GIS and the Social Sciences

LEARNING OBJECTIVES

▪ Introduce the use of GIS for the analysis of the geographic, socio-economic and demographic determinants of educational attainment

▪ GIS for school catchment area analysis

▪ GIS for the analysis of educational performance in market systems

▪ GIS for the analysis of social and spatial inequalities in higher education attainment

In this chapter we look specifically at GIS for education planning, in particular at the following topics:

▪ GIS and its potential use in the analysis of geographical issues in higher education.

The chapter will primarily address GIS applications in the context of school education in urban and regional environments. A main issue to note is that around the world more and more education systems are becoming market oriented, in which parental choice and school performance become the most important drivers of change, not the decisions of the local planning agencies. As Stillwell and Langley (1999) comment, planning in the UK education sector, for example, developed far greater complexity following the introduction of the 1988 Education Reform Act and the creation of a state education system in which parents were given free choice in deciding the schools to which they would send their children. The requirement for secondary schools to publish achievement information in the form of examination results, coupled with a geography of non-fixed catchment boundaries, has resulted in a market system in which schools must effectively compete for pupils. That said, in some countries the state or local planning agencies do maintain a higher level of jurisdiction. This chapter will demonstrate the methods through which GIS and spatial modelling can be used effectively by planners to monitor the education environment and to advise on decision-making processes within whichever of the education systems are in operation.

Pupil forecasting is undertaken around the world to ensure that sufficient accommodation exists for the students within a particular area. If a new housing development is built, or current pupil levels increase/decrease due to demographics, the local education planning agency needs to be aware of the likely impacts for education provision as a whole. These factors have an obvious impact on school location and catchment area policy, hence the long tradition and importance of using spatial analytical techniques to predict changes in pupil numbers through population data, geodemographics and population forecasting techniques.

At the most basic level, schools are able to roll their attendance figures forward year on year in a straightforward manner. In a secondary school (in the UK this is typically children aged 11–16/18), this operation would enable a rough estimate of year group size to be followed through for up to seven years. Primary intake (in the UK, children aged 5–11 typically) can be simulated using birth rate data (if available) providing a guide for pupil numbers within a four-year prediction period. A slightly more committed approach to the analysis of current roll data might involve a study of student population trends and a comparison of patterns in neighbouring authorities and on a national scale. For example, while nationally pupil numbers might be increasing, regionally there is likely to be a very marked difference between the number of pupils of primary and secondary school age and the magnitude of change over time. In the UK, for example, by using sub-national population projections produced by the ONS, the Department for Education reported that between 2012 and 2017 the primary aged population was forecast to increase by between 9% and 15% depending on region, while the population of secondary school aged pupils was set to decline.

Cohort-survival techniques are widely used in population forecasting generally. Cohorts in this context refer to age groups: 0–4, 5–11, 12–18, etc. In relation to school planning, Simpson (1987) argues survival rates work in the following way:

Compared to the births in area X, how many children will appear five years later in school Y; compared to the five-year-olds now in school Y, how many six-year-olds will be there next year; and so on. The survival rate would not usually be exactly 100 per cent because some children move home and school, some children attend private schools and after 16 children leave school. The birth-to-school survival rate is often called an ‘arrival rate’, while the post-15+ survival rates are usually called ‘staying-on rates’ … It represents the net effect of migration, dropping-out, mortality and transfers between the state system and other schools: it estimates the balance of children who are lost and gained to the system at each age.

A greater level of understanding can be achieved by incorporating other influences on student numbers, in particular migration (where the primary factor is the development of new housing). The process uses the same current school attendee information rolled forward by year as the basic approach for calculating the number of pupils by postcode-based catchment area. This number is then supplemented by forecasting for new housing developments from the first ‘future allocation’ of land to housing – this gives an approximate number of houses likely to come on stream around five years later. Exact numbers and house types to be built are only known when final planning permission is granted, and are used to provide a last revision to earlier indicators. It is this number, and the nature of the new housing, which are the greatest influences on likely pupil numbers. For example, a new development of three-bed semi-detached housing is likely to generate far greater pupil increases than a block of sheltered accommodation for the elderly. Table 11.1 shows the general rules for predicting student numbers adopted by one UK local authority: Norfolk County Council (in East England). The overall pupil number forecasting approach of Norfolk County Council is illustrated in Figure 11.1.

