ArticlePDF Available

Urban Revival by Millennials? Intra‐Urban Net Migration Patterns of Young Adults, 1980–2010

April 2019
Journal of Regional Science 59(3)

DOI:10.1111/jors.12445

Authors:

Yongsung Lee

Georgia Institute of Technology

Bumsoo Lee

University of Illinois Urbana-Champaign

Md Tanvir Hossain Shubho

University of Illinois Urbana-Champaign

This study investigates neighborhood scale net migration of young adults in the top twenty urbanized areas (UAs) in the U.S. between 1980 and 2010. Both descriptive and regression analyses show that Generation Xers and Millennials were more likely to net migrate into central locations and less aversive to high density at their young ages than late boomers were in the 1980s. Consumption amenities are a critical factor that distinguishes the net migration patterns between young and old adult groups, and became a more important location factor for young adults in the 2000s (late Gen Xers and older Millennials) than in the 1990s (early Gen Xers). There exists a considerable degree of heterogeneity across UAs and neighborhoods even within the same UAs. This article is protected by copyright. All rights reserved.

Content uploaded by Yongsung Lee

Content may be subject to copyright.

Accepted Article

Yongsung Lee ORCID iD: 0000-0002-1980-1225

Urban Revival by Millennials? Intra-Urban Net Migration Patterns

of Young Adults, 1980–2010

Yongsung Lee

Georgia Institute of Technology

Bumsoo Lee

University of Illinois at Urbana–Champaign

Md Tanvir Hossain Shubho

University of Illinois at Urbana–Champaign

Contact author: Yongsung Lee, School of Civil and Environmental Engineering, Georgia

Institute of Technology, 790 Atlantic Drive, Atlanta, GA 30332. Email:

yongsung.lee@gatech.edu

Abstract

This study investigates neighborhood scale net migration of young adults in the top

twenty urbanized areas (UAs) in the U.S. between 1980 and 2010. Both descriptive and

regression analyses show that Generation Xers and Millennials were more likely to net

This article has been accepted for publication and undergone full peer review but has not

been through the copyediting, typesetting, pagination and proofreading process, which

may lead to differences between this version and the Version of Record. Please cite this

article as doi: 10.1111/jors.12445.

Accepted Article

migrate into central locations and less aversive to high density at their young ages than

late boomers were in the 1980s. Consumption amenities are a critical factor that

distinguishes the net migration patterns between young and old adult groups, and became

a more important location factor for young adults in the 2000s (late Gen Xers and older

Millennials) than in the 1990s (early Gen Xers). There exists a considerable degree of

heterogeneity across UAs and neighborhoods even within the same UAs.

Keywords: Millennial generation, residential location choice, urban revival, net

migration, consumption amenities, urbanism

Introduction

After more than half a century of suburbanization, the central cities of the largest

metropolitan areas in the U.S. have seen a trend reversal; in recent years, cities have

grown faster than their suburbs (Acolin, Voith, & Wachter, 2016). Mirroring this turn

toward urban revival, there was a significant decline in driving between the mid-2000s

and mid-2010s, after decades of near-constant growth in automobile use (McDonald,

2015; Zhong & Lee, 2017). According to the literature, young and educated population

groups are at the center of these shifts in mobility and urbanization trends (Baum-Snow

& Hartley, 2016; Millsap, 2016).

While many journalists and researchers have focused on how today’s young

adults, the millennials, are driving urban renewal, others view the resurgence of central

cities as transitory. Skeptics believe that the urban concentration and reduced automobile

use of young adults are an outcome of their economic struggles and delayed lifecycle

events (Manville, King, & Smart, 2017) or merely the manifestation of the peak

Accepted Article

Millennial birth cohorts (Myers, 2016). As Millennials age and their economic conditions

improve, some suggest that they might move back to car-oriented suburbs, just as their

parents did (Blumenberg, Ralph, Smart, & Taylor, 2016; Klein & Smart, 2017).

However, if the urban concentration of Millennials is largely lifestyle-driven, reflecting a

generational shift in mobility and location preferences, more sustained urban revival

supported by multi-modal transportation systems is likely (Davis, Dutzik, & Baxandall,

2012; McDonald, 2015).

Millennials, born between 1981 and 2000, are the largest demographic cohort today,

surpassing their Baby Boomer parents (Myers, 2016), and occupy more than a third of the

American workforce, exceeding Generation X (Frey, 2018). Hence, the decisions that

Millennials make about where and how they live, work, and travel will greatly (re-)shape

American cities for decades to come. It is an important but difficult question to address

whether the city-bound and carless lifestyle of today’s young adults will persist or not,

because most Millennials have not reached the age or a stage in life at which members of

their parents’ generation typically settled in the suburbs. One way to infer how urban

revival might turn out in the future is to examine whether and to what extent the location

choices of millennials are different from those of previous generations at the same ages.

That is the approach taken in this paper. We investigate the net migration patterns of

young and old adult populations at the census tract level in the twenty largest U.S.

urbanized areas for the thirty-year period between 1980 and 2010.

This study contributes to the literature by enhancing our understanding of the location

choices of the millennial generation. It is the first paper to analyze the net migration, as

Accepted Article

opposed to simple population change, of Millennials at such a fine geographical scale. As

such, it can distinguish movers from stayers, and capture the location preference of

today’s young adults more accurately than previous studies. The difference between the

two approaches, net migration vs. population change, and contrasting results are further

discussed in the research design section.

Second, this study examines how various neighborhood characteristics influence the net

migration of young adults of different generations. In doing so, we develop a rich set of

carefully operationalized physical urbanism and consumption amenities variables.

Moreover, this study offers a framework to compare the location choices of young adults

in three different decades, with each representing a different generation. Although we do

not claim these as generational preferences because our aggregate analysis cannot control

for other demographic and socio-economic characteristics, the results still provide

important insights into how young adults’ location choice has been changing over

generations. Finally, we report a considerable degree of heterogeneity in young adults’

location choice across regions by studying the twenty largest UAs and address

heterogeneity within UAs by conducting quantile regression analyses.

Based on net migration, we show that Generation Xers and Millennials are more likely to

move into central locations and less averse to high density at young ages than late

boomers were in the 1980s. Consumption amenities are found to be a critical location

factor that explains the centralizing tendency of young adults that is distinctive from the

net migration pattern of older adult groups. The association between consumption

Accepted Article

amenities and young adults’ net migration became stronger in the 2000s (late Gen Xers

and older Millennials) than in the 1990s (early Gen Xers).

The next section reviews a recent debate on the location and mobility choices of

millennials, identifying research gaps. Section 3 explains how we build our unique

datasets and variables, and describes how we conduct the analysis. Section 4 visually

presents the net migration patterns of four adult cohorts—youngest (20-24), young (25-

34), midlife (35-44), and middle age (45-64)—for the past three decades, using maps and

descriptive analysis tools. Section 5 presents key findings from ordinary least square

(OLS) and quantile regression analyses. Section 6 discusses the findings and offers

suggestions for future research.

Generations and Location Choice of Young Adults

The resurgence of downtowns and central cities, led by Millennials, has drawn attention

to the impact of generational change on urban development. A generation, the aggregate

of individuals who are born in the same time period and age together experiencing similar

historical events, tends to share similar values, preferences, and behaviors, and hence is

an important vehicle of societal transformation (Ryder, 1965). Generation research

particularly emphasizes the role of experience from adolescence and early adulthood in

forming collective memories and shared traits (Mannheim, 1952; Schuman & Scott,

1989). Generations also leave their imprints on cities mainly through the shifts in

locational preferences tied with their lifestyles and changing demands for consumption

spaces and community facilities (Filion & Grant, 2017). For example, post-war prosperity

and consumerism largely shaped location and housing preferences of Baby Boomers for

Accepted Article

large suburban single family homes, leading to massive auto-oriented suburban

developments (Filion & Grant, 2017).

Now, urban lifestyles of a significant segment of Millennials, largely influenced by

economic restructuring and societal changes, are driving the resurgence of high-density,

central city neighborhoods with better transit access and walkable streets (Moos, 2014).

Delayed lifecycle events including marriage and child rearing among today’s young

adults, combined with the pursuit of higher education (Patten & Fry, 2015; Settersten Jr

& Ray, 2010), lead to living in amenity rich central cities where rental units and public

transit are readily available. These important social trends are often associated with long-

term economic restructuring toward a more knowledge-oriented economy that has

dramatically increased the returns to education (Goldin & Katz, 2007; Taylor, Fry, &

Oates, 2014) and educational attainment of more recent generations. Young and highly

educated workers tend to live in dense urban neighborhoods with a large labor market

and rich “learning externalities” (Duranton & Puga, 2004; Glaeser, 1999; Peri, 2002).

They also tend to have fewer children and smaller families (Roy, Schumm, & Britt, 2014)

that better fit with urban living. One question that has been examined in the literature is

whether the compositional effect of increasing share of young adults with higher

education or the changing marginal effect of education is driving urban revival (Lee,

2018; Millsap, 2016).

Research has shown that young and educated population groups also value consumption

amenities such as shopping, social, cultural, and recreational opportunities that

downtowns and lively urban neighborhoods offer (Couture & Handbury, 2016; Edlund,

Accepted Article

Machado, & Sviatschi, 2015; Glaeser & Gottlieb, 2006). An important question regarding

the role of consumption amenities in urban revival is whether the amenities available in

downtown have increased or valuation of the amenities has increased over time. While

several studies show that the demand for urban amenities have been rising (Couture &

Handbury, 2016; Glaeser et al., 2001; Rappaport, 2008), little evidence of increasing

amenities in central locations is found in the literature other than falling crime rates

(Ellen, Mertens Horn, & Reed, 2017; Glaeser & Gottlieb, 2006).

Urban neighborhoods of large metropolitan areas are increasingly delineated by

household demography and generational differences (Moos, 2014, 2016) and one cannot

fully understand today’s neighborhood changes without “a generational lens” (Vinodrai,

Moos, & Pfeiffer, 2017). Will the urban resurgence by Millennials be transitory? Many

researchers who believe it is mainly caused by the prolonged recession hold the view that

young adults will move out to suburbs as they get a job, form a family, and have children

(Kotkin, 2016). Further, Myers (2016) argues that urban revival will subside as the

Millennial cohort passes young adulthood and several recent reports seem to support his

prognosis (Frey, 2017; Kolko, 2016, 2017; Kotkin, 2016). However, the return of today’s

young adults to suburbs may not be automatic if new lifecycle related norms are not just a

short-term response to the recession, but also a result of long-term post-industrial and

societal restructuring (Moos, 2014; Moos & Revington, 2017).

Several recent empirical studies examine the location choices of generation cohorts.

Moos (2014) investigates the concentration of young adults to high density and transit

neighborhoods in Vancouver and Montreal, Canada, but his study period (1980 and 2006)

Accepted Article

only covers Baby Boomers and Gen Xers and the findings of a case study in two cities

may not generalize to a broader population. Couture and Handbury’s (2016) analysis of

all metropolitan areas in the U.S. shows college-educated young adults are more attracted

to amenities like theaters and bars than non-college-educated and older adult groups, and

the gaps in locational preferences are generally becoming wider. While they use

consumption amenity indices constructed in an innovative way, their regression models

fail to include many other neighborhood attributes such as transportation variables. We

also believe the change in young adults’ share of census tract population, their dependent

variable, might underestimate the location preference of young adults because it does not

distinguish stayers from movers as further discussed in the research design section.

