Models of leaving home: patterns and trends in Sweden, 1830–1959

ABSTRACT In this study, we examine the development of age at leaving the parental household in Sweden between the years 1830-1959. We utilize individual-level longitudinal data from two geographically and socioeconomically different regions: the county of Scania in the very south of Sweden, and Västerbotten to the north. We use descriptive and multivariate analyses to investigate how determinants, such as age at marriage and socioeconomic status, affected the age at leaving the parental household over time and between different subgroups, such as sex and rural-urban setting. We show that the age at leaving the parental household was initially low but increased strongly during industrialization but fell again during the interwar period and onwards. Regional and subgroup differences in age at leaving the parental household were small throughout the investigated period, indicating that the development was general in nature. Therefore, we argue that our results indicate that different models governed the structures and norms of home leaving during our investigated period. More specifically, a pre-industrial model gradually shifted into an industrial model, with the latter one becoming dominant in the 1920s. In the pre-industrial model, leaving home was shaped by the life-cycle service system. In the industrial model, age at marriage instead became a main determinant of home leaving.


Introduction
In this study, we examine the age at leaving the parental household in Sweden for a period of more than a century, 1830-1959, using longitudinal individual-level data for two regions with different social and economic structures: Scania (Skåne) in the very south of Sweden and Västerbotten to the north. We describe the long-term changes in the timing of home leaving during the period when Sweden was transformed from an agrarian to an industrial society and analyze the potential drivers of that development. We focus on differences across geographic context, gender, rural/urban environment, the socioeconomic status of the parent, and the family structure of the parental household.
We also relate the determinants of leaving home to the changes in age at marriage and the development of the economic structure of the regions.
The study period begins in the early 19 th century when Sweden was predominantly agrarian. During this time the paths out of the parental household were limited. As in other parts of Northwestern Europe, individuals primarily left the parental home for one of two reasons. The first was to form a separate household (i.e. via marriage). The second was to take up work in another household for a limited time, with the goal of earning enough to establish one's own household later; a system called life-cycle service (Dribe, 2000;Guinnane, 1992;Laslett, 1977;Steckel, 1996;Van Poppel & Oris, 2004;Wall, 1978Wall, , 1989. The study period ends in 1959, right before the modern pattern of home leaving emerges in Western Europe and North America. During the 1960s and −70s, the paths out of the parental household started to become increasingly more diversified than previously. The system of life-cycle service had become totally obsolete, and marriage an increasingly atypical reason to leave home, occurring at progressively later ages (G. Andersson & Philipov, 2002;Holland, 2017). Instead, leaving for post-secondary education became increasingly common. Furthermore, new forms of household arrangements also became more frequent and accepted. For example, one-person (or one-parent) households, living as a lodger or sharing a household with friends or family (e.g. F. K. Goldscheider et al., 1993;Murphy & Wang, 1998;White, 1994; see also Dribe & Stanfors, 2005 for Sweden).
While research on both the traditional, pre-industrial, home leaving patterns as well as the modern patterns has been conducted by historical demographers, demographers, and sociologists respectively, few studies have focused on the long-term development of home leaving during the transition from agricultural society in the early 19 th century to a fully industrial one that emerged after the Second World War (an exception being Gutmann et al., 2002). Therefore, it is uncertain if the changes in the home leaving patterns were gradual or sudden. Furthermore, we do not know how the development of the age at leaving home is related to other important developments in society during this time, such as changes in age at marriage, socioeconomic structure, and urbanization. Especially pertaining to the changes across subgroups such as men/women, different socioeconomic status (SES) groups, and people in rural/urban areas respectively.
To analyze the development and drivers of home leaving during this 130-year period, we investigate whether the development was related to structural changes in society, and if these relationships changed during the study period. More precisely, we analyze the relationship between home leaving on one hand and the timing of marriage and changes in structures of economic opportunity on the other, two major drivers of the timing of leaving home (as specified by Gutmann et al., 2002). 1 Our main research questions are: (1) Did the trend in age at leaving home follow the trend in age at first marriage, i.e. to what extent was home leaving marriage driven? (2) Were changes in the timing of home leaving related to changes in the socioeconomic structure (e.g. the decline in the agrarian life -cycle service system, new employment options in industry and trade, and various new housing options in urban areas)?
To analyze changes in the determinants of leaving home, we use descriptive measures such as Kaplan-Meier estimates of the median age at leaving home and the interquartile range of home leaving using non-parametric survival analysis. We also use multivariate event history analysis (Cox regression) of longitudinal individual-level data to evaluate the non-parametric results and assess how household composition, socioeconomic status, and living in a rural or urban context impacted home leaving over time.

