0.1% 0.002

Dest. CBSA Remote Work 355,527 32.2% 3.8% 19.3% 29.4% 33.0% 35.0% 41.2% 32.3% 32.0% -0.3% 0.000

Orig. Zip Remote Work 363,948 32.2% 8.5% 3.0% 26.3% 30.4% 36.2% 94.8% 32.2% 32.3% 0.1% 0.086

Dest. Zip Remote Work 363,761 32.0% 8.9% 3.0% 25.9% 29.9% 36.0% 89.7% 32.2% 31.6% -0.6% 0.000

Orig. Zip Density 362,297 4439.4 10761.7 0.0 580.8 1792.1 3985.6 144220.5 4399.7 4524.1 124.4 0.289

Dest. Zip Density 357,429 3388.5 8447.2 0.0 414.8 1347.8 3391.8 156666.7 3533.3 3078.4 -454.9 0.000

Orig. Zip Rent 360,489 1288.6 440.2 99.0 963.0 1207.0 1524.0 3501.0 1279.9 1307.0 27.1 0.000

Dest. Zip Rent 354,983 1238.5 416.3 256.0 938.0 1168.0 1445.0 3501.0 1243.7 1227.4 -16.3 0.000

Orig. Marginal Tax Rate 364,005 4.8 2.9 0.0 3.2 5.0 6.9 9.3 4.8 4.9 0.1 0.143

Dest. Marginal Tax Rate 364,003 4.3 3.0 0.0 0.0 4.9 6.3 9.3 4.3 4.2 -0.2 0.000

Orig. CBSA Population (Mil.) 364,000 4.3 5.2 0.0 0.6 2.2 5.6 18.9 4.3 4.5 0.2 0.016

Dest. CBSA Population (Mil.) 363,972 3.3 4.4 0.0 0.5 1.9 4.3 18.9 3.4 3.2 -0.1 0.000

Orig. July ’20 Stringency Index 364,005 63.2 7.6 38.0 57.4 64.8 67.6 83.3 63.2 63.3 0.2 0.002

Dest. July ’20 Stringency Index 364,003 62.8 7.5 38.0 57.4 62.5 67.6 83.3 62.8 62.7 0.0 0.753

Orig. July ’20 Cases Per Cap. 356,712 88.03 64.57 0.58 42.15 74.78 117.52 2137.83 87.42 89.33 1.9 0.103

Dest. July ’20 Cases Per Cap. 360,279 78.23 55.43 0.58 40.26 66.80 107.29 2284.64 78.66 77.31 -1.4 0.000

Orig. Stringency (ma) 125,255 48.18 18.66 0.00 35.29 51.82 62.04 93.52 10.77 51.06 40.3 0.000

Dest. Stringency (ma) 125,254 46.98 18.55 0.00 34.26 49.44 61.11 93.52 10.08 49.81 39.7 0.000

Orig. Cases Per Cap. (ma) 117,728 670.14 562.18 0.00 131.72 621.65 1109.18 11775.22 1.10 692.85 691.8 0.000

Dest. Cases per Cap (ma) 119,086 682.10 588.61 0.00 121.80 612.54 1131.67 11867.69 1.02 704.21 703.2 0.000

3 7

E le

ctro n ic co

p y a

va ila

b le

a t: h

ttp s://ssrn

.co m

/a b stra

ct= 3 8 0 8 3 2 6

Panel B: Survey Demographic Distributions

Age Bracket Count Pct. Income Bracket Count Pct. Household Size Count Pct.

18 to 24 1,153 1% Less than $15,000 300 0% 1 16,816 23% 25 to 34 11,306 13% $15,000 to $24,999 700 1% 2 38,108 52% 35 to 44 12,832 14% $25,000 to $34,999 1,218 1% 3 8,316 11% 45 to 54 12,985 15% $35,000 to $49,999 2,760 3% 4 6,985 9% 55 to 64 19,843 22% $50,000 to $74,999 8,065 9% 5+ 3,643 5% 65 to 74 18,094 20% $75,000 to $99,999 9,355 11% 75 or older 6,335 7% $100,000 to $149,999 14,536 16% Total 73,868 Prefer not to answer 6,517 7% $150,000 or more 23,207 26% Total 89,065 Prefer not to answer 28,937 32%

Total 89,078

3 8

E le

ctro n ic co

p y a

va ila

b le

a t: h

ttp s://ssrn

.co m

/a b stra

ct= 3 8 0 8 3 2 6

Table 2: Survey Response Statistics The table presents summary statistics for the survey responses. In Panel A, we present the response rate for each reason for moving using data from UniGroup over the period January 2017-December 2021. Customers can choose from a preset list of reasons for their moves. “Cost of Living” was added in the summer of 2020, so we cannot calculate a differential response for the pandemic period. In Panel B, we present the proportion of moves influenced by the COVID-19 pandemic. UniGroup began asking customers if their move was influenced by COVID-19 starting in August 2020. UniGroup re-surveyed previous survey respondents from earlier in 2020 to complete the COVID-19 influenced assessment. Month is defined as the month the customer completed their move with UniGroup. Panel C presents reasons for moving based on categorizing free responses to the question of how COVID-19 affected your move. Respondents were only asked conditional on answering “yes” to COVID-19 influencing their move.

