Source: MIDDLEBURY COLLEGE submitted to NRP
THE IMPACT OF THE COVID-19 PANDEMIC ON URBAN-RURAL MIGRATION SYSTEMS
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
ACTIVE
Funding Source
AFRI COMPETITIVE GRANT
Reporting Frequency
Annual
Accession No.
1029992
Grant No.
2023-67023-39288
Cumulative Award Amt.
$418,938.00
Proposal No.
2022-10198
Multistate No.
(N/A)
Project Start Date
Jul 1, 2023
Project End Date
Jun 30, 2026
Grant Year
2023
Program Code
[A1661]- Innovation for Rural Entrepreneurs and Communities
Recipient Organization
MIDDLEBURY COLLEGE
(N/A)
MIDDLEBURY,VT 05753
Performing Department
(N/A)
Non Technical Summary
This project examines the ways urban-rural migration systems have responded both spatially and temporally since the onset of the COVID-19 pandemic. We draw on unique individual level data recording the monthly home location of over 30 million mobile devices (10% of all devices in the US) to provide the first extensive analysis of the spatial and temporal character of migration responses to the COVID-19 pandemic. Tracking month-by-month changes in mobile devices' home locations from January 2019 through December 2021 will provide a spatially and temporally fine resolution picture of migration systems in the year prior to the pandemic and document the ways these systems have changed since January of 2020. The unique data allow us to study aspects of migration not possible with conventional migration data sources. The multi-stage research design focuses on two fundamental questions. RQ1 - How have migration systems responded spatially to the COVID-19 pandemic? In answering RQ1 we will examine changes in urban-rural migration flows between places along the rural-urban continuum and the characteristics of places with the most substantial migration shifts. RQ2 - How have migration systems responded temporally to the COVID-19 pandemic? Through RQ2 we will distinguish between temporary and permanent movers, identify new forms of cyclical migration between urban and rural places, and examine the characteristics of places where migration shifts have been temporary and contrast those with places with more permanent migration changes. The data allow for analysis at both county and census tract scales.
Animal Health Component
(N/A)
Research Effort Categories
Basic
100%
Applied
(N/A)
Developmental
(N/A)
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
80360992060100%
Knowledge Area
803 - Sociological and Technological Change Affecting Individuals, Families, and Communities;

Subject Of Investigation
6099 - People and communities, general/other;

Field Of Science
2060 - Geography;
Goals / Objectives
Goal #1-to better understandthe ways urban-rural migration systems have responded both spatially and temporally since the onset of the COVID-19 pandemic. We draw on unique individual level data recording the monthly home location of over 30 million mobile devices (10% of all devices in the US) to provide the first extensive analysis of the spatial and temporal character of migration responses to the COVID-19 pandemic. To achieve this goal, the project identifies the following research objectives.1. Describe how the timing and geography of migration responsesvaried throughout the first two years of the pandemic. When and where did urban-to-rural migration begin to accelerate and how did these flows change over time?2. Identify and mapplaces wheremigration responses werepermanent versus temporary.For some, the pandemic-induced migration initiated a permanent move to a new residential location while for others the migration response was temporary with an eventual return to their place of origin. Where were these temporally different migration responses more common in terms of regional location and position along the urban-rural hierarchy?3. Compare and contrast places where pandemic-induced migration was more permanent with those places where the migration was more temporary.What characterizes the places where temporary or permanent migration were more common in terms of racial, ethnic, economic, and educational composition?4. Estimate how the pandemic-induced migration is leading to population redistribution across the urban hierarchy. Which types of places have seen the largest relative increases/decreases in mobile devices since January of 2020?5. Identify geographically unique migration systems that have developed since the onset of the COVID-19 pandemic in early January of 2020. How do urban-to-rural flows originating from New York City differ from those leaving Chicago or San Francisco?Goal #2 - enhance our spatial understanding of urban-rural migration processes. Given data constraints, most scholarship to date focusing on urban-rural migration dynamics uses counties as the unit of analysis. The unique data set used in this project allows for migration analysis at the sub-county (census tract) level, so the tract-based analysis will bring to light migration processes impacting the rural areas withinmetropolitan regions and those rural communities on the exurban fringe.To achieve this goal, the project identifies the following research objectives.1. Replicate the analysis of migration flows at both county and tract scales to highlight new types of movement both between and within counties.2. Utilize the USDA's tract-based Rural-Urban-Commuting-Areas (RUCAs) to identify different migration responses impacting different types of rural areas located within metropolitan and micropolitan regions.Goal #3 - Bring to light new temporal forms of migration invisible within conventional migration data sources. Migration scholars have long acknowledged the existence of cycle or periodic migration flows (temporary workers, sunbird/snowbird, college students, migrant labor, etc.), yet most public data reports residence at a single point in time each year rendering these cyclical forms of movement invisible. The monthly panel data utilized in this project allows for more extensive identification of cyclical migration.To achieve this goal, the project identifies the following research objectives.1. Identify month-to-month change of residence for each mobile device in the data set over the duration of the three year study period.2. Distinguish between