University of Southern California Dissertations and Theses

The long-term impact of COVID-19 on commute, employment, housing, and environment in the post-pandemic era

(USC Thesis Other)

The long-term impact of COVID-19 on commute, employment, housing, and environment in the post-pandemic era

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content The Long-Term Impact of COVID-19 on Commute, Employment, Housing, and Environment in the Post-Pandemic Era By Bonnie Wang A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirement for the Degree DOCTOR OF PHILOSOPHY (URBAN PLANNING AND DEVELOPMENT) December 2024 ii Dedication This dissertation is dedicated to my family, mentors, friends, and everyone who has supported me throughout this journey. Your encouragement and belief in me have made all the difference. I would not be where I am today without you. iii Acknowledgements I would like to express my deepest gratitude to my chair and advisor, Dr. Marlon Boarnet, for his invaluable guidance, support, and patience throughout my PhD journey. His expertise and encouragement have been so impactful to my academic development, and I feel incredibly fortunate to have had the opportunity to learn under his mentorship. I am also deeply thankful to my committee members, Dr. Genevieve Giuliano and Dr. Lisa Schweitzer, for their insightful advice and unwavering support throughout my research. Their feedback and suggestions were crucial in shaping this dissertation, and I am grateful for their time and commitment to my work. A special thanks to my co-authors and research team members, Seva Rodnyansky and Andre Comandon. Working with them has been one of the most rewarding experiences of my academic journey. They have been both mentors and friends, always there to lift me up when I needed it most. I have learned so much from them, and I look forward to continuing our collaboration in the future. To my parents, Sam and Diane, thank you for always supporting my decisions and believing in me, even when the path was uncertain. Your trust and constant encouragement gave me the strength to keep going. I would not be where I am today without you, and for that, I am forever grateful. I am grateful to my colleagues for their support and friendship during these past five years. Their presence made this journey more enjoyable, and I truly appreciate all the encouragement and help they provided along the way. Lastly, I am grateful to my friends and everyone who supported me along the way, whether in life or by providing much-needed moments of joy and mental escape. Your support helped me stay grounded, reminded me to enjoy life beyond research, and ensured I maintained a healthy balance between work and life. iv Table of Contents Dedication....................................................................................................................................... ii Acknowledgements........................................................................................................................iii List of Tables ................................................................................................................................ vii List of Figures................................................................................................................................ ix Abbreviations.................................................................................................................................. x Abstract.......................................................................................................................................... xi Chapter 1. Introduction ................................................................................................................... 1 Chapter 2. Measuring the Impact of COVID-19 Policies on Local Commute Traffic: Evidence from Mobile Data in Northern California....................................................................................... 5 Abstract.....................................................................................................................................................6 1. Introduction...........................................................................................................................................7 2. Literature Review................................................................................................................................10 3. Data and Methodology........................................................................................................................12 3.1 Hypotheses and Analysis Plan ......................................................................................................12 3.2 COVID-19 stages, Pandemic Policy, and Traffic Volume ...........................................................14 3.3 Model Specification ......................................................................................................................16 3.4 Data and Descriptive Statistics .....................................................................................................18 4. Results.................................................................................................................................................22 5. Conclusion ..........................................................................................................................................32 7. References...........................................................................................................................................34 Chapter 3. The Relationship Between Remote Work and Workplace Traffic During and After COVID-19 in the Bay Area and Central Valley Region............................................................... 38 Abstract......................................................................................................................................... 39 1. Introduction............................................................................................................................... 40 2. Literature Review...................................................................................................................... 42 3. Data and Methodology.............................................................................................................. 45 3.1 Methodology.....................................................................................................................................45 3.1.1 Analytical Framework................................................................................................................45 3.1.2 O-D flow weights.......................................................................................................................48 3.2 Study Area ........................................................................................................................................51 3.3 Data...................................................................................................................................................52 3.4 Observation Period............................................................................................................................54 v 4. Results....................................................................................................................................... 56 4.1 Destination Characteristics ...............................................................................................................57 4.1.1 Traffic Recovery at Destination.................................................................................................57 4.1.2 Industry composition at Destination ..........................................................................................60 4.2 Origin Characteristics .......................................................................................................................63 4.2.1 Work-from-home Growth at Origin...........................................................................................63 4.2.2 Demographic Characteristics at Origin......................................................................................66 4.2.3 Migration at Origin ....................................................................................................................67 4.3 Work-From-Home Trip Reduction at Destination............................................................................70 4.4 Regression Analysis..........................................................................................................................71 5. Conclusion ................................................................................................................................ 76 References..................................................................................................................................... 78 Chapter 4. The Impact of Remote Work on GHG Emissions in the United States in the PostPandemic Era: A National Analysis of the Relationship Between COVID-19, Commuting, Residential Choice, and Emission Reduction ............................................................................... 82 Abstract......................................................................................................................................... 83 1. Introduction............................................................................................................................... 84 1.1 Background.......................................................................................................................................84 1.2 Research Purpose ..............................................................................................................................85 1.3 Contribution and Significance ..........................................................................................................86 2. Literature Review...................................................................................................................... 87 3. Research Design and Data ........................................................................................................ 91 3.1 Research Design................................................................................................................................91 3.2 Sample Definition and Field Period..................................................................................................93 3.3 Survey Methodology.........................................................................................................................95 4. Results....................................................................................................................................... 96 4.1 Work arrangement preferences.........................................................................................................98 4.2 Remote Workers’ Demographics and Vehicle Type ......................................................................102 Remote work Ability.........................................................................................................................102 Vehicle choice and work arrangement..............................................................................................107 4.3 Migration and Remote Work ..........................................................................................................109 4.4 Distance between work and home ..................................................................................................112 4.5 Environmental impact of Remote work ..........................................................................................116 Daily commute and GHG emissions.................................................................................................117 vi Personal weekly commute GHG emissions by work arrangement...................................................118 Commute distance, migration, and emissions...................................................................................119 5. Conclusion .............................................................................................................................. 125 6. References............................................................................................................................... 127 Chapter 5. Conclusion................................................................................................................. 131 Bibliography ............................................................................................................................... 133 Appendix..................................................................................................................................... 143 Appendix A. Survey Questionnaire ..................................................................................................144 Appendix B. Weighting Benchmark Distributions...........................................................................180 Appendix C. Detailed and additional summary data ........................................................................184 vii List of Tables Table 1. Blueprint for a Safer Economy initial tier and reopening rules example......................................15 Table 2. Summary statistics of variables ....................................................................................................21 Table 3. Regression Results........................................................................................................................29 Table 3. Regression Results (continued).....................................................................................................30 Table 3. Regression Results (continued).....................................................................................................31 Table 3.1.1. Summary of variables.............................................................................................................47 Table 3.1.2.1. Example of OD weights between origin and destination ZCTAs........................................50 Table 3.4.1. Observation week pairs in 2019 and 2021..............................................................................55 Table 4.1. Summary of Descriptive Statistics.............................................................................................56 Table 4.1. Summary of Descriptive Statistics (continued) .........................................................................57 Table 4.1.1.1. Traffic volume by region – average per ZCTA ...................................................................57 Table 4.1.2.1. Industry composition by region – average per ZCTA .........................................................60 Table 4.2.1.1. Share of remote workers by region – average per ZCTA....................................................63 Table 4.2.2.1. 2019 Demographic characteristics by region – average per ZCTA.....................................66 Table 4.2.3.1. Cumulative migration by region – average per ZCTA ........................................................68 Table 4.3.1. Descriptive statistics for WFH trip reduction – average per ZCTA .......................................71 Table 4.4.1. Descriptive statistics for all variables used in the regression..................................................73 Table 4.4.2. Regression results, dependent variable = percentage change in all-day traffic volume in destination ZCTAs, daily, 2019 to 2021.....................................................................................................75 Table 4.4.3. Regression results, dependent variable = percentage change in all-day traffic volume in destination ZCTAs, yearly average, 2019 to 2021......................................................................................75 Table 3.1.1 Unweighted responses by working arrangements and moving status......................................91 Table 3.2.1. Completion and qualification rates of the survey ...................................................................93 Table 3.2.2. Weighted shares (weighted responses divided by 2,124 total respondents)...........................95 Table 4. Summary of variables...................................................................................................................97 Table 4. Summary of variables (continued)................................................................................................98 Table 4.1.1. Share of different types of work arrangements from pre-COVID to post-COVID ................99 Table 4.1.2. Shift in work arrangement from pre-COVID to post-COVID..............................................100 Table 4.1.3. Current work arrangement preference - interior rows sum to 100% ....................................100 Table 4.1.4. Anticipated future work arrangement ...................................................................................101 Table 4.2.1. Regression results between remote work ability and industry types (Dependent variable: Remote work ability) ................................................................................................................................104 Table 4.2.2. Regression Results of remote work ability ...........................................................................106 Table 4.2.3. GHG emission of owned vehicles by work arrangement......................................................107 Table 4.2.4. GHG emission ranking of owned vehicles by work arrangement ........................................108 Table 4.2.5. Regression results of vehicle GHG emissions......................................................................109 Table 4.3.1. Moving status by current work arrangement ........................................................................110 Table 4.3.2. Moving status by change in work arrangement from pre-COVID to post-COVID..............110 Table 4.3.3. Moving distance (mile) by current work arrangement..........................................................111 Table 4.3.4. Future moving plan by work arrangement............................................................................111 Table 4.3.5. Future moving plan by work arrangement and housing tenure.............................................111 viii Table 4.4.1. Home-to-job distance (miles) by work arrangements...........................................................113 Table 4.4.2. Home-to-job distance (miles) by work arrangements excluding outliers (non-remote workers with home-to-job distance >= 300 miles).................................................................................................113 Table 4.4.3. Regression results of home-job distance...............................................................................115 Table 4.4.4. Commute mode from pre-COVID to post-COVID ..............................................................115 Table 4.5.1. Total daily commute trips, weighted and aggregated daily for pre- and post- COVID........116 Table 4.5.2. Total weekly commute VMT by work arrangement, in total weekly miles, excluding outliers (non-remote workers with home-to-job distance >= 300 miles)...............................................................118 Table 4.5.3. Total weekly commute GHG Emissions by work arrangement, excluding outliers (nonremote workers with home-to-job distance >= 300 miles) .......................................................................119 Table 4.5.4. Regression results of GHG emissions (without outliers: home-to-job distance >=300 for non remote workers) ........................................................................................................................................124 Table B1. Gender by Age Distribution .....................................................................................................182 Table B2. Race-Ethnicity Distribution......................................................................................................182 Table B3. Education Distribution .............................................................................................................182 Table B4. Income Distribution..................................................................................................................182 Table B5. Language Dominance Distribution ..........................................................................................183 Table C1. Vehicle GHG Category matching with original EPA Rating...................................................184 Table C2. Moving Frequency after COVID outbreak in March 2020......................................................184 Table C3. Moving frequency by current work arrangement.....................................................................184 Table C4. Home-to-job distance (miles) by change in work arrangements from pre-COVID to postCOVID......................................................................................................................................................185 Table C5. Commute mode change from pre-COVID to post-COVID .....................................................185 Table C6. Commute mode and work arrangement from pre-COVID to post-COVID.............................186 Table C7. Commute mode from pre-COVID to post-COVID..................................................................186 Table C8. Days of commute by work arrangements from pre-COVID to post-COVID ..........................187 Table C9. GHG ranking by commute days...............................................................................................187 Table C10. Home-to-Job distance by commute days................................................................................187 Table C11. Total weekly commute VMT by work arrangement, in total weekly miles...........................188 Table C12. Total weekly commute GHG Emissions by work arrangement.............................................188 ix List of Figures Figure 1. Map of the geographic extent of the study ....................................................................................8 Figure 2. Daily peak AM traffic volume from 03/04/2019 to 09/25/2021, without weekends and national holidays (StreetLight, 2019-2021)..............................................................................................................13 Figure 3. Daily number of new cases of COVID-19 per capita in the Bay Area and Central Valley (StreetLight, 2019-2021; California Department of Public Health, 2020-2021)........................................14 Figure 4. Weekly vaccination progress from January to June 2021 ...........................................................16 Figure 5. Estimated differences in trips per day by stage by income group, relative to lowest income ($0- 25k group). All estimates shown are statistically significant at the p<0.001 level (see Table 1). ..............27 Figure 6. Estimated differences in trips per day by stage by ratio of occupation share to Sales / Office occupation share. All estimates shown are statistically significant at the p<0.001 level (see Table 1)......28 Figure 3.1.2.1. Example of daily OD flows to ZCTA 94043......................................................................50 Figure 3.2.1. Study area counties................................................................................................................52 Figure 4.1.1.1. Left: 2019 Traffic volume at destination; Right: Traffic recovery at destination...............59 Figure 4.2.1.2. Work-from-home growth at origin (ACS)..........................................................................65 Figure 4.2.3.1. Cumulative net migration rate at origin..............................................................................69 Figure 3.2.1. Current home location of survey respondents.......................................................................94 Figure 4.5.1. Total daily commute GHG Emissions (CO₂ grams per mile) by day of week ....................117 Figure A1. Ratio of Biden to Trump votes in the 2020 Presidential election at ZCTA level...................143 x Abbreviations COVID-19 Coronavirus Disease 2019 GHG Greenhouse Gas OD Origin-Destination ZCTA Zip Code Tabulation Area VMT Vehicle Miles Traveled COA Change of Address WFH Work From Home xi Abstract The COVID-19 pandemic has profoundly transformed work and commuting patterns, reshaping urban dynamics in the United States. This dissertation investigates these shifts through three interrelated studies, focusing on the impacts of remote work on commuting behavior, residential choices, and environmental sustainability. The first study analyzes the effects of policy interventions and the widespread adoption of remote work on commute traffic in Northern California, emphasizing the inequitable outcomes for lower-income workers and essential sectors. The second study examines how remote work has influenced job and housing locations in the Bay Area and Central Valley, noting significant traffic volume reductions and migration toward more remote residential areas. The third study explores the environmental consequences of remote work, showing that fully remote workers reduce greenhouse gas (GHG) emissions by eliminating commutes, while hybrid workers also contribute to emissions reductions by commuting less frequently. Together, these studies offer insights into the evolving nature of commuting and residential dynamics, providing guidance for equitable and sustainable urban planning in the post-pandemic era. 1 Chapter 1. Introduction The COVID-19 pandemic has fundamentally reshaped how we work and live, leading to significant changes in commuting patterns and residential choices (Couture et al., 2021; Delventhal et al., 2021; Behrens et al., 2021; Mondragon and Wieland, 2022). The shift to remote work, the implementation of various policy interventions, and the subsequent changes in commuting behavior have had profound impacts on urban dynamics (Brynjolfsson et al., 2020, Barbieri et al., 2020; De Vos, 2020). This dissertation explores these changes through three interconnected studies, each focusing on different aspects of the pandemic's impact on urban life. The first two papers examine the changes in Northern California Megaregion, which includes the San Francisco Bay Area and California’s Central Valley, serving as an ideal study area due to its diverse demographics, economic conditions, and significant spatial inequalities. The third paper provides a broader perspective by looking at the entire country to offer comprehensive insights into the nationwide effects of remote work on commuting patterns, residential choices, and environmental sustainability. By examining the influence of remote work, commuting behavior, migration patterns, and environmental implications, this dissertation contributes to the understanding of how the pandemic has redefined urban dynamics and offers insights for future urban planning and policy. The COVID-19 pandemic saw one of the most significant changes in work and commute patterns ever experienced. The first confirmed case of COVID-19 in California was on January 26, 2020. The United States declared a national emergency on March 13, 2020, and state and local governments began issuing stay-at-home and social distancing orders by mid-March 2020. Commuting by the end of March 2020 was 75% lower in many U.S. metropolitan areas and even lower in many cities around the globe (Tomer & Fishbane, 2020; Sung & Monschauer, 2020; Bick, 2021). The disruption to public transit, fluctuations in traffic volume, and the equity implications that followed the interactions between private and public sector actions (e.g., remote work and stay-at-home orders) provide a unique opportunity to study how policy and economic structures influence travel behavior. Remote work, also known as work-from-home (WFH) and telecommuting, has become a prevalent norm, significantly altering job and housing locations. Approximately 4% of U.S. workers telecommuted in 2006, a share that increased to 6% by 2019 (U.S. Census American Community Survey). The pandemic caused a rapid shift to remote working, with many workers continuing to telecommute even after initial restrictions were lifted (Yilmazkuday, 2020; Srichan et al., 2020, Baker et al., 2020). This shift has implications for traffic, migration patterns, and environmental impact. Studies have shown that remote work reduces the need for proximity to job centers, leading to changes in residential location choices and increased demand for housing in less densely populated areas (Liu & Su, 2021; Delventhal, Kwon, & Parkhomenko, 2021). The shift to remote work also impacts commuting behavior and greenhouse gas (GHG) emissions. Remote workers tend to relocate farther from their workplaces, potentially reducing overall emissions due to fewer commute days. However, hybrid workers may still contribute to traffic congestion and emissions on the days they commute (Nilles, 1988; Bailey & Kurland, 2002; 2 Hopkins & McKay, 2019; Nguyen, 2021). Understanding these dynamics is crucial for developing sustainable transportation strategies and urban planning policies that accommodate the changing nature of work and commuting. This dissertation analyzes the impact of remote work on urban dynamics, specifically regarding commuting patterns, residential choices, and environmental sustainability. The COVID-19 pandemic has accelerated the adoption of remote work, leading to significant shifts in how people live and commute. The first paper examines how remote work and policy interventions influenced commute traffic, highlighting the unequal effects on different socioeconomic groups. The second paper explores how increased remote work has changed job and housing locations, resulting in decreased traffic volumes and more dispersed residential patterns. The third paper focuses on the environmental implications, showing that remote work can reduce greenhouse gas emissions by cutting down on commute-related travel. Together, these studies provide insights into the evolving nature of work and its implications for urban planning and policy in a post-pandemic era. Paper 1: Impact of Policy Interventions on Commute Traffic The first paper investigates how various policy instruments implemented by the State of California during the COVID-19 pandemic affected morning peak hour and home-based work traffic volume in the San Francisco Bay Area and California’s Central Valley. Using mobilederived traffic data from StreetLight Insights and neighborhood socioeconomic characteristics, the study models the variation in traffic volumes across different neighborhoods. The findings highlight that remote work, a policy initiated by the private sector, had a more significant and lasting impact on reducing commute traffic than state policies. Furthermore, the research underscores the inequities in commute burden, with lower-income workers and essential sectors experiencing less reduction in commute traffic. This paper contributes to the literature by providing a detailed empirical analysis of how policy and economic structure influenced travel behavior during the pandemic. Paper 2: Remote Work and Its Impact on Job and Housing Locations The second paper examines the implications of increased remote work on job and housing locations in the Bay Area and Central Valley regions. Utilizing datasets such as StreetLight for traffic data, LEHD LODES for job-related data, and USPS Change of Address for migration patterns, the study analyzes how work-from-home (WFH) trends have altered traffic volumes, migration, and workplace dynamics. The results indicate significant decreases in traffic volume, especially in the Bay Area, and suggest that remote work has led to a shift towards more remote living arrangements. The study finds that workplace industry characteristics are more influential on traffic volumes than worker demographics and remote working status, providing valuable insights for urban planning and transportation strategies in the post-pandemic era. 3 Paper 3: Remote Work, Commuting Behavior, and Environmental Impact The third paper explores the environmental implications of remote working by analyzing how changes in commuting behavior have affected greenhouse gas (GHG) emissions. Based on a survey conducted with a nationally representative sample, the research investigates whether remote workers tend to move farther from their workplaces and how this relocation impacts their commuting behavior and GHG emissions. The findings reveal that fully remote workers produce less GHG emissions due to the elimination of commutes, while hybrid workers, despite moving farther away, reduce their overall emissions by commuting less frequently. This paper highlights the potential of remote work to contribute to environmental sustainability and provides critical data for transit and planning agencies to model air quality and traffic congestion. This dissertation is among the first to examine the impact of working-from-home during a period when the initial shock of COVID-19 has subsided, but work and residential location dynamics are still in transition. While the long-term equilibrium is not yet clear, by 2023, most workers had settled into more permanent work arrangements. This dissertation provides insights into how the COVID-19 pandemic has reshaped commuting patterns, residential choices, and environmental impacts. The research underscores the need for equitable and sustainable urban planning policies that accommodate the changing nature of work and commuting in a postpandemic era. References Barbieri, P., Boffelli, A., Elia, S., Fratocchi, L., Kalchschmidt, M., & Samson, D. (2020). What can we learn about reshoring after Covid-19?. Operations Management Research, 13(3), 131-136. Bailey, D. E., & Kurland, N. B. (2002). A review of telework research: Findings, new directions, and lessons for the study of modern work. Journal of Organizational Behavior: The International Journal of Industrial, Occupational and Organizational Psychology and Behavior, 23(4), 383-400. Baker, M. G. (2020). Nonrelocatable occupations at increased risk during pandemics: United States, 2018. American Journal of Public Health, 110(8), 1126-1132. Behrens, Kristian, Sergey Kichko and Jacques-François Thisse, 2021. "Working from home: Too much of a good thing?." CESifo Working Paper No. 8831. Brynjolfsson, E., Horton, J., Ozimek, A., Rock, D., Sharma, G., & TuYe, H.-Y. (2020). COVID19 and Remote Work: An Early Look at US Data (No. w27344; p. w27344). National Bureau of Economic Research. https://doi.org/10.3386/w27344 Bick, A., Blandin, A., & Mertens, K. (2020). Work from home after the COVID-19 Outbreak. Couture, Victor, Jonathan I. Dingel, Allison Green, Jessie Handbury and Kevin R. Williams, 2021. "JUE Insight: Measuring movement and social contact with smartphone data: a real-time application to COVID-19." Journal of Urban Economics: 103328. 4 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. De Vos, J. (2020). The effect of COVID-19 and subsequent social distancing on travel behavior. Transportation Research Interdisciplinary Perspectives, 5, 100121. Hopkins, J. L., & McKay, J. (2019). Investigating ‘anywhere working’as a mechanism for alleviating traffic congestion in smart cities. Technological Forecasting and Social Change, 142, 258-272. Mondragon, J. A., & Wieland, J. (2022). Housing demand and remote work (No. w30041). National Bureau of Economic Research. Nilles, J. M. Traffic reduction by telecommuting: A status review and selected bibliography. Transportation Research Part A: General 22, 301–317 (1988). Liu, S., & Su, Y. (2021). The impact of the COVID-19 pandemic on the demand for density: Evidence from the US housing market. Economics letters, 207, 110010. Nguyen, M. H. Factors influencing home-based telework in Hanoi (Vietnam) during and after the COVID-19 era. Transportation 48, 3207–3238 (2021). Srichan, P., Apidechkul, T., Tamornpark, R., Yeemard, F., Khunthason, S., Kitchanapaiboon, S., ... & Upala, P. (2020). Knowledge, attitudes and preparedness to respond to COVID19 among the border population of northern Thailand in the early period of the pandemic: a crosssectional study. WHO South-East Asia journal of public health, 9(2), 118-125. Sung, J., & Monschauer, Y. (2020). Changes in transport behaviour during the Covid-19 crisis. Tomer, A., & Fishbane, L. (2020). Coronavirus has shown us a world without traffic. Can we sustain it?. Yilmazkuday, H. (2020). COVID-19‐19 and unequal social distancing across demographic groups. Regional Science Policy & Practice, 12(6), 1235-1248. 5 Chapter 2. Measuring the Impact of COVID-19 Policies on Local Commute Traffic: Evidence from Mobile Data in Northern California Author: Bonnie S. Wanga Co-authors: Seva Rodnyanskyb , Marlon G. Boarneta , Andre Comandona a Department of Urban Planning and Spatial Analysis, Sol Price School of Public Policy, University of Southern California, 650 Childs Way, Los Angeles, CA 90089, USA b Department of Urban and Environmental Policy, Occidental College, 1600 Campus Road, Los Angeles, CA 90041, USA Declaration of Interest The authors declare that they are no competing financial/personal interests or belief that could affect the objectivity of this paper. Declaration of Funding Source This research is funded by USDOT Grant: 69A3551747109 and Caltrans: 65A0674 Author Contributions Bonnie S. Wang: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Writing - original draft, Writing - review & editing, Visualization. Seva Rodnyansky: Conceptualization, Validation, Writing - review & editing, Supervision. Marlon G. Boarnet: Conceptualization, Validation, Writing - review & editing, Supervision. Andre Comandon: Conceptualization, Validation, Writing - review & editing, Supervision. Acknowledgments We thank StreetLight for making their data available for this project. 6 Abstract The COVID-19 pandemic upended the world economy and significantly altered how people work by imposing limits on in-person interactions. The policy response in places like California illustrate a range of scenarios that are applicable more broadly, from stringent restrictions on travel to flexible directives. This study focuses on the impact of various policy instruments the State of California implemented on morning peak hour and home-based work traffic volume in the San Francisco Bay Area and California’s Central Valley. Previous studies have not been able to measure COVID-19’s impact on commute traffic at a sufficiently local level, nor have they tested the impact of policy instruments on local commute traffic. We fill this gap by using mobile-derived morning peak hour traffic volume data from StreetLight Insights along with census-based neighborhood socioeconomic characteristics to model neighborhoodlevel variation in morning traffic volume and distance traveled. Remote work, a policy the private sector initiated, had a more significant and lasting effect than state policy. Morning traffic volume did not recover in response to policies aiming to provide businesses and people greater flexibility and predictability, and while vaccination was associated with a recovery of morning peak hour and home-based traffic, that recovery was short of pre-COVID levels. The effect of remote work undergirds inequities in commute burden. Economic sectors deemed essential and lower-income workers saw the lowest decreases in commute traffic and fastest recovery. Policymakers and planners ought to consider the occupation mix as a key ingredient in understanding the impact of telework ability and takeup on future transportation infrastructure improvements across various regions. Keywords: commute, COVID-19, travel behavior, socioeconomic inequality, regional disparity 7 1. Introduction The COVID-19 pandemic saw one of the most significant changes in work and commute patterns ever experienced. The first confirmed case of COVID-19 in California was on January 26, 2020. The United States declared a National Emergency on March 13, 2020. State and local governments began to issue stay-at-home and social distancing orders by mid-March 2020. Commuting by the end of March 2020 was 75% lower in many US metropolitan areas and even lower in many cities around the globe (Tomer and Fishbane, 2020; Sung and Monshauer, 2020). Anecdotal reports indicate that pandemic policies and the pandemic itself exacerbated inequities in commuting and is likely to have long-term effects on individuals and cities. Workers had vastly different experiences depending on their occupation, income level, and education, with the heaviest burden falling on lower-income people who lacked economic and social safety nets (e.g., Yilmazkuday, 2020; Srichan et al., 2020, Baker et al., 2020). The disruption to public transit, fluctuations in traffic volume, and the equity implications that followed the interactions between private and public sector actions (e.g. remote work and stayat-home orders) can inform how states and firms respond to future health and climate-induced emergencies. We use daily origin-destination traffic volume data collected during the peak morning hours (6 am - 10 am) over a time of rapid policy changes – from before the pandemic in 2018-2019 to the transition to endemic COVID-19 in 2021 – to study the joint impact of policy interventions, health concerns, and economic structure in a mega-region that shares characteristics with many other urban agglomerations worldwide. This study contributes to the need for detailed empirical research that disentangles how policy affected people differently and bolsters our ability to inform equitable policy decisions for future events (Harrington and Hadjiconstantinou, 2022, Hu and Chen, 2021). The Northern California Megaregion spans the San Francisco Bay Area east to California’s Central Valley, a demographically and economically diverse urban region, emblematic of large, interconnected megaregions worldwide with significant spatial inequalities (Figure 1). Our analysis focuses on policy interventions that range from the strictest stay-at-home orders during the early days of the pandemic to the implementation of various economic reopening strategies (such as California’s Blueprint for a Safer Economy and vaccination distribution) in the later stages of the pandemic. The Northern California Megaregion’s employment base is anchored by the high-tech clusters of San Francisco, San Jose, and Silicon Valley, with a hinterland of industrial, logistics, and local government services clusters in the Central Valley and Sacramento sub-regions. The two economies are separated by topography, agricultural land, and environmentally sensitive areas, but connected by key highway and rail corridors buttressed by residential sprawl. Despite its luster as “Silicon Valley,” home to the headquarters of the largest technology companies (e.g., Google, Apple, and Meta), the megaregion shares much with other regions around the world that rely heavily on their immediate hinterland to support the main urban centers. The resulting variation in remote work adoption and responsiveness to pandemic-driven policy makes this megaregion a worthwhile study area with the potential for generalization to other large, polycentric regions. This megaregion of 11 million people has experienced massive population growth and suburbanization in both its dense coastal cities and its low-rise hinterland towns over the past 30 8 years (Boarnet et al., 2023). The denser, more urbanized coastal portion of the megaregion is a major high-wage job creation engine, with regional GDP growth outpacing national GDP growth from 1997-2017, driven by productivity gains in the information sector (Bay Area Council Economic Institute and McKinsey and Co., 2018) and low unemployment rates (3.5%). Economic growth supported the highest median household incomes ($120,000 USD) and most expensive housing market (over $1 million USD median home value) in the U.S. as of 2019. The eastern, more rural portion of the megaregion focuses on basic industries (agriculture, food processing, manufacturing) and goods movement. The Central Valley has a slightly higher unemployment rate (5.7%), a lower median household income ($72,000 USD), and median home values one-third the level of the coastal part ($360,000).1 Yet, the coastal areas and hinterland are deeply integrated, with much cross-commuting, despite distances of over 100 kilometers between major urban centers. The coastal portion relies on the hinterland’s vast land supply for the logistics industry that supports the technology sector, but also to provide housing that the pricey coastal portion has failed to construct. The hinterland, including Sacramento has become an important destination for people and firms moving out of the high-priced and dense coastal area (Boarnet et al., 2023). Figure 1. Map of the geographic extent of the study 1 Based on the American Community Survey 2015-2019 5-year estimates. 9 The diverse and highly unequal economic structure of the Northern California Megaregion encompass a wide range of circumstances and, therefore, responses to the pandemic and related policies. Our analysis will answer the following questions: 1. How did traffic volume change as the prevalence of COVID-19 fluctuated and in response to the introduction of vaccines? 2. How did traffic volume change in response to the policy interventions the state and private sector used during COVID-19? 3. How did traffic volume vary across different income and occupation characteristics? While we cannot measure commute trips directly or individually, we analyze the change in daily neighborhood-level peak morning and home-based work trip volume to indirectly inform commute patterns and answer these questions. We analyze the variation in trip generation at the level of Zip Code Tabulation Areas (ZCTAs), a geographic area the US Census Bureau created based on the US Postal Service’s extent of mail delivery service areas. We use the ZCTA to capture the local environment people live in, making this one of the first papers to examine traffic volume longitudinally at a scale smaller than the city or county. We estimate daily trip generation across the study area using data from StreetLight Inc, which aggregates mobile phone data to provide information on trip volumes. The data were collected every day between March 4, 2019, and September 25, 2021, giving the analysis the full span of policy stages up to the most complete relaxation of restrictions. We link ZCTA-level traffic data with demographic and COVID-related data sources, to gain insights into the variations across different income and occupation categories. We find that, after controlling for local factors, the prevalence of COVID-19 locally had relatively little effect on traffic volume. Policy interventions had disparate effects. The results suggest that it was the voluntary adoption of remote work in the private sector, even before state policy interventions that had the greatest and most lasting effect. It appears that the strict stay-athome order cemented the effect of remote work so that the guidelines the state later introduced to increase flexibility and predictability for businesses had little direct effect on traffic volume. It was not until the introduction of vaccines that traffic volumes inched upward, but without ushering a complete recovery to pre-COVID volumes. Results focused on the ability to work remotely highlight the lasting effect or work from home options as areas with higher ability continued to have substantially lower commute volume after the introduction of the vaccine and despite their initial higher volume pre-pandemic. Differences between socioeconomic classes in commute behavior during COVID were one of the most striking results we uncovered. Those differences reflect the greater ability of higher-income people in managerial, professional, and scientific occupations to work remotely, indicating inequities in telework feasibility that will likely persist beyond the pandemic unless policymakers proactively work to address those inequities. 10 2. Literature Review The spread of COVID-19 to California in early 2020 led to a series of policies aiming to curb the spread of the disease. The restrictions on mobility and social interactions had an immediate and dramatic effect on the economy and, by extension, people’s travel behavior (Barbieri et al., 2020; De Vos, 2020). Within two weeks, workplace activity in the Northern California Megaregion plunged by about 55% on average with significant variation. Bay Area counties surrounding San Francisco, San Jose, and Oakland decreased by 64% but in Central Valley counties like Sacramento, its suburbs, and the Stockton area, the decrease was 44%, according to the analysis of Google Community Mobility Reports (McKinney, Morton, and Rodnyansky 2021). The variation in policy impact on commuting patterns is tied to the heterogeneity of people and economic structure. Higher-income and higher-educated populations, younger people, and people working in information and technology sectors were the most likely to shift to working at home and were relatively unaffected in their employment stability (Brynjolfsson et al. 2020, Yilmazkuday, 2020; Srichan et al., 2020, Baker et al., 2020). People who were able to shift to a work-from-home model decreased their travel dramatically in the early days of the pandemic as work, school, and non-essential shopping-related trips were eliminated. They were also slower to return to pre-pandemic travel habits as many workplaces allowed full or part-time remote working (Smart, 2022). In contrast, many lower-income workers faced compounded crises. Lower-wage workers in industries like hospitality and retail faced a much higher incidence of unemployment (Couch et al., 2020). The overrepresentation of people of color, especially Latinos, in these industries meant that the burden of the stay-at-home policies fell disproportionately on these populations (Gupta et al., Yasenov, 2020). The unemployment racial gap was lower for African Americans who tended to work in industries more sheltered from the economic downturn (Couch et al., 2020). Employment stability for low-wage workers, however, often did not come with the protection of remote work. Workers deemed essential were often at greater risk of exposure to the virus and continued to commute to work much like before the pandemic (Plyushteva, 2022). Essential workers who did not own a car had little choice but to rely on transit. Transit service was often reduced, limiting the options for transit-reliant people, many of whom switched to car transportation (Parker et al., 2021). Despite the reduction in transit and higher unemployment, declines in traveling were significantly smaller for lower-income people in the Seattle region (Brough et., 2021). The pandemic led to a shifting perception of occupation and a rapid re-organization of jobs that were considered critical and required constant in-person engagement (e.g., health care and grocery stores) and those that were primed for termination. These changes not only had immediate effects on people but are also likely to have long-term effects on job perception and people’s willingness to travel (Kramer and Kramer, 2020, Shamshiripour et al., 2020). The ability to work remotely, in particular, is likely to create permanent differences in travel behavior (Brough et al., 2021) 11 While economic structure had the most direct link to policy interventions, other factors were likely to affect travel behavior. COVID-19 itself affected willingness to travel. People developed an early awareness of risk and adjusted accordingly, but with limitations, based on available information. County-to-county travel decreased significantly in response to higher case rate (Brinkman and Mangum, 2022). Yet, changes in travel frequency were less clear for new cases in contrast to new deaths, with some variation based on the distance traveled (Truong and Truong, 2021). Perceived risk also varied depending on the built environment. The early peaks in infection and death in large, dense cities shaped people’s perceptions. People living in counties with more compact development reduced their trips to grocery stores and transit significantly more than people in less compact counties (Hamidi and Zandiatashbar, 2021). While vaccination rates plainly showed the partisan nature of the pandemic, the role of political ideology was visible early on. People in states where Donald Trump received greater support were less likely to respect stay-at-home orders and reduced traveling significantly less (Hill et al., 2021). Research on the impact of COVID-19 and commuting aligns well with a larger narrative surrounding the pandemic in the United States. The pandemic exacerbated health disparities and economic precarity. People of color and low-wage workers shouldered a disproportionate share of the burden of stay-at-home and social distancing policies imposed. There is, however, limited research to support these narratives with the kind of specificity that can inform more equitable policy decisions in the future. Most published research focuses on the early days of the pandemic at large geographic scales (state and county), and much of it relies on data for countries other than the United States. Research on COVID-19 and mobility specifically has used small survey data that are only representative at large scales or data that are not meant to measure travel behavior directly (e.g., SafeGraph). This study is the first to analyze travel behavior at a small scale, using data designed to measure trip volume for the entire period during which policies were actively implemented. 12 3. Data and Methodology 3.1 Hypotheses and Analysis Plan This paper tests three broad hypotheses regarding the impact of the COVID-19 pandemic on traffic volume. 1) Health risks and interventions: traffic volumes followed COVID-19 case rates as infected people were unable to work and the local population was afraid of catching and transmitting the virus. Once a health intervention (the vaccine) was available, people let down their guard and traffic volumes recovered. 2) Closing the economy: traffic volumes responded to the private sector encouraging remote work and the government mandating the shutdown of certain businesses (e.g., restaurants) and schools. Traffic volumes increased following the economy’s reopening. 