Abstract
The paper aims to show how and to what extent the system of compulsory education in Milan is affected by social and ethnic segregation. We argue that, despite being guided by the general criteria of universal access and equality of treatment, not only do Milan’s schools fail to counter socio-economic inequalities and differentiation along ethnic lines in an effective manner, but they actually tend to amplify and entrench them. We begin with a theoretical discussion of the main factors contributing to school segregation and a general overview of Italy’s compulsory education system. This is followed by a presentation of the empirical case of Milan, analysing social and ethnic segregation of children of primary school age (i.e. 6–10 years) by place of residence and school of enrolment. As a clear gap emerges between the ‘natural’ and the ‘actual’ school composition, our analytical focus then shifts to home-to-school mobility as an expression of parental choice. We show that 56% of all students in Milan do not enrol at local state schools and this is due to two main phenomena: families choosing private schools and families moving within the state school system. The analysis of these movements makes it possible to identify avoidance dynamics (i.e., in which disadvantaged or ethnic areas are avoided), as well as incoming mobility towards private schools and state schools located in affluent areas or with a lower intake of pupils of non-Italian ethnic backgrounds.
Introduction
School segregation occurs when there is a significantly higher concentration of a minority group in a school than exists in the city as a whole. There are three main drivers of school segregation: the residential distribution of the population, regulations concerning school placement and parental choice concerning their children’s education. This article analyses the impact of these drivers on ethnic and social segregation in the mandatory school system of Milan. More specifically, we analyse to what extent home-to-school mobility of children, which is considered an expression of parental choice, reshapes the characteristics of the compulsory education system, which is based on the principles of universal access and equality even though it allows freedom of choice. The main questions addressed in the paper are: to what extent does residential segregation mirror school segregation? What is the impact of home-to-school mobility of children on socio-economic inequalities and ethnic differentiation found in the school system? And finally, what is the association between social and/or ethnic characteristics of the territory and the probability of home-to-school mobility?
The case of Milan is important for two main reasons. First, school segregation in southern Europe has been a neglected issue to date, if compared with research on northern and continental countries (with a few exceptions, such as Maloutas and Lobato, 2015, Barberis and Violante, 2013, and Santangelo et al., 2018). Therefore, this article represents a step forward in the knowledge of school segregation dynamics of southern European cities. These cities are historically characterised not only by a peculiarly mixed socio-spatial structure (Arbaci, 2008), but also by comprehensive, universalistic school systems. In the last two decades, immigration and increased spatial inequality have hugely altered the ethnic and social composition of these cities, with still-unclear effects on home-to-school mobility and segregation in schools. Second, our local case is relevant to the study of school segregation as it provides a spatial and statistical account of the specific impact of students’ mobility on the composition of school intakes, taking territorial segregation into account. Furthermore, our analysis allows us to consider the association between urban segregation and school choice, shedding light on the flight of a large number of children away from poor and/or ethnicised urban areas. Ultimately, this analysis shows to what extent school choice is a crucial vector of higher urban inequality in a city characterised by low-to-medium residential segregation.
The article is organised in six sections. In the next section the main factors contributing to the spread of ethnic and social segregation in schools are discussed from a theoretical perspective. The following section examines institutional aspects of the compulsory education system in Italy and Milan, with a specific focus on enrolment regulations. While the next section presents the data and methods deployed, the penultimate section focuses on the empirical study. First, ethnic and social segregation in Milan along residential lines are reconstructed. The focus then shifts to differences between school-age populations in specific catchment areas and actual school student populations in those areas. The analysis shows significant home-to-school mobility, with 56% of all enrolments in Milan being at schools outside the catchment areas (CA hereafter) of residence. An analysis of these flows shows that mobility is mainly due to parents choosing private schools or a state school that is not located in the CA where they live. Two aspects are then analysed: on the one hand, urban areas in which flows out are most evident, that is, those neighbourhoods from which families ‘flee’ when choosing schools, and, on the other hand, the schools attracting the students moving out from their CA. Finally, we present an analysis of the probability of pupils enrolling at schools outside their CA as a proxy of school choice. To conclude, we summarise the main peculiarities of the case of Milan and highlight the value of such an analysis for studies of ethnic and social segregation in schools.
