American superstar cities fade as remote work sparks migration southward and inland, exacerbating urban sprawl

doi:10.21203/rs.3.rs-4831481/v1

Download PDF

Article

American superstar cities fade as remote work sparks migration southward and inland, exacerbating urban sprawl

https://doi.org/10.21203/rs.3.rs-4831481/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

The rise of remote work has impacted residential and employment location decisions significantly, as physical proximity is no longer necessary for many job functions. This study examines recent trends in employment and population distribution in the United States, focusing on the impact of work-from-home (WFH) in the post-COVID-19-pandemic era. Our findings reveal that the largest cities have experienced declines in their centres, while exurban and peri-urban counties – particularly in the Sunbelt – demonstrate resilience and growth. Counties with higher WFH adoption also perform better economically. Although WFH practices have not fully decoupled jobs and people, they have reinforced existing population growth patterns driven by lifestyle, affordability, taxation, and job opportunities.

Social science/Geography

Scientific community and society/Geography

The pandemic-induced surge in work-from-home (WFH) has shaken the foundations of existing regional development models (Althoff et al., 2022; Barrero et al., 2023; Lee, 2023; Ramani & Bloom, 2021). The concept that ‘people follow jobs’ wherever those jobs may be located (Østbye et al., 2018; Sassen, 2018) no longer holds true for jobs that can be remotely accessed. Likewise, the ‘creative class’ theory – which predicts that ‘jobs follow people’ to thriving, lifestyle locations (Florida, 2002) – has also been upended by WFH. The decoupling of jobs and population also has knock-on effects for corporate spatial strategy, drawing into question whether ‘global cities’ such as New York and Los Angeles remain ‘Superstar’ growth poles for both people and jobs. As described by Sassen (2013), global cities have historically played a pivotal role in the global economy, acting as major hubs for finance and corporate services. However, the landscape is changing. Kemeny and Storper (2024) highlight the evolving spatial income disparities and the shifting growth dynamics across U.S. cities, indicating that not all traditional growth poles continue to attract population and job growth as they once did.

The rapid rise of WFH has complicated the picture so much that a detailed analysis of the most recent data is needed to uncover any new trends, which may later coalesce into theory. Perhaps the disruption to conventional work patterns experienced during the pandemic has served to accelerate trends that were already set in motion during the preceding decades (Kemeny & Storper, 2024). This includes fully remote work, but also unconventional schedules that disrupt standard Monday-to-Friday routines (Allen et al., 2024; Hopkins & Bardoel, 2023). Even prior to the pandemic, those able to reduce their workplace presence were already relocating to suburbs with more affordable housing options at the expense of a longer but less frequent commute (Bjerke et al., 2024; Brueckner et al., 2023; Ory & Mokhtarian, 2005; Rhee, 2009). Others left their regions of employment entirely, moving to places best suited to their lifestyle and personal preferences (Akatsuka & Toyoda, 2023; Denham, 2021; Haslag & Weagley, 2022).

In this study, we analyse how the pandemic (and the resulting WFH) has reoriented the distribution of jobs and population in the United States at three scales: urban, regional, and national. We compare the immediate pre-pandemic period (2016-2019) to the pandemic and post-pandemic period (2019-2022), focusing more on the latter. More specifically, this study aims to:

1. Understand the relationship between regional changes in jobs and population.

2. Determine the relationship between WFH and the regional distribution of jobs and population.

3. Characterise the determinants of regional job and population changes.

For our analysis, we leverage comprehensive datasets from the American Community Survey and the Bureau of Economic Analysis to perform ‘K-means clustering’ and ‘robust linear regression’. We use counties as the unit of analysis, which we classify based on their size into (a) Large Central Counties, in other words, major cities (b) Large Fringe Counties, meaning the suburbs of large cities (c) Medium and Small Counties, known as towns or smaller cities in common parlance and (c) Non-core Counties, denoting predominantly rural regions. This approach allows us to understand the relationship between job availability and population movement in the context of growing WFH practices.