3) Differential impact of stay-at-home orders, health risks, and interventions by income and occupation: travel behavior and response to health and policy interventions followed income and occupation characteristics. Higher-income households and knowledgeintensive workers were more flexibly able to adjust their commute behavior to engage in social-distancing orders during COVID-19. Lower-income households and essential workers were less flexibly able to change the way they commuted and end up with less participation in the social-distancing orders. Unpacking these hypotheses, we believe, sheds light on the continued evolution of travel as pandemic recovery continues and provides guideposts for future pandemics or similar-scale economic disruptions. To test these hypotheses, we defined five stages to analyze changes in trip volume based on the chronology of the pandemic and policy interventions in California. The first confirmed case of COVID-19 in California was on January 26, 2020, and the number jumped to two digits around March 6th to 9th in the Bay Area counties. We set the beginning of the outbreak to the week of March 9th, 2020. Traffic volumes were already lower in early March 2020 compared to a year earlier, but the variation in the peak AM (6am-10am) traffic volume between December (when the first case was reported in Wuhan, China) and the second week of March (when the number of cases reached double digits in the Bay Area) was not significantly different from preCOVID levels (see Figure 2). The early adoption of remote work in the tech sector may explain the slightly lower traffic volume before the cliff-like drop starting on March 10th. By the time the state announced the first shelter-at-home order on March 19th, traffic volume was already 62% lower than the baseline set at March 2019. 2 The subsequent stages are set based on the timing of policy interventions and are as follows: • Stage 0. Pre-COVID-19: 03/04/2019 – 03/09/2020. This period is a full year before the COVID-19 outbreak, which is the baseline for pre-pandemic traffic patterns. 2 See Appendix for home-based work trip trends by county over time. 13 • Stage 1. COVID-19 outbreak: 03/10/2020 - 08/30/2020. This period is marked by the rapid state and private sector responses to the severity of COVID-19. Policies were blunt and extreme. • Stage 2. Start of Blueprint: 08/31/2020 - 12/31/2020. This period is between the launch of the Blueprint framework and the start of vaccination. Stringency varied by county and time. • Stage 3. Start of Vaccination: 01/01/2021 - 06/14/2021. This period is between the start of vaccination and the retirement of the Blueprint measurement. • Stage 4. Fully Reopen: 06/15/2-21 - 09/25/2021. This period is between the lifting of all statewide restrictions (retirement of Blueprint) and the last day of the collected data. This period marks the lifting of state interventions, but the continuation of private sector policies allowing remote working. We then set up a regression model to capture the change in trip volume across socioeconomic groups in different COVID-19 periods. The focus is on estimating the interaction between income, occupation, travel distance, and the five policy stages we defined while controlling for a variety of ZCTA characteristics. The rest of this section highlights the pandemic impact, policy response, and traffic volume data for the study area and provides detail on the regression and data used therein. Figure 2. Daily peak AM traffic volume from 03/04/2019 to 09/25/2021, without weekends and national holidays (StreetLight, 2019-2021) 0 20 40 60 80 100 120 Traffic Volume Normalized at March 2019 Daily Peak AM Traffic Volume Bay Area Central Valley 14 3.2 COVID-19 stages, Pandemic Policy, and Traffic Volume The main goal of the analysis is to disentangle the effect of policy interventions from the local socioeconomic and health environment. The state implemented an evolving policy framework that first aimed to curb infection aggressively and then focused on increasing flexibility for businesses and their customers. In light of the public health emergency, California Governor Gavin Newsom issued a broad stay-at-home order on March 19, 2020, for all residents outside of those working in the 16 federally-determined critical infrastructure sectors.3 This order required residents to shelter-inplace as much as possible and limit nonessential activities (e.g., attending social gathering and non-essential shopping or travelling). Nonessential businesses such as movie theaters, most retail stores, bars and restaurants, were also mandated to close. The order was extended several times and was lifted gradually starting in May 2020, with different regions of the state reopening at varying rates on a case-by-case basis. This stage should have had a uniform effect on traffic volume with variation coming from the occupational structure. Figure 3 shows that Central Valley workers endured a higher per capita average of daily new COVID-19 cases while continuing commutes on the road during the pandemic. The difference in traffic level between the Bay Area and Central Valley in Figure 2 may come from the higher share of tech workers in the Bay Area and essential workers in the Central Valley (particularly in the food industry). Figure 3. Daily number of new cases of COVID-19 per capita in the Bay Area and Central Valley (StreetLight, 2019-2021; California Department of Public Health, 2020-2021) 3 State of California Executive Order N-33-20. https://www.gov.ca.gov/wp-content/uploads/2020/03/3.19.20- attested-EO-N-33-20-COVID-19-HEALTH-ORDER.pdf 0.000% 0.020% 0.040% 0.060% 0.080% 0.100% 0.120% Daily new cases by total population Bay Area Central Valley 15 By August 31, 2020, the state government via the California Department of Public Health enacted the Blueprint for a Safer Economy, a flexible and data-driven approach to loosening stay-at-home restrictions county by county. Table 1 shows the initial reopening tiers and reopening rules based on new COVID-19 cases, test positivity, health equity, and eventually vaccine equity. Reopening tiers ranged from widespread risk (purple, Tier 1) to minimal risk (yellow, Tier 4). The lower the risk, the higher the possible resumption of in-person activity in the county. Counties’ tiers were evaluated weekly; counties could advance to a less restrictive tier after 3 weeks in the prior tier and returned to more restrictive tiers when COVID-19 conditions worsened.4 Each tier has specific rules for various sectors, such as retail, dining, and personal care services (Table 1). The Blueprint tiers were in effect until June 14, 2021, at which point all counties were fully reopened to all economic activity. Table 1. Blueprint for a Safer Economy initial tier and reopening rules example5 Tier Level Widespread Purple Substantial Red Moderate Orange Minimal Yellow New Cases per 100,000* More than 7 4 to 7 1 to 3.9 Less than 1 Positive Tests More than 8% 5 - 8% 2 - 4.9% Less than 2% Reopening Rules Gatherings Outdoor only: Max 3 households Outdoor: Max 25 people Indoor: Max 25% capacity Outdoor: Max 50 people Indoor: Max 25% capacity Outdoor: Max 100 people Indoor: Max 50% capacity All Retail Indoor: Max 25% capacity Indoor: Max 50% capacity Open indoors with Modifications Open indoors with modifications Restaurants Outdoor only with modifications Indoor: Max 25% capacity or100 people, whichever is fewer Indoor: Max 50% capacity or 200 people, whichever is fewer Indoor: Max 50% capacity Bars (no meal) Closed Closed Indoor: Max 25% capacity Indoor: Max 50% capacity Office Remote Remote Open indoors with modifications Encourage telework Open indoors with modifications Encourage telework * Case numbers are adjusted up or down based on testing volume above or below the state median. Figure 2 shows little variation in traffic volume between May 2020 and November 2020 when the state transitioned from a blanket shelter-at-home policy to its Blueprint variable framework. Peaks in transmission in the summer of 2020 and especially in January 2021 correlated with decreases in trip volume and times of lower transmission in the first halves of 2020 and 2021 show only timid recovery in trip volume (Figure 2). This lack of variation outside times of extremes and the close parallels between the Bay Area and Central Valley trends suggest that the policy may not have had much effect in the aggregate. 4 California’s Color-Coded County Tier System. California Department of Public Health. https://emd.saccounty.gov/EMD-COVID-19-Information/Documents/California-Color-Coded-Tier-System--en.pdf 5 California’s Color-Coded County Tier System. California Department of Public Health. https://emd.saccounty.gov/EMD-COVID-19-Information/Documents/California-Color-Coded-Tier-System--en.pdf 16 Traffic volume began to recover in late February 2021. At that point the vaccine was becoming more widely available. Figure 4 shows the weekly vaccination progress in the Bay Area and Central Valley. California started to deliver vaccines on January 1, 2021 but the majority of the general public did not receive their first dose until March 2021. The share of the population fully vaccinated reached 38% in the Bay Area and 30% in the Central Valley by late April 2021, while by that same date the share partially vaccinated (with one dose) was 20% and 10% respectively. By the end of June 2021, 70% of the Bay Area residents were fully vaccinated, 20 percentage points more than the Central Valley share. Figure 4. Weekly vaccination progress from January to June 2021 3.3 Model Specification To better understand the impact of the pandemic and related policies on traffic volume, we set up a regression model to capture the change in trip volume across socioeconomic groups in different COVID-19 stages. We use OLS regression with time-fixed effects to estimate traffic volume across ZCTAs in the study area. Time-fixed effects are applied to control for the variance that is constant across observations but varies over time. This is important because the response to COVID-19 was not based only on local and state factors, but also international and national trends. COVID-19 hit the northeastern US and Europe earlier than California, which could have influenced people’s response. The fixed effect captures this overall context. We also use robust standard errors to allow non-constant variance across observations. The regression model is defined as follows: 𝑉𝑜𝑙𝑖𝑡 = 𝛽0+ 𝛽1𝐶𝑗+ 𝛽2∑(𝐶𝑗 ∗ 𝐼𝑖 )+ 𝛽3∑(𝐶𝑗 ∗𝑂𝑖 )+𝛽4∑(𝐶𝑗 ∗ 𝑅𝑖 )+𝛽5∑(𝐶𝑗 ∗𝐷𝑖𝑡) + 𝛽6∑(𝐶𝑗 ∗ 𝑏𝑝𝑖𝑡)+ 𝛽7𝑤𝑖𝑡+𝛽8𝑣𝑖𝑡+ 𝛽9𝑥𝑖+ 𝛽10𝑛𝑖+𝜀𝑖𝑡 0% 10% 20% 30% 40% 50% 60% 70% 80% 5-Jan 12-Jan 19-Jan 26-Jan 2-Feb 9-Feb 16-Feb 23-Feb 2-Mar 9-Mar 16-Mar 23-Mar 30-Mar 6-Apr 13-Apr 20-Apr 27-Apr 4-May 11-May 18-May 25-May 1-Jun 8-Jun 15-Jun 22-Jun 29-Jun % Partially Vaccinated Bay Area % Partially Vaccinated Central Valley % Fully Vaccinated Bay Area % Fully Vaccinated Central Valley 17 The dependent variable, 𝑉𝑜𝑙𝑖𝑡, is measured as either daily peak AM trips or home-based work trips (see details below). The subscript i represents the ZCTA and t represents the day. 𝐶𝑗 indicates one of five COVID-19 stages. The pre-COVID-19 time period dummy is the omitted category. 𝐷𝑖𝑡 is the average travel distance of the traffic departing/leaving. 𝑤𝑖𝑡 represents the daily COVID-19 cases per 100,000. COVID-19 cases were reported daily at the county level. Each ZCTA within the same county is assigned the same county-level daily covid rate. 𝑣𝑖𝑡 represents the share of the population fully vaccinated. The vaccination rate is originally reported weekly at the ZCTA level and so each day within a week has the same vaccination rate. 𝑏𝑝𝑖𝑡 indicates the assigned blueprint tiers. Blueprint tiers are originally reported weekly at the county level and assigned to days and ZCTA accordingly. All socioeconomic variables are measured pre-COVID-19. Ii indicates the household income level divided into five categories: <$25k (omitted category), $25-49k, $50-74k, $75-99k, and > $100k. 𝑂𝑖 represents the ratio of the number of workers for each occupation to the number of workers in sales and office occupations. The American Community Survey (ACS) provides the number of workers (civilian employed population 16 years and over) under five occupation types based on the 2018 Standard Occupational Classification (SOC) system: 1) management, business, science, and arts occupations, 2) service occupations, 3) sales and office occupations, 4) natural resources, construction, and maintenance occupations, and 5) production, transportation, and material moving occupations. According to the U.S. Bureau of Labor Statistics, SOC describes the occupation held by individuals but not the industries (NAICS codes) in which people work. We choose sales and office occupation as the baseline for normalization because the share of workers in sales and office varies the least across ZCTAs among all five occupation categories. 𝑅𝑖 indicates the total share of workers with ability to perform their critical job function outside the workplace, usually from home. To assess the impact of health and policy interventions across incomes and occupations, we interact COVID-19 stage (Cj) with ZCTA income (Ii) and occupation (Oi). We interact COVID-19 stages with Blueprint tiers (bpit) to separate the role of vaccination versus economic reopening. We also interact COVID-19 stage with ZCTA average distance (𝐷𝑖𝑡) to test whether locations where people tend to have longer commutes showed greater or lesser resilience in terms of commute trips compared to locations with shorter average commutes. Previous research has suggested that workers with higher income tend to have longer commutes as they often reside in the suburbs (Boarnet et al., 2023). Additionally, higher-income workers are more likely to have jobs that offer remote work options, allowing them to commute less frequently. In contrast, lower-income workers often need to live in closer proximity to their workplace and are more likely to be employed in jobs that require more in-person and on-site operation. 𝑥𝑖 is a set of pre-COVID-19 demographic control variables at the ZCTA level which may affect traffic volume and/or response to health or policy interventions, including total population, total employment, median age, percentage of essential workers, percentage of Asians, African Americans, and Latinos, percentage of workers commuting by public transit, percentage of renters, percentage of population below poverty, percentage of people with a college degree or above, median number of rooms, median gross rent (USD) of renter-occupied housing units, and median home value (USD) of owner-occupied housing units, and the ratio of Biden to Trump votes in the 2020 presidential election. 𝑛𝑖 is a set of variables to control for the geographic 18 context of ZCTA, population density, distance from the closest job center, and distance to the closest primary city center. 3.4 Data and Descriptive Statistics Travel Data We use data from StreetLight InSight® to estimate the daily trip generation throughout the study area. StreetLight is a private firm specializing in mobility metrics and analysis, using Global Positioning System data from phones to create measures of flow between locations. The platform provides information about travelers’ origin and destination, travel distance, and travel purpose. The data is available from 2016 and is updated monthly. StreetLight Inc. is unique in providing reliable estimates of origin and destination traffic flows and their data is not publicly accessible. Other firms provide phone-based mobility data (e.g. SafeGraph, Cuebiq, etc.), some of which are freely available to researchers, but each firm specializes in different metrics and replication with a different source should take these differences into account. The data have gone through extensive validation against multiple public and private entities, including state and local governments, transportation agencies, enterprises, and transportation consulting firms across the U.S and Canada (StreetLight, 2022). Our analysis focuses on the variation in trip generation at the Zip Code Tabulation Areas (ZCTAs) level. Zip Code Tabulation Areas are the U.S. Census’ generalized areal representations that are equivalent to the U.S. Postal Service’s (USPS’s) 5-digit zip codes. USPS ZIP Codes are just a set of mail delivery routes while ZCTAs represent the actual areal features (U.S. Census Bureau, 2018). Our agreement with StreetLight made 500 spatial units available to aggregate their individual level data. StreetLight can aggregate individual data to any geographic unit, but in the interest of linking their data to Census information and to maximize geographic coverage of the megaregion, the ZCTA was the best unit. We included all ZCTAs that intersected with an urbanized area (as defined by the US Census Bureau) or that had population of at least 3,000 in 2010 resulting in a sample of 491 ZCTA. The included ZCTA have an average population of about 30,000 and, because of their tie to postal delivery routes, their area tends to be inversely proportional to their population density. StreetLight’s database includes about 65 million devices in the US and Canada, covering approximately 23% of combined adult population in these countries. The daily trip penetration rate could vary, but on average it is between 1% to 5% of all trips on any one specific day (StreetLight, 2018). We used the platform’s Origin-Destination capabilities of personal vehicles to estimate the daily total number of trips out of every ZCTA during peak AM time (6am-10am) and for all home-based work trips (excluding weekends and national holidays). The two definitions serve as proxies for commute trips. However, home-based work trips were identified by StreetLight’s own Location-Based Services algorithm which is not accessible to the users. Therefore, our analysis focus on Peak-AM traffic, which offers a more transparent and easier definition for replication. The definitions also address the shifting commuting pattern over the course of the pandemic. Data have shown a flattening of peak congestion in the morning without a decrease in the overall number of trips in the later phases of the pandemic. This flattening suggests that many workers retained some flexibility for when they drive, which may muddle the 19 connection between peak AM trips and commuting (Smith, 2021).6 COVID-19 Data The analysis integrates both the prevalence of COVID-19 throughout the period and vaccination rates. We measure prevalence using the California Department of Public Health daily counts of COVID-19 cases, deaths, and testing at the county level since February 1, 2020. The data is not reported on weekends or state holidays. We also include information from weekly updates on full, partial, and at least one vaccine dose coverage rate by ZCTA for the whole state since January 5, 2021. Socioeconomic Data We use Census data from U.S. Census Bureau’s ACS 5-year average for the years 2015- 2019 to examine the interactions between policy and demographics. The main variables focus on economic status as measured by income and occupation. We reduce the 16 income categories the US Census reports to 5 categories (below $25,000, $25,000 to $50,000, $50,000 to $75,000, 75,000 to $100,000, and above $100,000) to enable the interaction of income categories with the COVID-19 stages. Similarly, we use the more rudimentary five class occupation categorization from the census rather than the 36 available categories for the same reason. We also include survey data from the 2021 U.S. Bureau of Labor Statistics Occupational Requirements and California’s “Essential Critical Infrastructure Workers” guideline to account for the varying potential of remote work across different occupations.7 In our Study Area, this amounts to about 14% of workers having remote feasibility (Table 2) with higher rates in Bay Area (17%) than in the Central Valley (11%) (See Appendix table A2). About 24% of Study Area workers were labeled as “Essential” according to California’s guidelines. In addition to income and occupation, we include other variables that stand out in the literature: age, race and ethnicity, employment, educational attainment, reliance on public transit, housing tenure, and housing price. We add data on the 2020 election to capture the political leaning of ZCTA. The 2020 Election data from the Voting and Election Science Team at the University of Florida and Wichita State University is available at the precinct level for all States. We use the presidential election result in California and aggregate precinct-level data into ZCTA-level data (see Figure A1 in the Appendix). The ratio of Biden to Trump votes is one potential proxy for voters’ attitudes toward candidates’ COVID-19 policies. The dataset is publicly available on the Harvard Dataverse. We complement data on the demographic composition of ZCTA with contextual variables. Existing research shows that density and how urban a place was relevant to how people adjusted their behavior during the pandemic. We include population density as a measure of development compactness. We also calculate the distance between the population-weighted 6 These patterns were noted after significant recoveries in trip volume were recorded, most of which happened after our study period. 7 BLS Occupational Requirements in the United States 2021. Table 1: Percentage of workers with selected flexibilities: https://www.bls.gov/news.release/archives/ors_11182021.pdf 20 median center of every ZCTA to the closest job center and the city hall of all principal cities.8 Table 2 shows summary statistics, unit, time period, and data source for each variable. The observations are ZCTA-days. The full sample 491 ZCTA, each represented for 932 days. The combination of unit and time period provides the unit for the mean and standard deviation. For example, there were 13,701 daily trips on average out of every ZCTA, but the ZCTA weekly average vaccination rate was 11%. 8 Job centers are defined as block with employment density over 5,000 per km2 within a job cluster larger than 10,000 total jobs. We included the city hall location of all cities that were listed in the name of each metropolitan statistical area (MSA). For example, the San Francisco MSA lists San Francisco, Oakland, and Fremont. 21 Table 2. Summary statistics of variables Category Variable Unit Time period Source Obs. Mean Std. dev. Trips Peak AM trips ZCTA Daily, 2019-2021 StreetLight 225,178 13,701.38 10389.99 Home-based work trips (all day) 225,178 14,060.22 10350.49 Average travel distance (mile) 225,178 2.13 0.37 COVID-19 Daily COVID-19 cases per 100,000 ZCTA (Transferred from county) Daily, since March 2020 California Department of Public Health 225,178 10.36 17.58 % Population fully vaccinated ZCTA Weekly, since January 2021 225,178 0.11 0.24 Demographics Total population ZCTA 2019 U.S. Census Bureau’s American Community Survey (ACS) 5- Year Estimates: 2015-2019 225,178 30,761.41 19,491.63 Median household income 222,913 98,591.68 43,344.08 Median age 224,346 38.64 6.75 % Asian 224,346 0.20 0.17 % Black 224,346 0.06 0.07 % Hispanic 224,346 0.26 0.18 % Commute by public transit 223,963 0.08 0.10 % Renter 223,459 0.42 0.20 % Below poverty 224,346 0.11 0.09 % College graduated 224,346 0.41 0.22 Median number of rooms 223,842 5.31 1.03 Median gross rent (renter) 220,414 1,772.26 591.72 Median home value (owner) 218,710 711,756.80 458,616.60 Occupation Total employment ZCTA 2019 U.S. Census Bureau’s American Community Survey (ACS) 5- Year Estimates: 2015-2019 225,178 15,212.58 9,940.32 business/science/arts 225,178 0.45 0.18 Service 225,178 0.16 0.07 Sales and office 225,178 0.19 0.05 natural resource/construction 225,178 0.09 0.07 production/transportation 225,178 0.11 0.07 % worker with telework ability ZCTA 2021 U.S. Bureau of Labor Statistics Occupational Requirements 223,963 0.14 0.06 % essential worker ZCTA 2021 California Department of Public Health 223.963 0.24 0.09 2020 presidential election Total votes ZCTA (Transferred from precinct) 2020 2020 Election data, The Voting and Election Science Team at the University of Florida and Wichita State University 216,075 14,800.94 8,834.05 Biden votes 216,075 10,209.13 6,507.57 Trump votes 216,075 4,269.06 3,513.66 Biden/Trump ratio 216,075 3.95 4.65 Neighborhood Population density ZCTA 2019 Authors’ calculations based on ACS 2015- 2019 5-Year Estimates 224,346 2,057.28 2,780.82 Distance to nearest job center 224,346 9,249.98 14,273.77 Distance to nearest city hall of principal /secondary city of each MSA 224,346 15,856.76 14,482.39 22 4. Results Table 3 shows the regression results for all four commute traffic models with daily fixed effects. Models 1 and 2 focus on examining only the COVID-19-related variables. Models 3 and 4 run the full equation with all explanatory variables. The dependent variable for models 1 and 3 is peak AM volume, which we consider our primary model. The dependent variable for models 2 and 4 is the StreetLight-defined home-based work trip, which serves as a robustness check. The results of the explanatory variables are discussed below. COVID-19 To test the hypothesis regarding health risks and interventions, our analysis focuses on the impact of the different COVID stages, weekly blueprint assignments, daily COVID cases, and weekly vaccination rates on Peak AM and home-based work trip volume. Model 1 focuses only on analyzing the traffic impact from these four COVID-19-related variables. Using preCOVID-19 peak AM traffic volume (which corresponds to the constant 19,663) as the baseline, ZCTA peak AM trips decreased by 40% in stage 1 (-8,060), and stage 2 (-8,741). The number remained low (-8,557) in the first 6 months after vaccine distribution began. The peak AM traffic grew back to 80% of the pre-COVID-19 level in the fully reopened stage 4 (-3,904). The patterns are similar in models 2, 3, and 4. There was no anticipation effect of COVID-19 Stage 1: alternate specifications that begin Stage 1 at times between December 2019 and March 2020 show no significant traffic volume changes. The COVID-19 rate (daily COVID-19 cases per 100,000) is significantly negatively related to peak AM and home-based work trips in all four models. Fully vaccination rate is significantly positively related to peak AM and home-based work trips in three models (Model 1, 3, and 4). The magnitude of the coefficient of the COVID-19 rate is relatively small in the full models. A change from the Blueprint Tier 1 level (less than 1 new case per 100,000) to Tier 4 (more than 7 new cases) would result in a drop of 56 trips in Model 3 (7*-8.301). In contrast, vaccination has a substantial impact on travel volume. Going from no vaccine to 50% of the population vaccinated (which happened within 4 months in the Bay Area) corresponds to an increase of 4,000 trips based on model 3 estimates (0.5*7972.705). Aggregate traffic volume suggested a minor impact of policy interventions. When looking at the Blueprint assignment in stage 2, model 1 shows that traffic volume in the riskiest tier (Tier 1) decreased most (-8,349), followed by Tier 2 (-7,838) and then Tier 3 (-7,390), compared to the lowest risk Tier (Tier 4). The full equation of model 3 indicates similar results. However, in stage 3, after vaccines were available, the difference between each tier became less obvious. In model 1, peak AM traffic decreased by 11,319 for Tier 1, 9,293 for Tier 2, and 9,375 for Tier 3. In model 3, peak AM traffic decreased by 8,561 for Tier 1, 8,170 for Tier 2, and 8,941 for Tier 3. The first two hypotheses aimed to disentangle two of the main mechanisms explaining the drop in traffic volume. A higher prevalence of COVID-19 as measured by the daily new number of cases resulted in lower traffic volumes, but even extreme levels would not explain the variation in traffic volume, let alone the dramatic drop from baseline levels. The role of policy is ambiguous. The widespread adoption of remote working in the private sector predated any state 23 intervention and seemed to have enabled lower traffic volume. Subsequent policy interventions were modestly effective in encouraging greater activity. The effects of moving from a stricter tier to a more relaxed one were negligible to small compared to the magnitude of the drop in overall traffic. Vaccination is the only intervention that worked to help traffic volumes recover. If longterm traffic volume remains closer to the September 2021 levels than they were in March 2019, vaccination will have effectively restored traffic volume to a new baseline level. It would also mean that policy interventions should be measured against this new baseline rather than the preCOVID-19 baseline (reflecting the permanent increase in remote work). Income In the pre-COVID-19 period (stage 0), ZCTAs with higher median household income produced more peak AM and home-based work trips than ZCTAs with lower median household income. Based on the result of model 3, ZCTAs with median household income >$100k are associated with 2,789 more peak AM trips than ZCTAs with median household income <$25k in stage 0. Throughout COVID-19 (stages 1-4), ZCTA median household income was negatively associated with traffic volume. ZCTAs in the highest income band (>$100k) had the largest traffic volume decreases, relative to the lowest income band (<$25k) (Figure 5). These higherincome ZCTAs decreased traffic volume by a factor of two relative to those with a median income of $25-100k. In the post-vaccine period (stages 3 and 4), peak AM trips in ZCTAs with lower median household income recovered relatively more than ZCTAs with higher median household income. This result is in line with findings from previous research that higher-income households have higher flexibility to work from home and are more likely to engage in social distancing. Lower-income households are less likely to engage in social distancing due to fear of losing income (Yilmazkuday, 2020; Srichan et al., 2020, Baker et al., 2020, Austrian et al., 2020). While the difference in levels is unsurprising, the lack of difference in change between the pandemic stages suggests that higher-income communities did not respond more or less than lower-income communities to policy interventions. Occupation Before the COVID-19 outbreak, ZCTAs with higher ratios of service workers to sales / office workers had larger traffic volumes, while those with higher ratios of natural resources, construction, and maintenance had the lowest traffic volumes. Throughout COVID-19, ZCTAs with higher ratios of natural resource / construction relative to sales and office had higher traffic volumes (Figure 6). The results indicate that, although primary (agriculture, fishing, mining, etc.) and secondary industries (construction, manufacturing, etc.) generate fewer commutes pre-COVID-19, they were deemed essential during the pandemic and often required on-site operations. As a result, these industries were 24 more resilient and experienced fewer commute disruptions compared to other tertiary industries (service, finance, retail, etc.). In contrast, ZCTAs with higher ratios of service and production / transportation occupations had the largest decrease in peak AM traffic volume. The reason for such decline could relate to fact that service activities involve a lot of face-to-face interactions, which was highly restricted during the COVID-19 outbreak. Interestingly, the home-based-work results show service occupations more in line with sales and office, potentially indicating that service workers still commuted, but shifted their commutes outside of the peak AM time window. This suggests that service workers may have had more flexibility in their work schedules during the pandemic. Additionally, it is possible that the type of service they provided changed, such as restaurants shifting from dine-in to delivery services. Business / science / arts traffic volumes were largely in line with those for sales and office occupations. These occupations generally require a high level of education and skills and often involve office-based work, suggesting higher flexibility in making the switch to remote work. In the post-vaccine period, the peak AM trips of service occupations bounced back the most. The relative peak AM volume of service occupation grew back in stage 4. The peak AM trips of natural resources, construction, and maintenance occupations remain high in stage 3. The comparative difference between them and other tertiary industries was closer to the pre-COVID19 level in stage 4. Prior to the pandemic, ZCTA with higher share of workers with telework ability generally had more traffic, which could be that workers were not utilizing their ability to work from home and were still commuting. During the pandemic, ZCTA with higher remote share did indeed see lower traffic volumes. The average ZCTA had 14% telework-eligible workers; multiplying this by the coefficients yields about a 6800 daily trip drop in Stage 1 and over 8200 trips dropped by Stage 4. To the extent that telework continues to be an option, these findings suggest a potentially commute trip decrease into the future, though this may be moderated by new non-work trips. Travel Distance We examine whether ZCTA with longer commute distances is associated with a higher or lower level of resilience compared to those with shorter commutes throughout COVID-19. The results showed that ZCTAs with Peak AM and home-based work trips of longer average travel distance generated fewer peak AM and home-based work trips during COVID-19. This result supports our hypothesis that higher-income workers, who tend to have longer commutes, were more likely to have jobs that allow to be perform remotely, enabling them to commute less frequently. ZCTAs with shorter peak AM and home-based work trip distances recovered more than longer peak AM and home-based work trip ZCTAs in the post-vaccine period. In stage 0, the pre-COVID-19 period, one additional mile in average ZCTA travel distance is associated with – 2,888 peak AM trips. In stage 1, it is associated with an increase of 546 peak AM trips departing from a ZCTA. This number gradually decreased in stage 2 and stage 3. In the fully opened stage 4, one-mile addition on average travel distance is associated with a decrease of 377 peak AM 25 trips departing from a ZCTA. This result is also in line with our hypothesis that lower-income workers tend to live closer to their workplace and are more likely to work in jobs that require them to be physically present on-site as soon as the social distancing order was lifted. Political Orientation and Location Despite the evidence that political persuasion significantly affected individual response to safety measures such as mask-wearing and school closing, ZCTAs political leaning had relatively low magnitude: doubling the Biden to Trump ratio was associated with a decrease in 102 – 170 daily trips (models 3 and 4). This may be, in part, because in California political leaning correlates with population density. Population density was negatively associated with traffic volume, a 1000 person / km increase is associated with a 767 – 869 decrease in trips per day (models 3 and 4), in line with Hamidi and Zandiatashbar’s results (2021). Distance to the nearest job center, a measure of rurality which also correlates with political leaning, was not strongly statistically significant in the model. Distance to downtown (principal cities of each metropolitan area) was statistically significant, but signs diverged between peak AM and homebased-work trips and the magnitudes were extremely low (a 100-mile distance was associated with a -0.5 peak AM trips and -0.4 home-based work trips). Taken together, partisan voting and population density were more important in explaining ZCTAs traffic volumes than its location within a regional hierarchy. Yet, partisanship and population density magnitudes were generally lower than those of income, occupation, and COVID-19-related variables. Other Demographics Controls ZCTA population, employment, essential worker share, renter share, median gross rent, and median home value are significantly positively related to peak AM and home-based work traffic volume. Magnitudes are on the order of income and occupation effects for population size (a 10,000 person population increase is associated with +3910 daily trips) and is triple that of employment. Every percentage point increase in renter share is associated with an increase of 24 – 113 trips (models 3 and 4). In essence, larger ZCTAs with higher employment maintained proportionately higher levels of economic activity and thus traffic volume. Higher shares of essential worker were positively correlated with traffic volume. A one percentage point increase in essential worker share was associated with 128 – 141 more peak AM and home-based work trips. Higher college education shares were negatively correlated with traffic volume (1 p.p. increase led to 67 – 93 fewer trips (models 3 and 4). A higher median age in a ZCTA was associated with fewer peak AM trips, suggesting that older individuals are being more cautious when commuting during the pandemic. More rooms in a housing unit were also associated with fewer peak AM trips, indicating higher income and greater potential of remote working. Higher non-white proportions were negatively correlated with traffic volume. A one percentage point increase in African American share was associated with 120 – 159 fewer trips, for Hispanic share -40 – 46, and -17 – 28 for Asian share. Transit use plummeted during the pandemic, yet core transit users may in fact be essential workers and/or have low flexibility to work remotely. Signs diverged for public transit commute share: one percentage point increases were associated with 42 more peak AM trips but 69 fewer home-based work trips. This hints at the possibility that essential workers with no 26 private car option still commuted by transit during peak AM, but that ZCTAs with high transit share decreased overall commutes. 27 Figure 5. Estimated differences in trips per day by stage by income group, relative to lowest income ($0-25k group). All estimates shown are statistically significant at the p<0.001 level (see Table 1). -8000 -6000 -4000 -2000 0 2000 4000 Stage 0: preCOVID-19 stage 1: COVID-19 outbreak stage 2: blueprint started stage 3: vaccination started stage 4: fully reopen Peak AM Traffic Volume $25-49k $50-74k $75-99k > $100k -10000 -8000 -6000 -4000 -2000 0 2000 4000 6000 Stage 0: preCOVID-19 stage 1: COVID-19 outbreak stage 2: blueprint started stage 3: vaccination started stage 4: fully reopen Home-Based Work Trip Volume $25-49k $50-74k $75-99k > $100k 28 Figure 6. Estimated differences in trips per day by stage by ratio of occupation share to Sales / Office occupation share. All estimates shown are statistically significant at the p<0.001 level (see Table 1). -8000 -6000 -4000 -2000 0 2000 4000 6000 Stage 0: preCOVID-19 stage 1: COVID-19 outbreak stage 2: blueprint started stage 3: vaccination started stage 4: fully reopen Peak AM Traffic Volume business/science/arts Service natural resource/construction production/transportation -8000 -6000 -4000 -2000 0 2000 4000 6000 8000 Stage 0: preCOVID-19 stage 1: COVID-19 outbreak stage 2: blueprint started stage 3: vaccination started stage 4: fully reopen Home-Based Work Trip Volume business/science/arts Service natural resource/construction production/transportation 29 Table 3. Regression Results Category Description w/ daily dummy Y= Number of trips depart from origin ZCTA (1) Peak AM vol (2) HBW vol (3) Peak AM vol (4) HBW vol COVID-19 stages (base=stage 0: pre-COVID-19: 03/04/2019 – 03/09/2020) stage 1: COVID-19 (03/10/2020 - 08/30/2020) -8060.226*** -7460.071*** -5248.653*** -5902.183*** stage 2: blueprint started (08/31/2020 - 12/31/2020) -8741.211*** -13451.860*** -7030.359*** -7500.690*** stage 3: vaccination started (01/01/2021 - 06/14/2021) -8556.899*** -5141.784*** -9333.939*** -8449.489*** stage 4: fully reopen (06/15/2-21 - 09/25/2021) -3904.365*** -708.866 -2929.103*** -1071.509 Blueprint (base = Tier 4) stage 2: blueprint started Tier 1 392.515 3017.681*** -1164.744*** -2024.465*** Tier 2 903.438** 2891.764*** -645.740** -1371.683*** Tier 3 1351.034*** 2663.431*** -132.376 -709.040** stage 3: vaccination started Tier 1 -2761.763*** -1817.589*** 787.994*** 533.499*** Tier 2 -735.693*** -104.335 1213.069*** 1010.377*** Tier 3 -817.681*** -580.539** 545.624*** 605.986*** COVID-19 rate Daily COVID-19 cases per 100,000 -48.928*** -22.775*** -10.277*** -14.236*** Vaccination rate % Population fully vaccinated 1103.296*** -865.724** 7911.979*** 9151.715*** Travel distance stage 0: pre-COVID-19 average travel distance (mile) -2888.450*** -2251.959*** stage 1: COVID-19 average travel distance (mile) 3434.477*** 2661.088*** stage 2: blueprint started average travel distance (mile) 3336.895*** 2591.497*** stage 3: vaccination started average travel distance (mile) 2973.803*** 2221.200*** stage 4: fully reopen average travel distance (mile) 2511.219*** 1745.749*** COVID-19 stages * Household income (base= stage 0: pre- COVID-19* income<$25k) Stage 0: pre-COVID-19: (03/04/2019 – 03/09/2020) $25-49k -110.134 -561.270*** $50-74k 963.277*** 258.248 $75-99k 1382.738*** -133.247 > $100k 2991.271*** 1121.168*** stage 1: COVID-19 outbreak. (03/10/2020 - 08/30/2020) $25-49k -3168.344*** -1391.060*** $50-74k -3946.149*** -1660.818*** $75-99k -4437.770*** -1074.795*** > $100k -6984.689*** -2821.436*** stage 2: blueprint started (08/31/2020 - 12/31/2020) $25-49k -3249.121*** -1366.749*** $50-74k -3964.094*** -1426.367*** $75-99k -4276.825*** -746.300*** > $100k -6659.757*** -2233.320*** stage 3: vaccination started (01/01/2021 - 06/14/2021) $25-49k -3160.480*** -1491.729*** $50-74k -3949.927*** -1637.529*** $75-99k -4428.077*** -969.353*** > $100k -6699.022*** -2316.876*** stage 4: fully reopen. (06/15/2021 - 09/25/2021) $25-49k -2115.316*** -1035.784*** $50-74k -2769.688*** -1169.593*** $75-99k -3586.185*** -1014.643*** > $100k -4956.417*** -1718.273*** 30 Table 3. Regression Results (continued) Category Description w/ daily dummy Y= Trips depart from origin ZCTA (1) Peak AM vol (2) HBW vol (3) Peak AM vol (4) HBW vol COVID-19 stages * Occupation (base=stage 0: pre-COVID- 19*sales/office occupation) Stage 0: pre-COVID-19: (03/04/2019 – 03/09/2020) business/science/arts -1057.174*** -372.128*** Service 2618.040*** 70.348 natural resource/construction -6029.867*** -5840.064*** production/transportation 1998.124*** 3109.224*** stage 1: COVID-19 outbreak. (03/10/2020 - 08/30/2020) business/science/arts 1500.892*** 834.897*** Service -4554.112*** -1941.426*** natural resource/construction 5166.279*** 3534.642*** production/transportation -4117.884*** -4082.722*** stage 2: blueprint started (08/31/2020 - 12/31/2020) business/science/arts 1420.481*** 729.035*** Service -4232.510*** -2048.560*** natural resource/construction 4999.230*** 3247.036*** production/transportation -3949.885*** -3773.614*** stage 3: vaccination started (01/01/2021 - 06/14/2021) business/science/arts 1329.493*** 531.392*** Service -4317.678*** -2318.165*** natural resource/construction 5424.458*** 3919.942*** production/transportation -4360.275*** -4282.874*** stage 4: fully reopen (06/15/2-21 - 09/25/2021) business/science/arts 868.301*** 153.180*** Service -3238.641*** -2167.969*** natural resource/construction 3546.775*** 2592.965*** production/transportation -3349.430*** -3238.923*** Telework ability stage 0: pre-COVID-19 % workers with telework ability 46997.495*** 57880.313*** stage 1: COVID-19 % workers with telework ability -40312.313*** -39262.584*** stage 2: blueprint started % workers with telework ability -37085.075*** -39031.720*** stage 3: vaccination started % workers with telework ability -39638.234*** -38854.753*** stage 4: fully reopen % workers with telework ability -49689.041*** -49020.498*** 31 Table 3. Regression Results (continued) Category Description w/ daily dummy Y= Trips depart from origin ZCTA (1) Peak AM vol (2) HBW vol (3) Peak AM vol (4) HBW vol Demographic controls Total population 0.391*** 0.459*** Total employment 0.080*** -0.070*** Median age -83.850*** -84.453*** % essential workers 12792.018*** 14135.494*** % Asian -2766.989*** -1707.318*** % Black -11975.244*** -15891.469*** % Hispanic -3983.570*** -4571.988*** % Commute by public transit 6417.472*** -4921.905*** % Renter 2435.437*** 11372.689*** % Below poverty -1155.511*** -4406.056*** % College graduated -6799.484*** -9395.695*** Median number of rooms -468.358*** -598.542*** Median gross rent (renter) -0.018 0.621*** Median home value (owner) 0.001*** -0.001*** 2020 presidential election Biden/Trump ratio -170.012*** -101.881*** Neighborhood controls Population density -0.767*** -0.869*** Distance to nearest job center -0.001 -0.004* Distance to nearest city hall of principal city of each MSA 0.005** -0.004* Constant 19663.936*** 18351.999*** 13964.694*** 10900.649*** Observations 215510 215510 203813 203813 R-squared 0.147 0.092 0.837 0.818 Adjusted R-squared 0.145 0.089 0.837 0.817 *** p<0.001, ** p<0.01, * p<0.05 32 5. Conclusion Our study provided insights on how government policies have affected commute traffic patterns (morning peak hour and home-based work traffic) in an urban region with diverse demographics and economic conditions during the COVID-19 pandemic. We disentangled the impact of policies from the local socioeconomic and health environment, analyzing interventions ranging from strict stay-at-home orders during the initial phase of the pandemic to the implementation of diverse economic reopening strategies. We examined the impact of these policies on local morning peak /commute traffic volumes and tested how different population groups have adapted to them. Our analysis revealed that local prevalence of COVID-19 has minimal impact on morning peak and home-based work traffic volume, while policy interventions have a significant influence. The introduction of remote work policies by the private sector led to an initial drop in traffic, and state-imposed stay-at-home orders further reduced commute traffic levels well before COVID-19 became widespread in California. The most effective intervention in promoting morning peak/commute traffic recovery was the distribution of vaccines. The state guidelines and the Blueprint reopening policy that replaced stay-at-home orders showed little direct effect on traffic volume. Remote work continued to be associated with lower morning peak/commute traffic volumes in the post-vaccine period. Subsequent policy interventions had only moderate effectiveness in increasing activity, with the effects of relaxed tier measures being negligible compared to the overall drop in traffic. Our findings confirmed the existing income disparity that lower-income workers were more likely to be disproportionately affected by social distancing policies. Higher-income households were more likely to engage in remote work and social distancing, while lowerincome households are less likely to engage with social distancing orders due to fears of losing income. However, higher-income groups did not respond differently to economic reopening policy interventions, compared with the lower-income group. The results by occupational category highlight the limits of policy. Primary and secondary industries, which generated fewer commutes pre-COVID-19, experienced fewer commute disruptions compared to tertiary industries. Primary and secondary sectors have few opportunities to work remotely and are more likely to be deemed essential, while many tertiary occupations typically require a higher level of education and skills and often involve officebased work. When combined with the greater ability of office workers and professionals to work remotely (and their higher income), ZCTAs with high employment shares in these sectors had significantly lower morning peak/commute traffic volumes throughout the pandemic than other groups. Service occupations experienced the largest drop in morning peak/commute traffic and fastest recovery due to the nature of the occupation and its sensitivity to policy interventions. At the same time, essential workers (e.g., health care) had to find alternative ways to return to work, adapting through flexible schedules or operational changes. In short, much of the fluctuation in traffic volume appears to have been driven by economic structure and demands than by public policy. 33 While the private sector drives morning peak/commuting traffic volume, public policy is far from inconsequential. Stay-at-home orders were effective in enforcing social distancing and the vaccine did much to return morning peak/commuting traffic closer to pre-pandemic levels. As many cities and nations struggle with the aftermath of the pandemic (e.g., office occupancy and public transit usage rate) and ponder what policy to pursue, this study highlights the importance of the interactions between private and public sector policy. The disparities in how different groups and areas responded to the pandemic suggest a public policy approach that is more attuned to how the labor market is structured and able to respond to crises. This study contributed to understanding how the complex set of factors affecting morning peak/commuting traffic volumes adapts to a large scale and unprecedented shock. While we used cutting-edge mobility data, there are still many factors we were unable to model and that can advance this type of study. The longitudinal nature of the data could not be matched for neighborhood and population characteristics. Labor force participation is an important factor that we could not include, for example. There were significant disparities in who dropped out of the labor force and returned over the course of the pandemic, which likely contributed to traffic volume fluctuations. We were also unable to conduct tests at the individual-level where the primary mechanisms reside. With increased access to anonymized individual-level data, there are many opportunities for future research to test how people respond to economic shocks with greater granularity and in ways that get closer to the causal mechanism. 34 7. References • Anwari, N., Ahmed, M. T., Islam, M. R., Hadiuzzaman, M., & Amin, S. (2021). Exploring the travel behavior changes caused by the COVID-19 crisis: A case study for a developing country. Transportation Research Interdisciplinary Perspectives, 9, 100334. • Austrian, K., Pinchoff, J., Tidwell, J. B., White, C., Abuya, T., Kangwana, B., ... & Ngo, T. D. (2020). COVID-19 related knowledge, attitudes, practices and needs of households in informal settlements in Nairobi, Kenya. • Barbieri, P., Boffelli, A., Elia, S., Fratocchi, L., Kalchschmidt, M., & Samson, D. (2020). What can we learn about reshoring after Covid-19?. Operations Management Research, 13(3), 131-136. • Baker, M. G. (2020). Nonrelocatable occupations at increased risk during pandemics: United States, 2018. American Journal of Public Health, 110(8), 1126-1132. • Bay Area Council Economic Institute and McKinsey and Co. (2018). Continuing Growth and Unparalleled Innovation: Bay Area Economic Profile. Tenth in a Series. • Bick, A., Blandin, A., & Mertens, K. (2020). Work from home after the COVID-19 Outbreak. • Blau, F. D., Koebe, J., & Meyerhofer, P. A. (2021). Who are the essential and frontline workers?. Business Economics, 56(3), 168-178. • BLS. (2022). Occupational Requirements in the United States News Release. U.S. Bureau of Labor Statistics. Retrieved April 22, 2023, from https://www.bls.gov/news.release/archives/ors_11172022.htm • Brinkman, J., & Mangum, K. (2022). JUE Insight: The Geography of Travel Behavior in the Early Phase of the COVID-19 Pandemic. Journal of Urban Economics, 127, 103384. https://doi.org/10.1016/j.jue.2021.103384 • Brough, R., Freedman, M., & Phillips, D. C. (2021). Understanding socioeconomic disparities in travel behavior during the COVID-19‐19 pandemic. Journal of Regional Science, 61(4), 753-774. • Brynjolfsson, E., Horton, J., Ozimek, A., Rock, D., Sharma, G., & TuYe, H.-Y. (2020). COVID-19 and Remote Work: An Early Look at US Data (No. w27344; p. w27344). National Bureau of Economic Research. https://doi.org/10.3386/w27344 • Boarnet, M., Rodnyansky, S., Austin, B., Comandon, A., Wang, B., & Rodriguez, S. (2023). Bay Area to Central Valley Migration and its Impacts. Final Report prepared for Chan Zuckerberg Initiative. March. https://sites.usc.edu/ca-regional-impact/research/ • Ceccato, R., Rossi, R., & Gastaldi, M. (2021). Travel demand prediction during COVID-19 pandemic: Educational and working trips at the University of Padova. Sustainability, 13(12), 6596. • Chen, J. T., & Krieger, N. (2021). Revealing the unequal burden of COVID-19 by income, race/ethnicity, and household crowding: US county versus zip code analyses. Journal of Public Health Management and Practice, 27(1), S43-S56. • Couch, K. A., Fairlie, R. W., & Xu, H. (2020). Early evidence of the impacts of COVID-19 on minority unemployment. Journal of Public Economics, 192, 104287. https://doi.org/10.1016/j.jpubeco.2020.104287 • De Vos, J. (2020). The effect of COVID-19 and subsequent social distancing on travel behavior. Transportation Research Interdisciplinary Perspectives, 5, 100121. 