Panel A: Reason for Moving

Reason Count Proportion Pre-Covid Post-Covid Difference p-value

Job 37,878 42.5% 46.6% 33.1% -13.5% 0.000

Family 23,794 26.7% 24.7% 31.3% 6.6% 0.000

Retirement 17,974 20.2% 19.7% 21.2% 1.5% 0.090

Lifestyle 12,848 14.4% 13.7% 16.1% 2.5% 0.000

Health 5,242 5.9% 5.8% 6.0% 0.2% 0.454

Cost of Living 1,551 - - 5.7% - -

Partial Move 3,223 3.6% 3.8% 3.3% -0.5% 0.017

Panel B: Proportion of COVID-19 Influenced Moves over Time

2020 2021

March 0.0% January 16.6%

April 3.5% February 14.1%

May 3.0% March 18.6%

June 4.2% April 14.7%

July 5.6% May 12.5%

August 16.0% June 12.7%

September 18.6% July 12.3%

October 19.2% August 11.6%

November 19.3% September 9.7%

December 16.9% October 11.1%

November 10.4%

December 6.1%

39

Electronic copy available at: https://ssrn.com/abstract=3808326

Panel C: Proportion of COVID-19 Free Responses Mentioning Each Reason

Reason Proportion

Family 31.8%

Work From Home 17.2%

Job Loss 14.0%

Strong Restrictions 6.7%

Timing Sooner 6.4%

Political 5.0%

Cost of Living 4.0%

Climate 2.4%

Timing Later 2.4%

Death 2.0%

Infection Rate 2.0%

Density 1.7%

Lax Restrictions 1.4%

Unrest 0.7%

40

Electronic copy available at: https://ssrn.com/abstract=3808326

Table 3: Survey Trend Changes in Response to the COVID-19 Pandemic The table presents the proportion of pandemic-influenced moves and the change in the proportion of respondents citing particular reasons for moving, delineated by household demographic data. Households choose from a preset list of reasons for their move. The first column calculates the proportion of respondents within a given demographic bracket who answered “yes” to COVID-19 influencing their move. Similarly, we present the proportion of respondents selecting “Cost of Living” during the COVID-era (this option was added in the summer of 2020, so we cannot compare to the pre-COVID period). In the remaining columns, we calculate the proportion of households within each group who respond with a particular reason (e.g. “Job”) after the onset of the pandemic (April 2020-December 2021) minus the portion selecting that reason prior to the pandemic (January 2019-March 2020). Demographics are obtained from the survey data.

Panel A: Income Brackets

COVID Proportion COVID Proportion − Pre-COVID Proportion

Income Group COVID-Induced Cost of Living* Job Family Retirement Lifestyle Health

≤$49,999 10.18% 8.77% -2.81% -1.07% -0.74% -4.24% -1.02% $50,000-$99,999 10.29% 5.75% -9.94% 5.70% 0.13% 0.12% 1.15%

≥$100,000 13.85% 5.79% -13.80% 7.08% 2.00% 4.02% 0.26%

Panel B: Age Brackets

COVID Proportion COVID Proportion − Pre-COVID Proportion

Income Group COVID-Induced Cost of Living* Job Family Retirement Lifestyle Health

≤34 years 12.37% 3.88% -10.73% 5.52% -0.01% 2.38% 0.63% 35−54 years 16.31% 6.10% -16.90% 7.17% 1.07% 6.25% 0.78% ≥55 years 10.07% 6.01% -4.78% 3.37% -0.24% -0.62% -0.41%

Panel C: Household Size Brackets

COVID Proportion COVID Proportion − Pre-COVID Proportion

Income Group COVID-Induced Cost of Living* Job Family Retirement Lifestyle Health

1 member 11.97% 5.10% -11.55% 6.89% -0.22% 0.83% 0.92%

2 members 10.97% 6.18% -9.28% 5.24% 1.37% 1.46% 0.34%

3+ members 13.47% 6.40% -16.53% 7.56% 1.28% 5.49% 0.04%

41

Electronic copy available at: https://ssrn.com/abstract=3808326

Table 4: Origin-Destination Pair Analysis The table presents Poisson fixed-effect regression results which compare origin-destination pair differences in the number of moves before and after the onset of the crisis. The unit of observation is Origin CBSA-Destination CBSA-Year level and the panel is strongly balanced. The data cover all moves performed by UniGroup from January 2017-December 2021. Post takes the value of one if the date the move was completed occurs in 2020 or 2021, and zero otherwise. The characteristics are calculated as the Destination-level value less the Origin-level value. Remote work data is at the CBSA-level and sourced from Su (2020). Annual zip-code level Median rent is sourced from Manson et al. (2020). To aggregate to the CBSA-level, we take the simple average. To calculate Density, we sum zip-level population and land area to the CBSA and re-calculate. Zip-level population and land area are taken from the 2010 Census. The Stringency Index measures the number of state-level restrictions on mobility/closures and are calculated using data from Hale et al. (2020). Cases per capita is calculated using the number of cases at the county-level from the New York Times. We take the sum of all counties within a CBSA and divide by the 2010 Census CBSA population. We use the Stringency Index and Cases per Capita as of July 1, 2020 to capture the first wave within the post-pandemic period. State level marginal tax rates are taken from NBER’s 2018 tax data for households making more than $100,000. Regression analysis includes Origin-Destination pair and year fixed effects. Standard errors clustered at the origin CBSA level are shown below the estimates. ***, **, * indicates significance at the 1%, 5%, and 10% levels, respectively.

