Received: 3 August 2021 | Revised: 25 October 2021 | Accepted: 12 November 2021 DOI: 10.1002/psp.2542 RESEARCH ARTICLE Opting out of neighbourhood schools: The role of local education markets in student mobility Rosario Scandurra1 | Adrián Zancajo2 | Xavier Bonal1 1 Department of Sociology, Globalisation, Education and Social Policies (GEPS) Research Abstract Center, Universitat Autònoma de Barcelona & Open enrolment policies assume that students living in disadvantaged areas can Centro de Estudos Globais, Universidade Aberta, Barcelona, Spain access better schools outside their neighbourhood. However, characteristics of in- 2 Manchester Institute of Education, University dividuals, quality of schooling and neighbourhood characteristics interact in very of Manchester, Manchester, UK complex ways to produce heterogeneous patterns of school choice in local educa- Correspondence tional markets. This article analyses how the geography of educational opportunities, Rosario Scandurra, Department of Sociology, the socioeconomic background of students' families and the characteristics of their Globalisation, Education and Social Policies (GEPS) Research Center, Universitat residential areas, impact on the travel‐to‐school distance in Barcelona. Based on a Autònoma de Barcelona & Centro de unique data set of school and student registers from Barcelona's local education Estudos Globais, Universidade Aberta, Barcelona 08193, Spain. authority, our study shows that distances travelled by students with the same social Email: rosario.scandurra@uab.cat background vary depending on the characteristics of educational supply and the Funding information income of the neighbourhood. While socially advantaged students tend to travel Juan de la Cierva Incorporación Programme longer distances than their peers from lower socioeconomic backgrounds, geo- (Ref. IJC2019‐040056‐I); British Academy graphical and educational factors mediate to produce high heterogeneity in mobility Postdoctoral Fellowship (PF2\180006) patterns. Our findings cast doubt on the supposed virtue of school choice to reduce education inequalities and underline the need to consider the diversity of local microeducation markets in policymaking and planning. KEYWORDS distance, educational opportunity, school choice, school composition, spatial inequality 1 | INTRODUCTION The liberalisation of access to public services, irrespective of re- sidential location, is understood as an important strategy in over- A recent report by the OECD shows that despite the fact that private coming social barriers and promoting patterns of social mobility schooling has not increased substantially in most of the countries that (Ferrari & Green, 2013). Thanks to open enrolment policies, it is as- participated in PISA between 2000 and 2015, systems of student sumed that students living in disadvantaged areas may access better allocation to schools based on their place of residence are increas- schools outside their neighbourhood. By moving out, they can im- ingly less frequent (OECD, 2019). By decoupling place of residence prove their educational opportunities and life chances. from schooling, educational reforms have increased families' capacity Breaking the link between place of residence and schooling was for choice and increased competition between schools, changes that the objective of desegregation policies after Brown vs Board of Edu- are positively seen by some authors as a more efficient and equitable cation in 1954, when the US Supreme Court prohibited Southern policy (Chubb & Moe, 1990; Chumacero et al., 2011; Hoxby, 2003). States from separating students by race. Busing (transporting This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2021 The Authors. Population, Space and Place Published by John Wiley & Sons Ltd. Popul. Space Place. 2021;e2542. wileyonlinelibrary.com/journal/psp | 1 of 15 https://doi.org/10.1002/psp.2542 2 of 15 | SCANDURRA ET AL. students to different school districts from their residence) became family factors, by the characteristics of educational supply and by the the strategy for detaching housing and schooling and ensuring that neighbourhood context (Andersson et al., 2012). While more affluent equality of opportunities did not depend on income or race (Cascio students have the ability to travel longer distances to avoid local et al., 2008). Certainly, the geography of education has demonstrated schools, others do not have this possibility or might not need to travel that socio‐spatial differences in the structural concentration of pov- because they have ‘good schools’ near their homes. Analysing which erty and disadvantage has an impact on educational outcomes (Butler factors explain travel‐to‐school distance and how mobility patterns & Hamnett, 2007; Kuyvenhoven & Boterman, 2020; Sykes, 2011; impact on the socioeconomic composition of schools reveals how Sykes & Musterd, 2011). The quality of schooling in poor areas de- school choice shapes education inequalities in Barcelona. pends on the number of resources and the learning opportunities Examining the sources of heterogeneity in selecting geo- available. There are usually fewer resources and opportunities than in graphically ‘convenient’ schools and its implications is crucial for a wealthier neighbourhoods (Ledwith & Reilly, 2013; Lubienski & better understanding of how people conceptualise and enact place Dougherty, 2013). Beyond human and material resources, the quality and space in education settings. In this sense, even in European of schooling and the lower performance of students in socially dis- countries where residential segregation is deemed moderate advantaged areas are also affected by school composition (Dumay & (Musterd, 2005), school choice policy shifts might impinge on fa- Dupriez, 2008; OECD, 2012). The concentration of vulnerable chil- milies’ microdecisions and reflect or magnify potential gaps in ac- dren in certain schools neutralises the benefit of the peer effect and cessing to education provision. As we will show, these effects vary impacts negatively on their learning experience and performance. largely depending on the neighbourhood and educational character- From this perspective, open enrolment systems, which allow istics, resulting in unequal geographical differences in access to families to choose beyond their neighbourhood, would result in more schooling. Furthermore, the analysis of students’ mobility patterns opportunities for children living in disadvantaged areas. However, also provides useful insights to assess whether school choice policies other factors challenge the theory of change of this policy option. contribute to reducing school segregation and social stratification Individual factors, quality of schooling and neighbourhood char- between schools or, on the contrary, the higher capacity of choice acteristics interact in very complex ways to produce different pat- increases education inequalities. terns of school choice in local educational markets. Indeed, several The article is structured as follows. The following section reviews studies have shown that school choice rationalities do not respond to key aspects of the literature on socio‐spatial inequalities and school single and simple factors. ‘Choosers’ differ in their perceptions of choice and formulates the research questions regarding the drivers of school quality (Schneider et al., 2000), in the importance given to home‐to‐school distance in Barcelona. The third section of the article other school characteristics (security, values, school size, and describes the school admission system of Barcelona and the main in- composition) (Bonal et al., 2017), in their opportunity and willingness equalities in the geography of educational provision. The fourth section to travel greater distances (Easton & Ferrari, 2015; He & Giuliano, describes the data and methods used in the analysis, with particular 2018), or in the restrictions they must face (economic, geographical, attention given to the strategies used to characterise school supply in discriminatory) when deciding their choice set (Bell, 2009; areas of close proximity. The fifth section presents the results of our Ben‐Porath, 2009; Bonal & Zancajo, 2018). analysis of the patterns of student mobility in Barcelona. The last It is highly relevant to understand the expected and unforeseen section concludes and discusses the policy implications of our findings. effects of certain education policies, especially since policymakers promote choice and competition as drivers of school reform. Inter- national literature on school choice has widely demonstrated that 2 | SOC I O‐S PA T I A L I N E Q U A L I T I E S AN D choice strategies are socio‐spatially diverse and unequal. These dif- SCHOOL CHOICE ferences are not only vertical but also horizontal, as research on middle‐class strategies of school choice has demonstrated (Benson The literature on school choice has shown that proximity plays an et al., 2015; Boterman, 2013; Lareau, 2014). Diverse strategies may important role in parental decisions when selecting a school for their alter the expected effects of open enrolment policies, which can in- children (Alegre & Benito, 2012; Bosetti & Pyryt, 2007; Butler & van crease school segregation by means of different mechanisms Zanten, 2007; Schneider et al., 2000). Proximity is regarded as one of (Boterman, 2020; Makris, 2018; Zancajo & Bonal, 2020). the main factors that conditions