Previous research -the determinants of leaving home
In a study of leaving home in the 20 th -century US, Gutmann et al. (2002) identified four possible determinants. The first one was marriage timing. Until the 1960s, many young people stayed in the parental home until they married, and therefore, changes in marriage ages over time tended to correlate with changes in the ages at leaving home. Women left home earlier than men simply because they married earlier. Some scholars claim that home leaving was more or less a marriage driven phenomenon up until this period (F. K. Goldscheider & DaVanzo, 1989;Rodgers & Thornton, 1985). However, there are studies based on more detailed data indicating that not all people stayed in the parental home until they married but left earlier for alternative housing (Avery et al., 1992;Buck & Scott, 1993; F. K. Goldscheider & DaVanzo, 1989; F. K. Goldscheider et al., 1993;Mulder & Clark, 2000;Lundh & Öberg, 2018). The second factor influencing home leaving was the structures of economic opportunity. Young people stayed in the parental home if the household could offer financial or service support, but left if it could not, or if better earnings and employment terms could be found elsewhere. In urban or industrialized areas in the countryside, young people could possibly find a job nearby, so they did not have to leave the parental home, but in more sparsely populated areas a geographic move was often necessary. One specific type of employment, not mentioned by Gutmann et al. (2002), is life-cycle servanthood. In rural areas in 19 th -century North-Western Europe, a large proportion of young people from all social standings below the elite left home well before marriage to work as servants in another household. This most often meant leaving home at quite an early age, especially among the non-landed groups (Dribe et al., 2014;Dribe, 2000;Lundh, 1999b). The employment included board and lodging in the employer's household, so payment was both in money and in kind. Servants frequently moved between employers, saving money to eventually be able to leave servanthood, marry and set up a household of their own (e.g. Dribe & Lundh, 2005b;Kok, 1997;Lundh, 1999a). Peter Laslett (1977) named this occupational system life-cycle service since it was only for a distinct stage of the life-course that young people lived as unmarried servants (see also Fauve-Chamoux, 2004;Kussmaul, 1981;Mitterauer, 1992). This type of household formation system, typical of Northern Europe, was described by John Hajnal as resulting in neolocal household formation and late and non-universal marriages (Hajnal, 1965(Hajnal, , 1982. In urban areas, there was a similar arrangement in the crafts system. Unmarried young boys that were employed as apprentices or journeymen lived in the household of their master. The occurrence of servants and apprentices inherent in the master's household lowered the mean age at leaving home during the period this system was in place. The third important factor influencing home leaving was available alternatives to marriage and employment, such as education and military service (Gutmann et al., 2002). The trend of greater access to education had diverse effects on home leaving. More primary and secondary education led to a longer stay in the parental household. Expansion of primary education occurred in Sweden throughout the second half of the 19 th century, and secondary education started to expand rapidly after the turn of the century, especially after 1927, when girls were given access to publicly funded secondary schools on the same conditions as boys (Stanfors, 2003). The expansion of tertiary education after 1960 was associated with a decrease in the age at leaving home as the students were living away from their parents while at school. These changes related to the expansion of tertiary education occurred primarily for the cohorts born after 1945 in Sweden and are thus not covered by our analysis.
The fourth factor was migration. The vast internal migration from rural-agrarian to urban-industrialized areas led to homogenization in the age at leaving home over the 20 th century. In the US, international immigration also affected the pattern of home leaving, as some migrants came without their parents or brought their own values and practices. In Sweden, rural-to-urban migration also increased significantly during the investigated period. The number of immigrants was, however, negligible (Svanström, 2015).
Apart from the four factors outlined by Gutmann et al. (2002), the timing of leaving home has also often been described as structured by social norms (Aassve et al., 2013;Billari & Liefbroer, 2007). These norms could be about the proper age at leaving home itself, but norms regarding the family system, the household formation system, or the marriage pattern could also affect home leaving indirectly. Norms on a rather high proper marriage age, or an expectation that a married couple should form a separate household, also affected the timing of home leaving (Hajnal, 1965(Hajnal, , 1982van Poppel & Nelissen 1999). Both aggregate and micro-level studies indicate that the timing of home leaving varied with gender and time. Studies based on multivariate analysis of individual-level data, mostly for the post−1960s period but in some cases for earlier periods, often include variables on individual and household characteristics such as: gender, age and employment/earnings, father's human capital and earnings, and household composition, e.g. number of siblings by gender and age, or the absence of the father or mother in the household.
Results from these studies indicate that women tended to leave the parental home earlier compared to men, and that individuals in their 20s were the most prone to move out. A young person with employment and earnings was more likely to move out than someone without resources. Higher levels of parental resources (e.g. status or earnings) seem to have worked in both directions: on the one hand, increasing the possibility of keeping an adult child in the parental household, on the other hand, making it possible to finance the young person's education, away from the parents (Avery et al., 1992;Buck & Scott, 1993;Dribe & Stanfors, 2005;F. K. Goldscheider & DaVanzo, 1989;F. K. Goldscheider et al., 1993;Lundh & Öberg, 2018;Mulder & Clark, 2000).
Further results indicate that the effect of household composition on home leaving interacted with other factors. The death of a parent usually stimulated home leaving (see e.g. Rosenbaum-Feldbrügge, 2022). However, the need for financial aid or domestic service of the surviving parent could work in the opposite direction. The gender of the young adult also played a role. A widower may have more need of domestic service from a daughter, and a widow may have demanded financial aid from a son with higher earnings. Moreover, the number, age, and gender of siblings influenced the possible effect on home leaving, depending on the gender, age, and earnings of the young person who considered moving away from the parents.
In North-Western Europe, the traditional (pre-industrial) model of leaving home could be described in the following way. Unmarried people were included in an existing household, either the parental home or a household headed by the employer or somebody else (a relative or non -related person of legal status, e.g. a distant-related or nonrelated family father or widow). It was usually not possible for unmarried young people to leave the parents and form an independent, singleton household. There were two routes out of the parental home: leaving for marriage, including family and household formation, or leaving for work, living as an unmarried servant in the household of the employer or somebody else of legal status. The norms related to marriage (concerning proper age and ability of forming a separate, new household) and the lack of housing for unmarried and uncomfortable housing conditions for lodgers tended to increase the mean age at leaving home. The life-cycle service system worked in the opposite direction; it tended to lower the mean age at leaving home.
The development from the traditional home leaving model to the modern model, as described in the introduction, involved changes in the housing alternatives. Life-cycle service gradually diminished and had disappeared completely by the 1950s, which should have contributed to an increase in the mean age at leaving home. However, increasing possibilities for unmarried young people to form a separate household opened new routes out of the parental home, which probably decreased the age when young people moved out. It remains uncertain when this transition happened, and how the patterns of leaving home looked in the period between the traditional and the modern models.