Number of Moves Origin-Destination-Year

(1) (2) (3) (4) (5) (6) (7) (8) (9)

(StringencyDest. - StringencyOrig.) × Post -0.027** 0.004 (0.012) (0.008)

(COVIDCasesDest. - COVIDCasesOrig. )× Post -0.015*** -0.004 (0.002) (0.003)

(DensityDest. - DensityOrig.) × Post -0.023*** -0.001 (0.003) (0.006)

(MedianRentDest. - MedianRentOrig.) × Post -0.066*** -0.057*** (0.007) (0.010)

(TaxRateDest. - TaxRateOrig.) × Post -0.042*** -0.022*** (0.010) (0.008)

(PopulationDest. - PopulationOrig. )× Post -0.018*** -0.019*** 0.005 (0.001) (0.001) (0.006)

Constant 1.511*** 1.509*** 1.506*** 1.506*** 1.509*** 1.505*** 1.506*** 1.674*** -0.756***

(0.000) (0.000) (0.001) (0.001) (0.001) (0.001) (0.000) (0.001) (0.003)

Subset? NO NO NO NO NO NO NO High WFH Low WFH

Orig-Dest Pair FE? YES YES YES YES YES YES YES YES YES

Year FE? YES YES YES YES YES YES YES YES YES

Obs. 340,150 340,150 340,150 340,150 340,150 340,150 340,150 266,520 73,630

Pseudo R2 0.644 0.645 0.645 0.645 0.644 0.645 0.645 0.669 0.153

4 2

E le

ctro n ic co

p y a

va ila

b le

a t: h

ttp s://ssrn

.co m

/a b stra

ct= 3 8 0 8 3 2 6

Figure 1: Time-Series of Work from Home-Related COVID-19 Moves

The figure displays a times series of the percentage of quarterly COVID-related moves which cited remote

work as a key reason for their move. We include only pandemic-influenced moves completed over the period

April 2020–December 2021. We then calculate the percentage of movers each quarter who reference work

from home or remote work as a reason for their move in their included free response.

0

5

10

15

20

25

W or

k fro

m H

om e

(% )

2020q2 2020q3 2020q4 2021q1 2021q2 2021q3 2021q4

Year-Quarter

Rate of Work From Home Responses

43

Electronic copy available at: https://ssrn.com/abstract=3808326

Figure 2: Proportional Changes in Moves by State

The figure displays a map denoting the change in the proportion of moves from the pre-pandemic period to

the post-pandemic period. The proportion is calculated as the number of moves into/out of the state as

compared to the total number of moves across all areas in the pre-pandemic period (January 2019-March

2000) and the post-pandemic period (April 2020-December 2021). Proportions, and therefore differences,

are measured in percentage points.

Panel A: Proportional Change in Origin Following Covid-19 Outbreak

0.179 − 0.614 0.021 − 0.179 -0.033 − 0.021 -0.070 − -0.033 -0.172 − -0.070 -0.594 − -0.172 No data

Panel B: Proportional Change in Destination Following Covid-19 Outbreak

0.14 − 1.27 0.06 − 0.14 0.03 − 0.06 -0.06 − 0.03 -0.15 − -0.06 -0.95 − -0.15 No data

44

Electronic copy available at: https://ssrn.com/abstract=3808326

Figure 3: Changes in Origin and Destination by Income Group

This figure shows the average origin population by income groups on a quarterly basis. The sample

represents data for survey respondents which is approximately 25% of the entire sample from January

2017-December 2021. We define the groups as Low Income (<$50,000), Medium Income ($50,000-$99,999) and High Income ($100,000+). Populations are determined at the CBSA level using 2010 Census data. Panel B calculates the difference in CBSA-level population, comparing the destination population to the

origin. Green-shaded bars represent the pandemic period, defined as moves completed April 2020-December

2021, while the clear bars represent the pre-pandemic period (January 2017-March 2020).

Panel A: Average Origin CBSA Population by Income Group

3

3.5

4

4.5

5

A vg

. O rig

. P op

ul at

io n

2017q1 2018q1 2019q1 2020q1 2021q1 2022q1

Quater-Year

High Income Medium Income Low Income

Panel B: Destination Population Difference for High Income Households

0

10

20

30

40

P er

ce nt

-20 -10 0 10 20

Difference in Population (Mil.)

Post Pre

45

Electronic copy available at: https://ssrn.com/abstract=3808326

A Appendix A: Additional Figures and Tables

Figure A1: Moves through time

The figure displays a times series of the number of monthly moves completed over the sample period of

January 2017-December 2021. In Panel A, we include all moves completed over this period. For Panel

B, we break out the number of monthly moves by income group. Therefore, Panel B relies on the sample

where households provide income data, which represents approximately 25% of all moves. We define income

groups as Low Income (< $50,000), Medium Income ($50,000-$99,999) and High Income ($100,000+) to best approximate terciles.