Drivers of school segregation: A general discussion
Urban scholars have traditionally considered residential distribution of the population as the most significant driver of school segregation. A large body of studies has addressed the relationship between residential and school segregation, particularly in the American context, where the link between the two phenomena is relatively strong and where ‘school segregation’ is a recurrent topic in public debate (Hochschild and Scovronick, 2003; Orfield et al., 1997; Scott, 2005). Similar studies of European cities have produced contrasting findings. Spatial distribution of the population is the main determinant of school selection in these cities too (Boterman, 2019). However, the association is often weaker than in US cities and in many cities high levels of ethnic segregation in schools exist alongside relatively low levels of residential segregation (Gramberg, 1998; Rangvid, 2007). This is due to a number of factors linked to different residential segregation patterns which are more common in Europe. Historically, ethnic minority groups in European cities have been less marginalised than in the USA; the concept of residential segregation does not apply to European cities to the same extent (Tammaru et al., 2016). Moreover, greater diversity in the spatial organisation, population settlement patterns and migration history of European cities produces a greater variety of outcomes in terms of residential and segregation dynamics (Musterd, 2005). These aspects are even more important in southern European cities, where socio-spatial inequalities are generally lower and non-EU migration has emerged as a relevant issue only recently (Arbaci, 2008). A second key driver of school segregation is the institutional regulation of the school system, and specifically of school choice. Many cities over the last 20 years have introduced market criteria, thus reducing constraints on parental choice. Such reforms have generally been supported by an underlying ideological discourse based on competitiveness and performance. In the view of many scholars, this surge in liberalisation of parental choice has not only allowed greater geographical mobility but has also confirmed ‘the perpetuation of inequality and, in particular, the ongoing middle-class advantage in the education field’ (Benson et al., 2015: 2). These reforms have exacerbated class divisions by providing middle-class parents with greater school choice while reinforcing the weaker position of working-class families (Reay and Ball, 2007).
Nevertheless, research evidence is ambiguous about the overall impact of liberalisation of parental choice on school segregation. While many studies show that the introduction of free choice can exacerbate segregation dynamics (Le Grand and Bartlett, 1993; Warrington, 2005) and even lead to polarisation (Gewirtz et al., 1995), cases have been also reported in which free choice has mitigated exclusion dynamics (Finn, 1990; Merrifield, 2001). These apparently contradictory results are likely to depend on the scale at which the phenomenon is analysed: as Gibson and Asthana (2000) note, the appropriate geographical scale for verifying whether such a polarising effect exists is the local one. Indeed, if a neighbourhood perspective is adopted, a polarising effect is observed in areas already characterised by marked diversity in terms of population composition and choice set. The losers seem to be the already disadvantaged schools and families.
Finally, school segregation is also affected by parental agency. The literature on school choice shows that it is strongly class-based. According to Van Zanten and Obin (2010), it is mainly middle-class households who actively make choices, while working-class households are ‘passive choosers’ (Haylett, 2003; Oberti, 2007). The main criteria driving the school choice of middle-class parents seems to be the search for social homogeneity (Boterman, 2013), followed by the popularity of the school (Gibbons and Machin, 2003), academic standards (Allen and Burgess, 2013) and the school climate (Boterman, 2013). Household agency has more to do with avoidance strategy than deliberate choice of a specific school, giving rise to what has been labelled a ‘white flight’ in the USA (Crowder, 2000; Renzulli, 2014). Household agency has acquired greater importance with the introduction of free choice promoting ‘an ethical framework that encourages and legitimates self-interest in the pursuit of competitive familial advantage’ (Reay et al., 2011: 7). Parental choice is not simply a rational process based consciously on the goal of social reproduction; it is infused with emotions (Lobato and Groos, 2019), it is framed by norms and values and it is affected by the need for conformity to dominant ideas on what is a ‘good match’ (Noreisch, 2007; Van Zanten, 2013).