the school choice decisions of most This article analyses how the geography of educational opportu- disadvantaged families, who may face greater economic and geo- nities, the socioeconomic background of students' families and the graphical constraints and are less likely to travel longer distances characteristics of their residential areas, impact on the travel‐to‐school (Andersson et al., 2012). However, proximity is never a nonsignificant distance in Barcelona. The school admissions policy of the city is factor in relation to school choice, regardless of socioeconomic characterised by a controlled school choice system. However, with background, especially in preschool or primary education. Even in many choice options and many publicly subsidised private schools, the open‐enrolment systems, most parents make their decisions in a two‐ study of student mobility is particularly interesting. Travel‐to school stage process. First, they construct a choice set of those schools they distance is a good proxy on the families' capacity to ‘exercise’ choice are willing to consider. Second, they make the final decision within and the extent to which this capacity is influenced by individual and this smaller choice set (Bell, 2009; Burdick‐Will et al., 2020). The SCANDURRA ET AL. | 3 of 15 limits of this choice are the result of an interaction between geo- schools’ characteristics. Parents enrol their children in very different graphical factors (e.g., accessibility or travel costs), the institutional types of schools even if they live in the same neighbourhood in a design of admission policies, and the educational preferences of in- process of ‘disaffiliation and selective belonging’ (Boterman, 2020, p. 1). dividual choosers. White flight strategies and colonisation strategies may coexist in gen- The different geographies of education generate dissimilar trified neighbourhoods and produce different patterns of mobility and schooling opportunities for children from different neighbourhoods. different effects on school segregation (Candipan, 2019, 2020; Research has shown that schooling options are not homogeneously Mordechay & Ayscue, 2020; Pearman, 2019). spatially distributed. The number of schools available and their The extent to which families are able to activate these strategies is characteristics vary depending on the social and economic char- closely related to families' cultural or economic capital, but these practices acteristics of the territories (Bayona‐i‐Carrasco & Domingo, 2021). are not independent of context or location. Building on the Bourdieusian Families residing in the most disadvantaged areas tend to live close to concept of forms of capital, Jacques Lévy (1994) defined spatial capital as schools of lower quality with a higher proportion of disadvantaged those resources that actors can accumulate related to space to develop students (Burgess et al., 2011; Elacqua et al., 2011; Fjellman specific strategies. Spatial capital is, therefore, an unequally distributed et al., 2019; Holt et al., 2019). The configuration of local education asset, which provides different resources and capacities to different in- markets confirms a strong hierarchical ordering of urban spaces. The dividuals to use space in their own interest. According to Lévy, spatial best performing and most attractive schools are frequently con- capital has two components: positional and situational. Position capital is centrated in the most affluent areas, while the most deprived related to a place, whilst situation capital is related to an area. neighbourhoods have less access to quality educational resources (Oberti, 2007). There are several reasons to explain why geographical In the first case [position capital], it is the inclusion in a market imperfections of educational provision cannot easily be cor- space (a residential place or a place of work) that pro- rected. First, the regulation of educational markets is insufficient to vides an individual with spatial assets. In the second guarantee a balanced territorial and social distribution of educational case [situation capital], it relates to a space that the provision. They may refer to systems of authorisation of providers or individual appropriates globally via a complete range of accountability mechanisms, but have limited capacity to determine mobilities, in which distance, though still a factor, is where schools are open (Bonal et al., 2020). Second, schools with low controlled (Barthon & Monfroy, 2010, p. 178). demand and poor academic performance rarely disappear from the market (Burgess et al., 2011). Third, schools respond differently to Applied to school choice, these concepts of position capital and attract the best students who are less costly to educate and may situation capital express the relative advantage that some actors enjoy contribute to increasing the school's reputation (van Zanten, 2009). with respect to where they live and their mobility in specific local edu- Therefore, their geographical location is also a competitive factor cation markets. Therefore, it is not only the way actors are positioned in (Gulosino & Lubienski, 2011). the space that is important in understanding urban education inequal- This unequal distribution not only generates unbalanced edu- ities, but also the ways in which these actors appropriate the space. cational markets from the supply side, but also conditions the school The socio‐spatial dimension of educational inequalities enriches choice decisions of families and students. Some families can develop the idea of the different ‘circuits of schooling’ that are configured by different strategies to maximise their opportunities in the local class, cultural capital and choice (Ball et al., 1995). By considering the education market. They may have incentives to select where they spatial resources that individuals and groups may activate, it is possible live based on the location of the best schools (Ely & Teske, 2015; to identify differences in the school choice processes among in- Ramond & Oberti, 2020), they might develop white flight strategies dividuals that share the same class position. The spatial limits of their to avoid those schools with a higher concentration of disadvantaged choice set and the distance they are willing to travel to school, will vary students (Billingham & Hunt, 2016; Kye, 2018) or they may even depending on the educational opportunities available in their neigh- adopt strategies of social closure to prevent the participation of bourhood, the access restrictions to certain schools, and their educa- more disadvantaged students (Ichou & van Zanten, 2019; Zancajo tion preferences. School choice policies are affected by both the et al., 2021). characteristics of the available proximity schools and the distance As Wilson and Bridge (2019) have shown, school choice increases students are willing to travel. The capacity of open enrolment schemes school segregation independently of the choice mechanism in place. to provide better schooling options (particularly for socially dis- However, the factors causing observed increases are localised and advantaged students) depends largely on the extent to which unequal contextual. The ‘size of school districts/catchments, the number of educational geographies impinge on their potential and actual choices. schools and degree of competition in the area, and the existing socio- This body of research inspires the research questions of this economic geographies of the neighbourhoods involved’ (Wilson & study. These are: Bridge, 2019, p. 14) are among the causes of school segregation. Urban gentrification has made school choice strategies even more complex, as (1) What are the characteristics of the proximity schools available socio‐spatial strategies of school choice of newcomers and residents are for different social groups of students? Do more socially ad- affected by the social composition of the neighbourhood and the vantaged students tend to select more distant schools? 4 of 15 | SCANDURRA ET AL. (2) Is distance travelled to school related to the socioeconomic (1) All schools located in the same catchment area where the cluster characteristics of the students, neighbourhood characteristics of houses is located. and the characteristics of proximity schools? Do socially ad- (2) The three public schools and three privately subsidised schools vantaged students travel higher distances to access schools with closest to the cluster of houses. more affluent populations? (3) All schools less than 500 m from the cluster of houses. (3) What are the main factors that have an influence on the distance (4) If necessary, all the closest schools to achieve a minimum choice to school? What are the roles of individual characteristics, edu- set of six public schools and six private subsidised schools. cational effects, and neighbourhood effects? As a consequence of these new criteria regarding proximity priority and considering the oversupply of schools in certain areas of 3 | T H E LO C A L E D U C A T I O N MA R K E T O F the city, the estimated average number of schools for which families BARCELONA had geographical priority increased from 7.9 to 16.7 in 2012 (CEB, 2012). Regarding the type of educational institution, the In Spain, families apply for a school place in preschool education average number of local public schools by cluster of houses was 9.0, during the year in which the child turns 3 years old.1 