It is interesting to note that new housing developments will not necessarily result in increased pupil numbers in local schools. This is especially the case in countries that have a market system in education (see below for further discussion). Gulson (2007) gives a very good example of this in relation to urban renewal in inner-city Sydney in the 2000s. Such renewal brought new houses and a greater population to many inner-city Sydney communities (a process of regeneration and gentrification), but as these new residents were more middle class there was a tendency not to use the local schools with poorer exam scores and reputations, but to send children across the city to the better performing schools.

As detailed above, school networks and service provision are inextricably linked to the location and dimension of the student population. Due to the investment in schools often over tens or hundreds of years, and the long-established relationships between populations and primary and secondary schools, it is not feasible to completely redesign a school network on the basis of an optimisation approach (unless we are considering an unlikely new town scenario). However, such an exercise might be appropriate in developing countries, especially as plans to expand access to education come to fruition. The Kenyan School Mapping Project has used GIS to create a database of school locations, resources, teachers and pupil numbers across Kenya (Mulaku and Nyadimo, 2011). This baseline information captured in a GIS provides decision makers with the ability to answer questions such as: where should new schools be placed? Where are teachers required? Are the school resources adequate? Is current provision adequate? Hite (2008) looks at the potential for using a GIS system for educational micro planning in Nepal and Uganda, arguing that, given the increasing use of tools to visually represent the world, there will be

a move in educational micro-planning efforts toward more visually-conceptualised and oriented planning at all levels of decentralisation. In this regard, it is reasonable to expect that GIS and other user-controlled visualisation techniques and solutions should become more prominent in education micro-planning efforts in the near future.

As noted above, any changes to future provision must be based on the current school network and are driven by population changes and school performance indicators. Potential interactions in this respect include expansion, downsizing, openings, closures, mergers and divisions, all of which will have implications for demand and supply functions in other education facilities or catchments. GIS could be applied in the approach to education planning by offering a better understanding of relationships between pupil populations and school locations through visualisation and analytical techniques. An example of this application might be best realised in the planning of a new school location or the selection of an existing location most suitable for extension or investment, a situation considered by Cropper (2003). This commentary reviews potential uses of GIS facilities, with suggestions including location of student populations, and analysis of available and affordable land through aerial photos and data on acreage and current appraised values. As the author rightly declares, all of these operations minimise searching and planning time while ensuring an extremely high rate of accuracy. In addition, Cropper (2003) proposes GIS for the investigation of:

The fundamental concept when assessing the need for new schools is that of capacity. This refers to the number of pupils that can be accommodated at individual schools and within catchments and the region as a whole. This information is vital as decisions regarding locations, networks and resource allocation must be based on a sound understanding of the relationship between demand and supply (i.e. pupil population and available accommodation). An interesting investigation has been undertaken across the city of Chicago in response to concerns over disparities between education supply and demand. Capacity is calculated by the Chicago school board as a formula accounting for the number of classrooms, courses scheduled and the maximum class size recommended by teachers’ union contracts. Comparing information on capacity vs. enrolment provides an index of resource utilisation, revealing schools that enrol greater than 100% of design capacity as overcrowded. Further investigation has shown that several of these schools are over attended by choice and tend to attract students from outside of their attendance boundaries. Results of this trend have become apparent in other sectors whose schools are now suffering under-utilisation and forced closure, which is a pressing concern for the city’s education management authorities who must ensure that all areas and pupil populations are adequately served by educational facilities. Fortunately the phenomenon has now been identified and recognised as a geographical issue, offering great potential for solutions by GIS application. Information on student population, capacity and location, as well as data on school closures, renovations and upgrades can be used to better manage attendance catchments and future planning of facilities and resource management. The first stage might be to map existing data and recent shifts in attendance patterns to improve understanding of the changes in pupil/parental preference or forced movements due to closures/openings. Scenario applications could also be used to interactively manipulate school networks, considering situations for more favourable provision through adjustment of school location and facility or catchment boundaries.

Of course population change can involve massive population decline, putting great pressure on the future of schools in certain locations. Lubienski and Dougherty (2009) use GIS to plot population loss in New Orleans after the devastating Hurricane Katrina in 2005. Figure 11.2 shows how many schools (of different types) are located in the areas of high population loss, putting great strain on the future viability of each one.

