Measuring Residential Mobility: A Historical Overview of Novel Data and Methodological Approaches

Abstract

This review paper analyzes the evolution of data sources, methods, and challenges in measuring residential mobility in the United States since the turn of the 20th century, focusing on attempts to measure these aspects of residential mobility: (i) its magnitude; (ii) its “why”; and (iii) its spatial and temporal context. The expansion of data and methods has been instrumental in the development of theoretical frameworks and the ability to interrogate key empirical questions over the decades. While imperfect, the recent growth in alternative sources of individual- and household-level mobility data promises to extend the frontiers of residential mobility research.

Introduction

Households’ decisions to move their place of residence have continuously shaped and reshaped the built environment and social milieu, the ebb and flow comprising both historic phenomena and everyday patterns of importance. In the USA, the late 19th century saw waves of immigration and rapid growth of cities, resulting in early planners’ concerns about the spread of slums (Ravenstein 1889; Riis 1890) and early urban sociologists’ observations and theorizations about processes of urban growth and change (Dubois 1996 [1899]; Park 1914).

Today, while migration levels are much lower (Cooke 2013; Molloy, Smith, and Wozniak 2017), information on residential mobility continues to be important for planning practice. Some of the most crucial issues in planning are directly linked to patterns of residential mobility, including the lack of housing opportunities, especially for young adults (Mawhorter 2017); the migration to and from environmentally risky areas (Marandi and Main 2021; Sheldon and Zhan 2022); the epidemic of evictions across many low-income communities (Desmond 2016); neighborhood change and displacement associated with gentrification (Ding, Hwang and Divringi 2016; Freeman et al. 2024); and the importance of place for economic mobility and opportunity (Chetty and Hendren 2018; Chetty, Hendren, and Katz 2016).

Our methods for measuring residential mobility have advanced significantly over the past 125 years, from manual tabulation of residential directories to complex and nationwide analyses of big data. Indeed, recent years have witnessed a growth in novel data sources. Given the critical juncture of both methodological and substantive debates on residential mobility, now is an especially ripe time to catalog both past and present approaches to measuring and understanding residential mobility.

This paper aims to examine the evolution of residential mobility studies from the turn of the 20th century to the present. In doing so, we elucidate how both our methods for studying residential mobility and our understanding of the process itself have changed, driven by historical contexts, theoretical and scholarly advances, and technological breakthroughs. The remainder of the paper is organized as follows. We first define residential mobility and detail our method for identifying relevant studies and their methodological approaches. We then synthesize the literature around the major questions of interest to planners and scholars: (a) how many people moved; (b) why they moved; and (c) the spatial and temporal context of moving.

Within each of these subsections, we proceed chronologically, examining the relevant historical context, the major sources of data and methodological approaches, and the specific types of questions they addressed. We reference noteworthy examples and discuss the strengths and weaknesses of these approaches. We also note if the type of data is likely to be useful to professional planners, such as those working for a local planning department, and planning researchers working in academia or think tanks.

Scope and Method for Identifying Studies on Residential Mobility

For the purposes of our review, we define residential mobility as a change in one's place of primary residence within a particular political or administrative area such as a county or metropolitan area—in other words, local mobility as differentiated from long-distance migration (Hackworth and Smith 2001). For domestic moves, there is generally a distinction between local mobility and long-distance migration (Shryock 1964). We focus on local mobility for two reasons. First, since at least the middle of the 20th century, local moves have been the majority of moves in the USA. Second, the motivations for, and implications of, long-distance migration are more typically connected to changing regional economic fortunes and different than those related to local residential mobility (Quigley and Weinberg 1977; Shryock 1956).

Given our focus on local mobility, we also exclude international migration from the scope of our review. While migration and residential mobility are global phenomena, its causes and implications differ considerably, and the varying contexts for moves (e.g., war, governmental incentives, employer-provided housing) and sources of data cannot be adequately treated in this format.

To identify studies of residential mobility, we used the citation retrieval software Publish or Perish with Google Scholar as the database. We conducted keyword searches using the baseline terms “residential mobility” and “residential migration ¹ ” and manually filtered the results for relevant peer-reviewed, empirical, English-language studies whose geographical scope was within the USA. Moreover, we supplemented the search results with the selected studies’ references, the authors’ knowledge, and some non-peer-reviewed works such as books and government reports that, in our judgment, met standards for acceptable scholarship. For earlier years prior to the 1960s, we relaxed our search parameters as the nomenclature around residential mobility had not yet solidified.

From the resulting list, we filtered for studies that employed quantitative methodology, as our primary aim is to describe the evolution of numerical data and methods for measuring and explaining mobility. This unfortunately excludes important qualitative studies of residential mobility (e.g., DeLuca, Garboden, and Rosenblatt 2013; Desmond 2016) from our review. Qualitative studies are invaluable for the development and refinement of theoretical models and for understanding the process of residential mobility in a more nuanced way. Nevertheless, as they are fundamentally different from quantitative studies in their purpose and attributes, we focus on the latter.

With our filtered list of studies, we reviewed the existing research with a focus on identifying scholarship that employed data and methods that were considered contemporaneously novel for measuring residential mobility.

How Many Move

Cities in the late 19th and early 20th centuries experienced significant migration from overseas and from rural areas, as well as mobility across neighborhoods and to the suburbs. This population movement triggered neighborhood change and property value fluctuations, stressed public infrastructure, and highlighted the need for publicly wielded tools to manage this change—giving rise to zoning and ultimately planning (Hirt 2014).

High levels of residential instability and transiency were often associated with “pathologies,” such as crime, deviance, and disorder—the idea being that “dislocated” individuals no longer bound by community ties and social controls displayed “anomic” behavior (Rossi and Shlay 1982). Indeed, McKenzie (1921) wrote, “That the mobility of modern life is intimately connected with many of our social problems there is a general consensus of opinion” (157).

Given the impacts of residential mobility and concerns about its causes, there was great interest in measuring mobility rates to test the hypothesis of mobility as pathology. Before 1940, however, aside from asking respondents their place of birth, the Census Bureau collected no data on residential mobility. Consequently, several approaches were developed to infer it, including using publicly available administrative data and residual methods.

Early Uses of Administrative Data

Lacking census data with which to measure residential mobility, researchers resorted to various types of administrative data to fill the breach. Administrative data refers to “data collected and maintained by agencies or firms […] used to administer (or run) programs and provide services to the public” (U.S. Census Bureau 2023). In the early 20th century, publicly available administrative data such as voter registration records were examined for clues about residential mobility (Grebler 1952; Longmoor and Young 1936; McGarry 1935). For instance, McKenzie (1921) used the percentage of registered voters who failed to register the next year as an indicator of residential mobility in Columbus, Ohio, and estimated the rate of mobility for a given neighborhood using this ratio.