Parents freely while in the case of privately subsidised schools it was 9.6. By shifting express a set of school preferences, including any public or subsidised school choices through these criteria, the CEB tried to compensate private schools, and students are allocated a school using an im- for the unequal internal distribution of the different catchment areas. mediate acceptance algorithm (also known as the Boston mechan- These areas differ in the overall number of schools and in the pro- 2 ism). While families are free to apply to any public or private vision of public and privately subsidised schools. subsidised school in the city, in cases of oversubscription, applica- However, while the reform ensured a minimum number of po- tions are prioritised using three main criteria that were established by tential schools as proximity schools, geographical differences in the the 8/1985 Education Reform Act on the Organic Law of the Right to proximity schooling options remain significant. Figure 1 shows the Education. These criteria include residential proximity, the enrolment unequal distribution of choice opportunities in the territory per of siblings at the school and household income.3 cluster of houses (the territorial unit used by the CEB to calculate The case of Barcelona as a local education market has two in- local schools). There are significant differences in supply between the teresting characteristics that make the study of school choice and centre of the city and the periphery. Overall, there are fewer schools socio‐spatial inequalities particularly interesting. First, there are many in peripheral areas than in the city centre. In addition, there are no- private schools, most of them publicly subsidised. The private sub- table differences in the density of schools between and within sidised sector enrols 55% of primary and lower secondary students in catchment areas. The density of schools is particularly high in some the city, while 43% go to public schools and only 2% of students are areas in the north and northwest parts of the city. Moreover, access enroled in independent private schools (CEB, 2020). These numbers to a public or a private subsidised school is unevenly distributed are significantly different from those of the rest of Catalonia or Spain, geographically. Figure 2 divides the supply of public and private where only 35% and 28.2% of students are enroled in private sub- subsidised schools available in a 500‐m range for each cluster of sidised schools, respectively (MEFP, 2019). Second, the city is divided into 29 catchment areas or school districts. All residents have priority access to all public and subsidised private schools within their catchment area. However, in 2012, the Consorci d'Educació de Barcelona (CEB), the local education authority in charge of the city's educational planning, changed the school admissions system to in- crease school choice. With the reform, families could have priority access to more schools than just those included in the catchment areas, under the following conditions: 1 In Spain, compulsory education starts at the age of 6, when children begin primary education. However, most applications take place at the age of 3, since the system universally provides 3 years of preprimary education. Indeed, the net rate of enrolment at 3 years old is 96.8% (INE, 2020). 2 The immediate acceptance algorithm or ‘Boston mechanism’ is a student placement procedure, through which students (families) list their preferences. Given the reported preferences, the allocation of school places follows an algorithm that maximises students' preferences, subject to the prespecified priorities of students at each school. Places at each school are allocated based on students' rank calculated from the algorithm (Cantillon, 2017). 3 F I G U R E 1 Number of schools in a 500 m range per cluster of While these general criteria are set as a national regulation, regional and local educational authorities can establish their own indicators to define residential proximity and thresholds houses in Barcelona. Source: Own elaboration based on CEB for household income. register data SCANDURRA ET AL. | 5 of 15 F I G U R E 2 Number of public and private subsidised schools in a 500 m range per cluster of houses in Barcelona. Source: Own elaboration based on CEB register data houses. The maps reveal that both public and private schools are In this context, the high diversity of the education system and the unequally distributed in the city. The territorial imbalance is parti- marketisation of the educational landscape given by an almost de cularly marked in the case of private subsidised schools. Supply is facto open enrolment policy makes the study of the travel to school concentrated in the centre and northwest parts of the city where the distance particularly interesting. wealthier districts are located and where there are fewer public schools. In contrast, private subsidised schools are almost non- existent in peripheral areas of the city. This includes some catchment 4 | D A T A A N D ME T H O DS areas in the north that have a low population density. How this unequal distribution of school supply is translated into The data used in the analysis are retrieved from several sources. First, inequalities of schooling depends on a complex interaction between information from students and schools comes from the school reg- several factors. Whether families opt for their children to attend the ister of the CEB. This information is compiled from 11,557 students closest school, a school in the catchment area, or commute to a in their first year of school in 325 schools in the academic year school outside their area may depend on the geographical dis- 2017–2018. Variables include a wide range of individual character- tribution of schools in the city, the spatial distribution between istics, including the student's country of birth, whether the student is public and private subsidised schools, the level of oversubscription a recipient of Free School Meals (FSM) or Social Allowances (SA),4 of proximity schools and parental preferences for a certain type of and the geographic coordinates of the household. We extract the school. geolocation of schools from the open data service (Servei de dades Thus, Barcelona's local education market is based on a controlled obertes) of the Barcelona City Council. Second, the education register choice policy, but the particular way of considering proximity and the is linked to 1068 census tracts, 233 statistical unit areas,5 unequal geography of public and private schools may potentially 73 neighbourhoods and 29 catchment areas in the city of Barcelona. generate high levels of student mobility and impact on the socio- Contextual data on the neighbourhoods come from the open data economic composition of schools. The combination of residential service of the Barcelona City Council, while data on 29 catchment segregation, the unequal geography of education, the presence of areas come from the CEB. Third, CEB data are matched with parents' high number of private subsidised school (mainly Catholic) and the place of birth and parents' education level using city council and described institutional school choice arrangement produce a specific population registers. educational landscape of high school segregation in the city and Moreover, information on socioeconomic census tracts was re- significantly greater than residential segregation, especially with re- trieved from the Spanish Bureau of Statistics (INE, for its Spanish gard to the ethnic dimension (Bonal et al., 2019). Differently from acronym). This bureau collects information concerning the average Northern European or US cities, Barcelona has relatively low levels of residential segregation (as is the case of other Southern European cities). However, ‘the combination of catchment areas, school choice 4 SAs include students in a situation of severe poverty, who receive cash transfers from the mechanisms and segmentation of the school supply into public and City Council. 5 Statistical Unit Area (AEB by its acronym in Catalan), is a population, urban and private makes the system relatively segregated, even in the face of socioeconomic uniform unit area in which each of the 10 districts are divided. This is low levels of residential segregation’ (Boterman et al., 2019, p. 3065). composed of at least 500 voters. 