When considering school performance and catchment areas GIS can offer useful additional insights (although see the following section for a more detailed discussion of GIS and school performance in market systems). Peters and Hall (2004), for example, demonstrate the usefulness of GIS for representing and improving the quality of education provision in Peru. They provide an analysis of education quality and what they refer to as neighbourhood well-being. They first use a series of different indicators such as number of computers, student-to-teacher ratio, presence of a library, etc. to produce an education quality indicator for each school. They produced a similar indicator for neighbourhood quality based on 13 socio-economic inputs. Plotted together these indicators show clear spatial inequalities. When correlated, they demonstrate that a statistically significant relationship exists between the two. The analysis also allows for the identification of individual schools that urgently require more resources.

An interesting aspect of school catchment area analysis is planning transportation. In England, local education authorities are obliged to provide public transport to pupils from outside a two-to three-mile radius of the school (depending on pupil age). All pupils within this radius are expected to find their own way to school. GIS could be used here to:

▪ create spatial buffers at various distances from the school and to generate a list of pupil names and addresses for students entitled to transport services;

▪ work out the shortest walking route from one point to another (using network analysis described in Chapter 4);

▪ analyse the most efficient routes for buses to take in order to collect pupils who qualify for free transport.

A further advanced approach to GIS and spatial analysis application in the education planning field involves the use of gravity or spatial interaction models to analyse the impact of school openings and closures, and also in modelling optimum sizes, numbers and location of schools (see Chapter 8 for more details on these models in a retail context). York Local Education Authority in the UK has used such spatial models and GIS in consultation exercises with the public when revising an old system of catchments. In this project, four proposals were analysed for their impacts using GIS. The first stage in this process was to derive a set of mathematical functions, coefficients and constraints to effectively model the current situation of flows or trips from pupils’ homes to school facilities. This was best achieved by sub-dividing a study region into suitable geographical boundaries and considering flows to and from zone centroids. Following calibration and testing with observed data, such a model can be used to run scenarios or impact assessments to study how the network would respond to change, most likely in the form of adjustment to demand (e.g. new housing development) or supply (e.g. opening/closure of a school facility). Such a model could also react to changes in destination site attractiveness, for example, if a school were to improve performance and hence increase the attendance demand. The simplest form of attractiveness indicator might be school size or roll, which does act as a proxy for other school features such as facilities and funding. Another useful variable in this context would be a performance figure such as proportion of students achieving more than five A*–C grades at GCSE level. Stillwell and Langley (1999) also examined a spatial interaction approach to modelling the education system in Leeds. Having calibrated a model with existing flow data, they studied the effects of several scenarios:

▪ Removing 200 pupils from the population of one central city postal district revealed a significant impact on the city as a whole with declining student numbers in most facilities.

▪ A single large school closure in a suburban ward resulted in attendance increases across all city zones.

The authors recognise the unlikely situation of such large-scale events, but claim that the model evidence is useful to planners in providing confirmation that a number of small local shifts could lead to significant alterations to the system as a whole.

The remaining examples of geographical analysis in education planning refer to more advanced approaches of spatial modelling, scenarios and optimisation, and aim to broaden the understanding of the concerns and complexity of school planning issues. Pacione (1989) demonstrated the relative accessibility of secondary schools to the residents of 104 postcode sectors in Glasgow using a variation of the SIM (again see Chapter 8 for more details on these models). An index of accessibility extended the model by accounting for the relative access to cars and public transport, and the population of the neighbourhoods. Impacts of projected school closures on the resulting accessibility surface were evaluated by comparing provision prior to and after school closure. This research highlights the need to recognise the social as well as the spatial dimensions of accessibility, through the fact that two neighbours within the same locality can experience different levels of access depending on social grouping. Add to this the need for consideration of changing pupil populations and school capacities and the traditional element of school relationships that were mentioned in an earlier section, and the true complexity of the school planning strategy can be realised. Only with the aid of computational functions, data manipulation and mapping facilities can we realistically hope to account for all these potentially important factors in the management of educational planning.

Finally, work by Harland and Stillwell (2007) was concerned with analysis of the PLASC (Pupil Level Annual School Census) database, in particular to look at the movements of pupils between their homes and schools by ethnic group and over time. They note a number of challenges in adapting SIMs to account for the specific characteristics of the education sector. These include school capacity constraints, school admissions criteria and complex drivers of school choice decision making by pupils and parents. Nevertheless, they build a ‘Spatial Education Model’ which they demonstrate can be used to predict or forecast future school roll and capacity constraints (Harland and Stillwell, 2008).