These three drivers combine in different ways, fostering diverse geographies of school segregation. While the relationship between residential and school segregation is not so strong as in US contexts, the socio-economic characteristics of territories play an important role also in southern European cities, despite their relatively low levels of segregation.
While it is quite clear how different institutional rules affect and constrain households’ agency, the relation between parental choice and territory is less evident. What has been observed so far is that the direction and the intensity of agency, in terms of home-to-school mobility, change according to the socio-economic characteristics of the urban areas. This article, through a spatial analysis, attempts to enlighten this relationship.
The school system in Italy and in Milan: The institutional framework and policy developments
The Italian education system is based on two tiers: a five-year primary education and three-year lower secondary education programme with a common curriculum, followed by three to five years of upper secondary education with students choosing different study programmes (Barberis and Violante, 2013). Private schools (most of which are run by religious institutions) are granted equal status to state schools but cannot be publicly funded under the Italian Constitution (although in practice they actually receive some public funding if their curriculum conforms to the ministerial guidelines).
Until the mid-1980s, each primary or lower secondary state school was assigned a corresponding CA for its student intake. Parents who were not satisfied with schools in their CA could choose only to move their children to a private school. In the mid-1980s, parents were allowed to choose a school outside their CA unless the school in question was fully subscribed, in which case students from the CA had priority (a rare situation because of the steady trends in the size of Milan’s child population and the possibility for public schools with over-enrolment to employ more teachers). The aim of the reform was to foster greater mobility and reduce segregation. At the same time, since 2000, state schools have enjoyed the status of autonomous bodies in terms of administration, teaching and organisation. Within a national and regional regulatory framework, they can design their own educational offer and regulate student admission. The high degree of autonomy granted to schools has increased competition between them to select the ‘best students’, thus indirectly contributing to segregation (Barberis and Violante, 2013).
In Milan, the compulsory state education system is evenly spread across the city, with 137 primary schools and 87 lower secondary schools and an average of 330 students enrolled at each school (standard deviations: 142 for primary and 154 for lower secondary schools). In 2000, the Lombardy Region introduced a voucher system entitling families who enrol their children at private schools to a yearly benefit of 300 euros. Although this provides a very modest incentive to choose private schooling (reducing the total cost of the private school tuition by an average of just 10%), Milan has 75 private primary schools accounting for 26% of the total school population (with 51 private lower secondary schools representing 16% of students). The quite considerable provision of private schools is comparable to that of other southern European cities, such as Barcelona (Bonal, 2002) and Lisbon (Esteves Bea, 2018).
To conclude, in Milan the education system is designed according to general criteria of universal access, highly standardised educational programmes and a predominantly state-run system. Nevertheless, not only are private schools relatively numerous, but the reduced role of CA makes it impossible to effectively use them as a buffer against wider school segregation. The introduction of free choice and competition between schools has paved the way for widespread mobility in terms of school enrolments in the city.
Data and methods
Data about students and school population come from the 2015/2016 school registers (AnaSco – Anagrafe Scolastica) of Milan municipality, supplemented with 2011 national census data and standardised national Ministry of Education performance tests (INVALSI). We created an integrated database combining information regarding both schools and the population resident in the CAs.
CA is used as a geographical unit of analysis for maps and statistics on the distribution of school-age population, enrolments and home-to-school mobility. Though CAs are not mandatory since the mid-1980s (but still in use in specific circumstances as shown in the previous section), their geographical boundaries (designed by the education department of the municipality) delimit the area of residential proximity for each school, allowing an analysis of home-to-school mobility that is sensitive to the geographical distance from school of each student.
We focus in this article only on primary schools (though data on lower secondary schools are also available) to prioritise the school system with the best-developed territorial distribution (and smaller CAs). Moreover, the choice of primary school is the most affected by home-to-school distance. Data about lower secondary schools do not significantly vary and are available upon request.
The home-to-school mobility is analysed for both Italian and foreign pupils. Foreign children are identified by their citizenship status. When comparing Italians and foreigners, children from OECD countries are considered equivalent to Italians.