Other studies perform individual-level analyses using public use microdata. Millsap’s

(2016) linear probability model shows that the effect of education on young adults’

probability of living in a central city did not change significantly from 1990 to 2011.

Instead, increased educational attainment of the more recent generation accounts for the

increased urban presence of young adults. A more extensive analysis in a similar line by

Lee (2018) finds the effects of other generational characteristics including racial/ethnic

diversity and delayed life-cycle events as well as that of higher education. An individual

level analysis has certain advantages over an aggregate analysis. By controlling for

individual demographic and socio-economic characteristics, it can test a number of

hypotheses proposed to explain urban revival. In particular, one can distinguish the

changes in marginal effects of location factors from composition effects. A disadvantage

is that the lowest level of geographic aggregation identified in the data, the public use

microdata area, is too large to study the effects of neighborhood level characteristics. For

Accepted Article

example, the only neighborhood variable included in Millsap (2016) is a central city

dummy and distance from the CBD in Lee (2018).

The current study fills a gap in the literature by investigating the census tract level net

migration patterns, not a gross change of cohort size, of young and old adults in the top

twenty U.S. urbanized areas for the thirty-year period between 1980 and 2010. We use a

rich set of physical urbanism and consumption amenities variables and test how their

marginal effects have changed over generations.

Research Design

3.1: Research questions

The main purpose of this paper is to address the questions whether and to what extent the

Millennial generation is different from previous generations in residential location

choice. To address the key issue, we investigate the following research questions:

- When measured by the net migration of birth cohorts, not by gross change of the

same age groups, are Millennials moving into more urban neighborhoods in

central locations?

- What are the dominant location factors that attract Millennials (and young adults

in previous decades)?

- How significantly does the location preference of today’s Millennials differ from

those of early Generation Xers (Gen Xers) and Baby Boomers at the same ages in

previous decades?

- Is there a significant variation in the top twenty urban areas in intra-urban

relocation patterns of young adults?

Accepted Article

3.2: Study areas and age cohorts

We analyze the location choice of young adults in the largest twenty UAs in the U.S.—

i.e., the largest five UAs from each of the four Census Regions, according to the 2000

decennial U.S. Census. While Washington, D.C. and Baltimore UAs technically belong

to the Southern Region, they share more commonalities with Northeastern cities in spatial

structure and urban history. Thus, we include the two UAs in the Northeast. Previous

studies show that the resurgence of young populations in central cities is more

pronounced in large metropolitan areas. These UAs account for 33 percent of the total

U.S. population as of 2000 and include:

- Northeast: New York, Philadelphia, Boston, Washington, D.C., and Baltimore;

- Midwest: Chicago, Detroit, Minneapolis, St. Louis, and Cleveland;

- West: Los Angeles, San Francisco, Phoenix, Seattle, and San Diego;

- South: Miami, Dallas, Houston, Atlanta, and Tampa.

We use UA boundaries instead of metropolitan statistical areas (MSAs) used in most

previous studies mainly for two reasons. First, the use of UA boundaries allows us to

limit our analysis to urban and suburban census tracts. Since the building blocks of a

MSA are counties, MSAs include many rural communities within their boundaries, which

could distract the focus of the analysis. Second, UA boundaries are more consistently

delineated based on physical urban form than MSAs as the U.S. Census Bureau applies

the same minimum density thresholds across the regions.

UA boundaries are updated for each census year to reflect changing urbanization

patterns. To maintain the same study area for the thirty-year study period, we use 2000

Accepted Article

UA boundary definition. For each decade, we select beginning year census tracts whose

centroids are included in the 2000 UA boundaries. Then, we convert ending year census

data onto beginning year census tracts so that the data from the two census years are on

the same geography. We reorganize the data, using census tract relationship files

published by the U.S. Census Bureau for the 1990s and 2000s. But, since the relationship

file is not provided for the 1980s, we have to rely on area ratios between 1980 and 1990

census tracts.

Our analysis focuses on four adult population groups in each decade of our study period

from 1980 to 2010. In particular, the location choice of young adult cohort of age 25 to

34 (in the end year of each decade) is of the primary interest because this young age

group represents a different generation in each decade: the late Baby Boomers (defined as

those born between 1946 and 1964) in the 1980s, early Generation Xers (born between

1965 and 1980) in the 1990s, and late Gen Xers & early Millennials (born between 1981

and 2000) in the 2000s. This young adult group (25-34) also represents a stage in human

life in which they are most likely to have important life cycle events such as the first job,

home purchase, marriage, and childrearing. We have a separate age group of 20-24

because these youngest adults tend to be still in a college and/or are less likely to make

their own location choice than the young adult group of age 25-34. The location choice of

these two young adult groups will be compared with that of two older adult groups,

midlife (35-44) and middle age (45-65) groups.

Accepted Article

3.3: Estimation of net migration

In order to better capture the location decisions of adult population groups at different life

stages, we analyze the net migration patterns of birth cohorts at the census tract level as

opposed to a simple population change in the same age groups at varying points in time.

For example, Couture and Handbury (2016) compare the population share of the age

group 25–34 in 2010 (born in 1976–1985) with that of 2000 (born in 1966–1975).

Although convenient, this approach does not distinguish those who aged in the same

place from those who moved to the place; however, movers and stayers are likely to have

different location preferences. In particular, those young adults who aged in their parents’

home in the suburbs would be counted as an increase in young adult population by this

approach, even though their true location preference might be different. Thus, we

estimate and analyze the net migration of young adults who were 15 to 24 years old in the

beginning year and turned 25 to 34 years old in the ending year of each of the three

decades between 1980 and 2010. Figure A1 in the appendix highlights the contrasting

results of the two different approaches, net migration versus population change. When

location choice is measured by percent population change (panel A), the age 20-29 group

does not show a distinctive pattern from overall suburbanizing population (panel B) in

the 2000s in the Washington, DC UA. In contrast, the measure of net migration (panel C)

clearly shows a different location choice of older Millennials (20-29 in 2010). They have

concentrated into neighborhoods with rich consumption amenities that are mainly

clustered in the central city and along public transportation corridors (Panel D).

We estimate net migration using a modified cohort-component method because migration

data do not exist at sub-county levels. First, we estimate an expected population size in

Accepted Article

the ending (target) year for each age-sex cohort, using the cohort size in the beginning

(base) year and the age-and-UA-specific death rates published by the Centers for Disease

Control and Prevention (CDC). In the second step, we obtain the net migration of each

cohort by subtracting the expected cohort size from the actual cohort size in the ending

year. Positive net migration counts imply more in-migrants than out-migrants, and vice

versa. Since we operate with 5-year age cohort using decennial census data, we repeat the

same procedure twice for each decade.

We use the central county’s death rates for each UA since the CDC’s dataset has many

undisclosed numbers for smaller non-central counties. In general, the death rate of young

adult population (25-34) is higher in central urban areas than suburbs due to more violent

crimes, accidents, and poverty. This might lead to overestimated net migration of young

adults into suburban census tracts via overestimating their deaths and undercounting

stayers. As a result, the difference in young adults’ net migration between central cities

and suburbs may be presented in a slightly muted fashion in our analysis.

Pt, x+1 = Pb, x – Dx’ + IMx’ – OMx’

IMx’ – OMx’ = Pt, x+1 – (Pb, x – Dx’),

Where b is the base year; t is the target year; x is a sex-age cohort; x+1 is the next age cohort of

the same sex; x’ is the cohort x in the base year and aging to x+1 in the target year; Pb, x is the

population size of a 5-year age cohort in the base year; Pt, x+1 is the population size of the next age

cohort in the target year; IMx’ and OMx’ are the number of in-migrants and out-migrants during the

5 year period, respectively; and Dx’ is the number of death for the 5 year period.

Accepted Article

Maps in Figure 1 shows how the spatial patterns of net migration differ among the four

adult groups and changed over time, using the Washington, D.C. urbanized area as an

example. In all three decades, the youngest adult group of age 20 to 24 stands out in net

migrating toward more compactly developed neighborhoods along main public transit

corridors including downtown, whereas two older adult groups of age 35 to 64 show a

continued decentralization toward low-density suburbs. Interestingly, young adults of age

25 to 34 showed a not much different suburbanization pattern from that of midlife (35-

44) and middle age (45-64) groups in the 1980s, but their net-migration pattern became

similar to that of the youngest (20-24) group by the 1990s and even more so in the 2000s.

It seems that Gen Xers and Millennials (25-34 in 2000 and 2010, respectively) prefer

more urban neighborhoods than baby boomers (25-34 in 1990) at the same age at least in

the Washington, D.C region.

3.4: Measuring neighborhood level urbanism and consumption amenities

We operationalize urbanism in two dimensions: physical neighborhood characteristics

and consumption amenities. There has been an extensive effort over the past two to three

decades to define and quantify urban form components that facilitate sustainable

urbanism—density, diversity, design, distance to public transit, and destination access

(Cervero & Kockelman, 1997; Cervero, Sarmiento, Jacoby, Gomez, & Neiman, 2009;

Ewing & Cervero, 2010). We develop census tract level variables that correspond to

these urban form components except diversity for the beginning year of each decade—

i.e., 1980, 1990 and 2000—using various data sources including the Decennial Census,

historic Topologically Integrated Geographic Encoding and Referencing (TIGER) data,

the Census Transportation Planning Package (CTPP), and Homeland Infrastructure

Accepted Article

Foundation-Level Data (HIFLD). TIGER GIS files are used to calculate street density. To

identify the locations of rail transit stations in each census year, we use the public transit

stations GIS file from HIFLD and track each station’s opening year information from

transit agency websites and other sources. We define employment centers including the

CBD and subcenters in each UA, using census tract level employment data from the 2000

CTPP and a geographically weighted regression (GWR) procedure described in Lee

(2007) and Lee and Lee (2014).

Specific measurements include:

• Population density per acre;

• Street density to measure street design, street length (miles) per acre;

• Binary indicator of location within a half mile buffer from a frequent rail transit

station;

• Distance to the closest station for frequent rail transit in miles, Dstop=1/Distance;

• Binary indicator of location in the central city of the region;

• Distance to the central business district (CBD) in miles, DCBD=ln(Distance CBD);

We define the central city in two alternative ways. First, we define the largest

incorporated place in each UA as the central city and the rest as suburbs whether or not

they are incorporated. While the portion and role of the core central city varies from one

region to another, this political boundary still matters in terms of taxing and public

services provision—most importantly public transportation and school districts. Second,

Accepted Article

we also use the distance from the CBD to define an alternative set of inner cities census

tracts since many central municipalities include communities with suburban

characteristics such as San Fernando Valley in Los Angles. We experimented a range of

threshold distances used in previous studies (Baum-Snow & Hartley, 2017; Cortright,

2014; H. Lee, 2018) and settled with 5 miles. A descriptive analysis in Table 2 uses the

political definition of a central city and all regression models use the distance based

definition, 5 miles from the CBD. The results of regression models using alternative

central city definition will be presented in Table A5 in the appendix.

Using all input variables including the central city dummy and a factor analysis, we

extract final urbanism factors that we use in regression models for two decades, 1990-

2010. As shown in Table A1 in the appendix, we extract three variables, density, access

to transit, and central location. Any input variables showing low factor loadings are

excluded in the final factor analysis. We run three decade models (1980-2010) separately

with only available individual urban form variables because some of the input variables

are not available for 1980.