As noted in the introduction, many countries now have a more market-oriented education system where schools compete for pupils based on parental choice. At the heart of such systems are school performance indicators. These can include discipline record, extra-curricular activities, truancy rates, ambience, etc., but most important are league tables of school success (in many countries the outcomes of exams at various key stages in academic development). League tables are intended to consolidate information on school performance and education quality, and a web-based GIS system is used by the UK Department for Education, for example, to display this information geographically (Department for Education, undated).

In many countries the introduction of market systems seems to have led to increasing social segregation as the higher income groups choose the better schools (or are chosen by the schools themselves if they have too many applications for limited places). Waslander and Thrupp (1995) were among the first to demonstrate this following the introduction of markets in New Zealand education in 1988 (also see an update in Thrupp, 2007). Since then a wealth of UK studies have shown the same (Ball et al., 1995; Ball, 2003, 2013; Taylor, 2009; Butler et al., 2007; for US studies see Bell, 2009). In spatial terms, it is not surprising perhaps that the schools that do the best in educational attainment terms are those in middle and higher income areas where students are more motivated and have had a more advantageous upbringing, including better parental support, better knowledge of the education system by those parents in general, access to more financial resources to attend better schools and perhaps more chance to attend pre-school educational establishments. In contrast, such market systems tend to produce a set of schools that struggle to attract high performing pupils, and are normally in working-class areas, which lack many of the attributes described above for the wealthier areas. These schools have less impressive exam success and often face a spiral of decline as resources are shifted to the better schools, as resources normally follow pupils. Clarke and Langley (1995) show an example of the catchment areas for two schools in Leeds, UK, following the full market system coming into operation in the UK in 1988.

Figure 11.3 shows that the school in North Leeds (top diagram), which has high attainment scores, is attractive to the higher income residents right across the north of the city. The bottom diagram shows the opposite is the case for a poor performing school in working-class inner-city Leeds – this has pupils drawn from only the immediate catchment, pupils perhaps trapped into having very few effective alternative choices. Parsons et al. (2000) also use GIS to plot inflows and outflows in an anonymised UK local authority area and show the vast number of non-local flows to school each day, especially to better performing schools.

A number of other studies have linked GIS to regression models (also see Chapters 5 and 8) in order to try to show a more robust statistical relationship between school attainment (success) levels and various socio-economic small-area indicators (i.e. Gordon and Monastiriotis, 2007). A difficulty with such analysis is that it is hard to estimate the nature of catchment area boundaries without actual pupil address data. Yet, it is important to try this:

Using these polygons he was able to link census data to school catchment areas and undertake regression analysis (see also Flowerdew and Pearce, 2001). While the use of global regression analysis is common, Fotheringham et al. (2001) applied their GWR technique to school performance data in north-east England. They showed that the power of individual key socio-economic variables used to ‘explain’ poorer exam success (low income, unemployment, no cars, etc.) could vary spatially, thus producing a more powerful and interesting local set of regression models.

A number of other studies have examined school performance by geodemographics. One of the more powerful is Butler et al. (2007), in East London, UK (though see Hamnett et al. (2007) and Webber and Butler (2007) for similar discussion). They used individual pupil addresses from the six top-and bottom-performing schools and profiled those students by the Mosaic geodemographic classification (which contains categorisations such as ‘symbols of success’ for higher income households and ‘blue-collar enterprise’ to represent working-class households). The results showed that generally school results were better for those schools closer to the wealthier suburban edge, as we might expect, but that a number of schools seemed to be perhaps punching above their weight in less well-off areas of East London (showing of course that school teachers, school policy, etc. can make a difference and improve performance in working-class environments).

In such a market environment, and given the arguments of how the market is playing out in many countries outlined above, Mayston (2003) argued that the assessment of performance needs to adequately account for the complexities of the real world and that the simple performance league tables published by countries such as the UK apply pressure to individual schools, as slipping within the national tables can have adverse effects on their ability to generate future funding (again, note that normally resources follow pupils in such quasi-market systems). GIS can provide a more complete review by adjusting data for these catchment area factors. The aim with such work is to try to show these poorer performing schools in a better light which may help their future attractiveness – so the exam scores could be repackaged for example as ‘actual performance against expected performance’. The higher the number of low-income residents in a catchment, then the lower expectations of exam success probably are – so if a school can show a better actual performance against expected outcome then this will offer a more positive light for marketing purposes. This exercise can be facilitated by incorporating more data into a model environment and might use, for example, deprivation indicators to give greater insight to real conditions. Eligibility for free school meals is another useful variable for assessing socio-economic structure (in the UK) in this context.