Panel A: Number of Monthly Moves

2000

4000

6000

8000

10000

N um

be r o

f M ov

es

2017m1 2018m7 2020m1 2021m7

Month-Year

Panel B: Number of Moves by Income Group

0

500

1000

1500

N um

be r o

f M ov

es

2017m1 2018m7 2020m1 2021m7

Month-Year

High Income Medium Income Low Income

46

Electronic copy available at: https://ssrn.com/abstract=3808326

Figure A2: Proportion of Moves to Urban, Suburban and Rural Place by Origin Type The figure displays the proportion of moves to urban, suburban, and rural areas by the origin zip code type (e.g. urban, suburban, etc.) across years 2017-2021. The classifications for urban, suburban, and small town are determined by 2010 Rural-Urban Commuting Areas (RUCA) codes at the zip code level. Urban is defined as metropolitan areas with a primary flow within an urbanized area (RUCA=1). Suburban is defined as a metropolitan area where at least 10% of daily flow is to an urban area (RUCA=2,3). Finally, Rural is defined as micropolitan, small town, or rural area (RUCA=4-10).

5

6

7

8

9

10

84

85

86

87

P ro

po rti

on o

f M ov

es (%

)

2017 2018 2019 2020 2021

Year

Urban Dest. (Left Axis) Suburban Dest. (Right Axis) Rural Dest. (Right Axis)

(a) Urban Area Origins

8

10

12

14

16

18

72

74

76

78

P ro

po rti

on o

f M ov

es (%

)

2017 2018 2019 2020 2021

Year

Urban Dest. (Left Axis) Suburban Dest. (Right Axis) Rural Dest. (Right Axis)

(b) Suburban Area Origins

5

10

15

20

71

72

73

74

P ro

po rti

on o

f M ov

es (%

)

2017 2018 2019 2020 2021

Year

Urban Dest. (Left Axis) Suburban Dest. (Right Axis) Rural Dest. (Right Axis)

(c) Rural Area Origins

4 7

E le

ctro n ic co

p y a

va ila

b le

a t: h

ttp s://ssrn

.co m

/a b stra

ct= 3 8 0 8 3 2 6

Figure A3: Proportion of COVID-induced Moves by Household Size, Income Group, and Age Groups The figure displays the proportion of moves that were influenced by the pandemic across different demographic segments on a quarterly basis. For each quarter, we sum the number of moves indicated as COVID-influenced and divide by the total number of surveys completed. Demographic data is taken from survey responses.

.05

.1

.15

.2

.25

P ro

po rti

on o

f C O

V ID

-In du

ce d

M ov

es

2020q2 2020q3 2020q4 2021q1 2021q2 2021q3 2021q4

Year

Less than 50k 50k - 100k 100k+

(a) Income Group

0

.1

.2

.3

P ro

po rti

on o

f C O

V ID

-In du

ce d

M ov

es

2020q2 2020q3 2020q4 2021q1 2021q2 2021q3 2021q4

Year

34 and Under 35-54 55 and Older

(b) Age Group

.05

.1

.15

.2

.25

P ro

po rti

on o

f C O

V ID

-In du

ce d

M ov

es

2020q2 2020q3 2020q4 2021q1 2021q2 2021q3 2021q4

Year

1 Person 2 Persons 3+ Persons

(c) Household Size

4 8

E le

ctro n ic co

p y a

va ila

b le

a t: h

ttp s://ssrn

.co m

/a b stra

ct= 3 8 0 8 3 2 6

Figure A4: Annual Proportion of Reasons by Income Group This figure presents the proportion of moves influenced by particular reasons across time and by income groups. The reasons are chosen from a fixed menu of options. Cost of Living only entered the menu post-onset of the pandemic. We define the groups as Low Income (<$50,000), Medium Income ($50,000-$99,999) and High Income ($100,000+) to best approximate terciles. This data is restricted to moves that provide income information in their surveys over the period January 2017-December 2021.

.1

.2

.3

.4

.5

.6

P ro

po rti

on J

ob -R

el at

ed

2017 2018 2019 2020 2021

Year

High Income Medium Income Low Income

(a) Job-Related Move Proportion

.04

.06

.08

.1

.12

P ro

po rti

on F

am ily

-R el

at ed

2017 2018 2019 2020 2021

Year

High Income Medium Income Low Income

(b) Family-Related Move Proportion

.15

.2

.25

P ro

po rti

on R

et ire

m en

t-R el

at ed

2017 2018 2019 2020 2021

Year

High Income Medium Income Low Income

(c) Retirement-Related Move Proportion

0

.05

.1

.15

P ro

po rti

on H

ea lth

-R el

at ed

2017 2018 2019 2020 2021

Year

High Income Medium Income Low Income

(d) Health-Related Move Proportion

.1

.12

.14

.16

.18

.2

P ro

po rti

on L

ife st

yl e-

R el

at ed

2017 2018 2019 2020 2021

Year

High Income Medium Income Low Income

(e) Lifestyle-Related Move Proportion

0

.05

.1

P ro

po rti

on C

os t_

of _l

iv in

g- R

el at

ed

2017 2018 2019 2020 2021

Year

High Income Medium Income Low Income

(f) Cost of Living-Related Move Proportion

49

Electronic copy available at: https://ssrn.com/abstract=3808326

Figure A5: Proportional Change in Reason for Move by State This figure presents the change in the proportion of survey respondents who state they move for a particular reason. We create changes in these proportions by state during the pandemic period (April 2020-Decemeber 2021) as compared to before the onset of the pandemic (January 2019-March 2020). Proportions, and therefore differences, are measured in percentage points. Note, “Cost of Living” does not appear as an option until the summer of 2020. The figure highlights how different states’ entrants and exiters responded to the onset of the pandemic, defined as April 1, 2020. The data includes all survey responses over the period January 2019-December 2021.

