Would you Return?
A Quantitative Study Into the Determinants of
Interregional Return Migration in the Netherlands
Master’s Thesis Author: O.P.J. Leemhuis, BSc.
Student Number: 10371230
Supervisor: C. Kooiman, MSc. (Statistics Netherlands) Supervisor/First Examiner: Dr. A. Zorlu (University of Amsterdam) Second Examiner: Dr. C. Hochstenbach (University of Amsterdam)
Date: 2017/06/26
Programme: Master Human Geography Graduate School of Social Sciences
University of Amsterdam Statistics Netherlands
2
Pr e f ace
2017/06/26 Dear Reader,
Hereby I present to you my master’s thesis. This thesis is my proof of competence to earn the title of Master of Science in Human Geography. During this master’s programme I specialised in Urban Geography.
My thesis is, unlike many others, a quantitative thesis. People often think quantitative research is inferior to qualitative research as it does not require the researcher to collect his own data. After writing this thesis I can assure anyone that writing a quantitative thesis is equally as hard as writing a qualitative thesis.
The hardest part for me however, was not even the writing. It was the thinking. Thinking about your thesis topic starts around October. Unaware of my quantitative ambitions back then, I started thinking about a qualitative subject, until something odd happened. Let me take you back to the moment when I knew for myself that I was going to write a quantitative thesis.
It was October 2016 and we just started our master’s programme. After completing an intensive course on quantitative statistics, we only had to turn in the final assignment. I was at home, behind my laptop like I am now, working my way through the data, testing out every possible combination of variables, just to see what happened. I got hungry so I turned the stove on and began to boil two large eggs. As eager as I was, I continued working on my assignment while waiting for the eggs to boil. A few moments later, I thought to myself, what is this noise and where is it coming from. I figured it was coming from the kitchen and as I walked up to the kitchen, both eggs exploded. Seventy-five minutes had passed since I started boiling my eggs. The eggs exploded, eggshells hit my glasses, while the rest scattered throughout the kitchen. From this moment on I knew I was going to write a quantitative research.
I hope you enjoy reading my thesis as much as I enjoyed writing it, Oscar Leemhuis.
4
Ab st r act
The young, high potentials are leaving their municipality of origin in the peripheral regions to realise their aspirations for higher education and job opportunities in the Randstad area (Dam et al, 2006). After finishing their educational career, the majority moves on to another municipality after experiencing rapid upward social mobility in an urban environment (Findlay et al, 2008). If people do not move back to their municipality of origin, this could result in an increased level of segregation between the highly-educated living in the core areas, while the lower educated reside in the more peripheral regions.
Up until now, the phenomenon of return migration has received little to no attention in the Netherlands. Academics, as well as policymakers are unaware of the extent to which this specific form of migration occurs among the working population in the Netherlands. This report addresses this knowledge gap by presenting the results of a quantitative study into the individual characteristics that select return migrants, as well as the regional specific characteristics that define an attractive municipality of origin to return to.
This research uses individual-level register data (N=180,471) from the System of Social statistical Datasets (SSD) of Statistics Netherlands. Conducting a multinomial logistic regression analysis shows that the family structure alongside the level of urbanisation are recognised as the most influential individual determinants of return migration. Conducting a linear regression analysis on the 388 Dutch municipalities proved that, controlling for all other variables, a higher percentage of owner-occupiers results in a higher amount of return migrants to the municipality of origin. A higher average income in the municipality of origin results in less return migrants to that municipality.
Recommendations for future research strongly lie in the further unravelling of return migration in the Netherlands because characteristics influencing whether or not people return to their municipality of origin differ for specific groups in society.
5
TABLE OF CONTENTS
ABSTRACT ... 4
1. INTRODUCTION ... 7
2. RESEARCH QUESTION & CONCEPTUAL MODEL... 10
3. LITERATURE REVIEW ... 13
3.1. INTRODUCTION ... 13
3.2. MIGRATION VS.RESIDENTIAL MOBILITY ... 13
3.3. THEORIES ON MIGRATION ... 14
3.3.1. Disequilibrium vs. Equilibrium approach ... 15
3.3.2. Human Capital ... 16
3.3.3. Social Capital ... 16
3.3.4. Escalator Effect ... 18
3.4. WHO MIGRATES WHEN?–CHARACTERISTICS THAT SELECT MIGRANTS ... 18
3.4.1. Leaving the Municipality ... 19
3.4.2. Return Migration ... 21
4. LEAVING THE MUNICIPALITY – METHODOLOGY ... 25
4.1. DATA ... 25
4.2. RESEARCH POPULATION ... 25
4.3. DEPENDENT VARIABLE –‘MOVING OUT’ ... 26
4.4. INDEPENDENT VARIABLES ... 27
5. LEAVING THE MUNICIPALITY – RESULTS ... 32
5.1. DESCRIPTIVE STATISTICS –MOVING OUT ... 32
5.2. MULTINOMIAL REGRESSION RESULTS ... 38
5.2.1. Gender differences? ... 39
5.2.2. Education induces Leaving the Region ... 41
5.2.3. The importance of geography ... 41
5.2.4. The family influence ... 42
5.3. LEAVING THE MUNICIPALITY –DISCUSSION ... 42
6. RETURN MIGRATION – METHODOLOGY ... 45
6.1. RESEARCH POPULATION ... 45
6.2. DEPENDENT VARIABLE –RETURN ... 46
6.3. INDEPENDENT VARIABLES ... 46
7. RETURN MIGRATION – RESULTS ... 54
6
7.2. MULTINOMIAL REGRESSION RESULTS ... 58
7.2.1. The Personal and Location Model ... 59
7.2.2. Achieved – 2015 ... 60
7.3. RETURN MIGRATION –DISCUSSION ... 66
8. REGIONAL ANALYSIS – METHODOLOGY ... 68
8.1. DATA ... 68
8.2. RESEARCH POPULATION ... 68
8.3. DEPENDENT VARIABLE –‘RETURN PERCENTAGE’ ... 69
8.4. INDEPENDENT VARIABLES ... 72
9. REGIONAL ANALYSIS – RESULTS ... 79
9.1. DESCRIPTIVE STATISTICS ... 79
9.2. LINEAR REGRESSION RESULTS ... 80
9.2.1. Model 1 – Demographic & Economic Model ... 80
9.2.2. Model 3 – Environmental ... 82
9.3. REGIONAL ANALYSIS –DISCUSSION ... 83
10. CONCLUSION AND RECOMMENDATIONS ... 86
11. REFLECTION ... 89
12. EPILOGUE ... 91
13. BIBLIOGRAPHY ... 92
7
1 . In t r od uct io n
According to data from Statistics Netherlands (2017a), the Dutch population is growing. On average, the population increases by 200 residents per day (Statistics Netherlands, 2017b). From this growing population ten per cent moves every year. These moves are not evenly distributed across the country, however. The Netherlands shows an increasing number of regions where the local population is diminishing. According to a recent publication by Statistics Netherlands (2016), 25% of the 390 Dutch municipalities shrank in the first half of 2016. The young, high potentials are leaving their municipality of origin in the peripheral regions to realise their aspirations for higher education and job opportunities in the Randstad area (Dam et al, 2006). After finishing their educational career, the majority moves on to another municipality after experiencing rapid upward social mobility in an urban environment (Findlay et al, 2008). If people do not move back to their municipality of origin, this could result in an increased level of segregation between the highly-educated living in the core areas, while the lower educated reside in the more peripheral regions. Furthermore, academic literature shows that attracting highly skilled workers is essential for municipalities to