Besides voter registration records, other publicly available records utilized in the first decades of the 20th century included residential directories, school enrollment records, death, and marital records (Corbally 1930; Lind 1925; Longmoor and Young 1936), and they were frequently used in conjunction with one another to achieve greater completeness. Albig (1933) used directories to study residential mobility in four medium-sized Illinois cities between 1929 and 1930 and supplemented his consultation of directories with a comparison of school records to identify minors not listed in the residential directories. A study of population mobility in Austin, Texas, from 1929 to 1931 also compared changes in directory addresses in intervening years to infer residential mobility and additionally consulted death records to exclude those who had died (Sciences 1941).

Public utility companies made records available that revealed requests for new and cessation of service connections, including to and from where the household was moving (Green 1934). The availability of these records, along with effort provided by the Federal Emergency Relief Administration, allowed Green to produce Movement of Families within the Cleveland Metropolitan District for the year 1931, one of the most comprehensive studies of local residential mobility of the early 20th century.

Early 20th century use of administrative data had several advantages for studying residential mobility. First, address directories provided extensive coverage of the general population. Likewise, school records provided a near approximation of a census of school age children. This allowed for estimates of mobility for very granular levels of geography, including neighborhoods. Second, by using directories at two points in time, researchers could infer origins and destinations for those who moved within the study area during the intervening period.

However, most studies relying on directories and other publicly available administrative data suffered from important limitations. First, they did not directly measure residential mobility. For example, rates of voter re-registration fell as low as 30% in some neighborhoods McKenzie (1921) studied, which would imply an unrealistically high residential mobility rate of 70%. Second, the directories were not comprehensive and missed several data cases such as moves that took place before or after the two end points of the study period, and moves whose origin or destination locations lay outside the coverage area. Although attempts to verify the coverage of the directory approach suggested typical coverage rates of at least 95% (Goldstein 1958), directories also failed to account for those very transient or on the margins of society (Sciences 1941). Directories could also be problematic for tracing women's mobility because marriage almost always entailed a change of their last name. Thus, women could be “lost” in between the publication of two directories (Goldstein 1958).

Lastly, the aforementioned approaches were also resource-intensive in terms of both the time and effort necessary to compile the statistics. Consequently, professional planners would have been hard pressed to utilize these data sources. The use of administrative data to measure residential mobility fell out of favor in the mid-20th century after more direct and accessible measures became available. Recently, however, there has been a return to the utilization of a wide range of administrative data thanks to the digitization of operations and data and advances in computer technology. We discuss this resurgence later in the paper.

The Residual Approach

Although data directly measuring residential mobility was not collected in the early 20th century, the decennial census and vital statistics (i.e., births, deaths) were available at this time. Thus, one could infer how much population change was due to residential mobility by accounting for population changes due to births and deaths and attributing the remainder to migration. Such a residual approach was used by Hamilton (1934) to study rural–urban migration in North Carolina between 1920 and 1930. In addition to overall migration rates, Hamilton was able to produce estimates disaggregated by age, sex, and race.

Residual approaches are best suited for understanding net migration flows in larger geographies, as vital statistics are typically only available at larger geographies such as the county. The residual approach has several other shortcomings. It misses mobility that occurs within the unit of geography being studied. It cannot provide information on the origins and destinations of movers, individual or household traits associated with residential mobility, or motivations for moving. Despite these shortcomings, the lack of suitable alternatives for small geographies made the residual approach a mainstay of professional planners until the end of the 20th century (Myers 1992).

Although early 20th century planners and scholars were ingenious, their understanding of residential mobility was sorely lacking. Reliable estimates were not available, flows between origins and destinations were unknown, and motivations for movement little understood. The federal government was notably absent in efforts to collect data on residential mobility. It would take an economic calamity to prompt them to begin efforts measuring residential mobility.

The Federal Role

Until the Census Bureau began asking mobility-specific questions in 1940, the only data the Census Bureau collected that could be used to infer residential mobility was respondents’ state of birth. For those currently residing in a different state than their birth state, at least one move could be inferred. This method was used by Ravenstein (1889) in seminal studies of migration including between American states. The Great Depression of the 1930s and the ensuing New Deal highlighted the need for the federal government to collect more labor and housing data (Anderson 2015). Employment statistics were especially needed both to track the country's depths of economic despair and to gauge the effectiveness of the various New Deal programs. As such, in the following decades, the decennial census expanded to include more questions covering social, economic, and housing characteristics.

In 1940, the census began asking respondents to identify where they lived five years ago, including whether they lived in the same house, city, county, or state. For the first time, much more granular migration flows could be directly estimated (United States Bureau of the Census 1946) and reliable estimates of residential mobility could be produced for the nation and various subareas and subpopulations (United States Bureau of the Census 1957). The decennial census would hereafter be one of the most important sources to measure residential mobility.

Significant gaps in understanding residential mobility remained. Because the mobility question in the census asked the respondent of their residence five years ago, it only captured one move, potentially missing moves within and beyond this window. Moreover, origins and destinations of movers below the city level could not be identified. Although the Census Bureau collected information on whether the person lived in the same place, down to the same house, only data for intercounty moves was disseminated (Shryock 1964, 64). Our tabulation of the 1940 census data shows that, between 1935 and 1940, 11.4% of the population moved across county borders, while 42.2% moved within the county. Thus, the initial tabulations of residential mobility by the Census Bureau missed the largest type of residential mobility.

With the 1950 census, the Census Bureau changed the mobility question to ask about residence one year ago as opposed to five years ago, facilitating temporally finer analyses. (The Census Bureau returned to the five-year migration question starting with the 1960 census onward.) The Bureau also released tract-level data on residential mobility with the 1950 census. This was the first time that neighborhood-level residential mobility could be reliably estimated in the USA, though only the largest cities were tracted by 1950. Even without universal coverage, the collection and dissemination of residential mobility data at the tract level represented a major advance.

The temporal limitation of the decennial census was in part addressed by sample surveys conducted between censuses. One such survey was the current population survey (CPS), which began in the early 1940s as an instrument for tracking employment trends. The Monthly Report of Unemployment, which eventually became the CPS, began asking questions about residential mobility in 1940. The CPS had an important advantage over the decennial census: the residential mobility question was asked annually. Additionally, by including questions about where respondents lived both one and five years prior in the survey, the CPS allowed researchers to examine both short- and medium-term migration patterns. Using probability sampling methods, the CPS provided perhaps the first reliable estimate of annual residential mobility: 18.7% of people over the age of one moved between 1949 and 1950 (United States Bureau of the Census 1957).

