Elsevier

Social Science Research

Volume 52, July 2015, Pages 351-369
Social Science Research

Does residential mobility improve educational outcomes? Evidence from the Netherlands

https://doi.org/10.1016/j.ssresearch.2015.02.008Get rights and content

Highlights

  • Residential mobility increases school dropout probability in the short and long run.

  • In the medium run, residential mobility decreases the probability of school dropout.

Abstract

This paper explores the impact of residential mobility on educational outcomes. By considering a large Dutch city with substantial internal residential mobility, we examine how residential mobility influences the decision of students to drop out of school. The paper exploits a rich administrative dataset with extensive information on educational, individual, family, housing and moving characteristics of students. It combines a matching design with a multivariate regression analysis, such that the evaluation draws on a well-comparable control group for the treated students. Accounting for individual, family, educational, neighborhood and housing characteristics, as well as for school and year fixed effects, we observe that residential mobility increases the probability of school dropout in the first few years after moving. The estimated effect changes, however, to a lower risk of early school leaving after an initial period, and then changes again to a higher risk after 6 years. This effect remains, regardless the level of education the students attended, or whether the student moves to a better or a worse neighborhood.

Introduction

Residential mobility is a frequently occurring phenomenon, which might take place due to several reasons, such as change of jobs, marriage or divorce, addition to the family, or simply because people want a large house. Previous literature has shown that people’s economic situation or education level influences residential mobility, welfare and other forms of social mobility (Boschma and Fritsch, 2009, Haveman and Smeeding, 2006 and references therein). However, the effect of residential mobility on educational outcomes attracted only limited attention of scholars. This is mainly due to serious methodological issues. Measuring the causal effect of residential mobility on (educational) outcomes is for most settings difficult due to endogeneity issues arising from unobserved characteristics. The latter arise as individuals with residential mobility are likely to have other (observed and unobserved) characteristics and backgrounds compared with individuals without social mobility. By exploiting a setting in which residential mobility can be naturally observed and in which a proper control group can be defined, this paper constructs evidence on the causal effect of social mobility on educational outcomes.

The literature makes a distinction along different types of social mobility, including intergenerational mobility (e.g. Checci and Flabbi, 2007, Werfhorst van de, 2002), occupational mobility (e.g. Kambourov and Manovskii, 2009), financial mobility (e.g., thanks to winning lotteries), intellectual mobility (e.g. Burt, 2011) or residential mobility (e.g. Gasper et al., 2010). This paper focuses on the influence of residential mobility on education outcomes. Earlier literature is inconclusive on the sign and magnitude of this effect, mainly because the reasons for residential mobility are diverse. Positive effects of residential mobility on education have been related to the upward economic mobility after World War II (Gasper et al., 2010, Rossi, 1955), whereas harmful effects are argued to be caused by ‘forced’ mobility due to unstable families and divorce (Amato, 2000, McLanahan and Sandefur, 1994, Morrison and Coiro, 1999, Speare et al., 1975). Besides on education, the impact of residential mobility has been explored on health (Larson et al., 2004) and adolescent delinquency (Gasper et al., 2010), for which in both cases negative correlations have been observed. Using the longitudinal data of the National Education Longitudinal Study (NELS), Swanson and Schneider (1999) made a distinction between residential mobility and school mobility. They observed that both residential and school mobility are associated with higher rates of early high school dropout.

Earlier literature on the influence of residential mobility on educational outcomes has some serious flaws which this paper aims to tackle. First, endogeneity issues arise in many studies as social mobility and educational outcomes might be affected by common unobserved characteristics (see e.g. Gasper et al., 2010). In other words, residential mobility is not random among individuals. Tucker et al. (1998) avoid this endogeneity issue by making the analysis conditional on the family structure (e.g., divorced, two-parents). In doing so, they assume that the reason for the move is homogeneous within a family type. In contrast with Tucker et al. (1998), our study accounts for non-random residential mobility by conditioning on a wide set of individual and family characteristics in a matching design. This is a more comprehensive approach as we include all households who move in a certain year (and thus avoid selection and the necessity of strong underlying assumptions) and construct a proper control group such that the results can be interpreted in a causal way.

More precisely, to identify the effect of residential mobility on educational outcomes, we proceed in two steps. In a first step, we construct for the students with residential mobility in 2005 (i.e., treated students) a matched sample of students without residential mobility (i.e., untreated students). By matching on a wide range of observed characteristics from a rich data set, we argue that also the unobserved characteristics are similar for the matched sample. In a second step, we estimate by a simple multivariate regression analysis how the treatment status influences the education outcome status. The identification strategy allows us to study the following research question: Are the education outcomes for students with residential mobility different from the counterfactual students in the matched control group?