-7.27 − 3.48 -9.17 − -7.27 -10.06 − -9.17 -11.74 − -10.06 -14.30 − -11.74 -21.98 − -14.30 No data

(a) Origin: Job-Related Change

-5.67 − -1.54 -7.65 − -5.67 -10.18 − -7.65 -12.13 − -10.18 -16.23 − -12.13 -21.79 − -16.23 No data

(b) Destination: Job-Related Change

4.82 − 16.94 1.33 − 4.82 0.16 − 1.33 -0.98 − 0.16 -4.46 − -0.98 -12.14 − -4.46 No data

(c) Origin: Family-Related Change

4.87 − 16.01 1.98 − 4.87 0.74 − 1.98 -0.84 − 0.74 -2.58 − -0.84 -14.58 − -2.58 No data

(d) Destination: Family-Related Change

4.29 − 9.44 2.79 − 4.29 1.42 − 2.79 -0.03 − 1.42 -2.43 − -0.03 -5.99 − -2.43 No data

(e) Origin: Lifestyle-Related Change

3.75 − 7.15 2.50 − 3.75 1.72 − 2.50 0.63 − 1.72 -0.20 − 0.63 -10.69 − -0.20 No data

(f) Destination: Lifestyle-Related Change

7.40 − 18.01 4.21 − 7.40 2.30 − 4.21 1.68 − 2.30 1.01 − 1.68 0.45 − 1.01 No data

(g) Origin: Cost of Living-Related Change

8.27 − 12.86 7.05 − 8.27 4.60 − 7.05 3.65 − 4.60 2.46 − 3.65 0.29 − 2.46 No data

(h) Dest.: Cost of Living-Related Change 50

Electronic copy available at: https://ssrn.com/abstract=3808326

Figure A6: Proportional Change in Mover Demographic by State This figure presents the proportional change in the income group, age, and household size of the moving household by origin and destination states. The figure highlights how different states’ entrants and exiters responded to the onset of the pandemic, defined as starting April 1, 2020. Proportions are defined as the number of moves in the state in either the pre/post-pandemic period divided by the total number of moves across all locations in that period. Proportions, and therefore differences, are measured in percentage points. The data includes all moves with survey responses over the period January 2019-December 2021.“High income” is defined as those households who make more than $100,000. “Younger” is defined as those where the respondent is less than 55. Finally, “Family” captures any household that is three people or more. All groupings were meant to best approximate the median while using some judgement on an appropriate cut.

6.01 − 16.12 2.80 − 6.01 1.59 − 2.80 -0.19 − 1.59 -2.83 − -0.19 -36.84 − -2.83 No data

(a) Origin: High Income Change

8.40 − 18.41 5.14 − 8.40 2.99 − 5.14 0.51 − 2.99 -4.11 − 0.51 -9.55 − -4.11 No data

(b) Destination: High Income Change

1.37 − 12.62 -0.73 − 1.37 -2.24 − -0.73 -4.98 − -2.24 -7.80 − -4.98 -17.22 − -7.80 No data

(c) Origin: Younger Age Change

5.52 − 14.29 0.08 − 5.52 -1.54 − 0.08 -3.27 − -1.54 -5.56 − -3.27 -14.80 − -5.56 No data

(d) Destination: Younger Age Change

4.82 − 16.94 1.33 − 4.82 0.16 − 1.33 -0.98 − 0.16 -4.46 − -0.98 -12.14 − -4.46 No data

(e) Origin: Family Change

4.87 − 16.01 1.98 − 4.87 0.74 − 1.98 -0.84 − 0.74 -2.58 − -0.84 -14.58 − -2.58 No data

(f) Destination: Family Change

51

Electronic copy available at: https://ssrn.com/abstract=3808326

Figure A7: Origin and Destination Characteristics Through Time These figures present the average characteristics of origin and destination through time across years for the entire move sample (January 2017-December 2021). The Stringency Index measures the number of state-level restrictions on mobility/closures and are calculated using data from Hale et al. (2020). Stringency Index is measured as of July 1, 2020 to approximate the mid-point of the pandemic era in our data. Density is calculated from the 2010 Census at the zip-code level. Median rent at the zip-code level is sourced from Manson et al. (2020).

62.5

63

63.5

64

A ve

ra ge

S tri

ng en

cy

1 4 7 10 13

Month

2021 2020 2019 2018 2017

(a) Origin: Stringency Index

62.4

62.6

62.8

63

63.2

A ve

ra ge

S tri

ng en

cy

1 4 7 10 13

Month

2021 2020 2019 2018 2017

(b) Destination: Stringency Index

3500

4000

4500

5000

5500

A ve

ra ge

D en

si ty

1 4 7 10 13

Month

2021 2020 2019 2018 2017

(c) Origin: Population Density

2500

3000

3500

4000

A ve

ra ge

D en

si ty

1 4 7 10 13

Month

2021 2020 2019 2018 2017

(d) Destination: Population Density

1260

1280

1300

1320

1340

A ve

ra ge

R en

t

1 4 7 10 13

Month

2021 2020 2019 2018 2017

(e) Origin: Median Rent

1200

1220

1240

1260

1280

A ve

ra ge

R en

t

1 4 7 10 13

Month

2021 2020 2019 2018 2017

(f) Destination: Median Rent

52

Electronic copy available at: https://ssrn.com/abstract=3808326

Table A1: CPS Migration Survey Statistics The table presents an overview of demographics and reasons for moving drawn from the CPS ASEC migration surveys between 2017-2021. We use all surveys for interstate moves (migrate1=5) and the associated ASEC weight to re-weight observations and get closer to a representative sample. There is virtually no difference in the distributions when not weighting. We bracket age, household income, and household size to match the brackets used in the UniGroup survey. The Age bracket is determined by the respondent’s age. Income is defined as household income. Household size is taken from the number of survey respondents within the household. In Panel A, we present the demographic distributions. In Panel B, we present a breakdown of reasons for moving. Respondents are able to select from a variety of reasons for moving but may only select one choice.