facilitate economic development (Niedomysl & Hansen, 2010). If they are thus not able to attract or retain their highly-educated, they will struggle to keep up with the economic development of the core areas. Academics acknowledge the tension between the core and the peripheral areas as the phenomenon of shrinking cities has received a lot of academic attention in the Netherlands (for examples see: Elzerman & Bontje, 2015; Rocak et al, 2016). Furthermore, the academic debate is also concerned with how to deal with urban shrinkage (for examples see: Groβmann et al, 2013; Haartsen & Venhorst, 2010). However, until now, no research is conducted into the socio-economic as well as ecological characteristics that drive young adults back to the municipality they once left. In the Netherlands, there is no knowledge on why young adults, in the midst of their career, would return to their municipality of origin, if they return at all. Stimulating interregional return migration by supplying an attractive set of characteristics in a municipality can result in a declining amount of shrinking municipalities and potentially even facilitate economic development. The current academic debate lacks such a research.
Data show that interregional return migration back to the municipality of origin is occurring, however. Up until now, the phenomenon of return migration has received little to no attention in the Netherlands. Academics, as well as policymakers are unaware of the extent to which this specific form of migration occurs among the working population in the
8
Netherlands. In the present research, return migration involves the return of interregional migrants back to their municipality of origin after leaving the municipality in a previous phase of their life. This specific form of migration, like any form of migration, is influenced by a set of individual and place specific characteristics (Faggian et al, 2015). However, no research in the Netherlands examines the individual as well as place specific characteristics that determine the return of previous leavers. By following the 1980 birth cohort of the Netherlands, this research tries to find the individual socio-economic characteristics that determine the return of adults, aged 35, back to their municipality of origin through a life-course perspective. The research follows this group from the age of fifteen until the age of 35 to see who leaves the municipality of origin and who returns to their municipality of origin. Furthermore, by researching a set of ecological characteristics for every Dutch municipality, this research hopes to find a set of regional specific characteristics that explain why leavers do move back to one municipality, but not to another. This research systematically explains the moving behaviour of a complete birth cohort in the Netherlands. This is a new approach in the debate on interregional migration.
The current knowledge gap is filled by presenting the results of a quantitative study into the individual characteristics that select return migrants, as well as the regional specific characteristics that define an attractive municipality of origin to return to. For policymakers, especially in the more peripheral regions of the Netherlands, gaining this specific knowledge is essential in the process of returning the ones who have previously left the municipality. The results potentially help municipalities combat further shrinkage of their population and counter the growing divide between the Randstad and its periphery.
The quantitative data that are analysed in this report are derived from the System of social Statistical Datasets (SSD) of Statistics Netherlands This system contains over fifty interoperable, anonymous registers. Data are longitudinal as they offer the ability to follow a population over time. Additional data come from Statline, a publicly available data source also managed by Statistics Netherlands.
This research report is structured as follows. First the research question and the conceptual model will be introduced. Thereafter the most important theories on migration are discussed in the literature review. After the literature review, the following chapters answer the research questions. To keep a clear structure, the three research sub-questions all have their own methodology and results chapter, followed by an answer to the problem statement
9
in question and a discussion of the results. The general conclusion answers the main research question, followed by the recommendations for future research and a critical reflection of the conducted research.
10
2 . Re se ar ch Que st ion & Con ce p t ual Mo d e l
To find out what region specific characteristics can explain the return migration of adolescents who left their municipality of origin in earlier phases of their life, this research addresses one main research question. This question is divided into three research sub-questions. The sub-questions are addressed in separate chapters of this research report followed by the general conclusion answering the main research question.
The main research question that is answered in this research is:
To what extent do socio-economic and ecological characteristics influence and determine the return migration of Dutch adolescents back to their municipality of origin?
To fully capture the process of return migration, the main research question requires three research sub-question. The first research sub-question delimits the research population that potentially returns by researching the socio-economic characteristics that influence leaving the municipality of origin. The second research sub-question helps answering the main research question by addressing the socio-economic characteristics that define the return migrant. The third and final research sub-question helps answering the main research question by defining a set of ecological characteristics of the Dutch municipalities that potentially accelerate the return of previous leavers.
Applying these three stages to the research allows the researcher to follow the birth cohort of 1980 through their life-course. Such a research has not yet been conducted with regards to interregional return migration in the Netherlands.
In chapters four and five, sub-question one is addressed. Returning to the municipality of origin requires someone to have left it in a previous phase of his or her life. What makes people leave their municipality of origin and in what respect do these people differ from the ones who stay in the municipality? According to the literature, leaving home is influenced by a variety of socio-economic as well as regional factors. These factors are categorised in three distinct categories, personal, household, and contextual characteristics. This brings the researcher to the following research sub-question:
To what extent do personal, individual household and contextual characteristics influence the likelihood of leaving the municipality of origin before the age of 22 among Dutch adolescents?
11
The next two chapters, chapters six and seven, discuss the second sub-question. After leaving the municipality of origin and the parental home, only a small part decides to migrate back to the municipality of origin. Knowing that migration is a highly selective process, some people are more likely than others to return to their municipality of origin, although reasons to return are potentially different for every individual. Therefore, the second step of this research ought to clarify which personal, household and contextual characteristics are the most influential in the decision to return. The following research sub-question is addressed in chapter six and seven:
What are the major determinants of return migration?
