Remembrance of Cars and Buses Past: How Prior Life Experiences Influence Travel
Smart, Michael J., and Nicholas J. Klein. “Remembrance of Cars and Buses Past: How
Prior Life Experiences Influence Travel.” Journal of Planning Education and Research
38, no. 2 (June 1, 2018): 139–51. https://doi.org/10.1177/0739456X17695774.
Michael J. Smart Ph.D.,
Edward J. Bloustein School of Planning and Public Policy
Rutgers, the State University of New Jersey
33 Livingston Avenue, New Brunswick, New Jersey 08901
Nicholas J. Klein Ph.D.,
Visiting Assistant Professor
Urban Planning Program
Graduate School of Architecture, Planning and Preservation
Avery Hall, Columbia University, New York, NY 10027
Does growing up in a neighborhood with high-quality public transit influence travel behavior
later in life, even if you move to a neighborhood with worse transit service? To test this, we
construct residential histories of individuals using decades of data from the Panel Study of
Income Dynamics. We find that past experiences shape transportation futures. Exposure to
transit during young adulthood, in particular, associated with an auto-light lifestyle and greater
transit usage later in life. This research suggests a long-term benefit for encouraging transit at
younger ages to foster a “transit habit.”
DO PAST EXPERIENCES INFLUENCE TRAVEL?
Imagine a thirty-something born and bred New Yorker who packs up and moves across
the country to Los Angeles. As a stereotypical New Yorker, she used transit or walked most
places. But now that she lives in LA, does her experience of using public transit in New York
continue to influence her daily travel? Is she more likely to use transit than her Angeleno peers
who have lived their whole lives in auto-dependent neighborhoods? Or does she adapt
completely to the new reality, abandoning her transit habits in the context of a new built
environment? This article tests whether past experience living in areas with high-quality public
transit influences travel behavior later in life.
If today’s travel choices are influenced by our prior experiences, this would suggest a
long-term rationale for encouraging public transportation use. Rather than focusing solely on
relatively short-term cost-benefit analyses or economic development gains for new and existing
public transit service, planners, policy-makers, politicians and advocates might argue that by
providing these public services, we habituate residents to public transportation. Familiarity with
public transit can have longer-term payoffs that do not appear on a typical cost-benefit ledger,
and policies like those that provide fare-free public transit access for high school and college
students could have long term benefits (Brown, Hess, and Shoup 2001; Brown, Hess, and Shoup
Our work builds on research that highlights how social factors can mediate the
relationship between the built environment and travel behavior. While decisions about how,
when and where we travel are still largely explained by cost, time, and distance, other factors
also contribute to these decisions, including a positive utility of travel (travel for the sake of
travel) (Mokhtarian and Salomon 2001), the influence of social networks (Axhausen 2005; Smart
and Klein 2013; Tilahun and Levinson 2011; Blumenberg and Smart 2014) and perceptions of
travel modes (Klein 2016; Guiver 2007). Here, we argue that past experiences also shape today’s
decisions about travel. Our research improves on earlier studies (Weinberger and Goetzke 2010;
Macfarlane, Garrow, and Mokhtarian 2015; Chen, Chen, and Timmermans 2009) by using a
long-term national panel dataset with information on travel and the built environment and
evaluates the influence of exposure to public transit throughout one’s life.
We test whether living in a transit-rich area in the past leads to an auto-light lifestyle and
greater transit usage later in life, even when the new community is not particularly transit-
friendly. To do this, we use the Panel Study of Income Dynamics (PSID), which has surveyed
the same families and their descendants for nearly 50 years. We create residential histories for all
respondents and measure their transit “exposure” throughout their life. We then test whether
higher exposure during earlier years leads to more transit use and less car ownership later in life.
We find that exposure to transit early in life leads to greater likelihood of using public
transportation and decreases an individual’s level of auto ownership later in life. We note that
exposure to transit during one’s formative years (ages five to eighteen), when one has no say in
where one lives, is a strong predictor of later transit use and lower rates of auto ownership. We
further note that our models suggest that the strongest predictor of future transit use is the transit
environment in which one lives in one’s late 20s and early 30s, suggesting that policy
interventions targeted to these age groups may bear the most fruit.
In the next section, we describe the existing research on the relationship between travel
behavior and past experiences. Then we describe our data and research approach. Next, we
discuss the findings from our regression models. We conclude with a discussion of the model
results and implications for policy.
RESIDENTIAL RELOCATIONS AND TRAVEL BEHAVIOR
When households relocate, their daily travel patterns are influenced by the built
environment context of their new neighborhood, their economic and demographic circumstances,
their attitudes and preferences, and, as we test here, perhaps by their past experiences. Among
these factors, researchers have recently focused most of their attention on the role of the built
environment (e.g., Ewing and Cervero 2010). Residential relocations, along with changes in
employment and family composition, function as critical events in an individual or household’s
mobility biography that alter their circumstances leading to a behavioral shift in travel (Scheiner
2007; Müggenburg, Busch-Geertsema, and Lanzendorf 2015; Lanzendorf 2003; Beige and
Axhausen 2012). This work suggests that programs that aim to influence travel behavior should
focus on young people (under 35 or so) who have not yet settled down and after these critical
events in one’s lives (Beige and Axhausen 2012).
Scholars focus on movers as a mechanism to address concerns about residential self-
selection, namely that the statistical associations between travel and the built environment may
be manifestations of a sorting process by which people who have an affinity for transit or
walking chose to live in neighborhoods where they can realize those preferences. By controlling
for attitudes, preferences and socioeconomic attributes, researchers frame a residential move as a
“treatment” effect and compare movers to “control group” that did not relocate as a way to tease
out these self-selection effects (Cao and Ermagun 2016; Cao, Mokhtarian, and Handy 2007;
Handy, Cao, and Mokhtarian 2006; Krizek 2003; Scheiner and Holz-Rau 2012). Collectively,
this research finds that even when controlling for self-selection effects, movers do change their
travel behavior in response to the built environment, though there is less agreement on the
magnitude of these effects.
A smaller body of literature has examined how prior experiences both in the short and
long term can shape future travel behavior. In the short term, habits have a powerful influence on
daily travel behavior. Through repeated experiences, decisions about how, when and where to
travel become less calculating and more habitual, as long as the outcome is more or less
acceptable (Gärling and Axhausen 2003; Verplanken, Aarts, and Van Knippenberg 1997; Fujii
and Kitamura 2003). Some have criticized this formulation of travel habits as automatic and not
reflective (Schwanen, Banister, and Anable 2012) and have critiqued the narrow view of travel
habits as barriers to sustainable travel while ignoring the positive aspects of habits (e.g. freeing
individuals from the rational decision-making calculations and allowing them to be in the
moment) (Middleton 2011).
Prior experiences can also exert an influence over a longer time horizon, shaping car
ownership, commute time, and distance. Past experiences may influence later travel behavior if
these experiences shape attitudes and preferences for neighborhoods or travel modes (Chen,
Chen, and Timmermans 2009; Weinberger and Goetzke 2010; Macfarlane, Garrow, and
Mokhtarian 2015). For example, Weinberger and Goetzke (2010) use Census data to analyze
movers to and from cities with high-quality transit service to test whether households “learn
preferences” for a car-light lifestyle and then take those preferences with them when they move.
They find that when urbanites move to the suburbs, they have fewer vehicles than would be
expected and, conversely, when suburbanites and rural residents move to the central cities, they
own more vehicles than their urban peers. Chen et al. (2009) use data from the Puget Sound
Transportation Panel Survey to test the influence of the previous residential locations on
commuting distances of movers. Chen et al. find that exposure to lengthy commutes does indeed
“make them more tolerant of long commute distances” but that these “historical depositions” are
weakened by lifecycle factors, such as the birth of a child (p. 2773). Macfarlane et al. (2015)
construct residential histories for residents of the Atlanta region to test how past exposure to
neighborhood conditions (residential density and the share of workers commuting via non-auto
modes) influence auto ownership. While the authors find a significant effect of past built
environments, they conclude that “exposure to higher densities and non-vehicle transportation
options (either currently or in the past) has a relatively modest influence on vehicle ownership
decision” (Macfarlane, Garrow, and Mokhtarian 2015, 197). Lastly, Döring et al. (2014) find
multigeneration effects. Using a retrospective study of three generations, they find that parents’
and grandparents’ attitudes about travel and their residential location can influence the
subsequent generations’ travel behavior.
Past experiences can also influence how individuals make subsequent decisions about
their travel. For example, Simonsohn (2006) draws on the psychological concept of a “contrast
effect,” by which decisions are influenced by surrounding contexts, to test whether the average
commute duration in movers’ prior metropolitan area influences commute durations after the
move to a new region. Using data from the PSID, Simonsohn finds that “individuals coming
from cities with longer average commutes choose to commute significantly longer in their new
city than their peers coming from cities with shorter commutes” though this contrast effect
dissipates over time (p. 4).
Some have also speculated that past experiences may be one of the reasons why transit
use is higher among immigrants to the United States than among their US-born peers.
Immigrants from countries with lower car ownership rates may bring a “transit habit” with them
when they migrate (Blumenberg and Smart 2011). Additionally, many immigrants first settle in
transit-rich neighborhoods when they arrive in the US and they may develop a transit habit in
these contexts, which continue for many years (Chatman and Klein 2013).
Beyond the transportation literature, a host of social scientists have examined the role of
neighborhood effects on a range of outcomes including economic, academic, employment, and
health outcomes, to name a few. This literature has developed a rich conceptualization of the
mechanisms by which one’s residential location, at different stages in life, influences particular
outcomes (e.g., Ellen and Turner 1997; Galster 2012; Sharkey and Faber 2014). For example, the
role of family, schools, peers and institutions within a neighborhood may vary in importance
depending on an individual’s age and the outcome of interest (Ellen and Turner 1997). And
responses to a particular neighborhood effect may require a threshold, might not be universal,
could be mediated or buffered by other factors, or influenced by the frequency, duration and
consistency of exposure (Galster 2012).
Our study builds on and expands existing research about how exposure to transit can
influence travel behavior over many years. While others have examined residential histories and
exposure to transit, they have not isolated the effects of exposure at specific ages, as we do here.
We also extend this literature by testing the effects of living in an area with high-quality transit
on both auto ownership and transit use. Finally, because the PSID is a national data set that has
been following the same families for many years, we can study the effects of past experiences in
a broader range of built environments and over a longer time horizon than previous studies.
We use a restricted version of the Panel Study of Income Dynamics (PSID) to test
whether past experiences influence future transit use and auto ownership (“Panel Study of
Income Dynamics, Restricted Use Data” 2014). The PSID is a long-running panel survey that
has been following the same families since 1968, when it began with 5,000 families. Since then,
the PSID has grown to over 9,000 families (and 22,000 individuals) through births and the
addition of Latino and immigrant samples in the 1990s. While the survey focuses on earnings
and expenditures, the questionnaires have often asked families about auto ownership and
transportation expenses, including transit expenses. We use the responses to these questions for
We estimate models that test whether prior exposure to transit influences future spending
on transit and auto ownership. We measure transit spending using a dichotomous variable where
a value of one indicates that the family had some transit expenses during the past month, and a
zero value means the family reported spending no money on transit. We thus use a logistic
regression to model transit spending. We measure auto ownership as the ratio of cars to adults in
the family, and model this using ordinary least squares (OLS) regression. We use random-effects
models for both outcome measures and evaluate multiple measures of prior exposure to transit,
which we describe below. We chose a random-effects model, rather than a fixed-effects model,
because the fixed-effects would exclude time time-invariant variables, including most of our
measures of past exposure to transit. Using a fixed-effects model would also be problematic for
our model of transit use since this excludes 84 percent of our sample who either always or never
report transit expenditures in each survey wave.
We use data from the PSID dating back to 1968 to inform our measures of exposure to
transit but we limit our models to more recent waves. For our models of auto ownership, we use
the eight PSID waves from 1999 through 2013 since the PSID omitted questions about auto
ownership between 1986 and 1999. For the models of transit expenditures, we use six waves
from 2003 to 2013. While the PSID has included questions about transit expenses since 1999, the
question wording and responses changed considerably between the 2001 and 2003 surveys.
Our variable of interest is past exposure to transit. Since we do not have historical data on
transit service quality for the entire US, we use two proxies for transit service. First, we use US
Census journey-to-work data as a proxy for transit quality. For each census tract, we use the
share of workers commuting by transit from the 1970, 1980, 1990 and 2000 decennial census
and the 2008-2012 American Community Survey 5-Year Estimates. For each wave of the PSID,
we use the most contemporary decennial census (e.g. panel waves 1968 through 1974 we use the
1970 census and for the panel waves 1975 through 1984 we use the 1980 census).
Our second proxy for transit service is a composite database of transit accessibility. We
use data on transit accessibility from The Brookings Institute and The Accessibility Observatory
at the University of Minnesota, each of which provides partial coverage for the entire United
States (Owen and Levinson 2014; Tomer et al. 2011). From both datasets, we use the number of
jobs accessible by transit within 30 minutes (including access and egress on foot). In places
where data were available from both sources, the measures are very highly correlated, and we
used the average of both. We convert this transit accessibility measure to a z-score (standard
deviations from the regional mean, normalized separately for each Combined Statistical Area in
the United States) to help account for the large variation in labor market sizes across metro areas.
There are a few obvious limitations with this data. First, the coverage is incomplete for the US
(covering only about 60 percent of our PSID records). Second, the data are not historical, though
our transit access metrics are just as correlated with previous decades’ transit use as they are for
current transit use (r=0.52 for the 1970 and 1980 censuses, 0.54 for 1990, 0.55 for 2000 and 0.53
for the 2010 ACS), suggesting that while new transit services have opened in many locales, the
landscape of transit service throughout the nation has remained remarkably stable in recent
Like Macfarlane et al. (2015), we evaluated several different measures of past transit
exposure. First, we used the head of family’s average exposure to transit since 1990. Second, we
constructed an exponential decay function, to give more weight to recent experiences. Our third
measure is an average exposure during one’s 20s. And finally, we measure the head of family’s
exposure to transit during the formative years (five to 18), when one has no say in the location of
the family; we include this measure to help control for residential self-selection, in which those
with a strong preference self-select to live in neighborhoods where they can ride transit (for an
overview, see Cao, Mokhtarian, and Handy 2009). Finally, we estimate a series of models using
moving averages for transit exposure from birth to age 40 and present these results in a summary
While the PSID collects data on transit use (expenditures) aggregated for the entire
family, we are only able to estimate the head of family’s exposure to public transit. This
mismatch may lead to some error, though we expect the magnitude of the error to be small; we
assume that partnering and marriage behavior is unrelated to both partners’ prior exposure to
transit in any way that would bias our results.
To account for other factors that likely influence transit expenditure and auto ownership,
we include several control variables in our models. We include the total family income in the
previous year, residential population density of the census tract where the family lives (measured
in thousands of people per square mile), race and ethnicity of the head of the household, the
family’s poverty status, the student status of the household head and his/her spouse or partner
(where present) and number of children in the household. We also include the number of cars per
adults that the family owns in the model of transit expenses.
PAST EXPOSURE TO TRANSIT AND CURRENT TRAVEL BEHAVIOR
We analyze the determinants of transit use and car ownership, focusing on the influence
of prior exposure to transit. We use a large panel dataset, and families who move from one
transit environment to another are of particular interest. Table 1 shows descriptive statistics for
our full sample, families who live in low- and high-transit tracts, families who moved and non-
movers, and families who moved from low- to high- and from high- to low-transit environments.
We define low- and high-transit environments using our transit access data and set the cutoff
point at the regional mean (z-score of zero). Transit accessibility is considerably right-skewed
(since there are far more locations below the regional mean than above it) and there are over six
times as many observations in low-transit environments as in high-transit environments. We
separately examined differences among those in the bottom and top third of transit accessibility,
and the results were broadly similar.
Those who live in high-transit areas are quite different from those who live in lower-
transit environments, as we expect given the literature on residential location and transit. They
use transit much more (25 percent had used transit in the prior month compared with just six
percent of families in low-transit environments) and own fewer cars per adult in the family. They
also have lower average incomes, are more likely to be in poverty, and are considerably more
likely to be students, immigrants, and people of color.
Families living in high-transit areas are also more likely to have grown up and lived
recently in higher-transit areas than are those who live in lower-transit environments, though the
data suggest a nuanced story. While families in transit-rich environments live in tracts that are on
average 1.30 standard deviations above the regional mean for transit access (vs. -0.18 for
families in low-transit environments), these two groups were not as different earlier in their lives.
The gap between the two groups’ average exposure since 1990 is considerably smaller (0.66
vs. -0.11) and the gap between their access to transit as children is smaller yet (0.29 vs -0.07).
The differences between movers and non-movers tell a similar story; those who have
moved in the past two years (35 percent of records) are younger, are more likely to be students,
have lower incomes, and are more likely to be people of color. They also have fewer cars per
adult in the family and use transit more than non-movers do. Movers also live in areas with
somewhat better transit service, and have a slightly higher level of exposure in the recent past
and in their twenties. However, both movers and non-movers grew up in census tracts with
average transit service.
Finally, we examine the differences between two specific groups of movers: those who
move from a low-transit environment to a high-transit environment, and those who make the
opposite move. Here, we observe fewer differences, though some are remarkable. Notably, those
who move to from low- to high-transit areas are considerably more likely to live in poverty (22
percent do) than the full sample (column a, 11 percent), the sample of all movers (column e, 16
percent) and those who move from high- to low-transit areas (column f, 14 percent), echoing the
findings of Glaeser et al (2008). We also observe something notable about transit use and car
ownership: movers in both “directions” use transit more than their new neighbors (columns f vs.
b and g vs. c; 12 vs. 6 percent and 27 vs. 25 percent). We suspect this may in part be due to other
systematic differences between movers and non-movers, and we explore this in greater detail
TABLE 1. Descriptive Statistics by Transit Quality in Home Tract and Move Status, PSID,
Outcome Data 2003-2013, Exposure Data 1968-2011
When people with a history of living in transit-rich neighborhoods move to a low-transit
area, do they use transit more and own fewer cars than their neighbors? Our data suggest they do.
As table 2 shows, those who recently moved from a high-transit area own fewer cars (0.83 per
adult in the family vs. 0.98 per adult) and are more likely to have used transit in the prior month
(13 percent versus just eight) compared to those who have lived there for two or more years. On
average, these movers spend 40 percent more on transit per adult in the family than their
neighbors do. When we adjust these figures for the presence of children in the family, the results
(a) (b) (c) (d) (e) (f) (g)
Sig. (f) vs.
Used transit 8% 6% 25% *** 7% 11% *** 12% 27% ***
Ratio of cars to adults in family 0.96 1.00 0.67 *** 1.00 0.86 *** 0.83 0.69 ***
Transit access to jobs
Current z-score -0.02 -0.18 1.30 *** -0.06 0.07 *** -0.33 1.35 ***
Prior exposure (z-scores)
Average exposure -0.02 -0.11 0.66 *** -0.03 0.01 *** -0.03 0.18 ***
Decay -0.06 -0.28 0.97 *** -0.10 0.04 *** 0.12 -0.23 ***
Twenties 0.05 -0.04 0.84 *** 0.03 0.09 *** 0.00 0.34 ***
Formative Years -0.02 -0.07 0.29 *** -0.02 -0.01 -0.03 0.12 ***
Transit JTW share in tract
Current z-score -0.06 -0.22 1.19 *** -0.10 0.05 *** -0.05 1.09 ***
Prior exposure (z-scores)
Average exposure -0.04 -0.15 0.83 *** -0.07 0.03 *** 0.14 0.48 ***
Decay -0.06 -0.19 0.99 *** -0.10 0.04 *** 0.22 0.25
Twenties 0.06 -0.02 0.60 *** 0.03 0.12 *** 0.18 0.58 ***
Formative Years 0.02 -0.07 0.56 *** 0.00 0.04 ** 0.15 0.48 ***
Head and/or spouse is student 1.4% 1.3% 2.5% *** 0.8% 2.9% *** 3.1% 4.5%
Number of adults in family 1.7 1.7 1.5 *** 1.8 1.5 *** 1.5 1.4 ***
Children present in family 30% 31% 26% *** 29% 33% *** 33% 22% ***
Total family income (mean) $78,371 $79,412 $70,009 * $85,391 $61,445 *** $66,004 $69,337
Total family income (median) $49,000 $51,660 $35,000 *** $57,200 $37,800 *** $37,260 $32,918 ***
Family is below poverty line 11% 10% 17% *** 8% 16% *** 14% 22% ***
Immigrant family 8% 7% 16% *** 8% 8% 13% 11%
Age of family head 50.7 51.1 47.0 *** 54.9 40.4 *** 36.0 40.6 ***
Race/ethnicity of head
Non-Hispanic White 75% 78% 54% *** 78% 70% *** 61% 49% ***
Non-Hispanic Asian 2% 2% 3% * 2% 2% 3% 3%
Non-Hispanic Black 14% 12% 28% *** 12% 18% *** 23% 36%
Hispanic of any race 7% 6% 14% *** 7% 8% ** 11% 23%
Other 0.7% 0.8% 0.4% * 0.6% 1.0% * 1.0% 0.2% **
Residential density 4,569 3,089 16,440 *** 4,398 4,980 ** 5,221 12,539 ***
N (person-years) 64,559 56,297 8,262 41,724 22,817 4,351 893
Note: stars indicate statistical significance: * p<0.10; ** p<0.05; *** p<0.01
are the same (not shown here). We note that these are quite large differences, though the actual
magnitude of transit use here is quite small; five or seven dollars’ worth of transit spending
translates to just a couple of rides per month.
TABLE 2. Transit use and car ownership by presence in a low- or high-transit tract and
mover status, PSID, 2003-2013
Because movers and non-movers are systematically different in important ways (movers
are younger and earn less) we also show the statistics for two subsamples: those never in poverty
and young adults. Among families who live in low-transit tracts and never report incomes below
the poverty line, we find somewhat more muted differences for transit spending and auto
ownership. When we examine families headed by a person age 20-35, we also find similar
results; those who have moved from a higher-transit tract own fewer cars and use transit more.
And what of new residents of transit-friendly places? We find little difference between
Ratio of cars to adults 0.98 0.83 *** 0.70 0.69
Monthly spending, fares per adult $4.63 $6.50 *** $10.60 $11.01
Used transit last month 8% 13% *** 22% 27% **
N(person-years) 19,884 4,351 4,813 781
Ratio of cars to adults 1.06 0.95 *** 0.84 0.88
Monthly spending, fares per adult $4.62 $6.50 ** $8.95 $9.32
Used transit last month 7% 11% *** 22% 27% *
N(person-years) 9,169 1,258 1,290 178
Age 20-35 Sample
Ratio of cars to adults 1.002 0.855 *** 0.692 0.733
Monthly spending, fares per adult $4.74 $6.22 * $17.89 $12.70 *
Used transit last month 8% 11% *** 29% 31%
N(person-years) 3,511 2,773 1,031 411
Note: stars indicate statistical significance: * p<0.10; ** p<0.05; *** p<0.01
Lives in Low-Transit Tract
Lives in High-Transit Tract
those who have lived in high-transit tracts for two or more years and their new neighbors who
have moved from a transit-poor neighborhood. Surprisingly, a slightly higher percentage of
movers reported using transit in the past month (27 vs. 22 percent), and these findings hold up
for the sample of families never living in poverty. These differences are likely explained by the
fact that new residents of high-transit neighborhoods are more likely to be young and to be
students. When we examine only young residents of transit-rich neighborhoods and young
movers, we find no difference in car ownership rates or the likelihood of using transit. The data
suggest that those young people who have moved into these transit-rich neighborhoods from
low-transit places spend less on transit than longer-term young residents do. However, the
magnitude of transit use here is moderate, at two or three transit trips per week, on average.
To explore these differences further, we estimate a series of panel regression models of
transit use and auto ownership. In the following subsections, we present the results.
Prior Exposure Shapes Transit Use
Our analysis suggests that prior exposure to public transportation can influence later
decisions to use transit and to own one or more automobiles. Table 3 shows the results for
current use of public transportation, with prior exposure to public transportation measured four
ways each using two datasets, for a total of eight models.
Our transit measures include the current transit environment in which the family lives,
measured as region-specific z-scores (standard deviations from the mean access to jobs by
transit) and a measure of past exposure to transit. In all models, both the current transit
environment and prior exposure to transit matter. In some of our models, current transit exposure
is a stronger predictor of transit use than is past exposure to transit; in other models, the opposite
is true. In particular, exposure to transit in one’s twenties appears to be a strong predictor of later
transit use, with a standard deviation increase in transit exposure predicting a roughly 30 to 60
percent increase (Model 5: , model 6: ) in the odds of using transit.
While this is a large increase in odds, only 10 percent of families used transit at least once in the
previous month; thus, the model suggests that an across-the-board standard deviation increase in
transit exposure would only increase the number of families using transit in a given month to
about 12.8 to 16.2 percent of families.
Our control variables largely function as expected. Family composition influences the
decision to use transit in a number of ways: families in which the head or spouse/partner (where
present) is a student are far more likely to use transit than are other households, and larger
households (more adults or children) are associated with lower odds of using transit across all
eight models. Similarly, families in which the head or spouse/partner were born outside the US
are more likely to use transit than are US-born families. Controlling for other variables in the
model, non-Hispanic Asian and black families are more likely to use transit than are Non-
Hispanic white families. The model suggests that poor families are far more likely to use transit
than are non-poor families, but that for families not in poverty, the probability of using transit
increases with income, in line with previous research (Pucher and Renne 2003). Geography
matters for transit use, too. All else equal, families that live in denser areas are more likely to use
transit, with a roughly seven percent increase in the odds of using transit for each additional
thousand persons per square mile in the home census tract.
We include year-specific intercepts to control for other factors that vary over time but
which cannot be included in the model. We find moderate year-specific effects, with a peak in
2009 and low points in 2005 and 2013.
Overall, we observe consistent findings across eight separate measurements of “transit
exposure.” The strong effect of transit exposure during one’s formative years (ages five to 18)—
when one has little or no say in one’s residential location—suggest that the results are not simply
an artifact of the self-selection of individuals with a preference for riding transit into those
neighborhoods where they can, in fact, ride transit.
Prior Exposure Shapes Auto Ownership
Our models of auto ownership tell a similar story. As Table 4 shows, current transit
access to jobs as well as prior exposure to public transportation have a strong and statistically
significant effect on the level of car ownership in a family. In general, a standard deviation
increase in today’s transit access is associated with a one to four percent decrease in the ratio of
cars to adults in the family. A standard deviation increase in one’s prior transit exposure is
associated with a three to eight percent decrease in auto ownership depending on the measure of
exposure. We note, however, that in four of the eight models the coefficients for current and
prior transit exposure are not statistically different from one another; we thus suggest that the
effects of current and prior exposure to transit are “roughly equal” in our models. Again, we note
the strong consistency in model results across all eight models, and particularly in the “formative
years” (age 5 to 18) measure of transit exposure, which likely controls for residential self-
In these models, our control variables also perform as expected. The presence of children
is associated with an increase in the ratio of cars to adults in the household, while income
increases auto ownership. Living below the poverty line has a strong negative association with
car ownership, with a roughly ten or eleven percent decrease in cars per adults. Controlling for
other variables in the model, immigrant families own fewer cars, and all else equal, non-Hispanic
blacks have lower rates of car ownership in all eight models. Greater residential density is
similarly associated with lower levels of car ownership, with an increase of 1,000 persons per
square mile associated with a half percent decrease in cars per adult in the family. The temporal
trends are somewhat less clear, with slight peaks in the early 2000s and in 2007-9, though the
magnitude of the differences is small.
TABLE 3 Random-Effects Panel Logistic Regression Model of Transit Use in Family, United States, Outcome Data 2003-2013,
Exposure Data 1968-2011
Transit exposure metric:
Transit data source: Coeff. Sig. Coeff. Sig. Coeff. Sig. Coeff. Sig. Coeff. Sig. Coeff. Sig. Coeff. Sig. Coeff. Sig.
Current transit (z-score, measured two ways) 0.223 *** 0.237 *** 0.199 *** 0.275 *** 0.313 *** 0.266 *** 0.233 *** 0.333 ***
Transit exposure (z-score, measured four ways) 0.250 *** 0.462 *** 0.194 *** 0.221 *** 0.248 *** 0.480 *** 0.174 *** 0.260 ***
Head and/or spouse is student 0.865 *** 0.660 *** 0.948 *** 0.659 *** 1.130 *** 0.931 *** 0.923 *** 0.651 ***
Number of adults in family -0.129 *** -0.154 *** -0.107 *** -0.151 *** -0.117 *** -0.112 *** -0.103 *** -0.093 ***
Children present in family -0.262 *** -0.335 *** -0.267 *** -0.315 *** -0.140 *** -0.134 *** -0.151 *** -0.274 ***
Total family income (log-transformed) 0.136 *** 0.137 *** 0.146 *** 0.135 *** 0.140 *** 0.134 *** 0.169 *** 0.157 ***
Family is below poverty line 0.408 *** 0.323 *** 0.425 *** 0.346 *** 0.519 *** 0.540 *** 0.719 *** 0.557 ***
Immigrant family 0.503 *** 0.612 *** 0.478 *** 0.687 *** -0.417 *** -0.174 *** -0.128 ** -0.002
Age of family head 0.072 *** 0.071 *** 0.068 *** 0.071 *** 0.120 *** 0.121 *** -0.032 *** -0.006
Age of family head, squared -0.0010 *** -0.0010 *** -0.0010 *** -0.0010 *** -0.0014 *** -0.0013 *** 0.0003 *** 0.0000
Race/ethnicity of head (omitted: non-Hispanic white)
Non-Hispanic Asian 0.355 *** 0.620 *** 0.394 *** 0.667 *** 0.850 *** 1.010 *** 0.426 *** 1.100 ***
Non-Hispanic Black 0.644 *** 0.445 *** 0.668 *** 0.692 *** 0.740 *** 0.398 *** 0.367 *** 0.298 ***
Hispanic of any race -0.669 *** -0.570 *** -0.596 *** -0.511 *** 0.333 *** 0.095 * -0.203 *** -0.174 ***
Other -0.395 *** -0.754 *** -0.163 -0.752 *** -1.290 *** -1.670 *** -0.844 *** -0.868 ***
Residential density in thousands 0.063 *** 0.071 *** 0.062 *** 0.073 *** 0.075 *** 0.088 *** 0.076 *** 0.084 ***
Ratio of cars to adults in family -2.230 *** -2.160 *** -2.250 *** -2.180 *** -2.080 *** -1.960 *** -2.470 *** -2.380 ***
Year (omitted: 2003)
2005 0.088 *** 0.063 *** 0.103 *** 0.037 ** 0.049 * -0.027 -0.126 *** -0.085 ***
2007 0.079 *** 0.061 *** 0.061 *** 0.064 *** 0.133 *** 0.073 *** -0.178 *** -0.219 ***
2009 0.237 *** 0.256 *** 0.251 *** 0.260 *** 0.140 *** 0.194 *** -0.075 *** -0.078 ***
2011 0.135 *** 0.180 *** 0.161 *** 0.182 *** -0.042 -0.024 -0.133 *** -0.084 ***
2013 0.041 ** 0.089 *** 0.029 0.078 *** -0.021 0.010 -0.250 *** -0.194 ***
Intercept -4.550 *** -5.150 *** -4.580 *** -5.190 *** -6.270 *** -7.040 *** -2.460 *** -3.530 ***
N 31,510 52,163 29,870 50,587 16,887 27,061 15,823 23,149
Pseudo R-squared 0.140 0.137 0.142 0.138 0.140 0.136 0.119 0.117
Rho (proportion of variance explained by panel-level variance) 0.61 0.61 0.61 0.61 0.61 0.61 0.56 0.57
Note: stars indicate statistical significance: * p<0.10; ** p<0.05; *** p<0.01
Formative Years (5-18)
TABLE 4 Random-Effects Panel OLS Model of Ratio of Cars to Adults in Family, United States, Outcome Data 1999-2013,
Exposure Data 1968-2011
Transit exposure metric:
Transit data source: Coeff. Sig. Coeff. Sig. Coeff. Sig. Coeff. Sig. Coeff. Sig. Coeff. Sig. Coeff. Sig. Coeff. Sig.
Current transit (z-score, measured two ways) -0.024 *** -0.013 *** -0.031 *** -0.022 *** -0.035 *** -0.029 *** -0.036 *** -0.039 ***
Transit exposure (z-score, measured four ways) -0.066 *** -0.075 *** -0.030 *** -0.029 *** -0.064 *** -0.064 *** -0.047 *** -0.045 ***
Head and/or spouse is student 0.030 * 0.015 0.035 * 0.014 0.055 * -0.064 * 0.027 0.013
Number of adults in family -0.137 *** -0.150 *** -0.138 *** -0.151 *** -0.157 *** -0.064 *** -0.139 *** -0.153 ***
Children present in family 0.023 *** 0.037 *** 0.026 *** 0.038 *** 0.035 *** -0.064 *** 0.025 *** 0.034 ***
Total family income (log-transformed) 0.043 *** 0.044 *** 0.043 *** 0.044 *** 0.047 *** -0.064 *** 0.044 *** 0.053 ***
Family is below poverty line -0.122 *** -0.124 *** -0.119 *** -0.126 *** -0.129 *** -0.064 *** -0.144 *** -0.134 ***
Immigrant family -0.096 *** -0.087 *** -0.095 *** -0.096 *** -0.025 -0.064 -0.032 -0.006
Age of family head 0.031 *** 0.034 *** 0.031 *** 0.035 *** 0.024 *** -0.064 *** 0.023 *** 0.022 ***
Age of family head, squared -0.0003 *** -0.0004 *** -0.0003 *** -0.0004 *** -0.0002 *** -0.064 *** -0.0002 *** -0.0002 ***
Race/ethnicity of head (omitted: non-Hispanic white)
Non-Hispanic Asian 0.012 -0.034 0.001 -0.035 0.097 * 0.012 0.026 -0.035
Non-Hispanic Black -0.208 *** -0.177 *** -0.229 *** -0.219 *** -0.233 *** -0.179 *** -0.234 *** -0.174 ***
Hispanic of any race -0.036 ** -0.046 *** -0.046 *** -0.053 *** -0.023 -0.032 -0.037 -0.057 **
Other -0.033 -0.052 * -0.060 -0.075 ** -0.024 -0.057 -0.091 -0.113 **
Residential density in thousands -0.005 *** -0.006 *** -0.005 *** -0.006 *** -0.005 *** -0.006 *** -0.005 *** -0.006 ***
Year (omitted: 1999)
2001 0.022 *** 0.014 ** 0.019 ** 0.012 * 0.014 0.001 0.017 -0.005
2003 0.039 *** 0.035 *** 0.037 *** 0.033 *** 0.024 ** 0.024 ** 0.029 ** 0.025 **
2005 0.010 0.000 0.007 -0.002 -0.002 -0.012 0.010 0.007
2007 0.028 *** 0.027 *** 0.027 *** 0.026 *** 0.002 0.014 0.001 0.013
2009 0.038 *** 0.030 *** 0.039 *** 0.030 *** 0.022 * 0.013 0.025 * 0.018
2011 0.016 ** 0.007 0.016 * 0.008 -0.010 -0.017 0.002 -0.008
2013 0.012 0.000 0.009 -0.001 -0.013 -0.032 *** -0.012 -0.030 **
Intercept 0.067 * 0.041 0.077 ** 0.047 0.188 ** 0.167 ** 0.224 *** 0.190 ***
N 40,145 66,670 37,983 64,559 21,169 33,795 19,428 28,270
R-squared (within) 0.049 0.046 0.048 0.046 0.040 0.044 0.043 0.034
R-squared (between) 0.265 0.263 0.272 0.259 0.270 0.233 0.227 0.270
R-squared (overall) 0.265 0.167 0.196 0.166 0.151 0.161 0.132 0.185
Rho (proportion of variance explained by panel-level variance) 0.44 0.4 0.45 0.41 0.470 0.42 0.4 0.34
Note: stars indicate statistical significance: * p<0.10; ** p<0.05; *** p<0.01
Formative Years (5-18)
WHEN IS PAST EXPOSURE TO TRANSIT MOST INFLUENTIAL?
Given that our analysis suggests that previous exposure to high-quality public transit can
influence travel behavior later in life, we extended our analysis to examine how exposure at
different ages can influence later outcomes. Here, we use one exposure method, a moving
average of prior exposure to transit, and model only one outcome, transit spending. We also limit
our analysis to the Census-based journey-to-work measure of transit quality.
We estimated a series of models to test the effect of past transit exposure during various
periods in one’s life, from birth through 40 years of age. Our models are identical to those above,
but we substitute our new exposure measures. Figure 1 shows a conceptual diagram of these
models. We use three different moving average windows: three, five and seven years, and we
include these separately in distinct models (e.g. for the three-year average, we would use
exposure at ages 10, 11, and 12 in one model of current transit use, exposure at ages 11, 12 and
13 in another, and so forth). Our exposure measure is the average of the share of workers in the
person’s home census tract who used public transit to commute during the age for each moving
average window (e.g. for the five-year moving average, N through N+5 years old). We only
include observations in our model if the person is older than the moving average window
(otherwise these transit exposures would not be in the past); additionally, we estimated models
that restricted how recent the exposure could be (at least one year, at least three years, and at
least five years in the past), though, for most observations, the exposure metric lay much further
in the past. Using the three different moving average windows, the three time lags and the 40
different periods included in each moving average window, we ran 360 logistic regressions.
FIGURE 1 Conceptual Diagram of Moving-Window Models, Outcome Data 2003-2013,
Exposure Data 1968-2011
Figure 2 charts the coefficients for the past transit exposure and current transit
environment for the three moving average windows and the three time lags across the 40 moving
average periods. This analysis suggests that exposure to transit is particularly consequential
during one’s 20s and 30s. The points on the graph indicate the coefficients from all models,
while the lines plot the average of coefficients across our various models. The x-axis of this
graph indicates the midpoint for the moving average for the past exposure measure, not the age
of the person during the survey year. For example, the points at age 10 show the effect of
exposure to transit at age ten (in dark grey) along with the effect of the current transit
environment for all survey respondents (all of whom are older than ten). In this model, exposure
to transit at age 10 exerts a meaningful influence on current transit use, though the current transit
environment has a larger effect.
As the figure suggests, a person’s exposure to public transit during the period from their
early 20s to the late 30s has the strongest effect on their later travel behavior, even larger than the
effect of their current transit environment. For someone who is say 40, 50 or 60 years old, this
suggests that the quality of public transportation where they lived when they were 30 years old
has twice the effect on their current travel than the transit in their surrounding neighborhood. We
separately estimated a series of models using our alternative measure of transit access (jobs
accessible in 30 minutes on transit), and the results were similar, though the effect was muted, as
it was in our models presented above (Table 3).
FIGURE 2 Effect of Past Transit Exposure and Current Transit Environment on Transit
Use as a Function of Age used for Exposure Metric, Outcome Data 2003-2013, Exposure
Why might exposure to transit in one’s late 20s and 30s be of particular importance? We
hypothesize that several factors may be at play during this period. This is typically the time of
life when people begin to “settle down” by establishing long-term relationships, having children,
and, for some, settling into a longer-term job. Beige and Axhausen (2012) note that residential
moves, job changes and changes in transportation are much more common between ages 15 and
35 than after. This may enhance habit formation, or it may simply mean that there are fewer
“shocks” in the future that may cause a re-evaluation of location choices and travel patterns.
The figure appears to show that the importance of exposure to transit in one’s 40s and
one’s current transit environment are about equal. However, this convergence is likely an artifact
of our method; for earlier-exposure models (in one’s 20s, for instance), many of the people
included in the model are much older than 20 (because being 20 is, roughly, only one-quarter of
the way “through” one’s life); for our models using exposure during one’s 40s, a larger share of
people included in the model will be in their 40s or early 50s, and are much more likely to
continue to live in the same transit environment. Thus, the convergence we observe to the right
of our graph may be at least in part due to the shrinking “distance” between prior exposure and
one’s current environment.
In response to reviewer suggestions, we also estimated several alternative model
specifications, which we describe here. In each case, the results confirmed our preferred models
presented above, though there are several intriguing differences.
Because early exposure to transit may result in later residential self-selection into
neighborhoods with high-quality transit service, we estimated three models to explore this. First,
we estimated our models as above, but excluded the variables describing one’s current built
environment (transit access to jobs and residential density). In each these models, the effect size
of our variables describing prior exposure to transit roughly doubled compared to our preferred
models presented above. We take this to mean that when it comes to the effect of prior exposure
to transit on current transportation choices, some of the effect may be due to residential self-
selection, but experiences in earlier contexts can also shape transportation preferences and habits
independent of self-selection.
We also estimated models using a subset of respondents who live in radically different
environments from those where they grew up: city kids who live their adult lives in exurban and
rural communities, and rural and exurban children who moved to dense urban environments. In
these models, we find a strong effect of prior exposure on auto ownership. Rural kids who move
to the city consistently own more cars than their neighbors who grew up in the city and the
opposite is true for city kids who move to rural or exurban places.
Because a small number of urban areas account for nearly all transit ridership in the US,
we estimated two models with dummy variables for living in, or growing up in these metro areas.
We find that living in or growing up in New York, Chicago, San Francisco, Washington DC,
Boston, or Philadelphia provides an additional “boost” to transit use, though the results for the
rest of the nation remain essentially unchanged from the preferred models presented above.
Finally, we evaluated a panel logit model of car ownership which would be directly
comparable with our model of transit usage. In this version of the model, our outcome variable
was one if the family had “sufficient” automobiles (at least one per adult) and zero otherwise.
The model suggests a halving in the likelihood of having sufficient autos for each adult when
exposure to transit in the past increases by a standard deviation.
Exposure to high-quality public transportation during one’s life can encourage later
transit use and lower auto ownership, even if one lives in a less transit-friendly environment. Our
analysis suggests a strong and robust linkage, and this linkage holds even when we examine
exposure to transit at a young age, when one has no say in where one lives.
In this article, our primary aim was to examine the effects of where a person has lived on
their current travel behavior. While many researchers often seek to disentangle the “built
environment effect” for the broader population from the “self-selection effect” for those who end
up living in those places for reasons of preference, we are interested in ascertaining whether
current travel is related to prior experiences, regardless of whether this is the result of self-
selection or simply being more open to using transit or some other reason.
Our findings suggest something for policy. Transit agencies and advocates could “plant
the seed” for future ridership—in addition to providing an important social service—by
providing free or reduced-price transit passes for school or university students and targeted
programs for recent movers or new employees to encourage a transit habit. These types of future
payoffs may be difficult to quantify and incorporate in traditional cost-benefit analysis, though
our research suggests the payoffs may be substantial. Additionally, the growth in urban
populations over the past decade, particularly the increase young people living in cities (Myers
2016), could lay a foundation for transit use in the years to come. Our analysis suggests that
where someone lives during their 20s and 30s is particularly consequential for future travel
behavior. Even if many of the younger cohort do move out to the suburbs, our work suggests that
they will take some of their habits with them.
Although we find a relationship between past exposure to transit and future transit use
and car ownership, much about this relationship remains a mystery. First, we do not know how
this linkage forms. Perhaps, as some have suggested, preferences are developed through
exposure to transit-rich areas (Weinberger and Goetzke 2010; Macfarlane, Garrow, and
Mokhtarian 2015). Thus, the attitudes and preferences inherent in the self-selection effect
observed in many studies of transportation and land use may be shaped by prior experiences and
are likely mutable. Additionally, living in neighborhoods where many people use transit daily
could normalize transit ridership. Psychologists have long argued that individual behavior is
influenced to some degree from social learning and observations of other’s actions (e.g. Bandura
1977). Observing neighbors using transit could act as in the same way, providing a social model
that subtly encourages transit use. Alternatively, the processes could be more utilitarian rather
than social. Using transit at younger ages could be an individual learning process that individuals
carry with them later in life.
We do not know what dosage of transit is required to influence travel later in life. The
emerging consensus is that there is some effect, but we find a larger effect than previous studies
that used different data and measured different outcomes (Weinberger and Goetzke 2010;
Macfarlane, Garrow, and Mokhtarian 2015). Following Galster (2012), we suggest that there are
still many outstanding questions about the mechanisms for this “neighborhood effect.” We do
not know what levels of transit service are necessary early in life to lead to a lifelong habit of
transit use and decreases in car ownership. We also do not know if these effects are universal and
how the dosage of transit interacts with other factors. Chen et al. (2009) suggest at least one life-
cycle factor, parenthood, that moderates the effect of previous residential accessibility on
commute distance. Future research could shed light on the mechanisms and nuances of this
relationship to help guide policy.
We add to the growing body of knowledge that suggests that—while microeconomic
rationality likely drives the bulk of travel decisions—social factors work at the margins to shape
these decisions. Our work suggests that experiencing high-quality transit earlier in life can lead
to a decrease of a couple of percentage points in car ownership rates and a meaningful increase
in the likelihood (moving, roughly, from “very unlikely” to simply “unlikely” in the U.S.
context) of using transit once or more a month.
We obtained access to the restricted version of the Panel Study of Income Dynamics from
the Institute for Social Research at the University of Michigan, Ann Arbor. The collection of
data used in this study was partly supported by the National Institutes of Health under grant
number R01 HD069609 and the National Science Foundation under award number 1157698.
The Brookings Institution and the Accessibility Observatory at the University of Minnesota
provided data on job accessibility by public transit.
Axhausen, Kay W. 2005. “Social Networks and Travel: Some Hypotheses.” In Social
Dimensions of Sustainable Transport: Transatlantic Perspectives, edited by Kieran
Donaghy, Stefan Poppelreuter, and Georg Rudinger, 90–108. Aldershot, Hants, England ;
Burlington, VT: Ashgate.
Bandura, Albert. 1977. Social Learning Theory. Englewood Cliffs, N.J.: Prentice Hall.
Beige, Sigrun, and Kay W. Axhausen. 2012. “Interdependencies between Turning Points in Life
and Long-Term Mobility Decisions.” Transportation 39 (4): 857–72.
Blumenberg, Evelyn, and Michael Smart. 2011. “Migrating to Driving: Exploring the Multiple
Dimensions of Immigrants’ Automobile Use.” In Auto Motives: Understanding Car Use
Behaviours, edited by Karen Lucas, Evelyn Blumenberg, and Rachel Weinberger, 225–
51. United Kingdom: Emerald Group Publishing Limited.
———. 2014. “Brother Can You Spare a Ride? Carpooling in Immigrant Neighbourhoods.”
Urban Studies 51 (9): 1871–90.
Brown, Jeffrey, Daniel Baldwin Hess, and Donald Shoup. 2001. “Unlimited Access.”
Transportation 28 (3): 233–67. doi:10.1023/A:1010307801490.
———. 2003. “Fare-Free Public Transit at Universities An Evaluation.” Journal of Planning
Education and Research 23 (1): 69–82. doi:10.1177/0739456X03255430.
Cao, Xinyu (Jason), and Alireza Ermagun. 2016. “Influences of LRT on Travel Behaviour: A
Retrospective Study on Movers in Minneapolis.” Urban Studies, June,
Cao, Xinyu (Jason), Patricia L. Mokhtarian, and Susan L. Handy. 2007. “Do Changes in
Neighborhood Characteristics Lead to Changes in Travel Behavior? A Structural
Equations Modeling Approach.” Transportation 34 (5): 535–56. doi:10.1007/s11116-
———. 2009. “Examining the Impacts of Residential Self‐Selection on Travel Behaviour: A
Focus on Empirical Findings.” Transport Reviews 29 (3): 359–95.
Chatman, Daniel G., and Nicholas J. Klein. 2013. “Why Do Immigrants Drive Less?
Confirmations, Complications, and New Hypotheses from a Qualitative Study in New
Jersey, USA.” Transport Policy 30 (November): 336–44.
Chen, Cynthia, Jason Chen, and Harry Timmermans. 2009. “Historical Deposition Influence in
Residential Location Decisions: A Distance-Based GEV Model for Spatial Correlation.”
Environment and Planning A 41 (11): 2760–77. doi:10.1068/a41323.