35 • Dingil, A. E., & Esztergár-Kiss, D. (2021). The Influence of the COVID-19 Pandemic on Mobility Patterns: The First Wave’s Results. Transportation Letters, 13(5-6), 434-446. • Echaniz, E., Rodríguez, A., Cordera, R., Benavente, J., Alonso, B., & Sañudo, R. (2021). Behavioural changes in transport and future repercussions of the COVID-19 outbreak in Spain. Transport Policy, 111, 38-52. • Gao, S., Rao, J., Kang, Y., Liang, Y., Kruse, J., Doepfer, D., ... & Yandell, B. S. (2020). Mobile phone location data reveal the effect and geographic variation of social distancing on the spread of the COVID-19 epidemic. arXiv preprint arXiv:2004.11430. • Garnier, R., Benetka, J. R., Kraemer, J., & Bansal, S. (2021). Socioeconomic disparities in social distancing during the COVID-19 pandemic in the United States: observational study. Journal of medical Internet research, 23(1), e24591. • Gupta, R., Ghosh, A., Singh, A. K., & Misra, A. (2020). Clinical considerations for patients with diabetes in times of COVID-19 epidemic. Diabetes & metabolic syndrome, 14(3), 211. • Hamidi, S., & Zandiatashbar, A. (2021). Compact development and adherence to stay-athome order during the COVID-19 pandemic: A longitudinal investigation in the United States. Landscape and Urban Planning, 205, 103952. https://doi.org/10.1016/j.landurbplan.2020.103952 • Hanibuchi, T., Yabe, N., & Nakaya, T. (2021). Who is staying home and who is not? Demographic, socioeconomic, and geographic differences in time spent outside the home during the COVID-19 outbreak in Japan. Preventive Medicine Reports, 21, 101306. • Harrington, D. M., & Hadjiconstantinou, M. (2022). Changes in commuting behaviours in response to the COVID-19 pandemic in the UK. Journal of transport & health, 24, 101313. • Hill, T., Gonzalez, K. E., & Davis, A. (2021). The Nastiest Question: Does Population Mobility Vary by State Political Ideology during the Novel Coronavirus (COVID-19) Pandemic? Sociological Perspectives, 64(5), 786–803. https://doi.org/10.1177/0731121420979700 • Hu, S., & Chen, P. (2021). Who left riding transit? Examining socioeconomic disparities in the impact of COVID-19 on ridership. Transportation Research Part D: Transport and Environment, 90, 102654. • Jay, J., Bor, J., Nsoesie, E. O., Lipson, S. K., Jones, D. K., Galea, S., & Raifman, J. (2020). Neighbourhood income and physical distancing during the COVID-19 pandemic in the United States. Nature human behaviour, 4(12), 1294-1302. • Kalter, M. J. O., Geurs, K. T., & Wismans, L. (2021). Post COVID-19 teleworking and car use intentions. Evidence from large scale GPS-tracking and survey data in the Netherlands. Transportation Research Interdisciplinary Perspectives, 12, 100498. • Kim, J., & Kwan, M. P. (2021). The impact of the COVID-19 pandemic on people's mobility: A longitudinal study of the US from March to September of 2020. Journal of transport geography, 93, 103039. • Lou, J., Shen, X., & Niemeier, D. (2020). Are stay-at-home orders more difficult to follow for low-income groups?. Journal of transport geography, 89, 102894. • McKinney, E., Morton, I., and Rodnyansky, S. (2021). “The Uneven Workplace Recovery from COVID-19: Bay Area versus Central Valley”. • Pan, Y., & He, S. Y. (2022). Analyzing COVID-19’s impact on the travel mobility of various social groups in China’s Greater Bay Area via mobile phone big data. Transportation Research Part A: Policy and Practice. 36 • Parker, M. E. G., Li, M., Bouzaghrane, M. A., Obeid, H., Hayes, D., Frick, K. T., Rodríguez, D. A., Sengupta, R., Walker, J., & Chatman, D. G. (2021). Public transit use in the United States in the era of COVID-19: Transit riders’ travel behavior in the COVID-19 impact and recovery period. Transport Policy, 111, 53–62. https://doi.org/10.1016/j.tranpol.2021.07.005 • Plyushteva, A. (2022). Essential workers’ pandemic mobilities and the changing meanings of the commute. The Geographical Journal, geoj.12447. https://doi.org/10.1111/geoj.12447 • San Joaquin Council of Governments. (2013). Interregional multi-modal commute trip planning study. • Shamshiripour, A., Rahimi, E., Shabanpour, R., & Mohammadian, A. (Kouros). (2020). How is COVID-19 reshaping activity-travel behavior? Evidence from a comprehensive survey in Chicago. Transportation Research Interdisciplinary Perspectives, 7, 100216. https://doi.org/10.1016/j.trip.2020.100216 • Smart, T. (2022). The Great Return: Companies Are Calling Their Workers Back to the Office as Covid-19 Fades. U.S.News. https://www.usnews.com/news/economy/articles/2022- 03-04/the-great-return-companies-are-calling-their-workers-back-to-the-office-as-covid-19- fades • Smith, H. (2021). What will be new norm in L.A. traffic? Transit officials and motorists wait to see if patterns altered by Covid offer clues to future of commuting. Los Angeles Times. • Srichan, P., Apidechkul, T., Tamornpark, R., Yeemard, F., Khunthason, S., Kitchanapaiboon, S., ... & Upala, P. (2020). Knowledge, attitudes and preparedness to respond to COVID-19 among the border population of northern Thailand in the early period of the pandemic: a crosssectional study. WHO South-East Asia journal of public health, 9(2), 118-125. • StreetLight Insight (2022). StreetLight All Vehicles Volume Methodology and Validation White Paper. • Sun, Y., Hu, X., & Xie, J. (2021). Spatial inequalities of COVID-19 mortality rate in relation to socioeconomic and environmental factors across England. Science of the total environment, 758, 143595. • Sung, J., & Monschauer, Y. (2020, May 27). Changes in transport behaviour during the Covid-19 crisis. Iea. Retrieved April 20, 2023, from https://www.iea.org/articles/changes-intransport-behaviour-during-the-covid-19-crisis • Thomas, F. M., Charlton, S. G., Lewis, I., & Nandavar, S. (2021). Commuting before and after COVID-19. Transportation Research Interdisciplinary Perspectives, 11, 100423. • Tokey, A. I. (2021). Spatial association of mobility and COVID-19 infection rate in the USA: A county-level study using mobile phone location data. Journal of Transport & Health, 22, 101135. • Truong, D., & Truong, M. D. (2021). Projecting daily travel behavior by distance during the pandemic and the spread of COVID-19 infections – Are we in a closed loop scenario? Transportation Research Interdisciplinary Perspectives, 9, 100283. https://doi.org/10.1016/j.trip.2020.100283 • Tomer , A., & Fishbane, L. (2020). Coronavirus has shown us a world without traffic. Can we sustain it? Brookings. Retrieved April 20, 2023, from https://www.brookings.edu/research/coronavirus-has-shown-us-a-world-without-traffic-canwe-sustain-it/ • Yasenov, V. I. (2020). Who can work from home?. IZA Discussion Papers, No. 13197, Institute of Labor Economics (IZA), Bonn 37 • Yilmazkuday, H. (2020). COVID-19‐19 and unequal social distancing across demographic groups. Regional Science Policy & Practice, 12(6), 1235-1248. 38 Chapter 3. The Relationship Between Remote Work and Workplace Traffic During and After COVID-19 in the Bay Area and Central Valley Region Author: Bonnie S. Wang 39 Abstract The COVID-19 pandemic has significantly altered work and residential patterns, with remote working becoming a widespread norm. This study investigates whether the observed changes in traffic patterns at workplace destinations can be attributed to the increase in workfrom-home (WFH) phenomenon at residential locations. Specifically, I focus on all-day trip volumes to determine whether WFH impacts overall traffic at workplace destinations beyond the typical rush hours. My research focuses on the Bay Area and Central Valley regions, analyzing which industries experienced the most significant changes in traffic volume during COVID-19, the lasting impact of WFH on workplace traffic in the post-pandemic era, and the migration patterns during this period. The study utilizes five datasets: StreetLight for traffic data, LEHD LODES Workplace Area Characteristics (WAC) for job-related data, the American Community Survey (ACS) for demographic information, and USPS Change of Address (COA) and the US Census Current Population Survey (CPS) for migration patterns. The analysis covers 487 zip code tabulation areas (ZCTAs) in the Bay Area and Central Valley regions, employing methods such as destination (workplace) analysis, origin (residential) analysis, origin-destination (O-D) flow analysis, and regression analysis. My findings reveal significant decreases in traffic volume across all regions, with the Bay Area experiencing the largest drop. Increased WFH rates are associated with slower traffic recovery at workplace destinations, while migration patterns indicate a shift towards more remote living arrangements. However, workplace industry characteristics appear to exert a greater influence on traffic volumes than worker demographics or remote working status. These insights hold important implications for urban planning, housing policies, and transportation strategies in the post-pandemic era. Keywords: Work-from-home, traffic recovery, commute, migration, COVID-19 40 1. Introduction The concept of work-from-home (WFH) has become more prevalent, especially due to the COVID-19 pandemic. About 4% of U.S. workers working-from-home in 2006, a share that increased to 6% by 2019 according to the U.S. Census American Community Survey. The spread of COVID-19 in early 2020 and the physical distancing mandates that followed caused a rapid and massive shift to remote working. Although many workers have returned to the office, 39% of U.S. adults still substitute some or all of their typical in-person work for telework (2022 U.S. Census Bureau Household Pulse Survey). The changes in work and employment had an immediate impact on the economy and could lead to permanent shifts that last beyond the pandemic. Working from home, by eliminating or reducing the need to commute to an office, could affect workers' residential choices and increase housing demand in remote locations away from employment centers. This study is among the first to provide an early examination of the impact of working from home after the initial COVID-19 shock, during a period when work and residential location dynamics are still in flux and likely adjusting. While current studies analyzing commute patterns during the pandemic primarily focus on trip productions (i.e., trips originating from residential locations), none have effectively linked these changes to trip attractions at workplace destinations. This research fills the gap by examining how changes in trip volume at work locations are associated with work-from-home (WFH) rates at residential locations. We also examine workplace industry characteristics and their potential for remote work, as well as migration patterns in the residential areas, to determine whether the reduction in trip volumes at workplace destinations can be attributed to the increase in work-from-home rates. Specifically, we analyze all-day trip volumes during and after the COVID-19 pandemic to assess whether WFH affects overall traffic at workplaces beyond the typical rush hours in the combined Greater San Francisco Bay Area and Central Valley megaregion. To conduct this analysis, we utilize four datasets—StreetLight, the American Community Survey (ACS), LEHD Origin-Destination Employment Statistics (LODES), and USPS Change of Address (COA)—at the zip code tabulation area (ZCTA) level to explore how remote work has reshaped workplace dynamics in California’s Bay Area and Central Valley. This study aims to answer three key questions: 1. Where and in which industries did traffic volume change the most during COVID-19? 2. Did the increase in work-from-home still reduce all-day traffic at workplaces as we entered the post-pandemic era? 3. Which areas experienced the highest number of move-ins and move-outs during COVID19? The rest of the paper will provide a comprehensive understanding of the impact of remote work on workplace traffic patterns. Chapter 2 provides a literature review of the existing state of knowledge on remote workers’ attitudes and characteristics and the relationship between remote working, and traffic. Chapter 3 discusses the data sources and methodologies used in the study. Chapter 4 presents a detailed analysis of destination (workplace) characteristics, origin 41 (residential) characteristics, and an origin-destination analysis that weights destination traffic to residences to give insights into the relationship between work-from-home (origin) and workplace (destination) traffic flows. Chapter 4 also contains the main findings, including descriptive information on changes in traffic volumes, work-from-home growth, and migration patterns, and the results of our regression analysis. Finally, Chapter 5 concludes the report with a summary of findings, policy implications, data limitations, and suggestions for future research. 42 2. Literature Review The term "work-from-home," frequently used today, is a form of telecommuting. The concept of telecommuting was first introduced by Nilles in 1973 and has been a topic of interest for transportation researchers since then. Telecommuting describes working outside the traditional workplace during standard work hours (Ory and Mokhtarian, 2006; Bailey and Kurland, 2002). Other terms include telework, remote work, distance work, e-work, flexplace, and electronic cottage (Teo and Lim, 1998). Telecommuting has been seen as a potential solution to alleviate traffic congestion, reduce emissions, improve air quality, and create environmental benefits in urban areas (Hopkins and McKay, 2019; Nguyen, 2021). However, some scholars argue these forecasts might be overly optimistic (Gold, 1991). Before COVID-19, the share of telecommuters in the U.S. remained consistently low. According to the ACS, from 1997 to 2010, the number of people working at least one day a week from home increased by only 2.5 percentage points, from 7% to 9.5%. Zhu et al. (2018) also noted that only about 9% of the working population in the U.S. worked from home more than once a week. As of September 2023, the U.S. Household Pulse Survey indicated that 26% of adults in the U.S. were still substituted some or all of their typical in-person work for telework. 2.1 Remote work potential by demographic and industrial characteristics The shift to remote work has been uneven across employers within the same industry, even when hiring for similar roles. The pandemic significantly increased the prevalence of remote work, with the daily work-from-home rate jumping from 8% in February 2020 to 35% in May 2020 (Brynjolfsson et al., 2020). Much research has focused on the workers who transitioned to remote work during the COVID-19 pandemic (Barrero et al., 2020; Brynjolfsson et al., 2020; DeFilipis et al., 2020; Bick et al., 2020). Surveys by the Bay Area Council (2021) and Drucker (2021) indicate a strong preference among workers for a hybrid work model, allowing them to work from home 2-3 days per week. Tan et al. (2023) surveyed tech workers in the Bay Area between November 2021 and March 2022, revealing a dramatic shift: only 3% of respondents commuted to the office daily, while 66% were fully remote, and 31% followed a hybrid schedule. This contrasts sharply with pre-pandemic behavior when 74% commuted daily, and only 3% worked fully remotely. Remote work potential varies significantly across different industries, with several studies estimating the proportion of jobs that can be done from home based on occupation or industry (Dingel and Neiman 2020; Mongey et al., 2020). The impact of occupation type on the potential to work remotely has been well-documented in recent literature. The ability to work remotely is also significantly influenced by industry, specific job roles, and worker demographics (Brynjolfsson et al., 2020; Wang et al., 2023; Dingel and Neiman, 2020). Studies show that certain groups—such as those who are white, younger, more highly educated, higherincome, or employed in information work—are more likely to transition to remote work and are less likely to experience layoffs or furloughs. Bartik et al. (2020) further confirmed that workers 43 in office and desk-based roles are more likely to have the capability to work remotely, while those in service-related jobs and manual labor face more challenges. For instance, workers in information sectors and management positions are more likely to transition to remote work, whereas those in hands-on or customer-facing roles encounter more limitations. 2.2 Effects of Remote work on traffic volume The potential of remote work, particularly working-from-home (WFH), as a strategy to alleviate urban commuting peaks has been recognized for decades (Loo & Wang, 2018; Nilles et al., 1976). Research generally supports the idea that promoting WFH can reduce urban commuting congestion (Loo & Wang, 2018; Mitomo & Jitsuzumi, 1999). However, much of the literature has primarily focused on changes in individual travel behavior rather than directly examining the broader impact on traffic volume. Kitamura et al. (1990), through a questionnaire survey in California, found that remote workers tend to reduce trips during peak hours. Similarly, Stiles and Smart (2020), using data from the American Time Use Survey spanning 2003 to 2017, explored the relationship between daily work locations and travel patterns in the United States. Their findings indicate that full-day telework significantly decreases daily travel duration and increases the likelihood of avoiding peak hour travel for both work and non-work-related purposes. However, for part-day teleworkers—those who split their workday between home and the workplace—there is no reduction in daily travel time, and peak hour travel avoidance is limited to work-related travel. Additionally, working from alternative locations like cafés, libraries, or vehicles is associated with a higher likelihood of avoiding peak hours. The study also highlights that work location decisions tend to have a more pronounced impact on morning peak periods than on evening ones. The literature also notes that the demand for various activities differs between workers and non-workers, as well as between workdays and non-workdays (Yamamoto and Kitamura 1999). Telework, by enabling work activities to be conducted remotely, disrupts traditional constraints and tradeoffs associated with travel for both work and non-work activities (Dijst 2004). Despite these insights, the effectiveness of WFH in alleviating traffic congestion may have been overstated, partly due to the limited number of empirical studies directly examining traffic conditions. There is a common belief that promoting WFH would automatically lead to significant reductions in urban traffic congestion. However, there are lack of empirical evidence to support this claim. A recent study by Loo and Huang (2022) in Hong Kong, for instance, found that while WFH did lead to a statistically significant reduction in traffic congestion in the Central Business District (CBD) and some urban cores, the overall impact may not be as substantial as intuitively expected. 44 This body of research highlights the complex relationship between remote work and traffic, suggesting that while remote work can contribute to reducing congestion. However, its effectiveness may vary depending on the specific context and the extent to which it is adopted. In addition, there is a lack of post-pandemic research on traffic patterns, which leaves a gap in understanding how the long-term shift to remote work is impacting congestion. Further empirical studies are needed to fully understand and these effects. 45 3. Data and Methodology In this chapter, I outline the methods and data sources used to analyze the impact of remote work on job and housing locations in the Bay Area and Central Valley megaregion. I describe the study area, the five primary datasets (StreetLight, ACS, LODES, USPS COA, and CPS), the observation period, and the methodology. The methodology includes destination and origin analyses, origin-destination (O-D) flow analysis, and regression analysis to understand the relationships between remote work, traffic volumes, migration, and workplace dynamics. 3.1 Methodology 3.1.1 Analytical Framework I use four datasets—StreetLight, the LEHD LODES Workplace Area Characteristics (WAC), the American Community Survey (ACS), and USPS Change of Address (COA) data— to analyze how the increased remote working phenomenon may impact traffic at workplaces for the 487 ZCTA in the study area. In Table 3.1.1, I introduce the key variables used in the analysis, along with their respective units and sources. These variables will be explained in detail in the subsequent results section. My study comprises four major steps: 1. Destination (Workplace) Analysis: I analyze the trip volume change at workplace destination ZCTAs from the pre-COVID period (2019) to the post-COVID period (2021). This highlights areas where trip volumes dropped the most during COVID and recovered the fastest in the post-COVID period. I also examine the industry composition in these job destinations to understand the remote work potential for each area. 2. Origin (Residential) Analysis: This analysis assesses work-from-home rate, demographic characteristics, and migration patterns at residential locations. This involves examining factors such as household income, ethnicity, vehicle ownership, housing value, and other variables that can influence remote work trends. Examining migration patterns helps us understand whether increased flexibility in remote working leads to people moving away from job locations to more distant areas. 3. O-D (Origin-Destination) Analysis: I use StreetLight OD flow data (ZCTA to ZCTA) to link destination ZCTAs and origin ZCTAs together. Using OD flow allows us to determine the relative contribution of each origin ZCTA to the all-day traffic at destination ZCTAs. The more traffic sent by a specific origin ZCTA, the more important it is to the composition of workers at the 46 destination ZCTA. Therefore, I use OD flow as weights to weigh work-from-home rates and other origin demographic characteristics (see section 3.4.2 below for more details). 4. Regression Analysis: I conduct a regression analysis to identify residential and workplace characteristics associated with higher or lower traffic recovery at destinations. This analysis examines the connection between origin and destination ZCTAs to understand how changes in one area could impact the other, particularly in terms of traffic volume. I use the following equation to statistically test whether the increased work-from-home rate at residences is associated with the decrease in all-day traffic at work destinations. The details of the regression analysis will be discussed in Chapter 4.3. 2021 𝑇𝑟𝑖𝑝 𝑅𝑒𝑑𝑢𝑐𝑡𝑖𝑜𝑛 𝑎𝑡 𝑑𝑒𝑠𝑡𝑖𝑛𝑎𝑡𝑖𝑜𝑛𝑗𝑡 = 𝑊𝑗𝑡 = ∑{( 2021 % 𝑤𝑓ℎ𝑖 − 2019 % 𝑤𝑓ℎ𝑖) ∗ 2019 𝑂𝐷𝑖𝑗𝑡} 𝑛 𝑖=1 𝑖 = origin ZCTA 𝑗 = destination ZCTA t = day 𝑊= WFH trip reduction 𝑤𝑓ℎ= share of workers working-from-home 𝑂𝐷= origin-destination flow 47 Table 3.1.1. Summary of variables Category Variables Description Unit Source Traffic Destination traffic volume (2019) The total trips received by each specific destination ZCTA in 2019. ZCTA (total: 491) Daily StreetLight 2019, 2021 Destination traffic volume (2021) The total trips received by each specific destination ZCTA in 2021. % Change in destination volume The percentage change in trips received by each destination ZCTA between 2019 and 2021. 2019 OD trips The number of trips between each ZCTA origin-destination (OD) pair in 2019. ZCTA (Total: 241,081 from OD pairs 491*491) Daily 2021 OD trips The number of trips between each ZCTA origin-destination (OD) pair in 2021. Origin Characteristics % Work-from-home The percentage of workers within each ZCTA who worked from home. ZCTA (total: 491) Annually ACS 2015-2019, 2017-2021 Per capita income The average income per person within each ZCTA. Total population The total number of people living within each ZCTA. % Non-white The percentage of the population within each ZCTA that identifies as non-white. % Commute by public transit The percentage of workers within each ZCTA who commute using public transportation. Median number of rooms The median number of rooms in housing units within each ZCTA. % No vehicles available The percentage of households within each ZCTA that do not have access to a vehicle. Net migration rate The rate of net migration (inflow minus outflow) within each ZCTA USPS COA 2020, 2021 Destination characteristics % Agriculture The percentage of jobs within each ZCTA that are in the agriculture sector. ZCTA (total: 491) Annually LODES WAC 2019 % Construction and manufacturing The percentage of jobs within each ZCTA that are in the construction and manufacturing sectors. % Trade and transport The percentage of jobs within each ZCTA that are in the trade and transport sectors. % Office The percentage of jobs within each ZCTA that are in office-based industries, such as professional services, finance, and information technology. % Education and service The percentage of jobs within each ZCTA that are in education and service industries, including health care, social assistance, and public administration. 48 3.1.2 O-D flow weights In this section, I explain in detail the method used to apply origin-destination (O-D) flow weights in my analysis. The purpose of using O-D flow weights is to accurately attribute the characteristics of origin ZCTAs to their corresponding destination ZCTAs, based on the volume of traffic flowing between them. The basic principle is that the greater the share of people who travel from a residential location (origin) to a work location (destination), the more the workforce at the destination will resemble the population of the origin. Therefore, ZCTAs contributing greater shares of workers have a higher weight in estimating the composition of the workforce at the destination. This approach allows us to better understand the impact of residential demographics and behavior on workplace destinations. The application of the O-D flow weights and analysis results can be found in Chapter 4.3 WFH Trip Reduction at destination. To calculate the O-D flow weights we follow these steps: 1. Define variables • 𝐷𝑗 represents each destination ZCTA (where 𝑗 ranges from 1 to 487, the total number of destination ZCTAs) • 𝑂𝑖 represents each origin ZCTA (where 𝑖 ranges from 1 to 487, the total number of origin ZCTAs). • 𝑡 represents the time period in days (from March 2020 to October 2021). • 𝑉𝐷𝑗 ,𝑡 represents the total volume of traffic received at destination 𝐷𝑗 on day 𝑡. • 𝑉𝑂𝑖𝐷𝑗,𝑡 represents the volume of traffic from origin 𝑂𝑖 to destination 𝐷𝑗 (O-D flow) on day 𝑡.. • 𝑊𝑂𝑖𝐷𝑗,𝑡 represents the weight of traffic from origin 𝑂𝑖 to destination 𝐷𝑗 on day 𝑡. 2. Calculate the total traffic volume at each destination ZCTA for each day 𝑡. 𝑉𝐷𝑗,𝑡 = ∑𝑉𝑂𝑖𝐷𝑗,𝑡 487 𝑖=1 3. Calculate the O-D flow weight for each origin ZCTA to each destination ZCTA for each day 𝑡. 𝑊𝑂𝑖𝐷𝑗 ,𝑡 = 𝑉𝑂𝑖𝐷𝑗,𝑡 𝑉𝐷𝑗,𝑡 4. Applying O-D flow weights to characteristics at origin ZCTA for each day 𝑡. 𝐶𝑂𝑖 represent a characteristic of the population at origin 𝑂𝑖 (e.g., household income, work-from-home rates, etc.). The weighted characteristic 𝐶𝐷𝑗 at each destination 𝐷𝑗 for each day 𝑡 is calculated by summing the weighted contributions from each origin: 𝐶𝐷𝑗,𝑡 = ∑(𝑊𝑂𝑖𝐷𝑗,𝑡 × 487 𝑖=1 𝐶𝑂𝑖 ) 49 I use O-D flows to determine the relative importance of each origin ZCTA to the destination ZCTA. To better explain my methodology, I provide an example of how I traced back from destination ZCTAs to origin ZCTAs using StreetLight daily OD flow data. The example below details the steps taken to calculate O-D flow weights and apply them to my analysis. 1. Data collection: • Example Destination ZCTA: 94043 (Mountain View) • All day traffic volume received: 151,190 (Pre-COVID average daily traffic from 12 am to 12 am using StreetLight data) 2. Calculating O-D Weights: • I collected data on the daily volume of traffic going to ZCTA 94043 from all 487 ZCTAs in the study area. • The traffic volumes from different origin ZCTAs to the destination ZCTA were calculated to determine their respective shares of the total traffic received at destination ZCTA (Table 3.4.2.1). Table 3.4.2.1 shows the traffic volumes and corresponding O-D shares for the top sending ZCTAs to destination ZCTA 94043. Figure 3.4.2.1 visualizes the OD flows between origins and the destination ZCTA. The destination ZCTA 94043 (colored yellow in Figure 3.4.2.1) itself is the largest sending origin, contributing 43,866 trips, which accounts for 29% of the total trips received. This indicates that 29% of trips in ZCTA 94043 originate in the same ZCTA 94043 -- a significant amount of local traffic. The second-largest contributor is ZCTA 94040, sending 11,394 trips to 94043, corresponding to 8% of the total daily trips at ZCTA 94043. ZCTAs 95129, 95110, and 94027 each contribute 1% or less of the trips to ZCTA 94043, making their impact on the activity at ZCTA 94043 relatively minor. The cumulative share shows that the top 10 sending ZCTAs account for a little more than half of the total traffic volume. While this shows that most of the traffic is generated locally, it also highlights that almost half of all traffic flow comes from other ZCTAa, contributing very small shares of the total. This demonstrates the relative importance of high-contributing origin ZCTAs and their substantial influence on destination traffic patterns. For example, an increase in remote work in a highly influential origin ZCTA means fewer workers travelling to the destination ZCTA, potentially leading to a significant decrease in traffic volume. In contrast, the same increase in remote work in a low-contributing origin ZCTA will have a much smaller impact on the traffic at the destination. By using O-D flows to weight demographics and other characteristics at the origin, I capture fluctuations in these ZCTAs and their impact on destination traffic. In Chapter 4.3, "WFH Trip Reduction at Destination," I will demonstrate how remote work at the origin can be measured at the destination. In Chapter 4.4, "Regression Analysis," I will examine the association between destination traffic recovery and origin characteristics, applying the O-D weights to ensure an accurate representation of the origin ZCTAs' impact. 50 Table 3.1.2.1. Example of OD weights between origin and destination ZCTAs origin ZCTA Trips to 94043 Share of trips received at 94043 (O-D weights) Cumulative share 94043 43,866 29% 29% 94040 11,394 8% 37% 94041 6,401 4% 41% 94303 6,320 4% 45% 94089 4,834 3% 48% 94087 4,165 3% 51% 95129 1,319 1% 52% 95110 963 1% 53% 94027 258 0.01% 53.01% … … … … Total trips received at 94043 151,190 100% 100% Figure 3.1.2.1. Example of daily OD flows to ZCTA 94043 51 3.2 Study Area This study focuses on the San Francisco Bay Area metropolitan area, the Central Valley region (which extends from the Sacramento metropolitan area south to the northern San Joaquin Valley metropolitan areas of Stockton, Modesto, and Merced), and the adjacent counties. The Bay Area is the home to some of the country’s least affordable housing markets, highest incomes, and highest concentration of high-tech jobs. The Central Valley, separated from the Bay Area by a mountain range and river delta, faces higher unemployment and has a large agricultural and manufacturing base, with lower median incomes and housing costs. For this study, we analyze the remote working pattern before and after COVID-19 in the core Bay Area counties (Alameda, Contra Costa, San Francisco, San Mateo, and Santa Clara County) and nearby counties of the Central Valley (El Dorado, Merced, Placer, Sacramento, San Joaquin, Solano, Stanislaus, and Yolo County). I also included the populated area in the adjacent counties (Amador, Calaveras, Fresno, Madera, Marin, Monterey, Napa, Nevada, San Benito, Santa Cruz, Sonoma, Sutter, Tuolumne, and Yuba) as comparisons for the two contrasting regions. Figure 1 shows the study area counties and the ZCTA’s within those counties which I use for much of my analysis. 52 Figure 3.2.1. Study area counties 3.3 Data 1. StreetLight The traffic data I use for analyzing origin-destination flows before, during, and after COVID19 is from StreetLight Data Inc. The platform provides information about travelers’ origin and destination, travel distance, travel purpose, etc. StreetLight derives flow data from location-based services, apps that users enable and allow to share location data with vendors. The StreetLight database captures daily traffic at the ZCTA level on a daily basis and covers the general population based on a variable sampling rate that averages 8.5%. StreetLight uses 53 an algorithm to estimate daily trip volumes based on these sample trips. The trip volume estimates have been rigorously validated against real traffic volume data collected by state agencies. The ground-truth data primarily come from over 6,000 permanent traffic counter locations across the U.S., covering the years 2019 through 2022. This data coverage begins before COVID-19, I use the data from June 11, 2019, to October 25, 2019, and extends past the period of most severe COVID-19 shelter-in-place and movement restrictions. The latest data I used in this study covers the time between June 15, 2021, to October 29, 2021 to match the reference period before the pandemic. StreetLight data does not differentiate between trip purposes beyond providing an estimated share of trips categorized as home-based work, home-based other, or non-home-based trips. I use all trips to calculate traffic flow because StreetLight uses its own algorithm to estimate trip purposes. This algorithm is a "black box," making it difficult to understand and reproduce the process. StreetlLight also includes all modes in its estimate of traffic. Therefore, I am unable to differentiate between people who travel by car, transit, or active transportation. The data access was limited to 500 geographic units. Given this limitation, I will analyze traffic flows at the census zip code tabulation area (ZCTA) level to maximize geographic coverage. ZCTAs are the U.S. Census Bureau’s approximation of US Postal Service (USPS) zip codes. I use ZCTAs rather than zip codes because ZCTAs more easily connect to census and American Community Survey (ACS) data. I reduced the number of ZCTAs within the study area counties (which exceeded the 500 limit) by only including ZCTAs that overlapped with an urbanized area or had a population greater than 3,000 based on the American Community Survey 5-year estimate from 2015-2019 (see Figure 3.1.1 for a map of included ZCTAs). 2. American Community Survey (ACS) Demographic characteristics data are drawn from the U.S. Census Bureau's American Community Survey (ACS) 5-year estimates available at the Zip Code Tabulation Area (ZCTA) level from 2015-2019 and 2017-2021. Key demographic variables include age, sex, race, ethnicity, poverty status, and industry occupation. ZCTAs are the U.S. Census equivalent to the U.S. Postal Service’s (USPS’s) 5-digit zip codes (U.S. Census Bureau, 2018). In populated areas, ZCTAs have a high degree of overlap with zip codes (Hurvitz, n.d.; Langer, 2016) and have been used in geographic analyses (Grubesic and Matisziw, 2006). ZCTA to zip code crosswalks are available for public download (e.g., UDS Mapper, 2018). In addition to the limits on the number of geographic units for which I could pull data from StreetLight, ZCTA is the only unit for which I have small-scale migration data (see below). 54 3. LEHD Origin Destination Employment Statistics (LODES) I use the pre-COVID period, 2019 LODES WAC (Workplace Area Characteristic) data, to capture destination characteristics. The WAC data provides detailed job-related characteristics, including total jobs, worker age, earnings, job sector, ethnicity, and education level. All data is at the census block level, which I aggregate to the ZCTA level. 4. USPS Change of Address (USPS COA) USPS change of address data is used to serve as a proxy for migration patterns. The United States Postal Service has provided publicly accessible change-of-address (COA) data by month for the last four years (2018-2022) at the zip code level. For privacy protection purposes, the data is only reported when the COA request volume is greater than 10. I calculate total net migration for each zip code by subtracting the number of COA requests originating from the specific zip code (outgoing) from the number of COA requests to that same zip code (incoming). The COA data includes three categories: family, individual, and business. For my analysis, I used only family and individual move types, excluding business requests. I converted family moves into individual counts by multiplying the number of family COA requests by 2.5, based on the average household size from the Census Bureau. I also convert zip code level COA data to ZCTA level using the zip code to ZCTA crosswalk (e.g., UDS Mapper, 2018) for consistency with data from other sources. 5. US Census Bureau Current Population Survey (CPS) The CPS provides comprehensive data on population change in US counties, with statistics reflecting yearly domestic and international migration. I utilize two CPS datasets, from 2010–2019 and 2020–2023, publicly available from the Census Bureau website. Counties are linked to Metropolitan Statistical Areas (MSAs) through the Office of Management and Budget’s (OMB) core based statistical area delineation files, also available on the Census Bureau website. 3.4 Observation Period For this study, I used StreetLight data from 2019 and 2021, collected at the ZCTA (Zip Code Tabulation Area) level on a daily basis. The data were analyzed for all-day periods (12 am to 12 am) to capture the full daily traffic volume. The observation period consists of 20 weeks each for the pre-COVID and post-COVID time frames (see Table 3.3.1 for details). I measure the 20-week post-Covid period from the week of June 14, 2021 up through the week of October 24, 2021. June 15, 2021, marks the date when California fully reopened its economy and moved beyond the Blueprint for a Safer Economy, making it an ideal starting point for the post-COVID era. On the end timing, we are limited by data: the furthest date available for download with my StreetLight license was October 29, 2021. Additionally, to ensure a consistent comparison between the pre-COVID and post-COVID periods, we selected observation periods 55 from 2021 and then identified the corresponding weeks in 2019. Therefore, the post-COVID period extends from June 15, 2021, to October 29, 2021, while the pre-COVID period ranges from June 11, 2019, to October 25, 2019. Table 3.3.1 shows the week pairs for the post-COVID and pre-COVID periods, with 20 weeks each. The weeks were paired based on the calendar week of the year. For example, the week of June 14-18, 2021 (Week 24 of 2021), corresponds to the week of June 10-14, 2019 (Week 24 of 2019). This pairing method aligns the same weeks across different years, facilitating a direct comparison of traffic patterns before and after the COVID-19 pandemic while minimizing seasonal fluctuations. Table 3.4.1. Observation week pairs in 2019 and 2021 Week of the year Post-COVID (2021) Pre-COVID (2019) Monday Friday Monday Friday Week 24 6/14 6/18 6/10 6/14 Week 25 6/21 6/25 6/17 6/21 Week 26 6/28 7/2 6/24 6/28 Week 27 7/5 7/9 7/1 7/5 Week 28 7/12 7/16 7/8 7/12 Week 29 7/19 7/23 7/15 7/19 Week 30 7/26 7/30 7/22 7/26 Week 31 8/2 8/6 7/29 8/2 Week 32 8/9 8/13 8/5 8/9 Week 33 8/16 8/20 8/12 8/16 Week 34 8/23 8/27 8/19 8/23 Week 35 8/30 9/3 8/26 8/30 Week 36 9/6 9/10 9/2 9/6 Week 37 9/13 9/17 9/9 9/13 Week 38 9/20 9/24 9/16 9/20 Week 39 9/27 10/1 9/23 9/27 Week 40 10/4 10/8 9/30 10/4 Week 41 10/11 10/15 10/7 10/11 Week 42 10/18 10/22 10/14 10/18 Week 43 10/25 10/29 10/21 10/25 56 4. Results This chapter presents the findings from my analysis, focusing on destination and origin characteristics, defining the concept of WFH trip reduction (trip reductions at destinations which can be attributed, in a deterministic way, to work-from-home increases at origins), and a regression analysis to bring all the elements together. In my analysis, we separate between "destination" for workplaces and "origin" for residential locations. This distinction helps us study employment and living conditions separately, providing a better understanding of remote working dynamics. I then use Origin-Destination (O-D) flow weights to link destination and origin characteristics. Table 4.1 provides a summary of descriptive statistics for key variables in the study, capturing traffic volumes, population changes, work-from-home rates, migration patterns, and job characteristics across different regions. Each variable is presented with the number of observations (Obs.), mean, standard deviation (Std. dev.), percentile values (25%, 50%, 75%), and data source. The unit of analysis for each variable is ZCTA. For numeric variables such as destination traffic volume and total population, the mean represents the average value across all ZCTAs. For percentage change variables, such as Δ Destination traffic volume and Δ % WFH, I calculate the percent change for each ZCTA individually and then report the average of these individual changes. I will discuss each variable in detail in the following subsections. Table 4.1. Summary of Descriptive Statistics Variable Obs. Mean Std. dev. 25% 50% 75% Data Source Traffic Destination traffic volume (2019) 487 81,293 61,518 28,953 72,095 119,972 StreetLight 2019, 2021 Destination traffic volume (2021) 487 64,766 49,838 23,409 56,682 96,558 Δ Destination traffic volume 487 -18.3% 12.0% -24.0% -18.0% -13.0% Origin Total population (2019) 491 27,523 19,811 10,495 26,145 39,741 ACS 2015-2019, 2017-2021 Total population (2021) 488 27,992 19,988 10,717 26,960 40,925 Δ Total population 487 2.4% 18.4% -2.2% 0.9% 4.2% % WFH (2019) 486 6.5% 4.1% 3.9% 5.6% 8.6% % WFH (2021) 486 13.2% 8.3% 7.0% 12.1% 18.1% Δ % WFH 486 6.6% 6.8% 2.3% 5.5% 9.8% Cumulative Migration (2020 to 2021): Total from ZCTA 466 6,241 4,674 2,415 5,598 8,840 USPS COA 2020, 2021 Cumulative Migration (2020 to 2021): Total to ZCTA 466 5,517 3,941 2,279 5,205 7,781 Net Cumulative migration (2020 to 2021) 466 -724 1,438 -1,286 -341 12 * Numeric variables: the mean represents the average value across all ZCTAs. * Δ variables: the mean represents the average of individual (ZCTA) percent changes. 57 Table 4.1. Summary of Descriptive Statistics (continued) Variable Obs. Mean Std. dev. 25% 50% 75% Data Source Destination Total job (2019) 487 12,608 15,231 3,121 8,655 16,948 LODES WAC 2019 % Agriculture 487 5.4% 13.5% 0.01% 0.2% 3.2% % Construction 487 14.2% 12.6% 5.0% 10.7% 20.2% % Trade and Transport 487 15.2% 9.4% 8.8% 13.5% 20.2% % Office 487 21.2% 15.5% 10.7% 18.0% 27.2% % Education and Services 487 43.7% 19.7% 29.9% 43.1% 56.1% * Numeric variables: the mean represents the average value across all ZCTAs. * % variables: the mean represents the average of individual (ZCTA) percent. 4.1 Destination Characteristics In our analysis, "destination" serves as a proxy for workplace areas. Therefore, when I mention “destination,” I am primarily referring to characteristics and trends associated with locations where people go to work, although of course not all trips in those locations are work trips. By analyzing destination characteristics such as traffic recovery and industry composition, I can gain insights into the trip dynamics at places that are job centers and hence (but not exclusively) work locations. 4.1.1 Traffic Recovery at Destination Table 4.1.1.1 presents the average traffic volumes across different regions for 2019 and 2021, along with the percentage change between these two years, from pre-COVID to postCOVID. The results highlight a general decline in traffic volumes across all regions, with the Bay Area experiencing the largest drop of 20.5%, followed by the Central Valley at 16.4%, and the Comparison counties at 17.6%. Table 4.1.1.1. Traffic volume by region – average per ZCTA Variable Bay Area Counties Central Valley Counties Comparison Counties Destination traffic volume (2019) 97,446 81,519 59,962 Destination traffic volume (2021) 75,810 66,496 48,486 Δ Destination traffic volume (2019 to 2021) -20.5% -16.4% -17.6% Total ZCTAs 188 155 144 * Numeric variables: the mean represents the average value across all ZCTAs. * Δ variables: the mean represents the average of individual (ZCTA) percent changes. 58 Figure 4.1.1.1 shows the maps of traffic volume in 2019 (left) and traffic recovery in 2021 (right) at destination ZCTAs, based on Streetlight traffic volume data for the 2019 and 2021 time periods. The traffic recovery rate is based on the ratio of trip flows into the ZCTA, 2021 time period, compared to the 2019 baseline. This data reveals significant regional differences in post-COVID traffic recovery. In Figure 4.1.1.1, the left map shows higher baseline traffic volumes in the city centers of each county (darker area). The right map color codes indicate recovery levels: blue areas represent ZCTAs with less than 70% recovery, green areas show recovery between 71-80%, yellow areas show recovery between 81-90%, and orange areas show ZCTAs with recovery exceeding 90%. The results show that areas like South Bay, Oakland, and parts of San Francisco have seen a slower recovery, with 2021 traffic volumes remaining below 70% of the 2019 baseline. In contrast, more rural regions in the Central Valley (including San Joaquin, Sacramento, El Dorado, and others) have experienced a quicker recovery, with the 2021 traffic exceeding 90% of the 2019 levels. 59 Figure 4.1.1.1. Left: 2019 Traffic volume at destination; Right: Traffic recovery at destination 60 4.1.2 Industry composition at Destination The analysis of industry composition within destination ZCTAs is crucial for understanding the economic dynamics across Bay Area and Central Valley Counties. I use data from 2019 LODES WAC (Workplace Area Characteristic) to capture the employment pattern in the pre-COVID period. The results reveal distinct regional differences in employment sectors. Table 4.1.2.1 shows the total number of jobs and the average share of employment across different industry sectors by region. In 2019, Bay Area Counties had the highest average number of jobs (17,827) per ZCTA. This is significantly more than the 10,512 jobs reported in Central Valley Counties and the 8,049 jobs in Comparison Counties, highlighting the Bay Area's larger economic activity compared to the other regions. I found that agriculture is a minor sector in the Bay Area, with only 0.3% of jobs, but is much more significant in the Central Valley and the Comparison counties, where it accounts for 6.8% and 10.5% of jobs respectively. Construction jobs are relatively more evenly distributed, with 12.5% of Bay Area jobs being construction, 16.3% in the Central Valley, and 14.3% in other counties. Trade and Transport jobs are most prevalent in the Central Valley, accounting for 18.2% of employment. In contrast, this sector only represents approximately 14% of jobs in both the Bay Area and the Comparison counties. Office jobs are more common in the Bay Area, comprising 26.1% of employment there, compared to 18.6% in the Central Valley and 17.6% in the Comparison counties. The Education and Services sector are most prevalent in the Bay Area, comprising 46.9% of jobs, compared to 39.9% in the Central Valley and 43.6% in other counties. These statistics illustrate the different industrial characteristics of each region, revealing how economic activities are tailored to the urban or rural contexts of the respective areas. Table 4.1.2.1. Industry composition by region – average per ZCTA Variable Bay Area Counties Central Valley Counties Comparison Counties Total job (2019) 17,827 10,512 8,049 % Agriculture 0.3% 6.8% 10.5% % Construction 12.5% 16.3% 14.3% % Trade and Transport 13.9% 18.2% 13.8% % Office 26.1% 18.6% 17.6% % Education and Services 46.9% 39.9% 43.6% Total ZCTAs 188 155 144 * Numeric variables: the mean represents the average value across all ZCTAs. * % variables: the mean represents the average of individual (ZCTA) percent. The ability to work remotely depends largely on the type of work people do, which is, in turn, highly correlated with workplace characteristics. Understanding this association is crucial for addressing the intertwined relationship between remote working and workplace traffic. The effect of occupation type on the potential to work remotely has been documented in prior literature (Dingel and Neiman, 2020). Previous research (e.g., Brynjolfsson et al., 2020; Wang et 61 al., 2023) confirms that workers in office and desk-based roles are associated with a higher likelihood of being able to work remotely, while those in service-related jobs and manual labor are more likely to have a negative association with remote work. Figure 4.1.1.1 illustrates the potential for remote working based on industry types. The left map shows the share of office jobs (high remote working potential) in each ZCTA, while the right map shows the share of agriculture, construction, trade, and transport jobs (low remote working potential). In the left map, the dark blue areas indicate ZCTAs where office jobs constitute more than 27.3% of employment. These office jobs are concentrated in the city centers of each county and are more clustered in the Bay Area than in the Central Valley and the Comparison counties. The right map shows the share of agriculture, construction, trade, and transport jobs per ZCTA. The dark red areas represent ZCTAs where these jobs make up more than 47.2% of employment. Unlike office jobs, agriculture, construction, trade, and transport jobs are more concentrated in rural areas of the Central Valley and Comparison counties, with very few in the Bay Area. The spatial distribution of jobs indicates that the Bay Area has a higher potential for remote working due to the concentration of office jobs, while the Central Valley and the Comparison counties have a higher proportion of jobs that requires manual labor and are harder to perform remotely, such as those in agriculture, construction, trade, and transport. 62 Figure 4.1.1.1. Left: Share of office jobs; Right: Share of agriculture, construction, and trade and transport jobs 63 4.2 Origin Characteristics As mentioned in the previous section, the term "origin" is used to describe the residential locations from which individuals travel. By analyzing origin data, I can gain insights into the demographic characteristics of these residential areas. This information also helps us learn more about the socio-economic background of the residents. In this section, I will discuss trends in remote working (work-from-home), general demographics, and migration patterns. 4.2.1 Work-from-home Growth at Origin I collect work-from-home data from ACS's survey on “Journey to Work”, which asks respondents to select their usual mode of transportation to work from the previous week. Answer options include car, truck, van, bus, subway, commuter rail, light rail, ferryboat, taxi, motorcycle, bicycle, walking, and working from home. Table 4.2.1.1 presents the results from ACS. The average share of workers working from home (WFH) in 2019 (pre-COVID) and 2021 (post-COVID) across the study regions shows a substantial increase in remote work. In the Bay Area, WFH rates increase the most, by 10 percentage points, from 6.6% in 2019 to 16.6% in 2021. In Central Valley Counties, the increase was from 5.8% to 9.8%. In Comparison Counties, rates went up from 7.2% to 12.3%. These changes show how the Bay Area's technological and professional service industries, which adapted quickly to remote work, are associated with the increase in WFH rates. Table 4.2.1.1. Share of remote workers by region – average per ZCTA Variable Bay Area Counties Central Valley Counties Comparison Counties % WFH (2019) 6.6% 5.8% 7.2% % WFH (2021) 16.6% 9.8% 12.3% Δ % WFH (2019 to 2021) 10.0% 4.0% 5.0% Total ZCTAs 188 155 144 * % variables: the mean represents the average of individual (ZCTA) percent. * Δ variables: the mean represents the average of individual (ZCTA) percent changes. Figure 4.2.1.1 presents the share of work-from-home (WFH) rates at the ZCTA level for the study region, comparing pre-COVID in 2019 on the left with post-COVID in 2021 on the right. I grouped the 2019 WFH rates by quartile and applied the same categorization to 2021. The dark blue color indicates a WFH rate of more than 8.8%, which shows the most growth in the urban areas of each county. 64 Figure 4.2.1.1. Share of workers working from home in 2019 (left) and 2021 (right) 65 Figure 4.2.1.2 shows the Work-from-home growth from 2019 to 2021. Compared to preCOVID, the areas with WFH growth more than 10 percentage points clustered mostly in the South Bay Area, including the cities of Mountain View, Santa Clara, Sunnyvale, Fremont, etc., with pockets in relatively outlying locations that include suburbs/exurbs of Sacramento. This shift highlights the tech industry's role in advancing remote work. * pp.=percentage points Figure 4.2.1.2. Work-from-home growth at origin (ACS) 66 4.2.2 Demographic Characteristics at Origin Table 4.2.2.1 presents the general demographic characteristics of the study regions. The average total population per ZCTA in 2019 was approximately 19,000 in each region. Median ages varied slightly, with Central Valley being the youngest (37.5 years), followed by Bay Area (39.7 years), and the Comparison counties being the oldest (41.1 years). Bay Area ZCTAs also had the highest average percentage of non-white residents at 58.4%, compared to 52.8% in Central Valley and 46.2% in the Comparison counties. Bay Area ZCTAs reported the highest average of median income at $120,901, which is significantly more than Central Valley ($71,437) and the Comparison counties ($77,673). The poverty rates were lower in the Bay Area (an average of 9.0% per ZCTA) compared to 14.1% and 14.2% in Central Valley and Comparison counties. The average percentage of college graduates was also higher in the Bay Area at 53.5% per ZCTA, versus 26.2% in Central Valley and 31.6% in Comparison counties. This reflects the Bay Area's education and economic advantages, likely due to a concentration of high-paying jobs in tech and professional services. Housing costs also reflect this trend, with Bay Area's costs nearly double those of other regions: a median rent of $2,121 and median home value of $995,660, compared to $1,334 and $362,846 in Central Valley, and $1,465 and $543,242 in the Comparison counties. The median number of rooms per household remains consistent across all regions at about 5.5. Commute mode also varied, with an average of 13.5% of workers in Bay Area ZCTAs using public transit, compared to only 1.7% and 2.3% in Central Valley and Comparison counties respectively. The share of households without a vehicle was highest in the Bay Area at an average of 9.7% per ZCTA, likely due to more accessible public transit options, compared to 5.8% in Central Valley and 5.3% in Comparison counties. Table 4.2.2.1. 2019 Demographic characteristics by region – average per ZCTA Variable Bay Area Counties Central Valley Counties Comparison Counties Total population 19,532 19,232 18,427 Median household income 120,901 71,437 77,673 Median age 39.7 37.5 41.1 Median rent 2,121 1,334 1,465 Median home value 995,660 362,846 543,242 Median number of rooms 5.2 5.5 5.2 % Non-white 58.4% 52.8% 46.2% % Commute by public transit 13.5% 1.7% 2.3% % Poverty 9.0% 14.1% 14.2% % College graduated 53.5% 26.2% 31.6% % No vehicle 9.7% 5.8% 5.3% Total ZCTAs 188 155 144 * Numeric variables: the mean represents the average value across all ZCTAs. * % variables: the mean represents the average of individual (ZCTA) percent. 