School choice is a relevant aspect of European school systems and an integral part of market-oriented school policies. It is of practical relevance especially in densely populate areas, where choice between schools is actually viable. Considering this and the fact of a worldwide longstanding trend of urbanisation (GMDAC 2019), it is to expect that more and more people will be confronted with such choices and the effects of the underlying policies.

Simply put, the basic assumption of market and choice proponents is that parents, if given the opportunity, will choose good performing schools over bad performing ones (Chubb & Moe 1990; Hoxby 2001). Following the theory further, this choice behaviour is to create a tide lifting all boats by exerting pressure on all schools to improve and to free up students from being bound to their district school, thereby increasing social justice. On the other hand, critics dispute that the above-mentioned effects actually occur comprehensively (Ball 1993; Gewirtz et al. 1995; Bifulco et al. 2009). Instead, they warn of unintended consequences, such as increased disparities and segregation between schools. Their main argument is that many of the proponent’s assumptions (e.g. opting for performance) are empirically invalid or incomplete (e.g. disregard of the correlation between student composition and performance) and that theoretical concepts borrowed from the economic sciences (e.g. assumption of a market) cannot be extended to the field of education without far-reaching restrictions.

Obviously, it is crucial what parents consider to be a good school and which criteria they apply when choosing such. Research findings differ between studies and countries, due to the importance of societal and policy differences among them. However, there are reoccurring findings indicating the significance of academic aspects (performance, reputation, program), student composition, values, and pragmatic reasons (transportation, childcare) (Mayer 2018; Prieto et al. 2018). The neighbourhoods in which schools are located also seem relevant (Schuchart et al. 2011; Phillippo & Griffin 2016). The role of the student composition should be considered in the light of the disproportionally large share of migrants living in Europe’s major cities (GMDAC 2019) and the growing economic segregation within them (Musterd et al. 2016).

The paper aims to contribute to the research on primary school choice criteria in urban settings. This is done by adding an unconventional and supplementing perspective to the typical focus on chosen schools. Instead of only analysing the characteristics of schools chosen by parents, the schools that parents reject i.e., try to avoid, are also considered. Unlike a variety of studies that rely on stated preferences collected via surveys, this paper infers revealed preferences by analysing school authority data on parents' demand for and rejection of individual schools, which offers the advantage of independence from social desirability effects (Prieto et al. 2018). Demand is understood as by Zunker et al. as an “aggregate measure of individual educational or school choice decisions,” meaning “that demand for schools with certain characteristics can be viewed as a cumulative preference for schools with those characteristics” (2018, p. 592).

The research question is, how are characteristics of schools and their neighbourhoods related to differences between schools in demand and rejection?

In Berlin, where the study is located, school enrolment is regulated by school districts, but with the possibility of attending a chosen school upon a substantiated application. In the school year 2019/20, 37% of parents did file such a request. Depending on the city district, between 14% and 31% (Ø 25%, 12 districts) of all newly enrolled students did not attend their assigned school (own calculations).

The data and analyses are part of the author's dissertation project. The sample consists of all public elementary schools in three of Berlin's twelve districts in the 2018/19 school year (n = 93). Private schools are not included because data on them is not available. The focus is on primary schools, as similar studies have already been conducted for secondary schools (Zunker et al. 2018, Neumann & Maaz 2019). The approach chosen here represents a desideratum for the primary school sector. Due to the high population and school numbers, school density, and good public transportation connections, school choice in Berlin is not only theoretical but feasible, making Berlin an ideal study site. Furthermore, districts are diverse in terms of socioeconomic status and migrant proportions (SenAdUDE 2017). The data describing demand and rejection are collected using enrolment and school choice information provided by the local school authorities. The information describes all children to be enrolled in primary schools in the three districts in 2018/19 (n = 10509). If parents wish for their child to attend a non-district school, they can submit a school choice application in which they can indicate up to three choice schools. Building on this information, one variable each is created to describe demand and rejection. Demand (∈ [0, ∞), continuous variable) per school is operationalized as the number of first choice requests relative to enrolment capacity. Rejection (∈ [0, 1]) is operationalized as the percentage of students who attempt to avoid the school by transferring to a school of choice out of all students who were supposed to be enrolled in it, due to living in the school’s district. Due to properties of the data, generalized linear models (gamma with log-link and binomial with logit-link) are used for analysis (Dunn & Smyth 2018). All predictor variables are measured in percent and describe the school and its neighbourhood. Included are three student composition indicators (proportion of students of non-German heritage language, proportion of students whose family receive welfare, school absenteeism), two school quality indicators (sufficient staff, student performance measured as percentage of students with an academic track recommendation) and unemployment and migrant percentages of school neighbourhoods. Data sources are the Berlin Senate Administration for Education, Youth, and Family and the Bureau of Statistics Berlin Brandenburg. An additional factor variable is used to control for the city district. All models are checked for multicollinearity using the VIF.

