Valuing school quality using boundary discontinuities

https://doi.org/10.1016/j.jue.2012.11.001Get rights and content

Abstract

Existing research shows that house prices respond to local school quality as measured by average test scores. However, higher test scores could signal higher academic value-added or higher ability, more sought-after intakes. In our research, we show that both school value-added and student prior achievement – linked to the background of children in schools – affect households’ demand for education. In order to identify these effects, we improve the boundary discontinuity regression methodology by matching identical properties across admissions authority boundaries; by allowing for boundary effects and spatial trends; by re-weighting our data towards transactions that are closest to district boundaries; by eliminating boundaries that coincide with major geographical features; and by submitting our estimates to a number of novel falsification tests. Our results survive this battery of tests and show that a one-standard deviation change in either school average value-added or prior achievement raises prices by around 3%.

Introduction

Good schooling is frequently upheld as decisive in life, but empirical evidence remains quite ambiguous when it comes to pinning down what makes a ‘good’ school and what people value in education. Parents making school choices seem well aware of their preferences and go to great lengths to secure places for their children at their preferred schools. However, social scientists have had mixed success in eliciting general conclusions about the nature of these preferences.

Researchers in education have regularly used survey responses to learn about preferences for schools (e.g. Coldron and Boulton, 1991, Flatley et al., 2001, Schneider and Buckley, 2002). The evidence from this field shows that parents rank academic outcomes highly among the reasons for choosing a school, but other factors play an important role, such as distance from home, school composition, safety and wellbeing. More recently, parents’ actual choices of schools and teachers have been used as an alternative way to uncover preferences for school attributes (e.g. Hastings et al., 2005, Jacob and Lefgren, 2007).

Apart from these examples, other research has looked for evidence of the value of schools in the capitalisation of their benefits into housing prices – i.e. using the hedonic valuation method. This wide-ranging international literature has shown that the demand for school quality is at least partly revealed in housing prices whenever school places are assigned to neighbouring homes. Gibbons and Machin, 2008, Black and Machin, 2010, Nguyen-Hoang and Yinger, 2011 and Machin (2011) provide summaries of recent evidence, all suggesting a consensus estimate of around 3–4% house price premium for one standard deviation increase in school average test scores.

One limitation of previous work is that – with a few exceptions – it is confined to showing that prices follow headline school performance as measured by school average test scores. However, better test scores could occur through improvements in enrolment quality or through greater pupil progress – potentially driven by teaching quality, school resources, peer effects and school effectiveness. One possibility is that parents pay for school ‘value-added’ that represents the expected academic gains for their children. A second possibility is that parents pay for good peers and favourable school composition – i.e. school inputs – irrespective of the contribution of these factors to their own child’s achievements.1 While the first perspective is interesting from a policy point of view because it puts a price on interventions that raise academic standards, the second one is relevant because of its implications for school segregation (e.g. Epple and Romano, 2000).

A handful of papers have taken steps to disentangle these two channels of influence. Brasington and Haurin’s (2006) results show that that school value-added and initial achievements both have positive effects on prices, although this point is somewhat lost in their conclusions. Kane et al. (2005) also consider value-added and average test scores as alternative indicators of school performance. However, they do not present specifications that include both indicators and do not aim to provide evidence on the importance of value-added. In contrast, Clapp et al. (2008) show that pupil ethnicity seems more important than test scores to home buyers around Connecticut schools, although the authors do not have access to data on pupils’ academic progress.

Other papers have looked at the importance of school expenditure relative to test score outputs. For example, Downes and Zabel (2002) find that test scores are capitalised into local house prices, whereas measures of school expenditures are not. Cellini et al. (2010) use referenda outcomes in California’s school finance system to suggest that house prices respond to the level of capital expenditure per pupil and that this cannot be fully explained by changes in test scores. Occasionally other school attributes have been considered. For example, Figlio and Lucas (2004) find that state-assigned school ratings have a transient effect on prices, over and above test scores, suggesting that householders draw additional information about achievement from these grades, or else value the ratings in their own right. Finally, Gibbons and Machin (2006) suggest that popularity in itself raises prices, given that over-capacity schools command an additional premium relative to under-capacity schools with equal performance.

Our paper moves this literature forward in a number of important ways. Our first contribution is to use a convincing strategy to show that house prices respond causally to school age-7 to age-11 test score gains (value-added), indicating that parents value school educational output. Our results suggest that parents also value the average age-7 test score component of this value-added measure, which we interpret as a marker for students’ background characteristics. We argue that this result arises from parental demand for good school composition, rather than demand for school quality in the early years, even if school composition is not a productive input in the educational production function. This interpretation is supported by further evidence showing that the price effects from age-7 achievements are completely explained by students’ background characteristics, especially their eligibility for free meals (a proxy for low family income).

Our second contribution is to further refine, improve and test the boundary discontinuity regression method, which is the ‘state-of-the-art’ approach used in this field to mitigate potential biases induced by neighbourhood unobservables. We present several innovations and refinements, which can be summarised as follows: (a) We combine matching methods with the regression-discontinuity design to allow for flexibility in the way in which housing observables affect price differentials across boundaries; (b) We incorporate in our models a variety of boundary fixed effects and spatial trends to account semi-parametrically for between-district unobserved heterogeneity and trends in amenities across boundaries; (c) We inverse-distance weight our regressions such that identification comes from variation at the admission zone boundaries where neighbourhood heterogeneity is minimised; this refines previous studies which used samples restricted to fixed buffer-zones close to boundaries (e.g. 1/4 mile); (e) We perform a number of falsification exercises and a compelling placebo test which uses the quality of autonomous state schools that do not admit on the basis of residential location, but administer the same standard tests as the mainstream schools that prioritise admission on place of residence.