The outcome of the second research sub-question results in a percentage of return migrants for every municipality in the Netherlands. This percentage is then used in the analysis to answer the third research sub-question.
After analysing the combination of individual, household and contextual characteristics that makes people the most likely to leave from, and return to, the municipality of origin in the first two sub-questions, the chapters eight and nine answer the third research sub-question. Because migration decisions are potentially not only influenced by individual characteristics, but also by characteristics of the prospective residential location, a regional analysis of place specific characteristics is conducted to answer this research question. It uses municipal specific (ecological) characteristics to explain the amount of return migrants to that particular municipality. What place specific characteristics makes people return to their municipality of origin? Three categories are distinguished, demographic, economic and environmental characteristics. Analyses similar to the first and second step of this research have often been conducted by others, the third step however, approaches interregional return migration in a novel way. The third research sub-question that is addressed is the following:
12
To what extent can regional specific characteristics explain the return migration of Dutch adolescents who left their region of origin in the Netherlands?
The conceptual model below serves as a general overview of the research. It is divided into the three steps, all representing a different research question.
13
3 . Lit e r a t ur e Re v ie w
3 .1 . In t r od uct ion
To help answering the research questions as stated above, this chapter gives a broad overview of existing theories on migration and leaving home behaviour. The chapter first discusses the major theories of migration which relate to the third research sub-question. Thereafter, a more detailed discussion of individual characteristics inducing migration and return migration follows. Three important characteristics in the process of leaving the municipality and the parental home are discussed followed by a discussion of the three characteristics that induce return migration according to the theory.
The concept of migration is used differently in various studies. To get a clear understanding of what the present research means by migration, the difference between migration and residential mobility will be discussed prior to the other theoretical debates.
3 .2 . Mig r at ion v s. Re sid e n t ial Mob ilit y
Traditionally, migration has been conceptualized as a permanent affair. A person chooses to move from one location and settle in another. But not everyone who moves is considered a migrant. Migration is accompanied by a change in the daily activity space, that is, the spatial area contained within reasonable travel time from the residential location. Work, social contacts and every day activities are located within this space (Helderman, 2007). Earlier studies claimed that it is “the interactional system” of the migrant that must change for someone to be considered a migrant. This involves acquiring a new job, a new “set of friends, school for their children, and so on” (Petersen, 1978; 558). According to Petersen (1978) someone changing residence is only considered a migrant if it involves a transition into a new social environment. When one does not cross a cultural or societal boundary into a new interactional system, one is considered a mover (Petersen, 1978). According to early literature on residential mobility, it involves moving to a new residence because one is dissatisfied with its current one (Browne & Moore, 1970). The move is always over a short distance, often up to thirty kilometres, and is made to increase the residential satisfaction of the mover (Mulder, 2007).
Because of this change in the daily activity space, the costs of migration are significantly higher than the costs of a residential move. However, so are the potential results. Yet, the benefits of migration are not the same for everyone. Families will be more reluctant to migrate in comparison to people without children. For a family to migrate, one not only
14
changes their own daily activity space, but also the daily activity space of their spouse and children. Once children are in school, parents are less likely to migrate, because they are reluctant to make their children change schools (Helderman, 2007). As migration is accompanied by a change in environments, the costs are more than just financial. Migration often means a new working environment, a new social environment and a new daily environment. Replacing all those environments is what makes migration costly, and is therefore only undertaken if it renders more than it costs (Mulder, 2007).
Where more classical theories such as Petersen (1978) and Browne and Moore (1970) claim that motives for moving are predominantly concerned with residential dissatisfaction, more recent studies on both migration and residential mobility use a life course perspective to explain residential location choices. They claim that reasons for moving are not only, or always, related to dissatisfaction with the current residence. Dieleman and Mulder (2002) for example show that the probability of moving is strongly related to the stage in the life course. Events that induce migration do not occur randomly in life, but they are related to life course stages. Strong indicators for life course events are age and household composition (Helderman, 2007). The life course theory explains the current residential location by looking at previous residential locations, the origin of the individual, and major events that happened during someone’s life. “The life course of individuals is thus embedded in and shaped by the historical times and places they experience over their life-time” (Elder, 1998; 3). The present study uses a life course perspective to explain interregional migration patterns of individuals.
3 .3 . Th e or ie s o n Mig r at ion
The concept of migration is broadly studied. Multiple disciplines study people moving from one location to another. They explain migration from their own perspective, building upon established theories. This literature review discusses the geographic perspective on migration to find out what drives people to migrate. First, a classical theory on migration is discussed, the disequilibrium- versus the equilibrium approach. This theory still relates to current patterns of migration, as the present research will demonstrate. Thereafter, the human capital and the social capital theory are discussed, as well as the escalator effect theory. These ought to explain major migration flows and can be considered determinants of migration. After discussing those theories, this review briefly highlights the most important characteristics that select leavers and interregional (return) migrants respectively.
15
3 .3 .1 .Dise q uilib r ium v s. Eq uilib r ium a p p r oa ch
According to the disequilibrium model, interregional migration is simply a ‘by product’ of people searching for jobs (Schwartz, 1976). In a recent contribution to the migration literature, Faggian and colleagues (2015) show that, according to the disequilibrium model, wage differences between regions are the main cause of migration between those regions, where people move from low-wage regions to high-wage regions. People react to changes in unemployment and wage levels, by moving to these locations. Because of the inflow of workforce into the region, the disequilibrium restores itself, until a new disruption in the labour market causes a new disequilibrium in a different region (Faggian et al, 2015).
However, results from studies testing this hypothesis showed that people do not migrate in one direction. People also migrate from high-wage regions to low-wage regions and from low unemployment regions to high unemployment regions. The disequilibrium model is unable to explain these flows as it focusses solely on economic characteristics. Therefor an alternative view is established, which became known as the equilibrium approach (Faggian et al¸ 2015). This approach supports the idea that regional characteristics play a role in the explanation of migration flows. Location specific characteristics make one place more attractive than another. Regions with high-wages are often unattractive in the way that they cannot benefit from clean air or attractive landscapes, while low-wages regions can (Faggian et al, 2015). Wages thus compensate in a way for the attractiveness of the region. The present research builds upon this equilibrium approach in that it uses regional specific characteristics to explain interregional return migration of individuals. As Faggian and colleagues (2015) mention in their review article, previous empirical studies debunking the disequilibrium approach have shown that people do not only migrate to high-wage regions to make up the lack of workforce in a region. In other countries than the Netherlands, mostly the United States, multiple studies are conducted into the regional specific characteristics that determine migration to the specific region or municipality (Williams & Jobes, 1990). However, for the Netherlands such research is lacking. The present research is concerned with finding these regional specific characteristics for the Netherlands.