Segregation is measured through the calculation of dissimilarity indexes (DI). Residential segregation is calculated on children aged 6–10 years old considering all the 137 primary school CAs of the city, while school segregation is estimated on the distribution of enrolments in the 217 primary schools located in Milan.
Lack of citizenship from an OECD country is used as an indicator to estimate ethnic segregation and the DI is computed for each national group and overall for non-OECD foreigners.
Social segregation is measured considering the education level as an indicator of socio-economic status and comparing the distribution of parents (individuals living in households with children in the 6–10-years age group are used as a proxy of parents in the estimation of residential segregation) with tertiary education with that of parents without a degree. Finally, a regression analysis is carried out to estimate the likelihood to move away from the local state school based on the characteristics of the local CA. We ran a multinomial logistic model with the dependent variable being the school of enrolment (classified into three categories: local state schools, state schools outside the CA of residence, and private schools). Both the socio-economic characteristics of the CAs (a synthetic index based on a factor analysis considering education levels, unemployment rates, professions and share of population living in public housing) and their ethnic make-up were considered simultaneously in order to estimate their net effects. Variables relating to additional characteristics of each CA (age structure, presence of overcrowding) and to local schools (performance in standardised nationwide learning tests, presence of full-day classes, ICT equipment, share of disabled students) were also included. Finally, coefficients resulting from the regression were used to estimate the conditional probabilities of moving to a not local (either public or private) school.
Residential and school segregation
In this section we analyse to what extent the distribution of primary school attendance reflects the residential distribution of the child population in Milan. First, we consider the residential distribution of the child population (6–10 years old). Second, we compare it with school enrolments to identify differences in the socio-economic and ethnic distribution of students.
Following the pattern of other southern European cities (Arbaci, 2008; Malheiros, 2002), residential segregation in Milan is not particularly strong. The share of foreign population aged 6–10-years is equal to 19.8% in 2016 (increasing from 7% in 2001). The DI comparing the distribution of the foreign school population with that of Italian children is 0.35, an average value compared with other European cities (Tammaru et al., 2016). The same index calculated for specific national groups is the lowest for children from the Philippines, Sri Lanka, Peru and Ecuador, while it is higher for Moroccan and Chinese children, for which the DI is higher than 0.50 (see Table 1). A similar situation is observed regarding socio-economic status: the DI for individuals with tertiary education compared with those without a degree is 0.31.
DI for the ten largest national groups (reference group = Italians). Children aged 6–10 years.
Source: Our elaborations are based on AnaSco database.
The residential distribution of children of primary school age (see Figure 1) exhibits a twofold pattern. On the one hand, pupils are spatially dispersed according to a pattern of geographical polarisation along the centre–periphery axis. Foreign children and children of low socio-economic status are barely represented in the city centre (considering the two innermost rings of the city), where Italian children and children of higher socio-economic status are highly concentrated. On the other hand, in the outer, peripheral area, children from low-status households and foreign children are concentrated in specific areas, making for a ‘patchy’ distribution of social disadvantage and ethnic diversity and highly differentiated outer-city areas.

CAs by share of resident foreign children (2015–2016 school year) and by share of individuals with tertiary education in resident households with children of school age.
The distribution of children attending state primary schools partially differs from the situation described so far. The DI for all non-OECD foreign students rises from 0.35 to 0.43. Levels of segregation increase for every major ethnic group (see Table 1) and, particularly, for children from the Philippines and Albania, while the lowest increase is registered for Chinese children (whose residential segregation is very high).
To analyse the gaps between school and residential ethnic segregation, Figure 2 compares the share of foreign children in each CA (x-axis) and corresponding state school (y-axis). Results show that most of the schools located in areas with a low share of foreigners (0–10%) generally have a lower percentage of students of foreign origin (between 5 and 20%, with only two exceptions). As we saw in the previous section (Figure 1), almost all these schools are located in the central area of the city where social and ethnic homogeneity is very high. In contrast, in CAs with a high share of resident foreign students (between 30% and 50%), mostly located in peripheral areas (see Figure 1), the ethnicity level of the schools is consistently higher: in more than half of these schools it rises to over 50% and there are seven schools with more than 60% of foreign students. Finally, in areas where foreign children constitute 20–30% of the school-age population, the picture is interestingly mixed. In a significant number of schools, the percentage of students of foreign origin is lower than in the residential population of the corresponding CA. In other schools, by contrast, the percentage of foreign students is higher than in the corresponding CA.