2.1. Central cities decline; large suburban counties gain both jobs and population

A basic analysis of growth rates before and after the pandemic across various county types reveals considerable variations by geography (Table 1). Large Central Counties experienced population decline despite their infrastructure and amenities on offer. Job growth rates also decreased, with a weak recovery after the pandemic shock. The steepest population and job declines occurred in the Bronx, New York, and San Francisco, California, whereas Collin County, Texas, experienced the most growth. Large Fringe Counties exhibited the most robust growth in population and jobs, both before and after the pandemic. Again, Texas outperformed other states with Kaufman County and Rockwall County experiencing some of the largest increases. Medium and Small Counties also registered positive growth, albeit at a slower pace than Large Fringe Counties. Both extremes were in southern states, with Cameron Parish, Louisiana, declining the most and Long County, Georgia, experiencing the most growth. Meanwhile, Non-Core Counties faced considerable challenges, showing a slight decline in both population and job numbers. At the extremes, Loving County, Texas, saw its population drastically decline despite adding jobs, whereas Custer County, Colorado, grew both in terms of jobs and people.

Table 1

Comparison of average growth rates for population and jobs before and after the pandemic across different county types.
County Type	Name-State	Population Growth Rate (%)		Job Growth Rate (%)		Largest Population Change
County Type	Name-State	Pre-pandemic	Post-pandemic	Pre-pandemic	Post-pandemic	Largest Population Change
Large Central Counties		0.50	-0.25	1.71	0.34
	Bronx County-New York	-0.59	-2.45	2.54	-1.06	Largest Decrease
	San Francisco County-California	0.29	-2.41	2.76	-0.65
	St. Louis city-Missouri	-1.30	-2.12	0.99	-0.14
	Collin County-Texas	3.11	3.58	4.55	5.93	Largest Increase
	Wake County-North Carolina	1.99	1.73	3.03	2.79
	Hillsborough County-Florida	1.83	1.55	2.33	2.33
Large Fringe Counties		0.99	1.09	1.56	0.73
	St. John the Baptist Parish-Louisiana	-0.48	-2.23	0.05	-4.15	Largest Decrease
	Tunica County-Mississippi	-1.79	-1.74	-4.59	-7.35
	San Mateo County-California	-0.14	-1.60	2.08	0.68
	Kaufman County-Texas	4.95	8.07	3.25	5.54	Largest Increase
	Rockwall County-Texas	3.87	6.35	4.90	4.74
	Comal County-Texas	5.19	5.82	4.96	5.06
Medium and Small Counties		0.40	0.53	0.99	0.29
	Cameron Parish-Louisiana	0.45	-5.35	-6.37	-18.46	Largest Decrease
	DeKalb County-Missouri	-4.80	-5.15	-3.74	-3.22	Largest Decrease
	Long County-Georgia	2.22	6.39	2.69	3.42	Largest Increase
	Sumter County-Florida	3.05	5.17	5.01	5.09	Largest Increase
Non-core Counties		-0.18	-0.02	0.36	-0.25
	Loving County-Texas	12.14	-12.87	66.79	19.41	Largest Decrease
	Garza County-Texas	0.22	-8.09	1.10	-5.02	Largest Decrease
	Custer County-Colorado	3.77	5.42	-2.04	7.85	Largest Increase
	Broadwater County-Montana	2.34	5.40	0.11	0.55	Largest Increase

2.2. Towns see more growth; jobs are more stable than people

Based on a K-means clustering technique, we categorised population and job growth into five and four clusters, respectively. Figure 1 provides insights into regional disparities in growth rates before and after the pandemic. Large Central Counties showed either negative or modest population growth, accompanied by considerable declines in jobs post-pandemic. Conversely, Large Fringe Counties exhibited robust growth, reflecting suburban economic resilience and attractiveness. Medium and Small Counties presented a mixed picture, with growth in some counties offset by declines in others, depending on local conditions. Non-core Counties struggled the most, with a widespread decline in both population and jobs. This suggests that less urbanized regions are more vulnerable in case of major disruptions like the pandemic.