67 4.2.3 Migration at Origin Cumulative migration refers to the total number of individuals moving into and out of specific geographic areas over a set period. In the study, I examine cumulative migration at the ZCTA level to understand how residential patterns have shifted during the COVID-19 pandemic, from March 2020 to October 2021. I use USPS change of address (COA) data to track migration patterns at the ZCTA level. This dataset records the total number of COA requests submitted for each 5-digit ZIP Code monthly. To ensure consistency with the other data, I convert these ZIP codes into ZCTAs. The COA data have three categories on change of address requests: family, individual, and business. For my analysis, I only used Family and Individual move types, excluding Business COA requests as they do not represent residential moves. I converted family moves into individual counts by multiplying the number of Family COA requests by 2.5, using the average household size from the Census Bureau. This factor provides a more accurate representation of the total number of individuals moving. Individual COA requests were counted as one individual each. By summing the adjusted family requests and the individual requests, I obtained the total number of individuals who submitted COA requests each month over the past four years. The net migration rate during COVID-19 (from March 2020 to October 2021) is then calculated using the following equation: 𝐶𝑢𝑚𝑢𝑙𝑎𝑡𝑖𝑣𝑒 𝑛𝑒𝑡 𝑚𝑖𝑔𝑟𝑎𝑡𝑖𝑜𝑛 𝑟𝑎𝑡𝑒 = 𝑡𝑜𝑡𝑎𝑙 𝑚𝑜𝑣𝑒𝑠 𝑡𝑜 𝑍𝐶𝑇𝐴 – 𝑡𝑜𝑡𝑎𝑙 𝑚𝑜𝑣𝑒𝑠 𝑓𝑟𝑜𝑚 𝑍𝐶𝑇𝐴 𝑡𝑜𝑡𝑎𝑙 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑎𝑡 𝑍𝐶𝑇𝐴 𝑖𝑛 2020 Calculating net migration helps us compare the inflow of new residents to the outflow of existing residents relative to the total population within each ZCTA. A positive net migration rate indicates more people moving into the area than leaving, while a negative rate suggests a higher number of leaving than entering. Table 4.2.3.1 provides an overview of the cumulative migration patterns across different regions from March 2020 to October 2021. It includes both the total net migration (total persons) and the net migration rate (standardized by ZCTA population). This standardization allows comparisons across ZCTAs with varying population sizes. The total ZCTA population for the three regions is quite similar, with each having around 19,000 persons. The Bay Area Counties had the highest number of both move-outs (8,248) and move-ins (6,774), resulting in the largest net cumulative migration loss of 1,473 people. The Central Valley and Comparison counties experienced smaller inflows and outflows, resulting in a net loss of around 200 people each. In the Bay Area Counties, the standardized cumulative migration rate from 2020 to 2021 shows a net loss of 4 people per 100 residents. The result indicates a higher outflow from the Bay Area, potentially driven by factors such as high living costs and the more remote work, 68 which allow residents to relocate to more affordable regions. In contrast, the Central Valley and Comparison counties experienced smaller net losses, with a net loss of 1 person per 100 residents. Table 4.2.3.1. Cumulative migration by region – average per ZCTA Variable Bay Area Counties Central Valley Counties Comparison counties Number of Persons Cumulative outflow (2020 to 2021): Total from ZCTA 8,248 5,660 4,075 Cumulative inflow (2020 to 2021): Total to ZCTA 6,774 5,414 3,861 Cumulative Net migration (2020 to 2021) -1,473 -246 -214 Standardized by ZCTA population Cumulative outflow rate (2020 to 2021): Total from ZCTA 0.25 0.20 0.19 Cumulative inflow rate (2020 to 2021): Total to ZCTA 0.21 0.19 0.18 Cumulative Net migration rate (2020 to 2021) -0.04 -0.01 -0.01 Total ZCTAs 185 150 131 Figure 4.2.3.1 shows a map of the cumulative net migration rate at origin ZCTAs. The yellow color indicates a net migration loss of more than 4 people per 100. The orange color indicates a net migration rate between -3.9 and -1.7 people per 100. The brown color represents a net migration gain between -1.6 and 0.1 people per 100. The dark brown color indicates a net migration gain of more than 0.2 people per 100. I found that downtown areas and city centers in San Francisco, the South Bay area, and Sacramento experienced negative cumulative net migration, indicating more people moving out than moving in. This trend suggests that urban cores, which are traditionally high-density and often more expensive, are losing residents. Contributing factors could be the high cost of living, the shift to remote work, and the desire for more space and affordability. In contrast, the suburbs and exurbs of these counties saw a positive net migration, with more people moving in than exiting. These areas are likely attracting residents due to their relatively lower housing costs, larger living space, and better living environments. This shift reflects a trend of people prioritizing quality of life, especially with the increased flexibility of remote work arrangements in the post-pandemic era. 69 Figure 4.2.3.1. Cumulative net migration rate at origin 70 4.3 Work-From-Home Trip Reduction at Destination To establish a connection between the observed traffic recovery at destination ZCTAs (workplaces) and the increasing trend of remote work at trip origins (residences), I use OriginDestination (O-D) flows to weight the work-from-home characteristics back to origin ZCTAs. This methodology is detailed in Chapter 3.4.2 "O-D Flow Weights." My underlying assumption is that, in the absence of COVID-19, the O-D pattern in 2021 would be similar to 2019. Thus, any reduction in trips to the destination can be attributed to the increase in remote work at the origin. I use the following equation to estimate this phenomenon: 2021 𝑇𝑟𝑖𝑝 𝑅𝑒𝑑𝑢𝑐𝑡𝑖𝑜𝑛 𝑎𝑡 𝑑𝑒𝑠𝑡𝑖𝑛𝑎𝑡𝑖𝑜𝑛𝑗𝑡 = 𝑊𝑗𝑡 = ∑{( 2021 % 𝑤𝑓ℎ𝑖 − 2019 % 𝑤𝑓ℎ𝑖) ∗ 2019 𝑂𝐷𝑖𝑗𝑡} 𝑛 𝑖=1 𝑖 = origin ZCTA 𝑗 = destination ZCTA t = day 𝑊= WFH trip reduction 𝑤𝑓ℎ= share of workers working-from-home 𝑂𝐷= origin-destination flow Work-from-home (WFH) trip reduction is a construct—the estimated number of trips to the destination that would be reduced if changes in work-from-home rates directly result in changes in traffic flows. I use O-D flows to weight the change of workers working-from-home at origin ZCTAs. O-D flow weights provide the relative importance of each origin ZCTA to the destination ZCTA. By applying the O-D weights, I can accurately attribute changes in traffic volumes at destination ZCTAs to the increase in remote work at origin ZCTAs. For example, if origin ZCTA A contributes 30% of the total trips received at destination ZCTA C, and origin ZCTA B contributes only 1%, then an increase in remote workers would have a significantly different impact on the traffic volume at destination ZCTA C from these two origins. Specifically, under the assumption outlined above, a rise in remote workers in ZCTA A could reduce up to 30% of the traffic volume at destination ZCTA C, while a similar increase in remote workers in ZCTA B would only reduce up to 1% of the traffic. Table 4.3.1 presents the descriptive statistics for WFH trip reduction along with related information on destination traffic and work-from-home (WFH) rates from analyses in previous sections. Although destination traffic and origin WFH rates cannot be directly linked, I use WFH trip reduction (O-D weighted WFH growth) to establish this connection. The volume of WFH trip reduction, which represents the estimated number of trips that no longer occur due to increased remote working, was highest in the Bay Area at 10,183 per ZCTA, followed by 4,324 in the Central Valley Counties, and 2,873 in the Comparison Counties. When examining the share of WFH trip reduction as a percentage of the 2019 baseline destination traffic volume, the Bay Area Counties experienced the most significant impact with 13.6% of trips reduced (which is the average for the Bay Area ZCTAs). This is in contrast to the 71 Central Valley and Comparison counties, which saw ZCTA averages of 6.7% and 9.7% of their 2019 traffic volume reduction, respectively. The WFH trip reduction volumes align with the observed increases in the work-from-home (WFH) rates and the decreases in destination traffic volumes. The variable, "Share of WFH trip reduction to Traffic Drop," represents the proportion of WFH trip reduction relative to the overall reduction in traffic volume. It indicates how much of the traffic decline can be attributed to the increase in remote work, with Bay Area Counties showing the highest share (64.9% average for the Bay Area ZCTAs), suggesting that a significant portion of their traffic drop is due to increased work-from-home phenomenon. Central Valley Counties and Comparison Counties show lower shares, at 39.4% and 31.5% respectively, indicating a less impact of remote work on their traffic volumes. Table 4.3.1. Descriptive statistics for WFH trip reduction – average per ZCTA Category Variable Bay Area Counties Central Valley Counties Comparison Counties Destination Traffic Destination traffic volume (2019) 97,446 81,519 59,962 Destination traffic volume (2021) 75,810 66,496 48,486 Δ Destination traffic volume (2019 to 2021) -20.5% -16.4% -17.6% Origin WFH Rate % WFH (2019) 6.6% 5.8% 7.2% % WFH (2021) 16.6% 9.8% 12.3% Δ % WFH (2019 to 2021) 10.0% 4.0% 5.0% WFH trip reduction WFH trip reduction volume 10,183 4,324 2,873 Share of WFH trip reduction: WFH trip reduction/ 2019 destination traffic volume 13.6% 6.7% 9.7% Share of WFH trip reduction to traffic drop: WFH trip reduction / (2021 traffic volume – 2019 traffic volume 64.9% 39.4% 31.5% Total ZCTAs 188 155 144 4.4 Regression Analysis After weighting all trip origin demographic characteristics to the destination using OriginDestination (OD) flows (per the method in Chapter 3.4.2 “O-D Flow Weights”), I estimate the correlation between changes in traffic volume at the destination and the characteristics at their respective origin ZCTAs. The hypotheses I am testing in my regression analysis are: 1. Higher WFH rate at origin is associated with less traffic recovery (i.e., lower inbound traffic flows) at destination. 2. Higher in-migration at origin is associated with higher traffic recovery at the destination. 72 3. Traffic recovery at destination is associated with industry composition (e.g. tech vs. service). Traffic recovery is spatially uneven, related in part to the industry types at destinations. To test these hypotheses, I run an Ordinary Least Squares (OLS) regression using the equation: ∆ % 𝑉𝑜𝑙𝑗𝑡 = 𝛽0 + 𝛽1𝑊𝑗𝑡 + 𝛽2𝐶𝑖𝑗𝑡 + 𝛽3𝑂𝑗𝑡 + 𝛽4𝑀𝑖𝑗𝑡 + ε𝑗𝑡 (Week-fixed effects and robust standard errors are applied) % 𝑉𝑜𝑙𝑗𝑡 = % change in destination volume (Dependent variable) Observation: ZCTA, days (indexed by “t”) 𝑖 = origin ZCTA 𝑗 = destination ZCTA Pre-COVID sample period: 06/11/2019 – 10/25/2019 Post-COVID sample period: 06/15/2021 – 10/29/2021 𝑊 = WFH trip reduction at destination j (OD weighted wfh change) 𝐶 = demographic characteristics weighted by origin-destination flow i to j 𝑂 = share of industry composition at destination j 𝑀 = change in cumulative net migration (from USPS change of address data) = (total moves into ZCTA minus total moves out of ZCTA) / total population Table 4.5.1 shows the descriptive statistics for all variables. The dependent variable, percentage change in destination volume, shows an average decrease of 18.4% indicating a decreasing trend in traffic flow changes across destination ZCTAs. For origin characteristics, the weighted change in work-from-home (WFH) rates from 2019 to 2021 averages 12.8%, with a substantial range from 4.3% to 20.9% at the 25th and 75th percentiles, primarily due to the diverse range of industries across the Bay Area and Central Valle. The cumulative net migration rate, weighted by 2021 OD volumes, averages -2.3%, reflecting a net outflow in many areas. I weight the cumulative net migration (from March 2020 to October 2021) with 2021 OD flow rather than 2019 OD flow to more accurately capture the impact of recent migration patterns on current traffic volumes. By using 2021 OD flow, I ensure that the weighting accounts for any changes in origin-destination relationships that may have occurred due to the pandemic. In contrast, other demographic variables are weighted using 2019 OD flow to establish a baseline comparison to pre-COVID conditions. Per capita income and total population, both weighted by 2019 OD volumes, have means of 44,064 and 35,015, respectively. The share of non-white population, commute by public transit, households with no vehicles available and median numbers of rooms serve as control variables to provide additional demographic context. The destination industry composition variables at the destination level show that service and education sectors dominate, with an average of 43.8%, followed by office sectors at 21.3%, 73 trade and transport at 15.3%, and construction and manufacturing at 14.3%. Medians are similar to means in every case except for construction and manufacturing. This suggests that outside this latter employment category, occupation shares are distributed somewhat evenly about the mean. Table 4.4.1. Descriptive statistics for all variables used in the regression Variable Obs. Mean Std. dev. 25% 50% 75% Dependent Variable % change in destination volume 48,095 -18.4% 22.5% -26.7% -19.0% -11.9% Origin Δ % wfh * 2019 od vol 48,095 12.8% 9.8% 4.3% 12.3% 20.9% Δ % Net migration rate * 2021 od vol 48,095 -2.3% 2.6% -3.6% -1.9% -0.7% per capita income * 2019 od vol 48,095 44,064 16,729 31,775 39,913 53,940 Total population * 2019 od vol 48,095 35,015 10,881 27,850 35,729 41,498 % non-white * 2019 od vol 48,095 54.3% 17.6% 41.6% 56.3% 68.3% commute by public transit * 2019 od vol 48,095 6.6% 8.0% 1.3% 3.0% 9.0% median # room * 2019 od vol 48,095 5.23 0.52 4.92 5.26 5.54 no vehicles available * 2019 od vol 48,095 7.6% 5.6% 4.8% 5.9% 7.7% Destination % construction and manufacturing 487 14.3% 12.0% 5.0% 5.1% 20.3% % trade and transport 487 15.3% 9.5% 8.9% 13.6% 20.3% % office 487 21.3% 15.6% 10.7% 18.0% 27.2% % service and education 487 43.8% 19.7% 29.9% 43.1% 56.2% Note: The total number of observations (Obs.) is calculated by multiplying the number of days in the observation period by the total number of ZCTAs. However, it's important to note that not all destination ZCTAs received trips every day. Furthermore, due to privacy protection, StreetLight does not report trip counts for specific ZCTAs if the number of trips is identifiable. Table 4.5.2 presents the regression results, where the dependent variable is the percentage change in all-day traffic volume in destination ZCTAs from 2019 to 2021. The results from the regression analysis indicates that lower traffic recovery at the destination is associated with increasing remote workers at the origin and a higher share of office jobs at the destination. The findings from Model 1 indicate that a 1 percentage point increase in the work-from-home (WFH) rate is linked to a 0.28 percent decrease in destination traffic volume9 . Multiplying this coefficient by the mean share of WFH trip reduction derived from the descriptive analysis reveals that a 13-percentage point increase in WFH corresponds to a 3.64 percentage point decrease in destination all-day trip volume. This relationship remains significant in Model 2, 9 This is an elasticity that, if it holds in later analyses, is at the high end of elasticities of vehicle miles traveled with respect to land use variables. (See the summaries of VMT-land use elasticities available at https://ww2.arb.ca.gov/our-work/programs/sustainable-communities-program/research-effects-transportation-andland-use.) 74 though slightly reduced to -0.247, even when accounting for net migration rate changes. Model 2 also shows a significant positive effect of net migration rate on traffic recovery, with 1 percentage point change in the net migration rate is associated with a 2.756 percentage point change in traffic volume. Model 3 introduces industry composition variables, showing significant negative impacts of construction and manufacturing (-0.188), office (-0.513), and education and service (-0.156) sectors on traffic recovery. Although almost all job sectors see a decrease in volume, office jobs decrease the most traffic volume (-0.513) in terms of the magnitude of the coefficient. Model 4 includes the full set of controlling demographic variables. The results indicate that a higher share of non-white population and households with no vehicles available are associated with decreased traffic recovery. The coefficient for WFH trip reduction decreases to -0.102, although it remains significant. This suggests that, in terms of the magnitude of the coefficients, workplace industry characteristics are more influential on traffic volumes than worker demographics and remote working status. However, industry composition and residence demographics may be slower to change. To test the robustness of these findings, we conducted a robustness check in Table 4.5.3, using yearly average data instead of daily data. By averaging OD weights data over the entire period, results in Table 4.5.3 aimed to provide a more stable estimate of the relationships, minimizing potential noise from daily fluctuations in traffic volume. Upon comparison, the results from Table 4.5.3 largely corroborated those from Table 4.5.2. Despite the change in data frequency, the observed associations between the independent variables and the dependent variable remained consistent. Specifically, the negative relationship between WFH trip reduction and traffic volume, as well as the significant impact of the share of office jobs at the destination, persisted in Table 4.5.3. This consistency suggests that the identified relationships are robust to variations in the frequency of data aggregation. Overall, the results presented in Table 4.5.3 supports and strengthens the conclusions drawn from Table 4.5.2, providing confidence in the observed relationships between remote work, industry composition, and destination traffic volume changes. Despite the similarity in results between Table 4.5.2 and Table 4.5.3, the choice to use daily data in Table 4.5.2 offers advantages in terms of capturing the dynamic nature of traffic patterns, which may be masked when averaging data over longer time periods. 75 Table 4.4.2. Regression results, dependent variable = percentage change in all-day traffic volume in destination ZCTAs, daily, 2019 to 2021 * p<0.05 ** p<0.01 *** p<0.001 Table 4.4.3. Regression results, dependent variable = percentage change in all-day traffic volume in destination ZCTAs, yearly average, 2019 to 2021 * p<0.05 ** p<0.01 *** p<0.001 OD Variables 1 2 3 4 Origin WFH trip reduction : W = Δ % wfh * 2019 od vol -0.279*** -0.247*** -0.023 -0.102*** Δ % Net migration rate * 2021 od vol 2.756*** 0.341 per capita income * 2019 od vol 0.00001*** Total population * 2019 od vol 0.00001*** % non-white * 2019 od vol -0.096*** commute by public transit * 2019 od vol 0.019 median # rooms * 2019 od vol 0.031*** no vehicles available * 2019 od vol -0.287*** Destination (omitted: % agriculture) % construction and manufacturing -0.093*** -0.188*** % trade and transport 0.032 0.03 % office -0.444*** -0.513*** % education and service -0.088*** -0.156*** Constant -0.185*** -0.181*** -0.077*** -0.024 Observations 48095 48095 48095 48095 R-squared 0.036 0.039 0.11 0.144 Adjusted R-squared 0.036 0.039 0.11 0.143 OD Variables 1 2 3 4 Origin WFH trip reduction : W = Δ % wfh * 2019 od vol -0.230*** -0.083 -0.075 -0.032 Δ % Net migration rate * 2021 od vol 1.131*** -0.476 per capita income * 2019 od vol -0.000 Total population * 2019 od vol -0.082 % non-white * 2019 od vol -0.306 commute by public transit * 2019 od vol 0.102 median # rooms * 2019 od vol 0.054* no vehicles available * 2019 od vol -0.000 Destination (omitted: % agriculture) % construction and manufacturing 0.066 -0.008 % trade and transport 0.103 0.087 % office -0.264*** -0.286*** % education and service 0.014 -0.027 Constant -0.154*** -0.147*** -0.149*** -0.346* Observations 487 487 487 487 R-squared 0.035 0.078 0.172 0.247 Adjusted R-squared 0.033 0.074 0.163 0.229 76 5. Conclusion The COVID-19 pandemic has significantly reshaped commuting patterns and residential choices, with remote working becoming a new norm for many workers. This study analyzes the impact of remote work on job and housing locations in the Bay Area and Central Valley region. My research aims to understand how the rise in work-from-home (WFH), also known as remote working, or telecommuting, influences traffic volumes, residential migration, and workplace dynamics. I focused on three primary research questions: 1) Where and in which industries did traffic volume change the most during COVID-19? 2) Did the increase in work-from-home still reduce all-day traffic at workplaces as we entered the post-pandemic era? 3) Which areas experienced the highest number of move-ins and move-outs during COVID-19 To address these questions, I used four datasets: StreetLight for traffic data, LEHD LODES Workplace Area Characteristics (WAC) for job-related characteristics, the American Community Survey (ACS) for demographic data, and USPS Change of Address (COA) for migration patterns. The study area consisted of 487 ZCTAs in the Bay Area and Central Valley. My methodology included destination (workplace) analysis, origin (residential) analysis, origindestination (O-D) flow analysis, and a regression analysis to bring everything together. For destination characteristics, I found a notable decrease in traffic volumes across all regions from 2019 to 2021. The Bay Area experienced the largest drop at 20.5%, followed by the Central Valley at 16.4%, and Comparison counties at 17.6%. The industry composition of destination ZCTAs reveals distinct economic dynamics across the Bay Area and Central Valley. In the Bay Area, office jobs had the highest share at 26.1%, while agriculture had the lowest at 0.3%. In the Central Valley, the trade and transport sector dominated with 18.2%, whereas agriculture accounted for 6.8%. In the Comparison counties, the highest share was in agriculture at 10.5%, and the lowest was in office jobs at 17.6%. The education and service sectors were significant across all regions but had the highest share in the Bay Area at 46.9%. For origin characteristics, the Bay Area saw the most significant increase in WFH rates, rising by 10 percentage points from 6.6% in 2019 to 16.6% in 2021. Central Valley and Comparison counties saw increases of 4 and 5 percentage points, respectively. The Bay Area experienced a net cumulative migration loss of 1,473 people (approximately 4 people per 100 residents), indicating a higher outflow, potentially driven by high living costs and remote work flexibility. Central Valley and Comparison counties had smaller net losses of about 200 people each, representing a net loss of around 1 person per 100 residents. I used Origin-Destination (O-D) flow weights to connect the characteristics of destination and origin ZCTAs. This method helps us understand how residential demographics impact workplace locations by weighting the contribution of each origin ZCTA to destination ZCTAs based on traffic volume. This approach allows us to accurately link origin characteristics to their destinations, providing a clearer picture of how factors like work-from-home growth at origin affect traffic patterns at destination. 77 The concept of " WFH trip reduction " is used to estimate the reduction in trips due to increased remote working. The Bay Area experienced the highest volume of WFH trip reduction at 10,183, accounting for 13.6% of its 2019 traffic volume. The Central Valley and Comparison counties saw 4,324 (6.7%) and 2,873 (9.7%) WFH trip reduction, respectively. The WFH trip reduction align with the observed increases in WFH rates and the decreases in destination traffic volumes. The regression analysis showed that higher WFH rates at the origin are associated with less traffic recovery at the destination. A 1 percentage point increase in WFH rate is linked to a 0.28 percentage point decrease in destination traffic volume. In addition, the net migration rate at the origin has a positive effect on traffic volume changes, with a 1 percentage point increase in net migration linked to a 2.76 percent increase in destination traffic volume. However, industry composition at the destination is a more influential factor than origin characteristics. The analysis reveals that destinations with a higher share of office jobs experience the most significant declines in traffic volume. For example, office sectors show a significant negative impact, with a 1 percentage point increase in the share of office jobs associated with a 0.44 percent decrease in traffic volume. Other sectors, such as construction, manufacturing, education, and services, also negatively impact traffic recovery, though to a lesser extent. These findings highlight the critical role of workplace (destination) characteristics in shaping traffic patterns. The study period covers only up to October 2021, so ongoing changes might not be fully captured. In addition, the COA data does not allow us to fully analyze the migration patterns of remote workers. Future research could focus on where these remote workers have moved, where they came from, and how that affects urban structure. 78 References 1. Huws, U. Telework: Projections. Futures 23, 19–31 (1991). 2. Hopkins, J. L. & McKay, J. Investigating ‘anywhere working’as a mechanism for alleviating traffic congestion in smart cities. Technological Forecasting and Social Change 142, 258–272 (2019). 3. Nguyen, M. H. Factors influencing home-based telework in Hanoi (Vietnam) during and after the COVID-19 era. Transportation 48, 3207–3238 (2021). 4. Nilles, J. M. Traffic reduction by telecommuting: A status review and selected bibliography. Transportation Research Part A: General 22, 301–317 (1988). 5. Nilles, J. M. Telecommuting and urban sprawl: Mitigator or inciter? Transportation 18, 411–432 (1991). 6. Hughson, T. L. & Goodman, P. S. Telecommuting: Corporate practices and benefits. National Productivity Review 5, 315–324 (1986). 7. Mateyka, P. J., Rapino, M. & Landivar, L. C. Home-based workers in the United States: 2010. (U.S. Department of Commerce Economics and Statistics Administration, U.S. Census Bureau, 2012). 8. Pabilonia, S. W. & Vernon, V. Telework, Wages, and Time Use in the United States. Rev Econ Household 20, 687–734 (2022). 9. Barrero, J. M., Bloom, N. & Davis, S. J. Why Working from Home Will Stick. Working Paper at https://doi.org/10.3386/w28731 (2021) 10. Brynjolfsson, E. et al. COVID-19 and Remote Work: An Early Look at US Data. Working Paper at https://doi.org/10.3386/w27344 (2020) 11. Hansen, S. et al. Remote Work across Jobs, Companies, and Space. Working Paper at https://www.nber.org/system/files/working_papers/w31007/w31007.pdf (2023) 12. Bailey, D. E. & Kurland, N. B. A review of telework research: Findings, new directions, and lessons for the study of modern work. Journal of Organizational Behavior: The International Journal of Industrial, Occupational and Organizational Psychology and Behavior 23, 383–400 (2002). 13. Martin, B. H. & MacDonnell, R. Is telework effective for organizations? A meta-analysis of empirical research on perceptions of telework and organizational outcomes. Management Research Review 35, 602–616 (2012). 14. Ruppel, C. P. & Harrington, S. J. Telework: An innovation where nobody is getting on the bandwagon? ACM SIGMIS Database: the DATABASE for Advances in Information Systems 26, 87–104 (1995). 15. Storper, M. The Regional World: Territorial Development in a Global Economy. (1997). 16. Bartik, A. W., Cullen, Z. B., Glaeser, E. L., Luca, M. & Stanton, C. T. What jobs are being done at home during the COVID-19 crisis? Evidence from firm-level surveys. (National Bureau of Economic Research, 2020). at <https://www.nber.org/papers/w27422> 79 17. Aksoy, C. G. et al. Working from Home Around the World. Working Paper at https://doi.org/10.3386/w30446 (2022) 18. Bloom, N., Davis, S. & Barrero, J. M. 60 million fewer commuting hours per day: How Americans use time saved by working from home. CEPR (2020). at <https://cepr.org/voxeu/columns/60-million-fewer-commuting-hours-day-how-americansuse-time-saved-working-home> 19. Behrens, K., Kichko, S. & Thisse, J.-F. Working from home: Too much of a good thing? Regional Science and Urban Economics 105, 103990 (2024). 20. Delventhal, M. & Parkhomenko, A. Spatial Implications of Telecommuting. SSRN Scholarly Paper at https://doi.org/10.2139/ssrn.3746555 (2023) 21. Dingel, J. I. & Neiman, B. How many jobs can be done at home? Journal of Public Economics 189, 104235 (2020). 22. Harrington, E. & Kahn, M. E. Has the Rise of Work-from-Home Reduced the Motherhood Penalty in the Labor Market? Working Paper at https://drive.google.com/file/d/1Z1J2GHZjqkWzRV5ygA02yzSemwkAmzne/view?usp=sh aring (2023) 23. Adrjan, P. et al. Working from Home after COVID-19: Evidence from Job Postings in 20 Countries. SSRN Scholarly Paper at https://doi.org/10.2139/ssrn.4064191 (2022) 24. Alonso, W. Location and Land Use: Toward a General Theory of Land Rent. (Harvard University Press, 1964). doi:10.4159/harvard.9780674730854 25. Muth, R. M. Cities and Housing: The Spacial Pattern of Urban Residential Land Use. (University of Chicago Press, 1969). 26. Mills, E. S. An Aggregative Model of Resource Allocation in a Metropolitan Area. The American Economic Review 57, 197–210 (1967). 27. Kain, J. F. The journey-to-work as a determinant of residential location. Papers in Regional Science 9, 137–160 (1962). 28. Brueckner, J. K. in Handbook of Regional and Urban Economics (ed. Mills, E. S.) 2, 821– 845 (Elsevier, 1987). 29. Baum-Snow, N. Did highways cause suburbanization? The quarterly journal of economics 122, 775–805 (2007). 30. Mills, E. S. & Tan, J. P. A Comparison of Urban Population Density Functions in Developed and Developing Countries. Urban Studies 17, 313–321 (1980). 31. Ramani, A. & Bloom, N. The Donut Effect of Covid-19 on Cities. Working Paper at https://doi.org/10.3386/w28876 (2021) 32. Foroohar, R. US cities must beware the ‘donut effect’. (2023). at <https://www.ft.com/content/c7a3b822-0a98-4a95-9b49-63f0b96d9e98> 33. Kingson, J. A. ‘The donut effect’ is reshaping America’s cities. Axios (2023). at <https://www.axios.com/2023/12/15/cities-2024-donut-effect-crime-housing-conversioncongestion-pricing> 80 34. Lee, T. B. The Donut Effect: How the Pandemic Hollowed Out America’s Biggest Cities. Slate (2022). at <https://slate.com/business/2022/04/the-donut-effect-how-the-pandemichollowed-out-americas-biggest-cities.html> 35. Liu, S. & Su, Y. The impact of the COVID-19 pandemic on the demand for density: Evidence from the US housing market. Economics Letters 207, 110010 (2021). 36. Clark, T. N., Lloyd, R., Wong, K. K. & Jain, P. Amenities Drive Urban Growth. Journal of Urban Affairs 24, 493–515 (2002). 37. Glaeser, E. L., Kolko, J. & Saiz, A. Consumer City. Journal of Economic Geography 1, 27– 50 (2001). 38. Gupta, A., Mittal, V. & Van Nieuwerburgh, S. Work From Home and the Office Real Estate Apocalypse. SSRN Scholarly Paper at https://doi.org/10.2139/ssrn.4124698 (2023) 39. Ozimek, A. Remote Workers on the Move. SSRN Scholarly Paper at https://doi.org/10.2139/ssrn.3790004 (2020) 40. Ozimek, A. The New Geography of Remote Work. (Upwork, 2022). at <https://www.upwork.com/research/new-geography-of-remote-work> 41. Haslag, P. H. & Weagley, D. From L.A. to Boise: How Migration Has Changed During the COVID-19 Pandemic. SSRN Scholarly Paper at https://doi.org/10.2139/ssrn.3808326 (2022) 42. Mondragon, J. & Wieland, J. Housing Demand and Remote Work. San Francisco Fed Preprint at https://www.frbsf.org/economic-research/publications/working-papers/2022/11/ (2022) 43. Chapple, K. et al. The Death of Downtown? Pandemic Recovery Trajectories across 62 North American Cities. (School of Cities, University of Toronto; Institute of Governmental Studies, UC Berkeley, 2023). 44. Sassen, S. The Global City: New York, London, Tokyo. (University Press, 1991). 45. Allen, J. et al. Understanding the impact of e-commerce on last-mile light goods vehicle activity in urban areas: The case of London. Transportation Research Part D: Transport and Environment 61, 325–338 (2018). 46. Circella, G. & Mokhtarian, P. L. in The Geography of Urban Transportation (eds. Giuliano, G. & Hanson, S.) (Guilford Publications, 2017). 47. Mokhtarian, P. L. A conceptual analysis of the transportation impacts of B2C e-commerce. Transportation 31, 257–284 (2004). 48. Janelle, D. G. Metropolitan expansion and the communications-transportation trade-off. Geography of Urban Transportation (1986). at <https://trid.trb.org/View/238196> 49. Graham, S. & Marvin, S. Splintering Urbanism: Networked Infrastructures, Technological Mobilities and the Urban Condition. (Routledge, 2001). doi:10.4324/9780203452202 50. Golob, T. F. Travelbehavior.com: Activity Approaches to Modeling the Effects of Information Technology on Personal Travel Behavior. (2002). at <https://escholarship.org/uc/item/9t40s1mc> 81 51. Hamer, R., Kroes, E. & Van Ooststroom, H. Teleworking in the Netherlands: an evaluation of changes in travel behaviour. Transportation 18, 365–382 (1991). 52. Mokhtarian, P. L., Collantes, G. O. & Gertz, C. Telecommuting, Residential Location, and Commute-Distance Traveled: Evidence from State of California Employees. Environ Plan A 36, 1877–1897 (2004). 53. Obeid, H., Anderson, M. L., Bouzaghrane, M. A. & Walker, J. L. Does telecommuting reduce trip-making? Evidence from a US panel during the COVID-19 impact and recovery periods. Evidence From a US Panel During the COVID-19 Impact and Recovery Periods (2022). at <https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4213516> 54. Caros, N. S. & Zhao, J. Evaluating the travel impacts of a shared mobility system for remote workers. Transportation Research Part D: Transport and Environment 121, 103798 (2023). 55. Zheng, Y., Wang, S., Liu, L., Aloisi, J. & Zhao, J. Impacts of remote work on vehicle miles traveled and transit ridership in the USA. Nature Cities 1–13 (2024). 56. Loo, B. P., & Wang, B. (2018). Factors associated with home-based e-working and eshopping in Nanjing, China. Transportation, 45(2), 365-384. 57. Mitomo, H., & Jitsuzumi, T. (1999). Impact of telecommuting on mass transit congestion: the Tokyo case. Telecommunications policy, 23(10-11), 741-751. 58. Kitamura, T., Aso, Y., Kuniyoshi, N., Hara, K., & Yogo, Y. (1990). High incidence of urinary JC virus excretion in nonimmunosuppressed older patients. Journal of Infectious Diseases, 161(6), 1128-1133. 59. Stiles, J., & Smart, M. J. (2021). Working at home and elsewhere: daily work location, telework, and travel among United States knowledge workers. Transportation, 48(5), 2461- 2491. 60. Yamamoto, T., & Kitamura, R. (1999). An analysis of time allocation to in-home and outof-home discretionary activities across working days and non-working days. Transportation, 26, 231-250. 82 Chapter 4. The Impact of Remote Work on GHG Emissions in the United States in the Post-Pandemic Era: A National Analysis of the Relationship Between COVID-19, Commuting, Residential Choice, and Emission Reduction Author: Bonnie S. Wang 83 Abstract The transition to remote working post-COVID-19 has significantly altered commuting patterns and has implications for reducing greenhouse gas (GHG) emissions. This study examines how changes in commuting behavior impact the spatial relationship between homes and workplaces. I investigate whether workers with remote work capabilities tend to move farther from their workplaces and how this affects their commuting behavior. I conducted a survey using IPSOS’ KnowledgePanel in September 2023, to gather data from a nationally representative sample of 2,124 full-time working adults in the U.S. The survey included individuals with and without the ability to work from home, further divided by whether they relocated post-COVID and whether their commuting habits changed. This data provides insight into the spatial and environmental implications of remote work, aiding transit and planning agencies in modeling air quality and traffic congestion. The results confirm that fully remote workers tend to relocate farther from their workplaces and produce less GHG emissions as they no longer commute, while GHG emissions for fully in-person workers remain relatively stable. Hybrid workers tend to move farther away from their workplace relative to in-person workers. However, their overall weekly emissions significantly drop due to the reduction in commute days, offsetting any increased distance between their home and jobs. Hybrid workers, who typically commute on midweek days, witness a notable increase in greenhouse gas (GHG) emissions during these periods, surpassing prepandemic levels when their residences were closer to their workplaces. This could potentially worsen traffic congestion and air pollution on these days. Keywords: Commute, travel behavior, remote work greenhouse gas, environment 84 1. Introduction 1.1 Background The COVID-19 pandemic saw one of the most significant changes in work and commute patterns ever experienced. Telecommuting, or remote working, experienced large increases, stemming from the desire to prevent infection in-person. Pre-pandemic, the share of telecommuters remained persistently low, ranging from 4% in 2006 to 6% of employed people working from home full time in 2019 according to the Census American Community Survey. The Bureau of Labor Statistics (BLS) put the figure higher at 9% in February 2020.10 No matter the data source, the spread of COVID-19 in early 2020 led to a dramatic increase in remote working that has not been entirely reversed. Although many employees were required to return to the office starting in 2023, the U.S. Household Pulse Survey11, as of September 2023, reported that 26% of adults in the U.S. still substituted some or all of their typical in-person work for telework. Hybrid working schedules could become the “new norm”, enabling and prompting telecommuters to move farther away from job centers as they substitute what was previously a greater commute frequency (e.g., each weekday) for lower commute frequency and greater commute distance. Evidence of increased housing demand in rural areas and smaller metros is consistent with workers moving away from job centers (Arend et al., 2023). If work is hybrid, with persons working at home and in an office, vehicle miles traveled could increase or decrease depending on the balance of longer distances between home and office, lower frequency commuting, and the possibility that persons might substitute other non-work driving for free time that results from less commuting. No empirical studies have examined the joint impact of changes in working-from-home, jobs-housing re-location and the associated environmental impacts in the post-pandemic era. This project analyzes how changes in commuting behavior following COVID-19 have altered the spatial relationship between homes and workplaces. It aims to offer the best estimate currently available regarding the impact of remote work on greenhouse gas emissions, helping transit and planning agencies in modeling air quality and traffic congestion. Throughout the paper I use the COVID-19 pandemic as a reference period. The pandemic disrupted economies worldwide and led to the rapid adoption of remote working for millions. I use March 2020, when most U.S. states instituted physical distancing mandates that shut down most workplaces, as the beginning of the pandemic’s effect on work arrangements and refer to the time before as pre-COVID. The survey I use as the primary source of data was fielded in September 2023 and the questions asked survey respondents about their current situation and, occasionally, the pre-COVID time period. The World Health Organization declared the end of 10 The share of telecommuters varies with sources because of differences in definition and methodology. In this report we aim to use consistent points of reference and will use the Bureau of Labor Statistics figure because it separates full-time and part-time remote workers. Please see BLS article for detail: https://www.bls.gov/opub/mlr/2022/article/telework-during-the-covid-19-pandemic.htm 11 U.S. Household Pulse Survey database: https://www.census.gov/programs-surveys/household-pulse-survey.html 85 the COVID-19 pandemic in May 2023, so I refer to the time period at the time of the survey as post-COVID. 1.2 Research Purpose This project evaluates the spatial adaptation linked to remote working, how jobs-housing locations changed (i.e., by persons moving their residence or job location), and how the resulting configuration affected greenhouse gas (GHG) emissions in the post-pandemic era. I used a nationally representative survey that included retrospective and prospective questions asking respondents about the history of their work location and the most recent move over the last two years, their current residential location and commute patterns, and anticipated remote work and moves in the near future. While it will likely take years for the repercussions of the pandemic to play out in the housing and labor market, the survey in September 2023 was fielded at a time when trends in work arrangements had stabilized and the housing market had significantly slowed down due to higher mortgage rates. The survey, therefore, fills an important gap for research and policymaking. By using some prospective questions about anticipated behavior, I also get insights into possible future evolutions of remote work and commuting behavior. The project aims to answer three sets of questions: 1. During the pandemic, did workers who can work remotely move more frequently? Were they more likely to move farther away from job centers? For workers who work hybrid, as the frequency of in-person workdays increases, are persons more likely to live closer to job centers? 2. What factors affect the relationship between home location and work location for remote workers? Does the relationship vary by demographic and industry composition? 3. Does working remotely lead to less driving and a reduction of greenhouse gas emissions compared with other work arrangements, including hybrid in-person/remote models? Are reductions in driving higher for people who work full-time remotely? How does driving and related GHG emissions for hybrid in-person/remote workers compare with driving and GHG emissions for persons who work fully in-person? Additionally, I used retrospective and prospective questions to query past and anticipated moves, to assess if current patterns will persist or if, for example, remote workers are returning to the office. 86 1.3 Contribution and Significance This project is one of the first to provide an early look at the impact of working-fromhome during a time when the initial COVID shock has passed, but when work and residential location relationships are still in flux and likely adjusting. I understand that the long-run equilibrium might not yet be evident, but the initial shock has passed and by 2023 most workers had settled into a more permanent work arrangement as evidenced by the stability of the BLS monthly tracking of remote working rates.12 This report contributes insight into how people balanced the number of times they commute to work, how far they commute, and how the vehicle they drive impacts commute-related GHG emissions. Metropolitan planning organizations such as the Southern California Association of Governments (SCAG) and the San Diego Association of Governments (SANDAG) are already anticipating that remote work will reduce commute-based GHG emissions, and they are incorporating that into their plans to meet state mandated GHG emission targets and federally mandated air quality targets (True North Research, 2021). Yet, existing data and models give limited insight into the COVID-induced remote work phenomenon. The survey provides better evidence on the magnitude of changing remote work patterns on GHG emissions and on the circumstances under which hybrid working, specifically, can contribute to increased emissions. I estimate whether remote or hybrid workers actually generate lower GHG emissions on average than in-person workers. The results could offer the best estimate currently available regarding the impact of remote work on greenhouse gas emissions, helping transit and planning agencies in modeling air quality and traffic congestion. 12 Labor Force Statistics from the Current Population Survey. Telework or work at home for pay: https://www.bls.gov/cps/telework.htm. Share of persons who telework to some extent across various months is as follows: September 2023: 19.8%; May 2023: 18.9%; January 2023: 19.4%; October 2022: 17.9% 87 2. Literature Review The often-used term these days, work-from-home, is one of the forms of telecommuting. The concept of telecommuting was first formed by Nilles in 1973 and has been a field of interest for transportation researchers since then. Telecommuting is used to describe working outside of the workplace during standard work times (Ory and Mokhtarian, 2006; Bailey and Kurland 2002). Other terminologies include telework, remote work, distance work, e-work, flexplace, and electronic cottage (Teo and Lim 1998). Telecommuting has been considered as a potential mechanism to alleviate traffic congestion, reduce emissions, improve air quality, and create environmental benefits in our cities (Hopkins and McKay, 2019; Nguyen, 2021). However, some scholars argued that these forecasts could be overly optimistic (Gold, 1991). Prior to COVID-19, the share of workers working-from-home in the U.S. remained persistently low. According to the ACS, from 1997 to 2010, the percentage of people working from home at least one day a week increased by only 2.5 percentage points, from 7% to 9.5%. Zhu et al. (2018) also noted that only about 9% of the U.S. workforce worked from home more than once a week. The spread of the COVID-19 pandemic significantly shifted this trend, making work-from-home more prevalent. Although many employees were required to return to the office starting in 2023, the U.S. Household Pulse Survey reported that as of September 2023, 26% of adults in the U.S. still substituted some or all of their in-person work with telework. 2.1 Remote Work, Residential Location Choices, and Commute Distances The relationship between remote work and residential location choices presents a complex challenge in reducing overall travel distances. While teleworking can significantly reduce travel on days when employees work from home, the benefits may be offset by the tendency for teleworkers to choose residences farther from their workplaces (Ory and Mokhtarian, 2006; Graham and Marvin, 1996). When workers become more able to work remotely, the new flexibility to avoid commuting can affect workers' residential location choice (Liu and Su, 2021). The standard urban model (Alonso, 1964; Muth, 1969; Mills, 1967) predicts that households will trade long commutes for lower land prices (and hence lower housing prices) on the urban fringe. Similarly, Kain (1961) posits that households trade-off commute costs for residential site costs (as cited in Brueckner, 1987). Decades of urban economics research have verified those predicted patterns (see, e.g., Mills and Tan, 1980, for an early example). The standard urban model leads to predictions that persons will consume lower-cost housing far from the urban core and in effect trade longer commutes for more land or, on a per-unit basis, lower-cost housing. Telecommuting, by altering commute costs, can change residential location choices and equilibrium land uses in ways that can be predicted by the standard urban model. Several studies have explored how telecommuting frequency influences travel outcomes. Hu and He (2016) found that part-time telecommuters often have longer commutes than both full-time telecommuters and those who do not telecommute at all. This suggests that less frequent teleworkers may choose to live further from their workplaces, potentially diminishing 88 the travel savings that telework is expected to provide. Mokhtarian and Varma (1998) demonstrated that participation in neighborhood telecenter programs can lead to a significant reduction in travel distances for remote workers. However, frequent telecommuters may still travel greater distances annually by car than non-telecommuters. Chakrabarti (2018) noted that the longer commutes on non-teleworking days often outweigh the reductions achieved on teleworking days, leading to a net increase in travel for some remote workers. Research also shows that telecommuting can lead to substantial reductions in vehicle miles traveled (VMT). For instance, Choo et al. (2005) estimated that teleworking by 12% of the U.S. workforce once a week reduced annual VMT by 0.8%. Lari (2012) further supported these findings, showing that vehicle miles traveled per person were nearly 28 miles lower on teleworking days compared to traditional commuting days. Similarly, Giovanis (2018) reported that teleworking by 8.43% of the population in Switzerland was associated with a 1.9% reduction in traffic volume, along with corresponding decreases in pollutants. The overall impact of teleworking on commute distances appears to be contextdependent. While some teleworkers successfully reduce their travel distances, others may inadvertently increase their total travel due to residential location choices that prioritize living further from work (Janelle, 1986; Graham and Marvin, 1996). Henderson et al. (1996) found that U.S. teleworkers who use telecenters program have 91% higher VMT on non-teleworking days compared to non-teleworkers, suggesting that they may live farther from their workplaces. In contrast, on teleworking days, home-based teleworkers reduced their travel distance by 67% compared to traditional commuters. Mokhtarian et al. (2004) found that, on average, quarterly commute distances are 15% lower for teleworkers than for non-teleworkers, indicating that teleworking can effectively reduce travel when practiced regularly enough to offset longer oneway commutes. 2.2 Migration pattern during COVID-19 A series of research papers have developed spatial equilibrium models to analyze migration patterns during COVID-19. Many of these studies observed a decrease in housing demand in city centers and an increase in more distant areas during the pandemic. Liu and Su (2021) were among the first to investigate the impact of COVID-19 on housing demand at the neighborhood level, finding that the pandemic significantly reduced demand in central cities and higher-density neighborhoods. Delventhal, Kwon, and Parkhomenko (2021) conducted a spatial equilibrium model and found that job opportunities continued to concentrate in city centers even as residents increasingly chose to relocate away from urban areas. Couture et al. (2021) analyzed cell-phone data and identified a significant outflow of individuals from New York City. Meanwhile, Haslag and Weagley (2021) used cross-state moving data from a moving company to observe a trend where predominantly high-income individuals were relocating to smaller, more cost-effective cities. Ozimek (2020) conducted a survey revealing a notable increase in planned relocations, with over half of the respondents expressing a desire for more affordable housing due to the rise in remote work. 89 Ramani and Bloom (2022) shed light on this issue, revealing a shift in real estate demand. This shift is evident in both rents and property prices as people move away from major city centers toward less densely populated areas on the outskirts of cities. This phenomenon has been broadly observed but is more pronounced in larger cities. Both people and businesses have been relocating from major metropolitan areas to smaller cities and rural regions. One possible explanation for this trend is the greater flexibility of housing supply in less densely populated regions, which helps stabilize property prices even with changing population patterns. Ramani and Bloom (2022) discuss significant differences in rent and home price growth as well as population and business migration patterns between central business districts (CBDs) in the largest 12 US metropolitan areas and less densely populated areas during the COVID-19 pandemic. Rent growth and home price growth in CBDs lagged behind the less dense areas by around 15 to 20 percentage points compared to the growth observed in the least dense 50% of zip codes, after adjusting for pre-pandemic trends. Additionally, migration patterns based on USPS data revealed that CBDs experienced net population and business outflows, while less dense areas gained a small percentage of their pre-pandemic population and businesses. 