Comparison between share of foreign residents of school age by CA (horizontal axis) and share of foreign students in corresponding state school (vertical axis). Line is the bisector.
It is in the areas with an ethnic mix that polarisation within the school system takes place. This is the result of locally, micro-territorial dynamics of school segregation, which contrast with the ethnically mixed composition of the territory, making social and ethnic divisions in some schools much sharper than the residential distribution of the population. Figure 3 shows some examples of such dynamics. While some schools in these areas have a substantially higher concentration of foreign students compared with the local population, the share of foreign students decreases in other schools located in proximity to such areas, showing that Italian and foreign children who are resident in the same areas consistently attend different primary schools.

Difference between share of foreigners in state schools and share of foreigners in the CA in selected areas with 20–30% population of foreign children.
School segregation is higher than residential segregation if we also consider the spatial distribution of social inequalities. Our data enable an analysis of the highest educational attainment of the children’s parents. From Figure 1 we already know that the children of parents with a tertiary level of education are highly concentrated in the central areas of the city. School enrolment exacerbates this situation: the DI for these children rises from 0.31 (when calculated for the CAs of residence) to 0.37 when computed per school. Figure 4 compares the share of children with highly educated parents in CAs and in related schools. In relatively well-off areas (where the percentage of parents with a tertiary level of education is higher than 60%), schools frequently have a greater percentage of highly educated parents. In CAs where parents with a degree number less than 22%, the percentage of tertiary-educated parents registered at local state schools is generally even lower. In intermediate areas, higher levels of polarisation are again evident, with a contrast between schools with a higher percentage of well-educated parents and schools with a lower percentage of socially disadvantaged children.

Comparison between share of individuals with tertiary education in CAs (horizontal axis) and share of tertiary-educated parents of children by school (vertical axis). The line is the bisector.
To sum up, the composition of state primary schools reproduces, and to some extent reinforces, the spatial distribution of social disadvantage. In other words, in more disadvantaged and/or more ethnically mixed areas, mainly located in peripheral areas, ‘when it rains, it pours’.
The final effect of such dynamics is that in schools located in the most critical peripheral areas, the concentration of socially disadvantaged and/or foreign children is much higher than that found locally. Schools therefore contribute to the entrenchment of ethnic and social divisions within the students’ population in Milan, and the final impact of school attendance in terms of polarising or segregating populations is higher than the residential settlement of the population in the city. This result shows that school segregation does not depend only on residential segregation but is also due to the mobility of the student population within the city’s educational system, and that this dynamic increases the level of ethnic and social segregation of the child population. This is the focus of the next section.
School choice: Analysing home-to-school mobility
In Milan, 56% of Italian children and 40% of foreign children do not choose the state school located in their residential CA. The mobility of Italian children is the sum of two different choices: enrolment in private schools or choice of a state school located in a CA different from that of residence.
More than one-fifth (22%) of Italian students enrol at private schools. Many reasons may account for this choice, for example a higher incidence of traditional religious beliefs in more affluent areas of the city (where two-thirds of pupils attend religious schools), the widespread perception that state education is of lower quality, fear of urban and cultural heterogeneity, search for social identification or better quality. Whatever the subjective reasons for such a choice, most of the flows towards private schools originate from central areas of the city (see Figure 5). The social composition of the private-school population thus reflects the high social status of Italian residents in the central areas of the city, and private schools reinforce the wide socio-spatial separation of this social class from the rest of the population. Only 3% of private school students are of foreign origin and only a small number come from a lower social background. Students attending private schools experience therefore a highly homogeneous social environment.

CAs by percentage of Italian students attending a public school outside their CA (left) and by percentage of Italian students attending a private school; 2015–2016 school year.