Data and methods
In this paper, we investigate two regions in Sweden, Scania to the very south and Västerbotten in the north. The data are from the Scanian Economic-Demographic Database (SEDD) held by Lund University (see Dribe & Quaranta, 2020) and the POPLINK database held by Umeå University (see Edvinsson & Engberg, 2020). Both databases are part of the SwedPop data infrastructure, and the data have been harmonized in terms of main variables (e.g. occupation) and data structure (www.swedpop.se). We include data from 1830-1959 for both areas, and study individuals between age 15 and 39 living in their parental home.

Scania -SEDD
The SEDD-data consist of individual-level longitudinal information from five rural and semi-urban parishes for the period 1830-1959, Halmstad, Hög, Kävlinge, Kågeröd, and Sireköpinge, and a port town, Landskrona, after 1905; all located in Scania, Sweden's southernmost province Dribe & Quaranta, 2020;Quaranta, 2015Quaranta, , 2016. The five rural parishes had a combined population of 4,500 in 1830, 5,500 in 1900 and, together with Landskrona, a bit over 30,000 in 1950. Four of the five parishes were rural throughout, while one of them (Kävlinge) transformed into a small town with several factories following the building of one of the main national railroad lines through the town.
The study population is not a random sample of Sweden but is broadly representative by reflecting conditions shared by populations in similar areas during the time studied . More specifically, for the period 1830-1904, for which we only have data for the five parishes, the area reflects the population density, age, and occupational structure in Sweden outside the major towns. In fact, until about 1930, more than half of the Swedish population lived in rural areas.
These data consist of individual-level longitudinal information, provided from continuous population registers, husförhörslängder or församlingsböcker, household-based registers where information at the individual level is continuously updated, with information on demographic events, including migration to and from households for all individuals in the area. Birth and death registers have been used to add events not recorded in the population registers. Another important characteristic of the data is that migration, into and out of the study area, is comprehensively recorded, meaning that the population at risk is well defined.

Västerbotten -POPLINK
The POPLINK-data include the same type of population register information as in SEDD. Geographically, the POPLINK-data covers 15 parishes in the county of Västerbotten in northern Sweden. The individual-level longitudinal information includes both occupational information and detailed migration information (usually including both precise dates and the specific destination/origin of migration on a parish level). The database includes both coastal and inland rural parishes and the two towns of Umeå and Skellefteå situated along the coast of the Gulf of Bothnia. The data allows for the study of individual life-courses, spanning multiple generations. Due to the relatively low population turnover in the region, particularly before 1900, individuals can often be followed from birth to death with complete register coverage of the entire life course. This is ideal for studying various life-course transitions, such as leaving the parental home (Westberg et al., 2016). For the period 1830-1904, the total number of individuals found in the data is 49,219. For the period 1905-1959, 87,053 unique individuals are found in rural areas and 9,152 in urban areas.

Statistical method
The study population consists of unmarried persons in the age range 15-39 who were living in the parental home. The individuals were followed until they left the parents' household for the first time. Individuals who died, reached the age of 40 or were still living with the parents by the end of our study period were censored. The outcome on which we focus, the dependent variable in the regressions, is the move out of the parental household in terms of the date when the individual is registered as living in a household not headed by one of his/her parents.
The analysis combines two methodological approaches. First, to describe the patterns and long-term trends, we use non-parametric survival analysis, which provides averages and measures of variation, such as the median and interquartile range of the age at leaving home. Second, in the multivariate analysis, we use semi-parametric Cox regression to model the hazard of leaving the parental home. This is a proportional hazards model assuming that differences in the relative hazard between individuals with different covariate values are constant over time. In the regression analysis, we include several determinants of leaving home. All independent variables included, except gender, are time-varying such as the household composition of the parental household. Table 1 shows descriptive statistics for all variables included in the analysis. The time dependence of the hazard is modeled using age as the time scale. We found some evidence of non-proportionality in the effect of having parents present in the household with stronger effects of absence of parents at younger ages. However, models where we relax the assumption of proportionality for this variable by including an interaction with analysis time show that the effect is consistently positive throughout the period at risk. In addition, adjusting for non-proportionality does not meaningfully change the estimates for the other variables in the models. As the presence of parents is primarily included as a control variable, we therefore choose to present a simpler model that excludes a time-varying effect of parents for reasons of parsimony. In Table 1 we show the relative frequency of episodes across the independent variables included in the analysis.
The multivariate model includes variables for gender, urban-rural area, period, family context and SES. We define leaving home as a child moving out from the parental home for the first time. We follow all children that reside in the parental home at any time after 1 January 1830 from the date they turn 15 years of age until their first move out of the parental household. Individuals are censored and removed from the risk set if they die before moving out, reach age 40 or when our analysis ends on 31 December 1959, whichever occurs first. Gender is a categorical variable (male; female) defined based on direct information in the registers, or on first names. Period is also a categorical variable including three periods: 1830-1879; 1880-1929; 1930-1959. The period variable is used to analyze the development over time, and we define the periods based on the occupational makeup of the population in the two regions. The agricultural sector consistently comprised more than 80% of the workforce in the age groups in question (age 15-39) until the 1870s. Starting in the 1880s, a precipitous decline of the agricultural sector and corresponding increase of the industrial and service sectors ensued in both regions and continued until the 1930s when the industrial sector had outgrown the agricultural sector in both regions. Therefore, we define three periods indicating the overall economic makeup of society using the following cutoffs: a pre-industrial period (1830-1879), an industrializing period  and an industrialized period .
Rural and urban areas are defined at the parish level. In Scania, Landskrona constitutes the urban area, while the other five parishes are defined as rural even though one of them, Kävlinge, for most of the study period was a small semi-urban place with industries, trade, and crafts in addition to farmland (see Context). In Västerbotten, Umeå and Skellefteå town parishes are defined as urban, while the remaining parishes are defined as rural.