Panel A: CPS Survey Demographic Distributions

Age Bracket Pct. Income Bracket Pct. Household Size Pct.

18 to 24 12% Less than $15,000 7% 1 17% 25 to 34 30% $15,000 to $24,999 7% 2 31% 35 to 44 23% $25,000 to $34,999 8% 3 19% 45 to 54 13% $35,000 to $49,999 12% 4 19% 55 to 64 12% $50,000 to $74,999 17% 5+ 15% 65 to 74 6% $75,000 to $99,999 15% 75 or older 4% $100,000 to $149,999 16%

$150,000 or more 17%

Panel B: CPS Reason for Move Frequency

Reason Proportion

New Job or Job Transfer 33.8%

Other Family Reason 15.5%

To Establish Own Household 5.4%

Attend/leave College 4.7%

For Cheaper Housing 4.5%

Wanted New or Better Housing 4.2%

Other Reasons 4.2%

Change in Marital Status 3.8%

For Easier Commute 3.1%

Change of Climate 2.7%

To Look for Work or Lost Job 2.7%

Other Housing Reason 2.6%

Retired 2.5%

Want to Own Home, Not Rent 2.4%

Health Reasons 2.3%

Wanted Better Neighborhood 1.9%

Other Job-related Reason 1.7%

Relationship with Unmarried Partner 1.3%

Foreclosure or Eviction 0.4%

Natural Disaster 0.2%

53

Electronic copy available at: https://ssrn.com/abstract=3808326

Table A2: CPS Change in Migration Reasons by Mover and Move Type This table presents the changes in reasons for moving using data from the CPS ASEC migration surveys between 2017-2021. We calculate the proportion of moves for each group in the post period (ASEC 2021) and subtract the proportion of stated reasons for the pre-pandemic period (2017-2020). Note, 2021 surveys will capture migration over the previous year, so we only have one year of pandemic period reasons for moving. By using proportional changes, we account for any change in the overall number of moves within each period. In Panel A, we focus only on interstate moves (migrate1=5), while Panel B presents results only for the subset of within-state movers (migrate1=3,4). We use all surveys and the associated ASEC weight to re-weight observations and get closer to a representative sample. When splitting on household income, we define “High-income” as those households making more than $100,000 and all others as “Not High-Income.” “Family” denotes a serial observation that has more than three or more respondents, likely capturing a family. “Young” is defined as survey respondents whose age is below 55, while “Old” is any survey respondent 55 or older. These cuts are meant to approximate the cuts used for our analysis using the UniGroup data.

Panel A: Interstate Changes in Reasons

(1) (2) (3) (4) (5) (6) (7)

Overall High Income Not High Income Family Not Family Young Old

Change in Marital Status 0.2% 1.5% -0.6% 2.1% -1.8% 0.2% 0.2%

To Establish Own Household 0.6% 0.9% 0.9% -0.1% 1.4% 1.2% -2.3%

Other Family Reason 0.3% 4.8% -2.1% 1.6% -0.9% 0.1% 1.7%

New Job or Job Transfer -8.8% -17.4% -5.0% -13.3% -3.9% -8.8% -9.8%

To Look for Work or Lost Job 0.2% -0.9% 1.1% 0.6% -0.1% 0.4% -0.9%

For Easier Commute -1.5% -1.3% -1.6% -2.1% -0.8% -1.6% -0.9%

Retired 0.0% 1.0% -0.7% -0.1% -0.2% 0.4% -2.0%

Other Job-related Reason -1.3% -1.4% -1.2% -1.3% -1.2% -1.5% -0.4%

Want to Own Home, Not Rent 1.2% 2.9% 0.1% 0.5% 1.8% 0.9% 2.7%

Wanted New or Better Housing -0.5% 1.4% -1.7% 0.1% -1.1% -0.7% 1.1%

Wanted Better Neighborhood 3.7% 3.0% 4.2% 5.4% 1.9% 3.5% 4.8%

For Cheaper Housing 2.2% 1.5% 2.8% 2.9% 1.5% 2.4% 1.1%

Other Housing Reason 0.1% -0.3% 0.5% -0.5% 0.7% 0.3% -0.9%

Attend/leave College 0.0% -0.8% 0.6% 1.7% -1.9% 0.0% -0.4%

Change of Climate 0.0% 2.1% -1.4% 0.6% -0.8% -0.1% 0.6%

Health Reasons 0.9% 2.6% -0.1% 1.7% 0.0% 0.5% 3.0%

Other Reasons -0.2% -1.6% 0.5% -1.1% 0.5% -0.4% 0.4%

Natural Disaster 0.5% 0.5% 0.6% -0.1% 1.2% 0.4% 1.3%

Foreclosure or Eviction 0.3% 1.0% -0.1% 0.6% 0.1% 0.4% 0.0%

Relationship with Unmarried Partner 2.2% 0.6% 3.3% 0.9% 3.5% 2.5% 0.7%

54

Electronic copy available at: https://ssrn.com/abstract=3808326

Panel B: Within State Changes in Reasons

(1) (2) (3) (4) (5) (6) (7)