Moreover, the large flow of Italian, high social status children towards private schools contributes to a decrease in the social status of children attending state schools, so indirectly increasing the risk of higher ghettoisation of disadvantaged students within the state school system. If students enrolled in private schools went to the state local schools, the DI for ethnic school segregation would drop from 0.43 to 0.36.
Italian students also show high levels of mobility within the state education system. Over 30% of Italian children attend a school outside their CA. If such mobility is favoured by the city’s medium size (there is little incentive for families to apply a residential strategy to catch the best schools given the relative facility of daily commuting in the city), it is also driven by social and ethnic dynamics. The left-hand map in Figure 5 illustrates remarkably high outflows from peripheral areas with a higher concentration of foreigners and children from lower-educated households. The direction of such flows is, however, different between Italians and foreigners, confirming a relevant dynamic of ethnic polarisation in the school system of Milan. The correlation between the quota of Italian students not living in the CA of the school and the same quota of foreigners is indeed negative (−0.21). While Italian children moving away from disadvantaged areas mostly converge towards schools located in central areas (left-hand map, Figure 6), foreign students (right-hand map, Figure 6) show a different path with a high number of them moving into schools located in the intermediate ring of the city and in most cases ignored by Italian students.

State schools by share of Italian students living in another CA (left) and by share of foreign students living in another CA (right); 2015–2016 school year.
The probabilities of school mobility
To analyse school mobility, we calculated the probability of Italian students moving away from the local state school to either a private school or a state school outside their CA of residence and estimated to what extent this probability is associated with the level of social or ethnic segregation of the CA of residence (see Table 2). As previously explained, the estimation is controlled by characteristics of the local schools and further characteristics of the CAs of residence. The estimated probability is the result of an ecological analysis that does not consider the individual characteristics of students. Following a methodological approach consolidated in urban sociology research, the analysis therefore aims to assess to what extent the social and ethnic characteristics of the local contexts are significantly associated with school mobility, without properly considering the individual agency aspect that is specific to school choice.
Conditional probabilities of choosing a state school outside CA of residence or a private school; 2015–2016 school year.
Source: Our estimations based on AnaSco, INVALSI and national census data. Full regression model available in the Appendix.
The data show that primary school Italian students are more likely to attend state schools outside their CA of residence in neighbourhoods with a lower average socio-economic level than in areas with a high share of foreign residents. The probability of moving to another state school is 40.6% for children resident in areas with a very low average socio-economic level and drops to 26.4% for children who live in the most well-off areas. The mobility of Italians within the state school system is lower in areas with a higher share of foreigners (28.6% and 33.6%) than in areas with a lower share of foreigners (35.5%). On the other hand, private schooling is more likely to be chosen in areas with high shares of foreign households (28.1% as against 22.5%) as well as in areas with large proportions of households with a high socio-economic status (34.8% as against 10.7%). Foreign students are less likely to attend schools outside their CA of residence when the share of resident foreign children and socio-economic levels are the highest (41.4% versus 33.9% for the areas with highest quota of Italians). They are also more likely to move away from areas with low socio-economic level.
Two main flows in the school system are therefore identified. The first is within the state school system and includes students (both Italians and foreigners) living in poorer, peripheral areas, who avoid contact with students from comparable or lower socio-economic backgrounds. The second is a shift towards private education among Italian students from high-income households, living either in central, affluent areas, or highly ethnicised areas, whose parents seek a culturally and socially homogeneous school environment. Finally, although mobility among foreign students is also very high, it is limited to the state school system: only 3% of foreigners choose private (mainly religious) schools in Milan.
Conclusions
School segregation in urban contexts has generally been explained as the result of the combination of three factors: residential concentration of homogeneous groups of population, neoliberal policies promoting freedom of choice and competition between public and private schools, and the agency of households who take advantage of this freedom, thereby conveying values, habits and social representations.