We measure family context as a time-varying covariate with three separate categorical variables: presence of parents (both; only father; only mother), and presence of brothers and sisters respectively (only older; only younger; both; neither) that are updated when changes in the household composition occurs.
SES is a categorical variable based on the occupation of the ego's father. It is included as a time-varying covariate reflecting the current occupational status of the father. Occupational notations in the sources have been coded in an internationally comparable coding scheme for historical occupations: HISCO, the Historical International Standard Classification of Occupations (van Leeuwen et al., 2002). These standardized occupations have subsequently been coded into HISCLASS (Historical International Social Class Scheme), a 12-category occupational classification scheme based on skill level, degree of supervision, whether manual or non-manual, and whether urban or rural (van Leeuwen & Maas, 2011). In the analysis, we use a three-class version of the scheme. It includes the following classes: white-collar (non-manual) workers (HISCLASS 1-5), blue-collar (manual) workers (HISCLASS 6-7, 9-12), and farmers (HISCLASS 8). We also include individuals without a registered occupation as a separate category (N/A).

Context
In the pre-industrial period, the two study regions shared norm systems and institutional arrangements such as the family system (nuclear), household formation system (neolocal), marriage pattern (late marriages), and inheritance/succession (equal and partible) but differed in geographic location and their economic structure. In this section we briefly present and compare Scania and Västerbotten with regards to the development of the median age at marriage and the structural change of economic sectors and the labor market during industrialization.
Scania is the southernmost province of Sweden, and part of the former Danish provinces that were acquired by Sweden in 1658. The population included in this study is taken from the town of Landskrona on the west coast , and five rural and semi-urban parishes located about 15-20 kilometers inland from the town . In the 20 th century, Landskrona was a classic industrial city, dominated by manufacturing, shipbuilding, and textiles (Dribe & Svensson, 2019).
The five parishes differed in relation to geographical conditions and ownership structures. Halmstad, Sireköpinge, and Kågeröd were neighboring parishes, located in the transition area between the agricultural plains and more forested areas of Scania. They were historically dominated by land owned by the nobility and farmed by noble tenants or agricultural laborers working under the estates. Hög and Kävlinge were neighboring parishes about 20 kilometers south of the other three parishes. The parishes were located on the plains, dominated by freeholding peasants and tenants on crown land. During the 19 th century, the area transformed during the agricultural revolution with enclosures, extensive land reclamation, and new crops and farming methods. It was also a period of continued commercialization of the agricultural sector, leading to increased income first among the landholding class but from midcentury also among the agricultural laborers (see Bengtsson, 2004;Dribe, 2000 for more detailed descriptions of the parishes in the 19 th century). At the end of the 19 th century, Kävlinge was transformed into a small town with small industries, especially related to food processing and textiles. The transformation was to a large extent driven by the building of the main railroad line between Malmö and Göteborg, on which Kävlinge became a station and connected to other regional railroad lines (Hellborg, 2017). Hence, in the 19 th century, the SEDD sample was completely rural, while in the 20 th century it includes a rural area, a semi-urban locality, and one urban industrial area.
Västerbotten is the second most northern county in Sweden and is geographically much larger than Scania, stretching from the Baltic coast in the east to the Scandinavian mountain range in the west. By 1830, the only larger town was Umeå. While rural settlement was widespread along the coast and river valleys, large parts of the county were undeveloped forest. The county was sparsely populated, with around 50,000 inhabitants in 1830. However, during the 19 th and early 20 th century, the population grew rapidly, reaching 122,000 in the 1890s and 231,000 in 1950 (SCB, 1969). Settlement of previously uncultivated land increased from the late 1800s until the 1930s (Bäcklund, 1988;Morell, 2011). Sawmills also started to emerge in the late 1800s, spearheading industrial development in the county. This was later followed by large scale mining operations in the 1920s around the town of Skellefteå, which emerged in the mid−19 th century north of Umeå (see, e.g. Andersson, 1987 pp. 126-144;pp. 126;Östlund, 1995).
The expansion of the timber industry was crucial for the economic development of the region. Not only did the sawmills provide the option of industrial wage labor near the urban areas along the coast, but the extraction and transportation of the timber provided the agricultural population with an option to work part-time as wage laborers (Gaunitz, 1977;Lundgren, 1977). The combination of subsistence farming and wage labor became the foundation of the rural economy in most of the rural inland areas of Västerbotten. The availability of wage labor in rural areas could also partly explain the low rate of urbanization in Västerbotten during the early 20 th century compared to the rest of Sweden (Sundvall, 2023). However, by the 1940s, the number of small farmsteads began to decrease. By 1960, the population of the county had increased roughly 500% since the mid−19 th century, and the county was becoming increasingly urbanized.
The development of the median age at first marriage by gender and rural/urban context in both regions is shown in Figure 1. The trends are similar between the regions: Generally, men married later than women while living in a rural or urban parish had little effect on the median age at marriage. The average marriage ages were also similar in the two regions: 26-28 for men and 23-26 for women. The trends regarding age at first marriage were quite similar in both regions after 1920 when a general drop in median age occurred across the board as Sweden entered the mid−20 th century baby boom period (Sandström, 2014(Sandström, , 2017. The developments from 1830-1920 were not as similar, however. In Scania, the median age slightly decreased, for both men and women, during the 19 th century, whereafter it increased until the 1920s. In Västerbotten, the median age at first marriage increased slightly for men from 1830 until 1920, while for women it remained practically stable until the drop after the 1920s. Both regions were predominantly rural in the 19 th century (see Figure 2); in Västerbotten, the rural population exceeded 90%, while in Scania, it was closer to the national average of 90%. Because of internal migration to urban areas and changes in the classification of industrial agglomerations from rural to urban, the relative size of the rural population started to decline in the 1870s and 1900s respectively. Urbanization was more pronounced in Scania than in Västerbotten. By the end of our study period, the majority of the population in Scania was urban while two thirds were still rural in Västerbotten.