2.3 Environmental impact of Remote Work The traditional approach to evaluating transportation systems focuses on automobile traffic flow and congestion reduction. However, there's a noticeable shift underway. In response to the pressing need to combat climate change and reducing greenhouse gas (GHG) emissions, many cities, regions, and states are reassessing their priorities. Instead of solely focusing on metrics that measure congestion, there is a growing emphasis on reducing the overall amount of driving (Obeid et al., 2022; Brownstone, 2008). The shift to remote work has potential implications for environmental sustainability, particularly in reducing energy consumption and greenhouse gas emissions associated with commuting. The transportation sector, which accounts for about 33% of final energy use in the United States, stands to see significant changes as commuting patterns changes (Zhu & Mason, 2014). One key benefit of remote work is the reduction in commuter travel, which could lead to decreased energy consumption and lower carbon emissions (Hook et al., 2020). However, the environmental benefits of remote work are complex and not universally agreed upon. While some studies suggest substantial reductions in energy consumption—up to 77% primarily through avoided commuting (Koenig et al., 1996)—others indicate more modest gains or even paradoxical outcomes where energy use increases (Rietveld, 2011). This paradox may occur because teleworkers, who might live farther from their workplaces, could end up making longer trips on the days they do commute, offsetting the energy savings achieved on teleworking days (Bailey & Kurland, 2002). Many studies suggest that work-from-home or hybrid work arrangements has the potential to reduce greenhouse gas emissions. Wu et al. (2024) conducted a comparison between work-from-home (WFH) and traditional work-in-office (WIO) scenarios, finding that WFH led to a 29.11% reduction in greenhouse gas emissions. However, while WFH resulted in lower 90 emissions from commuting and workplace operations, it also led to an increase in residential emissions. In addition to work-related travel, the shift to remote work may also influence nonwork travel behavior. For instance, the time saved from commuting might be redirected to additional trips for leisure or social activities, potentially increasing overall travel and energy consumption (Lyons et al., 2008). This change in behavior suggests that the environmental impact of remote work could vary depending on how individuals use their saved time and whether remote work is adopted on a part-time or full-time basis. Despite these complexities, remote work still offers a notable opportunity for reducing energy use and emissions. For example, if 20% of workers telework one day a week, commuting energy consumption could decrease by 20% (Larson & Zhao, 2017). Similarly, teleworking one or more days per week by 3% of the U.S. workforce could lead to a reduction in annual primary energy consumption by up to 0.18% and a decrease in CO2 emissions by up to 0.23% (Roth et al., 2008). Kitou and Horvath (2003) estimated that teleworking between one and five times a week could decrease CO2 emissions by 2% to 80%. Moreover, Shabanpour et al. (2018) found that when 50% of workers adopt flexible work schedules, total daily vehicle miles traveled (VMT) and vehicle hours traveled (VHT) could be reduced by up to 0.69% and 2.09%, respectively, with corresponding reductions in greenhouse gas emissions. 91 3. Research Design and Data Large-scale, publicly available datasets are unsuitable to study the impact of changing work arrangements because they do not allow for the linking of individual records that enable the kind of before-and-after comparisons required for this research. In the absence of a suitable dataset, I developed a survey instrument that the firm IPSOS fielded in September 2023 using their nationally representative KnowledgePanel® panel of respondents. This section provides an overview of the research design and specifics about the survey. 3.1 Research Design The primary purpose of this research was to examine two main changes that would directly affect how people commute: whether they moved to a new location (and whether they concurrently changed jobs), and whether they were able to work remotely (part- or full-time). All survey respondents were adults (18 years old or older) working full-time at the time of the survey. This study aimed to collect 2,000 responses from a nationally representative sample that provide a sample of at least 500 responses in each of the outlined dimensions shown in Table 3.1.1. The 2x2 matrix divides the survey sample into workers who are able to work remotely or not, and for each type of worker, whether they moved after the Covid outbreak (March 2020) or not. These dimensions were used as selection devices so that respondents that did not fit into one of the matrix cells (or that fit into a cell for which I had already collected enough data) were excluded from answering further questions. I collected a total number of 2,214 responses, with each category meeting the target of 500 responses. This survey has been approved by the Institutional Review Board at the University of Southern California (Study ID: UP-23-00703). Table 3.1.1 Unweighted responses by working arrangements and moving status Type Moved at least once after covid Did not move after covid Total Able to work remotely 535 528 1,063 Not able to work remotely 527 534 1,061 Total 1,062 1,062 2,124 The survey collected information about how changes in work arrangements were associated with where people live and their daily routine. (Details of the survey questions are available in Appendix A.) The survey questions focused on three topics. The first set of questions gathered information about residential characteristics and their move history. I asked respondents about their current and past home zip codes, whether they moved, how many times they moved, when their last move occurred, why they moved, and how far they moved. The second set of questions focused on work arrangement and changes in employment. I asked respondents’ current and past job zip codes, how long have they worked for their current 92 employer, whether they changed jobs after March 2020, how many times they changed jobs, why they changed jobs, how many days in a week they worked remotely before and after the start of the pandemic, and which day(s) of the week they typically work remotely. The last set of questions focused on commute behavior, including information about the main vehicle respondents used. Questions covered the respondents’ commute mode, whether it changed after covid, vehicle make/model/year, and the frequency of driving to non-commute related activities. The questions provide a rich, but complex, picture of people’s trajectories from March 2020 to late 2023. I processed respondent’s responses to categorize people and create a simpler set of parameters to summarize data. Using respondents' work arrangements pre- and postCOVID, work arrangements are classified into three types: • Remote: Workers who work fully remote for 5 or more days per week. • Hybrid: Workers who commute to work in-person at least once a week, but no more than four days a week. • In-person: Workers who work fully in-person for 5 or more days per week. In addition to categorizing respondents, I used information about their zip code(s) of residence and work to calculate the distance they moved and the distance between their home and place of work pre- and post-pandemic. For all distance calculations, I was constrained to using the zip code information. Zip codes, because they are used for the purpose of mail delivery and can change inconsistently, are not set geographic areas the way other administrative units like counties and cities are. However, the US Census Bureau developed the Zip Code Tabulation Area (ZCTA) to approximate the boundaries of postal zip codes and to allow researchers to match zip code information with consistent geographic boundaries and census data. I match respondents’ zip code information to the ZCTA and use the geographic center of each ZCTA to calculate distance between locations. In cases where the respondents moved within the same zip code or lives and works in the same zip code, I am unable to calculate a distance and assign a value of zero. There are limitations to this method. First, zip codes vary in size, so distances between the centers of any two zip codes are less likely to be actual commute distances in rural and suburban areas where zip codes tend to be large. The net error from this, though, will balance under- and over-estimations of distance because I use the geographic centroid of ZCTA’s for all calculations. Second, because I have no information about respondents’ locations other than their zip code of residence, any moves within the same zip code are coded as a 0-mile move, leading to an underestimate of the mean move distance. Close to one in ten movers changed residence within the same zip code.13 Third, when respondents work and live in the same zip code, this is again coded as zero miles, also leading to potential undercounting of home-work distances. 13 The share of very local moves (within the same zip code) is in line with other research based on zip code moves using a larger dataset in California (Boarnet et al., 2023) 93 3.2 Sample Definition and Field Period The survey was conducted using KnowledgePanel®, the largest online panel in the United States, which employs probability-based sampling methods to recruit a representative sample of adults in the country. IPSOS invited one adult per household from a representative sample of households to participate in the survey. Selected panel members received an email invitation to complete the survey at their earliest convenience, with the subject and body of the email invitation provided in Appendix A. IPSOS translated the original English-language survey into Spanish and made both languages available to respondents. (The survey instruments in both languages are available in Appendix A). I piloted the survey internally with a sample of convenience to refine the instrument and gauge survey length before sharing the survey with IPSOS for their own pre-test (see Table 3.2.1 for sample details). The finalized survey was fielded to the full panel in late September 2023 and reached the target number of responses in each of the cells for the matrix (Table 3.1.1) in October 2023. The median completion time for the main survey was 6 minutes. Respondents were unable to complete the survey more than once and qualified respondents were entered into the KnowledgePanel® sweepstakes upon completion. Table 3.2.1 shows the timeline of the survey. I pretested the survey in August 2023 and fielded the main survey from September 22 to October 12, 2023. The Completion rate for the main survey is 63%. This number indicates the percentage of individuals who finished answering all the questions in the survey out of the total number of invited participants. The qualification rate, at 18%, is the percentage of completed surveys that met specific eligibility criteria. The lower qualification rate can be attributed to the need for a larger number of completed surveys to achieve the targeted 500 responses for each subcategory outlined in Table 3.1.1. The sample is nationally representative and includes respondents from most areas of the United States. The map (Figure 3.2.1) shows the current home locations of survey respondent. Table 3.2.1. Completion and qualification rates of the survey Field Start Field End N Fielded N Completed Completion Rate N Qualified Qualification Rate Pretest 8/9/2023 8/14/2023 200 102 51% 40 39% Main 9/22/2023 10/12/2023 19,000 12,011 63% 2,124 18% 94 Figure 3.2.1. Current home location of survey respondents 95 3.3 Survey Methodology I summarize key elements of the survey methodology based on the IPSOS KnowledgePandel Book.14 The KnowledgePanel® draws from a sample of 60,000 respondents to field nationally representative research surveys. Panel members are selected randomly through probability-based sampling. After accepting the panel invitation, participants complete a brief demographic survey. The information collected in this survey records demographic characteristics for all panel members. As a result, in my main survey, I don't need to ask extra demographic questions since I received the full set of demographic data for each respondent from IPSOS. IPSOS provides sample weights based on the geographic distribution of demographic information from national surveys such as the Current Population Survey and American Community Survey (see Appendix B for more details on benchmarking). However, studies like this one generally require a specific subset based on a set of criteria that compromise the representative nature of the sample. The difference between the completion and qualification rate in Table 3.2.1 illustrates some deviation from the original pool of respondents. IPSOS provides a second set of weights specific to my survey to ensure the results are still nationally representative. The probability-proportional-to-size (PPS) procedure yields demographically balanced and representative samples that match the weighing obtained for the full panel. Table 3.2.2 shows the 4x4 matrix containing weighted responses for my screening questions: 1) Are you currently able to work remotely? and 2) Have you permanently changed residence since the pandemic began (March 2020)? Table 3.2.2. Weighted shares (weighted responses divided by 2,124 total respondents) Able to remote work? Did you move at least once since March 2020? Yes No Total Yes 14% 35% 49% No 12% 39% 51% Total 26% 74% 100% (2,124) While IPSOS uses a rigorous method to reach respondents and weight the responses, there are certain limitations that survey cannot avoid. Errors from respondents’ answers (failure to recall an answer precisely, conscious or unconscious distortion of an answer, and lying are all possibilities) and limits in the ability to reach all relevant population means that no survey is perfect. 14 For detailed methodology, please refer to: https://www.ipsos.com/sites/default/files/KnowledgePanel_Book.pdf 96 4. Results My analysis aims to shed light on how people adapted their daily driving habits after the onset of the pandemic and following the normalization of remote work. People who switched to working remotely eliminated work-related traveling entirely, but many people switched to hybrid work arrangements and continued to commute at least once a week. The goal of the analysis is, in part, to disentangle the tradeoff between emission reduction from commuting to work less frequently and the increased driving distance due to living farther away from their work location (presumably because they commute less often). Vehicle emissions is the third element muddling this relationship. Someone traveling farther but using a clean-fuel car will have less impact than someone who continues to commute shorter distances alone in a highly polluting vehicle. I approach this by first looking at who works remotely full-time, part-time, or not at all, as well as how they get to work. I then use the results of the demographic analysis to test hypotheses relating to each aspect of the issue: how far people moved, how much longer is their commute, and, given the vehicle they use, did the balance of moving farther from work and driving less often to work result in a reduction of weekly GHG emissions for the average person? Throughout this chapter, I report results that have been weighted and are derived from specific questions. All tables and analyses presented in Chapter 4 have been weighted to align with the national population benchmark shown in Appendix B. For the exact wording of questions, please refer to Appendix A for the complete questionnaire. Summary tables reference the corresponding question numbers in the notes. Before delving into the results in detail, I have listed all the relevant variables extracted from survey questions that will be discussed in the following sections. 97 Table 4. Summary of variables Category Variable Related Survey Question Definition Work arrangement 2023 Remote Q1. Are you currently able to work remotely? As of September 2023, Workers who work fully remote for 5 or more days per week. Hybrid As of September 2023, Workers who commute to work in-person at least once a week, but no more than four days a week. In-person As of September 2023, Workers who work fully in-person for 5 or more days per week. Work arrangement 2020 Remote Q6_2. How many days per week did you actually work remotely in your job before March 2020? Before March 2020, Workers who work fully remote for 5 or more days per week. Hybrid Before March 2020, Workers who commute to work in-person at least once a week, but no more than four days a week. In-person Before March 2020, Workers who work fully in-person for 5 or more days per week. Commute days Commute days 2023 Q4_2. How many days per week do you actually work remotely in your current job? Days of commute to workplace per week in 2023 Commute days 2020 Q6_2. How many days per week did you actually work remotely in your job before March 2020? Days of commute to workplace per week in 200 ∆ Commute days Calculated from commute days 2023 and 2020 The difference between days of commute in 2023 and 2020. HomeJob Distance HomeJob Distance 2023 Q1a. What is your current home zip code: Q3. What is your current job zip code: Distance between home and workplace before March 2020. HomeJob Distance 2020 Q1b. What was your home zip code before the pandemic began (March 2020): Q6. What zip code did you work in most often before the pandemic began (March 2020): Distance between home and workplace in September 2023. ∆ HomeJob Distance Calculated from HomeJob Distance 2023 and 2020 Change in distance between home and workplace from pre-COVID (before March 2020) to postCOVID (September, 2023). Job Change Yes Q5. Have you changed jobs since the pandemic began (March 2020)? Workers who changed job after March 2020 No Workers who did not change job after March 2020 Moving status Moved Q2. Have you permanently changed residence since the pandemic began (March 2020)? Workers who moved after March 2020 Not moved Workers who did not move after March 2020 Moving distance The straight-line distance between the centroids of their home ZIP codes as of March 2020 and September 2023. *Unit of analysis is individuals. 98 Table 4. Summary of variables (continued) Category Variable Variable Related Survey Question Vehicle GHG emission Q7_2. What is the car make/model/year you use for your current commute most often? The make / model/ year Commute greenhouse gas emissions Weekly commute CO₂ per person 2020 Q7_2. What is the car make/model/year you use for your current commute most often? Calculated from HomeJob Distance 2023 and 2020 tailpipe CO₂ grams per person per week from commuting to work before March 2020. Weekly commute CO₂ per person 2023 tailpipe CO₂ grams per person per week from commuting to work in September 2023. ∆ weekly commute CO₂ per person change in tailpipe CO₂ grams per person per week from pre-COVID (before March, 2020) to postCOVID (September, 2023) from commuting to work Commute mode 2023 Q7. On days you go to work (for your current job), you commute by: Mode of transportation commuting to workplace, including: 1. Walking / biking 2. Car, single occupant (only yourself) 3. Car, multiple occupants (a carpool) 4. Ride share (Uber, Lyft), taxi, or vanpool 5. Bus 6. Train 7. Other (Please specify): Commute mode 2020 Q7_1. Pre-pandemic (before March 2020), on days you go to work, you commuted by: Remote work ability Survey demographic profile An index that summarizes workers’ ability to work remotely (factors include demographic and industry characteristics). *Unit of analysis is individuals. 4.1 Work arrangement preferences In this section, I examine the shifts in work arrangements before and after the COVID-19 pandemic, as well as workers' preferences regarding their work settings. My survey uses both retrospective and prospective questions to gather information on past and anticipated future work arrangements. This helps us determine whether current trends are expected to continue or if there are indications suggesting possible changes in the future, for example, remote workers planning to return to in-person work. I first focus on the three types of work arrangements I defined: • Remote: Workers who work fully remote for 5 or more days per week. • Hybrid: Workers who commute to work in-person at least once a week, but no more than four days a week. • In-person: Workers who work fully in-person for 5 or more days per week. 99 The total number of respondents is 2,124. Due to missing responses, the total valid responses for each analysis may vary slightly. For work arrangements, there are 2,105 responses for preCOVID and 2,114 for post-COVID. The composition of workers across various work arrangements changed between the pre-COVID and post-COVID periods. For example, hybrid workers from the pre-COVID era may not be the same individuals observed in the post-COVID period. Some may have changed jobs and become either post-COVID in-person or remote workers. Table 4.1.1 indicates that the proportion of remote workers nearly doubled compared to preCOVID, from 11.6% to 22.6%, with hybrid workers showing a similar trend. The figures are higher than those reported by the Bureau of Labor Statistics15 (19.8% of all workers involved teleworking at least some of the time, that is, hybrid or fully remote, in September 2023). Discrepancies of this magnitude are to be expected considering that the share reporting remote work arrangements is sensitive to sampling, survey design, and wording of the questions. In late 2020, researchers found a similar difference between their survey and BLS numbers (Brynjolfsson et al., 2020; Brynjolfsson et al., 2023). The proportion of in-person workers in my survey consequently decreased by about 20 percentage points, from 76.3% to 56.6%, from preto post-COVID. Table 4.1.1. Share of different types of work arrangements from pre-COVID to postCOVID Work Arrangement Pre-COVID Post-COVID Freq. Percent Freq. Percent Remote 244 11.6% 478 22.6% Hybrid 256 12.1% 440 20.8% In-person 1,605 76.3% 1,196 56.6% Total 2,105 100% 2,114 100% *Source: survey Question: Q4_2, Q6_2 Table 4.1.2 shows the shift in work styles moving away from in-person arrangements. While 71% of in-person workers remained in-person, 17% switched to working hybrid, and 11% switched to working fully remotely. Of those working under a hybrid arrangement, a third switched to being remote full time, and most remote workers before the pandemic were still remote in 2023. 15 Labor Force Statistics from the Current Population Survey, Telework or work at home for pay: https://www.bls.gov/cps/telework.htm 100 Table 4.1.2. Shift in work arrangement from pre-COVID to post-COVID Work Arrangement Post-COVID Remote Hybrid In-person Total Pre-COVID Remote 9% 1% 1% 12% Hybrid 4% 6% 1% 12% In-person 9% 13% 54% 76% Total 22% 21% 57% 2,097 (100%) *Source: survey Question: Q4_2, Q6_2 Worker preferences do not always align with the employers’ remote work policy. Some people may prefer to work in the office more often while others favor a greater share of remote work. Table 4.1.3 shows that most workers choose the maximum number of remote work days allowed by their employers, with few exceeding these limits (see percentages along the main diagonal and cells to the right of the diagonal). Among hybrid workers who are allowed to work remotely 1 to 3 days, 7% to 13% reported working more days remotely than allowed. In contrast, two in five hybrid workers who are allowed 1 to 3 days of working remotely commute to their workplace more days than they are required. The share of individuals working in-person more regularly than required by their employer decreases to 30% for those working remotely 4 days a week. Even among those permitted to work remotely five days a week, 14% of these remote workers still go into a physical work location to some extent. These results highlight the distinction between the ability to work remotely and the reality of working remotely, which vary depending on individual and workplace characteristics. I explore this further in section 4.2. Table 4.1.3. Current work arrangement preference - interior rows sum to 100% Remote work days Days actually working remotely 0 1 2 3 4 5 and up Share of Total Days allowed to work remotely 0** 98% 0.3% 0.1% 0% 0% 1% 1,098 (52%) 1** 44% 50% 3% 1% 1% 2% 101 (5%) 2 19% 19% 51% 8% 1% 2% 170 (8%) 3 6% 11% 15% 54% 8% 5% 138 (7%) 4 4% 2% 8% 13% 73% 0% 78 (4%) 5 and up 5% 1% 1% 2% 5% 86% 523 (25%) Share of Total 1,195 (57%) 109 (5%) 128 (6%) 106 (5%) 95 (5%) 475 (23%) 2,108 (100%) *Source: survey Question: Q4_1, Q4-2 **Not all interior rows sum to 100% due to rounding 101 The significant share of respondents who worked in person more days than they needed may reflect a preference for in-person work or tacit understanding within their workplace that the employers favor in-person presence. Although most workers anticipate maintaining their current work arrangement, Table 4.1.4 shows that hybrid workers are more likely to report anticipating working in-person within a year when compared to remote workers. Among hybrid workers, 20% indicate they expect to reduce or eliminate remote work for the coming year, whereas for remote workers, the corresponding share is only 10%. There is little uncertainty across all work arrangements. Only 11% of respondents stated that that they were unsure how their work arrangement would change in the future. Four out of five respondents working full-time remotely or in person stated that they expected to maintain this work arrangement one year later. Table 4.1.4. Anticipated future work arrangement Future work arrangement Current remote working status Total Remote Hybrid In-person Same amount of remote work 79% 63% 3% 33% Begin or increase remote work 1% 7% 3% 4% Reduce remote work 4% 13% 1% 4% No remote work 6% 7% 80% 48% Unsure / Don't know 9% 9% 13% 11% Total 478 (100%) 440 (100%) 1,190 (100%) 2,108 (100%) *Source: survey Question: Q4_2, Q9 102 4.2 Remote Workers’ Demographics and Vehicle Type Table 4.1.3 indicated that the ability to work remotely and its actual implementation may vary. This could be influenced by factors such as individual and employer characteristics. The ability to work remotely depends largely on the type of work people do, which is, in turn, highly correlated with people’s demographic characteristics (Brynjolfsson et al., 2020; Wang et al., 2023). Understanding this association is crucial for addressing the intertwined relationship between remote working and migration, commuting, and environmental impacts. Demographics also affect the types of vehicle people drive, another critical factor in understanding commuting behavior and its environmental impacts. This section examines how respondents’ demographic characteristics affect their likelihood of working remotely and the kind of vehicles they use when commuting by car. Remote work Ability My survey asked respondents two separate questions: (1) whether their employer allowed them to work remotely and also (2) whether they actually worked remotely. Previous research has shown, however, that the ability to work remotely was greatly influenced by the type of industry in which one works, the role on the job, and the demographics of those doing the job (Brynjolfsson et al., 2020; Wang et al., 2023) and Table 4.1.3 showed that ability to work remotely is not the same as actually working remotely, especially when examining the number of days working in person. I use a statistical model to create an index that summarizes respondents’ ability to work remotely as a way to condense many variables into a single value that represents the how likely someone is to be able to work remotely. I do so in two steps. First, I condense the 25 industry groupings provided by IPSOS into a categorical variable – industry potential - of whether the industry has positive, negative, or neutral impact on the ability to work remotely (see equation 1 and table 4.2.1 below). Second, I use a multivariate logit model to examine the dependent variable, which is from survey question 1: Are you currently able to work remotely? (Yes, any days=1, no=0), with independent variables including industry potential and worker demographics (also provided by IPSOS). I then obtain the predicted value conditional on the independent variables for predicted remote work ability. These predicted values of remote work ability serve as the control variable for all subsequent regression analyses in the remainder of this chapter. The effect of occupation type on the potential to work remotely has been documented in prior literature (Dingel and Neiman, 2020). The IPSOS respondent demographic information lists 35 industry types. Given this variable’s importance and to make the analysis more tractable, we estimate the impact of industry type on the self-reported ability to work remotely using the regression model in equation 1. 103 Equation 1: 𝑅𝑒𝑚𝑜𝑡𝑒 𝑤𝑜𝑟𝑘 𝑎𝑏𝑖𝑙𝑖𝑡𝑦 (1,0) = 𝑓(𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑡𝑦𝑝𝑒) The definition of each variable is listed below: • Remote work ability: Yes=1, No=0 • Industry type, 35 dummy variables, each 0,1 for working in these industries: Management (omitted); Business and Financial Operations; Computer and Mathematical; Architecture and Engineering; Life, Physical, and Social Sciences; Community and Social Services; Legal; Teacher, except college and university; Teacher, college and university; Other professional; Medical Doctor (such as physician, surgeon, dentist, veterinarian); Other Health Care Practitioner (such as nurse, pharmacist, chiropractor, dietician); Health Technologist or Technician (such as paramedic, lab technician); Health Care Support (such as nursing aide, orderly, dental assistant); Protective Service (such as firefighter, law enforcement worker); Food Preparation and Serving; Building and Grounds Cleaning and Maintenance; Personal Care and Service; Sales Representative; Retail Sales; Other Sales; Office and Administrative Support; Farming, Forestry, and Fishing; Construction and Extraction; Installation, Maintenance, and Repair; Precision Production (such as machinist, welder, baker, printer, tailor); Transportation and Material Moving; Armed Services; Other (Please specify); Business Operations (including Marketing); Financial Operations or Financial Services (including Financial Advisor, Broker); Education, Training, and Library; Arts, Design, Entertainment, Sports, and Media; Health Diagnosing or Treating Practitioner (such as physician, nurse, dentist, veterinarian, pharmacist); Sales Table 4.2.1 provides results of this regression (equation 1) and is organized based on the significance level, sign, and magnitude of the coefficient for each occupation. Table 4.2.1 shows the regression coefficients, sorted into groups of positive and statistically significant, insignificant, and negative and statistically significant. I use those groupings to form the categories of positive, neutral, and negative industry potential for remote work. Consistent with previous research (e.g., Bartik et al. 2020), my analysis confirms that workers in office and desk-based roles are associated with a higher likelihood of being able to work remotely, while those in service-related jobs and manual labor are more likely to have a negative association with remote work. Using these findings, I reclassify the detailed industries into three groups predicting remote work ability and call this new variable industry potential: those with positive potential, those with negative potential, and those with neutral potential for remote work. 104 Table 4.2.1. Regression results between remote work ability and industry types (Dependent variable: Remote work ability) Industry potential for remote work Detail occupation Coef. Positive potential Computer and Mathematical 0.355*** Business Operations (including Marketing) 0.343*** Legal 0.249** Architecture and Engineering 0.246*** Life, Physical, and Social Sciences 0.17** Financial Operations or Financial Services 0.17** Arts, Design, Entertainment, Sports, and Media 0.14* Neutral potential Sales 0.075 Armed Services 0.047 Office and Administrative Support 0.036 Community and Social Services -0.02 Negative potential Other (Please specify) -0.087* Education, Training, and Library -0.17** Building and Grounds Cleaning and Maintenance -0.197** Health Care Support -0.214*** Personal Care and Service -0.252*** Food Preparation and Serving -0.291*** Health Diagnosing or Treating Practitioner -0.291*** Health Technologist or Technician -0.305*** Transportation and Material Moving -0.328*** Installation, Maintenance, and Repair -0.384*** Precision Production -0.42*** Construction and Extraction -0.461*** Protective Service -0.484*** Farming, Forestry, and Fishing -0.523*** _cons 0.523 * p<0.05 ** p<0.01 *** p<0.001 The second step in creating an index for remote work ability is to model a composite variable – predicted remote work ability – that amalgamates demographic factors and industry potential into a single numerical value representing the ability to work remotely. I do this by regressing individual survey respondents’ answers to the question about whether they can work remotely (survey question 1: Are you currently able to work remotely? Yes, any days=1, No=0) on a set of demographic characteristics and industry potential variable from Table 4.2.1. I then run the logit regression shown in Equation 2 and get predicted values of remote work ability for each survey respondent based on their respective demographic characteristics, which I will use as a control variable in later regression analyses. 105 Equation 2: 𝑅𝑒𝑚𝑜𝑡𝑒 𝑤𝑜𝑟𝑘 𝑎𝑏𝑖𝑙𝑖𝑡𝑦 = 𝑓 ( 𝑖𝑛𝑐𝑜𝑚𝑒,𝑒𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛, ℎ𝑜𝑢𝑠𝑖𝑛𝑔 𝑡𝑒𝑛𝑢𝑟𝑒, 𝑎𝑔𝑒, 𝑔𝑒𝑛𝑑𝑒𝑟, 𝑟𝑎𝑐𝑒 𝑜𝑟 𝑒𝑡ℎ𝑛𝑖𝑐𝑖𝑡𝑦, 𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑝𝑜𝑡𝑒𝑛𝑡𝑖𝑎𝑙 𝑓𝑜𝑟 𝑟𝑒𝑚𝑜𝑡𝑒 𝑤𝑜𝑟𝑘) The definition of each variable is listed below: • Remote work ability: Yes=1, No=0 (omitted) • Income dummy variables (0,1) in three categories for: Less than $50,000 (omitted), $50,000 to $99,999, $100,000 or more • Education dummy variables (0,1) in three categories for: High school diploma or below (omitted), bachelor’s degree or Some college or Associate's degree, and Master’s degree or higher • Housing tenure dummy variables (0,1): Owned or being bought by you or someone in your household (omitted) and rented for cash or occupied without payment of cash rent • Age dummy variables (0,1) in four categories: 18-29 (omitted), 30-44, 45-59, and 60+ • Gender dummy variables (0,1): Male (omitted), female • Race/ethnicity dummy variables (0,1) in five categories: White (omitted), Black, Hispanic, Other, and 2+ Races • Industry potential dummy variables (0,1) in three categories: Neutral remote work potential (omitted), positive remote work potential, negative remote work potential Table 4.2.2 shows the result of the remote work ability estimation in equation 2 that combines the demographic variables with the industry potential variable derived above in table 4.2.1. I found that education level and industry potential have the most significant impact on remote work ability. Remote workers tend to have higher education levels and work in professional industries that typically involve more office and desk work. Those with graduate degrees in particular had higher ability to work remote. Renting one’s home was also slightly more correlated with the ability to work remotely compared to owning a home. Several demographic variables that the literature and/or popular perception previously considered important for remote work were not statistically significant in this model: notably, income, age, race/ethnicity, and gender had no correlation with the ability to work remotely, likely because those factors are working through the industry potential. Similarly, estimating the model (Equation 2) with a control for geographic region did not materially change the results. A likely explanation for the lack of significant coefficients associated with demographic factors is that they were already accounted for by the industry potential variable. However, I still believe these variables should be included in the model due to their theoretical value in understanding remote work dynamics. The Pearson chi-squared goodness-of-fit test suggests that the logistic regression model fits well when focusing on the significant predictors (Industry potential for remote work, education, and age), with a p-value of 0.0920, which is greater than 0.05. This indicates that 106 there is no significant evidence to reject the null hypothesis, meaning the model adequately fits the data. However, it's important to include additional variables like income, housing tenure, gender, and ethnicity because they are essential for understanding remote work dynamics. While including these variables reduce the statistical goodness-of-fit, the literature supports their inclusion, as they likely capture aspects of remote work ability that aren’t directly linked to the significant predictors but are still critical for a comprehensive understanding of the topic. In summary, while the simpler model shows good statistical fit, the full model with the additional variables provides a more complete and theoretically grounded explanation of the factors influencing remote work ability. This approach ensures that the model not only fits the data but also aligns with the broader theoretical framework of the research. For subsequent regression analyses in this report, I use the predicted value of remote work ability from this statistical model. The predicted value represents a continuous measure incorporating demographic and industry characteristics, in contrast to the binary indicator solely based on whether people responded to the survey question saying they can remote work. Table 4.2.2. Regression Results of remote work ability Dependent Variable Are you currently able to work remotely (1,0) Category Variable (1) Income (Omitted: <$50,000) $50,000 to $99,999 -0.158 > $100,000 0.337 Education (Omitted: High school diploma or below) Bachelor's / some college / Associate's 0.431** Master's or higher 0.773*** Housing tenure (Omitted: Owned) Rented 0.320* Age (Omitted: 18-29) 30-44 0.313 45-59 0.235 60+ 0.216 Gender (Omitted: Male) Female 0.073 Ethnicity (Omitted: White) Black 0.159 Other 0.383 Hispanic 0.387* 2+ Races 0.099 Industry potential for remote work (Omitted: no significant potential) Negative potential -1.126*** Positive potential 1.077*** Constant -0.779** Observations 2120 * p<0.05 ** p<0.01 *** p<0.001 107 Vehicle choice and work arrangement This project primarily examines how remote work influences commuting behavior and, consequently, how that affects greenhouse gas (GHG) emissions. A key component of greenhouse gas emissions from commuting is the type of vehicle used by the commuter. Like remote work ability, vehicle choice is often associated with socioeconomic status. The survey asked respondents to report the make, model, and year of their primary commute vehicle if they commuted by car. For those who commuted primarily by another mode (e.g., public transit), I asked which vehicle was most commonly available to them in their household, if they were to commute by car. About three quarters of respondents (1,629) commuted by car or owned a car and reported information on their vehicle. I identify typical greenhouse gas (GHG) emissions for each vehicle by cross-referencing respondents’ selfreported vehicle details (vehicle's make, model, and year) with the associated CO₂ tailpipe emissions (measured in grams per mile, gpm) reported in EPA's Fuel Economy Guide database for 2023.16 Table 4.2.3 reports the mean and quartiles of GHG emissions in CO₂ tailpipe grams per mile (gpm). There are no discernable differences between remote and in-person workers in terms of average vehicle efficiency choices. The main difference between the two categories is in the upper end (90th percentile) of the distribution (the top 10% of most polluting vehicles), where in-person workers tend to drive higher-emissions vehicles. Hybrid workers stand out as driving lower-emissions vehicles, about 8% lower on average, a difference that holds across the distribution. Table 4.2.3. GHG emission of owned vehicles by work arrangement Work arrangement CO₂ tailpipe (gpm) mean 25% 50% 75% 90% Remote 394 324 389 465 509 Hybrid 365 309 359 433 481 In-person 393 318 384 462 543 * Vehicle CO₂ tailpipe (gpm) source: EPA Fuel Economy Guide 2023, https://www.fueleconomy.gov/feg/download.shtml *Source: survey Question: Q4_2, Q7_2 For purposes of descriptive data analysis, I reclassified vehicle GHG emissions into three levels: high, medium, and low to better understand the relationship between workers and their vehicle. I use the 2024 GHG rating provided by the EPA (refer to Appendix C1 for classification) to categorize each vehicle according to its actual CO₂ tailpipe emissions obtained from EPA’s Greenhouse Gas Rating.17 Based on the result presented in table 4.2.4, compared to the overall distribution, I found that: Hybrid workers have the highest share (10%) of low GHG emission vehicles and lowest share of high GHG emission vehicles (7%). Fully in-person 16 Data is available for download from the EPA here: https://www.fueleconomy.gov/feg/download.shtml 17 Data can be downloaded here: https://www.epa.gov/greenvehicles/greenhouse-gas-rating 108 workers have the highest share (14%) of high GHG emission vehicles and lowest share (5%) of low GHG emission vehicles. Fully remote workers have a distribution of high, medium, and low emission vehicles similar to the overall distribution. Table 4.2.4. GHG emission ranking of owned vehicles by work arrangement My GHG Category EPA Rating CO₂ (g/mile) Remote Hybrid In-person Total High GHG emission 1-3 > 509 11% 7% 14% 12% Medium GHG emission 4-6 266-508 82% 83% 81% 82% Low GHG emission 7-10 0-265 7% 10% 5% 7% Total 126 (100%) 404 (100%) 1,099 (100%) 1,629 (100%) *Please refer to Appendix C1 for GHG emission category and detail EPA rating. Data can be downloaded here: https://www.epa.gov/greenvehicles/greenhouse-gas-rating *Source: survey Question: Q4_2, Q7_2 Considering the demographics of remote workers, which typically include individuals with higher income, higher education levels, and employment in professional industries that offer more flexibility for remote work, I found that workers who work remotely are more likely to own lower emission vehicles, while in-person workers are more likely to own higher emission vehicles. In other words, high-emission vehicle owners may be more likely to commute, possibly because they tend to have lower incomes and work in industries that require in-person attendance. Additionally, lower incomes may create barriers to upgrading to newer, loweremission vehicles. Conversely, lower-emission vehicle owners generally have higher incomes, work in industries with more flexibility for remote work, and can afford to purchase newer, more environmentally friendly vehicles. I use the following equations (3-6) to test the hypotheses. Equations 3 and 4 use Ability to remote work (from survey question 1: are you currently able to work remotely?) as the dependent variable whereas equations 5-6 use the number of actual reported commute days as an alternate measurement. Equation 3: 𝐴𝑏𝑙𝑒 𝑡𝑜 𝑟𝑒𝑚𝑜𝑡𝑒 𝑤𝑜𝑟𝑘 (1,0) = 𝑓(𝑣𝑒ℎ𝑖𝑐𝑙𝑒 𝐺𝐻𝐺 𝑟𝑎𝑛𝑘𝑖𝑛𝑔) Equation 4: 𝐴𝑏𝑙𝑒 𝑡𝑜 𝑟𝑒𝑚𝑜𝑡𝑒 𝑤𝑜𝑟𝑘 (1,0) = 𝑓(𝑣𝑒ℎ𝑖𝑐𝑙𝑒 𝐺𝐻𝐺 𝑟𝑎𝑛𝑘𝑖𝑛𝑔, 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠) Equation 5: 𝐶𝑜𝑚𝑚𝑢𝑡𝑒 𝐷𝑎𝑦𝑠2023 = 𝑓(𝑣𝑒ℎ𝑖𝑐𝑙𝑒 𝐺𝐻𝐺 𝑟𝑎𝑛𝑘𝑖𝑛𝑔) Equation 6: 𝐶𝑜𝑚𝑚𝑢𝑡𝑒 𝐷𝑎𝑦𝑠2023 = 𝑓(𝑣𝑒ℎ𝑖𝑐𝑙𝑒 𝐺𝐻𝐺 𝑟𝑎𝑛𝑘𝑖𝑛𝑔, 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠) The definition of each variable is listed below: • Able to remote work: Yes=1, No=0 (omitted) 109 • Vehicle GHG Ranking= Low (0-265 CO₂ g/mile), medium (266-508 CO₂ g/mile) (omitted), high (>508 CO₂ g/mile) • Commute days 2023 = Days of commute per week in 2023 • Controls: predicted remote work ability, regions (Northeast (omitted), Midwest, South, West) The regression results (table 4.2.5) from equation 3 show that when compared to workers owning medium GHG emission vehicles, those with low GHG emission vehicles are more likely (odds ratio +0.55) to remote work. In Equation 5, the results show that in comparison to workers with medium GHG emission vehicles, those with high GHG emission vehicles tend to have 0.29 more in-person workdays, while those with low GHG emission vehicles have 0.41 fewer inperson workdays. Models 4 and 6 include the controls (predicted remote work ability and region). Vehicle efficiency no longer achieves statistical significance with the addition of predicted remote work ability, implying a strong correlation between socioeconomic status and vehicle efficiency level. The stability in sign and magnitude of the efficiency variables point to these relationships likely being in the right direction. Table 4.2.5. Regression results of vehicle GHG emissions Dependent Variable Able to remote work (1,0) Able to remote work (1,0) Commute days 2023 Commute days 2023 Category Variable (3) (4) (5) (6) GHG emission rank (Omitted: Medium GHG emission) High GHG emission 0.000 0.000 0.293* 0.136 Low GHG emission 0.546* 0.336 -0.414* -0.241 Predicted remote work ability 0.975*** -0.500*** Region (Omitted: Northeast) Midwest -0.177 0.054 South -0.046 0.144 West 0.164 -0.027 Constant -0.393*** -0.349* 4.012 3.900*** Observations 1630 1627 1629 1626 R-squared 0.008 0.116 Adjusted R-squared 0.006 0.112 * p<0.05 ** p<0.01 *** p<0.001 4.3 Migration and Remote Work Migration was frequently in the news as the COVID-19 pandemic unfurled, especially as it pertained to the increase in flexible work arrangements. The results in Table 4.3.1 suggests that a little over a quarter of respondents migrated (changed their home location) in the period between March 2020 and September 2023. Among those who moved, 70% of respondents 110 moved once, 23% moved twice, and only 7% moved more than twice. Hybrid and remote workers were more likely to move than in-person workers. According to the Current Population Survey, about 8-9% of Americans moved annually (CPS, 2023), while the Public Use Microdata Sample (PUMS) reports a 14% annual move rate during this time period (Kerns D’Amore 2023). Once I divide the 26% of respondents who reported moving between March 2020 and September 2023 by the 3.5 years covered by the time period covered in the survey question, yield a 7.4% move rate. This estimate is in line in line with the national average, especially considering that some moved more than once. Table 4.3.1. Moving status by current work arrangement Moving status since COVID-19 began (March 2020) Remote Hybrid In-person Total Moved 29% 30% 23% 26% Did not move 71% 70% 77% 74% Total 478 (100%) 440 (100%) 1,196 (100%) 2,114 (100%) *Source: survey Question: Q2, Q2_1, Q4_2 Respondents who changed their work arrangement from pre-COVID to post-COVID were more likely to move (Table 4.3.2). Previously remote employees who switched to in-person work were almost twice as likely to move as those who stayed remote. While relatively uncommon, hybrid workers who went remote and in-person workers who went remote or hybrid were much more likely to move (nearly double) than their counterparts with less flexible arrangements. Table 4.3.2. Moving status by change in work arrangement from pre-COVID to postCOVID Change in work arrangement Moved Did not move Total Pre-COVID Post-COVID # Share Remote Remote 20% 80% 189 9% Hybrid 43% 57% 31 1% In-person 41% 59% 24 1% Hybrid Remote 33% 67% 89 4% Hybrid 25% 75% 136 7% In-person 37% 63% 28 1% In-person Remote 37% 63% 192 9% Hybrid 31% 69% 270 13% In-person 23% 77% 1,139 54% Share of Total 26% 74% 2,097 100% *Source: survey Question: Q2, Q2_1, Q4_2, Q6_2 111 The survey asked respondents to report current home and work zip codes and previous home and work zip codes. I then measured the distance between the center of their home zip code before and after moving to approximate move distance. Table 4.3.3. shows that the median in-person and hybrid mover moved a distance of about 10 miles, while the median remote mover moved more than 22 miles. Most movers did not move far, but 1 in 10 hybrid movers moved at least 698 miles, and 10% of remote movers moved over 1200 miles. Generally, at least 25% of hybrid and remote movers moved 191 and 364 miles away, which is well beyond their current metropolitan area; and in many cases likely across state lines. This suggests a much larger set of move locations from which to choose compared to in-person workers and makes commuting on a regular basis and by car an unlikely option for a significant share of workers. Table 4.3.3. Moving distance (mile) by current work arrangement Current work arrangement Mean 25% 50% 75% 90% Remote 330 4.98 22.3 364.22 1205.27 Hybrid 207 3.16 10.16 191.71 698.06 In-person 166 0 9.56 46.53 667.22 *Source: survey Question: Q1a, Q2a, Q4_2 Table 4.3.4 shows the results of respondents’ likelihood of moving in the next year. 13% indicated that they anticipated moving in the next year, slightly above the average annual move rate of 8-9% for this period. Table 4.3.5 shows the result of moving plan by work arrangement and house ownership. Renters were more likely to anticipate moving than homeowners, in line with national trends between these two groups. In-person workers were least likely to anticipate moving, followed by remote workers. Hybrid renters were the group most likely to anticipate moving. Table 4.3.4. Future moving plan by work arrangement Anticipated Future Move Status Remote Hybrid In-person Share of Total Move 15% 13% 13% 280 (13%) No move 85% 87% 87% 1,828 (87%) Share of Total 478 (100%) 440 (100%) 1,190 (100%) 2,108 (100%) *Source: survey Question: Q2, Q4_2 Table 4.3.5. Future moving plan by work arrangement and housing tenure Anticipated Future Move Status Remote Hybrid In-person Share of Owner Renter Owner Renter Owner Renter Total Move 9% 31% 7% 37% 7% 26% 13% No move 91% 69% 93% 63% 93% 74% 87% Total 331 (100%) 148 (100%) 338 (100%) 97 (100%) 813 (100%) 381 (100%) 2,108 (100%) *Source: survey Question: Q2, Q4_2 112 4.4 Distance between work and home Using current and past home and job locations, I calculated a measure for respondents’ commute distance based on the distance between the zip codes in which people reside and work. For people living and working in the same zip code, that distance is zero (see Chapter 3). Remote workers do not commute, but their distance to where they would otherwise go to work is indicative of their decision regarding where to live. The increase in average distance points to such a separation of residential and work location (Table 4.4.1 and Table 4.4.2). Table 4.4.1 presents distance statistics between home and job for all survey respondents. In Table 4.4.2, results are shown after excluding the outliers— non-remote workers (hybrid or in-person) with distances exceeding 300 miles between their home and job. These outliers represent the top 1 percentile. The lack of change in the distance distribution at different percentiles compared to the large increase in mean suggest that a small subset of hybrid and remote workers moved to locations very far from their employers. These outliers may involve in-person and hybrid workers who commute by air to another state once a week, then travel to their job by car on a daily basis. Therefore, either the distance or the number of people living far increased enough in that subset to skew the average. The percentile presented in Tables 4.4.1 and 4.4.2 refers to the percentile of different groups of workers, either before or after the COVID-19 pandemic. I use results that exclude outliers from Table 4.4.2 in the following explanation. Greater distances would be consistent with a view of remote workers as completely spatially divorced from where they work. In stark contrast to this view, however, most remote workers live and work in the same zip code. It's important to remember that the groups of remote, hybrid, and in-person workers pre-COVID and post-COVID are not identical. For the quarter of remote workers who lived the farthest from their place of work, persons increased the distance to their work location by 9 miles relative to pre-Covid from 9.9 to 18.9 miles; the 10% of remote workers who live the farthest increased their home-to-work distance by 1.5 times, suggesting a trend where post-COVID remote workers are more inclined to relocate to locations farther away from their jobs compared to their pre-COVID counterparts. Post-COVID Hybrid and in-person workers had lower magnitude changes in home-to-job distance relative to remote workers. Median distance between home and job increased by 1.1 miles for hybrid workers and decreased by 0.2 miles for in-person workers (Table 4.4.2). However, hybrid workers at the 90th, 95th, and 99th percentile of pre-Covid distance increased their home-to-job distance (which I as shorthand I call commute distance) by 1.8, 7, and 80 miles respectively. In-person workers did not have the same effect: in fact, the 99th percentile of inperson workers decreased their home-to-job distance (i.e., commute) by 60 miles. Given the difference in composition of workers across work arrangements pre- and post-COVID, one possible explanation for the large decrease could be that many in-person workers residing far from their workplace switched to some form of remote work. 113 Table 4.4.1. Home-to-job distance (miles) by work arrangements Percentiles Remote Hybrid In-person PreCOVID PostCOVID PreCOVID PostCOVID Pre-COVID Post-COVID mean 58.35 100.58 36.40 24.81 25.64 21.68 25% 0 0 0 3.8 0 0 50% 0 0 7.67 8.58 6.45 6.14 75% 9.933 18.84 16.85 16.98 14.26 14.01 90% 74.97 197.68 27.36 30.14 26.47 23.75 95% 296.53 548.30 49.14 59.77 46.11 33.32 99% 1440.14 2020.84 1137.88 500.19 609.32 552.07 Total workers 208 456 239 413 1,477 1,084 *Source: survey Question: Q1a, Q2a, Q4_2 Sample size: pre-Covid (1,924), post-Covid (1,953) Table 4.4.2. Home-to-job distance (miles) by work arrangements excluding outliers (nonremote workers with home-to-job distance >= 300 miles) Percentiles Remote Hybrid In-person PreCOVID PostCOVID PreCOVID PostCOVID Pre-COVID PostCOVID mean 58.35 100.58 12.43 14.78 10.93 9.46 25% 0 0 0 3.73 0 0 50% 0 0 7.41 8.54 6.21 6.07 75% 9.933 18.84 15.34 16.70 13.59 13.59 90% 74.97 197.68 25.36 27.15 23.69 22.75 95% 296.53 548.30 33.72 40.71 35.47 31.21 99% 1440.14 2020.84 105.32 185.31 111.46 58.8 Total workers 208 456 233 406 1,432 1,067 *Source: survey Question: Q1a, Q2a, Q4_2 Sample size: pre-Covid (1,873), post-Covid (1,929) I next conduct a regression analysis to statistically test the association between work arrangement and the change in distance between home and job, controlling also for the predicted remote work ability and geographic region (Equation 7). I hypothesize that, compared to preCOVID, remote and hybrid workers are more likely to live farther away from their workplace post-COVID. The change in distance could come from a change in residence (a move) a change in job, or a change in the location of employment. The test focuses on any change to reflect people’s choice with regards to their location, whether in-person workers are more likely to live closer to the job after the pandemic than remote and hybrid workers. 