A final advantage of our work is that we establish these findings using large scale administrative data for the whole of England, and not just for one city (e.g. Boston or San Francisco) as done by much of the previous research. The size and coverage of our data makes the above strategies feasible and the findings more representative.

To preview our results, we find that a one-standard deviation change in either age-7 to age-11 school average value-added or prior (age 7) achievement raises prices by around 3% for schools that prioritise students who live close by. Conversely, we show that there is no house price premium attached to properties close to high quality schools that do not prioritise local students. This finding – alongside other falsification exercises – demonstrates that our findings are causal and not spurious.2 Finally, various back-of-the envelope calculations show that the magnitude of this house price response to school quality is plausible as a parental investment decision given the expected return in terms of future earnings of their children.

The remainder of the paper has the following structure. Section 2 explains our methods. Section 3 discusses the context in which we apply our approach and the data setup. Section 4 presents our results and discussion. Finally, Section 5 concludes.

Section snippets

Methodological framework

Our empirical work uses a geographical boundary-based regression discontinuity design. This approach was initially popularised by the work of Black (1999), with several more recent examples (e.g. Bogart and Cromwell, 2000, Gibbons and Machin, 2003, Gibbons and Machin, 2006, Bayer and McMillan, 2005, Kane et al., 2005, Davidoff and Leigh, 2006, Fack and Grenet, 2010, Bayer et al., 2007, Ries and Somerville, 2010). Closely related studies investigate the effects of local taxes (Cushing, 1984,

National curriculum and assessment in England

Compulsory education in England is organised into five stages referred to as Key Stages. In the primary phase, pupils enter school at age 4–5 in the Foundation Stage then move onto Key Stage 1 (ks1), spanning ages 5–6 and 6–7. At age 7–8 pupils move to Key Stage 2, sometimes – but not usually – with a change of school.

Descriptive statistics

Table 1 presents some key descriptive statistics. The first two columns summarise the full data set, while the second two summarise the boundary sub-sample lying within 2500 m of an LA border (described above). The average price of sales in the full data is £182,730. In the boundary sub-sample the mean is about £13,000 or 7% higher. This is because administrative boundaries are more prevalent in and around towns and cities and there is a greater chance of finding matched pairs of sales in

Conclusion

A principal objective of this paper was to establish whether the well-documented response of housing prices to school-mean test scores represents a demand for the educational value-added output of schools, or demand for components of school quality that are desirable, but unlikely to raise a child’s achievements. Therefore, our first research aim was to go further than previous work in finding out if, why, and by how much people pay for homes near good schools. This is a crucial policy

Acknowledgments

We would like to thank Amy Challen and Anushri Bansal for excellent research assistance, and the Editor, two anonymous referees, Jaap Abbring, Victor Lavy, Erik Sørensen, and participants at numerous seminars and conferences for comments and suggestions. We are responsible for any errors or omissions.

References (52)

  • P. Nguyen-Hoang et al.

    The capitalization of school quality into house values: a review

    Journal of Housing Economics

    (2011)
  • Bayer, P., McMillan, R., 2005. Choice and Competition in Local Education Markets. NBER Working Paper...
  • P. Bayer et al.

    A unified framework for measuring preferences for schools and neighborhoods

    Journal of Political Economy

    (2007)
  • S. Berry et al.

    Automobile prices in market equilibrium

    Econometrica

    (1995)
  • S. Black

    Do better schools matter? Parental valuation of elementary education

    Quarterly Journal of Economics

    (1999)
  • Black, S., Machin, S., 2010. Housing valuations of school performance. In: Hanushek, E., Machin, S., Woessmann, L....
  • D. Brasington et al.

    Educational outcomes and house values: a test of the value added approach

    Journal of Regional Science

    (2006)
  • S. Cellini et al.

    The value of school facilities: evidence from a dynamic regression discontinuity design

    Quarterly Journal of Economics

    (2010)
  • D. Clark

    The performance and competitive effects of school autonomy

    Journal of Political Economy

    (2009)
  • J. Coldron et al.

    Happiness’ as a criterion of parents choice of school

    Journal of Education Policy

    (1991)
  • Z. Cooper et al.

    Does hospital competition save lives? Evidence from recent English NHS reforms

    Economic Journal

    (2011)
  • B. Cushing

    Capitalization of interjurisdictional fiscal differentials: an alternative approach

    Journal of Urban Economics

    (1984)
  • Davidoff, I., Leigh, A., 2006. How Much Do Public Schools Really Cost? Estimating the Relationship between House Prices...
  • J. DiNardo et al.

    Economic impacts of new unionization on private sector employers: 1984–2001

    Quarterly Journal of Economics

    (2004)
  • Duranton, G., Gobillon, L., Overman, H., 2006. Assessing the Effects of Local Taxation Using Micro Geographic Data....
  • Epple, D., Romano, R., 2000. Neighborhood Schools, Choice, and the Distribution of Educational Benefits. NBER Working...
  • Cited by (0)

    View full text