In terms of economic development, we find both similarities and differences between the two regions. Figure 3 shows the percentage of employees in three economic sectors:  Table 1. agriculture, industry, and services. Agriculture comprises farmers, agrarian workers, farmhands, and maids (including female domestic servants in rural areas). Industry includes production workers, transport equipment operators, shoemakers and leather goods makers, rubber and plastics product makers, forestry workers, and fishermen, hunters and related workers. Service comprises professional, technical, administrative, managerial workers, clerical, and sales and service workers (except female domestic servants in rural areas). The classification was made based on the HISCO codes derived from the occupational information in the data (for detalis, see the note toFigure 3). Figure 3 shows the change in economic structure in Sweden in the 19 th and 20 th century, which has been presented for the country as a whole in previous studies (Magnusson, 2000;Schön, 2012). Up to the industrial breakthrough in the 1870s, the vast majority of the population worked in the agrarian sector, but industrialization rapidly changed the balance between the economic sectors. By 1900 in Scania and 1930 in Västerbotten, there were more workers in industry than in agriculture. The service sector grew along with the growth of a mature industrialized economy. In the inter-war period, industry and service accounted for about 40% each of the total labor force in Scania, while as little as one tenth of the total labor force was working in agriculture by the end of the study period. Västerbotten lagged behind but developed in the same direction.
The system of life-cycle service was geographically widespread, but the size of the servant group is uncertain. Previous studies indicate that, measured as a cross-section, about 10% of the agrarian population were servants in 1800, a proportion that probably increased during the 19 th century before it started to decrease at the end of the century  (Wohlin, 1909, p 49). Harnesk (1990, pp. 9-10, 181-185) suggests that over half of the population worked as servants at some point during their lifetime, and Lundh (1999b) reports even greater degrees of transition into servanthood for a Scanian population in the 19 th century.
To investigate how the institution of life-cycle service developed during our study period, we generated a proxy for the size of the life-cycle servant group as the number of unmarried workers in the agricultural sector that did not live in the parental household. We divided the number of life-cycle servants by the total number of workers in the agrarian sector. Thereby, we obtained a measure of the relative size of the life-cycle servant group, which is presented in Figure 4. The development was strikingly similar in both regions. The proportion of life-cycle servants in the agricultural workforce was around 70-80% in the 1830s−1850s. The proportion of servants started to decline after the 1880s, a drop that continued throughout the study period. By the 1950s, the proportion of life-cycle servants in the agrarian workforce was around 10-15%.  Table 1. Note: Agricultural: HISCO minor groups 61, 62 (farmers; agricultural and animal husbandry workers), and for women in a rural context HISCO minor group 54 (maids). Industrial: HISCO major groups 7, 8, 9 (production and related workers; transport equipment operators; shoemakers and leather goods makers; rubber and plastics product makers) and HISCO minor groups 63, 64 (forestry workers; fishermen, hunters and related workers). Service: HISCO major groups 0/1, 2, 3, 4, 5 (professionals, technical and related workers; administrative and managerial workers; clerical and related workers; sales workers; service workers except minor group 54 for women in a rural context).

Results
In this section we present our main findings. First, we examine the patterns and development of home leaving in terms of Kaplan-Meier estimates of the median ages, then present the results from the multivariate Cox regressions. Figure 5 shows the median age at leaving home by gender and urban-rural setting. The general time trend is very similar for both regions; the median age rose significantly from the mid−19 th century up until the interwar period, peaked in the 1920s and declined thereafter. The pattern was similar for both genders, but the increase was greater for men. Furthermore, women generally left home at a younger age than men in Västerbotten during the entire period and in Scania from the 1890s onwards. Both the urban and rural populations in the two regions also followed the same general time trend. For men, we find very small urban-rural differences in the age patterns, while women in urban areas generally left home at a more advanced age than women in rural areas. Only in Scania we find that urban men, similarly to urban women, tended to move out later during the first two decades of the 20 th century. Figure 6 shows the difference between the median age at first marriage and the median age at leaving home by gender and rural-urban setting. Within Scania and Västerbotten respectively, the trends were very similar across the subgroups. Likewise, the general trends across both regions were also similar, i.e. the age gap diminished with time. However, the levels across the two regions differed a lot. There was a much larger gap between age at first marriage and leaving home in Scania during the 19 th century than in Västerbotten, with the Scanian gap being over 10 years for rural men at the start of  Table 1. Note: Differences based on medians for both first age at marriage and leaving home. the investigated period. However, by the turn of the century the difference between the two regions had in practice disappeared and trends during the first half of the 20 th century were more or less identical in both regions.
During the course of the 19th and early 20th century in Scania and Västerbotten, there was a noticeable trend towards a smaller gap between the age of marriage and the age at which individuals left the parental home. This trend was evident across both genders, and across rural and urban areas. Figure 7 describes the development of the median age at leaving home by gender and SES in the two regions. Children of blue-collar workers initially left home at markedly younger ages than children of farmers and white-collar workers. However, the age at leaving home for blue-collar children increased significantly during the 19 th century. The increase was gradual in Västerbotten from 1830 until the 1920s and more rapid in Scania between 1880 and 1920. The age at leaving home was consistently higher among children of farmers and white-collar workers than among children of blue-collar workers throughout the investigated period. The highest age at leaving home alternated between farmers and white-collar workers during the 19 th century. After the 1900s, sons of farmers consistently moved out at a higher age than white-collar sons, while the opposite was mostly true for daughters.