permanent movers - those devices that changed locations and then remained at that new location for the remainder of the study period - and temporary movers - those devices that changed locations at some point during the study period, and then returned to the earlier place of origin.3. Describe and map the different types of places that are nodes within these temporary or cyclical migration systems.Goal #4 - Provide communities with an accessible data resource that can be used to assess how varied migration systems are playing out in their areas.Community leaders need a better understanding of the various dimensions (quantity, origins, degree of permanence, etc.) of the 'COVID Refugees' they hear about anecdotally and in the news media. To achieve this goal, the project identifies the following research objectives.1. Develop a project website to distribute data sets, presentations, and results to audiences across the United States.2. Disseminate project results through a variety of different academic and non-academic publications and presentations.Goal #5- Strengthen regional research networks within Vermont.1. Leverage the skill sets and expertise of scholars from different institutions within Vermont throughout the duration of the project
Project Methods
MethodsOur research design involves four analytical goals 1) Build an individual level anonymized database of mobile device users recording their usual home location each month from 2019 to 2021 to identify devices that changed location at least once throughout the pandemic. 2) Identify different types of migration responses distinguishing between permanent, return, and onward movers. 3) Map and profile the varied migration responses to identify when and where certain migration responses were more common. 4) Construct a matrix of migration flows connecting specific types of origins and destinations to examine how specific types of flows vary over time and across space as well as how these flows are contributing to population redistribution across the urban-rural hierarchy.To accomplish these analytical goals, we draw on a unique individual level data set recording the monthly home location of mobile devices provided by Veraset, a location service firm, and our analytical strategy involves using these data in a multi-stage and multi-scale research plan.Stage 1 involves combining the 36 months of data containing home location information for each mobile device by joining each month's data based on the device ID. As we join the monthly data sets, we will keep records for all devices appearing in any of the 36 months to allow for devices to enter or exit the database over the three-year period.With the combined 36 months of data, Stage 2 will identify any device that changed home location at least once since the onset of the pandemic beginning in January 2020. Any device that recorded a change in the census tract home location will be flagged as a mover, and we will extract all devices recording at least one move to create a smaller dataset for further analysis. Stage 2 will also be structured to identify permanent, temporary, or onward/repeat moves. Permanent migration will be determined by identifying any device with a single change in usual home location since January 2020. In contrast, we can identify temporary movers as those devices with a change in usual home location and then an additional move back to an earlier origin. Finally, onward/repeat movers will be those devices with more than one change in location without any return to an earlier home location over the duration of the pandemic.Once the moves have been identified, we can aggregate these different migration responses to census tracts and counties and map them over time and across space exploring how the migration responses differ geographically (by urban-rural hierarchy, region, size of place, etc.) as well as across various socioeconomic and demographic communities defined along lines of income, education, race, and ethnicity. We will complete the Stage 2 analysis at both county and tract scales and draw on data from the American Community Survey 5-year sample to provide socioeconomic and demographic characteristics of tracts and counties. The results from Stage 2 will determine the timing of any migration response, how the migration responses link to changes in county levels of COVID-19 infections and/or place-based mitigation measures (mask mandates, quarantine requirements, etc.), and how different types of responses vary across different types of communities. For example, if temporary movement out of cities toward rural areas was more commonly associated with affluent rather than low-income neighborhoods, the results will suggest that migration was a COVID mitigation strategy marked by privilege. Or, if permanent migration is found to be more common from white neighborhood origins while temporary migration characterizes communities of color, then the migration response to COVID-19 is likely producing new racial geographies in both rural and urban spaces.The Stage 3 analysis also proceeds at both county and census tract levels and focuses on migration flows between origin-destination pairs. Aggregating counts of devices based on specific origin-destination pairs produces a series of month-to-month migration matrices tabulating flows of devices between tracts, counties, or classifications thereof (e. g. urban core tracts, suburban fringe counties, nonmetro noncore counties, etc.). Comparing the volume and direction of these migration flows since the onset of the pandemic with a comparable set from the 2019 monthly data will determine the extent to which the COVID-19 migration response followed established migration pathways. With these migration matrices, it will be possible to build a typology of migration flows at different geographic scales and identify which are most prominent. At the county level, we can identify flows directed down the urban hierarchy from large metro core counties into suburbs, smaller metro areas, micropolitan or nonmetro destinations. Alternatively, using the USDA's tract-level RUCA classifications, we can identify intra-county flows from urban core neighborhoods into suburban tracts or to rural regions on