Urban researchers have only begun to operationalize the notion of consumption

amenities. Bereitschaft (2014, 2017) provides rich descriptions of “creative-cultural

districts” as nodes of cultural production and consumption that appeal to young

professionals. Couture and Handbury (2016) develop a consumption amenities index

based on the number of retail, entertainment, sports, restaurant, and other service

establishments. However, developing this kind of index dating thirty years back is nearly

impossible or prohibitively expensive. Thus, we use the number of jobs in retail trade and

Accepted Article

entertainment, recreation, and food services sectors from the 1990 and 2000 CTPP data

as a proxy for these businesses. A complete set of variables is available only for 1990 and

2000; but street design variables and consumption amenities are missing in the 1980 data

set. Thus, our three-decade regression models do not include consumption amenities

while two-decade models utilize a complete set of urbanism variables.

3.5: Regression models

To examine the roles of various location factors in attracting different age groups in each

of the three decades, we conduct a series of regression analyses. The unit of analysis is a

census tract and the data will be pooled across the twenty UAs and over the three

decades. Interaction terms of location factors and period indicators will enable us to

investigate how the impacts of these urban neighborhood attributes on net migration of

the same age groups change over time and generations.

𝑌

𝑖𝑗𝑡 =𝛼+𝛽1𝑋1𝑖𝑗𝑡 +𝛽2𝑋2𝑖𝑗𝑡 +𝛽3𝑇+𝛽4𝑈𝐴 +𝛾1𝑋1𝑇+𝜀𝑖𝑗𝑡,

Yijt is the dependent, net migration into census tract i in UA j in decade t; X1 is neighborhood

urbanism variables including consumption amenities; X2 is other control variables, including many

socio-demographic variables of each census tract, population size, and population share of the

same age group in the beginning year; T is dummy variables for each of the three decades; UA is

dummy variables for each of the 20 UAs; X1T is interaction terms of key urbanism variables with

decade indicators; α, βs, and γ are the coefficients to be estimated; and ε is an error term.

The dependent variable of the regression model is the net migration of the four different

age groups in each decade. The right hand side covariates include socio-demographic

Accepted Article

location factors and other control variables as well as the aforementioned variables that

measure neighborhood level urbanism. These include log population size, population

share of the same age group, percent non-Hispanic white population, percent foreign

born, percent professionals, log median household income, percent unemployment rate,

and percent multifamily housing stock. While school quality plays a critical role in the

residential location choice of those with school-age children, including our target cohort

of 25-34, reliable measures across 20 UAs and three decades are not available. Thus, we

collected a school quality measure for individual census tracts as of 2017 from a non-

profit organization (www.greatschools.org), and we imputed their historical values with

tract-level socioeconomic attributes. We find the results for our key variables are

consistent with and without such imputed variables. However, we decided to drop this

variable from our final models because of high multicollinearity with other socio-

economic variables.

We employ both ordinary least square (OLS) and quantile regression models. OLS

models are to examine the average impacts of location factors on net migration of adult

populations. We use quantile regression approaches because we consider a possibility

that the qualities comprising lively urban neighborhoods might only matter or are more

important in popular destinations of young adults, often called ‘hipster hotspots’.

Quantile regression models, by revealing heterogeneous correlation patterns between

covariates and the outcomes of interest at various quantile points, shed light on

heterogeneous behaviors across individuals or locations as in this study, and offer richer

policy insights than a conditional expectation at the mean would do (Koenker and

Hallock, 2001).

Accepted Article

As indicated in the equation above, we pool the data across 20 UAs and two/three

decades and use interaction terms to identify decade-specific coefficients for key

variables. When running a pooled model, we standardize most variables including the

dependent by UA and decade except those with substantive meanings such as a binary

indicator of location within a half mile from transit stations. We standardize variables to

take into account varying growth trends and spatial contexts across UAs. For example,

not very attractive neighborhoods in Phoenix, AZ might have more in-migration than

popular locations in Detroit, MI. Further, the most compactly developed neighborhoods

in Houston or Dallas, TX may not be so by New York’s standard. Standardization of

variables will also allow us direct comparison of coefficients across birth cohorts and

UAs since they are estimated in the unit of standard deviation.

Finally, we also run spatial models to deal with likely spatial auto correlation in census

tract level net-migrations. However, we are allowed to run spatial models only for each

decade separately since a spatial pooled model requires too large a contiguity matrix.

Thus, we choose OLS and quantile regression as our main results and include the results

of decade-specific spatial error, spatial lag, and spatial quantile models in the appendix.

Descriptive Analysis Results

Table 2 presents net migration patterns of the four age groups in the central city versus

suburbs for the three recent decades. Apparently, the youngest group of age 20-24 shows

the most distinctive location choice, largely moving into central cities in all three decades

except only a few cities in the Midwest and Baltimore. Two older adult groups, midlife

(35-44) and middle-age (45-64), show an exactly opposite location pattern away from

Accepted Article

central cities across the board in all three decades. There seems no generational turn in

these three adult groups.

It is young adults of age 25-34 that show substantial changes in location choice over time.

The net migration pattern of young adults in the 1980s—late Boomers in this period—

was not much different from that of two older adult groups. However, the same age group

who are early Gen Xers in the 1990s and late Gen Xers plus early Millennials in the

2000s shows much weaker preference for suburban locations than the previous

generation. Suburbs still receive relatively more young adults than central cities even in

the 1990s and 2000s, but the gaps became much smaller than were in the 1980s. Thus,

the touted shift in location choice of young (25-34) adults seems to have come earlier

with Generation X than are generally speculated. While “downtown rebound” by young

adults in the 1990s was well reported in several earlier studies (Eugenie Ladner Birch,

2002; Eugenie L Birch, 2005; Sohmer & Lang, 2001), the current study suggests a

possibility that this trend was present across the central city beyond the downtown

boundary or the downtown rebound was strong enough to lead to a positive net migration

of young (25-34) adults, aggregately measured at the central city level.

It is also notable that there is considerable variation across UAs. The central city has

always been relatively more or equally attractive to young (25-34) adults in San

Francisco and New York; whereas a shift toward this trend has come with early Gen Xers

in the 1990s in Seattle and with late Gen Xers and Millennials in the 2000s for

Washington, D.C. and Miami. On the contrary, net-migration rates have been higher in

Accepted Article

suburban areas throughout the three decades in several UAs such as Boston, Cleveland,

Detroit, and Minneapolis.

Figure 2 also corroborates a similar trend over the three decades. Here, we estimate a

spline regression of percent net migration on distance from the CBD for each age group

separately by the Census Region and decade. Young adults’ (25-34) net migration rates

were an increasing function of the distance from the region’s CBD (except far outskirt

areas of the Eastern UAs) just as that of older age groups in all four regions in the 1980s.

However, the spline regression lines of young adults in the 1990s (early Gen Xers)

became much flatter and some significant peaks in downtown areas emerged in the 2000s

when late Gen Xers and older Millennials grew to be young (25-34) adults. Other three

age groups do not show a significant change in their location preferences for central

locations over the three decades.

Location factors driving intra-urban location choice of young adults

5.1: Average effects of neighborhood attributes on net migration

Table 3 shows how neighborhood attributes of individual census tracts in the beginning

year influence the net in-migration of four adult cohorts in each decade. Models (1) - (4)

present the results of three-decade models with three urban form variables—central city

dummy, density and transit access—and Models (5) - (8), two-decade models, include

three urbanism factor scores and a consumption amenities variable. We interact all key

variables with each of the three decade dummy variables to examine potential generation

effects. Overall, all age groups seem to be attracted toward central locations but avoid

high density neighborhoods. The impacts of three urbanism variables on net-migration,

Accepted Article

when viewed on average as presented by OLS regression models, seem to work in the

same direction across the four age groups although there is a variation in the magnitude.

It is the consumption amenities variable that shows more diverging preferences between

young and old adult groups.

Surprisingly, the central city dummy (defined as five miles from the CBD) and central

location factor score show mixed patterns for midlife group (35-44) and middle age (45-

64), often showing significant and positive effects on net migration. This counter-

intuitive result appears to conflict the well-established suburbanization trend in U.S.

metropolitan areas. We believe a part of the reason is spatial autocorrelation. The five

mile dummy variables show either a negative sign or insignificance for the two older

adult groups (35-44 and 45-64) in spatial regression results presented in table A3. It

should also be noted that the size of negative coefficients of density is substantially larger

than that of central city variable for the two older groups. What we observe in the

previous section’s descriptive analysis is the net effects. In case of two younger adult

groups (20-24 and 25-34), the relative coefficient sizes of central city and density are vice

versa.

OLS estimates under alternative definitions of the central city, the largest municipality

and the area within 3, 5, and 10 miles from the CBD, show similar results (Table A3).

Regardless of the definition, the youngest (20-24) adult group’s centralization pattern is

the strongest in the 2000s while young (25-34) adults show a significant shift toward

stronger centralization in the 1990s, which is consistent with the descriptive analysis

findings in the previous section. The net migration of young adults into central

Accepted Article

neighborhoods is stronger in the 1990s (early Gen Xers) when the central city is defined

broadly (the largest city and 10 miles from the CBD); it is stronger in the 2000s (late Gen

Xers and older Millennials) when the central neighborhoods are defined narrowly (3 and

5 miles).

Access to public transit is also valued especially by young (25-34) and midlife (35-44)

groups, but not by the youngest (20-24). The transit factor score actually shows a

negative sign for the youngest in the 2000s although it is not significant in spatial models

shown in Table A3. This is puzzling as the youngest adults are the most frequent transit

users (Brown, Blumenber, Taylor, Ralph, & Voulgaris, 2016; Buehler & Pucher, 2012).

It may be because our transit access variable only captures the access to rail transit but

not bus systems. Metro lines typically serve more prominent parts of the city including

downtowns and metro access is more likely to be capitalized into housing prices and rents

than bus access. Thus, high metro access neighborhoods can be unaffordable to poor

young households.

It is consumption amenities—proxied by the density of retail, entertainment, recreation,

and food services employment—that have the most contrasting effects on net migration

between young and old adult groups. Whereas both midlife (35-44) and middle age (45-

64) groups moved away from mixed-land use neighborhoods, the youngest (20-24) group

were attracted by these consumption, entertainment, and cultural opportunities in the two

recent decades. Young adults of age 25-34 were indifferent in the 1990s (early Gen

Xers), but came to value the access to consumption amenities in the 2000s (late Gen Xers

& older Millennials). The coefficient for the youngest (20-24) is also larger in the 2000s

Accepted Article

(Millennials) than in the 1990s (late Gen Xers). This finding is consistent with the

consumption amenities literature that shows the demand for urban amenities have been

rising over the years (Couture & Handbury, 2016; Glaeser et al., 2001; Rappaport, 2008).

The increasing marginal effect of consumption amenities on young adults’ location

decision has an important implication on urban development in the future although more

research needs to be done on whether the changes are largely composition effects or a

generational preference shift.

5.2: Heterogeneous effects of neighborhood attributes on net migration

The OLS regression results in the previous section shows that our target cohort of age 25-

34 at the end of each decade moved into the neighborhoods in central locations with

better transit access but with lower density. However, the OLS results do not identify a

clear generational difference in young (25-34) adult’s net migration between the 1990s

(early Gen Xers) and the 2000s (late Gen Xers & older Millennials) except that

consumption amenities became a significant location factor in the 2000s. We suspect that

outliers on both ends of our net migration measure might have substantially affected OLS

estimates. Thus, this section presents the results of quantile regression models for the

target cohort under the same specification, which would reveal varying effects of

urbanism attributes across the neighborhoods with different levels of young (25-34)

adults’ net-migration.