Residential segregation is no higher in Milan than in other European cities. Indeed, although the spatial distribution of the population is characterised by high levels of inequality along the centre–periphery axis, this does not translate into significant ghettoisation of the most deprived population groups. Socio-spatial marginalisation is prevented by the multi-dimensional, variegated character of peripheral areas, where ethnic concentration does not always overlap with the spatial distribution of poverty or social exclusion. Milan is traditionally a middle-class, socially mixed city, with relatively few clusters of ‘problematic’ population groups likely to be concentrated in some deprived areas (Mugnano and Costarelli, 2018; Torri and Vitale, 2010). Although urban poverty and socio-economic inequality are higher in Milan than in many other similar European cities (Cucca and Ranci, 2017), this fact is reflected less in the spread of huge peripheral areas than in micro-scale segregation dynamics (Balducci et al., 2017; Del Fabbro, 2017). However, recent economic and demographic trends leading to increased spatial inequality and higher ethnicisation of the population have paved the way for the emergence of higher residential segregation associated with poor housing conditions (Tosi, 2017), which is common to other southern European cities (see also Bonal, 2002).
Regarding the institutional context, the school system of Milan basically reflects prevailing national policy. The Italian education system has traditionally been based on criteria of universal access, mainly public provision of education, as well as highly standardised study programmes and teachers’ qualifications, with private schools playing a marginal role. However, this system has been radically changed in recent years through the introduction of free-market mechanisms, paving the way for intense competition between state and private schools and the recognition of freedom of school choice within the public-school system.
In this urban and institutional context, school segregation in Milan is the result not only of residential segregation, but also of outflows of Italian families.
Two main choice options for parents have emerged from our analysis. On the one hand, more affluent Italian families are more likely to take their children out of the state school system and send them to a private school. It is striking that, in a system traditionally dominated by a universalistic, comprehensive state school system, almost one-quarter of Italian students in Milan attend a private school, and that in the city’s central areas the share is close to 40–50% of total compulsory school attendance. Our analysis has shown that this ‘flight’ is highly concentrated in neighbourhoods with a high proportion of affluent Italian households. Their concentration in private schools makes this specific student population extremely homogeneous both socially and ethnically, and paves the way for social reproduction mechanisms based on self-segregation and a search for homogeneity among ‘élite’ classes in respect of the rest of the population.
On the other hand, we have identified a large ‘white flight’ of Italians (and, to a less extent, of foreigners) from peripheral areas in which households with very low socio-economic status predominate. School choice does not occur indifferently from locality but it results in a widespread home-to-school mobility exiting specific distressed areas. Indeed, our data show that what Italian households attempt to avoid is more a disadvantaged social context than concentration of individuals with a different nationality. Nevertheless, where socio-economic status is coupled with a share of foreign students in excess of the 30% threshold, ‘white flight’ dynamics are particularly strong. The ‘rush towards the centre’ is a manifestation of this avoidance of the most disadvantaged areas. Residents living in these neighbourhoods tend to move their children to schools located in more central areas. Finally, while a relevant share of foreign students moves within the state system, their number is less significant (4000 foreigners versus 14,600 Italians). Moreover, they follow different trajectories than Italian students, therefore increasing school segregation overall.
Considering the spatial aspects of the phenomenon, it is therefore quite clear that home-to-school mobility is more likely in disadvantaged neighbourhoods. This reflects trends found in other cities (Boterman, 2013; Gramberg, 1998; Oberti, 2007; Rangvid, 2007; Vergou, 2015) and confirms the failure of neoliberal reforms in fighting segregation mechanisms in Milan.
The overall impact of such trends is a strong risk of school segregation for students from the poorest or most marginalised segments of the population, that is, children with non-OECD nationality or origins, and/or children from deprived family backgrounds living in peripheral or marginal areas. For these children, the probability of having no choice other than to attend schools with a large concentration of migrants and disadvantaged students is very high.
More generally, what emerges clearly from the case of Milan is that the distribution of school attendance reflects both the spatial distribution of the population and school choice driven by socio-economic inequality and increasing ethnic differentiation. The composition of the school population thus constitutes a significant indicator of the divide between social classes in a city characterised by low-to-medium levels of spatial segregation. Home-to-school mobility has exacerbated these dynamics, as parents are released from the obligation to enrol their children at schools located within their residential CA. In Milan, freedom of choice has certainly fostered a ‘white flight’ dynamic. Our analysis shows that such flight has been highly driven by spatially shaped social and ethnic inequalities.