the exurban fringe. We can further isolate individual areas for this analysis focusing on origin-destination pairs. For example, during early stages of the pandemic, much was written about the exodus from New York City, as infection rates were severe in the Northeast. How does the timing, geography, and degree of permanence of these migration responses compare with those originating from Southern or Midwestern metro areas where infection rates spiked much later in the pandemic? Examining the direction and volume of these urban-rural flows over time and across space throughout the pandemic will shed light on the ways in which the COVID-19 migration response is leading to population redistribution across the urban hierarchy at different geographic scales. Further, tracing the change in these flow matrices from month-to-month will determine the degree of permanence of the redistribution and how migration responses have changed over the duration of pandemic.Combined, this multi-staged research approach will shed new light on the ways migration systems have responded both spatially and temporally to the large external shock of the COVID-19 pandemic. The research design will highlight differences in urban-rural migration responses across diverse origin and destination communities differentiated by geographical context and demographic/socioeconomic composition. The ability to trace migration responses between specific origin-destination pairs prior to the pandemic with subsequent responses since its onset will further reveal how established migration pathways shape the current migration response. Finally, the fine-grained temporal scale of the mobile device data set provides a unique opportunity to identify migration responses that are invisible within conventional migration data sources. The monthly panels illuminate longer term migration responses (e. g. aggregate moves from January 2020 through December 2021) along with short term and/or periodic migration responses that may coincide with particular moments in time (the early onset, the Delta surge, vaccine availability, etc.). The understandings gained through this research project will provide critical knowledge to rural communities about the long-term and continued impacts of COVID-19.

Progress 07/01/23 to 06/30/24

Outputs
Target Audience:September 2023 - Peter Nelson and John Cromartie attended the annual meeting for the W5001 Multi-state project, "Rural Population Change and Adaptation in the Context of Health, Economic, and Environmental Shocks and Stressors" in Nashville, TN. We announced the award and gave a brief project update to group members representing local, state, and federal government agencies as well as researchers at both public and private universities and colleges. January 2024 - Taught 12 undergraduate students a Data Science for Geography course using datasets developed as part of this project. April 2024 - John Cromartie gave a lunch time talk to members of his research branch at USDA's Economic Research Service April 2024 - John Cromartie and Peter Nelson both presented research papers at the American Association of Geographers Annual meeting in Honolulu, Hawaii. Attendance at the paper session included other academics from the US, Europe, Canada, and Asia along with community development professionals and policy makers. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?The data acquisition and pre-processing stage of the project required Peter Nelson and his two research assistants to develop new data science skills, particularly those related to data management and cloud-based storage/computing. Through this project, John Cromartie has transitioned to using RStudio as his primary data management, analysis, and visualization software. Using this open-source software facilitates easier collaboration across the research team and allows for a greater degree of replicability of research procedures. More researchers at ERS are using RStudio for analysis, so acquiring these skills through this project has benefits for Cromartie's work beyond this specific project. In January of 2024, Peter Nelson taught a Data Science for Geography course, enrolling 12 students, and these students worked extensively for four weeks on applications drawing on project data. These students developed data management, analysis, and visualization skills as well as techniques for building interactive web-based applications for querying and visualization these migration data. How have the results been disseminated to communities of interest?Thus far, we have disseminated results from this project through three public presentations. Working Research Seminar given to researchers in the Rural Economy Branch of USDAs Economic Research Service, April 2024. Cromartie, John (2024) Regional and rural-urban differences in migration efficiency before and during the COVID-19 pandemic. Presentation given at the American Association of Geographers Annual Meetings, Honolulu, HI, April 17. Nelson, Peter B. and Jorre Dahl (2024) Temporal shifts in urban-to-rural migration during the COVID-19 pandemic. Presentation given at the American Association of Geographers Annual Meetings, Honolulu, HI, April 17. What do you plan to do during the next reporting period to accomplish the goals?For project year #2, our efforts will be focused on the following objectives: Draft a manuscript for submission to Population, Space, and Place based on the results from Peter Nelson's AAG presentation. Begin to replicate the county-based analysis completed during Project Year 1 at the tract scale (Research Goal #2). Build more detailed case studies focusing on flows between specific places (Research Goal 1-5). For example, how did the volume and timing of flows between New York City and rural destinations in northern New England change throughout the pandemic? Or, from where did the primary flows into recreation counties originate, and how did these flows vary across regions? Once we have a prototype for these more specific place-based case studies, we can reproduce these case studies for any number of possible origins and destinations. Begin to build a template for the project website to more broadly disseminate results (Research Goal 4).