Overall, as shown in Table 4, there are considerable variations in the size of coefficients,

with the impacts of all key variables generally increasing with the level of net migration.

In other words, the effects of urbanism attributes are stronger in those neighborhoods that

Accepted Article

attract more young (25-34) adults. This is the case for nearly all urbanism variables

regardless of whether the impact is positive or negative. It should be also noted that the

OLS estimates are considerably larger than those at the median, which is suggestive of

the skewing effects of outliers on the OLS estimates.

The changes over time in the coefficients at the 75 percentile are not much different from

OLS results. The net migration of young (25-34) adults into central and mixed-use

neighborhoods became stronger in more recent decades. The negative impact of high

density has also weakened over time. The only variable whose coefficient is moving in

the opposite direction is transit access. As discussed above, this trend seems to indicate

that the access to metro stations are becoming increasingly unaffordable to young adults.

5.3: Heterogeneous effects in top twenty urbanized areas

Table 5 presents the coefficients of the consumption amenities variable from quantile

regression models of young adult’s (25-24) net-migration run for each of the top 20 UAs

and the last two decades separately. There is a considerable degree of heterogeneity both

in terms of the effects and their change over time. Interestingly, net migration’s

association with consumption amenities became much stronger in the 2000s in the four

Western UAs except Phoenix; however this trend is not present in the other three regions.

In fact, strong consumption amenities effects observed in the 1990s for many Eastern and

Midwestern UAs including New York, Washington, D.C., Detroit, and Minneapolis,

became less or non-significant in the 2000s. Further research is needed to better interpret

this result. One notable finding is that young adults in the South including a Western sun-

Accepted Article

belt city Phoenix are not significantly attracted to mixed-use neighborhoods in both

recent decades.

In sum, the recent urban revival by young adult population may not be present in all cities

and regions, but is more pronounced in certain types of urban areas, especially in the

West, that offer neighborhoods with concentrated urban amenities in central cities.

Conclusions

This paper investigates census tract level net migration patterns of younger and older

adult populations in the twenty largest UAs for the thirty-year period between 1980 and

2010. Descriptive analysis finds that young adults of age 25-34 show a considerable shift

in their location choice toward central locations in the 1990s, not the 2000s, indicating a

possibility that the shift in young adults’ location choice occurred earlier with Gen Xers

than is generally speculated. OLS and quantile regression analysis results also suggest

that young adults (25-34) in the 1990s share more commonality in location preference

with those of the 2000s (a mix of older Millennials and late Gen Xers) than with late

Boomers at their young ages in the 1980s. The two recent generations prefer more central

locations and are less averse to high density than late Boomers were in the 1980s.

The location choices of younger and older adult groups show the most divergence

with respect to the role of consumption amenities. Whereas both midlife (35-44) and

middle-age (45-64) groups moved away from mixed-land use neighborhoods, the

youngest adults of age 20-24 were attracted by consumption, entertainment, and cultural

opportunities in the two recent decades. Young adults of age 25-34 were indifferent in the

Accepted Article

1990s (early Gen Xers), but net-migrated into the neighborhoods with a higher access to

consumption amenities in the 2000s (late Gen Xers & early Millennials).

We also find a considerable degree of heterogeneity in the effects of urban neighborhood

attributes on young (25-34) adults’ net migration, both across UAs and neighborhoods.

Quantile regression results show that both the positive effects of central location and

transit access and the negative impact of density are substantially larger in the

neighborhoods and UAs that are popular among young (25-34) adults. The recent urban

revival by young population may not be present in all cities and regions, but is more

pronounced in certain types of cities that offer neighborhoods with concentrated urban

amenities in the central city.

A limitation of the paper is inherent in the aggregate level of observation. This study

cannot control for the varying demographic and socio-economic attributes of the

members in the four age cohorts in different decades. Thus, although this study can

explore aggregate changes in the relationships between location factors and net migration

over time, it cannot estimate the extent to which these changes are driven by generational

preferences or changes in cohort compositions. In contrast, an analysis of young adult’s

location choice at the individual level would enable us to examine interactions between

location factors and individual level characteristics. As the current study suggests, these

interactions may have changed over time with generations. Lee’s (2018) individual level

multinomial logit analysis is an example of this approach. However, a trade-off involved

in using publicly available microdata is a loss in spatial resolution. A study using

Accepted Article

restricted-use microdata at the Federal Statistical Research Data Center, with geography

identified at lower levels, seems to be a promising next step in this literature.

References

Acolin, A., Voith, R., & Wachter, S. (2016). city and usburbs--Hs there been a regime

change? Penn Institute for Urban Research White Paper. Retrieved from

http://www.penniur.upenn.edu/uploads/media/City_and_Suburbs_%E2%80%93_

Has_There_Been_a_Regime_Change_web.pdf

Baum-Snow, N., & Hartley, D. (2016). Causes and Consequences of Central

Neighborhood Change, 1970–2010. Paper presented at the Research

Symposium on Gentrification and Neighborhood Change.

Baum-Snow, N., & Hartley, D. A. (2017). Accounting for central neighborhood change,

1980-2010.

Bereitschaft, B. (2014). Neighbourhood change among creative–cultural districts in mid-

sized US metropolitan areas, 2000–10. Regional Studies, Regional Science,

1(1), 158-183.

Bereitschaft, B. (2017). Do “creative” and “non-creative” workers exhibit similar

preferences for urban amenities? An exploratory case study of Omaha,

Nebraska. Journal of Urbanism: International Research on Placemaking and

Urban Sustainability, 10(2), 198-216. doi: 10.1080/17549175.2016.1223740

Birch, E. L. (2002). Having a Longer View on Downtown Living. Journal of the American

Planning Association, 68(1), 5-21. doi: 10.1080/01944360208977188

Birch, E. L. (2005). Who lives downtown. Washington: Brookings Institution.

Accepted Article

Blumenberg, E., Ralph, K., Smart, M., & Taylor, B. D. (2016). Who knows about kids

these days? Analyzing the determinants of youth and adult mobility in the U.S.

between 1990 and 2009. Transportation Research Part A: Policy and Practice,

93, 39-54. doi: http://doi.org/10.1016/j.tra.2016.08.010

Brown, A. E., Blumenber, E., Taylor, B. D., Ralph, K., & Voulgaris, C. T. (2016). A taste

for transit? Analyzing public transit use trends among youth. Journal of Public

Transportation, 19(1), 49-67.

Buehler, R., & Pucher, J. (2012). Demand for Public Transport in Germany and the USA:

An Analysis of Rider Characteristics. Transport Reviews, 32(5), 541-567. doi:

10.1080/01441647.2012.707695

Cervero, R., & Kockelman, K. (1997). Travel demand and the 3Ds: density, diversity,

and design. Transportation Research D, 2(3), 199-219.

Cervero, R., Sarmiento, O. L., Jacoby, E., Gomez, L. F., & Neiman, A. (2009). Influences

of Built Environments on Walking and Cycling: Lessons from Bogotá.

International Journal of Sustainable Transportation, 3(4), 203-226. doi:

10.1080/15568310802178314

Cortright, J. (2014). The young and restless and the nation’s cities. City Observatory, 8.

Couture, V., & Handbury, J. (2016). Urban revival in America, 2000 to 2010. University

of Pennsylvania. Memo.

Davis, B., Dutzik, T., & Baxandall, P. (2012). Transportation and the new generation:

Why young people are driving less and what it means for transportation policy.

Accepted Article

Duranton, G., & Puga, D. (2004). Micro-foundations of urban agglomeration economies.

Handbook of regional and urban economics, 4, 2063-2117.

Edlund, L., Machado, C., & Sviatschi, M. M. (2015). Bright minds, big rent: gentrification

and the rising returns to skill: National Bureau of Economic Research.

Ellen, I. G., Mertens Horn, K., & Reed, D. (2017). Has falling crime invited gentrification?

US Census Bureau Center for Economic Studies Paper (No. CES-WP-17-27).

Ewing, R., & Cervero, R. (2010). Travel and the built environment--A Meta-Analysis.

Journal of the American Planning Association, 76(3), 265-294.

Filion, P., & Grant, J. (2017). The impact of generational change on cities The Millennial

City (pp. 29-42): Routledge.

Frey, W. (Producer). (2018, January, 2019). Millennials are the largest generation in the

U.S. labor force. Retrieved from http://www.pewresearch.org/fact-

tank/2018/04/11/millennials-largest-generation-us-labor-force/

Frey, W. H. (Producer). (2017, January, 2019). City growth dips below suburban growth,

census shows. Brookings Institution, Washington, DC. . Retrieved from

https://www.brookings.edu/blog/the-avenue/2017/05/30/city-growth-dips-below-

suburban-growth-census-shows/

Glaeser, E. L. (1999). Learning in cities. Journal of Urban Economics, 46, 254-277.

Glaeser, E. L., & Gottlieb, J. D. (2006). Urban Resurgence and the Consumer City.

Urban Studies, 43(8), 1275-1299. doi: 10.1080/00420980600775683

Accepted Article

Glaeser, E. L., Kolko, J., & Saiz, A. (2001). Consumer city. Journal of Economic

Geography, 1, 27-50.

Goldin, C., & Katz, L. F. (2007). Long-run changes in the US wage structure: narrowing,

widening, polarizing: National Bureau of Economic Research.

Klein, N. J., & Smart, M. J. (2017). Millennials and car ownership: Less money, fewer

cars. Transport Policy, 53, 20-29. doi:

http://dx.doi.org/10.1016/j.tranpol.2016.08.010

Kolko, J. (2016). Urban revival? Not for most Americans. Retrieved from

http://jedkolko.com/2016/03/30/urban-revival-not-for-most-americans/

Kolko, J. (Producer). (2017, January, 2019). Americans' shift to the suburbs sped up last

year. FiveThirtyEight. Retrieved from

https://fivethirtyeight.com/features/americans-shift-to-the-suburbs-sped-up-last-

year/

Kotkin, J. (2016). The human city: Urbanism for the rest of us. Evanston, IL: Agate

Publishing.

Lee, B. (2007). "Edge" or "edgeless" cities? Urban spatial structure in U.S. metropolitan

areas, 1980 to 2000. Journal of Regional Science, 47(3), 479-515.

Lee, H. (2018). Are Millennials Coming to Town? Residential Location Choice of Young

Adults. Urban Affairs Review, 0(0), 1078087418787668. doi:

10.1177/1078087418787668

Accepted Article

Lee, S., & Lee, B. (2014). The influence of urban form on GHG emissions in the U.S.

household sector. Energy Policy, 68(0), 534-549. doi:

http://dx.doi.org/10.1016/j.enpol.2014.01.024

Mallach, A. (2017). What does the march of the Millennials mean for the future American

city? In M. Moos, D. Pfeiffer & T. Vinodrai (Eds.), The Millennial City (pp. 267-

285): Routledge.

Mannheim, K. (1952). The problem of generations. In P. Kecskemeti (Ed.), Karl

Mannheim: Essays (pp. 276-322). London, UK: Routledge.

Manville, M., King, D. A., & Smart, M. J. (2017). The Driving Downturn: A Preliminary

Assessment. Journal of the American Planning Association, 83(1), 42-55. doi:

10.1080/01944363.2016.1247653

McDonald, N. C. (2015). Are Millennials Really the “Go-Nowhere” Generation? Journal

of the American Planning Association, 81(2), 90-103. doi:

10.1080/01944363.2015.1057196

Millsap, A. (2016). Location choice in early adulthood: Millennials versus Baby Boomers.