The trends for all three SES groups, across genders and regions, closely mirrored the overall development outlined in Figure 5, with the median age increasing during the late 19 th century, peaking around the 1920s and decreasing thereafter. Furthermore, the  Table 1. Note: White-collar .... Blue-collar (HISCLASS 6-7, 9-12); Farmers: HISCLASS 8. disparity in age at leaving home between the three SES groups gradually diminished over time. By the 1950s, there were only small differences between the three SES groups in both Västerbotten and Scania. Table 2 shows the ages at leaving home at the 1 st and 3 rd quartile of the population. The interquartile range (IQR; the age difference between reaching Q1 and Q3) is shown in Figure 8. The IQR indicates the age span during which most departures from the parental home occurred. For example, in 1900, 25% had left the parental home when they were 19.8 years (Scania) or 20.7 years (Västerbotten), and 75% had moved out by the age of 27.7 or 29.9 respectively. As shown in Figure 8, IQR varied between around 5 and 9 years in the two populations. It increased in the 19 th and early 20 th century, followed by a decrease from the 1920s (in Västerbotten it once again increased after the 1940s). The low age at leaving home in rural Scania in the 19 th century is explained by a very early departure by children of landless laborers (Dribe, 2000). At the age of 16, about 25% of the Scanian population had already left home; the corresponding age for Västerbotten was higher (about 18).
We now turn to the multivariate Cox regressions presented in Table 3 and 4. The estimated hazard ratios can be interpreted as the relative difference in the age-specific intensity of moving for the category in question compared to the reference category. Thus, values greater than one mean that individuals in the corresponding group have a higher hazard of moving out and on average do so at a younger age than individuals in the reference category, while values less than one indicate a higher age at leaving home. We estimate separate models for men and women by both region and period. In all models, we include controls for the SES of the father and family composition, namely, presence of parents, and presence of older or younger brothers and sisters respectively. The control for rural or urban area is measured only in the last two periods as data from urban areas are lacking in the data for the earliest period. All models also include a control for the decade that the individual is at risk of experiencing first move from the parental home.
We generally find similar estimates for the variables in Västerbotten and Scania. Figure 9 contains both the adjusted and unadjusted hazard ratios by decade, using 1830 as the reference category. 2 The development of the hazard ratios for the period closely mirrors the general time trend of the median age that we find using non-parametric Kaplan-Meier estimates that do not control for any compositional changes over time (see Figure 5). A significant reduction in the hazard of leaving home occurs throughout the 19 th century until 1920-1930, signifying progressively older ages at leaving home, after which the hazard increases again. This interwar increase in the hazard is apparent for both sexes, even though the increase among women was larger. The very modest differences between adjusted and unadjusted hazard ratios indicate that compositional changes in our control variables, such as parental socioeconomic status, sibling composition and urbanization, mattered little for the increased postponement of home leaving. This points to the importance of structural changes not measured by our individual level regression analysis.
Living in an urban area meant a significantly lower hazard of leaving home for men throughout our investigated period while differences for women minimal, as compared to the rural reference category.
As was shown in Figure 7, the father's SES influenced home leaving very differently at the start and end of the study period. More precisely, the difference in median age at leaving home between the SES groups markedly decreased over time. This standardization trend is also visible in the SES-specific development of the relative hazards. For example, the relative hazard of leaving the parental home for daughters of bluecollar workers in Scania was 2.45 in 1830-1879 (using the white-collar category as the reference) which diminished to 1.24 in the years 1930-1959. While still being statistically significant, the difference in the relative hazard had declined from being twice as large as the white-collar reference to only about 20% by the end of the investigated period.

Siblings present in household
No siblings Source: See Table 1. Models include a control for the decade that the individual is at risk of experiencing the first move from the parental home. Family composition was also important for the timing of leaving home. Having both parents present in the household was associated with a lower likelihood of leaving home throughout the study period, for both sexes, in both regions, and in every subperiod. In one-parent households, the hazard of leaving home was generally largest for children  Table 1. living with a parent of the same sex, i.e. with a missing father for sons and a missing mother for daughters. As noted in the methods section, one should keep in mind that there is a time-dependence in the effect of parents with a greater impact when children are relatively young compared to when they are older. Differences in effects for men and women are however also verified in time-interacted models.
The presence of siblings also affected the age at leaving home. Having no or only younger siblings present in the household entailed a higher hazard, while only having older sibling(s) present entailed a lower hazard. The effects of the variables pertaining to family composition were generally largest in the period 1830-1879. While still impactful in the later periods, the differences between the relative hazards generally became smaller.

Discussion
In this study, we examined the development of home leaving in two regions in Sweden for a very long period, 1830-1959. The overall time-trend is consistent across the two regions Scania and Västerbotten, and was also similar among men and women, across all SES groups and in both rural and urban areas. The age spans when most boys and girls left the parental home increased from 18-22 to 23-27 years between the start and end of the study period. The age at leaving home increased sharply from the 1870s/1880s when industrialization took off and continued up until the 1920s, but later dropped in the interwar period. The pattern of leaving home was strikingly similar in both regions and across subgroups during the entire investigated period. The time trend of leaving home in Sweden was quite similar to the pattern described by Gutmann et al. (2002) for the US during the same timeframe but different from the downward trend from the late 19 th century suggested by Stevens (1990) and Pooley and Turnbull (1997) for the US and Britain.
While we discovered no differences regarding the trends in the age at leaving home between the subgroups, there were some differences in levels. First, women generally left home a couple of years earlier than men, something also found in previous studies (e.g. Gutmann et al., 2002;Rodgers & Thornton, 1985). The gender difference in home leaving was probably related to differences in age at first marriage, with women usually marrying when they were a couple of years younger than men. In our populations, the median ages at marriage were 27.4/28.2 for men and 24.0/25.5 for women for the full period. The median age at leaving home was 24.3/25.8 for men and 22.0/21.7 for women.