Döring, Lisa, Janna Albrecht, Joachim Scheiner, and Christian Holz-Rau. “Mobility Biographies
in Three Generations – Socialization Effects on Commute Mode Choice.” Transportation
Research Procedia 1, no. 1 (2014): 165–76. doi:10.1016/j.trpro.2014.07.017.
Ellen, Ingrid Gould, and Margery Austin Turner. 1997. “Does Neighborhood Matter? Assessing
Recent Evidence.” Housing Policy Debate 8 (4): 833–66.
Ewing, Reid, and Robert Cervero. 2010. “Travel and the Built Environment.” Journal of the
American Planning Association 76 (3): 265–94. doi:10.1080/01944361003766766.
Fujii, Satoshi, and Ryuichi Kitamura. 2003. “What Does a One-Month Free Bus Ticket Do to
Habitual Drivers? An Experimental Analysis of Habit and Attitude Change.”
Transportation 30 (1): 81–95. doi:10.1023/A:1021234607980.
Galster, George C. 2012. “The Mechanism (s) of Neighbourhood Effects: Theory, Evidence, and
Policy Implications.” In Neighbourhood Effects Research: New Perspectives, 23–56.
Gärling, Tommy, and Kay W. Axhausen. 2003. “Introduction: Habitual Travel Choice.”
Transportation 30 (1): 1–11. doi:10.1023/A:1021230223001.
Glaeser, Edward L., Matthew E. Kahn, and Jordan Rappaport. 2008. “Why Do the Poor Live in
Cities? The Role of Public Transportation.” Journal of Urban Economics 63 (1): 1–24.
Guiver, J. W. 2007. “Modal Talk: Discourse Analysis of How People Talk about Bus and Car
Travel.” Transportation Research Part A: Policy and Practice 41 (3): 233–48.
Handy, Susan, Xinyu (Jason) Cao, and Patricia L. Mokhtarian. 2006. “Self-Selection in the
Relationship between the Built Environment and Walking: Empirical Evidence from
Northern California.” Journal of the American Planning Association 72 (1): 55–74.
Klein, Nicholas. 2016. “More than Just a Bus Ride: The Role of Perceptions in Travel
Behaviour.” Urban Studies, May. doi:10.1177/0042098016649324.
Krizek, Kevin J. 2003. “Residential Relocation and Changes in Urban Travel: Does
Neighborhood-Scale Urban Form Matter?” Journal of the American Planning
Association 69 (3): 265–81. doi:10.1080/01944360308978019.
Lanzendorf, Martin. 2003. “Mobility Biographies. A New Perspective for Understanding Travel
Behaviour.” In 10th International Conference on Travel Behaviour Research (IATBR),
Lucerne. Vol. 1015.
Macfarlane, Gregory S., Laurie A. Garrow, and Patricia L. Mokhtarian. 2015. “The Influences of
Past and Present Residential Locations on Vehicle Ownership Decisions.” Transportation
Research Part A: Policy and Practice 74 (April): 186–200.
Middleton, Jennie. 2011. “‘I’m on Autopilot, I Just Follow the Route’: Exploring the Habits,
Routines, and Decision-Making Practices of Everyday Urban Mobilities.” Environment
and Planning A 43 (12): 2857–77. doi:10.1068/a43600.
Mokhtarian, Patricia L., and Ilan Salomon. 2001. “How Derived Is the Demand for Travel?
Some Conceptual and Measurement Considerations.” Transportation Research Part A:
Policy and Practice 35 (8): 695–719. doi:10.1016/s0965-8564(00)00013-6.
Müggenburg, Hannah, Annika Busch-Geertsema, and Martin Lanzendorf. 2015. “Mobility
Biographies: A Review of Achievements and Challenges of the Mobility Biographies
Approach and a Framework for Further Research.” Journal of Transport Geography 46
(June): 151–63. doi:10.1016/j.jtrangeo.2015.06.004.
Myers, Dowell. 2016. “Peak Millennials: Three Reinforcing Cycles That Amplify the Rise and
Fall of Urban Concentration by Millennials.” Housing Policy Debate 0 (0): 1–20.
Owen, Andrew, and David M. Levinson. 2014. “Access Across America: Transit 2014 Data
[dataset].” Retrieved from the Data Repository for the University of Minnesota.
December 5. http://dx.doi.org/10.13020/D6MW2Q.
“Panel Study of Income Dynamics, Restricted Use Data.” 2014. Produced and distributed by the
Survey Research Center, Institute for Social Research, University of Michigan.
Pucher, J., and J.L. Renne. 2003. “Socioeconomics of Urban Travel: Evidence from the 2001
NHTS.” Transportation Quarterly 57 (3): 49–77.
Scheiner, Joachim. 2007. “Mobility Biographies: Elements of a Biographical Theory of Travel
Demand (Mobilitätsbiographien: Bausteine Zu Einer Biographischen Theorie Der
Verkehrsnachfrage).” Erdkunde 61 (2): 161–73.
Scheiner, Joachim, and Christian Holz-Rau. 2012. “Changes in Travel Mode Use after
Residential Relocation: A Contribution to Mobility Biographies.” Transportation 40 (2):
Schwanen, Tim, David Banister, and Jillian Anable. 2012. “Rethinking Habits and Their Role in
Behaviour Change: The Case of Low-Carbon Mobility.” Journal of Transport
Geography, Special Section on Theoretical Perspectives on Climate Change Mitigation in
Transport, 24 (September): 522–32. doi:10.1016/j.jtrangeo.2012.06.003.
Sharkey, Patrick, and Jacob W. Faber. 2014. “Where, When, Why, and For Whom Do
Residential Contexts Matter? Moving Away from the Dichotomous Understanding of
Neighborhood Effects.” Annual Review of Sociology 40 (1): 559–79.
Simonsohn, Uri. 2006. “New Yorkers Commute More Everywhere: Contrast Effects in the
Field.” The Review of Economics and Statistics 88 (1): 1–9. doi:10.2307/40042954.
Smart, Michael J., and Nicholas J. Klein. 2013. “Neighborhoods of Affinity: Social Forces and
Travel in Gay and Lesbian Neighborhoods.” Journal of the American Planning
Association 79 (2): 110–24. doi:10.1080/01944363.2013.883227.
Tilahun, Nebiyou, and David Levinson. 2011. “Work and Home Location: Possible Role of
Social Networks.” Transportation Research Part A: Policy and Practice 45 (4): 323–31.
Tomer, Adie, Elizabeth Kneebone, Robert Puentes, and Alan Berube. 2011. “Missed
Opportunity: Transit and Jobs in Metro America.” Washington, DC: The Brookings
Verplanken, Bas, Henk Aarts, and Ad Van Knippenberg. 1997. “Habit, Information Acquisition,
and the Process of Making Travel Mode Choices.” European Journal of Social
Psychology 27 (5): 539–60.
Weinberger, Rachel, and Frank Goetzke. 2010. “Unpacking Preference: How Previous
Experience Affects Auto Ownership in the United States.” Urban Studies 47 (10): 2111–