The equilibrium approach extends the human capital theory by highlighting the role of regional specific characteristics in the decision to migrate. However, human capital can still be considered one of the most important determinants of contemporary migration. The difference in mobility patterns of adolescents with high human capital and adolescents with low human capital is present in most studies (Faggian & McCann, 2009).
16 3 .3 .2 .Hum a n Ca p it a l
The accumulation of human capital throughout the educational and working career generally induces migration (Faggian & McCann, 2009). The term human capital, coined by Gary Becker in 1964, relates to a particular set of skills and characteristics acquired by an individual. Studies show that people with high human capital show increasing mobility as they are subject to higher opportunity costs of prospective locations (Faggian & McCann, 2006). Furthermore, high human capital individuals are more capable of collecting information from options elsewhere, which in turn reduces the risks of an unsuccessful move (Faggian & McCann, 2006). Businesses are constantly trying to attract human capital, as it is considered to increase its profit (Acemoglu & Autor, n.d.). The human capital theory thus implies that people with high human capital are more inclined to move to opportunity rich regions. Migrants are unlikely to move somewhere, unless there are relevant employment opportunities available at the selected location (Storper & Scott, 2009). The availability of human capital in the receptive region is considered a determinant of migration, as the agglomeration of human capital is proven to be a self-propelling process (Whisler et al, 2008). This complies with Faggian and McCann (2009) who claim that innovative regions are important recipients of human capital. Accordingly, a potential important task for regions is to develop an opportunity rich environment, as the people will follow. An increase in the years of schooling generally means an increase in human capital which in turn leads to an increase in mobility. Human capital is seen as an important determinant of migration. Next to human capital however, there is another type of capital that influences the (return) migration decision.
3 .3 .3 .Socia l Ca p it a l
As a child, one develops certain images of their residential environments which are often influenced by their feeling of belonging or attachment to the place (Blaauboer, 2011). Aero (2006) shows that people are more likely to return to particular residential locations if they have lived in similar type of housing before. Conducting a study in Denmark, he found that people who have lived in a particular type of housing before, were significantly more likely to return to similar neighbourhoods compared to their counterparts, who did not. It corresponds with earlier research from Van Dam and colleagues (2002), who also state that people who have experienced particular landscapes have other representations of those areas. Although Van Dam (2002) refers to the core periphery binary, as Aero (2006) shows, the same can be said for residential locations.
17
In the literature, this socialisation with residential locations is referred to as location-specific-capital (Mulder, 2007). This location specific capital is acquired during the residential period in a particular area. It consists of, but is not limited to, the attachment to the dwelling itself and the geographic location of the residential area with respect to work, friends and cultural facilities (Mulder, 2007). An increase in location specific capital potentially suggests a decrease in the propensity to migrate as one feels more attached to the location. Positive experiences during the childhood contribute to the increase of location specific capital, as is shown by Blaauboer (2011). She also proves that the residential environment during childhood has a strong influence on the current residential environment. Growing up in a suburban area increases the likelihood of living in a suburban area later in life by fourteen times, compared to people who grew up in a peripheral area (Blaauboer, 2011). Hypothetically speaking, location specific capital remains at the residential location, even after moving away. This suggests that when considering repeat migration, location specific capital can serve as a pull factor to the region of origin because of remaining ties there. DaVanzo and Morrison (1981) confirm this hypothesis. They conducted a longitudinal study among five thousand migrants in the United States and found that return migrants are twice as likely to return to their region of origin relative to a different region where they have lived before, presumably because the migrant has more location specific capital there than in other areas (DaVanzo & Morrison, 1981). However, the period between the initial and the return move due to remaining location specific capital is usually short. If the host region does not supply any location specific capital and the municipality of origin does, someone might decide to return. Yet in the present research there is a fourteen year time-period in which people can return, and therefore decisions are potentially not solely based on remaining location specific capital.
As strong as location specific capital may be, people often leave their municipality of origin to pursue their job- or educational-career. According to Storper and Manville (2006) adolescents are, more than other groups in society, pulled towards urban environments due to the availability of consumption based facilities, as will be discussed later. However, apart from consumption based facilities that potentially attract the young and college-educated to the cities, there are other motives for people to make the move to the city that are important to discuss in the light of return migration.
18 3 .3 .4 .Esca la t or Ef f e ct
According to the theory of the escalator effect introduced by Fielding in 1992, people who gain access to central highly-urbanised regions experience accelerated upward social mobility while living there (Venhorst, 2012). While there, young adults are able to increase their human capital at a faster rate than normal. A region would suffice as an escalator region if it fulfilled three conditions. The first condition is that it includes a region that attracts many young high potentials at the start of their working career. Second, through movement in the region’s labour and housing market, the in-migrants would be provided with accelerated upward social mobility (Champion, 2012). The third and final step of the escalator theory is stepping off the so-called escalator after gaining sufficient knowledge and skill (Findlay et al, 2008; Champion, 2012). This stepping of the escalator means moving out of the region to cash in on the relative prosperity one has acquired, in a low-cost but high-amenity region (Venhorst, 2012). Central regions serve as an escalator because of their opportunity rich environment, this can include the availability of jobs, but also having a university within municipal borders for students to start their career. Findlay and colleagues (2008) show that Scots move back from the South-East of England to Scotland in early stages of their career rather than later. After experiencing rapid upward occupational mobility in the 1990s, many Scots “no longer need to await retirement to be liberated from the expensive and constraining environment of life in the global city” (Findlay et al, 2008; 2184). They rather make use of the preferred quality of life amenities Scotland has to offer. This thus shows that migration decisions are not solely based on occupational opportunities, but also related to region specific amenities.
Cashing in on relative prosperity in a low-cost but high-amenity region suggests that people move away from places with a high average income to enjoy preferred quality of life amenities in other places, such as suburbs or peripheral municipalities. The lower income in such places is compensated by the higher level of amenities (Faggian et al, 2015). The remainder of this theoretical review is concerned with finding the individual characteristics that select the migrants that would use the opportunity to cash in on the potential relative prosperity they acquired in an escalator region.