Papers in Regional Science, n/a-n/a. doi: 10.1111/pirs.12240

Moos, M. (2014). “Generationed” space: Societal restructuring and young adults'

changing residential location patterns. The Canadian Geographer/Le Géographe

canadien, 58(1), 11-33.

Moos, M. (2016). From gentrification to youthification? The increasing importance of

young age in delineating high-density living. Urban Studies, 53(14), 2903-2920.

doi: 10.1177/0042098015603292

Accepted Article

Moos, M., & Revington, N. (2017). Will Millennials remain in the city? Residential mobility

in post-industrial, post-modern, post-suburban America. In M. Moos, D. Pfeiffer &

T. Vinodrai (Eds.), The Millennial City (pp. 183-199): Routledge.

Myers, D. (2016). Peak Millennials: Three Reinforcing Cycles That Amplify the Rise and

Fall of Urban Concentration by Millennials. Housing Policy Debate, 1-20. doi:

10.1080/10511482.2016.1165722

Patten, E., & Fry, R. (2015). How millennials today compare with their grandparents 50

years ago. Pew Research Center, 19.

Peri, G. (2002). Young workers, learning, and agglomerations. Journal of Urban

Economics, 52(3), 582-607. doi: http://dx.doi.org/10.1016/S0094-1190(02)00510-

Rappaport, J. (2008). Consumption amenities and city population density. Regional

Science and Urban Economics, 38(6), 533-552. doi:

https://doi.org/10.1016/j.regsciurbeco.2008.02.001

Roy, R. N., Schumm, W. R., & Britt, S. L. (2014). Voluntary Versus Involuntary

Childlessness Transition to Parenthood (pp. 49-68): Springer.

Ryder, N. B. (1965). The Cohort as a Concept in the Study of Social Change. American

Sociological Review, 30(6), 843-861. doi: 10.2307/2090964

Schuman, H., & Scott, J. (1989). Generations and collective memories. American

Sociological Review, 359-381.

Settersten Jr, R. A., & Ray, B. (2010). What's going on with young people today? The

long and twisting path to adulthood. The future of children, 19-41.

Accepted Article

Sohmer, R. R., & Lang, R. E. (2001). Downtown rebound. Redefining urban and

suburban America: Evidence from census 2000, 1, 63-74.

Taylor, P., Fry, R., & Oates, R. (2014). The rising cost of not going to college. Pew

Research Center.

Vinodrai, T., Moos, M., & Pfeiffer, D. (2017). Millennial cities of tomorrow. In M. Moos, D.

Pfeiffer & T. Vinodrai (Eds.), The Millennial City (pp. 287-294): Routledge.

Zhong, L., & Lee, B. (2017). Carless or car later? Declining car ownership of Millennial

households in the Puget Sound Region, Washington State. Transportation

Research Record: Journal of the Transportation Research Board, 2664, 69-78.

doi: 10.3141/2664-08

Tables

TABLE 1. Generations represented by Four Adult Cohorts in Each Census Year

Adult cohorts

Age

1980

1990

2000

2010

Youngest

-24

Baby Boomers

Generation Xers

Millennials

Young

-34

Baby Boomers

Generation Xers

Millennials & Gen Xers

Midlife

-44

Silent Generation

Baby Boomers

Generation Xers

Middle age

-64

Silent Generation

Baby Boomers

Notes: This study defines Millennials as those born between 1981 and 2000; Generation Xers between 1965

and 1980; Baby Boomers between 1946 and 1964; and Silent Generation between 1925 and 1945.

Accepted Article

TABLE 2. Percent Net Migration of Four Adult Groups to Central Cities vs. Suburbs,

1980-2010

a) 1980 to 1990

b) 1990 to 2000

CC Suburbs CC Suburbs CC Suburbs CC Suburbs

Baltimore -10 5-11 29 -14 4-13 -3

Boston 63 -3 -35 9-31 -4 -12 -8

New York 15 -22 16 15 -4 2-8 -9

Philadelphia 3-22 -19 14 -14 4-11 -6

Washinton DC 26 26 -17 43 -25 14 -16 -7

Chicago -1 -22 -10 13 -23 0-18 -9

Cleveland -11 -38 -26 8-24 -3 -17 -10

Detroit -17 -27 -32 7-26 -4 -21 -11

Minneapolis 61 -4 -25 32 -32 6-16 -8

St. Louis -10 -32 -23 12 -16 -3 -15 -10

Atlanta 10 34 -21 65 -23 30 -18 4

Dallas 36 26 865 -14 18 -13 0

Houston 23 3-16 67 -28 30 -20 4

Miami -4 11 548 629 -6 19

Tampa 8 6 -9 53 -1 42 -6 29

Los Angeles 59 28 15 19 -7 1-12 -9

Pheonix 13 49 28 48 10 45 -1 30

San Diego 60 49 936 729 -1 8

San Francisco 62 17 18 17 -10 3-8 -10

Seattle 63 5-5 35 -2 19 -8 -1

UA Name

Youngest (20-24)

Young (25-34)

Midlife (35-44)

Middle Age (45-64)

CC Suburbs CC Suburbs CC Suburbs CC Suburbs

Baltimore 15 14 -24 11 -28 10 -14 -1

Boston 80 6-29 0-45 3-11 -6

New York 39 -3 16 5-9 12 -5 -5

Philadelphia 21 -13 -19 2-20 7-9 -2

Washinton DC 46 24 -18 17 -32 8-17 -9

Chicago 32 -8 811 -24 8-10 -3

Cleveland 1-36 -7 10 -22 4-11 -1

Detroit -29 -17 -7 13 -25 2-19 -5

Minneapolis 94 14 -5 12 -44 7-17 -6

St. Louis 12 -16 -13 2-34 1-13 -6

Atlanta 49 45 10 50 -20 29 -13 4

Dallas 72 23 16 23 -20 10 -10 -3

Houston 47 -16 13 36 -18 15 -9 1

Miami 316 629 -1 27 019

Tampa 27 8-1 17 -5 21 -1 23

Los Angeles 38 14 -3 -2 -25 -10 -16 -12

Pheonix 56 60 30 41 939 145

San Diego 56 34 -10 -5 -16 0-7 -1

San Franci sco 75 35 44 16 -24 -6 -9 -7

Seattle 103 16 25 19 -29 12 -6 -2

UA Name

Youngest (20-24)

Young (25-34)

Midlife (35-44)

Middle Age (45-64)

Accepted Article

c) 2000 to 2010

Notes: Percent net migration = number of net migrants/cohort population in the beginning year*100. The

concept of central places for urban areas is no longer used since the 2010 Census. We define the largest

incorporated place in each UA as the central city and the rest as suburbs whether or not they are

incorporated.

TABLE 3. Net Migration of Four Adult Cohorts, 1980-2010 (OLS Regression)

Three-decade models Two-decade models

(1)

Youngest

(20-24)

(2)

Young

(25-34)

(3)

Midlife

(35-44)

(4)

Middle

(45-64)

(5)

Youngest

(20-24)

(6)

Young

(25-34)

(7)

Midlife

(35-44)

(8)

Middle

(45-64)

Central city '80s 0.025

** 0.078

*** 0.111

*** 0.065

***

Central city '90s 0.031

*** 0.123

*** -0.047

*** -0.085

***

Central city '00s 0.130

*** 0.175

*** 0.027

** -0.088

***

Central city '90s*

0.015

** 0.083

*** -0.005

0.062

***

CC Suburbs CC Suburbs CC Suburbs CC Suburbs

Baltimore 31 13 126 -19 -5 -5 7

Boston 109 7-28 12 -48 -9 -9 -2

New York 38 -7 14 14 -23 -4 -8 -4

Philadelphia 46 -21 -2 12 -17 -2 -5 1

Washinton DC 89 19 15 36 -28 -4 -6 -2

Chicago 25 -31 211 -34 -1 -19 -3

Cleveland -7 -53 -19 14 -31 -4 -15 -2

Detroit -33 -26 -38 9-30 -6 -29 -2

Minneapolis 82 -13 -9 17 -43 -15 -22 -6

St. Louis 32 -27 312 -27 -6 -10 1

Atlanta 69 7-1 20 -29 1-17 0

Dallas 32 -3 -3 17 -31 -2 -15 -3

Houston 34 -20 742 -21 18 -12 4

Miami 38 629 17 1 2 9 8

Tampa 31 13 219 -11 2-6 14

Los Angeles 36 02-1 -20 -11 -11 -8

Pheonix 223 -10 1-22 -6 -15 12

San Diego 64 22 -4 1-26 -6 -11 0

San Francisco 92 20 27 11 -36 -12 -7 -9

Seattle 99 -4 17 29 -26 1-15 -3

UA Name

Youngest (20-24)

Young (25-34)

Midlife (35-44)

Middle Age (45-64)

Accepted Article

Central city '00s*

0.057

*** 0.050

*** 0.083

*** 0.024

Population density '80s -0.059

*** -0.146

*** -0.145

*** -0.262

***

Population density '90s 0.037

*** -0.065

*** -0.283

*** -0.182

***

Population density '00s -0.021

*** -0.111

*** -0.269

*** -0.281

***

Population density '90s*

0.018

*** -0.061

*** -0.252

*** -0.189

***

Population density '00s*

-0.027

*** -0.079

*** -0.226

*** -0.269

***

Access to transit '80s 0.011

*** 0.021

*** 0.025

*** 0.023

***

Access to transit '90s 0.001

0.014

*** 0.004

** 0.005

***

Access to transit '00s 0.001

0.011

*** 0.007

***

Access to transit '90s*

0.005

0.085

*** 0.003

0.007

Access to transit '00s*

-0.015

** 0.053

*** 0.027

*** 0.008

Consumption amenity ' 90s

0.055

*** 0.003

-0.033

*** 0.004

Consumption amenity ' 00s

0.076

*** 0.033

*** -0.062

*** -0.010

decade '90s 0.000

0.027

*** 0.045

*** 0.035

***

decade '00s -0.027

*** 0.046

*** 0.071

*** 0.026

*** -0.010

* 0.023

*** 0.043

*** -0.005

ln(population) 0.016

*** 0.068

*** -0.061

*** -0.112

*** 0.019

*** 0.042

*** -0.106

*** -0.120

***

%Youngest (20-24) at beginning 0.281

***

0.256

***

%Young (25-34) at beginning

0.127

***

0.125

***

%Midlife (35-44) at beginning

-0.047

***

-0.038

***

%Middle (45-64) at beginning

0.011

***

0.044

***

Accepted Article

% non-Hispanic White 0.132

*** 0.108

*** 0.011

*** 0.097

*** 0.109

*** 0.137

*** 0.049

*** 0.149

***

% foreign born 0.077

*** 0.030

*** 0.027

*** 0.054

*** 0.052

*** 0.012

*** 0.005

0.033

***

% professionals 0.059

*** 0.068

*** -0.001

-0.085

*** 0.036

*** 0.069

*** -0.004

-0.128

***

ln(median household income) -0.118

*** -0.032

*** 0.122

*** 0.044

*** -0.095

*** -0.015

*** 0.146

*** 0.060

***

% unemployed -0.088

*** -0.027

*** 0.054

*** 0.007

-0.091

*** -0.032

*** 0.060

*** 0.011

% multifamily housing units 0.309

*** -0.044

*** -0.193

*** 0.097

*** 0.296

*** -0.051

*** -0.247

*** 0.054

***

Observations 52,331

52,331

36,866

36,847

36,848

36,835

R2 0.444

0.181

0.296

0.132

0.446

0.189

0.338

0.122

Adjusted R2 0.443

0.180

0.295

0.132

0.445

0.188

0.337

0.121

Notes: Models (1) to (4) present the results of three-decade models from 1980 to 2010 and models (5) to (8)

are two-decade models from 1990 to 2010. The dependent variable of all models are the net migration of

four birth cohorts, standardized for each decade and UA. Here, standardization helps compare coefficients

across cohorts and UAs while taking into account the varying contexts of individual UAs. The binary

indicator variables for each UA are included in the models, but not shown here for brevity. All key

variables are interacted with dummy indicators of each decade to examine generation effects. The sample

of each model excludes top and bottom 1% in terms of net-migration to minimize the influence of outliers.