6 of 15 | SCANDURRA ET AL. income of the census tracts (secciones censales), which are the (a) Students' socioeconomic background. We use two proxies of the smallest administrative units in Spain. Barcelona is divided into 1068 social and cultural characteristics of students. These are: foreign census tracts, which range between 1000 and 2500 inhabitants. origin (having both parents foreign‐born), and parents' education The exclusion of missing and unclassifiable data left 11,168 level (having at least one parent with secondary or lower edu- cases; this loss represents 2.1% of student records. This is due to cation or having at least a parent with upper secondary education incomplete information mainly on parents' education records and or higher). All these are measured as dummy variables. geolocation of the household. Additionally, for students who live (b) Neighbourhood educational resources. We use three variables to outside the city and who are enroled in a school in Barcelona, data account for educational resources near the student's residence: are incomplete, and the inclusion could have biased the results. Thus, number of schools within 500 m of the household, the share of these records are excluded from the analysis. public‐school places within 500 m of the household, and an index of school socioeconomic composition. The total number of schools and the number of public schools within 500 m are included to 4.1 | Dependent variable account for the difference in the number of proximity school op- tions across the different areas of the city, which can influence the This article analyses the distance to school travelled by 3‐year‐old distance to school travelled. The index of school socioeconomic preprimary students in Barcelona. Distance from home to school is an composition relies on four underlying variables: the proportion of indicator of the use that families give to open enrolment schemes, FSM receivers, the proportion of SA receivers, the proportion of and a good way to assess whether the enactment of school choice students with foreign parents, and the proportion of students with policies increases or reduces education inequalities. To examine parents with at most lower‐secondary education (ISCED 2). The educational choices, we calculate Euclidean and Manhattan distances index includes data for only 3‐year preprimary students. These from households to school.6 Since the difference between Euclidean indicators are combined through an iterated principal‐factor and Manhattan distance is not relevant, we use the Euclidean dis- method and the resulting factor is max/min transformed and tance travelled between home and school by 3‐year‐old students changed in sign. The indicator ranges between 0 and 1.7 Higher enroled in Barcelona city schools. Thus, we only use one address for values of the index mean better social composition of the school. each student‐school entry in our data. A shortcoming of drawing on (c) Neighbourhood socioeconomic characteristics. We use three proxies distance is that our data might not accurately reflect the student's of neighbourhood socioeconomic characteristics: census tract typical distance from school if a student frequently travels from a median income, the coefficient of variation of the census tract different address (e.g., from the home of another custodial parent or median income by statistical unit area,8 and the school composi- guardian). Additionally, for those students who made a permanent tion of the proximity area, which is the average of the school change to their address during the school year, but remained enroled composition index within 500 m range of the student's residence.9 in the same school, our results cannot reflect the distance they travel The census tract of median income is based on the 2017 Spanish to school since their parents might not report the address change Bureau of Statistics estimate. This defines a family as two or more until the following school year. related persons living in the same residence. A household includes all residences, even those with single people or unrelated groups of two or more. As for the school composition index within 500 m 4.2 | Independent variables range, using the geolocation of schools and households, we take the average of the school composition index of those schools We use a set of individual and contextual data such as educational within 500 m of the student's residence. Thus, the resulting item resources available in the neighbourhood (e.g., the number of schools indicates the school composition in the proximity area and we within a given area) and various school composition measures which interpret this as an indicator of the schools' socioeconomic char- are detailed below. To have a more accurate description of the acteristics available within a proximity area. proximity area in which the students and families operate, we take advantage of the geolocation of the household to construct a proxy within 500 m from the household. The main characteristics of the variables used in the analysis are summarised in Table 1. As the table presented above shows, the variables used in the 7 Moreover, we perform exploratory and confirmatory analysis and several sensitivity analyses using different factor iteration methods, principal component and analysis can be grouped into three different categories: maximum‐likelihood factor method. In the annexe we report the result of the factor analysis and a correlation matrix using four different specifications of the factor using principal components, iterated factor analysis and geometric mean and standard and robust Benefit of the Doubt weighting. Results show the reliability of the derived construct. 8 This measure indicates the variability of the census tract median income within the 6 Euclidean or Pythagorean distance is the smallest distance between two points, while Statistical Unit Area. 9 Manhattan is a distance metric between two points in an N dimensional vector space. It is School composition of proximity schools can also be understood as a variable related to the sum of the lengths of the projections of the line segment between the points onto the educational resources. However, in our analysis we consider it as a proxy of the coordinate axes. socioeconomic composition of the neighbourhood. SCANDURRA ET AL. | 7 of 15 TABLE 1 Variables description Variables Definition Nature Transformation Source Distance in metres Distance home to school Continuous Log CEB, register Students' socioeconomic Secondary of lower One of the parents has secondary or Dummy Census data background education lower education (ISCED 2<) Upper‐secondary education One of the parents has at least Dummy Census data upper secondary education (ISCED 3‐5B) Tertiary University Degree One of the parents has at least Dummy Census data university certificate (ISCED 5A‐8) Both parents foreign Both parents are foreigners Dummy CEB, register Educational resources Schools in 500 m Number of Schools in 500 m range Continuous Own elaboration of the residence Public school places Share of public‐school places within Continuous Own elaboration 500 m range of the residence Public school Student enroled in public school Dummy CEB, register School composition Factor of school composition: the Continuous Min/max Own elaboration share of FSM, the share of RSA and the share of students with parents with tertiary education Public supply Total public schooling supply in the Continuous Log CEB, register catchment area Neighbourhood Census tract household Census track residence median Continuous Share of the city INE characteristics income household income average CV Census tract household Coefficient of Variation of the Continuous Own income Census tract household income elaboration, in the statistical units' areas based on INE School composition Mean of the school composition of Continuous z score Own elaboration proximity area the schools in 500 m range of the residence 5 | U NEQ UAL OPPORT UN ITIES I N TH E higher number of private subsidised schools. On average, 31.6% of E D U C A T I O N MA R K E T the schools in a 500‐m range are public in the richest census tracts, while in the rest of the city, this percentage is greater than 56%. What are the characteristics of the proximity schools for different Although the number of schooling options does not vary sig- types of students? (Research question 1). Table 2 presents the nificantly with students' characteristics, there are important differ- average number of schools within 500 m of the student's residence, ences regarding the socioeconomic composition of the schools as well as the characteristics of their population. The table, inspired available in the proximity area. For instance, students with secondary by Burgess et al. (2011), describes the characteristics of schools that (or lower) educated parents have proximity schools where the aver- are located near different types of families. As the table shows, the age percentage of secondary‐level educated parents is 31.3%; for number of schools available within 500 m of the residence is very students with upper secondary or higher‐level educated parents, this similar for all the groups of students considered (on average 4.69). share is 17.7%. Similar differences are observed regarding other so- This is similar across the groups analysed, although a slight difference cioeconomic characteristics of proximity schools. Proximity schools is observed across income census tracts. On average, around 5.0 of those students with secondary educated parents tend to con- schools are located within 500 m in low‐ and middle‐income areas, centrate higher percentages of SA, FSM and foreign students than while in the richest areas, this is 3.8. In this regard, it is important to proximity schools of students with upper secondary or higher‐level consider that, in Barcelona, the richest census tracts are generally less educated parents. densely populated compared to the rest of the city. Moreover, as the Similarly, if we consider the characteristics of proximity second column shows, they have fewer public‐school places and a schools of foreign students, SA and FSM recipients, we see 8 of 15 | SCANDURRA ET AL. TABLE 2 Characteristics of schools within 500 m of the household % of school population Lower % of public Secondary or Schools schools less parental in 500 m in 500 m education SA FSM Foreigners All students 4.69 52.3 20.1 10 21.3 19.1 Lower secondary or less No 4.7 51.2 17.7 8.3 18.9 17.6 