The maps in Fig. 2 illustrate the spatial distribution of population and job growth clusters across all counties. The population growth map (Fig. 2a) shows a clear distinction between regions, with Texas, the Southeast, and Western states (apart from California) attracting more population than the Midwest and Northeast. Growth is concentrated around inland urban centres, as well as Florida cities such as Miami and Orlando, whereas former population magnets, including New York, Washington DC, Los Angeles, and San Francisco, are losing their pull. The Rust Belt continues its indomitable decline. The job growth trends (Fig. 2b) largely mirror the population growth trends. In California, there are some pockets of strong job growth but not in San Francisco or Los Angeles. Midwestern cities like Chicago and Detroit have been hit by the pandemic and are experiencing ongoing economic challenges. The Sunbelt region, particularly Texas, Florida, and Arizona is growing, with cities like Dallas and Miami exhibiting strong post-pandemic recovery. New York’s job market shows a mixed pattern, with inner suburbs lagging more than New York City.

2.3. Superstar cities fade; the nation shifts toward the Sunbelt

To better characterise growth and decline on a national scale, we also integrated the five population growth clusters and the four job growth clusters. Figure 3 shows only counties with substantial growth and/or decline in population and jobs; counties with moderate or inconsistent growth patterns have been discarded. Rather than visualising counties as a solid colour, we visualise them as dots. Each dot represents 10,000 inhabitants, reflecting the current population distribution.

The Midwest, particularly the Rust Belt, stands out as regressive, with widespread declines in both population and jobs. However, there are exceptions. Cities such as Indianapolis, Cincinnati, and Columbus seem to have experienced a recent turnaround. New York’s long period of growth may be slowing; while adding jobs, it is losing people, possibly due to high housing costs and lack of amenities. This may be an early sign of decoupling between population and job growth. Philadelphia seems to be absorbing some of New York’s population exodus.

In California, growth is moving inland (Sacramento, Modesto, Riverside, and San Bernardino) from the three major coastal cities (San Francisco, Los Angeles, and San Diego). On the coast, only Santa Barbara is adding jobs – yet people are leaving—likely due to high housing costs and taxes as in New York. High-growth places in the West and Southwest, including Denver, Salt Lake City, Boise, Las Vegas, and Phoenix, may be benefiting from an influx of people from California seeking more affordable living conditions. The Pacific Northwest also shows positive growth, with Portland standing out in Oregon. In all cases, these coastal cities are characterised by higher-than-average housing costs.

The South/Southeast, especially Florida and Texas, experienced the most observable growth in jobs and population. The inland cities of Nashville, Charlotte, Atlanta, and Raleigh are forming a high-growth ‘Piedmont’ triangle. Another high-growth triangle, formed by Austin, Houston, and Dallas, is visible in Texas. However, places like New Orleans and Jackson are struggling, possibly due to lingering effects of past natural disasters and structural transformation in the ‘Deep South’ economy (Bullard & Wright, 2009).

Nearly all of the locations losing jobs but gaining population were peri-urban counties of medium-sized cities. These include primarily exurbs of Seattle, Minneapolis, Denver, Washington, Oklahoma City and Dallas.

2.4. Impact of WFH on growth patterns

2.4.1. More WFH is associated with more growth

Robust linear regression was performed to test the impacts of WFH, demographic characteristics, economic indicators, industrial composition, housing & living conditions, well-being & social characteristics against population and job change, controlling for county type. The proportion of the population working from home correlates positively with population and job growth. The population bar chart (Fig. 4a) shows that higher WFH percentages are associated with more significant population growth, particularly in Large Central Counties. The job bar chart (Fig. 4b) indicates that WFH has less of an impact on job growth than population growth. Counties with growing job markets tend to have higher WFH percentages. Similarly, the combined clusters in the heatmaps (Fig. 4c) show that the highest WFH percentages are found in Large Central Counties with significant population and job growth. These figures highlight the importance of WFH in driving growth across various county types.