3 .4 . Wh o Mig r at e s Wh e n ? – Char act er ist ics t hat select m ig r an t s
The (dis)equilibrium approach and the theories of human and social capital, together with the escalator effect, explain the major flows of migration. However, when deciding to
19
migrate, the individual takes different opportunities into account. The decision is based on multiple individual, household and regional characteristics which are on their turn shaped by the life course. These characteristics will be discussed below. First, three, in the literature often mentioned, characteristics facilitating the start of the independent housing career will be discussed, followed by the discussion of three individual characteristics potentially defining the return migrant.
3 .4 .1 .Le a v in g t h e M un icip a lit y
A prerequisite of return migration is moving out of the municipality of origin, leaving social capital behind. To get an idea of the population that is able to make the return move, the first step of this research is concerned with leaving the municipality of origin and the parental home. This process has been studied extensively by other authors such as Goldscheider and DaVanzo (1986), Aquilino (1991) and Mulder (2013). Most studies are occupied with finding characteristics that expedite home-leaving before the average age of 22 (Stoeldraijer, 2014). Three characteristics, that are important for this research, stand out. The most undisputed characteristic that influences the timing of leaving the parental home is attending tertiary education. Furthermore, different studies have proven the influence of the geographic location in the decision to leave the parental home (Buck & Scott, 1993). The third and final characteristic that is discussed here is the family structure. Aquilino (1991) found sound results that it influences home-leaving behaviour.
Ed u cat io n
Leaving home is often seen as part of the transition to adulthood (Goldscheider & DaVanzo, 1986). One of the characteristics that shows an undisputable positive relationship to early home leaving is the continuation of education after secondary school. This characteristic is one of the most influential in the decision to leave the parental home before the age of 22. In the Netherlands, a majority of the adolescents attending higher education leaves the parental home because institutes for higher education in the Netherlands are located in the larger cities. This thus makes it necessary for most children pursuing higher education, to move out of the parental home at a relatively young age.
A study from Mulder and Hooimeijer (2002) confirms that Dutch people attending university show a significantly higher chance of moving out of the parental home without a partner, relative to people who only finished primary education. This goes for men, and even more so for women. A study conducted in Sweden by Nilsson and Strandh (1999) proves that
20
a share of the children pursuing university education leave the parental home in a relatively short period after secondary education. The need to be close to the educational institution and the access to financial resources through student loans increases the nest-leaving rate for this group (Nilsson & Strandh, 1999).
Geo g r ap h ic lo cat io n
Growing up in a particular region influences the timing of leaving that region and the parental home. Buck and Scott (1993) conducted a study in the United States and found that differences occurred in the timing of leaving the parental home between adolescents growing up in different regions of the country. Furthermore, Blaauboer (2011) showed that people still living in their region of origin, are most likely to live in cities. This indicates that people from suburban and peripheral areas have more incentives to move away. Storper and Manville (2006) explain this by suggesting that urban environments offer facilities that more peripheral areas do not. “Those consumption based facilities can explain why the young and college-educated would live urban areas rather than in rural ones” (Storper & Manville, 2006; 1253). Different authors find different results concerning the direction of migration flows. Some theories argue that migration is a mere by-product of the search for jobs (Schwartz, 1976). This would imply that migration flows are directed from more peripheral regions to the urban regions as more jobs are concentrated in urban areas. However, other authors show the importance of regional characteristics in the migration decision. The example given earlier from the study by Findlay and colleagues (2008) of Scots moving out of the escalator region to enjoy an environment with a higher quality of life, is an example of this. This potentially means that migration flows can also be directed from the more urban regions to the peripheral regions. A general agreement however seems to be that more people leave from non-urban regions compared to highly urbanised regions (Storper & Manville, 2006; Blaauboer, 2011).
Fam ily St r u ct u r e
The family structure is an important determinant of leaving the region. Different authors found similar results. Aquilino (1991) shows that the family structure is an important determinant of leaving the region before the age of 22. Having a non-biological or non-intact family promotes home leaving, this relationship is even stronger for women than for men. Furthermore, Aquilino (1991) also proves that children who grow up in a non-intact family are more likely to establish an independent household after leaving the municipality of origin, while less likely to leave to attend higher education. This implies that even when a child from
21
a non-intact family does not leave for higher education, it is still inclined to leave the parental home before the age of 22, potentially due to the relationship with the parents. Avery and colleagues (1992) also show that children who experienced disruption in their parental family, appear to be less-family orientated themselves. Children from divorced parents are more likely to leave the parental home at earlier ages, but not for ‘union formation’. Instead they try to delay marriage through cohabitation, potentially due to bad experiences with marriage (Thornton, 1991). Next to the fact that children who grow up in a non-intact family have an increased chance of leaving, they potentially also have a diminished chance of returning, compared to children who grow up in an intact family. This relates to the theories of residential location choice from Blaauboer (2011) and Mulder (2007) who claim that childhood experiences influence the representation of the area. Growing up in a non-intact family potentially results in less social capital in the municipality of origin.
3 .4 .2 .Re t ur n M ig r a t ion
As often as migration has been studied, as little has interregional return migration been studied. Especially in the Netherlands, little is known about the extent to which people migrate back to their region of origin and what motives underlie this return. The main focus of authors researching return migration has been the international return of economic labour migrants. However, there are international authors who focus on interregional return migration. This paragraph will give a brief overview of the literature concerning interregional return migration by establishing three potential determinants thereof.
Ret u r n v s. On w ard
Motives for return migration differ from motives of onward migration (DaVanzo, 1983). Faggian and others (2015) claim that return migration is often seen as a corrective move if the new situation did not work out as expected. Marinelli (2013) shows that Italian onward migrants are often more successful relative to their returning counterparts. However, literature on international, economic return migration claims the opposite. Successful migrants return to invest in their home community (Gmelch, 1980). Yet, theory agrees on the fact that interregional return migrants differ from onward migrants. For return migrants, the following characteristics are considered important determinants according to the theory.