A second issue in the previous literature arises from the diversity of educational outcomes. This study focuses on educational failure as an outcome. School dropout has been an important issue in national and international politics over the past decennium as it is a clear signal of an incomplete educational process. Within Western countries, governments aim for a significant reduction of early school leaving (e.g., The Horizon2020 Agenda within the EU, the No Child Left Behind Act in the US). By focusing on early school leaving as an outcome variable, we study the most disadvantaged students who are heavily ‘at risk’. For similar people with poor educational outcomes, the influence of social mobility can be most prevalent. Due to our focus on school dropout, the paper is close to earlier work by Astone and McLanahan (1994) who explored to which extent residential mobility among non-intact families leads to school dropout. They observed that “as much as 30% of the difference in the risk of dropping out between children from stepfamilies and children from intact families can be explained by differences in residential mobility” (p. 576). Our paper differs from Astone and McLanahan (1994) as we do not focus on households with negative decisions for residential mobility (e.g., due to a divorce), but on the contrary on both negative and positive decisions for residential mobility. We argue below that we can make this distinction by exploiting mobility in a city with clear and strong internal migration patterns.

Third, social mobility in general and residential mobility specifically are often imprecisely measured in the previous literature, giving rise to measurement errors. This paper draws on rich and unique administrative data, including neighborhood and housing information (e.g., house value and ownership). The advantage of focusing on residential mobility is that for each student a sound and uniform definition of social mobility emerges. Moreover, the data arise from a new town (i.e., a town specially constructed for low and medium income households at commuting distance of a large town) in which residential mobility is common and highly valued. Moving to another neighborhood with higher housing values is in the case of this new town perceived as an improvement in social hierarchy and, thus could be perceived as upward mobility (see Section 3 for a discussion).

We believe that residential mobility can change educational outcomes in the short run (i.e., reduce early school leaving) by three mechanisms. First, by moving to a new neighborhood, the student experiences the influence of new peers (see literature on peer and neighborhood effects, e.g. Black et al., 2010, Clark et al., 2012, Sharkley, 2012, Sharkley and Elwert, 2011, Staff and Kreager, 2008), which can work both positive and negative, depending on where the student moves. Second, by moving to a different neighborhood, the direct shock of moving and large changes can increase the chance of dropout. On the other hand, depending on where the student moves, moving into a new neighborhood might also allow students to ‘dream’ about a better position in life, motivates them for schooling, and can trigger changes in the student’s aspirations and behavior, which can have an influence on educational outcomes. In the long run, however, it could become clear that the better position in life that the student dreamt about is not happening, and the student could fall back to his/her old habits, regardless of the peers. Third, the decision to dropout may be motivated by the different skills associated with an education tract (e.g. vocational education, general education or pre-university education) and the labor market returns to those education tracts. As there is not much mobility between education tracts in the Netherlands, this lead to a negative effect of residential mobility.

The main outcomes of this paper indicate that, while controlling for individual, family, educational, neighborhood and housing characteristics, as well as for school and year fixed effects, we observe that residential mobility increases the risk of dropping out in the first few years after moving. This favorable outcome, however, changes to a lower probability of early school leaving as time passes, and changes back to a higher probability after six years. The effects are most visible for students that encounter only a little change in the housing value and the neighborhood characteristics. Our short term findings are in line with earlier work by Swanson and Schneider (1999) who observed that residential mobility increases dropout rates. Furthermore, we also find that in the long run residential mobility increases dropout rates again. Additional analyzes indicate that our results hold, regardless of whether the student has moved up or down (calculated by change of housing value), and for students in vocational and higher education. We do not find any results for students in primary and lower secondary education.

The remainder of this paper is structured as follows: Section 2 presents the literature and theoretical framework. Section 3 explains the Dutch educational system and the setting of this particular ‘new town’, which we exploit as a case for residential mobility. Section 4 describes the data, the indicators constructed from the data, and gives some descriptive statistics. Section 5 discusses the matching procedure and its use in the study at hand. Section 6 presents the results of the analysis and provides robustness analyzes. The paper ends with some concluding remarks and a discussion in Section 7.