* Urbanism variables for two-decade models are factor scores, derived from multiple urban form variables

(see Table A1 for details).

TABLE 4. Net Migration of Young (25-34) Adult, 1980-2010 (OLS and Quantile

Regression)

Three-decade models Two-decade models

(1) (2) (3) (4) (5) (6) (7) (8)

25th 50th 75th OLS 25th 50th 75th OLS

Central city '80s 0.063

*** 0.047

*** 0.038

*** 0.078

***

Central city '90s 0.054

*** 0.078

*** 0.096

*** 0.123

***

Accepted Article

Central city '00s 0.045

*** 0.093

*** 0.161

*** 0.175

***

Central city '90s*

0.056

*** 0.058

*** 0.057

*** 0.083

***

Central city '00s*

0.009

0.019

*** 0.044

*** 0.050

***

Population density '80s -0.038

*** -0.105

*** -0.218

*** -0.146

***

Population density '90s 0.002

-0.039

*** -0.128

*** -0.065

***

Population density '00s -0.036

*** -0.070

*** -0.137

*** -0.111

***

Population density '90s*

0.032

*** -0.034

*** -0.136

*** -0.061

***

Population density '00s*

0.006

-0.039

*** -0.112

*** -0.079

***

Access to transit '80s 0.010

** 0.026

*** 0.041

*** 0.021

***

Access to transit '90s 0.009

*** 0.018

*** 0.032

*** 0.014

***

Access to transit '00s 0.004

*** 0.007

*** 0.018

*** 0.011

***

Access to transit '90s*

0.067

*** 0.071

*** 0.090

*** 0.085

***

Access to transit '00s*

0.033

*** 0.031

*** 0.055

*** 0.053

***

Consumption amenity '90s

-0.009

** 0.006

0.017

*** 0.003

Consumption amenity '00s

-0.009

* 0.017

*** 0.033

***

decade '90s 0.031

*** 0.037

*** 0.021

*** 0.027

***

decade '00s 0.064

*** 0.071

*** 0.025

*** 0.046

*** 0.029

*** 0.027

*** 0.001

0.023

***

Observations 53,401

53,401

52,331

37,603

36,847

0.181

0.189

Adjusted R2

0.180

0.188

Accepted Article

Notes: Models (1) to (4) present the results of three-decade models from 1980 to 2010 and models (5) to (8)

are two-decade models from 1990 to 2010. The first three models of each study period present quantile

regression estimates at the 25th, 50th, and 75th percentiles, respectively. The specifications of all models in

Table 4 are the same as in models (2) and (6) in Table 3 although the results for control variables are not

shown here for brevity.

TABLE 5. Heterogeneity in Young (25-34) Adult’s Preference for Consumption

Amenities

1990s 2000s

25th 50th 75th 25th 50th 75th

Baltimore 0.025 -0.047 0.024 -0.019 -0.004 0.027

Boston -0.030 -0.005 -0.019 0.011 0.035 0.020

New York 0.016 0.027 0.092 -0.024 -0.010 0.005

Philadelphia 0.000 0.014 0.022 0.008 0.016 0.031

Washington, DC 0.011 0.012 0.035 -0.041 -0.006 0.048

Chicago -0.026 0.013 0.004 -0.037 -0.013 0.013

Cleveland -0.008 0.033 0.045 0.018 0.029 0.074

Detroit 0.027 0.026 0.053 0.032 0.006 -0.008

Minneapolis -0.007 0.007 0.093 -0.012 0.027 0.059

St. Louis -0.037 -0.050 -0.095 -0.038 0.001 -0.014

Atlanta 0.058 0.023 -0.005 -0.011 0.014 0.022

Accepted Article

Dallas 0.057 0.076 0.031 -0.011 0.006 -0.049

Houston 0.031 0.019 0.043 -0.025 -0.007 0.048

Miami -0.031 -0.033 -0.080 0.042 0.053 0.036

Tampa 0.016 0.026 -0.046 0.023 0.064 0.008

Los Angeles 0.004 0.020 0.030 0.018 0.033 0.058

Phoenix -0.022 -0.013 -0.016 -0.082 -0.054 -0.061

San Diego -0.110 -0.027 0.014 0.014 0.035 0.087

San Francisco -0.030 -0.016 0.080 0.070 0.063 0.068

Seattle -0.022 0.002 0.063 0.050 0.028 0.086

Notes: The table presents the coefficients of consumption amenities on the net migration of Young (25-34)

cohort from 120 quantile regression models (20 UA × three quantile points × two decades). The net

migration is standardized by UA and decade to help compare coefficients across models. The same set of

control variables as those in Model (6) in Table 3 are included but not shown for brevity. Colored cells

indicate statistically significant coefficients at least at p=0.1: darker red indicates significant at p<0.01 and

negative in sign, and darker green indicates significant at p<0.01 and positive in sign.

Accepted Article

1980s 1990s 2000s

a) Youngest (20-24)

1980s 1990s 2000s

b) Young (25-34)

1980s 1990s 2000s

c) Midlife (35-44)

Accepted Article

1980s 1990s 2000s

d) Middle Age (45-64)

FIGURE 1. Net Migration Patterns of Four Adult Cohorts in the Washington, DC

Urbanized Area, Notes: All maps present the percent net-migration (Number of net-migration/cohort

population in the beginning year) of the corresponding population cohort in each decade.

1980-1990 1990-2000 2000-2010

a) Youngest (20-24)

1980-1990 1990-2000 2000-2010

b) Young (25-34)

Accepted Article

1980-1990 1990-2000 2000-2010

c) Midlife (35-44)

1980-1990 1990-2000 2000-2010

d) Middle age (45-64)

FIGURE 2. Percent Net-Migration by the Distance from the CBD, 1980-2010, Notes: The

above lines are created from spline regression of percent net-migration on distance from the CBD by

Census Region. The data exclude outliers—top 2.5% and bottom 2.5%—and were pooled by Census

Region.

Accepted Article

Appendix

TABLE A1. Exploratory Factor Analysis Result for Neighborhood Characteristics

Density Access

to public

transit

Central

location

Final

communality

estimates

ln(street density) a 0.75

5 0.12

8 0.24

1 0.645

ln(population density) a 0.75

0 0.18

5 0.27

4 0.671

Half mile from the nearest transit facility b 0.17

1 0.66

3 0.24

3 0.528

1/(the distance to the nearest transit facility)

0.10

1 0.63

9 0.14

3 0.439

Census tract in the central city boundary b 0.31

1 0.31

8 0.59

9 0.556

ln(the distance to the CBD) a -

0.427 -

0.273 -

0.595 0.611

Variance explained by each factor 1.45

1 1.07

4 0.92

Notes: a indicates that variables are standardized by UA and decade before the exploratory factor analysis

(EFA), and b indicates binary variables. Because of data availability, EFA is conducted for the two

decades, 1990-2010. As for the rotation method, varimax is used to make the three factors orthogonal to

one another.

Accepted Article

TABLE A2. Summary Statistics of Key Variables

Variable names

1980's 1990's 2000's

mea

n s.d. min med

ian max mea

n s.d. min med

ian max mea

n s.d. min med

ian max

Net migration, 20-24

(count)a 33 414 -

2,0

44 -16 7,78

3 69 398 -

4,88

0 7 10,9

17 38 375 -

5,01

5 -12 6,95

Net migration, 25-34

(count)a 106 615 -

8,8

52 42 13,6

85 80 551 -

11,2

05 36 21,1

96 73 482 -

11,3

01 56 11,2

Net migration, 35-44

(count)a -1 303 -

3,5

92 -21 5,75

8 -4 408 -

2,38

1 -25 18,8

25 -74 307 -

2,82

2 -62 11,4

Net migration, 45-64

(count)a -67 243 -

2,1

62 -83 6,21

9 -37 287 -

1,13

0 -53 13,1

48 -39 248 -

1,90

5 -55 9,07

Distance to CBD

(mile) 13.3 10.5 0.0 10.3 83.6 13.4 9.8 0.0 10.8 79.9 14.4 10.8 0.0 11.5 81.1

Within the political

boundary 0.41 0.49 0.0

0 0.00 1.00 0.38 0.49 0.00 0.00 1.00 0.36 0.48 0.00 0.00 1.00

In 3 miles from CBD

(mile) 0.09 0.28 0.0

0 0.00 1.00 0.09 0.28 0.00 0.00 1.00 0.08 0.27 0.00 0.00 1.00

In 5 miles from CBD

(mile) 0.19 0.39 0.0

0 0.00 1.00 0.18 0.39 0.00 0.00 1.00 0.17 0.37 0.00 0.00 1.00

Population density (per

acre) 20.5 69.0 0.0 10.3 7,86

6.8 19.2 30.8 0.0 9.3 1,07

6.5 19.5 30.3 0.0 9.2 359.

1/(distance to nearest

transit) 0.6 2.0 0.0 0.0 119.

9 1.0 3.2 0.0 0.1 180.

5 1.0 3.0 0.0 0.2 180.

Factor score: central

cityb - - - - - 0.01 0.68 -

3.13 -

0.24 2.45 -

0.01 0.66 -

3.18 -

0.24 2.42

Accepted Article

Factor score:

population densityb - - - - - 0.00 0.81 -

5.09 0.16 4.12 0.00 0.82 -

4.21 0.13 3.30

Factor score: access to

transitb - - - - - 0.00 0.75 -

0.74 -

0.26 23.4

5 0.00 0.73 -

0.80 -

0.26 23.4

Consumption job

density (per acre) - - - - - 1.48 6.23 0.00 0.51 315.

80 1.33 6.48 0.00 0.40 270.

Population size 4,35

6 2,44

0 37 3,96

9 39,4

28 4,39

5 2,63

8 0 4,01

0 71,8

94 4,42

7 2,22

8 0 4,16

2 34,0

% 20-24 in the

beginning year 9.3

% 3.7

% 0.0

% 8.9

% 59.5

% 7.9

% 4.1

% 0.0

% 7.3

% 65.8

% 6.7

% 4.1

% 0.0

% 6.2

% 80.0

% 25-34 in the

beginning year 16.8

% 5.2

% 0.0

% 15.9

% 53.1

% 18.6

% 6.0

% 0.0

% 18.1

% 100.

0% 15.7

% 6.1

% 0.0

% 14.9

% 100.

% 35-44 in the

beginning year 11.5

% 2.9

% 0.1

% 11.0

% 26.9

% 15.2

% 3.8

% 0.0

% 14.9

% 100.

0% 16.2

% 3.3

% 0.0

% 16.2

% 99.9

% 45-64 in the

beginning year 20.9

% 5.6

% 0.2

% 20.8

% 57.7

% 18.8

% 5.6

% 0.0

% 18.5

% 66.7

% 21.4

% 5.7

% 0.0

% 21.2

% 99.9

% Non-Hispanic

White 68.4

% 34.0

% 0.0

% 85.7

% 100.