Over time, additional topical questions were added, and the CPS continues to be an important benchmark for national estimates of residential mobility to this day. For example, scholars have relied on CPS data to detect and understand sizable declines in rates of residential mobility over the mid- and late-20th century up to recent years (Fischer 2002). However, the CPS still did not address the question of why respondents moved; it would come in 1999, nearly half a century after its inception, leaving researchers to develop alternative instruments to answer the question in the meantime.

Why People Move

Early Theorizations with Bespoke Small-Scale Surveys

After being able to ascertain how many people moved, understanding why people moved remained perhaps the most important mobility question. For planners, understanding this could help with anticipating and responding to major demographic shifts, such as postwar suburbanization and population decline in central cities. As neither the decennial census nor the CPS asked respondents about their motivations for moving during the mid-20th century, researchers designed and implemented bespoke survey instruments to tackle this question.

The aforementioned linking of high rates of residential mobility with social disorder led those in government circles to conclude that mobility needed to be “cured,” which led to the funding of one of the most seminal works on residential mobility, Why Families Move (Rossi and Shlay 1982). Rossi's Why Families Move (1955) used 924 sample surveys across four Philadelphia census tracts with retrospective questions about residential mobility, examining households and their decision-making. In contrast to earlier neighborhood studies, the household-level study revealed that the family life cycle—i.e., stages through which people become adults, marry, have children, and so on—was the most important determinant of residential mobility. Paradigm-shifting, Rossi's findings repudiated the view of residential mobility as a harbinger of social instability and even pathology (Faris and Dunham 1939; Longmoor and Young 1936; Potter and Robert 1956; Schroeder 1942; Shaw and McKay 1942). Residential mobility was, in fact, a rational and adaptive response to changing household needs and motivations (Potter and Robert 1956).

While Rossi found the family life cycle was a key factor in residential mobility, there remained the question of why milestones associated with the life cycle induced people to move. Scholars of the mid-20th century sought to develop a general theoretical framework on residential mobility. For example, Sjaastad (1962) proposed a cost–benefit model of migration; Lee (1966) similarly posited that the mobility decision-making process concerned weighing the positives and negatives of the origin versus the destination. Others argued that migration was a response to stress concerning the needs of the household and the attributes of the environment (Brown and Moore 1970; Golant 1971; Wolpert 1965).

Building on these theories, one of the most influential works was that of Speare (1974), who developed a residential satisfaction model of mobility. Speare (1974) theorized that people moved when discrepancies arose between their current residential conditions and their desired residential conditions. To test this, Speare leveraged a bespoke survey instrument that asked respondents about their satisfaction with their housing and neighborhood and whether they had desire or plans to move in the next year. He then followed up a year later to see if they had moved. The study revealed that residential satisfaction was a predictor of a wish to move and actual mobility in the following year. The life cycle theory fit this framework, as changes such as additions of family members rendered one's current residence unsatisfactory.

Although these surveys were a great advancement, there remained significant limitations. Collecting survey data was resource-intensive and limited to a relatively small number of households. The surveys relied on the memory of respondents. Households could typically remember prior addresses, but the exact dates of moves were more elusive (Goldstein 1958). For these reasons, bespoke surveys are typically deployed by researchers as opposed to planning practitioners. Nonetheless, bespoke surveys measuring residential mobility will likely be utilized for the foreseeable future as there will always be research questions that can only be answered by the researcher crafting their own questionnaire.

The Federal Role Again: Large-Scale Panel Survey Data

Understanding residential mobility, including the role of housing and neighborhood conditions, continued to advance in the 1960s as the federal government's ambitions and responsibilities expanded with the Great Society—President Lyndon B. Johnson's legislative program of national reform for eliminating poverty and racial injustice. A central part of the Great Society reform was the War on Poverty. The Great Society programs put the nation on a path for studying poverty, attempting to eradicate it, and gauging the effectiveness of anti-poverty programs. Toward these ends, the federal government began funding projects that produced data later useful for studying residential mobility.

A confluence of several other developments amplified the involvement of the federal government. The tracting of the USA, which had previously been limited to the largest cities, began expanding dramatically after the 1960s due to legal and political battles over legislative redistricting during the 1950s, especially in consideration of the Voting Rights Act of 1965. Simultaneously, the development of computing technologies during World War II and the Cold War demonstrated the importance of scientific research. Thus, the federal government became one of the largest funders of research, including that carried out by social scientists.

For the 1960 Census, the Census Bureau, which had always disseminated data in printed books, began releasing electronic data, including anonymized individual records known as the Public Use Microdata Samples (PUMS) (Sobek et al. 2011). PUMS provided individual-level data, representing between 1% and 5% of the population for geographies ranging from states to public use microdata areas (PUMAs), areas comprising approximately 100,000 persons. With PUMS, although the smallest geography available, the PUMA, was much larger than a neighborhood, researchers could examine individual-level correlates of residential mobility. Combined with housing data collected in the decennial census since 1940, researchers could now link housing conditions to the propensity to move, illuminating motivations for residential mobility.

The War on Poverty and interest in tracking the success of social programs provided the impetus for the initiation of several large-scale panel surveys that would prove fortuitous for understanding residential mobility. Panel surveys include “repeated observations by following a sample of persons (a panel) over time and by collecting data from a sequence of interviews (or waves)” (De Keulenaer 2008, 571). One such panel study was the National Longitudinal Surveys (NLS), launched by the US Department of Labor (DOL) and Bureau of Labor Statistics (BLS) in 1966, which focused on education and employment dynamics of adolescents and young adults. Another was the Panel Study of Income Dynamics (PSID), launched in 1968 to “gauge the success of the War on Poverty and associated reforms and to track the economic well-being of U.S. families” (McGonagle and Sastry 2016, 186). Conducted by the Institute for Social Research at the University of Michigan, the PSID became the world's longest running, nationally representative, longitudinal household survey.

Also started in the wake of the War on Poverty was the American Housing Survey (AHS). Initiated in 1973 by the Census Bureau and the U.S. Department of Housing and Urban Development (HUD), the AHS sought to assess the general condition of the nation's housing stock and inform the design, implementation, and evaluation of government housing programs. To such ends, the AHS offered panel data on housing units. Moreover, the 1980s witnessed the launch of the Survey of Income and Program Participation (SIPP) and the Survey of Consumer Finances (SCF), which collected data on the use of public assistance programs and consumer finances, respectively. These surveys aimed to document households’ economic well-being and income patterns but also included information that was useful for studying residential mobility.

These panel surveys were instrumental in advancing the study of residential mobility in several ways, as they collected information that elucidated motives for moving. First, surveys including the AHS (created in 1973) and the PSID (created in 1968) directly asked recent movers why they had moved. With these queries, researchers could produce reliable national and, in the case of the AHS, metro-level estimates of self-reported motives for moving. Life cycle and other theories on residential mobility could now be tested using national samples (Fielding 1992).