Section snippets

Social mobility

Social mobility is a broad concept that is used for many types of changes that increase or decrease the social status of an individual. The literature on social mobility falls apart into two strands. First, it examines what influences social mobility. Similar studies conclude that a higher educational degree increases social mobility (see overviews in Hensen et al., 2009, Machin et al., 2012, Malamud and Wozniak, 2008) and that higher transaction costs and inequality decrease social mobility

School dropout in the Netherlands

In the Netherlands, pupils attend primary education between ages 4 and 12 and secondary education until a higher secondary degree is obtained. Fig. 1 depicts a self-constructed graphical representation of the Dutch educational system. One can distinguish five different levels of education: practical training education (pro), prevocational secondary education (vmbo), vocational education (mbo), general upper secondary education (havo) and pre-university education (vwo). Pupils enter a level of

Data

This study uses an unbalanced panel data set of 32,468 pupils at compulsory education age until 18, between the school years 2005 and 2010. Because the registration of dropouts between 18 and 23 is often not very accurate (see Cabus and De Witte, 2011), we have decided to only include individuals up to the age of 18 (in 2005) in our analysis, making them a maximum of 23 years old in 2010. To causally identify the effect of residential mobility on educational outcomes, we construct a control and

Constructing a valid reference group

The previous section indicated that there are differences in category of instruction, age, gender, family size, number of residency years in the Netherlands, and some types of ethnicity and education between students with and without a residential move. These differences might bias the analysis. To avoid this pitfall, we use a nearest neighbor (NN) matching procedure to identify for each student in the treatment group a similar student in the control group. Note that alternative matching

Results

The empirical results of estimating equations (3), (4), (5), (6) are presented in Table 4. As outlined in Section 5, each of the specifications has its own merits and assumptions. Although the underlying data originate from the matched sample, we control in the regressions for a wide range of control variables. In particular, we account for the age, gender, family size, ethnicity (i.e., native Dutch, Antillean, Surinam, Turkish, Moroccan, African, Other non-Western), type of education (i.e.,

Conclusion and discussion

The influence of residential mobility on education outcomes has attracted only limited attention in previous literature. This paper studied the effect of residential mobility on school dropout in a large new town in the Netherlands (more precisely, the seventh largest city in the Netherlands). As new towns are specially constructed for low and middle income groups and as they are characterized with continuous attempts to improve the living circumstances in the neighborhoods (e.g., by renewing

References (98)

  • P.R. Amato

    The consequences of divorce for adults and children

    J. Marriage Family

    (2000)
  • M. Astone et al.

    Family structure, residential mobility and school dropout: a research note

    Demography

    (1994)
  • M. Bartley et al.

    Increasing social mobility: an effective policy to reduce health inequalities

    J. R. Stat. Soc.

    (2007)
  • Black, S.E., Devereux, P.J., Salvanes, K.G., 2010. Under pressure? The effect of peers on outcomes of young adults....
  • R.A. Boschma et al.

    Creative class and regional growth: empirical evidence from seven European countries

    Econ. Geograpy

    (2009)
  • S. Bowles et al.

    The inheritance of inequality

    J. Econ. Perspect.

    (2002)
  • L.A. Brown et al.

    The intra-urban migration process: a perspective

    Geografiska Annaler B

    (1970)
  • C. Burt

    Intelligence and social mobility

    J. Stat. Psych.

    (2011)
  • Cameron, A.C., Trivedi, P.K., 2005. Microeconometrics – Methods and Applications. Cambridge University...
  • CBS, 2012. Statline Numbers on Internal Migration in Municipalities,...
  • Checci, D., Flabbi, L., 2007. Intergenerational mobility and schooling decisions in Germany and Italy: the impact of...
  • J. Chen

    Housing tenure, residential mobility and adolescents’ education achievement: evidence from Sweden

    Ann. Reg. Sci.

    (2013)
  • W.A.V. Clark et al.

    Households and Housing: Choice and Outcomes in the Housing Market

    (1996)
  • C. Clark et al.

    Mobility, housing stress and neighbourhood contexts: evidence from Los Angeles

    Environ. Plann. A

    (2006)
  • W.A.V. Clark et al.

    Residential mobility and neighbourhood outcomes

    Housing Stud.

    (2006)
  • C. Clark et al.

    Do peers influence achievement in high school economics? Evidence from Georgia’s economics end course test

    J. Econ. Educ.

    (2012)
  • J.S. Coleman

    Social capital in the creation of human capital

    Am. J. Sociol.

    (1988)
  • J. Crane

    The epidemic theory of ghettos and neighborhood effects on dropping out and teenage childbearing

    Am. J. Sociol.

    (1991)
  • De Witte, K., Van Klaveren, C., Smets, A., 2015. Selective migration in new towns: influence on regional accountability...
  • K. De Witte et al.