Dist an ce
When one moves across municipal borders into a new environment, the covered distance serves as a deterrent for return migration. According to Greenwood (1975), distance is the single most deterrent to migration. Increasing distance results in a decreasing ability to
22
gain sound information about the potential new location. Moves over great distances are therefore more likely to be followed by a ‘corrective’ move, either onward or return (DaVanzo, 1983). Distance furthermore “serves as a proxy for both the transportation and psychic costs of movement” (Greenwood, 1975; 398). The psychic costs that are involved with moving over greater distances prevent people from making the move. Psychic costs are related to the departure from friends and family and can be transformed to the time and absolute transportation costs involved in continuing to visit them, when deciding to move over greater distances (Schwartz, 1973). Therefore, according to Schwartz (1973), the increasing distance between two regions diminishes the migration flows to both regions. For the present research this implies that, although the distance is already covered once, the greater the initial distance covered when leaving the region, the less likely someone is to return. Research into return migration by Smeulders and colleagues (2009) supports this hypothesis by showing that increasing distance from the region of origin leads to less return migration by Dutch retirees. However, increasing distance leads to an increased probability of moving onwards (Smeulders et al, 2009). Nevertheless, DaVanzo (1983) finds an exception to the “distance-decay” relationship when researching longitudinal data on repeat migration of 5,000 American families. She found that the more distant the location of the return move is, the higher the probability that someone returns from this location. With regards to distance, the results of the present research show whether moving over great distance results in more or less return migrants to the municipality of origin. Such a study has not yet been conducted in the Netherlands for people aged 35.
Ed u cat io n
Increasing distance shows to be an important characteristic to influence return migration, while another important characteristic is the level of education. According to Schwartz (1976), after leaving the region of origin people invest in their level of education at the beginning of their career, in order to obtain better pay and more prestigious jobs in later stages of their career. In the same article based on classical theories, Aba Schwartz (1976; 706) claims that “migration in a homogeneous country is primarily a search for jobs.” Jobs that require higher educated individuals are more centralised in urban areas while the more peripheral regions supply jobs for other types of education. Because higher education requires more specialised jobs, these jobs are less common and people thus have to increase their search radius for a suitable job. This means that people with a higher level of education focus more on national or even international labour markets relative to people with lower levels of
23
education, who focus more on regional labour markets (Faggian et al, 2015). This implies that lower educated adolescents coming from more peripheral municipalities potentially show higher probabilities of returning versus the higher educated individuals coming from the same peripheral municipality.
Furthermore, as migration requires the collection of information from the prospective location, the lower educated are more likely to return in comparison to the higher educated. According to Faggian and colleagues (2015) this is due to the increased ability of the higher educated to gain sound information of prospective locations. For the present research, this means that the lower educated potentially show higher probabilities of returning.
Mar it al St at u s & Hav in g Ch ild r en
A third and final characteristic that influences migration probabilities is marital status combined with the family structure. Mobility is ‘triggered’ by marital-status change and even more by the addition of children (Clark & Huang, 2003; Kulu, 2008). Kley and Mulder (2010) claim that marriage and cohabitation are among the most important motives for long-distance moves in early adulthood. However once married, people migrate less compared to their unmarried counterparts. This is potentially due to married people more strongly relating to the dwelling and the environment since they have to take two careers into consideration when deciding to migrate (Kley & Mulder, 2010). Migration by married people requires the change of two daily activity spaces, therefore families and married couples are more reluctant to migrate (Helderman, 2007).
Kulu (2008) shows that having children induces the move to a more spacious dwelling. Conducting a study in Austria, Kulu (2008) found that couples expecting their first child show significant higher probabilities of moving to a more peripheral region. Furthermore, Kulu (2008) also showed that having children reduces the probability of moving to an urban area, but rather increases the probability of moving from an urban area to a more peripheral area. Both these findings are related to the family structure and the amount of people per household. The findings by Kulu (2008) are supported by Kim and colleagues (2005). They show that having one more child in a household increases the probability to migrate by 6.3%. However, not everyone agrees. Clark and Huang (2003) for example find no significant result for children raising the probability of migration on the British housing market.
24
Ot h er p o t en t ial in d iv id u al d et erm in an t s
Next to the three individual characteristics described above, there are other characteristics that require attention as well. The first one is tenure. Whether someone rents or owns their home, influences the probability to migrate, according to Helderman (2007). Homeowners are less likely to migrate as moving from an owner-occupied home results in increasing moving costs in comparison to moving from a rented home. However, research into Dutch return migration by retirees from Smeulders and colleagues (2009) shows no difference between owners and renters in the amount of return migrants.
The second additional characteristic is the level of urbanisation of the municipality. People who initially moved to municipalities with lower urban densities are more likely to return compared to people who moved to high urban density municipalities (Smeulders et al, 2009). According to Smeulders and colleagues (2009) this is potentially due to the amount of facilities that is offered by urban environments. This is similar to what Storper and Manville (2006) found, as discussed earlier.
After the discussion of the theory, it becomes clear that this research is predominantly concerned with finding the regional specific characteristics that influence return migration back to the municipality of origin as this has never been done before for the Netherlands. With regards to the individual characteristics determining leaving the municipality, this research aims to either confirm with the current literature or potentially find other characteristics which are not yet confirmed by the literature that expedite this process. The same is true for the individual characteristics surrounding the return of previous leavers.
The following chapters answer the research questions as they are stated above. In addition to an explanation of the methodological choices made during the research, the chapters also introduce the variables that are used in the model, taking additional theoretical considerations into account. After the methodological part, each chapter continues with the presentation of the results. First the descriptive statistics are presented followed by a discussion of the regression analysis. The chapters conclude by answering the research question and giving recommendations for future research.
25
4 . Le av in g t h e Mun icip alit y – Me t h od ology
This chapter validates all the methodological choices made for the first phase of this research, the phase of leaving the municipality of origin. The research sub-question that is answered in this chapter is the following:
To what extent do personal, individual household and contextual characteristics influence the likelihood of leaving the municipality of origin before the age of 22 among Dutch adolescents?
First the data and the research population are discussed followed by the introduction of the dependent variable. Thereafter the independent predictor variables are introduced. The next chapter discusses the results of the descriptive statistics as well as the regression results.
4 .1 . Dat a
To answer this sub-question the present research uses individual-level register data from the System of Social statistical Datasets (SSD) of Statistics Netherlands (Bakker et al, 2014). The SSD contains over fifty interoperable, anonymous registers. Every Dutch individual registered in the Dutch population register, can be found in the SSD. This results in a comprehensive collection of all sorts of information from the complete Dutch population, unregistered individuals excluded (Bakker et al, 2014). One of the biggest advantages of using this type of data to conduct scientific research is that it allows researchers to study rare events like return migration more easily than for example survey data would. Furthermore, register data is arguably the most reliable source of longitudinal data and it reduces the time and costs of collecting the data (Bakker et al, 2014). A potential downside to this type of data is that people deliberately register themselves somewhere else to profit from certain benefits, such as an increased financial benefit for students living away from their parents. Taking small potential errors like this into account, the benefits of using this type of data still outweigh the use of any other type of data.