Second, panel surveys such as the SCF and the SIPP enabled researchers to document the role of economic factors in predicting and shaping residential mobility. The impact of home equity on mobility choices for homeowners (Bricker and Bucks 2016), the influence of household debt on mobility decisions (Phillips et al. 2021), and how quickly tenants moved out of public housing (Dantzler 2021; Freeman 1998; Hungerford 1996) were but a few of the economic facets of residential mobility that could be explored with these data sources.

Third, by collecting data on respondents’ perceptions of their housing unit (especially the AHS), neighborhood, and their satisfaction with these conditions, researchers could further test the residential satisfaction framework. Beyond confirming the importance of the life cycle for predicting residential mobility (Clark and Davies 1990), the emergence of new panel data in the 1960s–1980s showed that housing inadequacies (including physical conditions and location), undue cost burdens, overcrowding, and neighborhood disamenities were also important explanations for why people might decide to move (Newman and Duncan 1979; Goodman 1976).

Panel surveys based on probability samples continue to be a useful tool for measuring different facets of residential mobility. The repeated nature of the data and the additional information these surveys collect make them indispensable. However, panel surveys are complex and resource-intensive to carry out. The surveys we describe typically provide estimates for the entire country or relatively large subregions (e.g., states). Consequently, while local planners may find panel survey research useful for general knowledge, it offers limited practical value in day-to-day planning.

By asking people why they moved, panel surveys like the AHS and PSID also allowed researchers to identify involuntary moves. By the late 20th century, involuntary movement or displacement had become a major concern (Grier and Grier 1980). Many cities were still reeling from the havoc caused by urban renewal and the construction of the interstate highway system that displaced thousands of households (Anderson 1964). Both the AHS and the PSID offered some of the earliest estimates of involuntary mobility. Studies in the late 1970s and early 1980s generally found that 1% (Newman and Owen 1982) to 3.3% (Lee and Hodge 1984) of households were displaced annually. These questions later proved valuable in studying displacement from events like the foreclosure crisis (Lee and Evans 2020).

Still, even when surveys asked about reasons for moving, some scholars argued they fell short in adequately capturing involuntary moves. For instance, Desmond and Shollenberger (2015) found residents often did not recognize their moving experience as displacement unless it aligned with formal, court-ordered eviction. Carlson (2020) noted that these surveys typically restricted respondents to selecting one reason for their move, even though mobility decisions were often shaped by multiple factors, with displacement pressures not always taking precedence. Nonetheless, these surveys continued to provide meaningful insights into involuntary mobility patterns, especially compared to older datasets that did not ask the “why” question.

The Spatial Context: To and From Where People Move

In the middle and latter decades of the 20th century, two types of neighborhood change figured prominently in the literature on residential mobility. The first was the continued Great Migration of Blacks from the southern USA to urban centers in the north and the west, and the second was the “white flight,” or the exodus of white households from central cities to suburbs, it appeared to trigger. These migration patterns altered the racial landscape of the USA and changed the very meaning of “urban” in a colloquial sense. Prior studies of the first waves of this migration circa World War I could only make crude estimates of the migrants’ origins and destinations as the mobility status of residents was unknown and tract-level data rare (Chicago Commission on Race Relations 1922; Kennedy 1930; Scott 1920). As the Great Migration reached its zenith after World War II, understanding where Black migrants would settle and which neighborhoods would experience the seemingly inevitable white flight drew the attention of students of residential mobility (Duncan and Duncan 1957; Taeuber and Taeuber 1965).

Attempts to measure white flight typically relied on census data. While white flight could be documented in the form of a shrinking white population in a given neighborhood, understanding the pace of the process, characteristics of white movers, and their specific destinations could only be inferred (Duncan and Duncan 1957; Taeuber and Taeuber 1965). Further, observing aggregate changes over 10-year periods resulted in rather crude estimates. The contemporary literature debated the drivers of residential racial turnover and the relative importance of factors such as neighborhood racial context, the structural conditions of the housing market, and other demographic and socioeconomic characteristics (Frey 1979; Guest and Zuiches 1971; Marshall 1979).

Starting in the 1960s and 1970s, several old and often poor urban core neighborhoods appeared to be bucking the trend of white flight and experiencing an influx of higher social economic status residents, many of whom were white—a type of neighborhood change that came to be known as gentrification. These patterns ran counter to the traditional ecological and economic models of neighborhood change, such as the Chicago School's invasion-succession and filtering models. Gentrification posed several questions for students of residential mobility.

One question was from where the gentrifiers were originating from. It was first hypothesized that the gentrifiers were coming from the suburbs where all the middle class had decamped in previous decades. However, research by Gale (1979) and Smith (1979), who documented the origins of owners making renovations in gentrifying neighborhoods, cast doubt on this thesis. In these instances, the overwhelming majority of the gentrifiers, defined as those filing for permits for building renovations in gentrifying neighborhoods, were from the central city. While these efforts were innovative, they were based only on a few neighborhoods and certainly did not include all gentrifiers, some of whom were not engaged in housing renovation.

Gentrification also sparked interest in how quickly people were moving from gentrifying neighborhoods and where they were going. Scholars relied on movement into and out of central cities as a proxy for gentrification and displacement and examined net changes in population characteristics at the neighborhood level over time (Grier and Grier 1980; Lee, Spain, and Umberson 1985; Sumka 1979). However, these approaches faced two key limitations. First, displacement was often inferred from any out-movement from gentrifying neighborhoods, which likely overestimated gentrification-induced displacement. Second, relying on neighborhood-level net change obscured the scale and direction of gross flows, masking the fact that large in-flows and out-flows can produce similar net changes to a neighborhood devoid of population change. Like the challenges posed in the studies of white flight, relying on net population characteristic changes or in-/out-migration to identify gentrifying neighborhoods limited early studies’ ability to understand household mobility behavior or to conclusively assess displacement.

Panel studies like the AHS and PSID, originally designed for other purposes, became instrumental for addressing the challenges in understanding the spatial context of mobility, including white flight and gentrification. The availability of neighborhood-level origin and destination data was especially fruitful for settling contentious debates, such as the role of selective residential mobility in creating neighborhoods with concentrated poverty (e.g., Crowder and South 2005; Massey, Gross and Shibuya 1994; Quillian 1999; South and Crowder 1997), measuring white flight from integrated neighborhoods (Crowder 2000; Crowder, Hall and Tolnay 2011), and the relationship between gentrification and displacement (Vigdor, Massey and Rivlin 2002).