Second, young people left the parental home about two years earlier in Scania compared to Västerbotten. Since the norms regarding marriage and family formation were similar in the two regions and life-cycle service was practiced in both, the difference could have been caused by differences in socioeconomic structure. In Scania, a large proportion of farmers were tenants and the workforce also included crofters and farm workers, groups for whom keeping an adult child at home represented a financial burden that could be relieved if the young person moved out. In Västerbotten, most farmers were freeholders and children could function as labor on the family farm and reduce the need to hire outside help that had to be paid a salary. Furthermore, opportunities for wage labor increased for large parts of the rural population in Västerbotten as the timber industry developed during the late 19 th century, which might have contributed to later moves as finding wage labor in the countryside of Västerbotten did not always necessitate moving. A larger proportion of blue-collar workers (an SES group that moved out early) in Scania also contributed to the lower median age at home leaving compared to Västerbotten during the 19 th century.
Third, in both regions, living in an urban area generally seems to have delayed home leaving among women. In Scania, this was also true for men up until the 1930s, while no significant difference between rural and urban men was found in Västerbotten. These results are comparable to the findings made by Rosenbaum-Feldbrügge (2022) for children born in rural and urban areas in the Netherlands, 1850-1922. While factors such as gender and socioeconomic structure affected age at leaving home, the change in age at leaving home over time was shared across regions, genders and SES groups throughout the investigated period. The strong similarity regarding the development of age at leaving home therefore indicates a shift in the general structures and norms determining this life course transition in Sweden during this period, at least outside of the most highly urbanized areas such as Stockholm and Gothenburg.
The big difference in our study is temporal: the change from relatively low ages at leaving home in the 19 th century to substantially higher ages in the early 20 th century, followed by a decline that started in the interwar period. We suggest that this development indicates a fundamental shift between different models governing home leaving in Sweden. Furthermore, we argue that these models were developed and sustained, mainly, due to structural developments in society, for example, changes relating to the prevailing social norms and socioeconomic structure.
From the start of our study period until the start of industrialization (1830-1880), home leaving was characterized by low ages. High rates of life-cycle service facilitated the early departures, especially among children with parents from the lower SES group. Generally, marriage occurred much later than home leaving during this period. We choose to call this the pre-industrial model of home leaving. This model would start to lose relevance in the late 19 th century as industrialization and urbanization reshaped the socioeconomic structure of Sweden, leading to a general rise in age at home leaving.
From the interwar years until the end of our study , a new industrial model of home leaving had emerged; where home leaving and marriage age were tied more closely. A similar pattern have been found for Gothenburg, Sweden's second largest city, in the period 1915-1943(Lundh & Öberg, 2018. The increased standard of living, due to industrial development, and the establishment of the first rudimentary welfare state subsequently led to a drop in both age at marriage and home leaving during this period. After the end of our study period, the modern, contemporary pattern of home leaving started to emerge in the 1960s. Ages at marriage and leaving home would once again start to diverge as new routes out of the parental home emerged (e.g. via post-secondary education). Keeping with the theme of the previous two, we call this the post-industrial model of home leaving. In the remainder of this discussion, we will expand further on these three models and the transitions between them.
We will start with the pre-industrial model of home leaving. Family and household formation systems, marriage patterns, and inheritance rules all imply social norms that shape the pattern of leaving home. In societies where the spouse of the child was brought into the parental household after getting married, or where one heir was chosen among the children to take over the parents' property, the process of leaving home was different than in societies with nuclear families, equal heritage among the children, and neolocal residence upon marriage (see, e.g. Dribe et al., 2014). Even though Sweden had equal inheritance for most of the study period and a dominance of nuclear families, it was common among farmers to transfer the farm to one of the sons/sons-in-law, even before the death of the previous household head. Other children/inheritors were then compensated, but before the development of a proper land market, the compensation was often below the productive potential of the farm (see, e.g. Dribe & Lundh, 2005a, 2005c. However, also in such situations, it was common that the chosen heir left home for a few years to work as a life-cycle servant before taking over (Dribe & Lundh, 2005b;Dribe, 2000). For most of the non-landed groups, the situation was different and more like a classic nuclear family system with neolocal household formation upon marriage. Marriage required savings and was supposed to be followed by the formation of a separate household (Hajnal, 1965(Hajnal, , 1983. While waiting to get married, unmarried men and women stayed in the parental home or moved out to live in the household of the employer or, less frequent, anyone who received boarders.
The system of life-cycle service was the solution to the need for boarding, earnings, and savings. A large part of the youth left the parental home to take up work on a local farm and circulated as farmhands or maids between peasant households for a period of their lives, resulting in departure from the parental home at comparatively young ages. For the parents, a child becoming a life-cycle servant eased the burden of provision, and it meant for the young person a chance to earn and save, achieve vocational training, and search for a partner. The North-Western European marriage pattern fits well into a system of such social norms, with late marriages and quite large celibacy rates. The time span between moving away from the parents and later marrying represented a prolonged youth in which taking up service was a first step and marriage was a later and final transition to adulthood.
The decline of the agricultural sector and rise of industry, urbanization, and the radical drop in the proportion of life-cycle servants in the workforce from the 1880s to the 1920s brought about profound changes in the conditions for home leaving. In 1880, about 70% of the agrarian workforce in Scania and Västerbotten were life-cycle servants; in 1940, the proportion was as low as 15%. Simultaneously, more and more young individuals had their first employment in the secondary or tertiary sector rather than in agriculture. During this period, the pre-industrial model of home leaving started to break down with the gradual disappearance of the life-cycle service system. Consequently, the age at home leaving drastically rose as no natural, early route from the parental household replaced life-cycle servanthood in the industrializing society. However, this drastic shift in the structure of the labor market also led to a closer relationship between the timing of home leaving and the timing of marriage over time.