4 .2 . Re se ar ch Pop ula t ion
The dataset contains information for every individual registered in the Dutch population register as a child living with its parents by 1995, born in 1980 somewhere in the world. The research population consists of everyone within this group still living in the Netherlands in 2001. This resulted in 180,471 units of analysis of which 82.3% is native Dutch, while the other 17.7% has a migratory background. Statistics Netherlands offers longitudinal data for the years 1995-2015. The birth cohort of 1980 is therefore the best
26
possible research population to answer research sub-question one and two, as this population reached the age that is best suitable for researching the specific concepts at the particular reference years. Older birth cohorts are more likely to have moved out of the parental home by 1995 already, while using younger cohorts would confine the researcher in conducting relevant analysis considering phase two, return migration, as the research population would simply be too young then.
Leaving the municipality of origin is operationalized by crossing municipal boundaries between 1995 and 2001 and having left the parental home within the same period. To research who is the most likely to leave the municipality of origin, the research population is researched at two points in time, 1995 and 2001. The research thus uses a longitudinal research design, and because the research population is constructed from a cohort, this research studies aging effects, “as all members of the sample are born at more or less the same time” (Bryman, 2012; 65). The first reference year is 1995. By this year, the research population has reached the age of 14 or 15. This has to do with the measurement date of the data, September 30th.
The second reference year is 2001. The research population reached the age of 20 or 21 by this time. Data shows that almost 35% of the research population moved out of the parental home by this year. The average age of moving out of the parental home in the Netherlands is 22 (Stoeldraijer, 2014). Motives for moving out at younger ages differ significantly from motives by other groups of leavers (Tang, 1997). Therefore, this research only researches leavers who left before the average age. This choice is furthermore made to give the research population the most time possible to decide to return or move onward. Already in 2001, a significant part of the research population moved out.
The following part of this chapter discusses the different variables that are used for this phase of the research, followed by the results of the descriptive statistics as well as the regression analysis.
4 .3 . De p e n d e n t Var ia b le – ‘Moving out ’
For the research question to be answered, the dependent variable ‘moving out’ is constructed. This variable is a categorical variable with three categories. Table 1 presents the distribution of the research population among those categories. In 2001, almost two thirds of the research population is still living with their parents. Twenty per cent moved out of the
27
municipality while 43% of the population that moved out of the parental home stayed within the municipality.
Ta b le 1 : Mo v in g o u t o f t h e m u n icip a lit y b y 2 0 0 1
Different conditions apply to the different categories of this newly constructed dependent variable ‘moving out’. The variable is constructed for everyone residing in the Netherlands in 2001. This excludes people who migrated to a foreign country, died in the period between 1995-2001 or for other reasons show missing data, hence the 180,471 observations. The variable has three categories. The first category ‘living with parents’ includes all people still living with their parents in 2001. The second category ‘Moved out within Municipal borders’ includes all the people not living with their parents by 2001, but who remained in the same municipality. The third and final category includes all the people not living with their parents and in a different municipality compared to 1995. The choice for a dependent variable with three categories, and not two, is made because motives for moving out within and across municipal borders differ significantly. The reference group would be too heterogeneous if the people who moved out of the parental home but within the municipality would be categorised with the people still living with their parents.
4 .4 . In d e p e n d e n t Var iab le s
The choice to move out of the parental home, either across or within municipal boundaries, is influenced by a variety of personal and individual household characteristics, which are introduced in the following section. A theoretical consideration is followed by an explanation of the variable and its categories to distinguish the importance of using the variable in the model.
Gen d er
The first predictor variable is gender. Statistics on leaving the parental home show differing numbers for men and women. In general, women leave the parental home at a younger age compared to their male counterparts (Stoeldraijer, 2014). According to Klimstra and colleagues (2009) this is due to the fact that women go through a process of maturation earlier than men do. This research explores the differences stated by the literature regarding gender and moving out across municipal borders.
N %
Living with Parents 118,073 65.42 Moved out within Mun. borders 26,672 14.78 Moved out across Mun. borders 35,726 19.8
28
Lev el o f Ed u cat io n – 2001
The most undisputed characteristic to influence the timing of moving out of the municipality of origin and leave the parental home is the level of education of the individual. A large part of the studies, using level of education as one of the predictor variables for moving out, finds similar results. If children pursue tertiary education after finishing secondary education, the likelihood of moving out of the parental home across municipal borders significantly increases (Nilsson & Strandh, 1999, Mulder & Hooimeijer, 2002). Furthermore, Goldscheider and DaVanzo (1986) proved that leaving the parental home is more related to higher education and gaining independence than it is to, for example, a full-time job or parenthood.
Data features different variables to measure educational attainment of the research population. For the year 2001, data contain the highest finished level of education as well as the highest level of education someone is enrolled in or has ever been enrolled in. Because the research population is only 20 or 21 years of age by the year 2001, it is not likely for someone to have finished a university degree or higher vocational training. For this reason, the level of education someone is enrolled in or has ever been enrolled in is used to measure the level of education. The variable contains four categories:
1) Primary Education
2) Intermediate Vocational Training (Dutch: Middelbaar Beroeps Onderwijs) 3) Higher Vocational Training (Dutch: Hoger Beroeps Onderwijs)
4) University Education
Universities are the least frequent in the Netherlands and therefore university students are triggered the most to leave the municipality of origin, followed by students of higher vocational training. For intermediate vocational training this trigging is near to nil as institutions for intermediate vocational training are equally spread across the country.
Lev el o f Ur b an isat io n o f t h e Mu n icip alit y – 1995
A third, important, predictor of leaving the municipality of origin is the level or degree of urbanisation of the municipality one moves out from. It is well-known that peripheral regions struggle with retaining their higher educated, also in the Netherlands. In 2006, Dam and colleagues showed that young, high potentials in peripheral regions are moving towards the Randstad to realise their aspirations for higher education and job opportunities. But not only in the Netherlands this periphery-core migration is an on-going process. Faggian and
29
McCann (2006) find similar results for the UK. Storper and Manville (2006) claim this is due to the facility level that cities and urban environments offer compared to the more peripheral environments.