In conclusion, the case of Milan offers a specific – and hence interesting – pattern of school segregation. Social inequalities partially mitigated by spatial heterogeneity clearly arise in the composition of school student populations through micro-dynamics of polarisation at the neighbourhood scale, huge outflows from poor neighbourhoods, and a wide choice of private schools for more affluent families. Further studies are surely needed to disentangle the reasons behind such household choices in order to fully understand the spatial dynamics highlighted in this article.
Footnotes
Appendix
Multinomial logistic regression model on being enrolled in a state school outside catchment area of residence or in a private school: 2015–2016 school year. Foreign children.
| State school outside catchment area | Private school | |
|---|---|---|
| Variables | ||
| Quota of foreign children in the catchment area. Reference category 0–10% | ||
| 10–20% | −0.203* | −0.369 |
| [0.106] | [0.257] | |
| 20–30% | −0.179 | −0.008 |
| [0.120] | [0.301] | |
| 30–40% | −0.314** | −0.169 |
| [0.126] | [0.317] | |
| >40% | −0.337** | −0.260 |
| [0.134] | [0.348] | |
| Socio-economic level of the catchment area (by quintiles). Reference category: first quintile (very low) | ||
| Low | −0.019 | −0.117 |
| [0.058] | [0.173] | |
| Average | 0.027 | 0.415** |
| [0.077] | [0.208] | |
| High | −0.212** | 0.454* |
| [0.098] | [0.259] | |
| Very high | −0.335** | 0.662* |
| [0.141] | [0.349] | |
| Born in a foreign country | 0.016 | 0.279** |
| [0.049] | [0.126] | |
| Average INVALSI score in Maths for the state school of the catchment area | 0.014** | 0.036* |
| [0.007] | [0.018] | |
| Average INVALSI score in Italian for the state school of the catchment area | −0.018** | −0.054** |
| [0.008] | [0.022] | |
| % of disabled pupils in the state school of the catchment area | −0.001 | 0.005 |
| [0.007] | [0.018] | |
| % of old age residents in the catchment area | 0.028*** | 0.018 |
| [0.006] | [0.016] | |
| % of unused buildings in the catchment area | −0.004 | −0.090*** |
| [0.011] | [0.033] | |
| % of unoccupied dwellings in the catchment area | −0.006 | 0.017 |
| [0.007] | [0.017] | |
| % of classrooms connected with wifi in the state school of the catchment area | 0.000 | −0.001 |
| [0.001] | [0.002] | |
| % of laboratories connected with wifi in the state school of the catchment area | 0.001 | −0.000 |
| [0.001] | [0.002] | |
| % of classrooms with IWBs in the state school of the catchment area | −0.015*** | −0.000 |
| [0.003] | [0.008] | |
| Number of computers in the state school of the catchment area | −0.003*** | −0.001 |
| [0.000] | [0.001] | |
| Presence in catchment area of census tracts with overcrowding situations | 0.202*** | 0.001 |
| [0.056] | [0.166] | |
| Constant | −0.770*** | −3.325*** |
| [0.215] | [0.594] | |
| Observations | 10,661 | 10,661 |
Notes: Base category: enrolled in the state local school. Standard errors in brackets. ***p < 0.01, **p < 0.05, *p < 0.1.
Source: Our estimations based on AnaSco, INVALSI (2011/12 data) and 2011 national census data.
Acknowledgements
We especially thank Sako Musterd, Willem Boterman, and Carolina Pacchi for their comments on previous versions of this paper, and Fabio Manfredini and his team for data management. We thank also the Educational Services Department of Milano municipality and INVALSI for the authorisation to use the data necessary for our analysis.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research presented in this article was funded through a 2016 FARB grant of the Department of Architecture and Urban Studies of the Politecnico of Milan. It was also supported by the Excellence Research Project ‘Fragile Territories’ of the same institution.