114 Equation 7: ∆ 𝐻𝑜𝑚𝑒𝐽𝑜𝑏 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 = 𝑓(𝑤𝑜𝑟𝑘 𝑎𝑟𝑟𝑎𝑛𝑔𝑒𝑚𝑒𝑛𝑡𝑠2023, 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠 ) The definition of each variable is listed below: • ∆ HomeJob Distance= Change in distance between home and workplace from preCOVID (before March 2020) to post-COVID (September, 2023) • Work arrangements 2023 = Remote, Hybrid, in-person (omitted) • Controls: predicted remote work ability, regions (Northeast (omitted), Midwest, South, West) I conduct another regression analysis (equation 8) to test whether more commuting days are associated with shorter distances between workers’ home and job locations to further delve into differences between different hybrid work arrangements. I use the following equations to test the hypothesis: Equation 8: 𝐻𝑜𝑚𝑒𝐽𝑜𝑏 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒2023 = 𝑓(𝑑𝑎𝑦𝑠 𝑜𝑓 𝑐𝑜𝑚𝑚𝑢𝑡𝑒2023, 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠) The definition of each variable is listed below: • HomeJob Distance 2023 = Distance between home and workplace in post-COVID (September, 2023) • Commute days 2023 = Days of commute per week in 2023 • Controls: predicted remote work ability, regions (Northeast (omitted), Midwest, South, West) Table 4.4.3 shows the regression results for both models. Model 7 indicates that postCOVID remote workers increased the distance from their job by 29 more miles compared to Post-COVID in-person workers. The hybrid worker coefficient was not statistically significant. The hybrid category includes people who commute nearly every day and those who commute once a week, which may muddle the results. Model 8 shows that an additional day of in-person work (commute) is associated with a reduction (-12 miles) in the distance between their home and the workplace post-COVID. This result helps explain why the difference in Model 7 between in-person and hybrid workers was not significant. Workers’ trade-off greater distances from their job for lower numbers of days they must commute. 115 Table 4.4.3. Regression results of home-job distance Dependent Variable Change in HomeJob Distance Post-COVID HomeJob Distance (2023) Category Variable (7) (8) Work arrangement 2023 (Omitted: in-person) Remote 28.984* Hybrid -19.577 Commute days per week in 2023 -12.059*** Predicted remote work ability 6.153 7.902 Region (Omitted: Northeast) Midwest 17.947 22.623 South 6.391 32.729* West 17.317 41.644** Constant -4.531 53.429*** Observations 1901 1950 R-squared 0.007 0.027 Adjusted R-squared 0.003 0.024 * p<0.05 ** p<0.01 *** p<0.001 Among respondents who commuted, the vast majority (~85%) drove alone. Walking / biking was the next most common, followed by transit, and then carpooling (Table 4.4.4). These shares are roughly in line with national statistics from ACS 2022 5-year estimates (U.S. Census Bureau, 2022); the survey has fewer carpoolers but slightly more walkers / bikers and those who drive alone. When asked whether they switched commute mode, including remote work options, compared to pre-COVID, 20% of respondents indicated that they changed commute modes (Appendix Table C4). The most common switch was from driving alone to working remote (1/3 of all mode switchers). There was little difference in commute mode and change thereof by work arrangements (omitted here for brevity, please refer to Appendix C6 for detail). Table 4.4.4. Commute mode from pre-COVID to post-COVID commute mode Pre-COVID Post-COVID Walking / biking 5% 4% Car, single occupant (only yourself) 84% 87% Car, multiple occupants (a carpool) 4% 4% Ride share (Uber, Lyft), taxi, or vanpool 1% 1% Bus 2% 2% Train 3% 2% Other (Please specify): 1% 1% Total 1,738 (100%) 1,636 (100%) *The difference in the total is caused by non-responses in the survey. *Note, table 4.4.4 excludes fully remote workers since they do not have a commute mode *Source: survey Question: Q7, Q7-1_1 116 4.5 Environmental impact of Remote work In this section, I look at how the changes in commute distance outlined in the previous section (daily and on a weekly basis, i.e., the sum of all commutes both ways) affected greenhouse gas emissions associated with each work arrangement. I analyze driving and related GHG emissions for hybrid in-person/remote workers compared with driving and GHG emissions for persons who work in-person full time. I found a 20% reduction in weekday commute trips (Table 4.5.1, for the aggregated, weighted data) compared to pre-COVID levels, along with a 3% decrease in weekly greenhouse gas (GHG) emissions (Figure 4.5.1). This suggests that much of the decrease in commute trips was offset by longer commutes. Moreover, there is nearly a doubling in the number of remote and hybrid workers post-COVID, and hybrid workers in post-COVID are more likely to remote work for more days than pre-COVID hybrid workers (Appendix C9). The greatest reduction in overall commute GHG emissions comes from the larger number of people who work full-time remotely and, therefore, do not commute. Hybrid workers show the least reduction in GHG emissions, likely due to living farther away from their workplace (see Table 4.4.3) and generating longer one-way commutes on the days they do commute. In-person workers' commute GHG emissions remained stable from pre-COVID to post-COVID periods. Table 4.5.1. Total daily commute trips, weighted and aggregated daily for pre- and postCOVID Time Period Monday Tuesday Wednesday Thursday Friday Total Pre-COVID 3,595 3,624 3,607 3,612 3,542 17,980 Post-COVID 2,899 3,036 3,004 3,021 2,795 14,755 Difference -19% -16% -17% -16% -21% -18% *Source: survey Question: Q4_2_1, Q6_2_1, Q7, Q7_1_1 GHG emissions typically refer to the total emissions of all greenhouse gases. However, according to the EPA18, when referring to vehicle GHG emissions, most cases focus primarily on CO₂ emissions, as CO₂ is the most prevalent greenhouse gas emitted by vehicles and is often used as a proxy for overall GHG emissions from vehicles. I calculate daily round-trip commute GHG emissions per vehicle using equation 9: Equation 9: 𝐷𝑎𝑖𝑙𝑦 𝑐𝑜𝑚𝑚𝑢𝑡𝑒 𝐶𝑂2 𝑒𝑚𝑖𝑠𝑠𝑖𝑜𝑛 𝑝𝑒𝑟 𝑣𝑒ℎ𝑖𝑐𝑙𝑒 = 𝐶𝑂2 𝑡𝑎𝑖𝑙𝑝𝑖𝑝𝑒 𝑔𝑟𝑎𝑚𝑠 𝑚𝑖𝑙𝑒 ∗ 𝐻𝑜𝑚𝑒𝐽𝑜𝑏 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ∗ 2 18 Vehicle Greenhouse Gas Emissions: https://www.epa.gov/ghgemissions/sources-greenhouse-gas-emissions 117 Daily commute and GHG emissions I use Equation 9 to estimate the daily commute CO₂ emission in grams for each individual in the sample, both pre- and post-COVID. Then, I calculate the aggregate daily greenhouse gas (GHG) emissions for the entire sample based on respondents' information about which day of the week that they usually commute. The results in Table 4.5.1 show the largest reduction in total commute traffic occurred on Friday (-21%) and Monday (-19%), while the smallest reduction was observed on Tuesday (- 16%), Wednesday (-17%), and Thursday (-16%). Total commute GHG emissions saw the most significant reductions on Monday (-7%) and Friday (-9%), followed by Thursday (-3%), compared to the pre-COVID period. Figure 4.5.1. Total daily commute GHG Emissions (CO₂ grams per mile) by day of week The survey findings suggest that hybrid workers are more likely to commute on Tuesday, Wednesday, and Thursday, shedding some light on the post-COVID results for those days in Figure 4.5.1. Despite the overall lower weekly commute greenhouse gas (GHG) emissions across all work arrangements, hybrid workers living farther away from their workplace may cause longer one-way commutes, potentially leading to higher daily total GHG emissions on these midweek days. Monday Tuesday Wednesday Thursday Friday Pre-COVID 22,644,348 22,590,235 22,704,322 22,657,406 22,351,550 Post-COVID 21,074,570 23,294,124 23,014,015 21,891,605 20,336,273 18,500,000 19,000,000 19,500,000 20,000,000 20,500,000 21,000,000 21,500,000 22,000,000 22,500,000 23,000,000 23,500,000 24,000,000 Aggregated total Daily Co2 Emissions (grams per day) Pre-COVID Post-COVID 118 Personal weekly commute GHG emissions by work arrangement After considering the broader context at the aggregated level, I next focus on weekly CO₂ emissions per person. I compute weekly CO₂ emissions per person by considering the number of car occupants and remote working days, using the following equation. Equation 10: 𝑊𝑒𝑒𝑘𝑙𝑦 𝐶𝑂2 𝑒𝑚𝑖𝑠𝑠𝑖𝑜𝑛 𝑝𝑒𝑟 𝑝𝑒𝑟𝑠𝑜𝑛 = ( 𝐶𝑂2 𝑡𝑎𝑖𝑙𝑝𝑖𝑝𝑒 𝑔𝑝𝑚 ∗ ℎ𝑜𝑚𝑒𝑗𝑜𝑏 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ∗ 2 ∗ 𝑐𝑜𝑚𝑚𝑢𝑡𝑒 𝑑𝑎𝑦𝑠) 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑣𝑒ℎ𝑖𝑐𝑙𝑒 𝑜𝑐𝑐𝑢𝑝𝑎𝑛𝑡𝑠 Table 4.5.2 shows total weekly vehicle miles traveled (VMT) by work arrangement. It's important to keep in mind that the groups of remote, hybrid, and in-person workers pre-COVID and post-COVID are not the same. I found that the weekly VMT of in-person workers is higher than hybrid workers, which is due in part to in-person workers’ additional days of commuting. However, when comparing the change in weekly VMT from pre-COVID to post-COVID periods, I found that both groups increased mean VMT. However, this was driven largely by VMT increases in the 99th percentile of commuters by VMT. The median in-person worker had no VMT increase. The median hybrid workers increased VMT by about 8 miles. Table 4.5.2. Total weekly commute VMT by work arrangement, in total weekly miles, excluding outliers (non-remote workers with home-to-job distance >= 300 miles)19 Work arrangement Remote Hybrid In-person PreCOVID PostCOVID PreCOVID PostCOVID PreCOVID PostCOVID Mean 0 0 73.64 129.6 110.9 164.8 25% 0 0 13.1 14.8 0 0 50% 0 0 37.0 44.7 65.4 65.5 75% 0 0 101.2 92.5 141.5 142.6 90% 0 0 180.2 182.8 233.0 234.1 95% 0 0 288.7 288.6 337.1 328.4 99% 0 0 436.1 2,000.8 1000.4 1120.5 Total workers 226 468 152 341 1,191 981 *Source: survey Question: Q1a, Q2a, Q4_2 Sample size: pre-Covid (1,569), post-Covid (1,790) Table 4.5.3 presents the weekly commute greenhouse gas (GHG) emissions by work arrangement during the pre- and post-COVID periods. GHG emissions are the combination of VMT and vehicle-specific emissions. For this analysis, I assume that people did not change 19 See Appendix Table D9 for total weekly commute VMT by work arrangement without excluding outliers 119 vehicles pre- to post-COVID.20 The table compares the total emissions for each group. Same as previous analysis, some individuals are in different groups pre- and post-COVID (i.e., some persons changed in-person/hybrid/remote work status; see Table 4.1.2) and, therefore, the difference in emissions can be the results of changes in behavior for people who did not change work arrangement and the inclusion of new people (in the case of hybrid workers) or loss of people (for in-person workers) that have different driving habits. In line with average VMT increases in Table 4.5.2 above, average GHG emissions increased (Table 4.5.4) for both hybrid and in-person workers. These appear largely driven by extreme outliers toward the top of the distribution. For all percentiles, hybrid workers generated substantially less GHG than in-person workers. Although post-COVID hybrid workers tend to relocate farther from their job (as shown in Table 4.4.2), they, on average, commute fewer days, which offsets the longer one-way commute distance. Table 4.5.3. Total weekly commute GHG Emissions by work arrangement, excluding outliers (non-remote workers with home-to-job distance >= 300 miles)21 Work arrangement Remote Hybrid In-person PreCOVID PostCOVID PreCOVID PostCOVID PreCOVID PostCOVID Mean 0 0 27,949 26,342 42,310 38,070 25% 0 0 3,916 4,692 0 0 50% 0 0 16,962 15,108 24,408 24,909 75% 0 0 36,630 31,379 51,503 53,402 90% 0 0 66,710 64,220 96,477 90,417 95% 0 0 97,920 81,016 139,195 135,074 99% 0 0 178,480 188,324 398,084 209,573 Total workers 226 468 123 335 1,090 956 *Source: survey Question: Q1a, Q2a, Q4_2, Q7_2 Sample size: pre-Covid (1,439), post-Covid (1,759) Commute distance, migration, and emissions Remote work promises large reductions in commute-related emissions. While industry projections expect remote work to remain an important work arrangement, hybrid may become relatively more common (e.g., McKinsey Global Institute 2023; Haan 2023). As such, there is ambiguity as to whether hybrid work will lead to decreases in emissions because the reduction in driving that comes from cutting the number of commute days can be offset by the increasing 20 We did not ask about changes in vehicles in the questionnaire, but car sales were generally much lower in 2020 and 2021 and average fuel efficiency of the car fleet from 2020 to 2023 did not improve enough to make a marked difference for the overall fleet. https://www.bts.gov/content/average-fuel-efficiency-us-light-duty-vehicles 21 See Appendix Table D10 for total weekly commute VMT by work arrangement without excluding outliers 120 distance people drive to get to work. This section focuses on disentangling the relationship between work arrangement, migration decisions, and emissions through statistical analysis. I test the hypothesis that more commute days are associated with higher weekly greenhouse gas emissions. I estimate this relationship directly pre- and post-COVID (equation 11; models 9 and 10). I also estimate the difference in weekly commute GHG emissions among the three types of work arrangement, since the relationship between commute days and weekly GHG emissions may not be linear (equation 12, models 11 and 12). To examine whether post-COVID hybrid workers are more likely to generate higher weekly commute GHG emissions than pre-COVID hybrid workers, I analyze the relationship between the change in weekly commute GHG emissions and the change in commute days from pre to post-COVID periods (Equation 13, models 13 and 14). The equations used are as follows: Equation 11 (Model 9 and 10): 𝑾𝒆𝒆𝒌𝒍𝒚 𝑮𝑯𝑮 𝒆𝒎𝒊𝒔𝒔𝒊𝒐𝒏 𝒑𝒆𝒓 𝒑𝒆𝒓𝒔𝒐𝒏𝒕 = 𝒇(𝒄𝒐𝒎𝒎𝒖𝒕𝒆 𝒅𝒂𝒚𝒔𝒕 , 𝒄𝒐𝒏𝒕𝒓𝒐𝒍𝒔) Equation 12 (Model 11 and 12): 𝑾𝒆𝒆𝒌𝒍𝒚 𝑮𝑯𝑮 𝒆𝒎𝒊𝒔𝒔𝒊𝒐𝒏 𝒑𝒆𝒓 𝒑𝒆𝒓𝒔𝒐𝒏𝒕 = 𝒇(𝒘𝒐𝒓𝒌 𝒂𝒓𝒓𝒂𝒏𝒈𝒎𝒆𝒏𝒕𝒔𝒕 , 𝒄𝒐𝒏𝒕𝒓𝒐𝒍𝒔) Equation 13: ∆𝑾𝒆𝒆𝒌𝒍𝒚 𝑮𝑯𝑮 𝒆𝒎𝒊𝒔𝒔𝒊𝒐𝒏 𝒑𝒆𝒓 𝒑𝒆𝒓𝒔𝒐𝒏 = 𝒇(∆𝒄𝒐𝒎𝒎𝒖𝒕𝒆 𝒅𝒂𝒚𝒔, 𝒄𝒐𝒏𝒕𝒓𝒐𝒍𝒔) The definition of each variable is listed below: • t= post-COVID (2023), pre-COVID (2020) • Commute days = Days of commute per week • Weekly GHG emission per person= tailpipe CO₂ grams per person per week • ∆ Weekly GHG emission per person= change in tailpipe CO₂ grams per person per week from pre-COVID (before March, 2020) to post-COVID (September, 2023) • Work arrangements 2023 = Remote, Hybrid, in-person (omitted) • Controls: predicted remote work ability, regions (Northeast (omitted), Midwest, South, West) Model 9 (Table 4.5.4) indicates that an additional day of commuting in the post-COVID period is associated with an increase of 7,895 weekly round-trip commute GHG emissions (CO₂ in grams), while in the pre-COVID period, the increase is 8,749 (CO₂ in grams) (Model 10). Model 11 shows that in the post-COVID period, compared to remote workers, hybrid workers produce 28,039 more weekly commute GHG emissions, and fully in-person workers produce 40,146 more weekly commute CO₂ emissions. Moreover, an additional mile increase in home-job-distance in post-COVID, there is an associated increase of 16.7 grams of CO₂ emissions, controlling for other characteristics, controlling for other characteristics. 121 The result of the pre-COVID period in model 12 shows a greater difference between hybrid and in-person workers. Hybrid workers produced, ceteris paribus, 32,814 more weekly GHG emissions than remote workers, while in-person workers produced 49,018 more in the preCOVID time period. For home-job-distance, an additional mile increase pre-COVID is associated with an increase of 115 grams of CO₂ emissions. Model 13 indicates that compared to pre-COVID commute days, an additional day of commuting in the post-COVID period is associated with an increase of 9,200 weekly commute GHG emissions. Model 14 shows that compared to pre-COVID home-to-job distance, an additional mile between home and job in post-COVID period is associated with an increase of 66 weekly commute GHG (CO₂) emissions. A few key takeaways from these model results. First, per capita weekly GHG emissions decreased relative to pre-Covid. Second, in-person workers have higher weekly GHG emissions than hybrid workers who in turn have higher weekly GHG emission than remote workers. Migration and commute GHG emissions The foregoing analysis showed that adding days of commutes adds substantially to GHG emissions (Model 13). This is to be expected. Given a fixed commuted distance, driving that distance more days will increase the worker’s emissions. The next set of models turn to the effect of changes in the distance between residential and work location. I hypothesize that compared to their pre-COVID home-to-job distance; workers who move farther from their workplace generate more weekly commute GHG emissions. I used the following equation to statistically test this relationship: Equation 14: ∆𝑊𝑒𝑒𝑘𝑙𝑦 𝐺𝐻𝐺 𝑒𝑚𝑖𝑠𝑠𝑖𝑜𝑛 𝑝𝑒𝑟 𝑝𝑒𝑟𝑠𝑜𝑛 = 𝑓(∆𝐻𝑜𝑚𝑒𝐽𝑜𝑏 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒, 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠) The definition of each variable is listed below: • ∆ Weekly GHG emission per person= change in tailpipe CO₂ grams/mile per person per week from pre-COVID (before March, 2020) to post-COVID (September, 2023) • ∆ HomeJob Distance= change in distance between home and workplace from preCOVID (before March 2020) to post-COVID (September, 2023) • Controls: 𝑅𝑒𝑚𝑜𝑡𝑒̂𝑤𝑜𝑟𝑘 𝐴𝑏𝑖𝑙𝑖𝑡𝑦, regions (Northeast (omitted), Midwest, South, West) The results in table 4.5.4 model 14 show that a one-mile increase in the distance from home to workplace compared to the pre-COVID home-to-job distance is associated with an increase in weekly CO₂ emissions (+66.7 grams) compared to the pre-COVID period. As with the change in number of days commuting, this increase is to be expected. The estimate, however, is skewed because it does not differentiate between types of workers. Remote workers who 122 moved much farther from their workplace will have a large decrease in emissions if they were working in-person pre-COVID, or no change in emissions if there were already working remotely. We, therefore, expand on the model to examine the differences across work arrangements. I use an interaction term to investigate how changes in home-job distance affect emissions for hybrid and in-person workers. The interaction term separates the effect of changes in home-job distance for hybrid and in-person workers and tells us if increases in distance are associated with greater emissions for each of the worker types. I use the following equation to answer this question: Equation 15: ∆𝑊𝑒𝑒𝑘𝑙𝑦 𝐺𝐻𝐺 𝑒𝑚𝑖𝑠𝑠𝑖𝑜𝑛 𝑝𝑒𝑟 𝑝𝑒𝑟𝑠𝑜𝑛 = 𝑓(∆𝐻𝑜𝑚𝑒𝐽𝑜𝑏 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒, 𝑤𝑜𝑟𝑘 𝑎𝑟𝑟𝑎𝑛𝑔𝑒𝑚𝑒𝑛𝑡2023, (∆𝐻𝑜𝑚𝑒𝐽𝑜𝑏 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 ∗ 𝑤𝑜𝑟𝑘 𝑎𝑟𝑟𝑎𝑛𝑔𝑒𝑚𝑒𝑛𝑡2023),𝐽𝑜𝑏 𝑐ℎ𝑎𝑛𝑔𝑒, 𝑚𝑜𝑣𝑖𝑛𝑔 𝑠𝑡𝑎𝑡𝑢𝑠, 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠) The definition of each variable is listed below: • ∆ Weekly GHG emission per person= change in tailpipe CO₂ grams/mile per person per week from pre-COVID (before March, 2020) to post-COVID (September, 2023) • ∆ HomeJob Distance= change in distance between home and workplace from preCOVID (before March 2020) to post-COVID (September, 2023) • Work arrangements 2023 = Remote (omitted), Hybrid, in-person • Job change= job changed after COVID (march, 2020): No (omitted), Yes • Moving status = Moved after COVID (march,2020): No (omitted), Yes • Controls: 𝑅𝑒𝑚𝑜𝑡𝑒̂𝑤𝑜𝑟𝑘 𝐴𝑏𝑖𝑙𝑖𝑡𝑦, regions (Northeast (omitted), Midwest, South, West) 123 Model 15 in Table 4.5.4 indicates that, compared to post-COVID remote workers who experienced no change in weekly GHG emissions pre- and post- COVID (due to my definition of VMT for remote workers), hybrid workers in post-COVID period reduced weekly GHG emissions by 11,220 grams compared to the pre-COVID period. Conversely, in-person workers in post-COVID period increased weekly GHG emissions by 5,983 grams compared to the preCOVID period. Next, I examine the interaction between current work arrangement and changes in hometo-job distance. In the post-COVID period, compared to remote workers, for hybrid workers, each additional mile in distance between home and workplace, changed from pre-COVID to post-COVID, is associated with an increase of 698 in weekly GHG emissions per person. For inperson workers, the increase is 3,562. Regression results from equation 9 to 15 confirm that fully remote workers tend to relocate farther from their workplaces and produce less GHG emissions as they no longer commute (if the worker status changed to remote), while GHG emissions for fully in-person workers remain relatively stable. Hybrid workers tend to move farther away from their workplace relative to in-person workers. However, their overall weekly emissions significantly drop due to the reduction in commute days, offsetting any increased distance between their home and job. 124 Table 4.5.4. Regression results of GHG emissions (without outliers: home-to-job distance >=300 for non remote workers) Dependent Variable Weekly CO₂ per person 2023 Weekly CO₂ per person 2020 Weekly CO₂ per person 2023 Weekly CO₂ per person 2020 ∆ weekly CO₂ per person ∆ weekly CO₂ per person ∆ weekly CO₂ per person Category Variable (9) (10) (11) (12) (13) (14) (15) Commute days Commute days 2023 7895.423*** Commute days 2020 8749.013*** ∆ Commute days 9199.633** * HomeJob Distance HomeJob Distance 2023 16.77373*** HomeJob Distance 2020 115.0457*** ∆ HomeJob Distance 66.72752*** -4.998176 Work arrangement 2023 (Omitted: Remote) Hybrid 28039.78*** -11220.03*** In-person 40146.79*** 5983.285*** Work arrangement 2020 (Omitted: Remote) Hybrid 32814.44*** In-person 49018.91*** Work Arrangement *∆ HomeJob Distance (Omitted: Remote) Hybrid 697.9075*** In-person 3562.659*** Job Change: Yes 61.62939 Moving status: Moved 2224.033 Predicted remote work ability 1735.019 3455.991 697.2536 2834.572 707.8239 -1651.371 300.1607 Region (Omitted: Northeast) Midwest 29.24338 5367.172 -654.5274 4669.459 -4397.641 -4980.022 -2409.551 South 1114.158 5519.595 775.0774 4204.788 -1667.337 -241.3287 -1757.146 West -2152.444 -903.0669 -3402.516 -2492.959 -826.4457 -2248.317 409.082 Constant 273.6008 -3890.297 -1392.669 -9536.828 853.7748 -2280.065 -5569.282 Observations 1,743 1,425 1,713 1,407 1,382 1,364 1,364 R-squared 0.1313 0.0522 0.1221 0.0763 0.0442 0.0172 0.7188 Adjusted R-squared 0.1288 0.0489 0.1185 0.0717 0.0408 0.0135 0.7165 125 5. Conclusion The transition towards remote working following the COVID-19 pandemic significantly reshaped commuting patterns and prompted discussions of the implications for reducing greenhouse gas (GHG) emissions. This project analyzes how changes in commuting behavior following COVID-19 have altered the spatial relationship between homes and workplaces. I investigate whether workers with remote work abilities tend to move farther away from their workplaces and how this affects their commuting behavior. I developed and field a survey in September 2023, using IPSOS’ KnowledgePanel to gather a nationally representative sample of adults (18 years old or older) living in the U.S. and working full-time at the time of the survey. A total of 2,124 responses were collected among four types of workers: individuals with the ability to work from home and those without, further divided by whether they relocated following the Covid outbreak and whether their commuting and driving habits changed. This survey data serves a crucial role in filling gaps in previous work and offers a better understanding of the spatial and environmental implications of remote work. My analysis aims to offer the best estimate currently available regarding the impact of remote work on commute patterns and associated greenhouse gas emissions, helping transit and planning agencies in modeling air quality and traffic congestion. My findings highlight shifting work arrangements, characterized by a notable increase in remote work since the pandemic's onset. Seventy-one percent of pre-covid in-person workers remained in-person in post-covid period, 17% switched to working hybrid, and 11% switched to working fully remotely. Of those working under a hybrid arrangement, a third switched to being remote full time, and most remote workers before the pandemic were still remote in 2023. While uncertainties persist regarding future work arrangements, most survey respondents expect their work relationships to remain the same going forward. However, disparities in remote work ability across industries are evident, with professions emphasizing office settings and those requiring higher educational qualifications being more conducive to remote work. Regarding residential relocations, remote and hybrid workers demonstrate a higher likelihood of relocation. A quarter of survey respondents reported moving at least once since March 2020. Hybrid and remote workers were about 7 percentage points more likely to have moved than in-person workers. In-person workers who switch to remote or hybrid work arrangements were almost twice as likely to move than those who did not. While most moves were short distance (about 10 miles, and 22 miles for remote workers), at least 25% of hybrid workers moved >190 miles away and, among remote workers, >360 miles away. The average distance between home and work has shifted notably across different work arrangements. For remote workers, it increased from 58.6 to 100.5 miles, while for hybrid workers, it decreased from 36.4 to 24.8 miles. Similarly, in-person workers experienced a decline, from 25.6 miles to 21.6 miles. While most remote workers live and work in the same zip code, a quarter of remote workers who lived the farthest from their place of work, increased the distance to their work location by 9 miles (from 9.9 to 18.9 miles). The top 10% of remote workers who live the farthest increased from 74.9 to 197.7 miles. The median commute distance 126 for in-person workers was 6.5 miles and decreased by about 0.4 miles post-COVID. For hybrid workers, this was 7.7 miles, but increased to 8.6 miles post-COVID. From the regression results, I also found that remote workers increased the distance from their job by 29 more miles compared to in-person workers and an additional day of in-person work (commute) is associated with a reduction (-12 miles) in the distance between their home and the workplace in postCOVID. In terms of emissions, reduced commuting frequency is associated with reduced weekly commute GHG, this effect is not offset by the increases in commute distance. I found that the weekly VMT of in-person workers is higher than hybrid workers, which is mainly due to inperson workers’ additional days of commuting. When comparing the change in weekly VMT from pre-COVID to post-COVID periods, I found an increase in the average weekly VMT for both in-person workers (111 to 165 miles) and hybrid workers (74 to 130 miles). Similarly, the weekly commute GHG emissions for in-person workers increased, from 42,614 to 65,850. Results also show that hybrid workers generated higher weekly commute GHG emissions in post-COVID period compared to pre-COVID period, but there is a compositional change in hybrid worker status pre- and post-COVID. The average went up from 27,949 tailpipe gram to 50,393 tailpipe gram. The choices regarding vehicle efficiency and commute distance among hybrid workers remain unclear. Individuals who work remotely for fewer days per week may opt to continue using older and less fuel-efficient vehicles. Since they still tend to relocate farther away from their workplace, this choice could result in increased emissions on those particular days. While uncertainties persist regarding the overall impact of remote work on emissions in the post-COVID era, my findings underscore the potential significance of these arrangements in reducing driving activities and associated emissions. Continued research and ongoing monitoring of hybrid workers’ behavior will be crucial to fully understand and optimize the environmental implications of evolving work patterns. 127 6. References Alonso, W. 1964. Location and Land Use. Harvard University Press, Cambridge, MA. Ahrend, R., et al. (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, OECD Publishing, Paris, https://doi.org/10.1787/cf591216-en. Baert, S., Lippens, L., Moens, E., Weytjens, J., & Sterkens, P. (2020). The COVID-19 crisis and telework: A research survey on experiences, expectations and hopes. Bailey, D. E., & Kurland, N. B. (2002). A review of telework research: Findings, new directions, and lessons for the study of modern work. Journal of Organizational Behavior: The International Journal of Industrial, Occupational and Organizational Psychology and Behavior, 23(4), 383-400. Bartik, A. W., Cullen, Z. B., Glaeser, E. L., Luca, M., & Stanton, C. T. (2020). What jobs are being done at home during the COVID-19 crisis? Evidence from firm-level surveys (No. w27422). National Bureau of Economic Research. Behrens, Kristian, Sergey Kichko and Jacques-François Thisse, 2021. "Working from home: Too much of a good thing?." CESifo Working Paper No. 8831. Bick, A., Blandin, A., & Mertens, K. (2020). Work from home after the COVID-19 Outbreak. Brynjolfsson, E., Horton, J. J., Ozimek, A., Rock, D., Sharma, G., & TuYe, H. Y. (2020). COVID-19 and remote work: An early look at US data (No. w27344). National Bureau of Economic Research. Brynjolfsson, E., Horton, J. J., Makridis, C., Mas, A., Ozimek, A., Rock, D., & TuYe, H. Y. (2023). How many Americans work remotely? a survey of surveys and their measurement issues (No. w31193). National Bureau of Economic Research. Brueckner, Jan K. 1987. Structure of Urban Equilibria: A Unified Treatment of the Muth-Mills Model. In Handbook of Regional and Urban Economics, vol. 2, edited by Edwin S. Mills, 821.45. Amsterdam: Elsevier. Brownstone, D. (2008). Key relationships between the built environment and VMT. Transportation Research Board, 7. Chakrabarty, D. (2018). Anthropocene time. History and Theory, 57(1), 5-32. Choo, S., Mokhtarian, P. L., & Salomon, I. (2005). Does telecommuting reduce vehicle-miles traveled? An aggregate time series analysis for the US. Transportation, 32, 37-64. Couture, Victor, Jonathan I. Dingel, Allison Green, Jessie Handbury and Kevin R. Williams, 2021. "JUE Insight: Measuring movement and social contact with smartphone data: a real-time application to COVID-19." Journal of Urban Economics: 103328. Davis, Morris A., Andra C. Ghent and Jesse M. Gregory, 2021. “The Work-at-Home Technology Boon and its Consequences,” NBER Working Paper 28461. Delventhal, Matthew J., Eunjee Kwon and Andrii Parkhomenko, 2021. "JUE Insight: How do cities change when we work from home?" Journal of Urban Economics: 103331. Ewing, R., Hamidi, S., & Grace, J. B. (2016). Compact development and VMT—Environmental determinism, self-selection, or some of both?. Environment and Planning B: Planning and Design, 43(4), 737-755. 128 Fang, K., & Volker, J. (2017). Cutting greenhouse gas emissions is only the beginning: a literature review of the co-benefits of reducing vehicle miles traveled. Giovanis, E. (2018). The relationship between teleworking, traffic and air pollution. Atmospheric pollution research, 9(1), 1-14. Golob, T. F. (2002). Travelbehavior. com-activity approaches to modeling the effects of Information Technology on Personal Travel Behavior. Jonathan I. Dingel & Brent Neiman, 2020. "How many jobs can be done at home?," Journal of Public Economics, vol 189 Kitou, E., & Horvath, A. (2003). Energy-related emissions from telework. Koenig, B. E., Henderson, D. K., & Mokhtarian, P. L. (1996). The travel and emissions impacts of telecommuting for the State of California Telecommuting Pilot Project. Transportation Research Part C: Emerging Technologies, 4(1), 13-32. Lari, A. (2012). Telework/Workforce flexibility to reduce congestion and environmental degradation?. procedia-social and behavioral Sciences, 48, 712-721. Larson, W., & Zhao, W. (2017). Telework: Urban form, energy consumption, and greenhouse gas implications. Economic Inquiry, 55(2), 714-735. Haan, K. (2023). Remote Work Statistics And Trends In 2024. Forbes Advisor. https://www.forbes.com/advisor/business/remote-work-statistics/#sources_section Hamer, R., Kroes, E., & Van Ooststroom, H. (1991). Teleworking in the Netherlands: an evaluation of changes in travel behaviour. Transportation, 18(4), 365-382. Haslag, Peter H. and Daniel Weagley, 2021. "From LA to Boise: How Migration Has Changed During the COVID-19 Pandemic," Available at SSRN 3808326. Hopkins, J. L., & McKay, J. (2019). Investigating ‘anywhere working’as a mechanism for alleviating traffic congestion in smart cities. Technological Forecasting and Social Change, 142, 258-272. Hook, A., Sovacool, B. K., & Sorrell, S. (2020). A systematic review of the energy and climate impacts of teleworking. Environmental Research Letters, 15(9), 093003. Hu, L., & He, S. Y. (2016). Association between telecommuting and household travel in the Chicago metropolitan area. Journal of Urban Planning and Development, 142(3), 04016005. IPSOS, 2022, Panel Book: KnowledgePanel Brings you a Probability-Based Online Sample. Kar, A., Le, H. T., & Miller, H. J. (2021). What Is Essential Travel? Socioeconomic Differences in Travel Demand in Columbus, Ohio, during the COVID-19 Lockdown. Annals of the American Association of Geographers, 1-24 Kazancoglu, Y., Ozbiltekin-Pala, M., & Ozkan-Ozen, Y. D. (2021). Prediction and evaluation of greenhouse gas emissions for sustainable road transport within Europe. Sustainable Cities and Society, 70, 102924. Liu, S., & Su, Y. (2021). The impact of the COVID-19 pandemic on the demand for density: Evidence from the US housing market. Economics letters, 207, 110010. Lu, Y., & Giuliano, G. (2022). Understanding Mobility Change in Response to COVID-19: A Los Angeles Case Study. Working Paper. 129 McKinsey Global Institute (2023). How hybrid work has changed the way people work, live, and shop. https://www.mckinsey.com/mgi/our-research/empty-spaces-and-hybrid-placeschapter-1#/ Mills, Edward. 1967. An Aggregate Model of Resource Allocation in Metropolitan Area. The American Economic Review, Vol. 27(2):197-210. Mills, Edwin S. and Jee Peng Tan. 1980. A Comparison of Urban Population Density Functions in Developed and Developing Countries. Urban Studies 17: 313-321. Mondragon, John and Johannes Wieland, 2022, “Housing demand and remote work”, Federal Reserve Bank of San Francisco mimeo. Mokhtarian, P. L., Collantes, G. O., & Gertz, C. (2004). Telecommuting, residential location, and commute-distance traveled: evidence from State of California employees. Environment and Planning A, 36(10), 1877-1897. Mokhtarian, P. L., & Varma, K. V. (1998). The trade-off between trips and distance traveled in analyzing the emissions impacts of center-based telecommuting. Transportation research part D: Transport and Environment, 3(6), 419-428. Muth, Richard. 1969. Cities and Housing. The University of Chicago Press. Chicago, IL. Nguyen, M. H. (2021). Factors influencing home-based telework in Hanoi (Vietnam) during and after the COVID-19 era. Transportation, 48(6), 3207-3238. Nilles, Jack M. "Traffic reduction by telecommuting: A status review and selected bibliography." Transportation Research Part A: General 22.4 (1988): 301-317. Ozimek, Adam, 2020a. "The future of remote work." Available at SSRN 3638597, July. Ory, D. T., & Mokhtarian, P. L. (2006). Which came first, the telecommuting or the residential relocation? An empirical analysis of causality. Urban Geography, 27(7), 590-609. Obeid, H., Anderson, M. L., Bouzaghrane, M. A., & Walker, J. L. (2022). Does Telecommuting Reduce Trip-Making? Evidence From a US Panel During the COVID-19 Impact and Recovery Periods. Evidence From a US Panel During the COVID-19 Impact and Recovery Periods. Ramani, A., & Bloom, N. (2021). The Donut effect of COVID-19 on cities (No. w28876). National Bureau of Economic Research. Shabanpour, R., Golshani, N., Shamshiripour, A., & Mohammadian, A. K. (2018). Eliciting preferences for adoption of fully automated vehicles using best-worst analysis. Transportation research part C: emerging technologies, 93, 463-478. Rosenthal, S. S., Strange, W. C., & Urrego, J. A. (2022). JUE insight: Are city centers losing their appeal? Commercial real estate, urban spatial structure, and COVID-19. Journal of Urban Economics, 127, 103381. Shin, E. J. (2020). Commuter benefits programs: Impacts on mode choice, VMT, and spillover effects. Transport Policy, 94, 11-22. Teo, T. S., Lim, V. K., & Wai, S. H. (1998). An empirical study of attitudes towards teleworking among information technology (IT) personnel. International journal of information management, 18(5), 329-343. True North Research. (2021). Remote Work Policies & Practices. Surveys of Businesses and Employees in the San Diego Region. Summary Report Prepared for SANDAG. 130 https://www.sandag.org/-/media/SANDAG/Documents/PDF/data-and-research/appliedresearch-and-performance-monitoring/surveys/remote-work-policies-and-practicessurvey-2021-04-16.pdf U.S. Census Bureau. (2022). "Sex of Workers by Means of Transportation to Work." American Community Survey, ACS 5-Year Estimates Detailed Tables, Table B08006, 2022, https://data.census.gov/table/ACSDT5Y2022.B08006?q=means of transportation to work. Accessed on April 12, 2024. U.S. Census Bureau, Current Population Survey (CPS), Annual Social and Economic Supplement 1948-2022 (CPS ASEC). Table A-1. Annual Geographic Mobility Rates, By Type of Movement: 1948-2022. https://www.census.gov/data/tables/timeseries/demo/geographic-mobility/historic.html Wang, B. S., Rodnyansky, S., Boarnet, M. G., & Comandon, A. (2024). Measuring the impact of COVID-19 policies on local commute traffic: Evidence from mobile data in Northern California. Travel Behaviour and Society, 34, 100660. Wu, H., Chang, Y., & Chen, Y. (2024). Greenhouse gas emissions under work from home vs. office: An activity-based individual-level accounting model. Applied Energy, 353, 122167. Zhu, P., Wang, L., Jiang, Y., & Zhou, J. (2018). Metropolitan size and the impacts of telecommuting on personal travel. Transportation, 45(2), 385-414. 131 Chapter 5. Conclusion This dissertation has explored the profound impact of the COVID-19 pandemic on commuting patterns, residential choices, and environmental impacts at both national and regional level. Through three interconnected studies, I have gained insights into how remote work, policy interventions, and commuting behavior have redefined urban dynamics and what these changes mean for the future of urban planning. The first study highlighted that during the initial stage of COVID-19, remote work, driven by private sector policies, had a more significant and enduring effect on reducing commute traffic than state-imposed restrictions. This research highlights a significant disparity in the benefits of remote work and inequalities in commute burden between higher-income and lower-income workers, emphasizing an important issue related to social equity. My findings confirm that lower-income workers were disproportionately affected by social distancing policy, and this exacerbated their living quality. The second study demonstrated that increased remote work led to substantial decreases in traffic volumes and a shift towards dispersed residential choices. Industry composition highlight the critical role of economic dynamics in shaping traffic patterns. Office-centric regions saw more significant decreases in traffic, while regions with higher shares of agriculture, manufacture, or trade and transport experienced smaller declines. This migration pattern future confirms that many workers took advantage of remote work flexibility to move away from highcost urban centers to more affordable or desirable locations. This study provided critical insights into how workplace industry characteristics influence traffic volumes and highlighted the importance of accommodating remote work trends in future housing policies. The third study examined the environmental benefits of remote work, showing that fully remote workers significantly reduce GHG emissions by eliminating commutes. Hybrid workers, despite moving farther away, also contribute to lower emissions due to reduced commute frequency. These findings underscore the potential of remote work to improve environmental sustainability and provide essential evidence for transit and planning agencies to model air quality and traffic congestion effectively. The COVID-19 pandemic has served as a catalyst for significant changes in how we work and live. The lessons learned from these studies extend beyond the individual findings to offer a broader understanding of the pandemic's long-term impact on urban life. The research highlights the potential of remote work in reshaping commuting patterns, residential choices, and environmental outcomes. As we move forward, it is crucial to develop strategies that are flexible and adaptive to these changes. The findings emphasize the need for equitable urban planning policies that address the disparities in commute burden and access to remote work opportunities. Policymakers must consider the diverse needs of different socioeconomic groups and ensure that infrastructure improvements and housing policies promote inclusivity and equity. The ongoing evolution of work patterns and commuting behavior calls for continued research to monitor and understand these trends. Future studies should focus on the long-term implications of remote work on urban structure, housing markets, and environmental 132 sustainability. Additionally, examining the potential for hybrid work arrangements to optimize productivity, commuting efficiency and environmental benefits will be critical. 133 Bibliography Adrjan, P. et al. Working from Home after COVID-19: Evidence from Job Postings in 20 Countries. SSRN Scholarly Paper at https://doi.org/10.2139/ssrn.4064191 (2022) Ahrend, R., et al. (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, OECD Publishing, Paris, . Aksoy, C. G. et al. Working from Home Around the World. Working Paper at https://doi.org/10.3386/w30446 (2022) Allen, J. et al. Understanding the impact of e-commerce on last-mile light goods vehicle activity in urban areas: The case of London. Transportation Research Part D: Transport and Environment 61, 325–338 (2018). Alonso, W. Location and Land Use: Toward a General Theory of Land Rent. (Harvard University Press, 1964). doi:10.4159/harvard.9780674730854 Anwari, N., Ahmed, M. T., Islam, M. R., Hadiuzzaman, M., & Amin, S. (2021). Exploring the travel behavior changes caused by the COVID-19 crisis: A case study for a developing country. Transportation Research Interdisciplinary Perspectives, 9, 100334. Austrian, K., Pinchoff, J., Tidwell, J. B., White, C., Abuya, T., Kangwana, B., ... & Ngo, T. D. (2020). COVID-19 related knowledge, attitudes, practices and needs of households in informal settlements in Nairobi, Kenya. BLS. (2022). Occupational Requirements in the United States News Release. U.S. Bureau of Labor Statistics. Retrieved April 22, 2023, from https://www.bls.gov/news.release/archives/ors_11172022.htm Baert, S., Lippens, L., Moens, E., Weytjens, J., & Sterkens, P. (2020). The COVID-19 crisis and telework: A research survey on experiences, expectations and hopes. Bailey, D. E. & Kurland, N. B. A review of telework research: Findings, new directions, and lessons for the study of modern work. Journal of Organizational Behavior: The International Journal of Industrial, Occupational and Organizational Psychology and Behavior 23, 383–400 (2002). Baker, M. G. (2020). Nonrelocatable occupations at increased risk during pandemics: United States, 2018. American Journal of Public Health, 110(8), 1126-1132. Barbieri, P., Boffelli, A., Elia, S., Fratocchi, L., Kalchschmidt, M., & Samson, D. (2020). What can we learn about reshoring after Covid-19?. Operations Management Research, 13(3), 131- 136. Barrero, J. M., Bloom, N. & Davis, S. J. Why Working from Home Will Stick. Working Paper at https://doi.org/10.3386/w28731 (2021) 134 Bartik, A. W., Cullen, Z. B., Glaeser, E. L., Luca, M. & Stanton, C. T. What jobs are being done at home during the COVID-19 crisis? Evidence from firm-level surveys. (National Bureau of Economic Research, 2020). at <https://www.nber.org/papers/w27422> Bartik, A. W., Cullen, Z. B., Glaeser, E. L., Luca, M., & Stanton, C. T. (2020). What jobs are being done at home during the COVID-19 crisis? Evidence from firm-level surveys (No. w27422). National Bureau of Economic Research. Baum-Snow, N. Did highways cause suburbanization? The quarterly journal of economics 122, 775–805 (2007). Bay Area Council Economic Institute and McKinsey and Co. (2018). Continuing Growth and Unparalleled Innovation: Bay Area Economic Profile. Tenth in a Series. Behrens, K., Kichko, S. & Thisse, J.-F. Working from home: Too much of a good thing? Regional Science and Urban Economics 105, 103990 (2024). Bick, A., Blandin, A., & Mertens, K. (2020). Work from home after the COVID-19 Outbreak. Blau, F. D., Koebe, J., & Meyerhofer, P. A. (2021). Who are the essential and frontline workers?. Business Economics, 56(3), 168-178. Bloom, N., Davis, S. & Barrero, J. M. 60 million fewer commuting hours per day: How Americans use time saved by working from home. CEPR (2020). at <https://cepr.org/voxeu/columns/60-million-fewer-commuting-hours-day-how-americans-usetime-saved-working-home> Boarnet, M., Rodnyansky, S., Austin, B., Comandon, A., Wang, B., & Rodriguez, S. (2023). Bay Area to Central Valley Migration and its Impacts. Final Report prepared for Chan Zuckerberg Initiative. March. Brinkman, J., & Mangum, K. (2022). JUE Insight: The Geography of Travel Behavior in the Early Phase of the COVID-19 Pandemic. Journal of Urban Economics, 127, 103384. https://doi.org/10.1016/j.jue.2021.103384 Brough, R., Freedman, M., & Phillips, D. C. (2021). Understanding socioeconomic disparities in travel behavior during the COVID-19‐19 pandemic. Journal of Regional Science, 61(4), 753- 774. Brownstone, D. (2008). Key relationships between the built environment and VMT. Transportation Research Board, 7. Brueckner, Jan K. 1987. Structure of Urban Equilibria: A Unified Treatment of the Muth-Mills Model. In Handbook of Regional and Urban Economics, vol. 2, edited by Edwin S. Mills, 821.45. Amsterdam: Elsevier. Brynjolfsson, E., Horton, J. J., Makridis, C., Mas, A., Ozimek, A., Rock, D., & TuYe, H. Y. (2023). How many Americans work remotely? a survey of surveys and their measurement issues (No. w31193). National Bureau of Economic Research. 135 Brynjolfsson, E., Horton, J. J., Ozimek, A., Rock, D., Sharma, G., & TuYe, H. Y. (2020). COVID-19 and remote work: An early look at US data (No. w27344). National Bureau of Economic Research. Caros, N. S. & Zhao, J. Evaluating the travel impacts of a shared mobility system for remote workers. Transportation Research Part D: Transport and Environment 121, 103798 (2023). Ceccato, R., Rossi, R., & Gastaldi, M. (2021). Travel demand prediction during COVID-19 pandemic: Educational and working trips at the University of Padova. Sustainability, 13(12), 6596. Chakrabarty, D. (2018). Anthropocene time. History and Theory, 57(1), 5-32. Chapple, K. et al. The Death of Downtown? Pandemic Recovery Trajectories across 62 North American Cities. (School of Cities, University of Toronto; Institute of Governmental Studies, UC Berkeley, 2023). Chen, J. T., & Krieger, N. (2021). Revealing the unequal burden of COVID-19 by income, race/ethnicity, and household crowding: US county versus zip code analyses. Journal of Public Health Management and Practice, 27(1), S43-S56. Choo, S., Mokhtarian, P. L., & Salomon, I. (2005). Does telecommuting reduce vehicle-miles traveled? An aggregate time series analysis for the US. Transportation, 32, 37-64. Circella, G. & Mokhtarian, P. L. in The Geography of Urban Transportation (eds. Giuliano, G. & Hanson, S.) (Guilford Publications, 2017). Clark, T. N., Lloyd, R., Wong, K. K. & Jain, P. Amenities Drive Urban Growth. Journal of Urban Affairs 24, 493–515 (2002). Couch, K. A., Fairlie, R. W., & Xu, H. (2020). Early evidence of the impacts of COVID-19 on minority unemployment. Journal of Public Economics, 192, 104287. https://doi.org/10.1016/j.jpubeco.2020.104287 Couture, Victor, Jonathan I. Dingel, Allison Green, Jessie Handbury and Kevin R. Williams, 2021. "JUE Insight: Measuring movement and social contact with smartphone data: a real-time application to COVID-19." Journal of Urban Economics: 103328. Davis, Morris A., Andra C. Ghent and Jesse M. Gregory, 2021. “The Work-at-Home Technology Boon and its Consequences,” NBER Working Paper 28461. De Vos, J. (2020). The effect of COVID-19 and subsequent social distancing on travel behavior. Transportation Research Interdisciplinary Perspectives, 5, 100121. Delventhal, M. & Parkhomenko, A. Spatial Implications of Telecommuting. SSRN Scholarly Paper at https://doi.org/10.2139/ssrn.3746555 (2023) Delventhal, Matthew J., Eunjee Kwon and Andrii Parkhomenko, 2021. "JUE Insight: How do cities change when we work from home?" Journal of Urban Economics: 103331. 136 Dingel, J. I. & Neiman, B. How many jobs can be done at home? Journal of Public Economics 189, 104235 (2020). Dingil, A. E., & Esztergár-Kiss, D. (2021). The Influence of the COVID-19 Pandemic on Mobility Patterns: The First Wave’s Results. Transportation Letters, 13(5-6), 434-446. Echaniz, E., Rodríguez, A., Cordera, R., Benavente, J., Alonso, B., & Sañudo, R. (2021). Behavioural changes in transport and future repercussions of the COVID-19 outbreak in Spain. Transport Policy, 111, 38-52. Ewing, R., Hamidi, S., & Grace, J. B. (2016). Compact development and VMT—Environmental determinism, self-selection, or some of both?. Environment and Planning B: Planning and Design, 43(4), 737-755. Fang, K., & Volker, J. (2017). Cutting greenhouse gas emissions is only the beginning: a literature review of the co-benefits of reducing vehicle miles traveled. Foroohar, R. US cities must beware the ‘donut effect’. (2023). at <https://www.ft.com/content/c7a3b822-0a98-4a95-9b49-63f0b96d9e98> Gao, S., Rao, J., Kang, Y., Liang, Y., Kruse, J., Doepfer, D., ... & Yandell, B. S. (2020). Mobile phone location data reveal the effect and geographic variation of social distancing on the spread of the COVID-19 epidemic. arXiv preprint arXiv:2004.11430. Garnier, R., Benetka, J. R., Kraemer, J., & Bansal, S. (2021). Socioeconomic disparities in social distancing during the COVID-19 pandemic in the United States: observational study. Journal of medical Internet research, 23(1), e24591. Giovanis, E. (2018). The relationship between teleworking, traffic and air pollution. Atmospheric pollution research, 9(1), 1-14. Glaeser, E. L., Kolko, J. & Saiz, A. Consumer City. Journal of Economic Geography 1, 27–50 (2001). Golob, T. F. Travelbehavior.com: Activity Approaches to Modeling the Effects of Information Technology on Personal Travel Behavior. (2002). At <https://escholarship.org/uc/item/9t40s1mc> Graham, S. & Marvin, S. Splintering Urbanism: Networked Infrastructures, Technological Mobilities and the Urban Condition. (Routledge, 2001). doi:10.4324/9780203452202 Gupta, A., Mittal, V. & Van Nieuwerburgh, S. Work From Home and the Office Real Estate Apocalypse. SSRN Scholarly Paper at https://doi.org/10.2139/ssrn.4124698 (2023) Gupta, R., Ghosh, A., Singh, A. K., & Misra, A. (2020). Clinical considerations for patients with diabetes in times of COVID-19 epidemic. Diabetes & metabolic syndrome, 14(3), 211. Haan, K. (2023). Remote Work Statistics And Trends In 2024. Forbes Advisor. 137 Hamer, R., Kroes, E., & Van Ooststroom, H. (1991). Teleworking in the Netherlands: an evaluation of changes in travel behaviour. Transportation, 18(4), 365-382. Hamidi, S., & Zandiatashbar, A. (2021). Compact development and adherence to stay-at-home order during the COVID-19 pandemic: A longitudinal investigation in the United States. Landscape and Urban Planning, 205, 103952. https://doi.org/10.1016/j.landurbplan.2020.103952 Hanibuchi, T., Yabe, N., & Nakaya, T. (2021). Who is staying home and who is not? Demographic, socioeconomic, and geographic differences in time spent outside the home during the COVID-19 outbreak in Japan. Preventive Medicine Reports, 21, 101306. Hansen, S. et al. Remote Work across Jobs, Companies, and Space. Working Paper at https://www.nber.org/system/files/working_papers/w31007/w31007.pdf (2023) Harrington, D. M., & Hadjiconstantinou, M. (2022). Changes in commuting behaviours in response to the COVID-19 pandemic in the UK. Journal of transport & health, 24, 101313. Harrington, E. & Kahn, M. E. Has the Rise of Work-from-Home Reduced the Motherhood Penalty in the Labor Market? Working Paper at https://drive.google.com/file/d/1Z1J2GHZjqkWzRV5ygA02yzSemwkAmzne/view?usp=sharing (2023) Haslag, P. H. & Weagley, D. From L.A. to Boise: How Migration Has Changed During the COVID-19 Pandemic. SSRN Scholarly Paper at https://doi.org/10.2139/ssrn.3808326 (2022) Hill, T., Gonzalez, K. E., & Davis, A. (2021). The Nastiest Question: Does Population Mobility Vary by State Political Ideology during the Novel Coronavirus (COVID-19) Pandemic? Sociological Perspectives, 64(5), 786–803. https://doi.org/10.1177/0731121420979700 Hook, A., Sovacool, B. K., & Sorrell, S. (2020). A systematic review of the energy and climate Hopkins, J. L., & McKay, J. (2019). Investigating ‘anywhere working’as a mechanism for alleviating traffic congestion in smart cities. Technological Forecasting and Social Change, 142, 258-272. Hu, L., & He, S. Y. (2016). Association between telecommuting and household travel in the Chicago metropolitan area. Journal of Urban Planning and Development, 142(3), 04016005. Hu, S., & Chen, P. (2021). Who left riding transit? Examining socioeconomic disparities in the impact of COVID-19 on ridership. Transportation Research Part D: Transport and Environment, 90, 102654. Hughson, T. L. & Goodman, P. S. Telecommuting: Corporate practices and benefits. National Productivity Review 5, 315–324 (1986). Huws, U. Telework: Projections. Futures 23, 19–31 (1991). IPSOS, 2022, Panel Book: KnowledgePanel Brings you a Probability-Based Online Sample. 