0% 61.6

% 34.6

% 0.0

% 76.2

% 100.

0% 53.2

% 33.8

% 0.0

% 61.5

% 100.

% Foreign born 12.2

% 11.8

% 0.0

% 8.4

% 85.7

% 14.9

% 14.8

% 0.0

% 9.8

% 100.

0% 20.0

% 16.9

% 0.0

% 14.8

% 100.

% Professionals 11.3

% 7.7

% 0.0

% 9.7

% 67.3

% 14.4

% 9.9

% 0.0

% 12.9

% 100.

0% 17.1

% 10.7

% 0.0

% 15.6

% 100.

Median household

income ($) 19,2

46 8,46

7 2,5

00 18,3

60 75,0

00 36,4

12 18,2

88 0 34,3

59 150,

001 50,2

52 25,0

45 0 46,1

31 200,

001

% Unemployed 7.1

% 5.3

% 0.0

% 5.6

% 45.5

% 7.6

% 6.9

% 0.0

% 5.4

% 100.

0% 7.2

% 6.9

% 0.0

% 5.0

% 100.

% Multifamily

housing units 43.5

% 32.0

% 0.0

% 36.8

% 100.

0% 40.6

% 32.0

% 0.0

% 33.6

% 100.

0% 39.4

% 31.4

% 0.0

% 32.4

% 100.

Accepted Article

a: Net migration counts are computed for the ten year period in each decade while all other variables show

the values for beginning years.

TABLE A3. Spatial Regression of the Net Migration of Four Adult Cohorts

(A) Spatial lag models

With urbanism variables With urbanism factors

Youngest

(20-24)

Young

(25-34)

Midlife

(35-44)

Middle

(45-64)

Youngest

(20-24)

Young

(25-34)

Midlife

(35-44)

Middle

(45-64)

Central city '80s -0.003

0.066

*** 0.024

-0.051

***

Central city '90s 0.033

** 0.114

*** 0.014

-0.079

***

Central city '00s 0.119

*** 0.173

*** 0.013

-0.024

Central city '90s (factor score)

0.010

0.074

*** 0.014

0.054

***

Central city '00s (factor score)

0.055

*** 0.055

*** 0.033

***

Population density '80s -0.073

*** -0.176

*** -0.213

*** -0.272

***

Population density '90s 0.021

*** -0.060

*** -0.286

*** -0.181

***

Population density '00s -0.004

-0.089

*** -0.228

*** -0.258

***

Population density '90s (factor score)

-0.001

-0.068

*** -0.262

*** -0.188

***

Population density '00s (factor score)

-0.011

** -0.069

*** -0.221

*** -0.264

***

Access to transit '80s 0.007

*** 0.020

*** 0.018

*** 0.014

***

Access to transit '90s 0.000

0.012

*** 0.005

*** 0.004

Access to transit '00s 0.003

** 0.012

*** 0.009

*** 0.012

***

Access to transit '90s (factor score)

-0.006

0.076

*** -0.005

-0.008

Access to transit '00s (factor score)

-0.007

0.060

*** 0.031

*** 0.020

Accepted Article

Consumption amenity '90s

0.041

*** -0.006

-0.042

*** -0.024

***

Consumption amenity '00s

0.086

*** 0.040

*** -0.048

*** 0.015

* Goodness of fit measure: the squared covariance of the actual and predicted net migration

1980s 0.463

0.188

0.177

0.182

1990s 0.432

0.185

0.351

0.108

0.433

0.190

0.334

0.105

2000s 0.465

0.208

0.349

0.161

0.474

0.205

0.346

0.153

(B) Spatial error models

With urbanism variables With urbanism factors

Youngest

(20-24)

Young

(25-34)

Midlife

(35-44)

Middle

(45-64)

Youngest

(20-24)

Young

(25-34)

Midlife

(35-44)

Middle

(45-64)

Central city '80s -0.002

0.065

*** 0.029

* -0.051

***

Central city '90s 0.032

** 0.115

*** 0.021

-0.076

***

Central city '00s 0.119

*** 0.173

*** 0.014

-0.025

Central city '90s (factor score)

0.010

0.075

*** 0.019

** 0.053

***

Central city '00s (factor score)

0.055

*** 0.054

*** 0.056

*** 0.032

***

Population density '80s -0.072

*** -0.178

*** -0.214

*** -0.273

***

Population density '90s 0.022

*** -0.060

*** -0.285

*** -0.183

***

Population density '00s -0.004

-0.091

*** -0.227

*** -0.260

***

Population density '90s (factor score)

0.000

-0.067

*** -0.263

*** -0.190

***

Population density '00s (factor score)

-0.011

** -0.070

*** -0.220

*** -0.266

***

Accepted Article

Access to transit '80s 0.007

*** 0.019

*** 0.016

*** 0.013

***

Access to transit '90s 0.000

0.012

*** 0.005

*** 0.004

Access to transit '00s 0.003

** 0.012

*** 0.009

*** 0.011

***

Access to transit '90s (factor score)

-0.005

0.076

*** -0.009

-0.008

Access to transit '00s (factor score)

-0.007

0.059

*** 0.031

*** 0.020

Consumption amenity '90s

0.041

*** -0.005

-0.041

*** -0.022

***

Consumption amenity '00s

0.086

*** 0.041

*** -0.047

*** 0.015

* Goodness of fit measure: the squared covariance of the actual and predicted net migration

1980s 0.461

0.186

0.182

0.181

1990s 0.431

0.178

0.356

0.108

0.433

0.184

0.339

0.103

2000s 0.465

0.205

0.350

0.159

0.474

0.201

0.347

0.150

Notes: Spatial models are run for each decade separately though they are presented in the same way as the pooled non-

spatial models in Table 3 so that they are more comparable. The results for other control variables are not shown here

for brevity. As a goodness of fit measure, the squared covariance of the actual and model-predicted net migration is

computed, which is an equivalent to the R squared of the OLS regression.

TABLE A4. Spatial Quantile Regression of the Net Migration of Young (25-34)

Cohort

With urbanism variables With urbanism factors

25th 50th 75th Spatial lag Spatial error 25th 50th 75th Spatial lag Spatial error

Central city '80s 0.018

* 0.042

*** 0.054

*** 0.066

*** 0.065

***

Central city '90s 0.040

*** 0.070

*** 0.097

*** 0.114

*** 0.115

***

Central city '00s 0.074

*** 0.096

*** 0.149

*** 0.173

***

Accepted Article

Central city '90s*

0.042

*** 0.046

*** 0.054

*** 0.074

*** 0.075

***

Central city '00s*

0.021

*** 0.033

*** 0.052

*** 0.055

*** 0.054

***

Population density '80s -0.056

*** -0.122

*** -0.244

*** -0.176

*** -0.178

***

Population density '90s 0.010

* -0.040

*** -0.129

*** -0.060

***

Population density '00s -0.016

*** -0.056

*** -0.117

*** -0.089

*** -0.091

***

Population density '90s*

0.028

*** -0.035

*** -0.149

*** -0.068

*** -0.067

***

Population density '00s*

0.020

*** -0.037

*** -0.101

*** -0.069

*** -0.070

***

Access to transit '80s 0.009

** 0.022

*** 0.041

*** 0.020

*** 0.019

***

Access to transit '90s 0.008

*** 0.014

*** 0.029

*** 0.012

***

Access to transit '00s 0.006

*** 0.011

*** 0.024

*** 0.012

***

Access to transit '90s*

0.055

*** 0.058

*** 0.072

*** 0.076

***

Access to transit '00s*

0.040

*** 0.041

*** 0.060

*** 0.059

***

Consumption amenity '90s

-0.025

*** -0.004

0.002

-0.006

-0.005

Consumption amenity '00s

-0.001

0.021

*** 0.040

*** 0.041

***

* Goodness of fit measure: the squared covariance of the actual and predicted net mi gration

1980s

0.188

0.186

1990s

0.185

0.178

0.190

0.184

2000s

0.208

0.205

0.201

Notes: Spatial models are run for each decade separately. The statistical software employed for the estimation of spatial

quantile regression, the qregspiv command of the McSpatial package in R, does not return goodness-of-fit measures.

Also, note that the spatial lag and error models above are borrowed from Table A3 for comparison.

Accepted Article

TABLE A5. Robustness Check under Alternative Definitions for the Central City Census

Tracts

Definition of Central City

Three-decade models

Youngest

(20-24)

Young

(25-34)

Midlife

(35-44)

Middle

(45-64)

The largest city

1980s 0.052

*** 0.043 *** 0.186 *** 0.219 ***

1990s 0.019 * 0.124 *** 0.042 *** 0.146 ***

2000s 0.085 *** 0.069 *** 0.116 *** 0.072 ***

3 miles from CBD

1980s 0.022 0.102 *** 0.115 *** 0.066 ***

1990s 0.013 0.143 *** -0.017 -0.119 ***

2000s 0.161 *** 0.263 *** 0.029

-0.067 ***

5 miles from CBD

1980s 0.025 ** 0.078 *** 0.111 *** 0.065 ***

1990s 0.031 *** 0.123 *** -0.047 *** -0.085 ***

2000s 0.130 *** 0.175 *** 0.027 ** -0.088 ***

10 miles from CBD

Accepted Article

1980s 0.023 ** 0.004

0.124 *** 0.099 ***

1990s 0.034 *** 0.094 *** -0.055 *** 0.034 ***

2000s 0.062 *** 0.050 *** 0.053 *** -0.040 ***

Notes: The table presents the coefficients of central city binary indicators from four OLS models for each

age group, each with an alternative central city definition. Model specifications are the same as those in

Table 3.

A) % change in population, age group 20-29 B) % change in the total population

C) % net migration of older Millennials (20-29) D) Consumption amenities index

FIGURE A1. Net Migration Patterns vs. Population Changes in the Washington, DC

Urbanized Area, 2000-2010, Notes: The tone of colors in each panel indicates the decile of individual

tracts regarding the four measures: the darkest denotes the top decile, and the lightest denotes the bottom

decile. Panels A) and B) show simple population changes (%) of age group 20-29 and the total population,

respectively. In comparison, Panel C) presents percent net-migration patterns of the birth cohort who grew

up to be 20 to 29 years old in 2010. Panel D) maps the consumption amenities index, which is estimated via

a factor analysis of the densities of amenities-related businesses including cafes, restaurants, bars, and night

clubs.

The Declining Impact of Walkability and Transit Accessibility on U.S. Home Values During the COVID-19 Pandemic

Article

Full-text available

Mar 2025

Bradley Bereitschaft

The Migration Trajectories of Young Adults in Prague's Hinterland

Article

Full-text available

Jun 2024
SOCIOL CAS

Otakar Bursa

Back to the suburbs? Millennial residential locations from the Great Recession to the pandemic

Article

Jan 2024

In the past decade, there has been a great deal of attention paid to and speculation about the residential mobility and location decisions of millennials. Academics and practitioners alike have been trying to determine where millennials are moving and why, including whether they are leading a ‘back to the city’ movement or whether they are moving to the suburbs as previous generations did at their age. Using US Census data, this article examines the geographical population distribution of young adults in the USA in recent decades. Categorising neighbourhoods by their urban or suburban character and by their central or peripheral location, we find that millennials lived in urban areas on the heels of the Great Recession at higher rates than previous generations. However, over the decade, the millennial population gradually shifted towards suburban areas: central urban and peripheral urban neighbourhoods largely lost millennial residents from 2011 to 2021, while peripheral suburban neighbourhoods experienced substantial gains. When it comes to neighbourhood amenities (e.g. restaurants and parks), millennials largely left amenity-rich areas for neighbourhoods with fewer amenities, though these amenities grew faster in the neighbourhoods that gained millennials the most. Millennial suburbanisation seems to be associated with housing affordability and demand for larger homes, as the population shift was more pronounced in the metros that have lower housing affordability and a lower share of larger homes in their central urban neighbourhoods. The results indicate the importance of affordable and right-sized housing, complemented with neighbourhood amenities, in attracting and retaining this population group.