2.4.2. Mixed effect of other variables on growth

We find that other variables including education, winter temperature, owner-occupancy, and higher house prices positively correlate with both jobs and population growth, and since the pandemic more so in fringe counties, and smaller and medium-sized cities. The coefficient plot in Fig. 5 visually represents the impact of various socioeconomic variables on pre- and post-pandemic population and job growth, based on regression coefficients from our robust linear regression models. These coefficients quantify the relationship between each independent variable (e.g., WFH percentage, median housing rent, state income tax) and the dependent variables (population and job growth rates), showing the direction and magnitude of their effects. Each point represents the estimated coefficient for a variable, with blue dots for population growth and red dots for job growth. The horizontal lines are confidence intervals, indicating uncertainty around the estimates. If a variable's line crosses the vertical 0 line, its effect is not statistically significant, meaning we cannot confidently say it impacts growth positively or negatively.

A central focus of the analysis is the role of WFH practices on growth. This turns out to be statistically significant, meaning that WFH is associated with regional growth patterns. The analysis also reveals that demographic factors, particularly race, affect growth rates. Counties with a larger non-white population faced more challenges in adapting during the pandemic. Educational attainment was unrelated to population growth before the pandemic but became statistically significant post-pandemic. This change highlights a post-pandemic preference for regions with higher education levels (and possibly more remote work opportunities). Areas with a larger retiree cohort appear to be more attractive overall, possibly due to their stable, service-oriented economies, and more amenable climate.

Economic indicators play a crucial role, with state income tax having a strong negative impact on job growth post-pandemic. Similarly, economic well-being indicators like ‘population percentage below the poverty level’ and ‘unemployment rate’ have statistically significant and negative correlations with growth. This implies that economic disadvantage is a critical barrier to regional prosperity. Interestingly, the median household income has a statistically significant but negative effect on population growth post-pandemic. It appears that high-paying urban environments have become less attractive, probably because the cost of living is higher here as well. The next variable reinforces this explanation: housing costs (mortgage or rent) have a statistically significant impact on regional attractiveness and economic stability.

The industrial composition further shapes regional dynamics. With secondary industries as the reference, primary industries typically show a positive influence on growth, suggesting a base-level economic activity that attracts the population despite not expanding employment significantly. In contrast, tertiary and quaternary industries, which include service-oriented sectors, show varied impacts with generally negative correlations in post-pandemic job and population growth, indicating shifts in the job landscape or post-pandemic industry contractions.

Regarding county types, Large Central Counties often show a negative impact on growth rates, suggesting a shift away from urban cores. In contrast, Medium and Small Counties exhibit positive effects on post-pandemic population growth, highlighting their attractiveness or resilience in adapting to new normal conditions influenced by changes in work and living preferences. Finally, climate variables, such as the average January temperature, are also statistically significant in the post-pandemic population growth model. This suggests that people are attracted to milder winters in the post-pandemic period.

3.1. The ‘donut effect’ returns in magnified form

Though the ‘donut effect’ caused by inner-city decline has been observed in the United States since the 1950s, the term has regained significance as urban populations are further decimated by large-scale desertion of conventional work arrangements. Ramani and Bloom (2021) characterise this donut in terms of the momentum of exurbs and suburbs vis-à-vis central cities and the decanting of the population from larger cities to smaller ones. This trend is not confined to the United States; recent research by the OECD shows that similar patterns of population movement and housing demand are emerging in other countries as well, highlighting the widespread impact of remote work on urban form and regional development (Ahrend et al., 2023). The difference is that in the United States, relatively lax land use allows for substantially more sprawl, leading to larger, and more nebulous, metropolitan footprints.