138 Janelle, D. G. Metropolitan expansion and the communications-transportation trade-off. Geography of Urban Transportation (1986). at <https://trid.trb.org/View/238196> Jay, J., Bor, J., Nsoesie, E. O., Lipson, S. K., Jones, D. K., Galea, S., & Raifman, J. (2020). Neighbourhood income and physical distancing during the COVID-19 pandemic in the United States. Nature human behaviour, 4(12), 1294-1302. Jonathan I. Dingel & Brent Neiman, 2020. "" Journal of Public Economics, vol 189 Kain, J. F. The journey-to-work as a determinant of residential location. Papers in Regional Science 9, 137–160 (1962). Kalter, M. J. O., Geurs, K. T., & Wismans, L. (2021). Post COVID-19 teleworking and car use intentions. Evidence from large scale GPS-tracking and survey data in the Netherlands. Transportation Research Interdisciplinary Perspectives, 12, 100498. Kar, A., Le, H. T., & Miller, H. J. (2021). What Is Essential Travel? Socioeconomic Differences in Travel Demand in Columbus, Ohio, during the COVID-19 Lockdown. Annals of the American Association of Geographers, 1-24 Kazancoglu, Y., Ozbiltekin-Pala, M., & Ozkan-Ozen, Y. D. (2021). Prediction and evaluation of greenhouse gas emissions for sustainable road transport within Europe. Sustainable Cities and Society, 70, 102924. Kim, J., & Kwan, M. P. (2021). The impact of the COVID-19 pandemic on people's mobility: A longitudinal study of the US from March to September of 2020. Journal of transport geography, 93, 103039. Kingson, J. A. ‘The donut effect’ is reshaping America’s cities. Axios (2023). at <https://www.axios.com/2023/12/15/cities-2024-donut-effect-crime-housing-conversioncongestion-pricing> Kitamura, T., Aso, Y., Kuniyoshi, N., Hara, K., & Yogo, Y. (1990). High incidence of urinary JC virus excretion in nonimmunosuppressed older patients. Journal of Infectious Diseases, 161(6), 1128-1133. Kitou, E., & Horvath, A. (2003). Energy-related emissions from telework. Koenig, B. E., Henderson, D. K., & Mokhtarian, P. L. (1996). The travel and emissions impacts of telecommuting for the State of California Telecommuting Pilot Project. Transportation Research Part C: Emerging Technologies, 4(1), 13-32. Lari, A. (2012). Telework/Workforce flexibility to reduce congestion and environmental degradation?. procedia-social and behavioral Sciences, 48, 712-721. Larson, W., & Zhao, W. (2017). Telework: Urban form, energy consumption, and greenhouse gas implications. Economic Inquiry, 55(2), 714-735. 139 Lee, T. B. The Donut Effect: How the Pandemic Hollowed Out America’s Biggest Cities. Slate (2022). at <https://slate.com/business/2022/04/the-donut-effect-how-the-pandemic-hollowedout-americas-biggest-cities.html> Liu, S., & Su, Y. (2021). The impact of the COVID-19 pandemic on the demand for density: Evidence from the US housing market. Economics letters, 207, 110010. Loo, B. P., & Wang, B. (2018). Factors associated with home-based e-working and e-shopping in Nanjing, China. Transportation, 45(2), 365-384. Lou, J., Shen, X., & Niemeier, D. (2020). Are stay-at-home orders more difficult to follow for low-income groups?. Journal of transport geography, 89, 102894. Lu, Y., & Giuliano, G. (2022). Understanding Mobility Change in Response to COVID-19: A Martin, B. H. & MacDonnell, R. Is telework effective for organizations? A meta-analysis of empirical research on perceptions of telework and organizational outcomes. Management Research Review 35, 602–616 (2012). Mateyka, P. J., Rapino, M. & Landivar, L. C. Home-based workers in the United States: 2010. (U.S. Department of Commerce Economics and Statistics Administration, U.S. Census Bureau, 2012). McKinney, E., Morton, I., and Rodnyansky, S. (2021). “The Uneven Workplace Recovery from COVID-19: Bay Area versus Central Valley”. McKinsey Global Institute (2023). How hybrid work has changed the way people work, live, and shop. Mills, E. S. & Tan, J. P. A Comparison of Urban Population Density Functions in Developed and Developing Countries. Urban Studies 17, 313–321 (1980). Mills, E. S. An Aggregative Model of Resource Allocation in a Metropolitan Area. The American Economic Review 57, 197–210 (1967). Mills, Edwin S. and Jee Peng Tan. 1980. A Comparison of Urban Population Density Functions in Developed and Developing Countries. Urban Studies 17: 313-321. Mitomo, H., & Jitsuzumi, T. (1999). Impact of telecommuting on mass transit congestion: the Tokyo case. Telecommunications policy, 23(10-11), 741-751. Mokhtarian, P. L. A conceptual analysis of the transportation impacts of B2C e-commerce. Transportation 31, 257–284 (2004). Mokhtarian, P. L., & Varma, K. V. (1998). The trade-off between trips and distance traveled in analyzing the emissions impacts of center-based telecommuting. Transportation research part D: Transport and Environment, 3(6), 419-428. 140 Mokhtarian, P. L., Collantes, G. O. & Gertz, C. Telecommuting, Residential Location, and Commute-Distance Traveled: Evidence from State of California Employees. Environ Plan A 36, 1877–1897 (2004). Mokhtarian, P. L., Collantes, G. O., & Gertz, C. (2004). Telecommuting, residential location, and commute-distance traveled: evidence from State of California employees. Environment and Planning A, 36(10), 1877-1897. Mondragon, J. & Wieland, J. Housing Demand and Remote Work. San Francisco Fed Preprint at https://www.frbsf.org/economic-research/publications/working-papers/2022/11/ (2022) Muth, Richard. 1969. Cities and Housing. The University of Chicago Press. Chicago, IL. Nguyen, M. H. (2021). Factors influencing home-based telework in Hanoi (Vietnam) during and after the COVID-19 era. Transportation, 48(6), 3207-3238. Nguyen, M. H. Factors influencing home-based telework in Hanoi (Vietnam) during and after the COVID-19 era. Transportation 48, 3207–3238 (2021). Nilles, J. M. Telecommuting and urban sprawl: Mitigator or inciter? Transportation 18, 411–432 (1991). Nilles, Jack M. "Traffic reduction by telecommuting: A status review and selected bibliography." Transportation Research Part A: General 22.4 (1988): 301-317. Obeid, H., Anderson, M. L., Bouzaghrane, M. A., & Walker, J. L. (2022). Does Telecommuting Reduce Trip-Making? Evidence From a US Panel During the COVID-19 Impact and Recovery Periods. Evidence From a US Panel During the COVID-19 Impact and Recovery Periods. Ory, D. T., & Mokhtarian, P. L. (2006). Which came first, the telecommuting or the residential relocation? An empirical analysis of causality. Urban Geography, 27(7), 590-609. Ozimek, A. Remote Workers on the Move. SSRN Scholarly Paper at https://doi.org/10.2139/ssrn.3790004 (2020) Ozimek, Adam, 2020a. "The future of remote work." Available at SSRN 3638597, July. Pabilonia, S. W. & Vernon, V. Telework, Wages, and Time Use in the United States. Rev Econ Household 20, 687–734 (2022). Pan, Y., & He, S. Y. (2022). Analyzing COVID-19’s impact on the travel mobility of various social groups in China’s Greater Bay Area via mobile phone big data. Transportation Research Part A: Policy and Practice. Parker, M. E. G., Li, M., Bouzaghrane, M. A., Obeid, H., Hayes, D., Frick, K. T., Rodríguez, D. A., Sengupta, R., Walker, J., & Chatman, D. G. (2021). Public transit use in the United States in the era of COVID-19: Transit riders’ travel behavior in the COVID-19 impact and recovery period. Transport Policy, 111, 53–62. https://doi.org/10.1016/j.tranpol.2021.07.005 141 Plyushteva, A. (2022). Essential workers’ pandemic mobilities and the changing meanings of the commute. The Geographical Journal, geoj.12447. Ramani, A., & Bloom, N. (2021). The Donut effect of COVID-19 on cities (No. w28876). National Bureau of Economic Research. Rosenthal, S. S., Strange, W. C., & Urrego, J. A. (2022). JUE insight: Are city centers losing their appeal? Commercial real estate, urban spatial structure, and COVID-19. Journal of Urban Economics, 127, 103381. Ruppel, C. P. & Harrington, S. J. Telework: An innovation where nobody is getting on the bandwagon? ACM SIGMIS Database: the DATABASE for Advances in Information Systems 26, 87–104 (1995). San Joaquin Council of Governments. (2013). Interregional multi-modal commute trip planning study. Sassen, S. The Global City: New York, London, Tokyo. (University Press, 1991). Shabanpour, R., Golshani, N., Shamshiripour, A., & Mohammadian, A. K. (2018). Eliciting preferences for adoption of fully automated vehicles using best-worst analysis. Transportation research part C: emerging technologies, 93, 463-478. Shamshiripour, A., Rahimi, E., Shabanpour, R., & Mohammadian, A. (Kouros). (2020). How is COVID-19 reshaping activity-travel behavior? Evidence from a comprehensive survey in Chicago. Transportation Research Interdisciplinary Perspectives, 7, 100216. Shin, E. J. (2020). Commuter benefits programs: Impacts on mode choice, VMT, and spillover effects. Transport Policy, 94, 11-22. Smart, T. (2022). The Great Return: Companies Are Calling Their Workers Back to the Office as Covid-19 Fades. U.S.News. Smith, H. (2021). What will be new norm in L.A. traffic? Transit officials and motorists wait to see if patterns altered by Covid offer clues to future of commuting. Los Angeles Times. Srichan, P., Apidechkul, T., Tamornpark, R., Yeemard, F., Khunthason, S., Kitchanapaiboon, S., ... & Upala, P. (2020). Knowledge, attitudes and preparedness to respond to COVID-19 among the border population of northern Thailand in the early period of the pandemic: a crosssectional study. WHO South-East Asia journal of public health, 9(2), 118-125. Stiles, J., & Smart, M. J. (2021). Working at home and elsewhere: daily work location, telework, and travel among United States knowledge workers. Transportation, 48(5), 2461-2491. Storper, M. The Regional World: Territorial Development in a Global Economy. (1997). StreetLight Insight (2022). StreetLight All Vehicles Volume Methodology and Validation White Paper. 142 Sun, Y., Hu, X., & Xie, J. (2021). Spatial inequalities of COVID-19 mortality rate in relation to socioeconomic and environmental factors across England. Science of the total environment, 758, 143595. Sung, J., & Monschauer, Y. (2020, May 27). Changes in transport behaviour during the Covid19 crisis. Iea. Retrieved April 20, 2023, from https://www.iea.org/articles/changes-in-transportbehaviour-during-the-covid-19-crisis Teo, T. S., Lim, V. K., & Wai, S. H. (1998). An empirical study of attitudes towards teleworking among information technology (IT) personnel. International journal of information management, 18(5), 329-343. Thomas, F. M., Charlton, S. G., Lewis, I., & Nandavar, S. (2021). Commuting before and after COVID-19. Transportation Research Interdisciplinary Perspectives, 11, 100423. Tokey, A. I. (2021). Spatial association of mobility and COVID-19 infection rate in the USA: A county-level study using mobile phone location data. Journal of Transport & Health, 22, 101135. Tomer , A., & Fishbane, L. (2020). Coronavirus has shown us a world without traffic. Can we sustain it? Brookings. Retrieved April 20, 2023, from True North Research. (2021). Remote Work Policies & Practices. Surveys of Businesses and Employees in the San Diego Region. Summary Report Prepared for SANDAG. Truong, D., & Truong, M. D. (2021). Projecting daily travel behavior by distance during the pandemic and the spread of COVID-19 infections – Are we in a closed loop scenario? Transportation Research Interdisciplinary Perspectives, 9, 100283. Yamamoto, T., & Kitamura, R. (1999). An analysis of time allocation to in-home and out-ofhome discretionary activities across working days and non-working days. Transportation, 26, 231-250. Yasenov, V. I. (2020). Who can work from home?. IZA Discussion Papers, No. 13197, Institute of Labor Economics (IZA), Bonn Yilmazkuday, H. (2020). COVID-19‐19 and unequal social distancing across demographic groups. Regional Science Policy & Practice, 12(6), 1235-1248. Zheng, Y., Wang, S., Liu, L., Aloisi, J. & Zhao, J. Impacts of remote work on vehicle miles traveled and transit ridership in the USA. Nature Cities 1–13 (2024). 143 Appendix Figure A1. Ratio of Biden to Trump votes in the 2020 Presidential election at ZCTA level 144 Appendix A. Survey Questionnaire Final Programmed English Main Survey Questionnaire Study Information Client University of Southern California Project Name Work from Home Survey Account Executive Sergei Rodkin Project Manager An Liu Ipsos Job Number 23-018040-01 SNO(s) 25579 - Pretest LOI 10 minutes Type of Study Ad-hoc, one shot Field Start Date (tentative is fine) Field End Date (tentative is fine) Teams Involved Enter all teams who will touch the project (e.g., Scripting, DP, Coding, IIS, Panel Relations) DP Team Scope Kickoff Meeting Date (tentative is fine) Comments Note: The study information below should be completed for all projects. Copy/paste the table into the internal project kickoff meeting invitation so all teams have it for reference. 145 Sample Variables • KP standard demographics • Xspanish • Xacslang • Xzip Quota Description Recent movers moved after Covid-19 (March 2020) Non recent movers Did not move after Covid- 19 (March 2020) Telework (any days) 500 completes(if Q1=1 and Q2=1) 500 completes(if Q1=1 and Q2=2) Full time in person 500 completes (if Q1=2 and Q2=1) 500 completes(if Q1=2 and Q2=2) Main Questionnaire (including screener, if applicable) Programming Notes: • Code all refusals as -1. • Use default instruction text for each question type unless otherwise specified. • Do not prompt on all questions. (Remove this instruction if sample is all opt-in, client list sample, or otherwise not KP.) 146 Base: All respondents Q0 [DISPLAY] Dear Participants, The expansion of remote working options has changed the way many people organize their daily activities. This survey will collect information about how this shift in work arrangements, whether you work remotely or not, has affected where you live and your daily routine. As expanded remote working becomes normalized, the consequences for how we plan cities, transportation, and access to services could be significant. The information we collect will help us inform policymakers and the public about the direction of these changes. This survey is completely anonymous and should take no longer than 5-7 minutes to complete. If you are unable to answer a question, skip to the next question. In all questions, remote work refers to the ability to work outside your employer’s physical location where you would usually commute to. Your participation is greatly appreciated and will make a significant contribution to our understanding of remote work, during a time when the initial Covid-19 shock has passed, but when work and residential location relationships are still in flux and likely adjusting. Thank you for your time and insights. 147 SCREENING QUESTIONS Base: all respondents [PPEMPLOY] QEMPLOY [Q] How many hours do you usually work for pay or profit per week? Please include hours you work for pay or profit at all your jobs if you have more than one job. If none, enter “0”. If less than an hour in a week, enter “1”. SCRIPTER: min.=0, max.=168. Show label to right of box: Hours per week. Do not allow decimals. Prompt following nonresponse. Create data only variable. IF QEMPLOY ≥ 35 PPEMPLOY = 1. IF QEMPLOY ≤ 34 AND QEMPLOY ≥ 1 PPEMPLOY = 2. IF QEMPLOY = 0 PPEMPLOY = 3. Variable name: PPEMPLOY [S] Variable Text: Current employment status Response list: 1. Working full-time 2. Working part-time 3. Not working Terminate if ppemploy=2 OR 3 148 Q1 [S] Are you currently able to work remotely? 1. Yes (any days) 2. No Q2 [S] Have you permanently changed residence since the pandemic began (March 2020)? 1. Yes 2. No Programming instruction: terminate if refused after prompt Base: Q2=1 Q2_1 [S] How many times did you move since the pandemic began (March 2020)? 1. 1 time 2. 2 times 3. 3 times 4. 4 times 5. More than 4 times Base: Q2=1 Q2_2 [DD] When was your last move? Month: [DropDown with Range January to December] Year: [DropDown with Range 2020 to 2023] Base: Q2=1 Q2_3 [S] In your last move, how far did you move? 1. Less than 20 miles 2. Over 20 miles Base: All respondents Prompt twice if refused Base: All respondents Prompt once if refused 149 Base: Q2=1 Q2_4 [S] Were you able to work remotely before your last move? 1. Yes 2. No Quota table: Please terminate if quota is full Recent movers moved after Covid-19 (March 2020) Non recent movers Did not move after Covid- 19 (March 2020) Telework (any days) 500 completes(if Q1=1 and Q2=1) 500 completes(if Q1=1 and Q2=2) Full time in person 500 completes(if Q1=2 and Q2=1) 500 completes(if Q1=2 and Q2=2) Jobs-Housing Questions Base: All respondents Q1a [Number box] What is your current home zip code: Insert NUMBER BOX with Range [00000 to 99999] Programming instruction: If in-field and profile zip codes match or QZIP is refused, create dov_match=1 If in-field and profile zip codes do not match but in-field zip codes exist in look-up table, create dov_match=2 If in-field and profile zip codes do not match but in-field zip codes do not exist in look-up table, create dov_match=3 150 If in-field and profile zip codes match or QZIP is refused, create new geocode DOVs as follow: • State (numeric) DOV = ppstaten • Metropolitan status DOV = ppmsacat • Census division DOV = re-code from ppstaten • Census region DOV = re-code from ppstaten If in-field and profile zip codes do not match but in-field zip codes exist in look-up table, set new geocode DOVs = geocode variables from the look-up table • State (numeric) DOV = state (numeric code) from look-up • Metropolitan status DOV = metropolitan status from look-up • Census division DOV = re-code from state (numeric code) from look-up • Census region DOV = re-code from state (numeric code) from look-up • Re-code ppreg4 from ppreg9 from state from look up table Recode ppreg9 from state from look-up table if (state ge 11 and state le 16) ppreg9=1. if (state ge 21 and state le 23) ppreg9=2. if (state ge 31 and state le 35) ppreg9=3. if (state ge 41 and state le 47) ppreg9=4. if (state ge 51 and state le 59) ppreg9=5. if (state ge 61 and state le 64) ppreg9=6. if (state ge 71 and state le 74) ppreg9=7. if (state ge 81 and state le 88) ppreg9=8. if (state ge 91 and state le 95) ppreg9=9. • • If in-field and profile zip codes do not match and in-field zip codes do not exist in zip-level crosswalk -> use profile data for all geocode variables • State (numeric) DOV = ppstaten • Metropolitan status DOV = ppmsacat • Census division DOV = re-code from ppstaten • Census region DOV = re-code from ppstaten ppreg9 ppreg4 1 1 2 1 3 2 4 2 5 3 6 3 7 3 8 4 9 4 151 Base: All respondents Q2a [Number box] What was your home zip code before the pandemic began (March 2020): Insert NUMBER BOX with Range [00000 to 99999] Base: Q2=1 Q2b [Ranking] What are the top 3 reasons that motivated your move? (Up to 3 reasons, Ranked 1, 2 and 3, with #1 being the top reason) Programming instruction: See this link - Preview - Online Survey Software | Qualtrics Survey Solutions (ipsossay.com). This is the most natural way in Qualtrics for sorting/ranking statements based on the drag and drop functionality. Please only show 3 numbers and hide the others. Scripter: Please allow respondents to select up to 3 responses 1. Able to work remotely 2. 12. Move closer to family 3. New job or job transfer 4. To look for work or lost job 5. To establish own household / Change in marital status 6. Wanted easier commute 7. Wanted newer / better / larger house or apartment 8. Wanted lower priced housing 9. Wanted better / safer neighborhood 10.Wanted better schools / environment for kids 11.Health related reasons 12.Other (Please specify): [TEXTBOX] Please drag-and-drop your preferred rankings that motivated your move. (Ranked 1, 2, and 3, with #1 being the top reason). 152 Base: All respondents Q2_2_1 [Number box] How large is your current home? (Please provide your best estimate in square feet) Insert NUMBER BOX (square feet) Base: Q2=1 Q2_2_2 [Number box] How large was your home before the pandemic began (March 2020)? (Please provide your best estimate in square feet) Insert NUMBER BOX (square feet) Base: All respondents Q2_2_3 [S] What is the total number of bedrooms in your current place of residence? 1. 0 bedrooms 2. 1 bedroom 3. 2 bedrooms 4. 3 bedrooms 5. 4 bedrooms 6. 5 or more bedrooms Base: Q2=1 Q2_3_4 [S] What is the total number of bedrooms in your previous place of residence before the pandemic began (March 2020)? 1. 0 bedrooms 2. 1 bedroom 3. 2 bedrooms 4. 3 bedrooms 5. 4 bedrooms 6. 5 or more bedrooms 153 Base: All respondents Q3 [Number box] What is your current job zip code: (either the location where you work in person or the location of the office where you would report to work if you primarily or always work remotely) Insert NUMBER BOX with Range [00000 to 99999] Base: All respondents Q4 [S] How long have you worked for your current employer? 1. Less than 1 year 2. 1 - 3 years 3. 3 - 5 years 4. 5 - 10 years 5. More than 10 years Base: All respondents Q4_1 [S] How many days per week are you allowed to work remotely in your current job? 1. 0 days (Not allowed to work remotely) 2. 1 day 3. 2 days 4. 3 days 5. 4 days 6. 5 or more days (Fully remote) Base: All respondents Q4_2 [S] How many days per week do you actually work remotely in your current job? 1. 0 days OR not allowed to work remotely 2. 1 day 3. 2 days 4. 3 days 5. 4 days 6. 5 or more days (Fully remote) Base: if Q4_2=2-5 Q4_2_1 [MP] 154 On which day(s) of the week do you typically work remotely? (multiple choices) 1. Monday 2. Tuesday 3. Wednesday 4. Thursday 5. Friday 6. Varies from week to week [S] Base: All respondents Q5 [S] Have you changed jobs since the pandemic began (March 2020)? 1. No. Same position, same employer. 2. Yes. Different position, same employer. 3. Yes. Similar position, different employer. 4. Yes. Different position, different employer. Base: Q5=2,3,4 Q5_1 [S] How many times did you change jobs since the pandemic began (March 2020)? 1. 1 time 2. 2 times 3. 3 times 4. 4 times 5. More than 4 times When was the last time you changed jobs? Month: [DropDown with Range January (1 ... December (12)] Year: [DropDown with Range 2020 (1 ... 2023 (4)] Base: Q5=2,3,4 Q5_2 [DD] 155 What are the top 3 reasons that motivated your job change? (Up to 3 reasons, Ranked 1, 2 and 3, with #1 being the top reason) Programming instruction: See this link - Preview - Online Survey Software | Qualtrics Survey Solutions (ipsossay.com). This is the most natural way in Qualtrics for sorting/ranking statements based on the drag and drop functionality. Please only show 3 numbers and hide the others. Scripter: Please allow respondents to select up to 3 responses 1. Better pay / benefits 2. Better career opportunity 3. Ability to work remotely 4. Loss of previous job 5. Residence relocation 6. Easier commute 7. Other (Please specify): [TEXTBOX] Q5_3_b Please drag-and-drop your preferred rankings that motivated your job change. (Ranked 1, 2, and 3, with #1 being the top reason). Q6 [Number box] What zip code did you work in most often before the pandemic began (March 2020): Insert NUMBER BOX with Range [00000 to 99999] Q6_1 [S] How many days per week were you allowed to work remotely in your job before March 2020? 1. 0 days (Not allowed to work remotely) 2. 1 day 3. 2 days 4. 3 days 5. 4 days 6. 5 or more days (Fully remote) How many days per week did you actually work remotely in your job before March 2020? Base: Q5=2,3,4 Q5_3 [Ranking] Base: Q5=2,3,4 Base: all respondents Base: all respondents Q6_2 [S] 156 1. 0 days OR not allowed to work remotely 2. 1 day 3. 2 days 4. 3 days 5. 4 days 6. 5 or more days (Fully remote) Base: if Q6_2=2-5 Q6_2_1 [MP] On which day(s) of the week did you typically work remotely before March 2020? (multiple choices) 1. Monday 2. Tuesday 3. Wednesday 4. Thursday 5. Friday 6. Varies from week to week [S] Commute Questions Base: Q4_2=1 to 5 Q7 [S] On days you go to work (for your current job), you commute by: 1. Walking / biking 2. Car, by yourself 3. Car, with others (a carpool) 4. Ride share (Uber, Lyft), taxi, or vanpool 5. Bus 6. Train 7. Other (Please specify): [TEXTBOX] 157 Base: All respondents Q7_1 [S] Did your commute mode change, compared with pre-pandemic (before March 2020)? 1. Yes 2. No Base: Q7-1=1 Q7_1_1 [S] Pre-pandemic (before March 2020), on days you go to work, you commuted by: 1. Walking / biking 2. Car, single occupant (only yourself) 3. Car, multiple occupants (a carpool) 4. Ride share (Uber, Lyft), taxi, or vanpool 5. Bus 6. Train Base: Q7=2 or Q7=3 Q7_2 [Drop down] What is the car make/model/year you use for your current commute most often? 1. Make [Drop down] 2. Model [Drop down] 3. Year [Number box with range 1900 to 2023] Base: Q7=3 Q7_3 [S] How many people do you usually carpool with? 1. 1 2. 2 3. 3 4. 4 5. 5 or more 158 Base: Q7=1, 4, 5, 6 Q7_4 [S] Is a car available to you (on typical circumstances for daily use, not just for commute)? 1. Yes 2. No Base: Q7_4=1 Q7_4_1 [TEXTBOX, NUMBER BOX] What is the car make/model/year for your daily use? 1. Make [Drop down] 2. Model [Drop down] 3. Year [Number box with range 1900 to 2023] Base: All respondents Q8_1 [S, Grid] How often do you drive to the following locations, in comparison to before the pandemic (before March 2020)? Statement in rows: 1. Driving for errands 2. Driving to the grocery store 3. Driving kids to school, activities, and events 4. Driving to recreational locations (beach, parks… etc.) 5. Driving to social events or activities Answers in columns: 1. Much more often 2. Somewhat more often 3. About same 4. Somewhat less often 5. Much less often 159 Base: All respondents Q8-2 [Grid, Accordion] Did you drive greater or lesser distance to the following locations, in comparison to before the pandemic (before March 2020)? Statement in rows: 1. Driving for errands 2. Driving to the grocery store 3. Driving kids to school, activities, and events 4. Driving to recreational locations (beach, parks… etc.) 5. Driving to social events or activities Answers in columns: 1. Much farther 2. Somewhat farther 3. About same 4. Somewhat shorter 5. Much shorter Base: All respondents Q9 [S] Do you anticipate working remotely a year from now? 1. Yes, I anticipate working remotely next year about the same amount as I work remotely now 2. Yes, I anticipate beginning or increasing remote work by next year. 3. Yes, but I anticipate reducing remote work by next year. 4. No, I anticipate not working remotely next year. 5. Unsure / Don't know 160 Base: Q9=2 Q9_1_1 [S] Why do you anticipate increasing your remote work time by next year? 1. I anticipate changing jobs to one which allows greater flexibility. 2. My employer will allow remote work indefinitely going forward. 3. Other (Please specify): [TEXTBOX] Base: Q9=3 Q9_1_2 [S] Why do you anticipate reducing remote work by next year? 1. My employer will require more in-person/office work. 2. I prefer in-person/office work. 3. Other (Please specify): [Textbox] Base: Q9=4 Q9_1_3 [S] Why do you anticipate not working remotely next year? 1. I don’t work remotely now. 2. My employer will require in-person/office work by next year. 3. I prefer in-person/office work. 4. Other (Please specify): [TEXTBOX] Base: All respondents Q9_2_1 [S] Do you anticipate moving within the next year? 1. Yes 2. No Base: Q9_2_1=1 Q9_2_2 [S] Where do you plan to move in the coming months or year? 1. Place that is closer to the office 2. Place that is farther away from office 3. Other (Please specify): [TEXTBOX] 161 Final Programmed Spanish Main Survey Questionnaire Study Information Client University of Southern California Project Name Work from Home Survey Account Executive Sergei Rodkin Project Manager An Liu Ipsos Job Number 23-018040-01 SNO(s) 25579 - Pretest LOI 10 minutes Type of Study Ad-hoc, one shot Field Start Date (tentative isfine) Field End Date (tentative is fine) Teams Involved Enter all teams who will touch the project (e.g., Scripting, DP, Coding, IIS, Panel Relations) DP Team Scope Enter DP requirements here (e.g., data clean, banner tables, client SPSS dataset, etc.) Kickoff Meeting Date (tentative is fine) Enter kickoff meeting date here Comments Note: The study information below should be completed for all projects. Copy/paste the table into the internal project kickoff meeting invitation so all teams have it for reference. 162 Sample Variables • KP standard demographics • Xspanish • Xacslang • Xzip Quota Description Recent movers moved after Covid-19 (March 2020) Non recent movers Did not move after Covid- 19 (March 2020) Telework (any days) 500 completes(if Q1=1 and Q2=1) 500 completes(if Q1=1 and Q2=2) Full time in person 500 completes(if Q1=2 and Q2=1) 500 completes(if Q1=2 and Q2=2) Main Questionnaire (including screener, if applicable) Programming Notes: • Code all refusals as -1. • Use default instruction text for each question type unless otherwise specified. • Do not prompt on all questions. (Remove this instruction if sample is all opt-in, client list sample, or otherwise not KP.) 163 Base: All respondents Q0 [DISPLAY] Estimados participantes, El aumento de opciones de trabajo remoto ha cambiado la forma en la que muchas personas organizan sus actividades cotidianas. Esta encuesta recopilará información sobre cómo este cambio en los preparativos laborales, ya sea que se trabaje de forma remota o no, ha afectado su lugar de residencia y su rutina diaria. A medida que se amplie y normalice la modalidad de trabajo remoto , las consecuencias en la forma en que planificamos ciudades, transporte y acceso a servicios podrían ser significativas. La información recopilada nos ayudará a informar a legisladores de políticas y al público sobre la dirección de estos cambios. Esta encuesta es anónima y no debe tardar más de 5-7 minutos en completarse. Si usted no puede responder a una pregunta, por favor continue a la siguiente pregunta. En todas las preguntas de esta encuesta, se refiere a trabajo remoto a la capacidad de trabajar en una ubicación externa que no sea la ubicación física de su empleo, ubicación a la que normalmente se trasladaría oviajaría. Se aprecia su participación en esta encuesta, su contribución será significativa para nuestra comprensión de trabajo rem durante un periodo en el que el shock inicial de Covid-19 ha pasado, pero en el cual las relaciones laborales y la ubicación residencial de empleados continua cambiando y es muy probablemente se ajusten. Gracias por su tiempo y su perspectiva sobre el tema. 164 SCREENING QUESTIONS Base: all respondents [PPEMPLOY] QEMPLOY [Q] ¿Cuántas horas suele usted trabajar por un salario o ganancias obtenidas semanales? Por favor incluya las horas que trabaja usted por un salario o ganancias obtenidas en todos sus empleos si usted tiene más de un empleo. Si en estos momentos usted esta desempleado, introduzca "0". Si esta laborandomenos de una hora en una semana, ingrese "1". SCRIPTER: min.=0, max.=168. Show label to right of box: Hours per week. Do not allow decimals. Prompt following nonresponse. Create data only variable. IF QEMPLOY ≥ 35 PPEMPLOY = 1. IF QEMPLOY ≤ 34 AND QEMPLOY ≥ 1 PPEMPLOY = 2. IF QEMPLOY = 0 PPEMPLOY = 3. Variable name: PPEMPLOY [S] Variable Text: Current employment status Response list: 1. Trabajando tiempo completo 2. Trabajando medio tiempo 3. No está trabajando o desempleado Terminate if ppemploy=2 OR 3 165 Q1 [S] ¿Actualmente puede usted trabajar de forma remota ? 3. Sí (cualquier día de la semana) 4. No Q2 [S] ¿ Desde el comienzó de la pandemia (Marzo de 2020) usted cambio permanentemente de residencia? 1. Sí 2. No Programming instruction: terminate if refused after prompt Base: Q2=1 Q2_1 [S] ¿ Desde que comenzó la pandemia (Marzo de 2020) cuántas veces se ha mudado usted? 1. 1 vez 2. 2 veces 3. 3 veces 4. 4 veces 5. Más de 4 veces Base: Q2=1 Q2_2 [DD] ¿Cuándo fue su última mudanza? Mes: [DropDown with Range Enero to Diciembre] Año: [DropDown with Range 2020 to 2023] Base: All respondents Prompt twice if refused Base: All respondents Prompt once if refused 166 Base: Q2=1 Q2_3 [S] En su última mudanza, ¿Cuál es la distancia que usted se mudó? 1. Menos de 20 millas 2. Más de 20 millas Base: Q2=1 Q2_4 [S] ¿ Antes de su última mudanza pudo usted trabajar de forma remota? 1. Sí 2. No Quota table: Please terminate if quota is full Recent movers moved after Covid-19 (March 2020) Non recent movers Did not move after Covid- 19 (March 2020) Telework (any days) 500 completes(if Q1=1 and Q2=1) 500 completes(if Q1=1 and Q2=2) Full time in person 500 completes(if Q1=2 and Q2=1) 500 completes(if Q1=2 and Q2=2) Jobs-Housing Questions Base: All respondents Q1a [Number box] ¿Cuál es el código postal de su residencia o vivienda? Insert NUMBER BOX with Range [00000 to 99999] 167 Programming instruction: If in-field and profile zip codes match or QZIP is refused, create dov_match=1 If in-field and profile zip codes do not match but in-field zip codes exist in look-up table, create dov_match=2 If in-field and profile zip codes do not match but in-field zip codes do not exist in look-up table, create dov_match=3 If in-field and profile zip codes match or QZIP is refused, create new geocode DOVs as follow: • State (numeric) DOV = ppstaten • Metropolitan status DOV = ppmsacat • Census division DOV = re-code from ppstaten • Census region DOV = re-code from ppstaten If in-field and profile zip codes do not match but in-field zip codes exist in look-up table, set new geocode DOVs = geocode variables from the look-up table • State (numeric) DOV = state (numeric code) from look-up • Metropolitan status DOV = metropolitan status from look-up • Census division DOV = re-code from state (numeric code) from look-up • Census region DOV = re-code from state (numeric code) from look-up • Re-code ppreg4 from ppreg9 from state from look up table Recode ppreg9 from state from look-up table if (state ge 11 and state le 16) ppreg9=1. if (state ge 21 and state le 23) ppreg9=2. if (state ge 31 and state le 35) ppreg9=3. if (state ge 41 and state le 47) ppreg9=4. if (state ge 51 and state le 59) ppreg9=5. if (state ge 61 and state le 64) ppreg9=6. if (state ge 71 and state le 74) ppreg9=7. if (state ge 81 and state le 88) ppreg9=8. if (state ge 91 and state le 95) ppreg9=9. ppreg9 ppreg4 1 1 2 1 3 2 4 2 5 3 6 3 7 3 8 4 9 4 168 If in-field and profile zip codes do not match and in-field zip codes do not exist in zip-level crosswalk -> use profile data for all geocode variables • State (numeric) DOV = ppstaten • Metropolitan status DOV = ppmsacat • Census division DOV = re-code from ppstaten • Census region DOV = re-code from ppstaten Base: All respondents Q2a [Number box] ¿Cuál era el código postal de su hogar o vivienda antes de que comenzara la pandemia (Marzo de 2020)? Insert NUMBER BOX with Range [00000 to 99999] Base: Q2=1 Q2b [Ranking] ¿Cuáles son las 3 razones principales que motivaron su ultima mudanza? (Por favor enumere hasta 3 razones, Clasificando 1, 2 y 3, siendo #1 la razón principal) Programming instruction: See this link - Preview - Online Survey Software | Qualtrics Survey Solutions (ipsossay.com). This is the most natural way in Qualtrics for sorting/ranking statements based on the drag and drop functionality. Please only show 3 numbers and hide the others. Scripter: Please allow respondents to select up to 3 responses 13.Poder trabajar de forma remota 14.12. Mudarse para estar más cerca de la familia 15.Nuevo empleo o translado laboral 16.Para buscar empleo o por perdida de empleo 17.Para establecer su propio hogar /Cambio en su estado civil/marital 18.Quería acceso o unviaje más fácil 19.Quería una casa o un departamento nuevo/mejor/más grande 20.Quería una vivienda económicamente accesible o de mejor precio 21.Quería una mejor colonia más segura 22.Quería acceso a mejores escuelas/entorno para los niños 23.Razones relacionadas con problemas de la salud 24.Otra razón (Por favor especifique): [TEXTBOX] Por favor arrastre y suelte sus clasificaciones preferidas que motivaron su mudanza. (Clasificando 1, 2 y 3, siendo #1 la razón principal). 169 Base: All respondents Q2_2_1 [Number box] ¿Cuál es la extensión de su actual hogar? (Por favor proporcione una estimación en pies cuadrados) Insert NUMBER BOX (pies cuadrados) Base: Q2=1 Q2_2_2 [Number box] ¿Qué tan grande era su hogar antes de que comenzara la pandemia (Marzo de 2020)? (Por favor proporcione una estimación en pies cuadrados) Insert NUMBER BOX (pies cuadrados) Base: All respondents Q2_2_3 [S] ¿ En su residencia actual, cuál es el número total de habitaciones? 1. 0 habitaciones 2. 1 habitación 3. 2 habitaciones 4. 3 habitaciones 5. 4 habitaciones 6. 5 o más habitaciones 170 Base: Q2=1 Q2_3_4 [S] ¿ Antes de que comenzara la pandemia (Marzo de 2020) en su lugar de residencia anterior, cuál era el número total de habitaciones? 1. 0 habitaciones 2. 1 habitación 3. 2 habitaciones 4. 3 habitaciones 5. 4 habitaciones 6. 5 o más habitaciones Base: All respondents Q3 [Number box] ¿Cuál es el código postal de su actual empleo? (Este puede ser la ubicación donde trabaja usted de modo presencial o la ubicación de la oficina donde se reportaría usted aa trabajar, si usted siempre trabaja de forma remota) Insert NUMBER BOX with Range [00000 to 99999] Base: All respondents Q4 [S] ¿Cuánto tiempo ha trabajado para su actual empleo? 1. Menos de 1 año 2. De 1 a 3 años 3. De 3 a 5 años 4. De 5 a 10 años 5. Más de 10 años 171 Base: All respondents Q4_1 [S] ¿ En su empleo actual cuántos días de la semana se le permite trabajar de forma remota? 1. 0 días (No le es permitido trabajar de forma remota) 2. 1 día 3. 2 días 4. 3 días 5. 4 días 6. 5 o más días (Tiempo completo de forma remota) Base: All respondents Q4_2 [S] ¿ Actualmente cuántos días a la semana trabaja ustedde forma remota en su actual empleo? 1. 0 días (No le es permitido trabajar de forma remota) 2. 1 día 3. 2 días 4. 3 días 5. 4 días 6. 5 o más días (Tiempo completo de forma remota) Base: if Q4_2=2-5 Q4_2_1 [MP] ¿En qué día(s) de la semana suele usted trabajar de forma remota? (múltiples opciones) 1. Lunes 2. Martes 3. Miércoles 4. Jueves 5. Viernes 6. Varía de una semana a otra [S] 172 Base: All respondents Q5 [S] ¿Ha cambiado usted de empleo desde que comenzó la pandemia (Marzo de 2020)? 5. No. Misma posición, misma compañía. 6. Sí. Una posición diferente, misma compañía . 7. Sí. Una posición similar, diferente compañía . 8. Sí. Diferente posición, diferente compañía . Base: Q5=2,3,4 Q5_1 [S] ¿ Desde que comenzó la pandemia cuántas veces ha cambiado de empleo (Marzo de 2020)? 1. 1 vez 2. 2 veces 3. 3 veces 4. 4 veces 5. Más de 4 veces ¿Cuándo fue la última vez que cambió de empleo? Mes: [DropDown with Range Enero (1 ... Diciembre (12)] Año: [DropDown with Range 2020 (1 ... 2023 (4)] Base: Q5=2,3,4 Q5_2 [DD] 173 ¿Cuáles son las 3 razones principales que motivaron su cambio de empleo? (Por favor enumere hasta 3 razones, Clasificando 1, 2 y 3, siendo #1 la razón principal) Programming instruction: See this link - Preview - Online Survey Software | Qualtrics Survey Solutions (ipsossay.com). This is the most natural way in Qualtrics for sorting/ranking statements based on the drag and drop functionality. Please only show 3 numbers and hide the others. Scripter: Please allow respondents to select up to 3 responses 1. Un mejor salario/beneficios 2. Una mejor oportunidad laboral 3. La posibilidad de trabajar de forma remota 4. Pérdida de su empleo anterior 5. La reubicación de su residencia 6. El traslado o viaje más fácil a su nuevo empleo 7. Otra razón (Por favor especifique): [TEXTBOX] Q5_3_b Por favor arrastre y suelte sus clasificaciones preferidas que motivaron su cambio de empleo. (Clasificando 1, 2 y 3, siendo # 1 la razón principal). Q6 [Number box] ¿ Antes de que comenzara la pandemia (Marzo de 2020) en qué código postal trabajó usted con más frecuencia? Insert NUMBER BOX with Range [00000 to 99999] Base: Q5=2,3,4 Q5_3 [Ranking] Base: Q5=2,3,4 174 All respondents Q6_1 [S] ¿ Antes de marzo de 2020 cuántos días a la semana se le permitía trabajar de forma remota en su empleo? 1. 0 días (No se le permitía trabajar de forma remota) 2. 1 día 3. 2 días 4. 3 días 5. 4 días 6. 5 o más días (Tiempo completo de forma remota) All respondents Q6_2 [S] ¿ Antes de marzo de 2020 cuántos días a la semana trabajaba realmente de forma remota en su trabajo? 1. 0 días (No se le permitió trabajar de forma remota) 2. 1 día 3. 2 días 4. 3 días 5. 4 días 6. 5 o más días (Tiempo completo de forma remota) Base: if Q6_2=2-5 Q6_2_1 [MP] ¿Antes de marzo de 2020 en qué día(s) de la semana solía usted trabajar de forma remota? (múltiples opciones) 1. Lunes 2. Martes 3. Miércoles 4. Jueves 5. Viernes 6. Variaba de una semana a otra [S] Commute Questions 175 Base: Q4_2=1 to 5 Q7 [S] En su actual empleo, os días que usted va a trabajar, como se traslada o viaja usted? : 1. Caminando o en bicicleta 2. En coche, individualmente (usted solo/a) 3. En coche, con otros ocupantes (coche compartido o carpool) 4. En viaje compartido (Uber, Lyft), taxi o vanpool 5. En autobús 6. En tren 7. Otro modo de transporte (Por favor especifique): [TEXTBOX] Base: All respondents Q7_1 [S] ¿Cambió usted su modo de traslado o viaje en comparación con su modo de transporte antes de la pandemia (antes de Marzo de 2020)? 1. Sí 2. No Base: Q7-1=1 Q7_1_1 [S] Antes de la pandemia (antes de Marzo de 2020), los días que iba a trabajar, cual era su modo de transporte para llegar a su lugar de empleo?: 1. Caminando o en bicicleta 2. En coche, individualmente (usted solo/a) 3. En coche, con otros ocupantes (coche compartido o carpool) 4. En viaje compartido (Uber, Lyft), taxi o vanpool 5. En autobús 6. En tren 7. Otro modo de transporte (Por favor especifique): [TEXTBOX] 176 Base: Q7=2 or Q7=3 Q7_2 [Drop down] ¿Actualmente cuál es la marca/modelo/año del coche que usted utiliza con más frecuencia para su traslado o viaje ? 1. Marca [Drop down] 2. Modelo [Drop down] 3. Año [Number box with range 1900 to 2023] Base: Q7=3 Q7_3 [S] ¿Con cuántas personas suele usted compartir su viaje en coche (carpool)? 1. 1 persona 2. 2 personas 3. 3 personas 4. 4 personas 5. 5 o más personas Base: Q7=1, 4, 5, 6 Q7_4 [S] ¿Usted tiene disponibilidad de un automóvil ( para uso diario, no solo para trasladarse o viajar a su trabajo)? 1. Sí 2. No Base: Q7_4=1 Q7_4_1 [TEXTBOX, NUMBER BOX] ¿Para uso diario cuál es la marca/modelo/año del coche que usted utiliza? 1. Marca [Drop down] 2. Modelo [Drop down] 3. Año [Number box with range 1900 to 2023] 177 Base: All respondents Q8_1 [S, Grid] ¿Con qué frecuencia conduce usted a alguno de los siguientes lugares, en comparación con la frecuencia que usted conducía a estos lugares antes de la pandemia (antes de Marzo de 2020)? 1. Conduce usted para hacer mandados 2. Conduce a la tienda de comestibles 3. Lleva usted a los niños a la escuela, actividades y eventos 4. Conduce usted a lugares recreativos (playa, parques... etc.) 5. Conduce usted a eventos o actividades sociales 1. Mucho más a menudo 2. Un poco más a menudo 3. Más o menos lo mismo 4. Un poco menos a menudo 5. Mucho menos a menudo Q8-2 [Grid, Accordion] ¿Actualmente conduce usted una distancia más lejana o más cercana a alguno de los siguientes lugares, en comparación con la distancia que usted conducía antes de la pandemia (antes de Marzo de 2020)? 1. Conduce usted para hacer mandados 2. Conduce a la tienda de comestibles 3. Lleva usted a los niños a la escuela, actividades y eventos 4. Conduce usted a lugares recreativos (playa, parques... etc.) 5. Conduce usted a eventos o actividades sociales 1. Mucho más lejos 2. Un poco más lejos 3. Más o menos la misma distancia 4. Algo más cercano 5. Mucho más cercano Statement in rows: Answers in columns: Base: All respondents Statement in rows: Answers in columns: Base: All respondents 178 Q9 [S] ¿ Dentro de un año prevé usted trabajar de forma remota? 1. Sí, el próximo año anticipo que trabajare de forma remota aproximadamente la misma cantidad de tiempo que trabajo de forma remota actualmente. 2. Sí, anticipo que comenzara o aumentara mi trabajo remoto para el próximo año. 3. Sí, anticipo que se reducirá mi tiempo de trabajo remoto para el próximo año. 4. No, anticipo que no trabajare de forma remota el próximo año. 5. No estoy seguro/a. No lo sé Base: Q9=2 Q9_1_1 [S] ¿Cuál es la razón por la cual usted anticipa que aumentara su tiempo de trabajo remoto para el próximo año? 1. Anticipo cambiar de trabajo a uno que permita una mayor flexibilidad. 2. En un futuro cercano mi empleo permitirá el trabajo remoto de forma indefinida . 3. Otra razón (Por favor especifique): [TEXTBOX] Base: Q9=3 Q9_1_2 [S] ¿Por qué prevé usted que se reducira el tiempo de su trabajo remoto para el próximo año? 1. Mi actual empleo requerirá más trabajo en persona o en la oficina. 2. Prefiero el trabajo en persona o en la oficina. 3. Otra razón (Por favor especifique): [TEXTBOX] 179 Base: Q9=4 Q9_1_3 [S] ¿Por qué anticipa usted que no trabajara de forma remota el próximo año? 5. Actualmente no trabajo de forma remota. 6. Para el próximo año mi actual empleo requerirá más trabajo en persona o en la oficina. 7. Prefiero el trabajo en persona o en la oficina. 8. Otra razón (Por favor especifique): [TEXTBOX] Base: All respondents Q9_2_1 [S] ¿Anticipa usted mudarse dentro del próximo año? 1. Sí 2. No Base: Q9_2_1=1 Q9_2_2 [S] ¿A dónde planea usted mudarse en los próximos meses o próximo año? 1. A un lugar que este más cerca de la oficina 2. A un lugar que está más lejos de la oficina 3. Otra razón (Por favor especifique): [TEXTBOX] 180 Appendix B. Weighting Benchmark Distributions Below is the detailed weighting procedure quoted directly from IPSOS methodology documents (IPSOS, 2022).22 1. In the first step, design weights for KnowledgePanel (KP) assignees are computed to reflect their modeled selection probabilities. 2. The above design weights for respondents who reported working full-time are adjusted to align with the geodemographic distributions of the full-time employed population aged 18 and over using an iterative proportional fitting (raking) procedure. Geodemographic benchmarks are sourced from the 2023 March Supplement of the CPS, except for language dominance within Hispanics, which is sourced from the 2021 ACS. The dimensions used for weighting include: • Gender (Male/Female) by age (18–29, 30–44, 45–59, and 60+) • Race/Hispanic ethnicity (White/Non-Hispanic, Black/Non-Hispanic, Other or 2+ Races/Non-Hispanic, Hispanic) • Census Region (Northeast, Midwest, South, West) by Metropolitan Status (Metro, Non-Metro) • Education (Less than High School, High School, Some College, Bachelor and beyond) • Household income (under $10k, $10K to <$25k, $25K to <$50k, $50K to <$75k, $75K to <$100k, $100K to <$150k, and $150K+) • Language Dominance (non-Hispanic and English Dominant, Bilingual, and Spanish Dominant Hispanic) when survey is administered in both English and Spanish The resulting weights are trimmed and scaled to match the number of screened full-time employed respondents. 3. Next, all screened full-time employed respondents belonging to specific worker subgroups are isolated, and benchmarks for the respective populations are created using their screener weight. These benchmarks are used to weight the final qualified respondents. Four worker subgroups: • Telework - Recent movers • Telework - Non recent movers • Full time - Recent movers • Full time - Non recent movers 22 A full and detailed overview of the IPSOS methodology is available at: https://www.ipsos.com/enus/solutions/public-affairs/knowledgepanel, accessed 01/16/2024 181 4. In the final step, screener weights for final qualified respondents are raked to align with weighted geodemographic distributions of the full-time employed population aged 18 and over, ensuring that final qualified respondents within each subgroup are representative of their respective population, and the subgroups are weighted proportionally. The dimensions for weighting include: • Gender (Male, Female) by Telework (Yes, No) by Recent Mover (Yes, No) • Age (18-29, 30-44, 45-59, 60+) by Telework (Yes, No) by Recent Mover (Yes, No) • Race-Ethnicity (White/Non-Hispanic, Black/Non-Hispanic, Other or 2+/NonHispanic, Hispanic) by Telework (Yes, No) by Recent Mover (Yes, No) • Census Region (Northeast, Midwest, South, West) by Telework (Yes, No) by Recent Mover (Yes, No) • Metropolitan Status (Metro, Non-Metro) by Telework (Yes, No) by Recent Mover (Yes, No) • Education (Less than High School, High School, Some College, Bachelor or higher) by Telework (Yes, No) by Recent Mover (Yes, No) ** collapse LS/HS within Telework and Full time - Recent movers • Household Income (under $25K, $25K-$49,999, $50K-$74,999, $75K-$99,999, $100K-$149,999, $150K and over) by Telework (Yes, No) by Recent Mover (Yes, No) ** collapse under $50 K within Telework • Language Dominance (English Dominant or Spanish Dominant Hispanic, Bilingual Hispanic, Non-Hispanic) by Telework (Yes, No) by Recent Mover (Yes, No) The resulting weights are trimmed and scaled to align with the number of qualified respondents. Detailed information on the demographic distributions of the national benchmarks can be found in Appendix G. The analysis presented from this point onward is based on weighted results. We use Analytic Weights for our analysis, where each observation is treated as the mean of a group corresponding to its assigned weight. 182 18+ full-time employed Population Benchmarks Detailed weighting distribution quoted directly from IPSOS methodology (IPSOS, 2022).23 Table B1. Gender by Age Distribution Gender by age Frequency Percent 18-29 Male 14,780,067 11.18 18-29 Female 11,777,123 8.91 30-44 Male 26,976,804 20.4 30-44 Female 21,099,799 15.96 45-59 Male 22,758,081 17.21 45-59 Female 18,362,394 13.89 60+ Male 9,569,033 7.24 60+ Female 6,885,083 5.21 Table B2. Race-Ethnicity Distribution Race-Ethnicity Frequency Percent White, Non-Hispanic 78,517,043 59.39 Black, Non-Hispanic 16,329,127 12.35 Other, Non-Hispanic 10,625,238 8.04 Hispanic 24,706,283 18.69 2+ Races, Non-Hispanic 2,030,694 1.54 Table B3. Education Distribution Education Frequency Percent Less than HS 7,820,383 5.92 HS 34,301,567 25.95 Some college 32,377,201 24.49 Bachelor or higher 57,709,234 43.65 Table B4. Income Distribution Income Frequency Percent Under $25,000 3,822,227 2.89 $25,000-$49,999 13,269,333 10.04 $50,000-$74,999 18,846,781 14.26 $75,000-$99,999 18,871,264 14.27 $100,000-$149,999 30,898,053 23.37 $150,000 and over 46,500,726 35.17 23 A full and detailed overview of the IPSOS methodology is available at: https://www.ipsos.com/enus/solutions/public-affairs/knowledgepanel, accessed 01/16/2024 183 Table B5. Language Dominance Distribution Language Dominance percent English Dominant Hispanic 5.2 Bilingual Hispanic 10.41 Spanish Dominant Hispanic 3.08 Non-Hispanic 81.31 184 Appendix C. Detailed and additional summary data Vehicle greenhouse gas emissions categories Table C1. Vehicle GHG Category matching with original EPA Rating Our GHG Category EPA Rating MPG (gas) CO₂ (g/mile) Low 10 >=92 0-97 9 59-91 98-152 8 43-58 153-209 7 34-42 210-265 Medium 6 28-33 266-323 5 22-27 324-413 4 18-21 414-508 High 3 16-17 509-573 2 14-15 574-658 1 <=13 >=659 Data source: https://www.epa.gov/greenvehicles/greenhouse-gas-rating Moving Frequency after COVID-19 Table C2. Moving Frequency after COVID outbreak in March 2020 Moving frequency % 1 time 69% 2 times 23% 3 times 5% 4 times 2% More than 4 times 1% Total 550 (100%) Table C3. Moving frequency by current work arrangement Work arrangement Move frequency Mean 25% 50% 75% Remote 1.42 1 1 2 Hybrid 1.36 1 1 2 In-person 1.37 1 1 2 185 Home-to-job Distance Table C4. Home-to-job distance (miles) by change in work arrangements from pre-COVID to post-COVID Change in work arrangement mean 25% 50% 75% 90% Pre-COVID Post-COVID Remote Remote 66.23 0.00 0.00 9.35 182.29 Hybrid 99.35 0.00 6.07 18.42 276.46 In-person 131.00 0.00 4.17 21.54 154.56 Hybrid Remote 51.18 0.00 9.55 17.73 35.98 Hybrid 25.53 4.23 7.88 15.34 31.35 In-person 21.31 4.06 9.12 17.12 27.97 In-person Remote 7.56 3.91 7.42 12.87 14.36 Hybrid 54.70 0.00 5.92 16.95 32.98 In-person 21.22 0.00 6.14 14.01 23.75 Detailed commute mode change from pre- to post- COVID Table C5. Commute mode change from pre-COVID to post-COVID Mode change Freq. Percent No response 7 0.33% Yes 436 20.52% No 1,681 79.15% Total 2,124 100% 186 Table C6. Commute mode and work arrangement from pre-COVID to post-COVID commute mode Pre-COVID Post-COVID Hybrid In-person Hybrid In-person Walking / biking 4% 5% 4% 4% Car, single occupant (only yourself) 69% 86% 80% 89% Car, multiple occupants (a carpool) 10% 3% 5% 3% Ride share (Uber, Lyft), taxi, or vanpool 2% 0% 2% 0% Bus 4% 2% 2% 2% Train 10% 2% 6% 1% Other (Please specify): 1% 1% 2% 1% Total 204 (100%) 1,523(100%) 440 (100%) 1,196 (100%) Table C7. Commute mode from pre-COVID to post-COVID Commute mode now Fully Remote Walking / biking Car, single occupant Car, multiple occupants Ride share, taxi, or vanpool Bus Train Other Total Pre-COVID Fully Remote 5% 0% 5% 0% 0% 0% 0% 0% 11% Walking / biking 2% 0% 6% 0% 0% 0% 0% 1% 10% Car, single occupant 32% 3% 11% 4% 1% 1% 1% 0% 53% Car, multiple occupants 3% 0% 4% 1% 0% 0% 0% 0% 8% Ride share, taxi, or vanpool 0% 0% 1% 0% 0% 0% 0% 0% 1% Bus 2% 0% 3% 0% 0% 0% 0% 0% 7% Train 4% 1% 3% 0% 0% 2% 1% 0% 10% Total 48% 5% 33% 5% 1% 4% 3% 1% 100 % 187 Commute days by work arrangements Table C8. Days of commute by work arrangements from pre-COVID to post-COVID Commute Days Pre-COVID Post-COVID Remote Hybrid In-person Remote Hybrid In-person 0 100% 0% 0% 100% 0% 0% 1-2 0% 34% 0% 0% 46% 0% 3-4 0% 66% 0% 0% 54% 0% 5 0% 0% 100% 0% 0% 100% Total 244 (100%) 256 (100%) 1,605 (100%) 478 (100%) 440 (100%) 1,196 (100%) GHG emissions and frequency of commuting Table C9. GHG ranking by commute days Work arrangement Commute days GHG Ranking High GHG Total emission Medium GHG emission Low GHG emission Remote 0 11% 82% 7% 126 (100%) Hybrid 1-2 7% 80% 13% 180 (100%) 3-4 6% 85% 8% 224 (100%) In-person 5 14% 81% 5% 1,099 (100%) Total 12% 82% 7% 1,629 (100%) Table C10. Home-to-Job distance by commute days Work arrangement Commute days Mean 25% 50% 75% 90% Remote 0 100.59 0.00 0.00 18.84 197.68 Hybrid 1-2 25.90 3.57 7.89 16.77 30.73 3-4 23.87 4.06 8.74 17.15 27.97 In-person 5 21.68 0.00 6.14 14.02 23.75 188 VMT and GHG by Work Arrangement Table C11. Total weekly commute VMT by work arrangement, in total weekly miles Work arrangement Remote Hybrid In-person PreCOVID PostCOVID PreCOVID PostCOVID PreCOVID PostCOVID Mean 0 0 103.2 138.1 239.8 185.7 25% 0 0 13.1 14.8 0 0 50% 0 0 39.3 44.7 66.9 67.0 75% 0 0 101.2 92.5 144.1 143.7 90% 0 0 181.9 182.8 252.4 237.5 95% 0 0 290.8 334.6 428.0 329.8 99% 0 0 540.7 2,000.8 5,878.9 4,403.8 *Source: survey Question: Q1a, Q2a, Q4_2 Sample size: pre-Covid (1,924), post-Covid (1,953) Table C12. Total weekly commute GHG Emissions by work arrangement Work arrangement Remote Hybrid In-person PreCOVID PostCOVID PreCOVID PostCOVID PreCOVID PostCOVID Mean 0 0 37,551 53,039 79,285 73,366 25% 0 0 3,916 4,692 0 0 50% 0 0 16,962 15,136 24,730 25,432 75% 0 0 37,497 34,957 53,403 54,637 90% 0 0 66,710 68,616 106,682 97,489 95% 0 0 97,920 100,005 155,225 150,189 99% 0 0 178,480 790,301 1,670,437 1,014,180 *Source: survey Question: Q1a, Q2a, Q4_2, Q7_2 Sample size: pre-Covid (1,924), post-Covid (1,953)