Impacts
What was accomplished under these goals? The first month plus of the project was spent acquiring the mobile phone location data from the vendor (Veraset) and developing procedures for processing this unusually large dataset containing well over 1 billion observations. After this acquisition and pre-processing state, the majority of effort in this initial project year focused on Research Goals #1 and #3, and in doing so, we have also made progress toward Goal #5. Below we briefly describe specific results pertaining to these particular research goals. Additional detail, including slide decks with images of results is available upon request. Goal 1-1 Describe how the timing and geography of migration responsesvaried throughout the first two years of the pandemic. When and where did urban-to-rural migration begin to accelerate and how did these flows change over time? Initial county-level analysis of these mobile phone data utilizes the concept of 'migration efficiency'. The analysis reveals how the pandemic sparked an earlier onset of urban-to-rural migration than in a typical year. Data from 2019 show, as expected, urban to rural migration increases in June and July corresponding to summer holiday movements, and migration flows return to rural-to-urban movements as summer ends. In 2020, at the onset of the pandemic, urban to rural migration peaked in April, two full months before the typical summer shift and there was sustained urban to rural migration over a longer time-period in 2020 persisting further into the fall months. By late 2020 and through the first 6 months of 2021, there was a prolonged period of urban-to-rural migration, though not as pronounced as the initial surge during Spring of 2020 as the pandemic swept across space. Goal 1-2 Identify and mapplaces wheremigration responses werepermanent versus temporary.For some, the pandemic-induced migration initiated a permanent move to a new residential location while for others the migration response was temporary with an eventual return to their place of origin. Where were these temporally different migration responses more common in terms of regional location and position along the urban-rural hierarchy? We disaggregated the results pertaining to Goal 1-1 for each US Census Region and for the various county typologies defined by the Economic Research Service. These earlier peaks in urban-to-rural movements were consistent across all US Census Regions, yet the peaks were most pronounced in recreation dependent counties. The more pronounced flows arriving in recreation dependent counties at the onset of the pandemic areconsistent with other work on this topic. Goals 3-1, -2, -3 Bring to light new temporal forms of migration invisible within conventional migration data sources in order to distinguish betweenpermanent movers- those devices that changed locations and then remained at that new location for the remainder of the study period - andtemporary movers? We successfully developed procedures to identify temporary movers who cycle back and forth between origin-destination pairs, and our initial analysis focuses on those who moved between sets of urban and rural destinations. This new window into temporal dimensions of migration shows the rhythms of temporary urban-to-rural migration shifted during the pandemic. For example, the volume of temporary urban-to-rural movers increased two-fold in April of 2020 when compared with April of 2019, and these April 2020 movers remained in their rural destinations for longer durations (by several months) before returning to their urban origins than similar temporary movers from spring of 2019. Goal 5-1 Strengthen regional research networks within Vermont by leveraging the skill sets and expertise of scholars from different institutions within Vermont throughout the duration of the project. Collaborators from Middlebury College and University of Vermont worked closely together during the initial data acquisition stage of the project. Specifically, Eric Clark, Data Scientist at University of Vermont established the procedures for secure data transfer from the data vendor to Peter Nelson's Amazon Cloud Storage so Nelson and his research assistants could begin the processing and analysis. The data transfer would not have been possible without Clark's collaboration.

Publications