parental education Yes 4.66 57.1 31.3 16.8 31.7 25.6 Social allowances No 4.69 51.6 18.9 9.2 20.2 18.4 Yes 4.58 58.7 33.4 19.3 34.1 28.3 Natives parents No 4.66 51 18.3 8.7 19.7 17.2 Yes 4.73 55.2 24.5 13.1 25.3 24.2 Both parents foreigners No 4.67 51.6 9.4 9.4 20.7 18.3 Yes 4.77 57.5 14.9 15 27.1 27.3 Free school meals No 4.67 51.2 17.9 8.6 19 18 Yes 4.73 56.7 30.1 16.5 31.9 24.9 Census tract income Bottom 4.68 56.9 41.7 23.7 41.3 33.2 quantiles 2 4.87 56.7 20.9 10.5 23.8 18.8 3 5.09 57.8 16.3 7.6 19 16.5 4 4.95 57.8 14.3 5.7 15.3 16 Top 3.8 31.6 7.27 2.4 7.1 11.5 Source: Own elaboration based on CEB register and INE data. dramatic differences. The proximity schools in the areas where SA re- cipient students live enrol on average 28.3% of foreign students com- pared to the average of 19.1% in the whole city, and the proximity schools of those who are not SA recipients enrol on average 18.4% of foreigners' students. Likewise, the characteristics of the proximity schools of SA students' place of residence have a high proportion of SA students (as well as a high proportion of FSM students), significantly higher than the city's average. Differences follow the same pattern for all variables, with FSM beneficiaries' proximity schools enroling an average of 31.9% FSM students. The unequal distribution of schools with children whose parents are upper secondary level educated, receive SA or FSM, or are foreign stu- dents, is remarkable across the income quintiles of the census tracts. For students from the lowest income quintile, the average FSM rate is 41.3%, compared to 7.1% in the richest quintile. A similar but slightly less stark difference is reported for SA and foreign students. Figure 3 complements the information in Table 2 by plotting F I G U R E 3 Average SES schools in a 500 m range per cluster the average socioeconomic composition of schools within 500 m of houses in Barcelona. Source: Own elaboration based on CEB of a household per block of houses. The figure shows the highly register data unequal distribution of schools composition in the city. This is the result of both the unequal spatial distribution of social groups and affluent area, the school composition is very high and homo- their school choices, which results in highly unequal school char- geneous. Additionally, the maritime area (shown in the bottom‐ acteristics within proximity areas. The darker areas in the figure right of the map), is one of the more recently developed city dis- show that in the northwest of the city, which is also the more tricts and is becoming increasingly gentrified. This area also shows SCANDURRA ET AL. | 9 of 15 FIGURE 4 Distance travelled to school in kilometres by individual characteristics. Source: Own elaboration based on CEB register data high and homogeneous school composition. In contrast, the old distance travelled by students who have both parents born abroad is inner city of Barcelona (shown in the centre‐bottom of the map), 0.41 km compared with 0.53 km travelled by those with at least one together with more peripheral areas, are the poorest in terms of native parent. Additionally, the length of the bar inside the violin school composition. indicates the interquartile range, which provides a measure of the To summarise, we find stark differences between social groups spread of the distribution. We find that the interquartile range is in the composition of proximity schools and related educational smaller for students with foreign‐born parents indicating that supply in the local area. The figures presented in this section these families also have smaller variability compared to those with at together with related evidence (Bonal et al., 2019) suggest that least one native parent. Students from families with a tertiary‐level unequal distribution of school supply regarding its socioeconomic education tend to travel further (median 0.56 km) compared to those composition is considerably higher than residential segregation in from families without tertiary‐level education (0.40 km). Similar Barcelona. differences are observed for FSM and SA recipients compared to their peers who do not receive these social benefits. The median distance travelled by students who attend public schools is 5.1 | Describing the drivers of mobility to school 0.47 km, which is lower than the median of those students enroled in an independent private or a private subsidised school (0.58 km). Given the unequal distribution of educational opportunities in the Once again, the variability is smaller for those who attend a public city associated with the spatial differences in school composition, in school. this section we assess whether distance travelled to school is related Figure 5 analyses the relationship between the distance travelled to students' socioeconomic characteristics, neighbourhood socio- to school against each of the independent variables included in the economic characteristics or the characteristics of proximity schools analysis. In the case of the social composition of the school where the (Research question 2). As reported in Figures 4 and 5,10 different student is enroled, the figure shows a positive difference across individual and context characteristics are associated with the dis- school composition deciles, indicating that those who attend schools tance travelled to school. The figures show distance in kilometres with higher school composition travel longer distances than their 11 against student characteristics using eleven violin plots. The median peers enroled in more disadvantaged schools. For instance, the median distance travelled by students who attend schools in the bottom decile of school composition is 0.34 km, whereas for 10 Table A3 in the annex provides descriptive statistics for Figures 4 and 5, while A1 provides those at the top the median is 0.84 km. Interestingly, the relationship univariates statistics for all the variables. is positive with the dispersion of the distance travelled: the 11 We choose this way to report more clearly the shape of the distribution of each variable higher the school composition, the larger the variability in the used in the analysis. We cut the plot at 2 km distance to increase readability. The full tables are reported in A3. distance travelled. 10 of 15 | SCANDURRA ET AL. FIGURE 5 Distance travelled to school in kilometres by contextual characteristics. Source: Own elaboration based on CEB register and INE data The school composition of the proximity schools (within 5.2 | Estimating the effect of socioeconomic, 500 m of their place of residence) is also positively related to the educational, and neighbourhood characteristics distance travelled. However, the average of school composition on mobility to school in the proximity area captures all the available options of schooling within the area of residence and therefore dis- To carry out an in‐depth analysis of the individual, neighbourhood criminates less compared to the actual school where students are educational resources and neighbourhood socioeconomic character- enroled. This difference underlines the fact that while family istics that influence the distance travelled to school (third research preferences reveal a linear relationship between distance tra- question), we have estimated six multilevel models in which the velled and school composition, school choice varies depending on distance travelled to school is the dependent variable (Table 3). Such the specific characteristics of microlocal education markets. In- models are well equipped to consider the nested structure of the terestingly, the variability is higher in the 9th and 10th deciles, data, since in our case, students are grouped into 325 schools. and also in the 3rd and 4th deciles of the average social com- The analysis starts by estimating the unconditional model to show the position of proximity schools. Likewise, we find stark differences relevance of using such a framework. The intraclass correlation (ICC) when examining the average distance to school across deciles of measures the ratio of between‐school variation to the total variation. income of the census tracts. Additionally, when examining the We find that variance in the mean distance travelled across students, public‐school supply within 500 m of the household, we find that but within schools, is higher than between school variance. The total the median distance travelled by those who live in the bottom variance attributable to differences between schools accounts for decile is higher by approximately 0.3 km compared to the 34.5% of the total variance. rest. Finally, there is a negative relationship, as might be ex- Regarding socioeconomic and educational individual character- pected, between the distance to school and the number of istics (Model 2), results show that having both parents from a foreign schools available in the local area. Those who have one school country is significantly and negatively related to the distance tra- within 500 m of their home travel 0.84 km, whereas the median velled to school. The estimate is consistent across all the models, but home‐to‐school distance of the normative group (i.e., those it is mediated by the socioeconomic composition of the school where who have five schools within 500 m) is 0.49 km. At descriptive the student is enroled, and the composition of the schools located in level, there is not clear difference in distance to school travelled his/her residential