From the 1920s and until the end of our study period, the industrial model of home leaving emerged. In parts of Europe where the life-cycle service system had not been widespread, young people had also previously been more closely tied to the parents and family. Here, moving out was more closely linked to marriage even before industrialization. For Sweden, and other parts of North-Western Europe, the marriage driven industrial model was instead first introduced along with the dissolution of the life-cycle service system. When the life-cycle service system was widely spread and large proportions of young people lived in the households of their employers, the relationship between age at leaving home and age at first marriage was rather weak. In Scania home leaving and marriage age only followed the same general trend from the 1900s onwards (a general increase in age until 1930 and a decrease thereafter). In Västerbotten, the trends were already similar from the 1830s (also a general increase in age until 1930 and a decrease thereafter), however, the increase in marriage age was minimal compared to the increase in age at home leaving during the period 1830-1900 (1 and 6 years respectively). During the 20 th century, and especially after the 1920s, the trends in home leaving and marriage age would become more closely correlated.
During the industrial model of home leaving, young people lived in the parental home while working outside it, as urbanization and growing trade and industry increased the number of jobs available without having to move from the parents (Mitterauer, 1992 p. 74). However, just like in the pre-industrial model, marriage was still the predominant way of establishing a new household. Marriage would therefore become the prevalent norm for leaving the parental home during this time and both marriage and home leaving ages would decline in conjunction (see also F. K. Goldscheider & DaVanzo, 1989;Rodgers & Thornton, 1985;Sandström, 2017;Van Bavel et al., 2013). This can be explained by the economic growth during the interwar period allowing marriage and household formation (and therefore also home leaving) to occur at progressively earlier ages. Especially after the depression in the 1930s, when population policies and welfare state expansion were put into place to promote possibilities for family formation (Sandström & Marklund, 2019;Sandström, 2017). During this time, the socioeconomic development (i.e. new jobs and better earnings in the growing industrial and service sectors) provided the working class with an opportunity of early household formation. Years of wealth accumulation were no longer necessary to start a new household, as had previously been the case. These households were often the male breadwinner-female homemaker type and were commonplace up until the end of our study period (Sandström, 2017;Van Bavel et al., 2018). While the emergence of the industrial model coincided with earlier departures from the parental home, like in the previous pre-industrial model, the nature of the departures was very different between the models. The prolonged period of youth (i.e. living unmarried without an own household outside the parental home) declined in significance as marriage and household formations could occur at younger ages.
After the end of our study period another model of home leaving would become dominant. The post-industrial model would offer alternatives to living with the parents or moving out for marriage and emerged as an integral part of the 'second demographic transition' (e.g. Lesthaeghe, 1995;Lesthaeghe & Van de Kaa, 1986;Lesthaeghe, 2010;van de Kaa, 1987). Most notably, departure for education or employment, living as a lodger, sharing a household with relatives or unrelated persons, or forming a singleton household. The modern patterns of home leaving are not characterized by any strong single norm dictating how a departure from the parental home should occur, but rather by a multitude of options. This also relates to the diversification of households, and household formation, as new types of households would become normalized and increasingly common. For instance: single-person households, one-parent households, and cohabitation without marriage.

Conclusion
Our findings strongly indicate that the development of the age at leaving home in Sweden in 1830-1959 was closely related to changes in socioeconomic structure, primarily related to industrialization. Both the economic and normative framework regarding age at leaving home shifted radically as the agrarian, pre-industrial patterns of home leaving disappeared and new norms and factors influencing age at home leaving emerged. Marriage age was initially not a good predictor of the age when young men and women left their parents' household but became increasingly important as the previous life-cycle service system gave way for new industrial patterns of home leaving. From the 1900s until the end of the study period, the trends in marriage and home leaving closely followed each other. In previous research, both the traditional (what we refer to as pre-industrial) and modern (post-industrial) models have been the focus of numerous studies, we propose that between these two models another model dominated home leaving: the industrial model. While it was socioeconomic changes that brought about this model, it was marriage age that drove the development in home leaving during this period.
So how much do we know about the pattern of moving out of the parental household in a historical perspective? And what are the challenges for future studies on age at leaving home? Let us start with what we know from studies based on Swedish sources. The pre-industrial model of home leaving has been explored and examined in several local studies (Dribe, 2000, Dribe & Lundh, 2005b, Lundh, 1999a, 1999b, and a marriage driven industrial model during the interwar period has been suggested for Gothenburg (Göteborg) (Lundh & Öberg, 2018). The marriage-driven home leaving patterns of -what we call -the industrial model and the rise of the modern post-industrial model have also been examined in previous studies (Dribe & Stanfors, 2005;Nilsson & Strandh, 1999). For Sweden, the knowledge gap concerns two fields. First, we do not know how leaving home was structured before the lifecycle servant system. Was the pattern the same with early departures from the parental household, or was it more like the industrial model, with departure linked to the timing of marriage? Second, were there great differences between urban and rural areas prior to the 20th century in age at leaving home? In this study we find little difference, but we lack information on bigger cities that were truly urban during the pre-industrial period. Given the very strong similarities between two very distant, and somewhat socioeconomically different, areas of Sweden, our results could probably be generalized to other countries in North-Western Europe that underwent industrialization in the 19th century. However, more research is needed to ascertain this.

Notes
1. Besides timing of marriage and structures of economic opportunity, Gutmann et al. (2002) also mention two other potential drivers of the timing of leaving home: education and migration. Since post-secondary education only became a common phenomenon after our study period, we decided to leave this aspect out of our analysis. We do however investigate the effects of living in urban vs rural areas on home leaving, thereby investigating if the large scale rural-urban movement of the late 19 th century and onwards affected home leaving differently in areas with different socioeconomic structures and as well as different net levels of migration. 2. The adjusted hazard ratios include controls for the following variables: Socioeconomic status of ego's father, parents present in household, brothers present in household, sisters present in household and urban/rural parish.