The Long-Lasting Impact of Past Mobility Dependence on Travel Mode Share in a New Neighborhood: The Case of the Seoul Metropolitan Area, South Korea

Article

Full-text available

Oct 2023

Travel behavior researchers have dominantly explored the influence of increase in development densities with mixed pattern of land-uses, and investment in infrastructures related to public transit toward more sustainable-transportation policies. However, little has been known about the long-term interdependencies between people’s decisions on travel behavior and individual biographies relating to residential relocation and habitual behavior over a longer time period. To fill this gap, the present study aims to investigate the long-lasting impact of past travel behavior on current travel behavior after a process of residential relocation. For this purpose, aggregate analysis at a neighborhood level was carried out, focusing on cause-effect relationships between current travel mode share and the size of in-migrants dependent on a certain mode in the past by using Household Travel Surveys (HTS) and Internal Migration Statistics collected during 2006–2015 (10 years). Accordingly, the size of in-migrants who have their pre-determined travel behavior in the past play an important role in explaining the mode share of a neighborhood on the current state. Further, this study attempted to divide the influence of residential relocation from the influence of habitual behavior on travel behavior after residential relocation. The finding suggested that the habitualized travel pattern can affect the travel patterns in the new neighborhood even after separating the effect of self-selection. Specifically, the past dependency on public transportation represented significant influences on the current travel mode share. This study on travel behavior informs consideration of role of habitual qualities during the process of residential relocation.

Out‐of‐area home purchase and U.S. internal migration

Article

Aug 2024

This study demonstrates that out‐of‐area (OOA) property transactions can serve as a proxy for migration. Using micro‐level transaction data, we document that about 35% of migrants make OOA property purchases. The goodness‐of‐fit between migration and OOA purchases is higher for aggregate migration measures and lower for migration flows between disaggregated areas. Furthermore, in most specifications, a one percent increase in OOA purchases is associated with an approximately one percent increase in migration. We characterize the monthly out‐migration from NYC zip codes to surrounding areas after the outbreak of the COVID‐19 pandemic to demonstrate the high temporal and spatial resolution of OOA transaction data.

Are young adults’ choices temporary or permanent? Navigating economic recession with residential location and household expenditure

Article

Apr 2024

Jaeyong Shin

PhD stipends and program placement success in political science

Article

Apr 2024

A key grievance of the student labor movement is that across much of academia, and especially in the social sciences and humanities, stipends tied to PhD assistantships fall short of a living wage. In this article, we consider the issue from a pedagogical perspective, expecting that higher pay may lead to stronger program outcomes. We collect and validate data on assistantship stipends in political science from PhDStipends.com, and on tenure-track placements from an analysis of departmental placement pages. Graduate pay is significantly associated with tenure-track placements in the job market cycles spanning 2019–2021, independently of program size, rank, student unionization, location, and institution type and endowment. Across model specifications, a US$5,000 increase in student pay corresponds with 2.7 more placements per 100 enrolled students (or 34% of the median rate) over this period.

Young workers, jobs-housing balance, and commute distance: Findings from two high-housing-cost U.S. regions

Article

Apr 2024
CITIES

Da convergência para a divergência regional: impactos da terceira revolução industrial nas desigualdades regionais brasileiras

Article

Apr 2023

João Carlos Ramos Magalhães

A desigualdade regional no Brasil é um tema de grande complexidade, refletindo contrastes socioeconômicos em todo o país. Analisando dados desde 1920 até 2016, Magalhães e Alves (2021) destacam uma trajetória não linear da desigualdade, com um período de aumento seguido por uma queda a partir de 1970. Apesar disso, a região Sudeste apresentou um preocupante aumento da disparidade a partir de 1996. O ensaio levanta questões sobre a continuidade ou reversão dessas tendências, contrastando teorias que preveem uma redução contínua das desigualdades regionais com dinâmicas que indicam um possível retorno à divergência espacial. O estudo também explora teorias que relacionam desigualdades regionais com estágios de desenvolvimento econômico. Este ensaio oferece uma análise detalhada desses padrões e suas implicações para o futuro do desenvolvimento regional no Brasil.

Neighborhood and housing choices of out-of-town versus local movers: The case of Charlotte, North Carolina

Article

Nov 2023

Carless or Car Later?: Declining Car Ownership of Millennial Households in the Puget Sound Region, Washington State

Article

Full-text available

Jan 2017

Driving has been declining since the mid-2000s after six decades of constant growth, and millennials are allegedly leading the trend. This study investigates the decline of car mobility by using vehicle ownership data in the Puget Sound region of Washington State, where transit ridership is increasing faster than in most other large metropolitan areas. Poisson regression and simple simulation results show that millennial-specific impacts such as changing location and mobility preferences were associated with about 15% to 32% of the reduction in car ownership by young adults between 2002 and 2014; this impact is comparable to the effect of sociodemographic shifts (20% to 28%) and the impact of lowering income and employment (16% to 24%). The combination of the increasing influence of urban form and the increasing share of young adults living in compact neighborhoods was associated with the remaining 6% to 7% of the change. These findings suggest that declining car mobility observed in the past decade or so is likely to persist in coming decades, and the trend will be reinforced where smart growth policies to make more livable and sustainable cities receive further momentum.

The Driving Downturn: A Preliminary Assessment

Article

Full-text available

Jan 2017

Problem, research strategy, and findings: We examine why American driving fell between 2004 and 2013, weighing two explanations: that Americans voluntarily moved away from driving (“peak car”), and that economic hardship reduced driving. We analyze aggregate data on travel, incomes, debt, public opinion, and Internet access. These data lack the precision of microdata, but unlike microdata are available annually for years before, during, and after driving’s decline. We find substantial evidence for the economic explanation. During the downturn the cost of driving rose while median incomes fell. The economy grew overall, but did so unequally. Mass driving requires a mass middle class, but economic gains accrued largely to the most affluent. We find less evidence for “peak car.” If Americans voluntarily drove less, they would likely use other modes more. However, despite heavy investment in bicycle infrastructure and public transportation in the 2000s, demand for these modes remained flat while driving fell. Takeaway for practice: If Americans were voluntarily abandoning automobiles for other modes, planners could reduce investments in automobile infrastructure and increase investments in alternative mobility. Driving’s decline, however, was not accompanied by a transit surge or substantial shift to other modes. The lesson of the driving downturn is that people drive less when driving’s price rises. Planners obviously do not want incomes to fall, but they should consider policies that increase driving’s price. Planners might also rethink the current direction of U.S. transit policy; transit use did not rise even when driving fell at an unprecedented pace.

Millennials and car ownership: Less money, fewer cars

Article

Full-text available

Jan 2017
TRANSPORT POLICY

Americans are driving less. The changes are most pronounced among Millennials, those born in the 1980s and 1990s. Much ink has been spilled debating whether these changes in travel behavior are due to changing preferences or economic circumstances. In this paper, we use eight waves of data from the Panel Study of Income Dynamics (PSID) to examine recent changes in auto ownership among US families with a particular focus on Millennials. We find that today's young adults do own fewer cars than previous generations did when they were young. However, when we control for whether young adults have become economically independent from their parents, i.e. left the nest, we find that economically independent young adults own slightly more cars than we would expect, given their low incomes and wealth. We caution planners to temper their enthusiasm about “peak car,” as this may largely be a manifestation of economic factors that could reverse in coming years.

Will Millennials remain in the city?

Chapter

Aug 2017

The impact of generational change on cities

Chapter

Aug 2017

(Millennial) cities of tomorrow

Chapter

Aug 2017

Accounting for Central Neighborhood Change, 1980-2010

Article

Dec 2019

After 1980-2000 population decline and economic stagnation, downtown neighborhoods in most large US cities experienced 2000-2010 population growth and gentrification. Stark racial differences in valuations of downtown amenities and suburban labor market opportunities among those with less than college that only emerged after 2000 are the primary drivers of these downtown neighborhood dynamics. As college-educated whites moved in, increases in valuations of downtown amenities encouraged other whites to remain in downtowns, a reversal from prior decades. Continued departures of less than college educated minorities were driven both by relative improvements in suburban employment opportunities for this group and their continued declines in valuations of downtown amenities. Our evidence highlights the importance of racial differences in valuations of potentially endogenous local amenities for understanding recent downtown gentrification.

Are Millennials Coming to Town? Residential Location Choice of Young Adults

Article

Jul 2018

Hyojung Lee

How widespread is the downtown resurgence among the current generation of young adults 25 to 34 years old, often referred to as the Millennials? Answering the question, this article provides a detailed analysis of the past, current, and future geographic population distribution, focusing on the young adults. It first documents the demographic shifts from 1990 to 2015 by distance from the central business districts (CBDs) of the 50 largest U.S. metros. The results of the intrametropolitan analysis reconcile the back-to-the-city thesis with sustained suburban growth, explaining how the two sides have reached different conclusions. Then this article empirically examines the generational residential patterns using a multinomial logistic regression, suggesting the generational characteristics of the Millennials, at least partially, explain their inclination to choose city center over suburbs. The simulations based on the model estimates indicate that ongoing demographic shifts might contribute to more population growth in urban centers over the next two decades.

Has Falling Crime Invited Gentrification?

Article

Jan 2017

Who knows about kids these days? Analyzing the determinants of youth and adult mobility in the U.S. between 1990 and 2009

Article

Nov 2016
TRANSPORT RES A-POL

The 2000s was a decade of transitions for teens and young adults. In comparison with previous generations of youth, those living in the developed world (i) faced the harshest economic climate in decades, (ii) lived with their parents longer and were more likely to return back home as young adults, (iii) used information and communication technologies (ICTs) extensively, and (iv) in the U.S., were subject to increasingly stringent graduated driver’s licensing (GDL) regulations. All were dramatic societal changes to be sure, but how did they affect youth travel behavior? Some argue dramatically and enduringly, but usually with fragmentary evidence. We examine data from the three most recent U.S. national travel surveys and find that, with one exception, after controlling for personal, household, locational, and travel factors, the effects of factors associated with various societal trends on person-kilometers traveled (PKT) are surprisingly muted. The exception is that decreased employment is associated with substantially lower PKT; however, this effect is 32% greater among older (ages 27–61) than younger (ages 20–26) adults, suggesting that economic factors, rather than changes in youth travel preferences, were at the root of declines in personal travel in the U.S. during the 2000s.

Urban Revival by Millennials? Intra‐Urban Net Migration Patterns of Young Adults, 1980–2010

Abstract

Recommended publications

Housing Environment Important for Retirement

Religión y construcción simbólica de territorios identitarios urbanos en la ciudad de Guadalajara: E...

Declining Number of Slums: Nature of Urban Growth

SUBURBAN TRANSFORMATION AND NEW URBANISM: THE EVALUATION OF NEW SETTLEMENT AREAS IN ISTANBUL ACCORDI...