neighbourhood. Indeed, there is a reduction in the among deciles of the coefficient of variation of the median size of the estimate between Models 3 and 6 of 21%. Parents' edu- household income. cation does not appear to be related to the home–school distance SCANDURRA ET AL. | 11 of 15 TABLE 3 Multilevel model of individual and local education market drivers of distance to school Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Both parents foreign −0.103*** −0.0942*** −0.0953*** −0.0949*** −0.0864*** −0.0753** −0.0261 −0.0259 0.0254 0.0253 0.025 0.0249 Parental education Secondary or lower −0.00176 −0.0159 −0.0519** −0.0522** −0.0674** −0.0489** −0.023 −0.0229 0.0226 0.0226 0.0223 0.0223 Upper secondary 0.028 0.0217 0.00699 0.00585 −0.00837 −0.0023 −0.0189 −0.0187 0.0184 0.0183 0.0181 0.0181 Tertiary ref. ref. ref. ref. ref. ref. Attending public school −0.230*** −0.273*** −0.337*** −0.340*** −0.359*** 0.0872 −0.0626 −0.0636 0.0724 0.0729 0.0813 0.0751 Schools in 500 m 1 0.293*** 0.398*** 0.395*** 0.348*** 0.352*** −0.0439 0.0435 0.0434 0.0429 0.0427 2 0.188*** 0.231*** 0.238*** 0.203*** 0.204*** 0.0303 0.0298 0.0298 0.0294 0.0293 3 0.114*** 0.122*** 0.133*** 0.116*** 0.116*** 0.027 0.0264 0.0265 0.0261 0.0261 4 0.0465 0.0598* 0.0616* 0.0563* 0.0576* 0.026 0.0255 0.0255 0.0251 0.025 5 ref. ref. ref. ref. ref. 6 −0.0324 −0.0313 −0.0302 −0.0311 −0.0314 0.0253 0.0248 0.0248 0.0244 0.0244 7 −0.0509 −0.051 −0.0486 −0.0388 −0.0393 0.0282 0.0276 0.0276 0.0272 0.0271 8 or more −0.0993** −0.0965** −0.0952** −0.0794* −0.0779* 0.0332 0.0326 0.0325 0.032 0.032 Ratio of public supply 0.337*** 0.216*** 0.214*** 0.101** 0.0827** 0.0318 0.0319 0.0319 0.0322 0.0321 Tract hh Income −0.655*** −0.644*** −0.451*** −0.452*** 0.0336 0.0337 0.0356 0.0355 Tract hh Income, cv AEB −0.724*** −0.595*** −0.590*** 0.163 0.161 0.16 School comp. −1.269*** 2.114*** 0.0765 0.167 School comp. proximity area −1.392*** 0.0772 Constant 6.194*** 6.317*** 6.293*** 6.298*** 6.295*** 6.294*** 6.147*** −0.032 −0.046 0.0495 0.0553 0.0557 0.0615 0.0527 sd(schools) 0.550*** 0.536*** 0.545*** 0.626*** 0.631*** 0.710*** 0.575*** −0.0234 −0.023 0.0235 0.0269 0.0271 0.0307 0.0247 (Continues) 12 of 15 | SCANDURRA ET AL. TABLE 3 (Continued) Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 sd(residuals) 0.756*** 0.752*** 0.742*** 0.726*** 0.726*** 0.714*** 0.714*** −0.00513 −0.00523 0.0052 0.00509 0.00508 0.005 0.005 ICC 0.3464 0.3364 0.3499 0.4264 0.4307 0.4972 0.3938 0.0195 0.0195 0.0199 0.0214 0.0214 0.0219 0.0208 N 11,168 10,644 10,527 10,527 10,527 10,527 10,527 Note: All the variables are mean centred apart from both parents foreigner and parental education. Source: Own elaboration based on CEB register and INE data. travelled when only socioeconomic and educational characteristics that schools' composition plays in attracting students and influencing are taken into account. However, in Model 4, when we account for their mobility patterns. However, when introducing the neighbour- neighbourhood socioeconomic characteristics, the distance travelled hood's school composition (Model 7), this variable mediates and by students with low educated parents (secondary or lower) is sta- strongly reduces the attraction effect of the composition of the tistically significantly lower than students with higher educated par- school attended. In fact, living in a more affluent educational area ents. Model 2 also shows a significant negative effect of attending a produces a 66% reduction in the effect exerted by school composi- public school. However, when school and neighbourhood educational tion to travel long distances. and socioeconomic characteristics are included (Model 7), its effect To sum up, and in response to the third research question, the size is sharply reduced, and it is no longer statistically significant. results of the multilevel model show that the distance travelled to In Model 3, we add a set of dummy variables indicating the school is certainly affected by students' socioeconomic character- number of schools available within 500 m of the student household. istics (immigrant origin or level of parental education). However, the The category of reference is the normative one having five schools social composition of the attended school and the neighbourhood within a 500‐m range. We find a positive and significant association characteristics have a stronger effect on student mobility patterns. with the distance travelled to school among students with four or What these results show is that high choice of schools (as in the case fewer schools available compared to those with five available in a of Barcelona) does not translate automatically into higher levels of 500‐m range. Nevertheless, only those who have eight or more mobility. Distance travelled to school is mediated by the educational schools available within their proximal area show a statistically sig- and social contexts of the areas where students reside. nificant reduction in the distance travelled compared to the reference category. The ratio of public supply is also positively associated with the distance travelled to school. However, its size effect is sig- 6 | D IS CU SS IO N A N D CO N CL US I O N nificantly reduced when neighbourhood characteristics are taken into account. This can be explained by the fact that neighbourhood Research has provided significant evidence on how unequal geo- characteristics not only affect the distance travelled to school, but graphies of educational supply and demand interact to produce dif- also the possibility of accessing the most socially advantaged schools ferent choice opportunities and different student mobility patterns within the public sector. Furthermore, we find a negative effect be- (Burgess et al., 2017; Kuyvenhoven & Boterman, 2020; Oberti & tween the level of tract median household income and the distance Savina, 2019). This body of evidence has challenged the supposed travelled. This shows that people living in more affluent areas tend to virtues of open enrolment policies, and it has highlighted the diffi- travel shorter distances to school. As shown in Models 5, 6, and 7, the culties in balancing school choice and equity (OECD, 2019). However, effect of the tract median household income is mediated by the for more accurate explanation of how spatial education inequalities composition of the proximity school area, since the size of this are reproduced, we need to understand the interaction between coefficient diminishes by 30% in Model 6 compared to Model 4. We student social background and the characteristics of the school and find also that those living in areas with higher dispersion of income the neighbourhood. This article has demonstrated how this interac- tend to travel shorter distance to school (Model 5). This effect is tion produces highly heterogeneous mobility patterns. Distance tra- mediated by the school composition and the school composition of velled to school certainly depends on the social characteristics of the proximity area, which reduce a 19% its effect size. choosers and their bounded rationalities of school choice. Socially When we introduce the composition of the attended school in advantaged students tend to travel longer distances compared to Model 6, we find a strong and positive correlation between this their peers from lower socioeconomic backgrounds. However, this variable and the distance travelled to school. This shows that those relationship is significantly mediated by educational and neighbour- students attending more socioeconomically advantaged schools are hood factors. For instance, living in wealthier areas eliminates mo- more likely to travel longer distances, which demonstrates the role bility incentives, as schools around residential areas have socially SCANDURRA ET AL. | 13 of 15 advantaged and more homogeneous social composition than average. derived from the different possibilities to mobilise both sorts of ca- Likewise, students facing more mobility restrictions (either geo- pital. Paradoxically, while open enrolment policies may attempt to graphical or economic) may have fewer mobility opportunities, de- neutralise inequalities in positional capital, different situational assets spite living in areas with a more disadvantaged school composition. end up making inequalities even greater. Together, these situations produce many combinations resulting from The unequal capacity to mobilise spatial capital become even the neighbourhood characteristics, the provision of public schooling more determinant in contexts of increasing gentrification. However, options, the type of school supply, and the socioeconomic status of the relationship between student mobility and neighbourhood gen- the student. Since families' choice sets are geographically con- trification is not a linear one. While some studies show higher levels strained, microlocal educational markets produce high heterogeneity of white flight in gentrified neighbourhoods (Candipan, 2019; in mobility patterns. Our findings are aligned with several studies that Pearman, 2019), others provide evidence of simultaneous processes have underlined the crucial role of the geography of education to of white flight and ‘integration’ in local schools by middle classes in understand school choice and distances travelled to school in other gentrified neighbourhoods (Boterman, 2020; Posey‐Maddox et al., cities. Access to good schools combined with mobility restrictions 