Abstract (if available)

Abstract The COVID-19 pandemic has profoundly transformed work and commuting patterns, reshaping urban dynamics in the United States. This dissertation investigates these shifts through three interrelated studies, focusing on the impacts of remote work on commuting behavior, residential choices, and environmental sustainability. The first study analyzes the effects of policy interventions and the widespread adoption of remote work on commute traffic in Northern California, emphasizing the inequitable outcomes for lower-income workers and essential sectors. The second study examines how remote work has influenced job and housing locations in the Bay Area and Central Valley, noting significant traffic volume reductions and migration toward more remote residential areas. The third study explores the environmental consequences of remote work, showing that fully remote workers reduce greenhouse gas (GHG) emissions by eliminating commutes, while hybrid workers also contribute to emissions reductions by commuting less frequently. Together, these studies offer insights into the evolving nature of commuting and residential dynamics, providing guidance for equitable and sustainable urban planning in the post-pandemic era.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

The built environment, tour complexity, and active travel

PDF

The demand for reliable travel: evidence from Los Angeles, and implications for public transit policy

PDF

The impact of demographic shifts on automobile travel in the United States: three empirical essays

PDF

Urban air pollution and environmental justice: three essays

PDF

Essays on congestion, agglomeration, and urban spatial structure

PDF

The flexible workplace: regional tendencies and daily travel implications

PDF

Urban consumer amenities and their accessibility

PDF

Productive frictions and urbanism in transition: planning lessons from traffic flows and urban street life in Ho Chi Minh City, Vietnam

PDF

Household mobility and neighborhood impacts

PDF

Location of warehouses and environmental justice: Three essays

PDF

Healthy mobility: untangling the relationships between the built environment, travel behavior, and environmental health

PDF

Property and labor formalization in the age of the sharing economy: Airbnb, housing affordability, and entrepreneurship in Havana

PDF

Who learns where: understanding the equity implications of charter school reform in the District of Columbia

PDF

Environmental justice in real estate, public services, and policy

PDF

Unraveling decentralization of warehousing and distribution centers: three essays

PDF

Spatial and temporal expenditure-pricing equity of rail transit fare policies

PDF

Active travel, outdoor leisure, and neighborhood environment: path analysis, Los Angeles County

PDF

Children’s travel behavior in journeys to school

PDF

Population and employment distribution and urban spatial structure: an empirical analysis of metropolitan Beijing, China in the post-reform era

PDF

The history of autonomous vehicle development and its likely futures and consequences

Asset Metadata

Creator Wang, Bonnie S. (author)

Core Title The long-term impact of COVID-19 on commute, employment, housing, and environment in the post-pandemic era

School School of Policy, Planning and Development

Degree Doctor of Philosophy

Degree Program Urban Planning and Development

Degree Conferral Date 2024-08

Publication Date 10/09/2024

Defense Date 07/31/2024

Publisher Los Angeles, California (original), University of Southern California (original), University of Southern California. Libraries (digital)

Tag commute,COVID-19,Environment,greenhouse gas,OAI-PMH Harvest,travel behavior

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Boarnet, Marlon (committee chair), Giuliano, Genevieve (committee member), Schweitzer, Lisa (committee member)

Creator Email bonniewang222@gmail.com,wangbonn@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC11399C88Y

Unique identifier UC11399C88Y

Identifier etd-WangBonnie-13589.pdf (filename)

Legacy Identifier etd-WangBonnie-13589

Document Type Dissertation

Format theses (aat)

Rights Wang, Bonnie S.

Internet Media Type application/pdf

Type texts

Source 20241011-usctheses-batch-1218 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu