Exploring Life-Course Trajectories in Local Spatial Contexts Across Sweden

Abstract

This article explores typical life-course trajectories based on annual observations of educational participation, employment, and establishing a family from age sixteen to age thirty. Using latent class analysis, we identify seven different trajectory classes that capture the different life courses experienced by individuals born in 1986. Examples of trajectory classes are (1) an early partner and childbearing trajectory; (2) a trajectory that mixes employment and a long postsecondary education into the later twenties; and (3) a trajectory involving low activity, very little employment, very little postsecondary education, and not starting a family. The classes identified correspond closely to trajectories found in earlier qualitative studies using life-history interviews, but in contrast to these studies that each encompass a few dozen individuals or less, our approach identifies trajectories for the individuals of an entire birth cohort. This allows for analysis of the geographical distribution of trajectories across regions, municipality types, and neighborhoods. Individuals following long postsecondary education trajectories were heavily concentrated in metropolitan areas and university towns. At the same age, individuals following early childbearing trajectories were concentrated instead in peripheral, rural areas. Individuals from nonmetropolitan areas also tend to follow more gender-polarized trajectories. Moreover, we find that there is more trajectory-based segregation at age thirty than at age fifteen. Theoretically, our study gives support to the idea that places are structured on the basis of life-course trajectories. Local context influences how individuals are linked into different trajectories and, at the same time, the spatial sorting of trajectories will shape local contexts.

基于对16至30岁群体的教育、就业和组建家庭的年际观察, 本文探索了典型人生轨迹。通过潜在类型分析, 我们确定七个类型的轨迹, 记录了1986年出生群体的不同生活历程。轨迹类型包括：（1）早期伴侣和生育轨迹；（2）20至29岁后期的就业和长期大学教育轨迹；（3）低活力、非常少的就业、非常低的大学教育水平、未组建家庭。这些类型, 与基于生活史采访的早期定性研究确定的轨迹, 密切对应。但是, 与数十人或更少的研究相比, 我们确定了整个同龄群体的个人轨迹。这使得我们能分析跨区域、跨城市类型和跨社区的轨迹地理分布。拥有长期大学教育轨迹的个人, 主要集中在大都市和大学城。在同一年龄段, 拥有早期生育轨迹的个人, 集中在周边农村地区。来自非大都市地区的个人轨迹, 在性别上更加极化。此外, 30岁比15岁有更高的轨迹隔离程度。从理论上讲, 我们的研究支持了这样一种观点：场所的构建是基于生活历程轨迹。局部环境影响了个人如何关联到不同的轨迹。同时, 轨迹的空间排序能塑造局部环境。

Este artículo explora las trayectorias típicas del curso de la vida basadas en observaciones anuales de la participación educativa, el empleo y el establecimiento de una familia desde los dieciséis hasta los treinta años. Usando el análisis de clases latentes, identificamos siete diferentes clases de trayectorias que captan los diferentes cursos de la vida experimentados por los individuos nacidos en 1986. Ejemplos de clases de trayectorias son (1) una trayectoria de pareja y maternidad temprana; (2) una trayectoria que mezcla el empleo con una larga educación postsecundaria hasta finales de los veinte; y (3) una trayectoria en la que se incluye poca actividad, muy poco empleo, muy poca educación postsecundaria y la ausencia de formar una familia. Las clases identificadas corresponden estrechamente con las trayectorias halladas en estudios cualitativos anteriores que aplicaron entrevistas relacionadas con historias de vida, pero en contraste con estos estudios, cada uno de los cuales abarca unas pocas docenas de individuos o menos, nuestro enfoque identifica trayectorias de los individuos de toda una cohorte de nacimientos. Esto permite el análisis de la distribución geográfica de las trayectorias a través de las regiones, tipos de municipalidades y vecindarios. Los individuos que discurrieron largas trayectorias de educación postsecundaria aparecieron altamente concentrados en áreas metropolitanas y ciudades universitarias. Con la misma edad, los individuos que siguieron trayectorias de maternidad temprana se concentraron, por el contrario, en áreas periféricas y rurales. Los individuos de áreas no metropolitanas registraron también trayectorias más polarizadas por género. Aún más, encontramos que hay una mayor tendencia a la trayectoria basada en segregación a la edad de treinta que a la de quince años. Teóricamente, nuestro estudio proporciona soporte a la idea de que los lugares se estructuran con base en las trayectorias vitales. El contexto local influye el modo como los individuos se vinculan a diferentes trayectorias y, al mismo tiempo, la ordenación espacial de las trayectorias configura los contextos locales.

Key Words:

• early adulthood
• geographical context
• latent class
• life course
• Sweden
• trajectories

关键词：:

• 成年初期
• 地理背景
• 潜在类型
• 生活历程
• 瑞典
• 轨迹。

Palabras clave::

• adultez temprana
• clase latente
• contexto geográfico
• curso de vida
• Suecia
• trayectorias

In this article we propose that an identification of life-course trajectories in longitudinal, individual-level, geocoded register data can be helpful for assessing urban and regional patterns in ways of life. We focus on young adults, and on how their lives are structured in terms of life-course trajectories with respect to education, employment, and establishing a family. Our approach can be seen as a merging of a traditional view of regions as displaying specific genres-de-vie—dating back to Vidal de la Blache (1911), but also present in the works of, for example, Pred (1984) and Massey (1994)—with a more recently emerging biographical paradigm in human geography that puts individual life courses at the center of geographical analysis. In fact, Massey demonstrates that genres-de-vie can be transformed into a tool for feminist analysis. This can be seen as a sociospatial dialectic approach—as developed by Knox and Pinch (2010) but originally from Soja (1980)—to trajectories, where they not only influence local social contexts, but are also influenced by them.

A strength of the life-course approach is that it bridges the gap between qualitative and quantitative methods (Elder, Kirkpatrick Johnson, and Crosnoe 2003). As evidenced in the review of earlier studies that follows, the life-course approach can be applied both in the analysis of longitudinal data and in studies based on in-depth interviews. In this article, a starting point will be five earlier qualitative studies with a focus on the life plans of young adults carried out in different geographical contexts by Swedish sociology-of-education researchers. These studies, to a large extent, are guided by Bourdieu’s (1977) view that typical, socially recognized trajectories are based on a specific habitus that “could be considered as a subjective but not individual system of internalized structures, schemes of perception, conception, and action common to all members of the same group or class” (86). The habitus is acquired by the individual in a processes of socialization that takes place in the family, in peer networks, in schools, in local institutions, and in the neighborhood. Using latent class analysis (LCA), we explore to what extent typical life courses identified in these qualitative studies can also be identified in longitudinal register data and if these trajectories can be understood using Bourdieu’s concept of habitus.

The life phase we focus on is entry into adulthood, ages between sixteen and thirty years old. This is a demographically eventful period in terms of education, starting work, leaving the family home, intense residential mobility, early career steps, and establishing a family. It is also a life phase that has been in focus for a number of qualitative, life-course-oriented studies. Therefore, we will be able to compare the trajectories that we find in register data with the results of interview-based studies.

Inspired by Simonsen (2016), who argued that “places are meeting points, moments or conjunctures, where social practices and trajectories, spatial narratives and moving or fixed materialities meet up and form configurations that are continuously under transformation and negotiation” (22), we proceed to an analysis of spatial variation in the representation of different life-course trajectories. Such an analysis can be compared to the analysis of segregation patterns using indicators of ethnicity, socioeconomic status, or education. Using the principal factor procedure and a multitude of variables, studies in factor ecology have grouped together areas displaying the most similarities (Murdie 1969). This approach to classifying data has its roots in the Chicago School of Sociology in the 1920s (McKenzie, Park, and Burgess 1967), but has lately seen a revival in that multiscalar and composite spatial contexts can be created using the enhanced speed and capacity of today’s computers (Andersson and Malmberg 2018; Petrović 2020).

Shifting the focus to spatial sorting according to life-course trajectories has advantages. Trajectories give a composite view of a person’s life course, and can be said to reflect an individual’s direction in life. This is interesting in relation to theories that consider social class and the local spatial context to have a structuring influence on the life of individuals that goes beyond their socioeconomic situation at a particular point in time. It has been suggested that local cognitive and motivational structures differ between places, and thereby also contribute to spatial variation in lifestyles (Simonsen 1991; Forsberg 1998, 2001, 2019). It could be argued that factor ecology seeks to capture such spatial variation, albeit in a very abstract way. Ethnographic studies of particular places, in contrast, can provide an in-depth, comprehensive understanding of local spatial contexts (e.g., Waara 1996; Palme 2008; Jonsson 2010), including local cognitive and motivational structures. The relation of a particular place is often discussed and described in relation to other places and through a time perspective, with a Bourdieusian lens on young individuals’ educational trajectory. For example, many ethnographic studies elaborate on differences between the lifestyle in a rural area and a metropolitan area (e.g., Stockholm in Sweden vs. a rural or peripheral area; Forsberg 2019). A limitation of ethnographic studies, however, is that they do not describe a large variety of contexts throughout a country. We do not argue that we will be able to bridge the gap between ethnographic place studies and the factorial ecology approach. Rather, we are interested in exploring if an analysis of variation in the spatial representation of life-course trajectories can help to shed light on processes of spatial differentiation and spatial polarization of lifestyles. In particular, we are interested in if spatial patterns in trajectory composition can give support to ideas about spatial variation in local cognitive and motivational structures, the systems of durable, transposable dispositions (through which we act, think, judge, and acquire specific tastes) that Bourdieu (1977, 19) called habitus.

Our analysis of spatial variation in the representation of life-course trajectories will look at spatial sorting across different spatial scales. First, we look at the distribution of trajectory classes across municipality types, relying on a classification made by the Swedish Association of Local Authorities and Regions (SALAR). This classification into metropolitan cities, metropolitan suburbs, major cities, suburbs to major cities, and different categories of more peripheral municipalities including commuting, manufacturing, tourism, and low-density municipalities captures differences along the metropolitan and nonmetropolitan dimension and is based on municipality size, sectoral composition, population density, and municipality location. Thus, the SALAR classification captures a relatively nuanced differentiation of local contexts, and it has been used to good effect in different types of studies. Second, we look at variation across Swedish NUTS-3 regions and, finally, at small-scale variation using an individualized neighborhood approach (Östh, Malmberg, and Andersson 2014). Thus, the aim of this article is to explore the segmentation of young adults’ life-course trajectories using longitudinal register data, and to analyze how this trajectory segmentation is linked to spatial differentiation. Recently, Centner (2021) made a plea for more ambitious spatial theorizing in a Nordic context. We would argue that identifying and mapping trajectories can provide a starting point for a theoretical understanding of the processes that shape urban and rural contexts, and that this article, therefore, can be seen as one step toward empirically grounded theorizations.

Theoretical Background

Life-course literature has discussed whether life courses have become less standardized over time (Macmillan 2005). It can be difficult to determine trends in this regard. Based on both the results of qualitative studies of life histories and theories about the reproduction of social classes, however, the starting point for this article is that clear differentiation in life courses will be highlighted when trajectories are identified in register data. This differentiation reflects existing social inequalities, as well as an occupational structure that divides the workforce, for instance, into categories with different status, power, and skill requirements. At the same time, it can be argued that geographical factors can contribute to life-course differentiation. Based on a spatial division of labor, different locations will provide structures in terms of opportunity that favor certain types of career over others. There is also a role for lifestyle factors, though. Following a specific life course is a way of expressing identity and demonstrating allegiance to particular norms. It is therefore conceivable that local patterns develop over time in terms of what constitutes an ideal life course (Pred 1984; Massey 1994).

Spatial differences in opportunity structures can clearly influence the life courses into which young adults are directed (Andersson 2001). In metropolitan areas a broad range of education is available, and the occupational structure is highly diversified. Metropolitan areas also have a concentration of upper-middle-class families who will encourage and influence their children’s ambitions in terms of careers (Andersson, Abramsson, and Malmberg 2021). In contrast, peripheral areas are characterized by less occupational diversity, especially in high-status occupations. They also offer few local options for higher education. Instead, in many cases, they are characterized by sectoral specialization in terms of agriculture or forestry, mining, specific aspects of manufacturing, and tourism, for example. On the other hand, the opposite is true with respect to the ubiquity of work in schools, nursing, and care (Brandén 2013). Moreover, in the Swedish context, peripheral areas have fewer upper-middle-class families (Kawalerowicz and Malmberg 2021). These differences are likely to influence how young adults who grow up in a geographical context are distributed across life-course trajectories.

Furthermore, irrespective of where people grow up, the life-course trajectory they follow during young adulthood can be expected to influence their geographical career. Taking part in tertiary education will require them to migrate to university and college towns, and the labor market for university graduates will be concentrated in larger cities. On the other hand, affordable housing for individuals who establish a family early will be easier to find in less central locations. It might also be the case that people actively look for residential areas with high concentrations of people in the same trajectory as themselves. These factors will lead to spatial sorting based on a person’s life-course trajectory.

Thus, it is not only the case that spatial context influences life-course trajectories. The geographical sorting of these trajectories will also, in turn, shape spatial contexts. A long-standing idea in human geography suggests that spatial contexts can develop special or even unique characteristics. This article suggests that, by first identifying life-course trajectories in register data, and then exploring how local contexts are constituted as assemblages of life-course trajectories, it becomes possible to provide an analysis of Swedish regions and urban areas based on genres-de-vie. We argue that this is possible because the life-course trajectories that can be identified statistically clearly overlap with trajectories that can be demonstrated through qualitative life-history interviews. Qualitative results therefore help to interpret the trajectories and allow them to be seen as an expression of life plans. Moreover, life-course trajectories are linked to specific local understandings of goals in life, and of normative ideas about how lives should be structured. They also influence the kind of experiences and standards that individuals will acquire through life. This implies that the genres-de-vie of particular places can be hypothesized to reflect the types of life-course trajectory represented there.

In association with geographical contexts and gender structures, some studies have analyzed partnerships and establishing a family (Forsberg 1998, 2001; Haandrikman, Webster, and Duvander 2021). Forsberg’s (2021) article notes, “Gender contracts are unwritten rules that regulate relations between sexes, and re-create and reform relations as everyday actions within the framework of these local structures. Together these various local contracts construct a regional structure” (161). Forsberg (1998) identified three gender contracts in principle. The first is the traditional gender contract, common in forestry and industrial areas, with high gender labor market segregation and in which families play an important role. The second is the modernized gender contract in metropolitan areas, with integrated labor markets and more women in public life. Regions with modernized gender contracts are dominated by the service sector. According to Forsberg, these regions are sometimes ”escalator regions” for women. The third, involving nontraditional gender contracts, is prevalent in some peripheral and rural areas in which there is a traditional economic base but where gender relations are more equal (in politics, the labor market and everyday life) than in other, similar regions (Forsberg 1998).

More recent than the work by Forsberg (1998, 2001) is the work of Haandrikman, Webster, and Duvander (2021), who tested regions as a geographical unit for gender contracts. Using a geographical multiscalar approach, they argued that local gender contracts vary substantially, and that there are no dominant regional gender contracts. They highlight, however, local variance in gender contracts (Haandrikman, Webster, and Duvander 2021). Nevertheless, even with their improved multiscalar approach and several measures of gender relations in a cluster analysis, the spatial patterns they produce do not, in essence, contradict the coarser spatial patterns in Forsberg (1998).

Life-Course Trajectories in Early Adulthood, Identified in Swedish Qualitative Studies

A fundamental difference between longitudinal register data and life-course information derived from interviews is that register data, as such, is silent about motivations, intentions, and personal influences that have played a role in individual life stories. Nonetheless, both interview narratives and the records found in register data refer to life-course trajectories in a shared spatiotemporal life world (Hägerstrand 1978). Much of qualitative, Swedish, life-course research has been published in research monographs, often available only in Swedish. In what follows, the results of five such studies with a focus on young adult trajectories are summarized. The studies have been selected to have examples from both metropolitan and nonmetropolitan settings, and to include individuals from both affluent contexts, middle-income contexts, and low-income contexts.

Lundqvist interviewed students from two classes in an upper secondary school in Stockholm for her PhD dissertation (Lundqvist 2010; Lundqvist and Olsson 2012) probably, given the ethnicity of the students, schools in one of Stockholm’s migrant-dense suburbs. The interview sample consists, on the one hand, of students in a program with an orientation toward retailing and office work (RO) and, on the other hand, students in a program oriented toward continued education in social sciences and economics (SE). Both groups have a general idea that further education after upper secondary school is of value, but the educational plans for the SE group are more concrete and specific, whereas students in the RO group have less detailed plans. Their immediate future is more linked to getting a job. These different orientations are to some extent linked to differences in parental background. Parents of the SE students tend to be more highly educated, and also tend to be more determined that their children should have a tertiary education. Parents of the RO students acknowledge the importance of education but, possibly linked to their own lack of tertiary education, are less able to provide their children with more definitive suggestions.

Palme’s (2008) dissertation is based on interviews with thirteen individuals from an affluent elite area north of central Stockholm, and it is based on Bourdieu’s theory. The interviewees were very much aware of what was expected of them in terms of behavior and career plans. They would not necessarily go directly into tertiary education after graduating from upper secondary school, and most of them had plans for a gap year. At least until they were twenty-five years old, though, their life would be dominated by education. Both parents and neighbors had elite occupations: doctors, lawyers, civil engineers, and civil servants. The interviewees can be characterized as resourceful individuals with well-developed verbal skills that allowed them to adapt the way they spoke to different circumstances. They greatly valued “culture” and seemed to have a sense of culture that could help them navigate education and careers related to the media, cultural communication, or the arts. Their current abilities not only reflected input from parents, but also from parental networks of neighbors and friends.

Lindblad’s (2016) dissertation focuses on migrant youth. In this case, twenty individuals who had failed to graduate from upper secondary schools, representing different spatial contexts, were selected for interviews. The selected individuals were from residential areas dominated by rental dwellings. Three of the selected areas were characterized by high unemployment, low levels of education, high poverty rates, and high proportions of foreign-born residents.

Few of the interviewees had moved away from the area where they had grown up, and many were still in their twenties, still living with their parents. The interviews indicate that the areas where they lived had a high concentration of migrants, and that the Swedish language was less well represented there. It was also suggested that the environment had not been helpful, and had offered opportunities to become involved in criminal activities, for example. They had had substantial problems establishing themselves in the labor market, and they were also participating in a variety of programs aimed at supplementing their education. In general, they emphasized the important role played by their family, but at the same time family obligations in the form of work can be seen as a factor behind their difficulties in finishing school. An interesting contrast is the group interviewed by Palme. This group was characterized by very clearly laid-out life plans, whereas Lindblad’s interviewees reported that they had difficulties finding their way.

Jonsson’s (2010) dissertation studied more than thirty young people in a small-to medium-sized Swedish city. His theoretical tool is life-mode analysis, originally proposed by (Hojrup 1983). Life-mode analysis is a class-based scheme that distinguishes between career-oriented life modes and wage-labor-oriented life modes. Jonsson added a gender dimension: female career-oriented, male career-oriented, male wage-labor-oriented, and female family-oriented life modes. Both female and male career-oriented individuals had higher education as an important goal in life, but differed in how education fit into their life plans. For female career-oriented individuals, it was important to excel at school, whereas male career-oriented individuals in Jonsson’s sample did not strive for the highest grades. Women in this group also expressed less satisfaction than men in terms of the town in which they were living. In addition, men with a wage-labor orientation were comfortable with the place they were living in, and were aiming for a relatively rapid transition to full adulthood, with the acquisition of a single-family home as an important goal in life. For individuals in the female family-oriented life mode, establishing a family stood out as a life project that was at least as important as education was for career-oriented females. Equally, there were strong contrasts between the groups in terms of their geographical orientation and in patterns of social interaction.

Waara’s (1996, 1998) dissertation involved interviews with twenty-six students in a small- to medium-sized city in northern Sweden. They attended a three-year program with no occupational orientation, implying that the students were expected to move into tertiary education. Waara used group interviews and his work was not based on a preconceived theoretical frame. Instead, he used a grounded theory approach, and the class perspective was not emphasized in the study. Moreover, group interviews might move the focus away from how ideas about the life course are shaped by parental influences toward an interest in shared ideas about possible trajectories in young adulthood. One finding is that there was a shared understanding of the different phases the students would go through in the years ahead. First, this involved a phase of building competence through education and work, then a phase of establishing a family, and finally a settling-down phase. The geographical dimension is important in the analysis. How did the students relate to and evaluate their local community? What were their plans for moving to other places, and the extent to which they had an interest in returning to their home areas? This analysis also emphasizes the gender dimensions. What gender roles were represented in their home area? How did ideas about gender influence the future plans of the students, and how did these conceptions differ between boys and girls?

Two interrelated dimensions seem to be the most important in these studies: education and establishing a family. Geography is relevant, too, with postsecondary education a more important option in metropolitan areas, whereas establishing a family is a more central concern in nonmetropolitan areas. In addition, outside metropolitan areas, postsecondary education is seen as an option associated with leaving the place where a person has grown up.

Earlier Quantitative Trajectory Research

Life-course trajectory research in human geography, concerning neighborhoods, residential mobility, and housing, has long been recognized as a fascinating, telling research tool and approach (Elder 1985; Damhuis et al. 2019). At the same time, life-course trajectory research is currently becoming important in human geography and related research fields (Bäckman and Nilsson 2011; Virtanen et al. 2011; van Ham et al. 2014; de Vuijst, van Ham, and Kleinhans 2016; Almquist and Brännström 2018; Ilmakunnas and Moisio 2019; Vogiazides and Chihaya 2020; Sugahara and Nordvik 2022). Studies use LCA and sequence analysis (SA), but spatial dimensions are included less frequently (see, however, Gruebner et al. 2016). This section discusses a number of studies that have used LCA and SA.

Our study is, in many ways, similar to a study using LCA by Amato et al. (2008) on the precursors for young women’s pathways in terms of establishing a family. Their study, using the National Longitudinal Study of Adolescent Health 1995 to 2002, studied young women in the United States in three waves, when they were between eighteen and twenty-three years of age. Precursors from the study of these women were subjected to a factor analysis in which three factors were extracted: (1) personal and social resources; (2) socioeconomic resources in the family and adolescent academic achievement; and (3) conservative values and behavior. The issue was then whether the values and parental background precursors in these factors could help explain the pathways of the women in question. For the pathways, the authors used five indicators: cohabitation, marriage, births, education, and income from employment. Examples from the seven types of trajectory include one in which women went to college and did not establish a family, another in which women cohabited without children, another involving single mothers, and one in which women were inactive. Finally, they tested for associations between the life paths and the background value factors. (Amato et al. 2008).

As a response to the idea that transitions to adulthood could show more complex patterns with time, Elzinga and Liefbroer (2007) analyzed the life-course trajectories of young adults in nineteen European countries. They used SA and concluded that the family-life trajectories of young adults had not become more complex or turbulent. They found, however, that the life courses were less similar to one another, and that there were more types of family trajectory than before. On the other hand, the geographical analysis, which in this case was country-wide, did not show clear differences (Elzinga and Liefbroer 2007).

Lorentzen et al. (2019) analyzed pathways to adulthood using SA of the school-to-work transition in Finland, Norway, and Sweden. Because the Nordic welfare states are fairly similar, only small differences were detected in the school-to-work trajectories. One difference involved a fairly strong link between early parenthood and workforce exclusion for Finnish women. The authors stated that school-to-work trajectories and related events have generally become more diversified. According to the authors, events that have tended to take place in a certain order, such as “leaving school, entering the workforce, leaving one’s family, marrying and having children” (Lorentzen et al. 2019, 1286) now take place in a different order. In line with our study, they found for individuals born in Sweden in 1975 that the most common sequencing of school-to-work was short education into work (see below, trajectory Class 5 in our study). Moreover, this trajectory was dominated by men (35 percent, with women at 11 percent). The second most common (and the most common sequencing for individuals born in Norway and Finland in 1977) was medium education into work (Lorentzen et al. 2019). The authors also indicated a trajectory of long period of education into work, in which individuals remain in education for a long time, but interrupted by periods of work until their mid- to late twenties (compare our trajectory Class 7). Finally, and in line with our study, they also found an exclusion trajectory, which involves 13 to 15 percent of Swedes and Norwegians but fewer Finns, although Finns appear to be further away from the labor market (compare our trajectory Class 3, low activity).

Methods and Data

This study first identifies latent classes in trajectory data and then describes these trajectory classes using in a life-course perspective. Second, these trajectory classes are mapped to highlight possible geographical patterns.

Many studies use SA to identify life-course patterns in longitudinal data (see, e.g., Svensson et al. 2015; de Vuijst, van Ham, and Kleinhans 2017), whereas other studies use LCA for this purpose (Virtanen et al. 2011; Almquist and Brännström 2018; Ilmakunnas and Moisio 2019). According to Piccarreta and Studer (2019), SA does not make “assumption on the data-generating process or any predefined judgment about the relevant features of the life course” (3). Furthermore, in SA “changes within sequences might have medium-term effects on future evolution,” and does not “simplify how past experience impacts the trajectory’s subsequent unfolding” (3). This flexibility allows SA to identify life-course patterns that results from those complex interdependencies that are often emphasized in life-course theory (Spini et al. 2017; Bernardi, Huinink, and Settersten 2019). LCA, in contrast, is a model-based approach. That is, it is based on assumptions that there is an underlying data generating process, the parameters of which can be estimated from the observed data (Henry and Muthén 2010). In a sense, LCA is a more restricted approach than SA, and it could be seen as less appropriate if the data to be analyzed cannot be assumed to be influenced by underlying latent classes. In this case, both existing theory and the results from existing qualitative studies suggest the possibility that young adult trajectories are influenced by an underlying biographical script or biographical schema; that is, a “socially objectivated oral or literary scheme or model for the course of an individual’s life” (Luckmann 1991, 163). The use of LCA is thus warranted. Moreover, an argument for using LCA is that it allows trajectories to be identified using indicators from different life domains. Doing this with SA requires the use of a multichannel approach (Gauthier et al. 2010), which is an interesting option. Compared to LCA, which is a standard method, especially in psychology and related fields, multichannel sequence analysis is still at a relatively early stage in its development. Thus, because the main aim of this article is not to compare different methods for life-course trajectory analysis (see Barban and Billari 2012), we have chosen to use LCA.

Data

The analysis focuses on the 1986 birth cohort, which is followed from age sixteen to age thirty. The data consist of a collection of longitudinal register data, released for research through Statistics Sweden’s platform for access to microdata, MONA. This collection encompasses the years 1990 through 2016, and contains detailed individual-level information on family status, household composition, country of birth, education, employment, different components of income, and different types of social benefit.

From these data, we constructed three life-course indicators: in-education, employed, and family. The in-education indicator is based on the variable study participation (StudDelt) contained in the annual LISA register (Longitudinal Integration Database for Health Insurance and Labor Market Studies). The in-education variable is coded with 1 for taking part in any education and 0 for not doing so.

The employed indicator is based on reported earned income (Forvink), where an income above 40 percent of the median earned income has been used as the cutoff value for employment. The motivation being the (un)employment variable at Statistics Sweden counts persons having employment in November as employed, which is crude for our purposes. The family indicator is coded as 1 if the family type involves a husband and wife with children, or cohabiting partners with children, where the individual is a married or cohabiting partner. This implies that single parents do not fall into the family category, and makes this an indicator of the normative family. We acknowledge that this is a limitation because union dissolution and repartnering cannot be observed and could be of interest to study (Jalovaara and Fasang 2017). In this explorative study, though, neither demographic, educational, nor working life indicators are in any way fully represented. Rather, the approach is broad and the result will be presented in geographical terms. The three binary variables registered for fifteen years make forty-five indicators for the LCA. In addition to the indicator variables, we also use background variables for the individuals to assess ethnic, gender, and parental background characteristics of different latent classes (see Table 1).

Table 1. Descriptive statistics, variables for latent class analysis and background variables

Latent class analysis variable	Observations	Minimum	Maximum
In education	93,355	11%	96%
Employed, > 40% of median income from work	93,355	6%	84%
Married/partnered with child/ren	93,355	0.02%	37%
Background variable	Proportion
Geographical background of parents	Sweden (78.771)	Europe (15.187)	Outside Europe (8.293)
Gender	Male (52.529)	Female (49.702)
Single-mother household (2001)	Yes (22.029)	No (80.222)
Social allowance household (2001)	Yes (7.435)	No (94.816)
Employed parents (2001)	Two (59.473)	One (35.554)	None (9.224)
Parents with tertiary education (2001)	Two (60.966)	One (26.150)	None (15.135)

The 1990 through 2016 database includes a total of 147,826 unique individuals born in 1986. Of these, there are full records for 93,355 individuals for the 2002 to 2016 period. Reasons for an absence of full records could involve emigration or death before 2016, or immigration after 2002. In other words, the study population consists of individuals who lived in Sweden without interruption from 2002 to 2016.

Number of Latent Classes

How many different classes are needed to describe the trajectories followed by the 1986 cohort from sixteen years of age to age thirty? We ultimately agreed on seven different classes. To arrive at this number, we tested models with two, four, six, seven, eight, ten, twelve, and twenty different latent classes. Going from two classes to four classes greatly reduces the Akaike’s information criterion (AIC), and an increase from four to six also appreciably reduces the AIC (Appendix). From six to eight classes the reduction in AIC is smaller, and going from eight to more classes yields even lower changes in the AIC, indicating that there is less to be gained from increasing the number of classes (see Appendix). As a result, we chose to work with seven latent classes, called trajectory classes. With this number of classes, the posterior probabilities are all above 0.9, and we deem the number to be manageable for further analysis and readability.

Results

Youth Trajectories Identified in Register Data

The seven trajectory classes we have identified are presented in Figure 1 and Table 2. In Figure 1, the probabilities of being in employment, being in education, and having a family, for the seven trajectory classes are presented as graphs or so-called trajectory profiles. Table 2 provides shorthand descriptions of the profiles and reports the size of the different trajectory classes.

Figure 1 Trajectory profiles for latent classes.

Table 2. Latent trajectory classes identified

Trajectory class	Description	No. of individuals	Proportion of cohort	Average posterior probability (most likely LC)
1	Postsecondary education after a gap year	17,382	19.3%	91.4%
2	Secondary education, employment, then establishing a family	9,348	10.4%	93.5%
3	Low activity	7,551	8.4%	94.8%
4	Establishing a family early	6,904	7.7%	94.2%
5	Employment only	24,564	27.3%	94.9%
6	Postsecondary education, work, family	10,624	11.8%	91.8%
7	Mix of postsecondary education and work	13,738	15.2%	90.3%
Total		90,111	100.0%

Note: Latent class (LC) based on 93,355 observations, but for some of these information on covariates is lacking.

The largest latent class is trajectory Class 5, to which 27 percent of the cohort were assigned. Figure 1 indicates that this class is characterized by high levels of employment after the end of secondary school, with very little participation in postsecondary education, and very few in this class establish families in the studied ages. We label this class employment only.

The second largest class is trajectory Class 1 with close to 20 percent of the cohort. Trajectory Class 1 starts postsecondary education relatively soon after the age of eighteen, and finishes before or around age twenty-five, after which they are in employment. The label chosen for trajectory Class 1 is postsecondary education after a gap year. Trajectory Class 7, the third largest class, also takes part in postsecondary education, but start their postsecondary education somewhat later, and their studies go on for longer. In addition, there is around a 50 percent probability that they will be employed while they are studying. We use the label mix of postsecondary education and work. Together, trajectory Class 5, trajectory Class 1, and trajectory Class 7 make up more than 60 percent of the cohort.

The next largest, trajectory Classes 6 and 2, both constitute around 10 percent of the cohort, and both are characterized by establishing a family. Trajectory Class 6 follows what could be called a normative pattern, with employment and postsecondary education after secondary school, a transition to employment, and then starting a family after the age of twenty-five. Those in trajectory Class 2 make the transition to starting a family somewhat earlier, after five to six years of employment, and they do not participate in postsecondary education (see Figure 1).

Trajectory Class 3 and trajectory Class 4 are the smallest, each with around 8 percent of the cohort. Class 3 follows a trajectory known in labor-market-oriented studies as no employment, education, or training (NEET; Lorentzen et al. 2019). In this case, no family could be added to this list although about 4 percent (varying with age) are single parents in trajectory Class 3, labeled low activity. Because trajectory Class 3 is a socially vulnerable group, it could be of interest to explore in more detail what subtypes of trajectories are found in this group. We note, however, that the group is not consisting of single parents, which we attribute to the Swedish welfare policy protecting single parents.

Trajectory Class 4, involving childbearing before twenty-five years of age, also represents an outlier to some extent in the Swedish context, where having children is postponed until well after twenty-five years of age for the majority of people. It is labeled establishing a family early. The marginality of trajectory Class 3 is evidenced by the high proportion of this group receiving a social allowance at thirty years of age (see Table 3). The only other class worth mentioning in terms of social allowance is trajectory Class 4, early family, characterized by low levels of education and about 60 percent in employment.

Table 3. Proportion with social allowance at age 30

	Trajectory class
	1	2	3	4	5	6	7
	Post- secondary education after gap year	Secondary education, employment, then family	Low activity	Establish family early	Employment only	Postsecondary education, work, family	Mixed postsecondary education and work	Overall
Proportion with social allowance at age 30	0.4%	0.2%	18.4%	2.4%	0.7%	0.2%	1.5%	2.3%

Table 4 shows the demographic characteristics of the trajectory classes. The column on the far right shows that women are overrepresented in trajectory Class 4 (early childbearing) and trajectory Class 6 (establishing a family after postsecondary education), but underrepresented in trajectory Class 2 (establishing a family around age twenty-five) and in trajectory Class 5 (employment only), compared with the bottom row. A characteristic of these trajectory classes is therefore that they are gender-polarized. On the other hand, trajectory Class 1 and trajectory Class 7 (postsecondary education without establishing a family) have a near equal representation of men and women, and therefore involve a gender-neutral trajectory.

Table 4. Demographic characteristics of trajectory classes

Trajectory class	Description	Swedish background (%)	European background (%)	Non-European background (%)	Women (%)
1	Postsecondary education, work	76.8	13.9	9.4	51.6
2	Work, then starting a family	84.5	11.9	3.5	42.3
3	Low activity	67.1	20.7	12.2	44.1
4	Establishing a family early	77.4	15.1	7.5	78.0
5	Employment only	81.5	13.2	5.4	33.3
6	Postsecondary education, work, family	80.8	13.0	6.3	70.3
7	Mixed postsecondary education and work	77.2	14.2	8.6	52.9
	Overall proportion	78.6	14.1	7.3	49.5

Note: Bold represents overrepresentation, italics represents underrepresentation.

The first three columns of Table 4 show, for each trajectory class, the proportion of individuals with a Swedish background, a European background, and a non-European background. One observation here is that individuals with a Swedish background are overrepresented in three of the gender-polarized classes: trajectory Class 2, trajectory Class 5, and trajectory Class 6 (shown in bold, first column). Trajectory Class 2 shows the largest proportion of Swedish-born individuals with Swedish parents, at 84.5 percent. On the other hand, individuals with a Swedish background are underrepresented in trajectory Class 3 (low activity). In contrast, individuals with a non-European background are overrepresented in the gender-neutral trajectory classes and underrepresented in the gender-polarized classes, with the exception of trajectory Class 4 (establishing a family early). Individuals with a non-European background are also heavily overrepresented in trajectory Class 3 (low activity), as are individuals with a European background. That Swedish background individuals are overrepresented in gender-polarized trajectories, and migrant background individuals are overrepresented in gender-neutral trajectories stands out as a finding that has not been reported before, but it resonates with the results from qualitative studies of migrant background and Swedish background trajectories, respectively. The overrepresentation of migrant-background individuals in trajectory Class 3 is in line with earlier studies.

Table 5 provides information about the socioeconomic characteristics of the different trajectory classes using family and parental background variables. It shows that parental background strongly influences the trajectory to which individuals are allocated. Thus, individuals who have parents with postsecondary education have a strong tendency to follow trajectory Classes 1 and 7. Overall, 41 percent of the individuals in our sample have at least one parent with postsecondary education, but the proportion is much higher in Classes 1 and 7. In contrast, individuals who follow trajectory Classes 2 and 4, establishing a family relatively early, are less likely to have parents with tertiary education. These are well-known features of individuals with highly educated parents and individuals whose parents lack tertiary education, respectively.

Table 5. Socioeconomic characteristics of individuals in trajectory classes at age 15

		Family/parental background at age 15
Trajectory class	Description	Proportion from postsecondary education households	Proportion from single- mother households	Proportion from households without employment	Proportion from households receiving a social allowance
1	Postsecondary education, work	57.7%	16.1%	5.9%	4.2%
2	Work, then establishing a family	22.8%	19.6%	6.5%	5.1%
3	Low activity	33.8%	31.1%	17.2%	13.9%
4	Establishing a family early	24.3%	25.1%	13.1%	11.3%
5	Employment only	29.7%	21.9%	6.9%	5.3%
6	Postsecondary education, work, family	47.8%	17.7%	6.2%	5.0%
7	Mixed postsecondary education and work	61.5%	20.2%	6.5%	4.3%
	Overall proportion	41.3%	20.8%	7.9%	6.0%

Note: Bold represents overrepresentation, italics represents underrepresentation.

The last three columns of Table 5 show that individuals from households with a single mother, households without employment, and households receiving a social allowance are overrepresented in the low activity type of trajectory (trajectory Class 3). They are also overrepresented in trajectory Class 4, involving individuals who establish a family early. In both cases the overrepresentation is strongest for households without employment and those receiving a social allowance. Again, this is a well-known pattern from research on young adults. Overrepresentation is more modest for individuals from single-mother households. Individuals from single-mother households are also similar to the overall sample in other respects, for example, in relation to trajectory Class 7 (mixed postsecondary education and work) and trajectory Class 5 (employment only).

Spatial Sorting of Trajectory Classes

The spatial allocation of these trajectory classes across municipality types, across counties (NUTS-3 regions), and across individualized neighborhoods is analyzed here based on where a subject lived at age fifteen. There will also be a short discussion on how individuals are allocated in early adulthood, based on the trajectory they follow.

Types of Municipality

Table 6 shows geographical patterns in the distribution across trajectory classes. Table 6 uses the SALAR municipality-type classification (SALAR 2010). To facilitate interpretation, the columns have been arranged so that trajectory types that are overrepresented in the more metropolitan municipalities appear on the left, and trajectory types that are overrepresented in the more peripheral municipalities appear on the right. For some of the trajectories, center–periphery patterns are quite evident. For example, trajectory Class 1 and trajectory Class 7, with high rates of educational participation, involve higher proportions in Sweden’s metropolitan areas and larger cities, but lower proportions in more peripheral municipalities. A converse pattern is found for trajectory Class 4, early childbearing, where the highest proportion is in sparsely populated municipalities and the lowest proportions are in the metropolitan areas. This is also the case for trajectory Class 2, childbearing in the mid-twenties, with much higher proportions in more peripheral municipalities and lower proportions in the metropolitan regions and larger cities. In trajectory Class 5, employment only, proportions are more similar across different geographical regions, but there are higher values in suburban municipalities compared to large urban-center municipalities, to some extent reflecting a gradient in educational participation. Trajectory Class 6, which involves individuals who start a family in their late twenties, has an even more uniform representation across the different types of municipality. This is also true of trajectory Class 3 (low activity), with the exception of metropolitan municipalities, where the proportion of individuals following trajectories involving low activity is much higher.

Table 6. Distribution of sample individuals across trajectory classes, by type of municipality

Municipality type	Trajectory class							Gender difference in distribution across classes, DI
	7	1	3	5	6	2	4
	Mixed postsecondary education and work	Postsecondary education, work	Low activity	Employment only	Postsecondary education, work, family	Work, then establishing a family	Establishing a family early
Metropolitan cities	20.2%	22.7%	11.6%	24.0%	10.1%	6.0%	5.4%	17.0%
Suburbs of metropolitan cities	16.6%	21.5%	7.8%	28.4%	11.1%	9.0%	5.7%	18.5%
Major cities	16.7%	20.3%	8.3%	25.3%	12.5%	9.5%	7.4%	20.6%
Suburbs of major cities	13.3%	16.5%	7.6%	29.9%	11.4%	12.4%	9.0%	22.9%
Commuting	12.5%	17.0%	7.8%	29.3%	11.8%	12.6%	9.0%	24.4%
Tourism	13.4%	15.2%	7.8%	31.7%	10.3%	13.4%	8.3%	25.9%
Manufacturing	10.7%	16.7%	7.1%	29.3%	13.0%	13.4%	10.0%	25.6%
Low density	11.0%	15.4%	8.5%	29.4%	11.9%	13.6%	10.3%	25.8%
In dense regions	12.8%	17.6%	8.1%	27.9%	12.1%	12.8%	8.7%	23.4%
In low-density regions	13.3%	16.6%	7.8%	28.2%	12.8%	11.5%	9.8%	25.3%
Overall	15.2%	19.3%	8.4%	27.3%	11.8%	10.4%	7.7%
Proportion of women in class	53.0%	52.0%	44.0%	33.0%	70.0%	42.0%	78.0%
Proportion with Swedish background	77.2%	76.8%	67.1%	81.5%	80.8%	84.5%	77.4%

Note: Bold represents overrepresentation, italics represents underrepresentation.

At the bottom of Table 6, two auxiliary rows have been inserted: the proportion of women and the proportion of Swedish-born individuals in each trajectory class. Here, it can be observed that gender-marked trajectory classes tend to be overrepresented in more peripheral areas. On the other hand, more metropolitan municipalities are characterized by trajectory classes directed toward postsecondary education, trajectory Class 7 and trajectory Class 1, and the low activity trajectory Class 3, a trajectory with underprivileged individuals. There is also a tendency for metropolitan areas to involve latent classes where individuals with a Swedish background are underrepresented, whereas trajectories with high proportions of individuals with a Swedish background tend to be found in more peripheral municipalities, although trajectory Class 4 is an exception to this pattern.

Taken together, this suggests that Sweden is characterized by gendered patterns of polarization, linked to different geographical and ethnic contexts. On the one hand, this involves gender differentiation, which is more clearly seen in the population with a Swedish background and in more peripheral areas (Waara 1996, 220). On the other hand, it involves a relatively gender-neutral pattern in metropolitan areas, larger cities, and suburbs of metro areas, characterized by higher proportions of individuals with a migrant background. A stronger gender polarization is also evidenced by higher dissimilarity indexes in nonmetropolitan regions for the distribution of men and women across trajectory classes (Table 6).

NUTS3 Regions

The geographical distribution of trajectory classes across NUTS3 regions (Figure 2), or so called counties, is presented in Table 7. The column on the far right contains the share of the sample population living in the county in 2001. The counties have been arranged so that those with an overrepresentation of individuals in trajectories involving starting a family early appear at the lower end of the table, and counties with an overrepresentation of postsecondary education trajectories appear at the upper end of the table.

Figure 2 The twenty-one NUTS3 regions in Sweden. Source: Statistics Sweden.

Table 7. Proportion of different trajectory classes in counties

	Proportion of different trajectory classes in NUTS3 regions/counties
	Mixed postsecondary education and work	Postsecondary education, work	Low activity	Employment only	Postsecondary education, work, family	Work, then starting a family	Establishing a family early
NUTS3	7	1	3	5	6	2	4	Overall
01 Stockholms län	22.1%	21.8%	19.0%	18.7%	15.7%	14.2%	13.3%	18.6%
03 Uppsala län	4.3%	3.8%	3.2%	3.7%	3.4%	3.4%	3.6%	3.7%
12 Skåne län	12.1%	12.9%	16.5%	11.9%	13.0%	11.9%	12.9%	12.7%
23 Jämtlands län	1.9%	1.1%	1.6%	1.7%	1.5%	1.6%	1.6%	1.6%
13 Hallands län	2.9%	3.3%	2.8%	3.6%	3.6%	3.4%	2.6%	3.3%
14 Västra Götalands län	17.1%	16.2%	16.2%	17.2%	17.5%	17.2%	16.2%	16.9%
24 Västerbottens län	3.3%	2.8%	2.8%	3.1%	3.3%	3.0%	3.2%	3.0%
05 Östergötlands län	4.5%	5.1%	5.2%	4.3%	5.3%	5.1%	5.2%	4.8%
17 Värmlands län	2.8%	2.5%	3.8%	2.9%	3.1%	3.4%	3.4%	3.0%
08 Kalmar län	2.3%	2.6%	2.4%	2.6%	3.4%	3.5%	3.3%	2.8%
07 Kronobergs län	1.9%	2.1%	1.4%	2.3%	2.3%	2.5%	2.1%	2.1%
09 Gotlands län	0.7%	0.5%	0.7%	0.7%	0.7%	0.9%	0.6%	0.7%
06 Jönköpings län	3.4%	3.9%	2.4%	4.4%	4.9%	5.6%	4.9%	4.2%
04 Södermanlands län	3.0%	2.9%	2.8%	3.0%	2.8%	3.5%	3.4%	3.0%
18 Örebro län	3.0%	3.0%	3.2%	3.1%	3.0%	3.4%	3.6%	3.1%
19 Västmanlands län	2.6%	2.7%	2.9%	2.5%	2.7%	3.1%	3.6%	2.7%
20 Dalarnas län	2.8%	3.2%	3.3%	3.7%	3.2%	3.6%	3.9%	3.4%
21 Gävleborgs län	2.7%	2.6%	3.0%	3.1%	3.1%	3.2%	3.8%	3.0%
10 Blekinge län	1.6%	1.8%	1.5%	1.6%	1.8%	1.8%	2.2%	1.7%
22 Västernorrlands län	2.6%	2.6%	2.8%	2.7%	2.7%	2.7%	3.2%	2.7%
25 Norrbottens län	2.5%	2.7%	2.5%	3.1%	3.1%	3.0%	3.6%	2.9%

Note: Values in bold indicate an overrepresentation of a class in a county; values in italics represent an underrepresentation.

Table 7 shows that counties can be broadly divided into five groups. First, at the lowest end of the table there are seven counties with a clear overrepresentation of individuals who start a family very early. These are counties with some areas that are quite a long way from major urban areas. Second, there are five counties with an overrepresentation of trajectory Class 2, involving starting a family around age twenty-five. This is possibly a result of slightly stronger urban influences on rural areas. The third group consists of four counties (No. 13, 14, 24, 05) with overrepresentation of trajectory Class 6, involving starting a family after postsecondary education. These are counties with major universities or, in the case of Halland, major adjacent universities. In two to three counties belonging to the fourth group, in the uppermost part of Table 7, there is underrepresentation of trajectories involving establishing a family (No. 01, 03, 12). In Stockholm and Uppsala, and possibly Jämtland (note Jämtland with nontraditional gender contract according to Forsberg [1998]), this can be linked to the overrepresentation of postsecondary education trajectories (trajectory Classes 7 and 1). In Skåne, along with Värmland, trajectory Class 3 (low activity) is overrepresented. One reason for this could be a combination of a weak local labor market and opportunities for working in neighboring Denmark or Norway.

Individualized k = 100 Neighborhoods

Detailed spatial sorting for three selected trajectory types is shown in Figure 3. It uses individualized neighborhoods encompassing the 100 nearest members of the age cohort in relation to where they lived at age fifteen. We considered that the scale level k = 100 was the smallest that, given that we compute the proportion of seven different trajectories, would not generate too much random variation. Note that the total population in such neighborhoods can be around 8,000. It illustrates the proportion of individuals of trajectory Class 5 employment only, which is the largest class, establishing a family early, trajectory Class 4, and postsecondary education, trajectory Class 7, two trajectories with contrasting spatial patterns.

Figure 3 Proportion in quintiles of trajectory Class 5 (employment), trajectory Class 4 (establishing a family early), and trajectory Class 7 (postsecondary) among the 100 nearest cohort peers at age fifteen.

At age fifteen, individuals following postsecondary education trajectories were already heavily concentrated in metropolitan areas and in college towns. On the other hand, at the same age, individuals following early childbearing trajectories (trajectory Class 4) were concentrated in peripheral, rural areas. A similar pattern is the case for trajectory Class 5 (employment only) but instead closer to cities. Note that the gender balance of these two classes is different. Trajectory Class 4 (early childbearing) is dominated by women, and trajectory Class 5 (employment only) by men. With this in mind, the maps therefore illustrate a gender-based polarization between trajectories in the countryside. At the same time, there are differences in the geographical pattern of trajectory Class 4 and trajectory Class 5, in that trajectory Class 5 (employment only) is better represented in perimetropolitan areas, in other words nonmetropolitan areas relatively close to the major metropolitan centers, such as Stockholm, Göteborg, and Malmö. High concentrations of trajectory Class 4 (establishing a family early) are found instead in locations that are relatively distant from the metropolitan centers. A possible interpretation of this is that metropolitan influences can deter individuals from starting a family early.

How Much Sorting?

The preceding analysis shows patterns in how the trajectory classes are linked to specific geographical contexts. Is it the case that individuals who follow different trajectories grow up in different places, or are life-course trajectories mixed in a local area? The answer, illustrated in Table 8, is that there are substantial differences but there is no complete segregation. Some trajectory classes are relatively well represented in most neighborhoods, such as trajectory Class 1, trajectory Class 5, and trajectory Class 6. Here, the contrast between the 25th percentile and the 75th percentile is relatively small. The differences are greater for trajectory Class 2, trajectory Class 3, trajectory Class 4, and trajectory Class 7. One way to summarize this is to say that the different trajectories are represented to some extent in most spatial contexts, but in some cases a certain trajectory is in a relative minority, and in other contexts the same trajectory is more dominant.

Table 8. Proportion of different trajectory classes among the 100 nearest cohort peers, at age 15 at different percentile values

Trajectory class	Description	10th percentile	25th percentile	75th percentile	90th percentile
1	Postsecondary education, work	13%	15%	23%	27%
2	Work, then starting a family	5%	7%	13%	16%
3	Low activity	4%	6%	11%	14%
4	Establishing a family early	3%	5%	10%	13%
5	Employment only	19%	23%	31%	35%
6	Postsecondary education, work, family	8%	9%	14%	17%
7	Mixed postsecondary education and work	8%	11%	19%	24%

Table 9 illustrates spatial sorting of trajectories at age thirty, and indicates a pattern of increased sorting for most of the trajectories. For Classes 1, 2, 4, and 7, the 10th and 25th percentile values have declined substantially from the context at age fifteen to the context at age thirty. This implies that these trajectories are poorly represented in a substantial proportion of the k = 100 neighborhoods. The trajectories involving establishing a family early are particularly poorly represented in 10 to 25 percent of the neighborhoods, but postsecondary education trajectories are poorly represented in the lower part of the distribution. Instead, many individuals who follow trajectory Class 1 and trajectory Class 7 are found in a spatial context where there are large concentrations of individuals following the same trajectories.

Table 9. Proportion of different trajectory classes among the 100 nearest cohort peers, at age 30 at different percentile values

Trajectory class	Description	10th percentile	25th percentile	75th percentile	90th percentile
1	Postsecondary education, work	8%	11%	28%	36%
2	Work, then starting a family	1%	3%	15%	20%
3	Low activity	3%	5%	12%	15%
4	Establishing a family early	1%	2%	11%	15%
5	Employment only	18%	23%	33%	37%
6	Postsecondary education, work, family	4%	7%	15%	19%
7	Mixed postsecondary education and work	5%	8%	23%	30%

Concluding Discussion

In this article, our motivation for assuming that individuals are assigned to different types of life-course trajectories was Bourdieu’s theory of habitus, which asserts that individuals by their socialization, adaptation to the structures of the social worlds that they inhabit, will be linked into specific biographical trajectories.

Nearly half of the 1986 cohort consists of young individuals who go into postsecondary education, and are allocated to trajectory Classes 1, 6, and 7. The reproduction of educational length, often tertiary education, from their parents is clear, as 48 to 62 percent grew up in a tertiary-educated household. A future of employment is also the norm of the three trajectories. As is often a way of life in combination with tertiary education, the phase of forming a family comes into these person’s life courses at later ages. Trajectory Class 6, however, includes forming a family after the age of twenty-five. Nevertheless, it seems individuals in these classes show trajectories of habitus suitable for acquiring higher education in the educational system.

For trajectory Classes 2, 4, and 5, it might be argued that the distance is too large between the habitus of academic language and productive pedagogical work for tertiary education, and the habitus of family life and earlier school experiences. Among these trajectories the proportion coming from tertiary-educated parental households is only 22 to 30 percent. Therefore, these classes describe trajectories of no further education after secondary education. Despite the similarity of no tertiary education, the trajectories display differences concerning employment and family formation. Trajectory Class 5 (employment only) and trajectory Class 2 (secondary education, employment, then establishing a family) can be considered working class in the true sense of displaying steady employment. Also, the majority of those in trajectory Class 2 and trajectory Class 5 are men. It could be argued that trajectory Class 4 (early family formation) is also working class, but with a majority of women, who do not reach the same high proportions of employment, instead the trajectory includes some tertiary education in the family formation phase.

Trajectory Class 3 can be assigned a trajectory for underclass, at least when individuals are in the studied life phase, and taking into consideration their socioeconomic and demographic position. As mentioned, the low activity recorded for education, employment, and family formation seem to concur with the NEET concept. The cultural capital seen as the social assets of a person including education, style, knowledge, and taste that can (or cannot) promote social mobility in a segregated society is presumably low among individuals in trajectory Class 3. The trajectory also shows the largest proportion of individuals in need of support with social allowance, and a majority are men. In addition, we find an overrepresentation of individuals being born outside Sweden in this trajectory.

Using the results of the LCA we have also analyzed the sorting of life-course trajectories across different spatial contexts. This analysis complements earlier studies that have compared young adult trajectories across different countries (Billari and Liefbroer 2010). This analysis shows a relatively strong sorting both across a metropolitan–nonmetropolitan gradient and within metropolitan areas. Our main finding is that trajectories involving postsecondary education are overrepresented in metropolitan areas, whereas trajectories that are not in this category are overrepresented in nonmetropolitan areas. Another finding is that nonmetropolitan trajectories are more gender-polarized than the metropolitan trajectories. These patterns we see as evidence of spatial variation in local cognitive and motivational structures. That is, growing up in metropolitan areas implies a clear inculcation of behavioral patterns and attitudes that guide individuals toward seeking out life courses involving a career based on participation in postsecondary education. In contrast, growing up in nonmetropolitan areas guides individuals into careers involving an earlier entry into working life and, for many, also early family formation. The contrast between metropolitan areas and nonmetropolitan areas is not absolute, though. Across local contexts it is possible to find a mixture of trajectories, but the composition of the mixture varies. Moreover, if one analyzes how trajectories are sorted at the end of the sixteen- to thirty-year age windows, it is clear that the sorting of trajectories becomes more pronounced with increasing age, which suggests a mechanism that can explain spatial variation in cognitive and motivational structures. It looks as if one’s life-course trajectory is a factor that governs where one chooses to live, and that this influence of trajectories on residential locations can have the effect of producing local contexts that are relatively homogeneous with respect to life-course patterns.

If the results of the LCA are compared to the findings reported in Swedish, interview-based accounts of young adult life courses, there is certainly a clear overlap. The qualitative studies we have referred to also point to a clear division between life courses involving postsecondary education and those that do not. These studies also report a clear differentiation by gender for life courses that involve an early entry into the labor market. These qualitative studies report similar patterns with respect to spatial variation in life-course patterns across local contexts to our LCA-based study: a stronger gender polarization in nonmetropolitan contexts and a strong orientation toward postsecondary education in metropolitan areas. This result is of interest because it shows that LCA using register data can provide an efficient tool for surveying regional and local differences in life-course patterns across a large number of local contexts. It suggests that qualitative studies can be used to get information about the attitudes that influence life-course behavior, and that such information can be used to provide an understanding of life course that is identified in register data. That is, by combining qualitative approaches with the results of latent class modeling, it is possible to provide a new understanding of regional differentiation as linked to a differentiation of life-course trajectory landscapes. This, we would argue, can be seen as akin to a genre-de-vie-oriented approach to regional analysis.

Data Availability Statement

The data that support the findings of this study are available from Statistics Sweden. Restrictions apply to the availability of these data, which were used under license for this study.

Additional information

Funding

We acknowledge financial support from the Swedish Foundation for Humanities and Social Sciences (Riksbankens Jubileumsfond, RJ), grant registration number M18-0214:1 and from the Swedish Research Council for Health, Working Life and Welfare (FORTE), grant number 2016-07105.

Notes on contributors

Bo Malmberg

BO MALMBERG is a Professor in the Department of Human Geography, Stockholm University, 106 91 Stockholm, Sweden. E-mail: [email protected]. His research interests include population geography, time geography, residential sorting, neighborhood effects, economic development, political geography, and methodology.

Eva K. Andersson

EVA K. ANDERSSON is a Professor in the Department of Human Geography, Stockholm University, 106 91 Stockholm, Sweden. E-mail: [email protected]. Her research interests include urban geography, neighborhood effects, residential and school segregation, housing, inequality, socioeconomic careers, rural gentrification, residential mobility, and older people.

Appendix

Figure A.1 Akaike’s information criteria for different numbers of latent classes.