2016). These complex responses are contingent on the neighbour- are key factors to understand choice opportunities (Andersson hood and educational resources, a result that is better aligned with et al., 2012; Burgess et al., 2011; Hamnett & Butler, 2011; the many mediating factors that we find in our study. Oberti, 2007). Certainly, people's place of residence is a determinant Our findings are undoubtedly relevant for educational policy of educational opportunities, but the neighbourhood and the edu- and planning. First, ensuring a minimum number of schools as cational characteristics close to where people's live may activate ‘neighbourhood’ or proximity schools to equalise choice opportu- different strategies of school choice that add high complexity to the nities does not improve education equity. Differences in the spatial relationship between choice and distance (Malmberg et al., 2014). distribution of schools and differences in spatial capital increase the The capacity of parents to work the school admissions system or the polarisation of the social composition of schools. If schools are not information or transport costs of exercising choice enable some fa- equivalent in their quality and social composition, having a minimum milies better than others to bypass local schools (Burgess & number of options appears to be insufficient as a fair policy. Second, Briggs, 2010, p. 647). and as a consequence, policies that boost market mechanisms in The interaction between individual, educational and neighbour- education, such as increasing school choice opportunities, would hood characteristics and distance is complex and is influenced by need to take into account key aspects that are rarely included when many mediating factors. Factors that appear significant in a bivariant policies are designed, such as the heterogeneous geographies of analysis lose their effects when included together with the educa- education, the unequal spatial capital of the choosers and the dif- tional and social characteristics of proximity schools, and vice versa. ferent school choice rationalities. If equity and school choice have This reveals the crucial role of family strategies in school choice. to be balanced, then other variables must be accounted for in the Different rationalities of school choice (Bosetti, 2004) are activated policy design equation. Finally, the heterogeneity of educational depending on people's social position and specific socialisation stra- geographies, the microdecisions of choosers and the cleavage be- tegies in specific urban settings (Lareau, 2014). These differences tween social groups puts into question the appropriateness of un- make people to activate their spatial capital in different ways, and opt ique policy instruments for different local educational markets in or opt out of local school accordingly. within the same city. The diversity of mechanisms that drive Although Barcelona adopts a controlled system in an attempt choosers to opt out of neighbourhood schools, choose a public or to equalise school choice options, educational opportunities and private school, or travel longer distances, explains why the same travel‐to‐school patterns are far from being equally distributed policy devices impact differently in different territories and for among social groups. These mobility patterns generate higher different social groups. The fact that choosers behave in many ways levels of school segregation than would result from choosing under the same policy conditions underlines the need for more neighbourhood schools or schools within the catchment area flexible and context‐sensitive policy designs to neutralise or com- (Bonal et al., 2020). Exerting school choice increases education pensate for the unintended effects on education inequalities. This inequalities and at the same time generates conditions for its flexibility may affect decisions that alter educational supply or reproduction. As our analysis shows, the search for better school constrain or widen school choice when needed (e.g., by designing composition drives decisions on travel‐to‐school distance. These catchment areas under different criteria). decisions, in turn, increase segregation processes that influence This article shows the potential of spatial analysis to better un- subsequent school choice processes. derstand how the interaction between educational supply and de- The concepts of positional and situational capital coined by Lévy mand produce different mobility patterns. Despite some limitations (1994) appear to be relevant to interpret our findings. The unequal to our approach (such as the absence of data on transportation costs), geographies of education, as shown in Figure 3, provide choosers our findings underline the complex effects of controlled choice sys- with different positional opportunities. In addition, differences ob- tems and their shortcomings as a policy strategy to achieve education served in mobility patterns illustrate how situational capital is also equity. As recent evidence indicates, the combination of more free- unequally distributed. There is, therefore, a double spatial inequality dom for families to choose a school and different opportunities to 14 of 15 | SCANDURRA ET AL. access different school types can negatively impact on equity, par- Bonal, X., Verger, A., & Zancajo, A. (2017). Making poor choices? Demand ticularly in terms of school segregation (Eurydice, 2020, p. 15). This rationalities and school choice in a chilean local education market. Journal of School Choice, 11(2), 258–281. https://doi.org/10.1080/ combination is clear in the case of Barcelona, where high levels of 15582159.2017.1286206 school choice coincide with significant diversity of schools regarding Bonal, X., & Zancajo, A. (2018). Demand rationalities in contexts of their public or private ownership and their social composition. poverty: Do the poor respond to market incentives in the same way? Therefore, the success of market‐led educational reforms as a path International Journal of Educational Development, 59, 20–27. https:// doi.org/10.1016/j.ijedudev.2017.10.001 towards equity is not supported by evidence. This finding indicates Bonal, X., Zancajo, A., & Scandurra, R. (2019). Residential segregation and that policymakers should reconsider what have become mainstream school segregation of foreign students in Barcelona. Urban Studies, policy options over the past decades. 56(15), 3251–3273. https://doi.org/10.1177/0042098019863662 Bonal, X., Zancajo, A., & Scandurra, R. (2020). Student mobility and school segregation in an (un)controlled choice system: A counterfactual A C KN O W L E D G E ME N T S approach. British Educational Research Journal. https://doi.org/10. We would like to thank the Consorci d'Educació de Barcelona for their 1002/berj.3694 support and access to databases. Rosario Scandurra also acknowledges Bosetti, L. (2004). Determinants of school choice: Understanding how the support of the Juan de la Cierva Incorporación Programme (Ref. parents choose elementary schools in Alberta. Journal of Education IJC2019‐040056‐I). Adrian Zancajo also acknowledges the support of the Policy, 19(4), 387–405. Bosetti, L., & Pyryt, M. C. (2007). Parental motivation in school choice: British Academy Postdoctoral Fellowship (PF2\180006). Seeking the competitive edge. Journal of School Choice, 1(4), 89–108. Boterman, W., Musterd, S., Pacchi, C., & Ranci, C. (2019). School segregation in A U T H O R C O N TR I B U T I O N contemporary cities: Socio‐spatial dynamics, institutional context and All authors contributed to the manuscript equally. urban outcomes. Urban Studies, 56(15), 3055–3073. Boterman, W. R. (2013). Dealing with diversity: Middle‐class family households and the issue of “Black” and “White” schools in D A TA A V A I L A B I L I T Y S T A T E M E N T Amsterdam. Urban Studies, 50(6), 1130–1147. The data that support the findings of this study are available from Boterman, W. R. (2020). School choice and school segregation in the Consorci d'Educació de Barcelona. Restrictions apply to the avail- context of gentrifying Amsterdam. Housing Studies, 1–22. https:// ability of these data, which were used under license for this study. doi.org/10.1080/02673037.2020.1829563 Burdick‐Will, J., Grigg, J. A., Nerenberg, K. M., & Connolly, F. (2020). Data are available upon request to the authors with the permission of Socially‐structured mobility networks and school segregation Consorci d'Educació de Barcelona. dynamics: The role of emergent consideration sets. American Sociological Review, 85(4), 675–708. ORCID Burgess, S., & Briggs, A. (2010). School assignment, school choice and Rosario Scandurra http://orcid.org/0000-0003-1756-2694 social mobility. Economics of Education Review, 29(4), 639–649. Burgess, S., Greaves, E., & Vignoles, A. (2017). Understanding parental Adrián Zancajo http://orcid.org/0000-0002-4431-8155 choices of secondary school in England using national administrative data. Discussion Paper 17/689. Department of Economics. University REFERENCES of Bristol. http://www.efm.bris.ac.uk/economics/working_papers/ Alegre, M. A., & Benito, R. (2012). ‘The best school for my child?’ Social pdffiles/dp17689.pdf geographies of school choice in Barcelona. British Journal of Sociology Burgess, S., Greaves, E., Vignoles, A., & Wilson, D. (2011). Parental choice of of Education, 33(6), 849–871. primary school in England: What types of school do different types of Andersson, E., Malmberg, B., & Östh, J. (2012). Travel‐to‐school family really have available to them? Policy Studies, 32(5), 531–547. distances in Sweden 2000–2006: Changing school geography Butler, T., & Hamnett, C. (2007). The geography of education: with equality implications. Journal of Transport Geography, 23, Introduction. Urban Studies, 44(7), 1161–1174. 35–43. Butler, T., & van Zanten, A. (2007). School choice: A European Ball, S. J., Bowe, R., & Gewirtz, S. (1995). Circuits of schooling: A perspective. Journal of Education Policy, 22(1), 1–5. sociological exploration of parental choice of school in social class Candipan, J. (2019). Neighbourhood change and the neighbourhood‐ contexts. The Sociological Review, 43(1), 52–78. school gap. Urban Studies, 56(15), 3308–3333. Barthon, C., & Monfroy, B. (2010). Sociospatial schooling practices: A Candipan, J. (2020). Choosing schoolsin changing places: Examining spatial capital approach. Educational Research and Evaluation, 16(2), school enrollment in gentrifying neighborhoods. Sociology of 177–196. Education, 93(3), 215–237. Bayona‐i‐Carrasco, J., & Domingo, A. (2021). School segregation of migrants Cantillon, E. (2017). Broadening the market design approach to school and their descendants in a dual school system: The case of Barcelona. choice. Oxford Review of Economic Policy, 33(4), 613–634. Population Space and Place, 27. https://doi.org/10.1002/psp.2446 Cascio, E., Gordon, N., Lewis, E., & Reber, S. (2008). From brown to busing. Bell, C. (2009). Geography in parental choice. American Journal of Journal of Urban Economics, 64(2), 296–325. Education, 115(4), 493–521. CEB. (2012). Inici del curs escolar 2012‐2013. Dossier de premsa. Consorci Ben‐Porath, S. R. (2009). School choice as a bounded ideal. Journal of d'Educació de Barcelona. https://www.edubcn.cat/rcs_gene/extra/ Philosophy of Education, 43(4), 527–544. 01_sala_de_premsa/0_Dossier_premsa_inici_curs_2012_13.pdf Benson, M., Bridge, G., & Wilson, D. (2015). School choice in London and CEB. (2020). Alumnat curs 2019‐20. Tots els ensenyaments. Consorci Paris—A comparison of middle‐class strategies. Social Policy and d'Educació de Barcelona. Administration, 49(1), 24–43. Chubb, J., & Moe, T. (1990). Politics, markets & America's schools. The Billingham, C. M., & Hunt, M. O. (2016). School racial composition and Brookings Institution. parental choice. Sociology of Education, 89(2), 99–117. SCANDURRA ET AL. | 15 of 15 Chumacero, R. A., Gómez, D., & Paredes, R. D. (2011). I would walk 500 Makris, M. V. (2018). The chimera of choice: Gentrification, school choice, miles (if it paid): Vouchers and school choice in Chile. Economics of and community. Peabody Journal of Education, 93(4), 411–429. Education Review, 30(5), 1103–1114. Malmberg, B., Andersson, E. K., & Bergsten, Z. (2014). Composite Dumay, X., & Dupriez, V. (2008). Does the school composition effect geographical context and school choice attitudes in sweden: A matter? Evidence from Belgian data. British Journal of Educational study based on individually defined, scalable neighborhoods. Annals Studies, 56(4), 440–447. of the Association of American Geographers, 104(4), 869–888. Easton, S., & Ferrari, E. (2015). Children's travel to school‐the interaction MEFP. (2019). Sistema estatal de indicadores de la educación. Ministerio de of individual, neighbourhood and school factors. Transport Policy, 44, Educación y Formación Profesional. 9–18. Mordechay, K., & Ayscue, J. B. (2020). Does neighborhood gentrification Elacqua, G., Martínez, M., Salazar, F., & Santos, H. (2011). Geografía de las create school desegregation? Teachers College Record, 122(5), 111–135. oportunidades educacionales en Chile. Instituto de Políticas Publicas, Musterd, S. (2005). Social and ethnic segregation in Europe: Levels, Universidad Diego Portales, 1–16. causes, and effects. Journal of Urban Affairs, 27(3), 331–348. Ely, T. L., & Teske, P. (2015). Implications of public school choice for Oberti, M. (2007). Social and school differentiation in urban space: residential location decisions. Urban Affairs Review, 51(2), 175–204. Inequalities and local configurations. Environment and Planning A, Eurydice. (2020). Equity in school education in Europe: Structures, policies 39(1), 208–227. and student performance. Eurydice report. Publications Office of the Oberti, M., & Savina, Y. (2019). Urban and school segregation in Paris: The European Union. complexity of contextual effects on school achievement: The case of Ferrari, E., & Green, M. A. (2013). Travel to school and housing markets: A middle schools in the Paris metropolitan area. Urban Studies, 56(15), case study of Sheffield, England. Environment and Planning A, 45(11), 3117–3142. 2771–2788. OECD. (2012). Equity and quality in education: Supporting disadvantaged Fjellman, A. M., Yang Hansen, K., & Beach, D. (2019). School choice and students and schools. Author. implications for equity: The new political geography of the OECD. (2019). Balancing school choice and equity: An international Swedish upper secondary school market. Educational Review, perspective based on Pisa. Author. 71(4), 518–539. Pearman, F. (2019). Gentrification, geography, and the declining enrollment of Gulosino, C., & Lubienski, C. (2011). School's strategic responses to neighborhood schools. Urban Education, 55(2), 183–215. competition in segregated urban areas: Patterns in school locations Posey‐Maddox, L., Kimelberg, S. M. D., & Cucchiara, M. (2016). Seeking a in metropolitan Detroit. Education Policy Analysis Archives/Archivos ‘critical mass’: Middle‐class parents' collective engagement in city public Analíticos de Políticas Educativas, 19, 1–29. schooling. British Journal of Sociology of Education, 37(7), 905–927. Hamnett, C., & Butler, T. (2011). “Geography matters”: The role Ramond, Q., & Oberti, M. (2020). Housing tenure and educational distance plays in reproducing educational inequality in East opportunity in the Paris metropolitan area. Housing Studies, 1–21. London. Transactions of the Institute of British Geographers, 36(4), https://doi.org/10.1080/02673037.2020.1845304 479–500. Schneider, M., Teske, P., & Marschall, M. (2000). Choosing schools: He, S. Y., & Giuliano, G. (2018). School choice: Understanding the trade‐ Consumer choice and the quality of American school. Princeton off between travel distance and school quality. Transportation, 45(5), University Press. 1475–1498. Sykes, B. (2011). Spatial order and social position. University of Amsterdam. Holt, L., Bowlby, S., & Lea, J. (2019). Disability, special educational needs, Sykes, B., & Musterd, S. (2011). Examining neighbourhood and school class, capitals, and segregation in schools: A population geography effects simultaneously: What does the Dutch evidence show? Urban perspective. Population Space and Place, 25, e2229. https://doi.org/ Studies, 48(7), 1307–1331. 10.1002/psp.2229 Wilson, D., & Bridge, G. (2019). School choice and the city: Geographies of Hoxby, C. M. (2003). The economics of school choice. University of Chicago allocation and segregation. Urban Studies, 56(15), 3198–3215. Press. Zancajo, A., & Bonal, X. (2020). Education markets and school segregation: Ichou, M., & van Zanten, A. (2019). France: The increasing recognition of A mechanism‐based explanation. Compare: A Journal of Comparative migration and ethnicity as a source of educational inequalities. In P. and International Education, 1–18. https://doi.org/10.1080/ Stevens, & G. Dworkin (Eds.), The Palgrave handbook of race and 03057925.2020.1858272 ethnic inequalities in education. Palgrave MacMillan. Zancajo, A., Fontdevila, A., Verger, A., & Bonal, X. (2021). Regulating INE. (2020). Las Cifras de la Educación en España. Estadísticas e Indicadores. public‐private partnerships, governing non‐state schools: An equity Instituto Nacional de Estadística. perspective. Background paper prepared for the 2021 Global Kuyvenhoven, J., & Boterman, W. (2020) Neighbourhood and school Education Monitoring Report. effects on educational inequalities in the transition from primary to van Zanten, A. (2009). Competitive arenas and schools' logics of action: A secondary education in Amsterdam. Urban Studies. https://doi.org/ European comparison. Compare, 39(1), 85–98. 10.1177/0042098020959011 Kye, S. H. (2018). The persistence of white flight in middle‐class suburbia. Social Science Research, 72, 38–52. SUPP ORTING INFO RM ATION Lareau, A. (2014). Schools, housing, and the reproduction of inequality. In Additional supporting information may be found in the online version A. Lareau, & K. Goyette (Eds.), Choosing homes, choosing schools. Russels Sage Foundation. of the article at the publisher’s website. Ledwith, V., & Reilly, K. (2013). Two tiers emerging? School choice and educational achievement disparities among young migrants and non‐ migrants in Galway city and urban fringe. Population Space and Place, 19, 46–59. https://doi.org/10.1002/psp.1708 How to cite this article: Scandurra, R., Zancajo, A., & Bonal, X. Lévy, J. (1994). L'espace légitime Sur la dimension géographique de la (2021). Opting out of neighbourhood schools: The role of fonction politique. Les Presses de Sciences Po. local education markets in student mobility. Population, Space Lubienski, C., & Dougherty, J. (2013). Mapping educational opportunity: Spatial analysis and school choices. American Journal of Education, and Place, e2542. https://doi.org/10.1002/psp.2542 115(4), 485–491.