This research uses the ‘level of urbanisation of the municipality 1995’ as an independent, predictor variable potentially influencing the likelihood of people moving out of the municipality, away from their parents. The variable contains five categories
1) Highly Urbanised (>= 2,500 addresses per sq. kilometre) 2) Urbanised (1500-2500 addresses per sq. kilometre)
3) Moderately Urbanised (1000-1500 addresses per sq. kilometre) 4) Little Urbanised (500-1000 addresses per sq. kilometre)
5) Not Urbanised (<500 addresses per sq. kilometre)
According to the literature, expected results are, the lower the degree of urbanisation, the higher the percentage of out-movers.
NUTS 1 Reg io n s – 1995
A predictor variable that is related to the degree of urbanisation in the Netherlands, is the NUTS 1 region (Dutch: Landsdeel). The Netherlands is divided into four NUTS 1 regions, north, east, south, and west. The western part contains the Randstad, the only mega-city-region of the Netherlands. Logically, this mega-city-region supplies more jobs compared to other regions. Furthermore, the western part of the Netherlands houses multiple universities and all four of the big Dutch cities (Amsterdam, Rotterdam, Utrecht and The Hague). Because of the geographic advantage of growing up in the western part of the Netherlands, for students as well as for employees, it is important to use the NUTS 1 region as a control variable. The table shows the population distribution among the four regions. Almost half of the population is living in the western part of the Netherlands.
Fam ily St r u ct u r e – 1995
The fifth predictor variable is the type of household one descents from. According to the literature, the type of household influences the timing of leaving the parental home. Growing up in an intact family with married parents reduces the likelihood of leaving the
Ta b le 2 : NUTS 1 Re g io n s N % North 20,641 11.1 East 41,228 22.17 West 83,322 44.81 South 40,756 21.92
30
municipality early, while growing up with a stepparent or in a single parent household, the urge to leave the parental home is found to turn into actual behaviour, earlier (Aquilino, 1991).
Data provide a variable with different types of households. However, data are limited to the extent that parents are not linked to their children. The present research is unable to study whether growing up with a stepparent increases the likelihood of moving out. Yet the present research identifies differences between unmarried and married couples with children, as well as single parent households. Other categories are omitted from the variable as one of the conditions for the first research sub-question is that the children live with their parents in 1995.
Et h n icit y
The sixth predictor in the model is the ethnicity of the research population. A study from Zorlu & Gaalen (2016) showed that Dutch adolescents with a Turkish or Moroccan background leave the parental home earlier, compared to their Dutch native counterparts. However, the groups with a migratory background often migrate within their municipality, while native Dutch more often move across municipal borders. The study from Zorlu & Gaalen (2016) also showed that the reasons to leave the parental home differ among the ethnic groups. The ‘ethnicity’ variable as it is used in the present research contains five categories.
1) Dutch 2) Moroccan 3) Turkish 4) Surinamese 5) Other
In the Netherlands, migrant populations are concentrated in the four big cities. This makes it relatively easy to migrate within municipal borders. Growing up in a small peripheral municipality makes moving across municipal borders easier for the people living there. Therefore, it is important to control for the next variable in the model.
Size o f t h e m u n icip alit y – 1995
It is important to control for the (population) size of a municipality as different patterns of leaving the municipality of origin and the parental home occur when children grow up in a city. As explained for ethnicity, the motives for leaving the parental home can differ
31
significantly when growing up in one of the four big cities, or one of the other relatively big municipalities in the Netherlands. Logically speaking, the larger a municipality, the more options exist for moving out within municipal borders.
Therefore, a new variable is constructed controlling for the size of the municipalities. It contains three categories:
1) One of the big four cities (Amsterdam, Rotterdam, Utrecht, The Hague) 2) One of the other 37 relatively big municipalities
32
5 . Le av in g t h e Mun icip alit y
– Re sult s
This chapter presents and discusses the results of the analysis conducted to answer research sub-question one as stated above. An extensive description of the statistics will be presented to get a good sense of how the data is constructed. Thereafter the results of the multinomial logistic regression analysis will be discussed. The dependent variable’s categories relate to the year 2001, presenting results of adolescents born in 1980, thus moving out before the age of 22.
5 .1 . De scr ip t iv e St at ist ics – Mov ing out
This paragraph discusses the descriptive statistics for moving out of the parental home as they are presented in Table 3. For the dependent variable ‘moving out’, 65.4% did not move out of the parental home by 2001, at age 20/21. Of all the adolescents leaving the parental home, 57.3% decides to move across municipal borders, while the others, making up 14.8% of the total population, decide to move within the municipality. The remainder of this paragraph discusses the variables in Table 3 separately. Extra attention is paid to the important characteristics.
Ta b le 3 : De scr ip t iv e st at ist ics f o r t h e d e p e n d e n t v a r ia b le 'm o v in g o u t ' a cco r d in g t o it s ca t e g o r ie s.
Living with Parents Moved Within Moved Across
Moving out by the year 2001 N % N % N %
118,073 65.42 26,672 14.78 35,726 19.8
Gender
Male 68,747 74.64 9,660 10.49 13,701 14.87
Female 49,320 55.82 17,009 19.25 22,022 24.93
Level of education (weighted) 2001
Primary Schooling 14,557 70.78 4,041 19.65 1,969 9.57
Intermediate Vocational Training 61,063 71.33 14,134 16.51 10,414 12.16 Higher Vocational Training 33,831 63.94 5,984 11.31 13,097 24.75
University 7,030 35.93 1,067 5.45 11,470 58.62 Level of Urbanisation 1995 Highly Urbanised 19,350 59.3 9,195 28.18 4,084 12.52 Urbanised 34,352 65.16 8,792 16.68 9,578 18.17 Moderately Urbanised 22,316 67.58 3,581 10.84 7,126 21.58 Little Urbanised 29,696 67.98 3,723 8.52 10,264 23.5 Non-Urbanised 12,350 67.16 1,381 7.51 4,659 25.33 NUTS 1 Region 1995 North-Netherlands 11,440 56.69 3,094 15.33 5,647 27.98 East-Netherlands 26,388 65.63 5,406 13.45 8,411 20.92 West-Netherlands 53,279 66.21 13,294 16.52 13,902 17.27 South-Netherlands 26,957 68.1 4,878 12.32 7,751 19.58 Family Structure 1995