The findings from this body of research suggested the neighborhood context was indeed important for residential mobility, although perhaps not in ways previously presumed. Using geocoded PSID data, Kyle and Crowder (2000) showed that greater nonwhite presence in a neighborhood increased the likelihood of white out-movement. While this finding was consistent with the “tipping point” theory of neighborhood change, their findings also revealed a more complex relationship that was shaped by factors like racial diversity and the pace of demographic change. Moreover, Crowder and South (2005), using the same PSID data, found the racial composition of the neighborhoods surrounding a respondent's neighborhood was an even stronger predictor of white flight than the composition of the respondent's own neighborhood.

Geocoded panel data has also proved illuminating in the study of gentrification and residential mobility. Using the AHS, Vigdor, Massey and Rivlin (2002) found that the rates of movement for poorer incumbent residents were no higher in gentrifying areas. Using a similar approach and the New York City Housing and Vacancy Survey, a panel survey modeled on the AHS for New York City, Freeman and Braconi (2004) found the rate of residential turnover to be lower in gentrifying neighborhoods. Freeman (2005), using the PSID, found no difference in mobility rates between gentrifying and non-gentrifying neighborhoods, and displacement rates were only modestly higher in gentrifying neighborhoods. More recently, researchers have exploited geocoded panel data to compare the destinations of movers from gentrifying neighborhoods with those of movers from non-gentrifying neighborhoods, finding only small differences (Freeman et al. 2024).

These results show that the availability of spatially granular, individual- and household-level data could elucidate the dynamics of residential mobility. One drawback of the PSID is that geocoded panel data is restricted-use, meaning researchers need to apply for access through valid U.S. research and governmental institutions. Nevertheless, the ability to track mobility flows allowed researchers to distinguish neighborhood change driven by newcomers from that stemming from incumbent residents, as well as identify specific driving factors of mobility and neighborhood change. By the end of the 20th century, the literature on residential mobility had laid a solid foundation on critical demographic phenomena, including those discussed above. In addition to measuring the magnitude of residential mobility, recurring surveys like the AHS and PSID provided national-level estimates pertaining to the motivations for residential mobility. The CPS also began asking respondents why they moved in 1999. Moreover, these surveys included a plethora of demographic, economic, and social data from which researchers could infer motivations for moving beyond those described by the respondents themselves.

While professional planners rarely used panel data like the PSID to study neighborhood dynamics, they did benefit from the increasing font of knowledge about the role of residential mobility and neighborhood change. Certainly, community-based organizations looking to preserve integrated neighborhoods long made use of demographic studies of tipping points and white flight (DeMarco and Galster 1993; Perkiss 2014). Gentrification remains a contested issue, and it is unclear to what extent scholarly research on the topic has influenced planning and policy directions. But at least a few cities have attempted to adapt their land-use practices to recognize the potential risk posed by displacement due to gentrification (Freeman 2023).

The Spatial and Temporal Context in the Era of Big Data

In the 21st century, the rise of big data and the concomitant increase in computing power have enabled students of residential mobility to continue probing the subject at unprecedented levels of granularity. Newly available data products have allowed researchers to identify moves spatially at the address level and temporally at much finer intervals. For planners, they have the potential to facilitate better-informed neighborhood-level planning. Rather than relying on the decadal release of neighborhood-level data from the decennial census, these new data sources allow for annual (and in some instances more frequent) updates in neighborhood mobility patterns. We highlight three developments that have enhanced the study of residential mobility in both spatial and temporal aspects below, including further advances in census data, the increasing digitization of public and commercial life, and the advent of smartphone technology.

Further Advances in Census Data

The utility of Census Bureau products for studying residential mobility further expanded with the advent of the ACS in 2005. Census mobility questions were transferred from the long form of the census to the ACS. As the ACS was conducted on a monthly rolling basis and released annually, estimates of residential mobility could also be obtained annually (except for small areas of geography where five-year moving averages were released annually). Moreover, versions of the ACS released through the PUMS allowed for individual-level analyses of residential mobility down to levels of geography as small as 65,000 persons. Free and in a format that is relatively accessible, the annually updated ACS data was especially fortuitous for professional planners who might lack the resources to utilize other data described here. It should be noted, however, that the inconsistencies between the ACS (capturing one-year mobility patterns) and the decennial census (capturing five-year mobility patterns in 1940 and from 1960 to 2000) have made it challenging for researchers to draw direct comparisons or link the two data sources.

The increased spatial and temporal granularity afforded by the ACS has been crucial particularly in light of the recent disquieting decline in residential mobility. While the traditional view of the USA as a “restless nation” (Jasper 2000) has been challenged by the steady decline in migration and local mobility over the decades (Fischer 2002), the recent acute downturn in the latter has raised concern for the country's declining economic and political dynamism and its potential impacts (Klein and Thompson 2025). Myers, Park and Cho (2023) leverage the ACS to identify the phenomenon's key correlates and find that lagging housing supply and the large size of the Millennial generation are contributing factors of suppressed mobility (Myers, Park and Cho 2023). Compared to the CPS (which has been leveraged to examine the decline in mobility up to 2000 by Fischer (2002)), the ACS features more detailed housing variables and a significantly larger sample size. Further, research has also found that mobility rates computed using the ACS track those based on IRS migration records and the AHS more closely (Masnick, 2013; Myers, Park and Cho 2023).

Another significant development toward expanded access to census data was the establishment of Census Research Data Centers (RDC), where qualified researchers could access census microdata with safeguards to protect data confidentiality. The first RDC was established at the Census Bureau headquarters in 1982 and the first remote RDC was in Boston, MA, in 1994. In 1998, the Census Bureau partnered with the National Science Foundation (NSF) to expand the RDC program, which has since been rebranded as the Federal Statistical Research Data Centers (FSRDC) and grown to 34 centers across the country. Confidential microdata from decennial censuses and the ACS offered the benefits of the census—geographic coverage and sample size—together with household- or individual-level granularity, with place of residence (and previous place of residence if a household moved in the last 12 months in the ACS) identified to the census block. In addition, from 2000, the Census Bureau began to provide data on individuals linked to their domiciles in the Master Address File Extract (MAF-X) and the Auxiliary Reference File (MAF-ARF) on a yearly basis, further enabling the construction of residential histories for those who appear in the census surveys (Sullivan and Genadek 2024).

Confidential census microdata has been leveraged in various forms in recent years, primarily in addressing research questions related to neighborhood change and gentrification. Using the data in cross-sectional form, McKinnish, Walsh, and White (2010) employed the confidential 1990 and 2000 Decennial Census Long Form data and created synthetic cohorts to compare their sizes over time and produce estimates of out-movement. They found little evidence of displacement of low-education or minority householders in gentrifying neighborhoods. McKinnish and White (2011) used the same data products to examine the types of households that moved into gentrifying neighborhoods. Similarly, Lee and Perkins (2022) leveraged the confidential 2011–2019 ACS one-year microdata sample, with a focus on individuals’ one-year mobility outcomes, to analyze the relationship between gentrification and mobility outcomes and how it varied by metropolitan area types.