    Dropout prevention measures in the Netherlands, an explorative evaluation

    Educational Rev.

    (2013)
  • K. De Witte et al.

    Dropout from secondary education: all’s well that begins well

    Eur. J. Educ.

    (2012)
  • F.M. Dieleman

    Modelling residential mobility; a review of recent trends in research

    J. Housing Built Environ.

    (2001)
  • Education Inspectorate, 2010. De staat van het onderwijs – Onderwijsverslag 08/09. Utrecht: Inspectie van het...
  • Ellen, I., 2000. Sharing America’s Neighborhoods Cambridge. Harvard University...
  • P. Feijten et al.

    Neighbourhood change. Reason to leave?

    Urban Studies

    (2009)
  • Galster, G., Hedman, L., 2012. Neighborhoods, neighborhood effects, and residential mobility: a holistic view and...
  • G. Galster et al.

    Who is affected by neighborhood income mix? Gender, age, family, employment and income differences

    Urban Stud.

    (2010)
  • Gasper, J., DeLuca, S., Estacion, A., 2011. Switching schools: reconsidering the relationship between school mobility...
  • S. Gibbons et al.

    Everybody needs good neighbours? Evidence from students’ outcomes in England

    Econ. J.

    (2013)
  • J. Hagan et al.

    New kid in town: social capital and the life course effects of family migration on children

    Am. Sociol. Rev.

    (1996)
  • Hall, P., Tewder-Johnes, M., 2010. Urban and regional planning:...
  • M. Ham van et al.

    Neighbourhood mobility in context: household moves and changing neighbourhoods in the Netherlands

    Environ. Plann. A

    (2009)
  • M. Ham van et al.

    Who wants to leave the neighbourhood? The effect of being different from the neighbourhood population on wishes to move

    Environ. Plann. A

    (2008)
  • Hango, D.W., 2003. The effect of neighborhood poverty and residential mobility on child well-beling. Doctor of...
  • D. Harding

    Counterfactual models of neighborhood effects: the effect of neighborhood poverty on high school dropout and teenage pregnancy

    Am. J. Sociol.

    (2003)
  • R.H. Haveman et al.

    The role of higher education in social mobility

    Future Children

    (2006)
  • R.H. Haveman et al.

    Childhood events and circumstances influencing high school completion

    Demography

    (1991)
  • J.J. Heckman

    Varieties of selection bias

    Am. Econ. Rev.

    (1990)
  • J.J. Heckman et al.

    Choosing among alternative nonexperimental methods for estimating the impact of social programs: the case of manpower training

    J. Am. Stat. Assoc.

    (1989)
  • Cited by (13)

    • Occupational status benefits of studying abroad and the role of occupational specificity – A propensity score matching approach

      2018, Social Science Research
      Citation Excerpt :

      Instead, self-selection of units into treatment and control condition takes place and a simple comparison of units in the treatment and control condition would be biased because covariates are not at balance and associated with the treatment. A prominent procedure for balancing covariates related to the observational units in the treatment and control conditions is propensity score matching (PSM) (e.g., Morgan and Harding, 2006), which is used with increasing frequency in the literature assessing the effects of education and educational contexts (e.g., Brand and Halaby, 2006) or residential mobility (e.g., Haelermans and Witte, 2015). In this method, the first step is to reduce the multi-dimensional differences in covariates between units of treatment and control condition to one dimension, the propensity score.

    • Reconsidering residential mobility: Differential effects on child wellbeing by race and ethnicity

      2017, Social Science Research
      Citation Excerpt :

      Using three waves of PHDCN data I fit pooled ordinary least squares (OLS) models, replicating results from prior studies, and then first difference models that reduce bias in estimating the effect of mobility compared to previous cross-sectional research. Previous work has largely based its findings on OLS regressions (e.g., Hagan et al., 1996; Ziol-Guest and McKenna, 2014) or logistic regressions (South et al., 2005) without accounting for the potential that different background characteristics or underlying factors of children and families could lead to a spurious association between residential mobility and child outcomes (a recent exception is Haelermans and De Witte, 2015). First difference models do not completely eliminate selection bias, but these analyses reduce concern about omitted variable bias by accounting for background and time-invariant characteristics related to both mobility and children's emotional and behavioral wellbeing.

    View all citing articles on Scopus

    The authors would like to thank participants from the 2012 TIER lunch seminars, the 2012 ZEW workshop on Education and Equality of Opportunity and the Second Lisbon Research Workshop on Economics, Statistics and Econometrics of Education 2013, for their useful comments on previous versions of this article.

    View full text