deltie.net archive

Extracted abstract

There is increasing evidence that social networks matter not only for long-distance moves but also for short-distance residential mobility. And the emerging structural sorting perspective is integrating networks into understandings of segregation processes. We add to this literature by considering how former school peers influence residential choices. We use Swedish register data describing the residential histories of cohorts of students who attended the same primary or secondary schools in Sweden. We trace their residential choices in young adulthood and estimate the effect of distance to peers on these choices. To account for selection, we use the spatial configuration of older cohorts who attended the same schools to adjust for peer similarity on unobserved preferences and attitudes. Using conditional logistic regression models of residential destinations, we find that individuals are more likely to choose a neighbourhood close to former school peers. Drawing on a linked lives perspective, we also consider how the peer effects change over the early adult life-course. The models imply that other networks can displace the social influence of primary and secondary school peers. While our analysis does not consider segregation as an outcome, our results suggest that schools may play a role in reproducing patterns of segregation within and between generations.

Introduction

Social influence-the effect people have on each other's behaviour-recurs across many areas of social life, guiding consumer, labour market, and health decisions (Granovetter, 1995; Grinblatt, Keloharju and Ikäheimo, 2008; Schaefer, Adams and Haas, 2013; Bernardi and Klärner, 2014; Arvidsson, Collet and Hedström, 2021) . Interest in social influence has been driven largely by its far-reaching implications for patterns of diffusion and inequality (Granovetter, 1978; Salganik, Dodds and Watts, 2006; Centola and Macy, 2007; Young, 2009) . The joint processes of residential mobility and segregation furnish a canonical example of how social influence drives inequality. In Schelling's (1971) segregation model, people belonging to two distinct groups affect each other's residential choices. Influence that is skewed in favour of in-group members can set off mobility cascades that lead to city-wide segregation, even if in-group preferences are relatively weak.

In Schelling's model, social influence is exerted between neighbors with superficial relationships. But many social relationships-within families, workplaces, schools, etc.are more intimate, durable, and institutionalized (Fischer, 1982; Wellman and Wortley, 1989) . There is evidence that these social ties, particularly kin relationships, affect internal migration and residential mobility (Massey and Espinosa, 1997; Palloni et al, 2001; Dawkins, 2006; Belot and Ermisch, 2009; Pettersson and Malmberg, 2009; Dahl and Sorenson, 2010a, 2010b; Hedman, 2013; Mulder and Malmberg, 2014; Spring et al., 2017; Mulder, 2018; Ermisch and Mulder, 2019; Jeon, 2020; Mulder, Lundholm and Malmberg, 2020; Gillespie, 2022; Ilyés et al., 2022; Blumenstock, Chi and Tan, 2023) . But we still know little about how non-kin tiesbetween friends, peers, classmates, etc.-influence residential choices within cities. What evidence we have is drawn from phone call records. These data suggest that non-kin ties matter-people are more likely to move to neighbourhoods that are more proximate to ties (Büchel et al., 2020 )-but provide scant information about the nature of these ties. Without knowing more about the ties, it is difficult to understand their broader social implications, including their role in segregation processes.

This paper investigates how proximity to former school peers affects residential mobility. We focus on schools because they are key sites for forming friendships that often persist into adulthood: In the US, for example, approximately 30 per cent of 30-year-olds' friendships originate from school (Thomas, 2019) . By concentrating on school peers, we address three gaps in previous research. First, knowing the genesis of the social ties enables us to better theorize their effects, including changes over the life-course. Second, whereas previous studies have used data spanning few years, we use longitudinal data extending from our study population's adolescence through their thirties, allowing us to assess long-term effects. Finally, our data contain rich demographic and socioeconomic information, allowing us to account for other sources of residential sorting. They also allow us to compare peer proximity effects to the effects of proximity to work and residential origins.

We estimate the effect of former school peers on young adults' residential destinations by applying discrete choice models (McFadden, 1978; Bruch and Mare, 2012) to Swedish register data. These data track individuals' residential locations and the schools they attended using precise geographic coordinates. Because social influence and homophily are confounded (Shalizi and Thomas, 2011) , we adopt an identification strategy from previous studies of the effect of social ties on internal migration (Dahl and Sorenson, 2010a, 2010b) , but reconfigured in explicitly spatial terms. We sample a set of 9 th and 12 th grade peers and track their residential locations over time. We also identify the residential locations of students from older cohorts who attended the same schools. We use the older cohorts as a natural reference group for our focal cohorts, estimating the effect of proximity to peers net of the spatial distribution of the older cohorts. We find that young adults are more likely to move to neighbourhoods near to former school peers. These effects are stronger for 12 th grade peers than 9 th grade peers, but vary over the young adult life-course, diminishing with university attendance and union formation.

While we do not analyse segregation dynamics, our analysis highlights schools as potential institutional engines of residential stratification. Because schools draw from local catchments, there is often a tight link between residential and school segregation for school-aged children (Rivkin, 1994; Frankenberg, 2013) . If young adults' moves keep them near childhood peers who disproportionately share the same socioeconomic and ethnic backgrounds, this may perpetuate residential segregation within and across generations (Sharkey, 2013) .

Background

Our analyses are grounded in the emerging social structural sorting perspective on residential mobility and segregation (Krysan and Crowder, 2017) . This perspective critiques and revises two classical models of mobility: the push-pull model (Rossi, 1955; Sabagh, Van Arsdol and Butler, 1969) and the housing search model (Wolpert, 1965; Brown and Moore, 1970) . Classical models have typically started with 'assumptions that individuals approach decisions about when, and to which neighbourhoods, to move with a high level of rationality and with comprehensive knowledge of all residential options (Krysan and Crowder, 2017, p.259) . The social structural sorting perspective, in contrast, posits that 'when considering a move, individuals approach the housing search with biases about specific neighbourhoods and limited knowledge about large swaths of metropolitan space' (p. 260). All three frameworks imply that people are more likely to move near to their friends and peers, but the mechanisms differ between them. Push-pull models view nearby network ties like amenities that increase the attractiveness of residential locations. The housing search model focuses on peers and friends as informants who influence which housing units are included in the housing search. The social structural sorting perspective suggests that peers also inform broader perceptions of the existence and suitability of neighbourhoods.

In the push-pull framework, having nearby friends is attractive because spatial proximity improves access to social capital (Coleman, 1988; Portes, 2000; Ryan et al., 2008) . This nearby social capital can be converted into social support-emotional, instrumental, or financial (Fischer, 1982; Wellman and Wortley, 1990; Wellman, 1992; Amato, 1993; Carbery and Buhrmester, 1998; Mok and Wellman, 2007; Weiner and Hannum, 2013; Martí, Bolíbar and Lozares, 2017) -and reduces residential mobility (Kan, 2007) . Having local network ties can also help individuals sustain and grow their networks. Geographic proximity increases face-to-face contact that cements desired social relationships and prevents tie dissolution (Ebbesen, Kjos and Koněcni, 1976; Rose, 1984; Fehr, 1996; Johnson et al., 2004; Mesch and Talmud, 2006; Mok and Wellman, 2007) . And proximity to friends enables the formation of new friendships through mechanisms of triadic closure (Verbrugge, 1979; Jackson and Rogers, 2007) . Interests in drawing on, preserving, and growing social capital all imply that people will be more likely to move to residential destinations near friends.

Friends and peers not only offer social support, but also provide information that guides housing searches (Herbert, 1973; Farley, 1996; Krysan, 2008) . This information takes specific and diffuse forms. The housing search model focuses on information about particular housing units that are available to rent or buy (DeSena, 1994; DiMaggio and Louch, 1998; Murdie, 2008; Skobba and Goetz, 2013) . If people have more knowledge of vacancies in their residential vicinity, EFFECT OF SCHOOL PEERS ON RESIDENTIAL MOBILITY IN YOUNG ADULTHOOD then relying on friends to identify vacant units will increase the likelihood that people will rent or buy near to them, even absent a strong preference to do so. In contrast, the social structural sorting perspective emphasizes how friends and peers foster a more diffuse awareness of select neighbourhoods, influencing the 'mental maps' that frame housing searches (Gould and White, 1986; Krysan and Crowder, 2017) . These mental maps may be superficial and biased, as with parents who glean neighbourhood reputations from conversations with other parents (Lareau, 2014; Bader, Lareau and Evans, 2019) or riddled with residential blind spots about which people have little or no information (Krysan and Bader, 2009) . We expect that mental maps will be more complete with respect to neighbourhoods near to friends and peers, leading people to favour these areas when conducting housing searches. Both the specific and diffuse roles of network ties in providing housing market information imply that people will move to neighbourhoods near friends and peers.

Variations in peer effects during early adulthood

While we have ample reason to believe that proximity to peers will matter for residential mobility, life-course theory and the 'linked lives' perspective (Elder, Johnson and Crosnoe, 2003; Antonucci et al., 2010; Coulter, van Ham and Findlay, 2016) suggest that even as peer relationships from adolescence persist into adulthood (Thomas, 2019) , their salience for life choices will shift. When adults are young and single, peers should be important sources of support and connection and highly salient for residential choices (Amato, 1993; Carbery and Buhrmester, 1998) . As people settle into careers and form families, the influence of friends and peers will wane as other considerations become more salient (Gillespie and Mulder, 2020) . This waning peer influence will have two sources: changing needs for social support and network replacement.

Changing needs for social support will most often involve young adults and their parents (Silverstein and Bengston 1997) . Adults approaching middle age may come to rely on parents as they balance family and labour market demands (Silverstein et al. 2002; Compton and Pollack 2014) . And parents may need their own help as they enter retirement and encounter health problems or widowhood (Silverstein 1995) . Given that provision of assistance is enabled by geographic proximity, it is not surprising that proximity to kin becomes more important for mobility decisions as people age and begin to form families (Michielin, Mulder and Zorlu, 2008; Pettersson and Malmberg, 2009; Mulder and Malmberg, 2014; Chan and Ermisch, 2015; Mulder, 2018) . Given this evidence, we expect that the effect of proximity to peers will also attenuate as people grow older and transition into new social roles.

People also form new relationships during adulthood that can crowd out old friendships. One key relationship is that between romantic partners. People who are married report de-emphasizing friendships relative to kin relationships (Haggerty et al., 2023) . In couples, both partners must also agree on residential choices, which might further dilute the social influence of any one partner's friends. We thus expect that proximity to friends will be less important for the residential choices of cohabitating couples. Beyond romantic relationships, people will also form more informal relationships in new social settings or 'foci' (Feld, 1981 (Feld, , 1982)) , which may displace previous friendships and weaken their importance for residential choices. University is one significant source of disruption where people will make many new friends (Rose, 1984; Johnson et al., 2004) . We thus expect that previous school peers will matter less for the residential choices of those who complete university. At the same time, people enrolled in university will also be constrained by the need to live close to campus and participate in campus social life. We thus expect that people enrolled in university will be less sensitive to proximity to compulsory and high school peers.

Social influence vs. homophily

Peer effects like those we theorize above are confounded with homophily (Lazarsfeld & Merton, 1954; McPherson, Smith-Lovin and Cook, 2001) . Similarity between friends in terms of attitudes, preferences, and habits always looms as an alternative explanation when socially connected people make similar choices (Marsden and Friedkin, 1993; Shalizi and Thomas, 2011; Veenstra et al., 2013) . In other words, people may move near peers because friends happen to prefer the same kinds of neighbourhoods, rather than because of peer effects per se. Selection into lower and upper secondary schools could drive the resemblance between peers in terms of preferences, where families with similar means, status orientations, and cultural capital send their children to the same schools. Alternately, students' preferences and cultural habits may converge through collective socialization that occurs within schools (Bourdieu, 1986; Aschaffenburg and Maas, 1997; Nagel, Ganzeboom and Kalmijn, 2011) . And students may face stigmatization based on mannerisms and affectations developed among school peers (Sernhede, 2011) , incurring housing market discrimination that shunts peers into stigmatized neighbourhoods. In any case, a naive analysis might infer a peer effect where none exists.

To disentangle social influence from underlying homophily, we must account for conditions, attitudes, and constraints shared among peers and resulting in selective mobility patterns. To do so, we adopt an approach from Dahl and Sorenson (2010a, 2010b) . We compare the geographic distribution of cohorts of students who have attended the same schools, but at different times, using older cohorts as a natural reference group. We assume that the selection and socialization pressures will be the same across cohorts, inducing shared preferences, attitudes, and constraints between cohorts, but that there will be few or no direct ties between cohorts. Controlling for the geographic distribution of prior cohorts allows us to isolate the effect of peers per se.

The Swedish context

We test for peer effects in Sweden, a country with a diverse population and heterogeneous settlement patterns. Nearly 50 per cent of the population lives in Sweden's three largest urban areas, Stockholm, Gothenburg, and Malmö. The rest of the population lives in lower density rural areas and in small and midsized cities that dot the countryside. The three major urban centres are coastal, and, particularly in the case of Stockholm, cross cut with water ways. However the most populated areas are well connected by roads, bridges, tunnels and extensive public transit networks, including commuter rail, subway lines, light rail, buses, and ferries. However, the main public transit arteries are not equally accessible to all residents (e.g. for Stockholm: Rokem and Vaughan, 2018) . In rural areas, long distance trains connect population centres, but the car remains the main mode of transportation outside of urban cores. The physical and modal fragmentation of the county poses some challenges for assessing distances in ways that reflect the actual experiences of Sweden's residents. We rely on straight-line distances in our analysis but consider limitations to this approach in our discussion.

Sweden's housing market is not only divided geographically, but also socially. While in rural and peripheral areas in Sweden, housing primarily consists of owner-occupied dwellings (Wimark, Andersson and Malmberg, 2020) , urban areas have traditionally had more mixed systems, combining ownership tenures and private rentals with a social housing sector where rent-controlled apartments are allocated through housing queues. Over the past few decades, housing in larger cities has shifted away from rentals towards homeownership (Christophers, 2013; Andersson and Turner, 2014; Wimark, Andersson and Malmberg,, 2020) . Access to the remaining rental housing, particularly close to city centers, often requires yearslong waits in urban housing queues. These long wait times, when coupled with high entry costs for home ownership, have pushed younger, lower-income, and immigrant households to the outskirts of cities (Malmberg et al., 2018; Haandrikman et al., 2021) . At the same time, younger adults, who are more likely to live in cities (Kulu, Lundholm and Malmberg, 2018) , often struggle to secure housing, given Sweden's early ages of nest leaving (Angelini, Laferrère and Pasini 2011) . In this context, friends and peers can become critical sources of support and information.

Data and methods

Our analysis relies on Swedish register data provided by Statistics Sweden (SCB). These data describe life events, demographic attributes, and socioeconomic outcomes at the microlevel for the entire Swedish population between 1990 and 2017. Crucially, our data contain individual school enrolment and residential histories geo-coded at 100 m resolution. These data are longitudinal, with individuals, schools, and residential locations tracked using unique, time-consistent identifiers. We target populations of 1,633,492 unique 9 th graders enrolled in Swedish schools from Spring 1991 to 2003, and 1,242,792 unique 12 th graders enrolled from Spring 1998 to 2007. In Sweden, 9 th grade marks the end of compulsory education (Grundskola) and students are generally 14-15 years old when they enrol. Over 90 per cent of students also go on to threeyear upper secondary schools (Gymnasieskola) that are organized by area of study (e.g. natural sciences, social sciences, vocational) (Holmlund and Böhlmark, 2017) . 12 th grade is the terminal year for upper secondary schooling and students are generally 17-18 years old when they enrol.

We limit our analysis to these cohorts to ensure that we observe post-schooling (i.e. from age 19) residential histories for our target population. To minimize potential problems stemming from the overlap between the peer networks of temporally adjacent school cohorts, we select cohorts at three-year intervals within schools. This makes sense because students typically attend upper secondary schools for three years, while students in grades 7-9 (högstadiet) are often located in different schools than those in grades 1-6 (lågstadiet and mellanstadiet). Thus, students in our focal, analytical cohorts often do not attend school at the same time as those in the three-years older reference cohorts that serve as natural controls in our adjustment strategy. Three years is the smallest possible gap that allows us to minimize social influence between peer and lagged cohorts, while maximizing the comparability of school environments.

We define residential origins and destinations using 100 m × 100 m blocks to avoid relying on arbitrary administrative units (SCB, 2010) . We analyse residential choices within metro areas using SCB's local labour market areas (LMAs), which are derived by aggregating

EFFECT OF SCHOOL PEERS ON RESIDENTIAL MOBILITY IN YOUNG ADULTHOOD

together municipalities connected by large numbers of commuters. Over 90 per cent of residential choicesdecisions to remain in or change houses-are made within these metros. We limit our analysis to within metro moves because these make up the bulk of moves and because the residential choices of between-metro movers may be idiosyncratically related to changes in employment or family circumstances.

Main explanatory variable and identification strategy

We operationalize peer effects based on the proximity of potential residential destinations to the time-varying residential locations of individuals' former school peers, with both characterized by their coordinates on a 100 m x 100 m square grid. Our approach is illustrated in Figure 1 . For a given square j and a given individual i from school s and cohort c observed at time t, we count the number of peers, n isckt , living in each square, k = 1,2,...,K. We also calculate the Euclidean distance, d jk , measured in metres, between the square j and all other squares k using their centroids. Our measure of distance to peers for square j is simply the mean of the log-distances weighted by the number of peers in each residential square:

mis(c-3) jt = K k=1 log d jk n isckt K k=1 n isckt (1)

Repeating this calculation for all potential destinations, j = 1,2,...,K, results in a spatial profile of distance to peers covering the entire city, as shown at the bottom of Figure 1 . Our use of the log transformation accounts for the diminishing marginal effect of distance. It creates an islanding effect for the peer proximity function, where the lowest values of mean log-distance occur in the areas immediately adjacent to peers. We likewise operationalize unobserved shared preferences and constraints using distance to peers from prior cohorts, or what we term, 'distance to lagged peers'. Specifically, we calculate mis(c-3) jt , a lagged version of miscjt for students from the same school, but from three cohorts prior. Note that both miscjt and mis(c-3) jt are calculated based on the observed distribution of each person's own and prior cohort at time t, the beginning of each mobility interval. By controlling for mis(c-3) jt , we isolate the effect of miscjt net of selection and socialization pressures in the school.

The residential mobility model

We use conditional logistic regression models to examine how proximity to peers affects where people move during early adulthood (McFadden, 1978; Bruch and Mare, 2012) , tracing the residential histories of individuals from three years after they attend 9 th grade up until 2017 in two separate samples. Specifically, we model the probability that an individual i who attended school s in cohort c moves in the interval [t,t + 1] to a specific residential destination j drawn from a 'choice set' of possible residential destinations C it as follows:

P isct+1 ( j | C it ) = exp β 1 miscjt + β 2 mis(c-3) jt + β 3 X ijt -ln(q ijt ) k∈Cit exp(β 1 misckt + β 2 mis(c-3)kt + β 3 X ikt -ln(q ijt ))

(2)

Figure 1 Measuring distance to peers Note: Squares are potential residential locations and the dots mark locations of school peers. In the first step, the weighted mean distance from a focal square, j, and all other squares, k = 1,2,...,K, is calculated, where the weights are the number of school peers, n, living in k. This calculation is repeated for all potential destination squares, j = 1,2,...,K, to compute the average distance to all school peers for each potential destination square.

Where β 1 and β 2 are scalars, miscjt and mis(c-3) jt are our measures of distance-to-peers in each person's own school cohort and in the cohort three years older, β 3 is a row vector of coefficients, X ijt is a column vector of control variables, and -ln(q ijt ) is a sampling correction we discuss below. We also add an interaction term between miscjt , mis(c-3) jt , X ijt and a dummy variable identifying the origin neighbourhood (see below) to distinguish pull from retention effects. Retention effects relate to whether a person is immobile. In contrast, pull effects relate to where people will move when they do decide to move. Our focus is on the pull effects.

We can interpret the coefficients attached to logtransformed independent variables, including our distance-to-peers measures, as approximate, quasielasticities. A coefficient of -1 for a log-transformed covariate would indicate that, compared to an otherwise equivalent alternative in the same choice set, an alternative with a 1 per cent proportionally higher value of the associated independent variable would have an approximately 1 per cent lower probability of being chosen, proportionally speaking. Appendix B contains a proof.

We apply the mobility model to residential choices made within each metro area. We assume that each individual chooses a destination from the full choice set, C, of residential locations within their metropolitan area, i.e. C it = C ∀i,t. For the sake of computational tractability, we randomly sample the choice set down to 100 residential destinations per person-year of residential history, making sure to always include the origin and destination squares. We sample additional alternatives from within the metro independently for each person-year, doing so without replacement within person-years. Including -1 times the log of the sampling fraction, -ln(q ijt ), as an offset in the model yields estimates that are consistent and unbiased compared to estimates that would be produced using the full choice set (Bruch and Mare, 2012; Jarvis, 2019) . Choice-set sampling is standard practice when dealing with large choice sets (Bruch, 2014; Gullickson, 2021; Mouw, Kalleberg and Schultz, 2024) . A fuller explanation of our choice set sampling strategy as well as a description of the data structure used to estimate the models is presented in Appendix A. We estimate the models in Stata using the clogit command with unique person-year identifiers serving as the stratifying variable (StataCorp LLC, 2021).

Control variables

Control variables in our models account for the sorting of people with different attributes into residential locations with different geographic positions and sociodemographic compositions. Location-level covariates are the primary explanatory variables in our models. We represent patterns of residential sorting through the specification of individual-by-location interactions, by which we mean any non-linear function of select individual-and location-level variables, including multiplication, absolute differences, ratios, etc. We include two groups of interactions: spatial and sociodemographic. These are summarized in Table 1 . In the table, Column A describes the individual-level covariates, Column B presents the corresponding location-level covariates, and Column C indicates the functional form for the interaction between variables from Column A and B. Tables 2 and 3 contain the corresponding descriptive statistics for neighbourhood level and individual level variables.

Spatial attraction to familiar areas, whether because of current residence, work, or childhood experience, is a key aspect of residential sorting. We account for immobility and inertia by including a dummy variable indicating if the potential destination is the origin, as well as the distance from the residential origin at t to the potential destination. We account for the tendency to minimize commuting time by including the distance from each person's workplace at t to the potential destination. When a person does not have a workplace, we set this variable to an arbitrary constant. This has no effect on our estimates, as variables that are constant within choice sets cancel out in the numerator and denominator in Equation 2. To represent lingering place attachment from childhood, we account for the distance from the relevant 9 th or 12 th grade school as an additional control. In all cases, we calculate simple Euclidean distances in metres and transform them using the natural logarithm, as with our measures of distance-to-peers. Table 3 summarizes these distances, expressed in log-metres, separately for the 9 th and 12 th grade samples and separately for stayers (i.e. those choosing their origin), movers' origins, movers' destinations, and then randomly sampled alternative locations.

Finally, we include three additional spatial controls. All else equal, people will be more likely to move to areas with more housing units. We proxy the number of housing units by controlling for the natural logarithm of the population in each residential 100 by 100 m square, calculated separately in each year. At the same time, accessing housing in major urban centres like Stockholm, Gothenburg, and Malmö often requires either a high income or seniority in public housing queues, which young people often lack. To capture the challenges of relocating close to urban cores, we calculate the mean log-distance to all other individuals in the location's metro area and the location's municipality, as well as the interaction between the two. Combined, these controls effectively account for the distribution and density of the housing around the key population centres in each metro area.

EFFECT OF SCHOOL PEERS ON RESIDENTIAL MOBILITY IN YOUNG ADULTHOOD

We also consider several socio-demographic controls at the neighbourhood-level. For these, we rely on Small Areas for Market Statistics or SAMS, rather than 100 m x 100 m squares. SAMS are mutually exclusive, bounded units defined by SCB. They have time-consistent boundaries that follow natural and man-made geographic features. They are relatively geographically compact and contain, on average, approximately 2,000 residents (Amcoff, 2012) . There is typically a many-to-one relationship between 100 m x 100 m squares and SAMS. When there are many-tomany relationships, we assign each square to a single SAMS based on where the majority of its population lives. We use SAMS for socio-demographic controls Shares of residents in micro-, meso-, and macro-ancestry groups.

B where Anc (A) = Anc (B) at micro, meso, and macro levels.

Note: All locations are geocoded to squares on a 100 m ×100 m grid (SWEREF99 TM). AB denotes the Euclidean distance between the points A and B. To account for selection into mobility, all listed variables are interacted with the dummy variable identifying the origin location. Models also included interactions between the origin dummy and the individual's education, university enrolment status, age, partnership status, and presence of children. Source: Authors' calculations using Swedish population registers provided by SCB.

Downloaded from https://academic.oup.com/esr/advance-article/doi/10.1093/esr/jcaf002/8071692 by Conservatorio di Musica di Como G Verdi user on 21 May 2025

because their populations are sufficiently large to justify the calculation of compositional covariates. 100 m x 100 m squares frequently have such small populations that it is inappropriate to characterize their demographic compositions directly. We consider two controls to account for sorting based on the socioeconomic make-up of individuals and neighbourhoods. To account for income sorting, we use information about time-varying disposable incomes of families, measured in hundreds of Swedish crowns (SEK). This yearly measure combines income from all sources, including wages, capital gains, pensions, government benefits, etc. and subtracts taxes paid. We have this measure for all individuals in our samples, but also use it to calculate each neighbourhood's median income. We include the log of the neighbourhood median family income as a covariate, as well as the log of the absolute difference in the median income and the individual's own income, measured in hundreds of SEK. We also control for the potential sorting of people with tertiary education into neighbourhoods with more educated residents. This can be related to the cultural capital and long-term earnings potential that tertiary education signals. We use individual-level education data, coded into ISCED categories in SCB's labour market registers, to classify people as college educated. We treat those with ISCED codes 5 or higher, (i.e. > 14 years of education) as college educated. At the neighbourhood level, we calculate the share of residents ages 18 and older who have a college degree based on this definition. In our models, we then include the share of college educated residents and the multiplicative interaction between this measure and a dummy variable indicating if the potential mover has a college degree.

Demographic controls account for the sorting of people into neighbourhoods based on their life-stage and family situation. Ongoing tertiary education is common among young adults in Sweden. To accommodate these students, cities set aside a portion of the housing stock for student residences, which they allocate through student-specific housing queues. Students may prefer to live in these residences for budget and convenience considerations and to participate in student social life. To address this, we use the university registers and identify, in each year, whether individuals are enrolled in university classes or programs. We then calculate the share of residents ages 18 and older who are enrolled in university in each neighbourhood. Our models include the main effect of the neighbourhoodlevel share of university enrolled residents, and the interaction between this and whether the potential mover is enrolled. Outside of enrolment, we also expect that young adults will tend to live in neighbourhoods where other young adults live. To account for this, we consider a set of age-based controls. We sort Table 3. Continued Downloaded from https://academic.oup.com/esr/advance-article/doi/10.1093/esr/jcaf002/8071692 by Conservatorio di Musica di Como G Verdi user on 21 May 2025 EFFECT OF SCHOOL PEERS ON RESIDENTIAL MOBILITY IN YOUNG ADULTHOOD

households with children in the neighbourhood, and the interaction between the share of households with children and a dummy indicating if the potential mover has a child in the household. Entry into cohabitation is another important lifecourse transition that typically precedes the birth of a child. While we do not expect that people sort into neighbourhoods based on cohabitation per se, cohabitation complicates residential choices. A partner will bring along with them their own place attachments and networks ties. Ignoring these cross-cutting pressures in our models could obscure whether the effect of peers in our focal sample varies over the life-course. To address this confounding, we control for the locations of cohabiting partner's peers in each year. To identify premarital cohabitation, we identify all individuals who are ever registered as married or who have children together. For these couples, we review residential history records to identify the years in which they are living on the same property. Those who are living on the same property as an (eventual) partner are classified as 'cohabiting'. In person-years when we identify that a person is cohabiting, we control for the residential locations of the partner's peers from 9 th or 12 th grade, using the same approach presented in Equation 1 . When we cannot identify a partner, the corresponding distance variable is set to an arbitrary constant, which has no effect on our estimates.

There is ample evidence that neighbourhood ethnic composition influences residential moves in Sweden. While we do not observe ethnicity in our data, we can account for residential sorting based on ancestry, a closely related concept, using country of birth data for individuals and their parents, classified into 51 country groups. The largest sending countries (e.g. Finland, Somalia, Turkey, and Iraq) have their own separate categories, but other countries are grouped based on geography and linguistic similarity (e.g. Japan, Korea, and Taiwan; Canada and the United States). We assign all individuals in the register to a single ancestry category based first on their mother's country of birth, then their father's country of birth, and finally, if both parents are missing in our data, then the person's own place of birth. We designate the 51 country groups as micro-ancestries, but also group these into coarser meso-(18 categories) and macro (6 category) ancestry groups, to account for linguistic and cultural similarities that may span borders. We calculate the shares in each micro-, meso-, and macro-ancestry group in each SAMS. We then include the micro-, meso-, and macro-ancestry shares matching our sample members' own micro-, meso-, and macro-ancestry assignment as covariates. The ancestry assignment is time-invariant for individuals, but ancestry composition is timevarying for neighbourhoods.

Results

Figure 2 reports select coefficients from our conditional logit models of residential mobility for the 9 th grade (left) and 12 th grade (right) samples. Complete estimation results are presented in Appendix Table C-1. Our main result is that people are significantly less likely to move to neighbourhoods that are further away from peers. We infer this based on the negative sign for the distance-to-peers coefficients in both the 9 th and 12 th grade samples. In the 9 th grade sample, the coefficient estimate indicates that a neighbourhood that is 1 per cent further from peers will have, proportionally speaking, a 0.2 per cent lower probability of being chosen as a destination relative to an otherwise equivalent reference neighbourhood (P < 0.05, twotailed test). In the 12 th grade sample, a neighbourhood that is 1 per cent further from peers compared to an otherwise equivalent reference neighbourhood will be 0.8 per cent less likely to be chosen, proportionally speaking (P < 0.01).

To obtain the results in Figure 2 , we also accounted for log-transformed distances between potential destinations and the residential origins and workplaces of individuals at each time t and the locations of the schools they attended in 9 th or 12 th grade. While we cannot interpret these effects causally, they serve as useful points of reference when considering the magnitude of the peer effects. The distance from the origin neighbourhood is the strongest proximity-based predictor of residential choice. Individuals in both samples have, proportionally speaking, a roughly 0.9 per cent lower probability of choosing a neighbourhood that is 1 per cent further away from their residential origin than an otherwise equivalent reference neighbourhood. In the 9 th grade sample, the distance-to-peers effect is roughly similar to the distance-to-work (-0.15 per cent) and distance-to-school (-0.25 per cent) effects. The peer effects in the 12 th grade sample are the second strongest among the proximity effects, and significantly larger than the distance-to-work and distance-to-school effects. The distance-to-school effect is not significant in the 12 th grade sample, while the distance to work effect is similar to that in the 9 th grade sample, at about -0.17 per cent. These results imply that the peer effects are at least on par with, and perhaps greater than, the distance-to-work and distance-to-school effects.

We next consider variations in the effect of peer proximity over the early adult life-course. We expected that the peer effect would attenuate with age, university enrolment, university completion, cohabitation, and when having children. We test these predictions in our conditional logit models by interacting individual-level indicators of these life-course statuses and milestones with the distance-to-peers variable. If our expectations are correct, then we should obtain negative estimates for the main effect and positive estimates for interaction effects.

We find that university enrolment and having a partner attenuate peer effects in both samples. Attenuation is indicated by positive coefficients for the interaction between these individual-level covariates and the distance-to-peers measure (Appendix Table C -2 contains full coefficient estimates). The university enrolment interactions are positive and significant (P < 0.05) in both samples. The partner interaction is positive and significant in the 12 th grade sample (P < 0.05), but only marginally so (P < 0.10) in the 9 th grade sample. However, when we work out the total, implied distance-to-peers effect for those who have a partner or are enrolled in university, we find that the effects are not significantly different from 0 in either the 9 th or the 12 th grade sample. This implies that having a partner or attending university fully attenuates the peer effects in both samples. Contradicting our other expectations, peer influence does not vary significantly with age, completion of a university degree, or having children in either the 9 th or 12 th grade samples, net of enrolment and partnering.

We graphically re-interpret the key interaction effects in Figure 3 . We plot cumulative predicted choice probabilities as a function of distance for hypothetical movers with different early life-course attributes searching for housing within a 10 km radius of a peer. We focus on the probability of moving within half of this radius, or within 5 km of the peer. Among those lacking a partner and having no university enrolments, members of our 9 th grade sample would have a 30 per cent probability of moving within 5 km of a peer in our hypothetical scenario. Someone choosing randomly would have a 25 per cent probability of moving with 5 km of the peer. But for those in the 9 th grade sample enrolled in university or living with a partner, the implied probability of moving within 5 km of a peer is slightly less than 25 per cent. In the 12 th grade sample, the peer effects are stronger: a single, non-enrolled person from the 12 th grade sample would have a little less than a 50 per cent chance of moving within 5 km of a peer in our hypothetical scenario, compared to 25 per cent for someone choosing at random. But the curves for a university enrolled or partnered person are virtually indistinguishable from the random choice line. These plots underscore how the peer effects we estimate are attenuated by university enrolment and partnering.

Discussion

In this article, we estimated the degree to which residential proximity to school peers matters for residential Figure 2 Coefficients from conditional logit models of location choice for 9 th and 12 th grade samples Note: Models were estimated using person-year-location formatted data. The dependent variable is a binary indicator coded, in each person-year, as 1 for the chosen location and 0 otherwise. Distance to school is measured from the compulsory school for 9 th graders and the high school for 12 th graders. Because all distances were logged, the coefficients can be interpreted directly as proportional, percentage changes in choice probabilities, as described in the text. Full coefficient estimates are presented in Appendix Table C- Downloaded from https://academic.oup.com/esr/advance-article/doi/10.1093/esr/jcaf002/8071692 by Conservatorio di Musica di Como G Verdi user on 21 May 2025 for processes of selection and socialization that induce similarity in peers' unobserved tastes and preferences. Our findings show that people are significantly more likely to move to neighbourhoods near to their former 9 th and 12 th grade peers. The effects are as strong or stronger than the effects of proximity to work and proximity to school on residential choices. We shore up these findings by demonstrating that social influence wanes in two theoretically expected ways, declining with university enrolment and cohabitation with a partner.

Beyond demonstrating that networks matter for residential mobility, our results suggest that life-course events in early adulthood embed individuals in different network structures that may replace or displace relationships to childhood peers. In this sense, we contribute to the 'linked lives' perspective on residential mobility (Coulter, van Ham and Findlay, 2016) . However, our analysis partly reframes this perspective, suggesting that life-course changes can also un-link lives. While comparing coefficients across non-linear models can be fraught, we find substantially larger peer effects in the 12 th grade as compared to 9 th grade sample. This suggests that transitions into upper secondary schooling can sever bonds for peers who do not make this transition together, while cementing old bonds or forging new bonds among those who do attend 12 th grade together. In a similar vein, we find that peer effects attenuate for those cohabiting with a partner and attending university. Those who experience these life-course transitions are less likely to move near to peers. This suggests that new network ties become more salient to residential choices as people move through the early adult life-course. People with partners may need to de-emphasize some prior relationships to make room for a partner's preferences, including their preferences to live near their own closest friends and family. And university students may, at least temporarily, de-emphasize old friendships in favour of new social milieus.

While these findings emphasize shifts in social networks, our study remains unable to distinguish the mechanisms driving these effects. On one hand, information sharing between peers, or the lack of it, could explain the patterns of interactions we observe. For example, living with a partner may crowd out communication with friends, thus reducing access to information that would have otherwise informed housing choices. On the other hand, the formation of a partnership may also change the need for and expected sources of social support, with parents or other relatives viewed as most likely to offer support for young adults considering raising their own families. Sorting out the precise mechanisms driving the waxing and waning of peer effects over the life-course will likely require different data than we employ here, including qualitative interviews (Owens and Clampet-Lundquist, 2017) , focus groups (Krysan and Crowder, 2017) , innovative survey instruments (Krysan and Bader, 2009) , or digital trace data (Bailey et al., 2018) . But we should emphasize that it need not be one or the other-provision of support vs. information sharing-driving peer effects. They likely go together. Provision of information about the residential lay-of-the-land may be a key aspect of social support, and provision of social support may engender the sharing of information about local neighbourhoods.

Apart from partner-related interactions, we also find evidence of network displacement for university enrolment: distance-to-peer effects are weaker for those enrolled in university. Because our models explicitly control for the presence of university students at the neighbourhood level, we believe this attenuation effect represents a change in networks, not the spatial clustering of student housing. However, we find no evidence that having a university degree reduces the social influence effect of former 9 th and 12 th grade peers, indicating that the network replacement during tertiary education is only temporary. Enrolment at a university might temporarily reduce the salience of former school peers for residential choices, without really causing these ties to atrophy.

Although we believe the peer effects we estimated are likely causal, we acknowledge weaknesses in our identification strategy. First, we assume that processes of school-level selection and socialization are the same across cohorts separated by three years. If this assumption does not hold, it could bias our estimates in unknown ways. Second, we assume that social influence occurs only within, not across, cohorts. If there are ties across cohorts, then our lagged peers measure may encode additional peer effects, not just common selection and socialization pressures at the school level. This may lead us to understate the true peer effects. Third, while our approach accounts for processes of socialization and selection that might induce similarity in unobserved preferences between all students attending the same school, we do not account for the sorting of students into friendships within schools. This could lead us to both over-and understate the true peer effects. Peer effects may be understated because we assume all school peers affect mobility, whereas it is likely that only those with whom people have direct friendships will exert influence. On the other hand, true friends among observed peers may be homophilous on tastes and preferences that our identification strategy does not address because they do not result from school-wide selection and socialization pressures. So, while our measures of peer ties may simply be noisy measures of friendship ties, thus biasing our estimates EFFECT OF SCHOOL PEERS ON RESIDENTIAL MOBILITY IN YOUNG ADULTHOOD towards zero, the underlying friendship ties may be more contaminated by unobserved homophily, hence biasing our estimates away from zero. Different data or identification strategies may be needed to address this ambiguity. Finally, we lack comprehensive data on housing market prices and rents, hindering our ability to account for affordability constraints. In particular, the recent choices of peers and prior cohorts may be associated with local, neighbourhood-level changes in housing supply and demand that are driving changes in prices or rents. The implications of this for our peer effects are unclear. It would depend on the degree to which peers and prior cohorts are converging on or avoiding neighbourhoods with changing supply or demand.

By focussing only on residential choices in Sweden, it is also unclear if our findings would generalize to other European countries. We foresee two potential problems. First, the Swedish education and welfare systems have historically encouraged early nest-leaving, for example by providing free tuition, generous living stipends, and access to housing for students enrolled in tertiary education (Billari, Philipov and Baizán, 2001; Angelini, Laferrère and Pasini, 2011) . The policy mix in other European countries may be less conductive to home leaving in early adulthood. At the same time, the increasing marketization of Sweden's housing stock and the withdrawal of the government from direct provision of housing may be shifting the housing model closer to some of its European peers (Arbaci, 2007) . These policy changes have led to housing shortages and rising costs-particularly in urban centres-that most affect the young, the economically disadvantaged, and immigrants (Christophers, 2013) . This may delay nest-leaving, push leavers to peripheral neighbourhoods, and increase the importance of network ties.

Second, Sweden's physical and social geography is varied, spanning dense urban centres, mid-sized cities, and sparsely populated rural regions. While we believe that our main finding-that peers matter-would hold up in other contexts, it seems likely that the organization of urban housing systems could moderate the effects of peers. Exploring this variation is an exciting avenue for future research, but it would necessitate a reconsideration of our approach to measuring proximity. Our straight-line distance measures may not be systematically biased, but they are almost certainly noisy measures of practical distances between places, which are mediated by transportation infrastructure and the built environment. This would imply that the significant peer effects we identify are underestimates of the true peer effects. In other contexts, with different physical geographies and built infrastructures, straight-line distances may be more or less noisy measures of the practical distance between places. Without using measures of proximity that are more readily comparable across places, differences in the effects of peers between places may result from differential measurement error. A cross-country comparative analysis should adopt measures of proximity that are more 'human-centred', perhaps using road networks alongside information about public transit routes (Roberto, 2018) .

If we set these concerns aside, the strong network effects we observe provide support for aspects of the social structural sorting view of residential segregation (Krysan and Crowder, 2017) . Our findings suggest that social networks, and the neighbourhood institutions in which they are forged, may contribute to the perpetuation of residential segregation. If students attend schools near their homes where their peers have similar ethnic or class backgrounds, and if the peers they meet at school influence residential choices during adulthood, school assignments can provide a blueprint for adult residential choices, binding people to segregated patterns of settlement that then persist across time. Additionally, these early social networks can guide selection into other social contexts across the subsequent life-course-not just neighbourhoods, but also tertiary education, occupations, and workplaces. These contextual experiences and affiliations can, in turn, shape the ongoing evolution of social networks, engendering not only residential segregation but also segregation in other life domains, such as marriage and the labour market. While our paper does not examine the dynamics of segregation explicitly, this is an important area for future research. Knowing more about the link between residential segregation and social networks can help to address the persistence of social and spatial stratification and develop policies to break enduring cycles of segregation.

References

1. Urban-rural differences in helping friends and family members P R Amato Social Psychology Quarterly 56 1993
2. Hur bra fungerar SAMSområdena i studier av grannskapseffekter? En studie av SAMSområdenas homogenitet Jan Amcoff 10.3384/svt.2012.19.2.2450 Socialvetenskaplig tidskrift SOCVET 1104-1420 2003-5624 19 2 2012 Förbundet för forskning i Socialt arbete
3. Segregation, gentrification, and residualisation: from public housing to market-driven housing allocation in inner city Stockholm Roger Andersson Lena Magnusson Turner 10.1080/14616718.2013.872949 International Journal of Housing Policy International Journal of Housing Policy 1949-1247 1949-1255 14 1 2014 Informa UK Limited
4. Nest leaving in Europe V Angelini A Laferrère G Pasini The Individual and the Welfare State: Life Histories in Europe A Börsch-Supan M Brandt K Hank M Schröder Berlin, Heidelberg Springer 2011
5. Convoys of social relations: integrating life-span and life-course perspectives T C Antonucci The Handbook of Life-Span Development R M Lerner M E Lamb A M Freund 2010 Ltd New York, NY
6. Ethnic segregation, housing systems and welfare regimes in Europe S Arbaci European Journal of Housing Policy 7 2007
7. The Trojan-horse mechanism: How networks reduce gender segregation M Arvidsson F Collet P Hedström Science Advances 7 6730 2021
8. Cultural and Educational Careers: The Dynamics of Social Reproduction Karen Aschaffenburg Ineke Maas 10.2307/2657427 American Sociological Review American Sociological Review 0003-1224 62 4 573 1997 SAGE Publications
9. Talk on the playground: the neighborhood context of school choice M D M Bader A Lareau S A Evans City & Community 18 2019
10. The economic effects of social networks: evidence from the housing market M Bailey Journal of Political Economy 126 2018
11. Friendship ties and geographical mobility: evidence from great britain M Belot J Ermisch Journal of the Royal Statistical Society. Series A (Statistics in Society) 172 2009
12. Social networks and fertility L Bernardi A Klärner Demographic Research 30 2014
13. Leaving home in europe: the experience of cohorts born around 1960 F C Billari D Philipov P Baizán International Journal of Population Geography 7 2001
14. Migration and the value of social networks J E Blumenstock G Chi X Tan The Review of Economic Studies 92 2023
15. The forms of capital P Bourdieu Handbook of Theory and Research for the Sociology of Education J G Richardson New York Greenwood Press 1986
16. The intra-urban migration process: a perspective L A Brown E G Moore Geografiska Annaler. Series B, Human Geography 52 1970
17. How population structure shapes Neighborhood segregation. AJS E E Bruch American Journal of Sociology 119 2014
18. Methodological issues in the analysis of residential preferences, residential mobility, and Neighborhood change E E Bruch R D Mare Sociological Methodology 42 2012
19. Calling from the outside: The role of networks in residential mobility K Büchel Journal of Urban Economics 119 103277 2020
20. Friendship and need fulfillment during three phases of young adulthood J Carbery D Buhrmester Journal of Social and Personal Relationships 15 1998
21. Complex contagions and the weakness of long ties D Centola M Macy American Journal of Sociology 113 2007
22. Residential proximity of parents and their adult offspring in the United Kingdom, 2009-10 T W Chan J Ermisch Population Studies 69 2015
23. A monstrous hybrid: the political economy of housing in early twenty-first century Sweden B Christophers New Political Economy 18 2013
24. Social Capital in the Creation of Human Capital James S Coleman 10.1086/228943 95-S120 American Journal of Sociology American Journal of Sociology 0002-9602 1537-5390 94 1988 University of Chicago Press
25. Family proximity, childcare, and women's labor force attachment J Compton R A Pollak Journal of Urban Economics 79 2014
26. Re-thinking residential mobility Rory Coulter Maarten Van Ham Allan M Findlay 10.1177/0309132515575417 Progress in Human Geography Progress in Human Geography 0309-1325 1477-0288 40 3 2016 SAGE Publications
27. The migration of technical workers M S Dahl O Sorenson Journal of Urban Economics 67 2010a
28. The social attachment to place M S Dahl O Sorenson Social Forces 89 2010b
29. Are social networks the ties that bind families to Neighborhoods? C J Dawkins Sociological Inquiry 21 2006. 1994 DeSena, J. N. Local gatekeeping practices and residential segregation* Housing Studies
30. by Conservatorio di Musica di Como G Verdi user on 21 May 2025 EFFECT OF SCHOOL PEERS ON RESIDENTIAL MOBILITY IN YOUNG ADULTHOOD P Dimaggio H Louch E B Ebbesen G L Kjos V J Koněcni 10.1093/esr/jcaf002/8071692 Journal of Experimental Social Psychology 63 1998. 1976 Socially embedded consumer transactions: For what kinds of purchases do people most often use networks? Spatial ecology: Its effects on the choice of friends and enemies American Sociological Review
31. The emergence and development of life course theory G H Elder M K Johnson R Crosnoe Handbook of the Life Course J T Mortimer M J Shanahan Boston, MA Springer 2003
32. Migration Versus Immobility, and Ties to Parents John Ermisch Clara H Mulder 0000-0003-0152-2225 10.1007/s10680-018-9494-0 European Journal of Population Eur J Population 0168-6577 1572-9885 35 3 2019 Springer Science and Business Media LLC
33. Racial differences in the search for housing: Do whites and blacks use the same techniques to find housing? Housing Policy Debate R Farley 1996 7
34. B Fehr The focused organization of social ties Sage S L Feld Thousand Oaks, CA 1996. 1981 86 Friendship Processes
35. Social structural determinants of similarity among associates S L Feld American Sociological Review 47 1982
36. To Dwell among Friends: Personal Networks in Town and City C S Fischer 1982 University of Chicago Press Chicago, IL
37. The role of residential segregation in contemporary school segregation E Frankenberg Education and Urban Society 45 2013
38. Family and friends living nearby, neighborhood satisfaction, and residential mobility B J Gillespie City & Community 21 2022
39. Nonresident family as a motive for migration B J Gillespie C H Mulder Demographic Research 42 2020
40. P Gould R White Mental Maps London, UK Routledge 1986
41. Threshold Models of Collective Behavior Mark S Granovetter 10.1086/226707 American Journal of Sociology American Journal of Sociology 0002-9602 1537-5390 83 6 1978 University of Chicago Press
42. M S Granovetter Getting a Job: A Study of Contacts and Careers Chicago, IL University of Chicago Press 1995 nd ed.
43. Social influence and consumption: evidence from the automobile purchases of Neighbors M Grinblatt M Keloharju S Ikäheimo The Review of Economics and Statistics 90 2008
44. A counterfactual choice approach to the study of partner selection A Gullickson Demographic Research 44 2021
45. Socio-economic segregation in European cities. A comparative study of Brussels, Copenhagen, Amsterdam, Oslo and Stockholm Karen Haandrikman 0000-0002-1246-2427 Rafael Costa 0000-0003-4523-0275 Bo Malmberg 0000-0001-7345-0932 Adrian Farner Rogne 0000-0003-2617-161X Bart Sleutjes 10.1080/02723638.2021.1959778 Urban Geography Urban Geography 0272-3638 1938-2847 44 1 2021 Informa UK Limited
46. Stability and change in newlyweds' social networks over the first years of marriage B B Haggerty Journal of Family Psychology : JFP : Journal of the Division of Family Psychology of the American Psychological Association (Division 43 2023
47. Moving near family? The influence of extended family on neighbourhood choice in an intraurban context. Population, Space and Place L Hedman 2013 19
48. The residential mobility process: some empirical observations D T Herbert Area 5 1973
49. Hur ska högstadiet organiseras? Effekter av övergången till 1-9-skolor H Holmlund A Böhlmark 2017 Uppsala, Sweden Tech. Rep. No. 2017 Institutet för arbetsmarknads-och utbildningspolitisk utvärdering
50. How to enter high-opportunity places? The role of social contacts for residential mobility V Ilyés Journal of Economic Geography 23 2022
51. Meeting strangers and friends of friends: HowRandom are social networks? M O Jackson B W Rogers American Economic Review 97 2007
52. Estimating multinomial logit models with samples of alternatives B F Jarvis Sociological Methodology 49 2019
53. Moving away from opportunity? Social networks and access to social services J S Jeon Urban Studies 57 2020
54. The process of relationship development and deterioration: Turning points in friendships that have terminated A J Johnson Communication Quarterly 52 2004
55. Residential mobility and social capital K Kan Journal of Urban Economics 61 2007
56. Does race matter in the search for housing? An exploratory study of search strategies, experiences, and locations M Krysan Social Science Research 37 2008
57. Racial Blind Spots: Black-White-Latino Differences in Community Knowledge Maria Krysan Michael D M Bader 10.1525/sp.2009.56.4.677 Social Problems Social Problems 0037-7791 1533-8533 56 4 2009 Oxford University Press (OUP)
58. Cycle of Segregation: Social Processes and Residential Stratification M Krysan K Crowder 2017 Russell Sage New York, NY
59. Is spatial mobility on the rise or in decline? An order-specific analysis of the migration of young adults in Sweden Hill Kulu Emma Lundholm Gunnar Malmberg 10.1080/00324728.2018.1451554 Population Studies Population Studies 0032-4728 1477-4747 72 3 2018 Informa UK Limited
60. Schools, housing, and the reproduction of inequality A Lareau Choosing Homes, Choosing Schools A Lareau K A Goyette New York, NY Russell Sage 2014
61. Friendship as a social process: A substantive and methodological analysis P F Lazarsfeld R K Merton Freedom and Control in Modern Society M Berger T Abel C H Page New York, NY Van Nostrand 1954
62. Residential segregation of european and non-european migrants in Sweden: 1990-2012 B Malmberg European Journal of Population 34 2018
63. Network studies of social influence P V Marsden N E Friedkin Sociological Methods & Research 22 1993
64. Network cohesion and social support J Martí M Bolíbar C Lozares Social Networks 48 2017
65. What's driving Mexico-U.S. migration? A theoretical, empirical, and policy analysis D S Massey K E Espinosa American Journal of Sociology 102 1997
66. Modeling the choice of residential location D Mcfadden 1978
67. Spatial Interaction Theory and Planning Models A Karlqvist L Lundqvist F Snickers J W Weibull North-Holland 3
68. Birds of a feather: Homophily in social networks M Mcpherson L Smith-Lovin J M Cook Annual Review of Sociology 27 2001
69. 10.1093/esr/jcaf002/8071692 by Conservatorio di Musica di Como G Verdi user on 21 May 2025
70. The quality of online and offline relationships: The role of multiplexity and duration of social relationships G Mesch I Talmud 2006 The Information Society 22
71. Distance to parents and geographical mobility F Michielin C H Mulder A Zorlu Population, Space and Place 14 2008
72. D Mok B Wellman Did distance matter before the Internet?: Interpersonal contact and support in the 1970s 2007 29
73. “Stepping-Stone” versus “Dead-End” Jobs: Occupational Structure, Work Experience, and Mobility Out of Low-Wage Jobs Ted Mouw 0000-0002-6657-1565 Arne L Kalleberg Michael A Schultz 0000-0003-1651-4295 10.1177/00031224241232957 American Sociological Review Am Sociol Rev 0003-1224 1939-8271 89 2 2024 SAGE Publications
74. Putting family centre stage: Ties to nonresident family, internal migration, and immobility. Demographic Research C H Mulder 2018 24
75. Local ties and family migration C H Mulder G Malmberg Environment and Planning A: Economy and Space 46 2014
76. Young adults' migration to cities in Sweden: Do siblings pave the way? C H Mulder E Lundholm G Malmberg Demography 57 2020
77. Pathways to housing: The experiences of sponsored refugees and refugee claimants in accessing permanent housing in Toronto R A Murdie Journal of International Migration and Integration / Revue de l'integration et de la Migration Internationale 9 2008
78. Bourdieu in the network: The influence of high and popular culture on network formation in secondary school I Nagel H B G Ganzeboom M Kalmijn Koelner Zeitschrift fuer Soziologie und Sozialpsychologie 63 2011
79. Housing mobility and the intergenerational durability of neighborhood poverty A Owens S Clampet-Lundquist Journal of Urban Affairs 39 2017
80. Social capital and international migration: A test using information on family networks A Palloni American Journal of Sociology 106 2001
81. Adult children and elderly parents as mobility attractions in Sweden Anna Pettersson Gunnar Malmberg 10.1002/psp.558 Population, Space and Place Population Space and Place 1544-8444 1544-8452 15 4 2009 Wiley
82. The two meanings of social capital A Portes Sociological Forum 15 2000
83. Residential segregation and school integration S G Rivkin Sociology of Education 67 1994
84. The spatial proximity and connectivity method for measuring and analyzing residential segregation E Roberto Sociological Methodology 48 2018
85. Geographies of ethnic segregation in Stockholm: The role of mobility and co-presence in shaping the 'diverse' city J Rokem L Vaughan Urban Studies 56 2018
86. How friendships end: patterns among young adults S M Rose Journal of Social and Personal Relationships 1 1984
87. Why Families Move P H Rossi 1955 The Free Press Glencoe, IL
88. Social Networks, Social Support and Social Capital: The Experiences of Recent Polish Migrants in London Louise Ryan Rosemary Sales Mary Tilki Bernadetta Siara 10.1177/0038038508091622 Sociology Sociology 0038-0385 1469-8684 42 4 2008 SAGE Publications
89. Some deteriminants of intrametropolitan residential mobility: Conceptual considerations G Sabagh M D Van Arsdol E W Butler Social Forces 48 1969
90. Experimental study of inequality and unpredictability in an artificial cultural market M J Salganik P S Dodds D J Watts Science 311 2006
91. Lokala arbetsmarknader -egenskaper, utveckling och funktion Scb 2010 SCB-Tryck Örebro, Sweden
92. Social networks and smoking: Exploring the effects of peer influence and smoker popularity through simulations D R Schaefer J Adams S A Haas Health Education & Behavior 40 2013 The Official Publication of the Society for Public Health Education
93. Dynamic models of segregation T C Schelling The Journal of Mathematical Sociology 1 1971
94. School, youth culture and territorial stigmatization in swedish metropolitan districts O Sernhede 2011 19
95. Homophily and contagion are generically confounded in observational social network studies C R Shalizi A C Thomas Sociological Methods & Research 40 2011
96. P Sharkey Stuck in Place: Urban Neighborhoods and the End of Progress Toward Racial Equality Chicago, IL University of Chicago Press 2013
97. Stability and Change in Temporal Distance between the Elderly and Their Children Merril Silverstein 10.2307/2061895 Demography 0070-3370 1533-7790 32 1 1995 Duke University Press
98. Reciprocity in parent-child relations over the adult life course M Silverstein The Journals of Gerontology: Series B 57 2002
99. Intergenerational solidarity and the structure of adult child-parent relationships in American families M Silverstein V L Bengtson American Journal of Sociology 103 1997
100. Mobility decisions of very low-income households K Skobba E G Goetz Cityscape 15 2013
101. Influence of proximity to kin on residential mobility and destination choice: Examining local movers in metropolitan areas A Spring Demography 54 2017
102. clogit -Conditional (fixed effects) logistic regression Llc Statacorp Stata Base Reference Manual: Release 17. College Station TX Stata Press 2021
103. Sources of friendship and structurally induced Homophily across the life course. Sociological Perspectives R J Thomas 2019 62
104. R Veenstra Network-behavior dynamics 2013 23
105. Multiplexity in adult friendships L M Verbrugge Social Forces 57 1979
106. Differences in the quantity of social support between geographically close and long-distance friendships A S B Weiner J W Hannum Journal of Social and Personal Relationships 30 2013
107. Which types of ties and networks provide what kinds of social support B Wellman 10.1093/esr/jcaf002/8071692 by Conservatorio di Musica di Como G Verdi user on 21 May 2025 EFFECT OF SCHOOL PEERS ON RESIDENTIAL MOBILITY IN YOUNG ADULTHOOD 1992 9 Advances in Group Processes
108. Brothers' Keepers: Situating Kinship Relations in Broader Networks of Social Support Barry Wellman Scot Wortley 10.2307/1389119 Sociological Perspectives Sociological Perspectives 0731-1214 1533-8673 32 3 1989 SAGE Publications
109. Different strokes from different folks: community ties and social support B Wellman S Wortley American Journal of Sociology 96 1990
110. Tenure type landscapes and housing market change: A geographical perspective on neoliberalization in Sweden T Wimark E K Andersson B Malmberg Housing Studies 35 2020
111. Behavioral aspects of the decision to migrate J Wolpert Papers in Regional Science 15 1965
112. Innovation Diffusion in Heterogeneous Populations: Contagion, Social Influence, and Social Learning H P Young 10.1257/aer.99.5.1899 American Economic Review American Economic Review 0002-8282 99 5 2009 American Economic Association