Overall High Income Not High Income Family Not Family Young Old

Change in Marital Status -0.6% -1.0% -0.4% 0.3% -1.8% -0.6% -0.2%

To Establish Own Household -1.2% -0.9% -0.8% -2.4% 0.0% -1.5% 1.0%

Other Family Reason -1.3% -1.0% -1.5% -1.0% -1.6% -1.8% 1.9%

New Job or Job Transfer -1.1% -0.7% -1.4% -0.5% -2.0% -1.1% -1.2%

To Look for Work or Lost Job 0.1% 0.2% 0.1% 0.0% 0.2% 0.0% 1.0%

For Easier Commute -1.3% -1.0% -1.5% -1.1% -1.7% -1.4% -0.3%

Retired -0.2% -0.4% -0.1% -0.4% 0.1% -0.2% -0.2%

Other Job-related Reason -0.4% -0.5% -0.4% -0.6% -0.3% -0.5% -0.3%

Want to Own Home, Not Rent 0.2% -0.1% 0.0% -0.1% 1.0% 0.5% -1.6%

Wanted New or Better Housing 1.4% 3.7% -0.1% 1.6% 1.8% 1.7% -0.5%

Wanted Better Neighborhood 2.7% 2.4% 2.9% 3.7% 1.4% 2.6% 3.4%

For Cheaper Housing 1.0% 0.9% 1.4% 1.2% 0.6% 0.8% 2.2%

Other Housing Reason -1.4% -2.0% -1.1% -2.2% -0.3% -1.4% -1.3%

Attend/leave College 0.6% 0.6% 0.8% 1.2% -0.1% 0.8% -0.1%

Change of Climate 0.2% 0.3% 0.2% 0.2% 0.2% 0.3% -0.1%

Health Reasons -0.6% 0.0% -0.9% -0.2% -1.3% -0.2% -3.3%

Other Reasons -0.3% -1.3% 0.1% -0.4% -0.2% -0.5% 0.8%

Natural Disaster -0.2% -0.2% -0.2% -0.1% -0.4% -0.1% -0.9%

Foreclosure or Eviction -0.4% -0.4% -0.4% -0.6% -0.2% -0.3% -1.1%

Relationship with Unmarried Partner 2.8% 1.7% 3.4% 1.5% 4.5% 3.1% 0.7%

55

Electronic copy available at: https://ssrn.com/abstract=3808326

Table A3: Proportional Changes in State to State Moves

The table displays the 25 state pairs with the highest and lowest proportional change in moves following the onset of the COVID-19 pandemic. The pre-period is defined as January 2017-March 2020, while the post period is April 2020 through December 2021. The proportion for each period is calculated as the number of moves to/from that state pair divided by the total number of moves across all states in that period.

Panel A: Highest Proportional Change In Moves Rank Origin State Destination State Pre-Pandemic Proportion (%) Post-Pandemic Proportion (%) Change in Proportion (%)

1 CA TX 1.253 1.630 0.377

2 NY FL 1.050 1.324 0.274

3 NJ FL 0.671 0.941 0.271

4 IL FL 0.661 0.910 0.249

5 CA TN 0.313 0.542 0.228

6 CA FL 0.705 0.887 0.183

7 CA NC 0.421 0.554 0.133

8 CO FL 0.209 0.330 0.121

9 NJ SC 0.162 0.282 0.120

10 CT FL 0.272 0.389 0.117

11 MA FL 0.314 0.430 0.116

12 WA TX 0.229 0.319 0.090

13 PA FL 0.440 0.528 0.088

14 CA ID 0.334 0.421 0.087

15 CO TX 0.296 0.379 0.083

16 WA AZ 0.241 0.322 0.080

17 WA FL 0.146 0.215 0.069

18 WA TN 0.042 0.109 0.067

19 IL TN 0.220 0.287 0.067

20 AZ FL 0.189 0.254 0.064

21 IL TX 0.453 0.517 0.064

22 NY SC 0.240 0.300 0.060

23 NY TX 0.383 0.442 0.059

24 CA PA 0.252 0.308 0.056

25 IL SC 0.117 0.172 0.055

56

Electronic copy available at: https://ssrn.com/abstract=3808326

Panel B: Lowest Proportional Change In Moves

Rank Origin State Destination State Pre-Pandemic Proportion (%) Post-Pandemic Proportion (%) Change in Proportion (%)