For other interesting attempts to model school catchment areas in market systems of education see Harris and Johnston (2008) and Singleton et al. (2011).

As mentioned at the beginning of the chapter, the issue of planning for higher education is associated more with regional and national student flows than with the local residential catchment area. This seems especially so in the UK where students are more unlikely to stay at home and attend higher education establishments in comparison to other European countries and the US and Australia. In terms of GIS analysis, many issues for higher education establishments are the same as for primary and secondary schools. For example, higher education establishments follow a similar approach for student forecasting to those adopted by schools and local planners. The University of Leeds, for example, recognises the importance of understanding the student population to support planning across departments and campuses. Estimates are based on current student counts and historical projection rates as well as forecast intakes for each individual course. These forecasts reflect market intelligence, growth targets, specific initiatives and promotional activities, and are used to predict fee incomes and financial margins as well as academic teaching space and accommodation needs. The university is determined to maintain its commitment to offering accommodation for all first-year students and projections are essential to meet these requirements.

In addition to roll forecasting, as with primary and secondary schools, many higher education establishments might be interested to use GIS to plot their catchment areas. In the UK, as noted above, these are likely to extend across the country. Figure 11.5 maps the home locations of students at the University of Manchester in 2005–2006.

Once the data are mapped as in Figure 11.5, the marketing team at the university can perhaps try to figure out why flows are high from certain feeder locations compared to others. This might relate to the spatial distribution of students with the highest educational attainment rates at secondary schools. It is not surprising therefore that there are a number of studies linking GIS with geodemographic profiling (Singleton, 2010; Brunsdon et al., 2011). They could also calculate indicators such as market shares (as the original totals or outflows will be known) and thus build decision support systems similar to those found in retailing (see Chapter 8).

Despite the challenges faced by education and planning departments, this chapter has illustrated a strong case for the implementation of GIS and spatial analysis in the education sector. Such technologies, when combined with the appropriate data, offer advantages to increase understanding of past trends and current patterns, as well as the potential for modelling of future scenarios. To this end, we have considered how spatial analysis, modelling and GIS can be applied to analyse pupil populations and forecasts, school networks and facilities to replicate interactions of demand and supply in primary, secondary and higher education.

This chapter is accompanied by Practical 7: Education provision and planning, giving you the opportunity to apply your spatial analysis skills to a series of data sets related to education provision. We guide you through the process of handling spatial data related to school catchment areas, school performance and access to educational opportunities, using data related to a specific administrative area within the UK. The data sets used contain more complex variables than some of our previous practicals, giving you the opportunity to develop your spatial and attribute data handling skills.

Ball, S. J. (2003) Class Strategies and the Education Market: The Middle Classes and Social Advantage, Routledge, London.

Ball, S. J., Bowe, R., & Gewirtz, S. (1995) Circuits of schooling: a sociological exploration of parental choice of school in social class contexts. The Sociological Review, 43(1), 52–78.

Bell, C. (2009) Geography in parental choice. American Journal of Education, 115(4), 493–521.

Brimicombe, A. (2000) Constructing and evaluating contextual indices using GIS: a case of primary school performance tables. Environment and Planning A, 32(11), 1909–1934.

Brunsdon, C., Longley, P., Singleton, A., & Ashby, D. (2011) Predicting participation in higher education: a comparative evaluation of the performance of geodemographic classifications. Journal of the Royal Statistical Society: Series A (Statistics in Society), 174(1), 17–30.

Butler, T., Hamnett, C., Ramsden, M., & Webber, R. (2007) The best, the worst and the average: secondary school choice and education performance in East London. Journal of Education Policy, 22(1), 7–29.

Clarke, G. P., & Langley, R. (1995) A review of the potential of GIS and spatial modelling for planning in the new education market. Environment and Planning C, 14, 301–323.

Flowerdew, R., & Pearce, J. (2001) Linking point and area data to model primary school performance indicators. Geographical and Environmental Modelling, 5(1), 23–41.

Fotheringham, A. S., Charlton, M. E., & Brunsdon, C. (2001) Spatial variations in school performance: a local analysis using geographically weighted regression. Geographical and Environmental Modelling, 5(1), 43–66.