Though at the turn of the twenty-first century, there was a notable resurgence in central (inner) cities after decades of suburbanisation, a lack of affordable housing options in major cities soon began to cast doubts on the long-term ‘triumph of the city’ (Kotkin, 2010). Pre-pandemic, the debate around urban deconcentration did not significantly account for WFH as a major factor (Costello, 2009; Florida, 2017; Gordon & Richardson, 1996; Hansen & Aner, 2017). The pandemic-induced lockdowns, however, made WFH a necessity, showcasing its potential to deconcentrate cities by allowing people to live farther from their workplaces without compromising their job opportunities (Althoff et al., 2022; Couture et al., 2022; Delventhal et al., 2022). Our results confirm this trend. Before the pandemic, many central counties experienced moderate growth at best, whereas post-pandemic, they declined. Now they are facing substantial challenges in retaining their populations and jobs. As WFH is expected to spread further (Barrero et al., 2021), urban centres may need to reinvent themselves to remain relevant in the new economic landscape. ‘Superstar’ cities like New York and San Francisco still dominate in terms of high-quality employment opportunities, but population growth has been shifting towards more affordable and lifestyle-friendly regions. The latter is experiencing a ‘hyperdonut’ effect where suburban and exurban growth is outpacing urban core growth (Kotkin, 2010).

3.2. The Sunbelt and Inter-Mountain West stand out from the Rust Belt and the Coasts

The Sunbelt region, encompassing Southern states like Texas and Florida, has shown significant resilience and growth. This aligns with broader narratives about the Sunbelt’s economic dynamism and its attractiveness for both businesses and individuals (Glaeser, 2012; Glaeser & Tobio, 2008). The Sunbelt’s warmer climate, lower cost of living, and business-friendly policies have made it a prime destination for both people and businesses looking to relocate. To some extent, the same patterns are observable in the Inter-Mountain West. In contrast, the Rust Belt and parts of the Midwest, regions traditionally reliant on manufacturing, have continued to fall behind. These areas have struggled to maintain growth due to their less adaptable industries and higher dependence on physical presence. Similarly, the Pacific region, particularly California, has faced challenges despite being home to many tech hubs. Challenges here are likely due to unaffordable living costs and high state taxes.

3.3. Americans are drawn to cheap housing, tax benefits, and warm weather

During our study’s timeline, counties with higher WFH percentages tended to perform better economically. These counties have been more responsive to the needs of remote workers beyond mere job proximity. They provide affordable housing, better quality of life, and robust digital infrastructure. In terms of socioeconomic determinants of growth, housing affordability emerged as a critical factor, pushing people out of expensive central cities and toward more affordable suburban and rural areas (Florida, 2017). This pattern of migration is particularly evident in regions with high housing costs, where the median housing rent coefficient was significantly negative at -0.161 for post-pandemic population growth and − 0.085 for post-pandemic job growth. Fiscal policies, such as state income tax rates, also played a crucial role in determining where people chose to live. States with lower or no income tax have become more attractive, as indicated by the state income tax coefficient of -0.178 for post-pandemic population growth and − 0.086 for post-pandemic job growth. Finally, climatic conditions proved significant, with warmer weather being a noteworthy attractor, reflected by a positive coefficient of 0.141 for post-pandemic population growth (Glaeser, 2012; Glaeser & Tobio, 2008). These findings underscore the idea that when people are free to choose where they live independently of their job location, they are driven by both lifestyle and economic factors (Åberg & Tondelli, 2021; Haslag & Weagley, 2022; McManus, 2022; Ory & Mokhtarian, 2005; Ramani & Bloom, 2021). The inability of all workers to relocate to their desired locations is creating new geographical disparities and forms of spatial segregation.

3.4. Conclusion

In contrast to narratives suggesting that urban sprawl has been arrested, or that inner-cities have regained appeal vis-à-vis suburbs, the post-pandemic metropolitan geography of jobs and population suggests that Americans are pushing toward lower-density cities, and the far-flung suburbs of larger cities, often characterised by lower taxes, favourable climate, and a robust jobs market.