1 CA OR 0.806 0.647 -0.159

2 FL CA 0.461 0.328 -0.134

3 TX CA 0.761 0.631 -0.130

4 CA WA 0.942 0.834 -0.108

5 CA AZ 0.770 0.679 -0.091

6 CA NV 0.413 0.328 -0.086

7 VA CA 0.327 0.248 -0.079

8 FL VA 0.356 0.280 -0.075

9 OR CA 0.247 0.174 -0.074

10 TX AZ 0.298 0.225 -0.072

11 FL GA 0.445 0.373 -0.071

12 OK TX 0.197 0.128 -0.069

13 FL TX 0.521 0.455 -0.066

14 WA CA 0.428 0.362 -0.066

15 AZ CA 0.286 0.221 -0.065

16 NC CA 0.229 0.168 -0.060

17 OH CA 0.280 0.220 -0.060

18 NY CA 0.750 0.690 -0.060

19 IL CA 0.587 0.528 -0.060

20 NC GA 0.193 0.133 -0.059

21 MI CA 0.251 0.193 -0.059

22 WA OR 0.170 0.112 -0.058

23 IN CA 0.136 0.082 -0.054

24 FL CO 0.263 0.213 -0.050

25 GA CA 0.220 0.171 -0.049

57

Electronic copy available at: https://ssrn.com/abstract=3808326

Table A4: Origin-Destination Pair Analysis Across Origin Incomes and Time Period The table presents Poisson fixed-effect regression results which compare origin-destination pair differences in the number of moves before and after the onset of the crisis, differentiated on CBSA income per capita or definition of the post period. The unit of observation is Origin CBSA-Destination CBSA-Year level and the panel is strongly balanced. The data cover all moves performed by UniGroup from January 2017-December 2021. Post takes the value of one if the date the move was completed occurs in 2020 or 2021, and zero otherwise. For Panel A, we run the specification subsetted on whether the origin zip code was below or above the median per capita income, as designated in the table. For Panel B, we define the post period as either 2020 or 2021 (omitting 2020) as designated in the table. The characteristics are calculated as the Destination-level value less the Origin-level value. Remote work data is at the CBSA-level and sourced from Su (2020). Annual zip-code level Median rent is sourced from Manson et al. (2020). To aggregate to the CBSA-level, we take the simple average. To calculate Density, we sum zip-level population and land area to the CBSA and re-calculate. Zip-level population and land area are taken from the 2010 Census. The Stringency Index measures the number of state-level restrictions on mobility/closures and are calculated using data from Hale et al. (2020). Cases per capita is calculated using the number of cases at the county-level from the New York Times. We take the sum of all counties within a CBSA and divide by the 2010 Census CBSA population. We use the Stringency Index and Cases per Capita as of July 1, 2020 to capture the first wave within the post-pandemic period. State level marginal tax rates are taken from NBER’s 2018 tax data for households making more than $100,000. Regression analysis includes Origin-Destination pair and year fixed effects. Standard errors clustered at the origin CBSA level are shown below the estimates. ***, **, * indicates significance at the 1%, 5%, and 10% levels, respectively.

Panel A: Splits on Origin Income Number of Moves Origin-Destination-Year

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)

(StringencyDest. - StringencyOrig.) × Post -0.009 -0.031* 0.007 0.003 (0.014) (0.013) (0.013) (0.009)

(COVIDCasesDest. - COVIDCasesOrig. )× Post -0.017*** -0.014*** -0.010** -0.003 (0.002) (0.002) (0.005) (0.003)

(DensityDest. - DensityOrig.) × Post -0.020*** -0.024*** 0.020 -0.001 (0.006) (0.003) (0.010) (0.006)

(MedianRentDest. - MedianRentOrig.) × Post -0.051*** -0.069*** -0.040*** -0.059*** (0.010) (0.007) (0.014) (0.009)

(TaxRateDest. - TaxRateOrig.) × Post -0.025** -0.045*** -0.015 -0.023** (0.012) (0.011) (0.011) (0.009)

Subset? Lower Income Higher Income Lower Income Higher Income Lower Income Higher Income Lower Income Higher Income Lower Income Higher Income Lower Income Higher Income

Orig-Dest Pair FE? YES YES YES YES YES YES YES YES YES YES YES YES

Year FE? YES YES YES YES YES YES YES YES YES YES YES YES

Obs. 88,320 251,795 88,320 251,795 88,320 251,795 88,320 251,805 88,320 251,795 88,320 251,795

5 8

E le

ctro n ic co

p y a

va ila

b le

a t: h

ttp s://ssrn

.co m

/a b stra

ct= 3 8 0 8 3 2 6

Panel B: Splits on Period of the Pandemic Number of Moves Origin-Destination-Year

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)

(StringencyDest. - StringencyOrig.) × Post -0.044*** -0.011 -0.001 0.008 (0.012) (0.011) (0.009) (0.009)

(COVIDCasesDest. - COVIDCasesOrig. )× Post -0.015*** -0.015*** 0.001 -0.009** (0.002) (0.002) (0.003) (0.004)

(DensityDest. - DensityOrig.) × Post -0.030*** -0.017*** -0.009 0.007 (0.003) (0.003) (0.006) (0.008)

(MedianRentDest. - MedianRentOrig.) × Post -0.082*** -0.052*** -0.067*** -0.047*** (0.007) (0.008) (0.009) (0.013)

(TaxRateDest. - TaxRateOrig.) × Post -0.041*** -0.043*** -0.014* -0.029*** (0.010) (0.012) (0.008) (0.010)

Subset? 2020 2021 2020 2021 2020 2021 2020 2021 2020 2021 2020 2021

Orig-Dest Pair FE? YES YES YES YES YES YES YES YES YES YES YES YES

Year FE? YES YES YES YES YES YES YES YES YES YES YES YES

Obs. 243,484 242,676 243,484 242,676 243,484 242,676 243,488 242,684 243,484 242,676 243,484 242,676

5 9

E le

ctro n ic co

p y a

va ila

b le

a t: h

ttp s://ssrn

.co m

/a b stra

ct= 3 8 0 8 3 2 6