More recently, the Census Bureau has been investing in the creation of a longitudinal data infrastructure for research through projects such as the Census Longitudinal Infrastructure Project (CLIP) and the Decennial Census Digitization and Linkage Project (DCDL) (Alexander and Genadek 2023; U.S. Census Bureau 2025). These efforts seek to maximize the utility of the microdata held at the Census Bureau by linking person records across the decennial censuses, surveys, and administrative records using Protected Identification Keys (PIKs). The resulting longitudinal data are made available to qualifying researchers through FSRDCs.

One of the pilot research projects for CLIP examined the effect of gentrification on residential mobility of original residents. Brummet and Reed (2019) constructed a national panel of individuals from the 2000 Decennial Census Long Form data and the 2010–2014 ACS data. They yielded approximately 3 million matched individuals and identified individuals’ residence to the census block. They found that gentrification increased less-educated renters’ out-of-neighborhood moves by approximately 4–6 percentage points but that movers from gentrifying neighborhoods were not more likely than those from non-gentrifying neighborhoods to settle in poorer neighborhoods or to commute farther.

While the FSRDCs are a great resource for researchers, accessing the confidential census microdata requires an application process, data access fees (unless the researcher is affiliated with an institution that hosts a FSRDC), and, in most cases, on-site work at one of the 34 centers around the country. The research must also present some benefit to the Census Bureau, aside from the researcher's aims. These requirements—of resources, both time and monetary, and of physical presence at a FSRDC—significantly limit the accessibility of confidential census data to researchers and, especially, professional planners.

Digitization of Commerce and Public Life

The increasing digitization of our world, which has produced prodigious quantities of data, the “open data” movement that has encouraged the public sector and some private institutions to make their data publicly available, and the increasing accessibility of computing power all have significantly advanced the research on residential mobility. In terms of public sector data, more administrative data and census products at the household or individual level have become available (with due process for access and safeguards in place for data protection). In terms of private sector data, there has been a growth in commercial data products, including large consumer reference data (also referred to as consumer trace data), which synthesize various public and private datasets, such as change-of-address data and utility bills, and smart device data, which provide device locations, online activity, and more. While much of the private sector data have traditionally been collected and sold for commercial purposes, a growing number of researchers are leveraging the datasets in their research.

Administrative Data

The use of administrative data has come full circle since the early 20th century, from utility records and directories being the only local sources to measure residential mobility to today, where administrative data from various sources can provide individual-level residential location data with nearly complete coverage for certain populations of interest. For example, researchers have used HUD's internal administrative data to analyze the residential mobility and neighborhood outcomes of households with Housing Choice Vouchers (HCV) (e.g., Collinson and Ganong 2018; Reina and Winter 2019; Reina, Acolin, and Bostic 2019). Such administrative data captures the universe of the population of interest (e.g., all voucher holders) and allows researchers to measure residential mobility precisely from both a spatial perspective (i.e., the address from and to where households are moving) and a temporal perspective (i.e., exactly when households move).

Another example of the granularity allowed by administrative data is a study by Dragan, Ellen, and Glied (2020). This study drew on New York State's Medicaid claims records from January 2009 to December 2015 to examine whether children living in gentrifying neighborhoods moved more frequently and longer distances than their counterparts in non-gentrifying neighborhoods and whether their move outcomes were different. The Medicaid data included exact addresses, enabling the researchers to identify precise buildings, restrict the sample to children who live in multifamily rental buildings, and distinguish between market-rate and subsidized housing. Using the data, Dragan, Ellen and Glied (2020) did not find evidence of elevated mobility rates for children living in gentrifying neighborhoods, but gentrification was associated with slightly farther moves.

Consumer credit data collected and disseminated by credit bureaus, such as the Federal Reserve Bank of New York/Equifax Consumer Credit Panel (CCP), have also proved a useful type of sampled administrative data for studying residential mobility. The CCP is a longitudinal panel of an anonymized 5% random sample of individuals, approximately 12 million people, in the US with a credit history and whose credit file contains their social security number. Comprising loan account data, public record and collection agency data, and some limited individual background information, it provides individuals’ census geographies of residence (down to block) (Lee and Van der Klaauw 2010). Its large sample size produces reliable estimates for small levels of geography, while quarterly updates enable finer longitudinal analysis (Ding, Hwang and Divringi 2016; D. Lee and Van der Klaauw 2010, 3). Leveraging these strengths, Ding, Hwang, and Divringi (2016) studied the relationship between gentrification and residential mobility in Philadelphia. They found that vulnerable residents (i.e., low credit scores and without mortgages) in gentrifying neighborhoods did not exhibit greater mobility rates than their counterparts in non-gentrifying neighborhoods, but that, when they did move, they moved to lower-income neighborhoods.

There are, however, caveats and drawbacks to the CCP. Only individuals with a credit history and a social security number are sampled, which is likely to exclude certain groups (e.g., minority, lower-income consumers) more than others (Brevoort, Grimm, and Kambara 2016). Moreover, while it features information on credit performance and history, the CCP does not include key demographic and socioeconomic attributes that may be of interest to researchers, such as race, educational attainment, and income. Finally, it does not provide street addresses of individuals, making hyper-local moves, such as within a census tract, invisible.

Lastly, federal income tax records have been leveraged in some of the most influential works in the study of neighborhood effects, highlighting their potential for residential mobility research. Federal income tax records from the Internal Revenue Service (IRS), as utilized by Chetty and colleagues in a series of papers, include residential location (at the time of their writing, only available to the zip code level) and income, as well as other useful information such as college attendance and quality, marital status, and fertility of individuals. From income tax returns (1040 forms) and third-party information returns (e.g., W-2 forms), the data cover the earnings of both those who file tax returns and those who do not, and the availability of social security numbers (SSNs) facilitates data linkage with other sources with SSNs as the identification key, such as government programs. For instance, Chetty, Hendren, and Katz (2016) match the IRS data with 99% of the participants in the Moving to Opportunity (MTO) demonstration. In another paper, Chetty and colleagues capture millions of children and parents using the IRS data, exploit moves to identify the causal effect of place, and find that place indeed matters for intergenerational socioeconomic mobility, primarily via duration of exposure in childhood (Chetty and Hendren 2018). However, similar to the CCP, the IRS data does not include race.