Residential and employment locations have not yet fully decoupled in the era of remote work, though lifestyle factors – including affordability – are increasingly relevant. While work-from-home (WFH) practices have introduced some changes, the fundamental link between job availability and population movement remains strong. However, the Pew Research Center estimates that roughly a third of U.S. workers who can work from home do so (Parker, 2023), indicating significant potential for further decoupling of jobs and population. Furthermore, growth in both people and jobs has shifted away from coastal superstar cities, which are now hyper-gentrified and unaffordable to the middle- and lower-income groups (Florida, 2017). On a national scale, there has been a significant movement toward the Sunbelt region, especially Texas and Florida. This shift has magnified the ‘donut’ phenomenon whereby central cities decline as suburban and exurban areas grow. Future research should continue to explore these dynamics, particularly the long-term impacts of WFH and the role of additional socio-economic variables in shaping the new geography of jobs and population in the United States.

Below we provide more detailed information about the methodology, analysis, and statistical testing performed in this study.

4.1. Data sources

We employed cross-sectional data. The main data sources were the American Community Survey (ACS) 5-Year Estimates and Bureau of Economic Analysis (BEA) Wage and Salary Employment Data. The first dataset provides a comprehensive and reliable picture of the present demographic and socioeconomic conditions in US counties. While not as current as the 1-year estimates, the 5-year estimates have the benefit of being applicable to all counties, including those with less than 65,000 inhabitants (roughly half of all counties). The second dataset provides insight into the average annual number of full-time and part-time jobs by place of work in US counties. It covers all jobs for which wages and salaries are paid.

4.2. County classification

U.S. counties were categorized based on their population size into the following classifications, adapted from the National Centre for Health Statistics (NCHS) Urban-Rural Classification Scheme: Large Central Counties (central counties in metropolitan areas with a population of 1 million or more), Large Fringe Counties (fringe counties in metropolitan areas with a population of 1 million or more), Medium and Small Counties (counties in metropolitan areas with a population of less than 1 million), and Non-core Counties (micropolitan and non-core counties).

4.3. Analysis

We used two methods to explore the association between WFH, socioeconomic variables, and population and employment shifts: (1) cluster analysis and (2) robust regression.

4.3.1. K-means clustering

K-means clustering was applied to analyse population and job growth rates over two timeframes: pre-pandemic (2016–2019) and post-pandemic (2019–2022). The ‘elbow method’ was used to determine the optimal number of clusters. This method yielded five clusters for population and four clusters for jobs, as shown in Fig. 1, presenting a balance between granularity and simplicity. They adequately capture the complex economic landscape of the US without overfitting the data. Seven Small and Non-core County outliers with extreme population and job growth rates skewed the results significantly and were therefore removed from cluster analysis. Some observations sit at the borders of clusters, indicating that the complex realities of economic development do not always conform neatly to categorical boundaries.

To examine the potential decoupling of population and job growth, the clustering was performed separately for population and jobs using pre-pandemic and post-pandemic growth rates.

4.3.2. Robust linear regression

Robust linear regression is an alternative to ordinary least squares regression, which avoids the effect of outliers (e.g., a few non-core counties experiencing extreme population or job growth rates). The primary focus of our regression analysis is to explore the impact of work-from-home (WFH) practices, quantified as the percentage of the working population engaged in WFH. This key variable is complemented by a comprehensive set of control variables to encapsulate the diverse influences shaping county-level dynamics.

The dependent variables in our models include:

Pre-pandemic population growth (annual population growth between 2016–2019)
Pre-pandemic job growth (annual job growth between 2016–2019)
Post-pandemic population growth (annual population growth between 2019–2022)
Post-pandemic job growth (annual job growth between 2019–2022)

Data for the independent variables were taken from specific years corresponding to the periods of interest. For post-pandemic models, the independent variables belong to the year 2022, while for pre-pandemic models, the independent variables belong to 2019. These independent variables include:

Work-from-home percentage
Demographic characteristics: retiree population growth, sex ratio, high school graduate percentage, bachelor's or higher degree percentage, non-white population percentage
Economic indicators: share of employment in the tourism industry, below the poverty level percentage, employment-population ratio, unemployment rate, median household income, state income tax at the $90K income level
Industrial composition: share of primary industries, secondary industries, tertiary industries (trade and transportation, services), quaternary industries
Housing and living conditions: housing units with a mortgage percentage, median housing value, average January temperature, owner-occupied housing units percentage, travel time to work, vacant housing units percentage, median housing rent
Wellbeing and social characteristics: household with children percentage, single-parent household percentage, with health insurance coverage percentage
County type (using dummy variables for Large Central County, Large Fringe County, Medium and Small County, and Non-core County)

The analysis was enriched by considering interaction effects between county types and socioeconomic variables to highlight regional disparities. To test for multicollinearity, we calculated the Variable Inflation Factor (VIF) for all variables and those with high VIF values were removed from the regression analysis. Variables susceptible to non-normal distribution, such as income levels and percentages, were log-transformed to mitigate skewness and stabilize variance, enhancing the model's assumption adherence. However, variables that inherently include negative values (i.e., 'Average January Temperature' and 'Retiree population growth') were not log-transformed to avoid computational errors. Categorical variables representing county types were encoded as dummy variables without transformation.