Gordon, I., & Monastiriotis, V. (2007) Education, location, education: a spatial analysis of English secondary school public examination results. Urban Studies, 44(7), 1203–1228.

Gulson, K. N. (2007) Repositioning schooling in inner Sydney: urban renewal, an education market and the ‘absent presence’ of the ‘middle classes’. Urban Studies, 44(7), 1377–1391.

Hamnett, C., Ramsden, M., & Butler, T. (2007) Social background, ethnicity, school composition and educational attainment in East London. Urban Studies, 44(7), 1255–1280.

Harland, K., & Stillwell, J. (2008) Commuting to School: A New Spatial Interaction Modelling Framework for the Education Sector. NCRM Research Methods Festival 2008, 30 June– 3 July, St Catherine’s College, Oxford. (Unpublished.)

Harris, R., & Johnston, R. J. (2008) Primary schools, markets and choice: studying polarization and the core catchment areas of schools. Applied Spatial Analysis and Policy, 1(1), 59–84.

Higgs, G., Bellin, W., Farrell, S., & White, S. (1997) Educational attainment and social disadvantage: contextualising school league tables. Regional Studies, 31, 775–789.

Hite, S. (2008) School mapping and geographical information systems in education micro-planning. Paper presented at the Symposium on Directions in Educational Planning, UNESCO International Institute of Educational Planning (IIEP), Paris, 3–4 July.

Lubienski, C., & Dougherty, J. (2009) Mapping educational opportunity: spatial analysis and school choices. American Journal of Education, 115(4), 485–491.

Mayston, D. (2003) Measuring and managing educational performance. Journal of the Operational Research Society, 54(7), 679–691.

Mulaku, G., & Nyadimo, E. (2011) GIS in education planning: The Kenyan School Mapping Project. Survey Review, 43, 567–578.

Nayati, M. A. K. (2008) School bus routing and scheduling using GIS. Available from: www.diva-portal.org.

Pacione, M. (1989) Access to urban services: the case of secondary schools in Glasgow. Scottish Geographical Magazine, 105, 12–18.

Park, J., & Kim, B. I. (2010) The school bus routing problem: a review. European Journal of Operational Research, 202(2), 311–319.

233Parsons, E., Chalkley, B., & Jones, A. (2000) School catchments and pupil movements: a case study in parental choice. Educational Studies, 26(1), 33–48.

Pearce, J. (2000) Techniques for defining school catchment areas for comparison with census data. Computers, Environment and Urban Systems, 24(4), 283–303.

Peters, P. A., & Hall, G. B. (2004) Evaluation of education quality and neighbourhood well-being: a case study of Independencia, Peru. International Journal of Educational Development, 24, 85–102.

Simpson, S. (1987) School roll forecasting methods: a review. Research Papers in Education, 2, 63–77.

Simpson, S. (1988) The use of school roll forecasts in LEA administration: the allocation of resources to schools. Cambridge Journal of Education, 18(1), 89–98.

Singleton, A. D. (2010) The geodemographics of educational progression and their implications for widening participation in higher education. Environment and Planning A, 42(11), 2560–2580.

Singleton, A. D., Longley, P. A., Allen, R., & O’Brien, O. (2011) Estimating secondary school catchment areas and the spatial equity of access. Computers, Environment and Urban Systems, 35(3), 241–249.

Singleton, A. D., Wilson, A. G., & O’Brien, O. (2012) Geodemographics and spatial interaction: an integrated model for higher education. Journal of Geographical Systems, 14(2), 223–241.

Stillwell, J. C. H., & Langley, R. (1999) Information and planning in the education sector, in J. C. H. Stillwell, S. Geertman, & S. Openshaw (eds) Geographical Information and Planning, Springer, Heidelberg, 316–333.

Taylor, C. (2009) Choice, competition, and segregation in a United Kingdom urban education market. American Journal of Education, 115(4), 549–568.

Thrupp, M. (2007) School admissions and the segregation of school intakes in New Zealand cities. Urban Studies, 44(7), 1393–1404.

Waslander, S., & Thrupp, M. (1995) Choice, competition and segregation: an empirical analysis of a New Zealand secondary school market, 1990–93. Journal of Education Policy, 10(1), 1–26.

Webber, R., & Butler, T. (2007) Classifying pupils by where they live: how well does this predict variations in their GCSE results? Urban Studies, 44(7), 1229–1253.