Consumer Reference Data

Recent vintages of consumer reference data have several features suggesting great potential for studying residential mobility. Perhaps most important is the sheer volume of these datasets. Two main providers of consumer reference data, Infutor and Data Axle, claim to include 80%–90% percent of the U.S. population (according to various sources including Verisk and Bernstein et al. (2022)) and up to the last 10 addresses of people in a dataset. These address histories are compiled from public and commercial vendors such as change-of-address notifications, tax assessment information, utility connections, voter registrations, and white page directories. They also contain the first and last names of individuals and inferred demographic information of variable quality (Bernstein et al. 2022, as cited in Greenlee 2019). Importantly, these datasets are typically available across the USA, which allows researchers to measure longer-distance domestic migration using the same methods as they might measure local residential mobility.

A small but growing body of literature utilizes such consumer reference data to answer questions related to residential mobility with enhanced spatial and temporal granularity, including how residential mobility contributes to neighborhood change and how housing policy and market conditions affect the residential mobility patterns of different groups. For instance, Greenlee (2019) uses InfoUSA data to attain yearly snapshots of households’ residential locations between 2006 and 2015 and trace neighborhood-level residential mobility flows in Cook County, IL. He finds a net flow from gentrifying neighborhoods to working-class neighborhoods.

Regarding housing policy and market conditions’ impact on residential mobility, Diamond, McQuade and Qian (2019) study the effects of rent control in San Francisco and leverage Infutor data to identify persons living in buildings covered by the implementation of rent control in 1979 and their mobility behavior before and after rent control adoption. They find that rent control limits displacement and renters’ mobility by 20%, while reducing rental supply by 15%. In a recent study using data from Data Axle to measure household-level in- and out-movement, Song and Chapple (2024) find that lower-income households are more likely to leave gentrifying areas than non-gentrifying areas, while they are less likely to enter gentrifying areas than higher income households. In another study, also using Data Axle data, the authors find that market-rate housing supply helps to slightly alleviate displacement pressures in Los Angeles and to increase in-migration in weaker market neighborhoods in San Francisco (Chapple and Song 2025).

Consumer reference data offer several advantages for researching residential mobility. First, they have large sample sizes, which, depending on the area, can capture most of the population. Second, they provide address-level information for individuals’ residential locations. Such granular spatial resolution, combined with the large sample sizes, allows researchers to formulate and answer bespoke questions related to residential mobility that are not limited to administrative boundaries. Third, their temporal resolution likewise enables analyses associated with specific events in time that might influence residential mobility, rather than needing to align a study's temporal scope to survey intervals. Finally, consumer trace data combine large sample size with a panel that follows individuals over time. These characteristics allow researchers to make inferences about residential mobility behavior for subsets of the population that traditionally could only be made using panel surveys with small samples (e.g., PSID).

Consumer reference data is not without important limitations. First, although the data coverage is generally high, research comparing the datasets with estimates from the ACS finds considerable discrepancies, particularly when it comes to certain populations, including lower income, younger, racial minority, and renter individuals. Such systematic undercounts could lead to bias and erroneous conclusions and policies (Ramiller et al. 2024). Second, the data typically lacks reliable demographic and socioeconomic information of individuals. Some private providers do offer sources to link to Census data, though we found linking to be inconsistent and were generally unable to establish many successful connections. Alternatively, demographic and socioeconomic traits can be inferred probabilistically using the census-provided neighborhood characteristics of home locations, but this is less desirable than that found in other self-reported data (e.g., ACS and PSID). Consumer reference data, in their current form, cannot substitute census microdata and require optimization or adjustment strategies, such as population weighting, and greater transparency in how the data providers construct the datasets (Ramiller et al. 2024).

Smart Device Data

With the development of GPS-equipped smart devices with geo-location features, the use of geospatial traces to understand mobility has become more common. Geo-tagged social media (e.g., Twitter, Instagram) posts and data from third-party data aggregators—such as Cuebiq, Safegraph, X-Mode, and Veraset, who collect location data from a bundle of mobile phone applications—provide opportunities to advance the study of residential mobility, especially with relation to spatial context. These data have four especially attractive features. First, device-derived data, including that from social media, can be spatially precise to around 10 m on average. Thus, mobility can be measured at virtually any scale, from the building level and up. Second, locations can be measured at precise points in time, down to the exact time of day. Third, device-derived data reveal day-to-day mobility in addition to, sometimes, residential mobility. Finally, the data points can be measured in the millions per day—generating truly “big data” and large sample sizes for analysis. Together, these characteristics allow researchers to not only measure residential mobility in innovative ways but also to gain insights into the relationship between other types of mobility (e.g., everyday) and residential mobility and observe other factors like actual or potential social interactions.

Indeed, several scholars have utilized smartphone-derived data to study residential mobility. Poorthuis, Shelton and Zook (2022) used Twitter data to examine changes in residential location of users in Lexington, Kentucky between 2012 and 2017 to understand factors associated with gentrification. They use the most “significant base location” of a user as the “home” location, defined by the volume, length, and regularity of a tweet from a particular location (Poorthuis, Shelton and Zook 2022, 10). Given the large sample size of Twitter data, they are able to estimate local movement patterns for specific neighborhoods, something impossible to do with nationally representative panel surveys like the PSID. Other scholarship has investigated the links between social networks, isolation, and where households move when relocating leveraging Twitter data (Phillips et al. 2021; Wang et al. 2018).

Three shortcomings of this data are worth noting. First, they lack socioeconomic characteristics. Second, they are only available for relatively short time periods: the earliest data coincide with the adoption of smartphones, which precludes longer-term research. Moreover, more recent data only trace individuals over limited spans; for instance, Cuebiq's device IDs change every six months. Third, they follow only individuals, without indicating whether they are heads of households, renters, and/or homeowners, which risks overcounting moves when multiple members of the same household relocate together.

The turn of the 21st century has witnessed the expansion of what we might call “big data”—generally characterized by its volume, velocity, and variety that makes it too large to be handled by standard software and hardware (Favaretto et al. 2020). In the public sector, the Census Bureau's investments into a linked longitudinal data infrastructure across different census products have meaningfully enhanced the scope of research using census data, while federal and state agencies making internal databases available to some researchers has demonstrated the value of administrative data that can follow a comprehensive and targeted swathe of the population over time. These data sources, like the CCP and the federal income tax records, are also valuable in their inclusion of individual characteristics that may shed light on the motivations for, and effects of, moving.

Appropriately, these public sector data sources have very restricted access, from the FSRDC's requirement for physical presence and a special sworn status, to more incidental forms of access for administrative datasets. This heightens the appeal of private data sources, such as consumer reference and smart device data. While these datasets are certainly monetarily costly—they can cost tens or hundreds of thousands of dollars for a national panel—they are nevertheless available to almost anyone who has the resources. Though these datasets have tremendous potential to examine specific periods of time or geographies, especially sudden shocks that might result in residential mobility, they are challenging to handle, were not intentionally created for research purposes and, thus, are not representative of most study populations, and lack reliable individual characteristics such as demographic and socioeconomic information. More research on these datasets themselves—biases and representativeness, temporal consistency, and other ground-truthing—is necessary.