Åberg, H. E., & Tondelli, S. (2021). Escape to the Country: A Reaction-Driven Rural Renaissance on a Swedish Island Post COVID-19. Sustainability, 13(22), 12895.
Ahrend, R., Banquet, A., Bétin, M., Caldas, M. P., Cournède, B., Ramirez, M. D., Pionnier, P.-A., Sanchez-Serra, D., Veneri, P., & Ziemann, V. (2023). Expanding the doughnut? How the geography of housing demand has changed since the rise of remote work with COVID-19. OECD Regional Development Papers, No. 54. https://doi.org/doi:https://doi.org/10.1787/cf591216-en
Akatsuka, H., & Toyoda, M. (2023). Analysis of the relationship between urban dynamics and prevalence of remote work based on population data generated from cellular networks. Scientific Reports, 13(1), 20139.
Allen, K. S., Grelle, D., Lazarus, E. M., Popp, E., & Gutierrez, S. L. (2024). Hybrid is here to stay: Critical behaviors for success in the new world of work. Personality and Individual Differences, 217, 112459.
Althoff, L., Eckert, F., Ganapati, S., & Walsh, C. (2022). The geography of remote work. Regional Science and Urban Economics, 93, 103770.
Barrero, J. M., Bloom, N., & Davis, S. J. (2021). Why working from home will stick.
Barrero, J. M., Bloom, N., & Davis, S. J. (2023). The Evolution of Work from Home. Journal of Economic Perspectives, 37(4), 23–50. https://doi.org/10.1257/jep.37.4.23
Bjerke, L., Bond-Smith, S., McCann, P., & Mellander, C. (2024). Work-from-home, relocation, and shadow effects: Evidence from Sweden.
Brueckner, J. K., Kahn, M. E., & Lin, G. C. (2023). A new spatial hedonic equilibrium in the emerging work-from-home economy? American Economic Journal: Applied Economics, 15(2), 285-319.
Bullard, R. D., & Wright, B. (2009). Race, place, and environmental justice after Hurricane Katrina: Struggles to reclaim, rebuild, and revitalize New Orleans and the Gulf Coast. Westview Press.
Costello, L. (2009). Urban–Rural Migration: housing availability and affordability. Australian Geographer, 40(2), 219-233.
Couture, V., Dingel, J. I., Green, A., Handbury, J., & Williams, K. R. (2022). JUE Insight: Measuring movement and social contact with smartphone data: a real-time application to COVID-19. Journal of Urban Economics, 127, 103328.
Delventhal, M. J., Kwon, E., & Parkhomenko, A. (2022). JUE Insight: How do cities change when we work from home? Journal of Urban Economics, 127, 103331.
Denham, T. (2021). The limits of telecommuting: Policy challenges of counterurbanisation as a pandemic response. Geographical Research, 59(4), 514-521.
Florida, R. (2002). The rise of the creative class. Basic Books.
Florida, R. (2017). The new urban crisis: How our cities are increasing inequality, deepening segregation, and failing the middle class-and what we can do about it. Hachette UK.
Glaeser, E. (2012). Triumph of the city: How our greatest invention makes us richer, smarter, greener, healthier, and happier. Penguin.
Glaeser, E. L., & Tobio, K. (2008). The rise of the sunbelt. Southern Economic Journal, 74(3), 609-643.
Gordon, P., & Richardson, H. W. (1996). Employment decentralization in US metropolitan areas: Is Los Angeles an outlier or the norm? Environment and Planning A, 28(10), 1727-1743. https://doi.org/10.1068/a281727
Hansen, H. K., & Aner, L. G. (2017). On the location dynamics of highly educated people migrating to peripheral regions of Denmark. Population, Space and Place, 23(8), e2076.
Haslag, P., & Weagley, D. (2022). From LA to Boise: How migration has changed during the COVID-19 pandemic. Journal of Financial and Quantitative Analysis, 1-31.
Hopkins, J., & Bardoel, A. (2023). The future is hybrid: how organisations are designing and supporting sustainable hybrid work models in post-pandemic Australia. Sustainability, 15(4), 3086.
Kemeny, T., & Storper, M. (2024). The Changing Shape of Spatial Income Disparities in the United States. Economic Geography, 100(1), 1-30. https://doi.org/10.1080/00130095.2023.2244111
Kotkin, J. (2010). The next hundred million: America in 2050. Penguin.
Lee, K. (2023). Working from home as an economic and social change: A review. Labour Economics, 102462.
McManus, P. (2022). Counterurbanisation, demographic change and discourses of rural revival in Australia during COVID-19. Australian Geographer, 1-16.
Ory, D. T., & Mokhtarian, P. L. (2005). An empirical analysis of causality in the relationship between telecommuting and residential and job relocation.
Østbye, S., Moilanen, M., Tervo, H., & Westerlund, O. (2018). The creative class: do jobs follow people or do people follow jobs? Regional Studies, 52(6), 745-755. https://doi.org/10.1080/00343404.2016.1254765
Parker, K. (2023). About a third of US workers who can work from home now do so all the time. Pew Research Center. https://www.pewresearch.org/short-reads/2023/03/30/about-a-third-of-us-workers-who-can-work-from-home-do-so-all-the-time/
Ramani, A., & Bloom, N. (2021). The Donut effect of COVID-19 on cities.
Rhee, H.-J. (2009). Telecommuting and urban sprawl. Transportation Research Part D: Transport and Environment, 14(7), 453-460.
Sassen, S. (2013). The Global City: New York, London, Tokyo. Princeton University Press.
Sassen, S. (2018). Cities in a world economy. Sage Publications.

There is NO Competing Interest.

Download PDF

Version 1

posted

You are reading this latest preprint version

American superstar cities fade as remote work sparks migration southward and inland, exacerbating urban sprawl

Status:

Version 1

Abstract

Figures

1. Introduction

2. Results

2.1. Central cities decline; large suburban counties gain both jobs and population

2.2. Towns see more growth; jobs are more stable than people

2.3. Superstar cities fade; the nation shifts toward the Sunbelt

2.4. Impact of WFH on growth patterns

2.4.1. More WFH is associated with more growth

2.4.2. Mixed effect of other variables on growth

3. Discussion

3.1. The ‘donut effect’ returns in magnified form

3.2. The Sunbelt and Inter-Mountain West stand out from the Rust Belt and the Coasts

3.3. Americans are drawn to cheap housing, tax benefits, and warm weather

3.4. Conclusion

4. Methodology

4.1. Data sources

4.2. County classification

4.3. Analysis

4.3.1. K-means clustering

4.3.2. Robust linear regression

References

Additional Declarations

Status:

Version 1