Different models were calculated for demand and rejection, containing either student composition, school quality, or neighbourhood characteristics. In a few models, there is potentially problematic multicollinearity with respect to some predictors (5

Ball, S. J. (1993). Education Markets, Choice and Social Class: the market as a class strategy in the UK and the USA. British Journal of Sociology of Education, 14(1), 3–19. Bifulco, R., Ladd, H. F. & Ross, S. L. (2009). Public school choice and integration evidence from Durham, North Carolina. Social Science Research, 38(1), 71–85. Chubb, J. E. & Moe, T. M. (1990). Politics, markets, and America's schools. Washington: Brookings Inst. Dunn, P. K. & Smyth, G. K. (2018). Generalized Linear Models With Examples in R. NY: Springer. Gewirtz, S., Ball, S. J. & Bowe, R. (1995). Markets, choice and equity in education. Buckingham: Open Univ. Press. GMDAC. (2019). Data in urbanization and migration, Global Migration Data Analysis Centre. https://migrationdataportal.org/themen/urbanisierung-und-migration Hoxby, C. (2001). Rising tide: New evidence on the competition and the public schools. Education Next, 1(4), 69–74. Mayer, T. (2018). "Das ist einfach dieses, was wir nicht kennen oder nicht wollen". Einzelschulwahl öffentlicher und privater Grundschulen im Kontext von Distinktion, Segregation und sozialer Un-gleichheit. Dissertation. McFadden, D. (1977). Quantitative Methods for Analyzing Travel Behaviour of Individuals: Some Recent Developments. Cowles Foundation Discussion Papers 474. Cowles Foundation for Research in Economics at Yale University. Musterd, S., Marcińczak, S., van Ham, M. & Tammaru, T. (2016). Socioeconomic segregation in Eu-ropean capital cities. Urban Geography, 38(7), 1062–1083. Neumann, M. & Maaz, K. (2019). Nachfrageunterschiede zwischen weiterführenden Schulen. Eine Untersuchung auf Basis schulstatistischer Daten im Land Bremen. In D. Fickermann & H. Weishaupt (Hrsg.), Bildungsforschung mit Daten der amtlichen Statistik. Die Deutsche Schule. Beiheft 14 (S. 197–214). Phillippo, K. L. & Griffin, B. (2016). The Social Geography of Choice. Neighborhoods’ Role in Students’ Navigation of School Choice Policy in Chicago. The Urban Review, 48(5), 668–695. Prieto, L. M., Aguero-Valverde, J., Zarrate-Cardenas, G. & van Maarseveen, M. (2018). Parental preferences in the choice for a specialty school. Journal of School Choice, 5(1), 1–30. Schuchart, C., Schneider, K., Weishaupt, H. & Riedel, A. (2011). Welchen Einfluss hat die Wohnumgebung auf die Grundschulwahl von Eltern? Schumpeter Discussion Papers, 2011-009, 1–23. SenAdUDE, Senate Administration for Urban Development and Environment (2017). Monitoring Social Urban Development 2017. www.stadtentwicklung.berlin.de/planen/basisdaten_stadtentwicklung/monitoring/de/2017/index.shtml Zunker, N., Neumann, M. & Maaz, K. (2018). Angebot und Nachfrage bei der Einzelschulwahl. Der Einfluss von Schulmerkmalen und der Zusammensetzung der Schülerschaft auf die Nachfrage nach weiterführenden Schulen in Berlin. Zeitschrift für Pädagogik, 64(5), 586–611.