While scholars have already begun utilizing these novel data to answer many questions, their use by planning professionals can be considered emergent at best, given the technical challenges and costs typically associated with these types of data. Some providers such as Cuebiq and Safegraph have “data-for-good” programs in place, through which they make their data free to use for researchers and government agencies. Beyond this, however, local city planning departments or state-level agencies typically do not have the resources to fund the data acquisition or to use and maintain these datasets.

Conclusion

As residential mobility is one of the most important components of population change, understanding and measuring it is foundational to the planning profession and will likely continue to capture the attention of planning practitioners, policymakers, and social scientists alike. Our review covers the major developments in the sources of data used to measure residential mobility in the USA, together with the theoretical developments in residential mobility and some of the most important topical questions that such data developments have helped answer, from the early 20th century to the present.

Today, the data available to study residential mobility is richer and more abundant than ever before. For most geographies and time periods, the ACS provides reliable residential mobility estimates. With the advent of bespoke surveys and panel datasets that query respondents of their motivations for moving, as well as a plethora of data on respondent characteristics, we also have a much better understanding of why people move, though such detailed data may not be available at the smallest levels of geography or time frames. Finally, our ability to understand the spatial and temporal contexts under which moves occur is unprecedented: we can trace individuals’ residential histories at the neighborhood level using panels like the PSID or even down to the building level using consumer reference data or smartphone data. This growth in data sources has also enabled cross-validation of estimates of residential mobility, both among the different datasets and within, across different population groups and geographies. Despite these developments, there are no standalone data sources that measure all the spatial, demographic, housing, and neighborhood dimensions of residential mobility that these datasets, combined, can illuminate.

All the datasets together have empowered researchers to answer numerous questions related to residential mobility, not only those that advance our knowledge on it per se but also those that concern some of the most critical and pressing issues in planning and society more broadly. As our historical overview has illustrated, the developments in novel data and methodological approaches have consistently been marshaled to study phenomena that were contemporaneously transforming urban landscapes, including suburbanization and gentrification. Today, as the U.S. grapples with an unprecedented housing affordability crisis, researchers are utilizing the rich mobility data to produce cutting-edge insights and policy recommendations, including those that pertain specifically to planning and housing. These include using the Census Bureau's MAF data to construct vacancy chains to understand the impact of new suburban housing supply on urban housing affordability (French and Gilbert 2025); leveraging consumer reference data, together with other real estate data, to examine the effects of new market-rate housing on rents and migration flows in low-income areas (Asquith, Mast and Reed 2023); and utilizing the ACS microdata to analyze young adults’ “boomerang” moves back to their parents’ (Chan, Liao and O’Regan 2025).

Yet, as challenges and limitations remain with the data discussed in our review, there is ample room for further work and improvements. For public-sector data sources, striking the balance between enhancing data access and safeguarding privacy and confidentiality will be key to ensuring that the data are utilized to their full potential and are able to catalyze research as private-sector datasets have done, while offering a level of data standards, protections, and transparency that does not exist with the private-sector products. For private-sector data sources, researchers should spend more effort in description and ground truthing, as well as infrastructure creation and collaboration. Research on the representativeness of consumer reference data for residential mobility, for instance, is only nascent. The same is true for the space–time representativeness of cell phone data. These are involved endeavors. Researchers should consider building out pipelines that can be reused and, potentially, shared with others to collectively build expertise.

While academics have already begun taking advantage of relatively novel data sources to better understand patterns of residential mobility, these data sources and methodological approaches have unevenly penetrated professional practice due to reasons including lack of time and resources. Big data, such as consumer reference data, present potential to serve local planners who may seek information on granular and real-time residential mobility patterns, for example, to assess the impact of policy changes like rezoning or of new housing or infrastructure developments on residential mobility. We call on planners and academics to work more closely on such topics for two reciprocal reasons. First, researchers can offer access (e.g., to a FSRDC or to a commercial dataset), data experience, and methodological expertise to address planners’ critical questions and needs. Second, the space and time specificity of planners’ use of such data sources may compel researchers to ground truth their estimates in a more thoughtful manner.

Lastly, our review points to several remaining gaps in the literature and fruitful directions for future research. One is that measuring the residential mobility of the transient and most marginalized members of society continues to be a challenge as most of the data and methods that we have covered are premised on the assumption that people have addresses and that their moves are registered in the form of an address change. Following people without addresses is immensely difficult and a task for which there is not yet reliable methodology. Smartphone data offers promise given the pervasive ownership of cell phones across the general population. Alternatively, qualitative methodologies that engage deeply with individuals facing such predicaments may be the more appropriate approach (e.g., Desmond 2016). Diverse and innovative approaches will need to be marshaled to unlock insights on the residential mobility of those difficult to observe in traditional data.

Novel data have always been critical in allowing scholars of residential mobility to theorize, conceptualize, and test their frameworks. As discussed, residential mobility was considered a pathology until a new theoretical framework established that it was, in fact, a rational decision that resolved discrepancies between a household's current residential situation and their desired housing and neighborhood conditions. Since then, in the late-20th century, the life course perspective has emerged (Elder Jr, Johnson and Crosnoe 2003), which, by taking a relational view and emphasizing an individual's kinship and social networks as well as broader structural forces, has sought to better capture the diversity, dynamism, and de-standardization of life in the 21st century (Coulter, van Ham and Findlay 2016). Novel longitudinal data and methodological approaches should be instrumental in operationalizing the theoretical framework and uncovering links between residential mobility and structural conditions, particularly those that planners can address through planning and land-use regulation such as housing supply.

Indeed, as the U.S. navigates an unprecedented housing crisis and an associated drop in residential mobility rates (Myers, Park and Cho 2023), local mobility (and immobility) will continue to command much attention, for both theoretical advances and empirical analyses. While it is difficult to anticipate from where the next major theoretical and conceptual development on residential mobility might arise, the avalanche of data becoming available with which to study residential mobility certainly portends such a shift.

Supplemental Material

sj-xlsx-1-jpl-10.1177_08854122251382934 - Supplemental material for Measuring Residential Mobility: A Historical Overview of Novel Data and Methodological Approaches

Supplemental material, sj-xlsx-1-jpl-10.1177_08854122251382934 for Measuring Residential Mobility: A Historical Overview of Novel Data and Methodological Approaches by Lance Freeman, Yeonhwa Lee, Yining Lei and Wenfei Xu in Journal of Planning Literature