Causal Effects of Changes in Ofsted School Ratings on Local Housing Sale and Rent Prices

Causal Effects of Changes in Ofsted School Ratings on Local Housing Sale and Rent Prices Xiang Ao University of Oxford Daniel Valdenegro University of Oxford Douglas Leasure University of Oxford Wenlan Zhang University of Oxford/University College London Melinda Mills University of Oxford/University of Groningen December 2025 Mapineq deliverable D5.1 Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 2 Mapineq – Mapping inequalities through the life course– is a three-year project (20222025) that studies the trends and drivers of intergenerational, educational, labour market, and health inequalities over the life course during the last decades. The research is run by a consortium of eight partners: University of Turku, University of Groningen, National Distance Education University, WZB Berlin Social Science Center, Stockholm University, Tallinn University, Population Europe, and University of Oxford Website: www.mapineq.eu The Mapineq project has received funding from the European Union’s Horizon Europe research and innovation programme under the grant agreement No. 101061645. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union, the European Research Executive Agency, or their affiliated institutions. Neither the European Union nor the granting authority can be held responsible for them. Acknowledgement: The content of the document, including opinions expressed and any remaining errors, is the responsibility of the authors. Publication information: This work is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) license. You are free to share and adapt the material if you include proper attribution (see suggested citation), indicate if changes were made, and do not use or adapt the material in any way that suggests the licensor endorses you or your use. You may not use the material for commercial purposes. Summary history Version Date Comments 1.0 23.12.2025 Manuscript for review 1.1 23.12.2025 Reviewed for submission Suggested citation: Ao, X., Valdenegro, D., Leasure, D., Zhang, W. and M.C. Mills. (2025). Causal Effects of Changes in Ofsted School Ratings on Local Housing Sale and Rent Prices. Turku: INVEST Research Flagship Centre / University of Turku. DOI: https://doi.org/10.5281/zenodo.18032475 Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 3 Executive summary This study examines whether sustained improvements or deterioration in Office for Standards in Education, Children’s Services and Skills (Ofsted) school ratings affect nearby housing prices in England. Using a large sample of UK housing listings from Zoopla and applying the Individual Synthetic Control (ISC) method, we analyse changes in school quality and their relationship to local rental and sales markets across thousands of primary and secondary schools. Key takeaways − Strong geospatial inequalities in schools and housing markets: there are notable regional disparities regarding the number of schools, students, students per school, and school ratings in England, especially between Southern England and other parts of England. − Housing markets show sharper regional divergence than schools: sale and rental prices are substantially higher in London and the South East, following a clear monocentric pattern, while other regions exhibit much lower and more heterogeneous prices. Local housing markets vary greatly in scale, especially in rentals. − School inspection ratings are highly persistent: over 98% of schools retain their rating across inspections. When changes occur, they are typically one-grade shifts. Rating changes are spatially dispersed but cluster in high-density areas, reflecting school concentration rather than regional dynamics. − School rating changes are rare shocks: the stability of ratings limits the number of treated observations and implies that inspection changes represent infrequent events rather than regular signals to households or markets. − Positive association between school ratings and prices: after controlling for regional differences in prices, housing prices are lower near lower-rated schools. A rating difference of one point (on a four point scale) was associated with 4.3% lower sale prices and 1.4% lower rental prices. − No observed causal price response to rating changes: using the Individual Synthetic Control method, we find no evidence that sustained upgrades or downgrades in school ratings cause nearby housing prices to change. Estimated effects remain close to zero across sales and rental markets, for both primary and secondary schools. − Interpretation and limitations: Housing and school choices reflect multiple factors beyond school ratings, including safety, space, affordability, and commute constraints. Consistent with prior research, we find that while school quality is correlated with housing prices, the causal impact of discrete rating changes is much smaller. Inspection outcomes likely convey limited new information, with school quality capitalised into prices gradually through reputation and neighbourhood sorting rather than immediate adjustment. Future work could incorporate broader neighbourhood and household factors to better capture these trade-offs. Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 4 Abbreviations DiD Difference-in-Differences Model IDW Inverse Distance Weighted ISC Individual Synthetic Control method ITL International Territorial Level LA Local Authority NUTS Nomenclature of Territorial Units for Statistics Ofsted The Office for Standards in Education UBDC Urban Big Data Centre Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 5 Content EXECUTIVE SUMMARY 3 ABBREVIATIONS 4 1. INTRODUCTION 7 2. INSTITUTIONAL CONTEXT AND DATA 9 2.1. PRIMARY EDUCATION IN ENGLAND 9 2.2. DATA 10 2.2.1. OFSTED INSPECTION RATINGS 10 2.2.2. ZOOPLA PROPERTY LISTING DATA 11 3. METHODOLOGY 11 3.1. MEASURES 11 3.1.1. STARTING POINT: PROPERTY LISTINGS AND PRICE-PER-BEDROOM 11 3.1.2. INFLATION ADJUSTMENT 12 3.1.3. DISTRICT-LEVEL CONTEXTUAL NORMALISATION 12 3.1.4. ASSIGNING LISTINGS TO SCHOOLS 12 3.1.5. SCHOOL RATINGS 13 3.1.6. TREATMENT IDENTIFICATION 13 3.2. ESTIMATION PROCESS 14 3.2.1. CONSTRUCTING A COMMON TIMELINE AROUND THE INTERVENTION 14 3.2.2. THE DONOR POOL 14 3.2.3. MATCHING ON PRE-TREATMENT DYNAMICS 14 3.2.4. CONSTRUCTING THE SYNTHETIC COUNTERFACTUAL 15 3.2.5. ASSESSING UNCERTAINTY 15 4. RESULTS 15 4.1. DESCRIPTIVE ANALYSIS 15 4.1.1. NUMBER OF SCHOOLS AND STUDENTS 16 4.1.2. DISTRIBUTION OF SCHOOLS WITH DIFFERENT RATINGS 17 4.1.3. SCHOOL RATINGS CHANGE 18 4.1.4. DISTRIBUTION OF SALE AND RENTAL PRICES 20 4.1.5. CORRELATION BETWEEN SCHOOL RATINGS AND HOUSING PRICES 22 4.1.6. TOTAL LISTINGS PER AREA (SALES) 23 4.1.7. TOTAL LISTINGS PER AREA (RENTS) 24 4.1.8. MOST DELINES AND MOST INCREASES (SALES) 25 4.1.9. MOST DECLINES AND MOST INCREASES (RENTS) 26 4.2. MAIN ISC ANALYSIS: IMPACT OF CHANGES IN SCHOOL QUALITY ON HOUSE PRICES 27 4.2.1. DOWNWARD TREATMENT: RENTS AND SALES 28 4.2.2. PRIMARY SCHOOLS DOWNWARD TREATMENT: RENTS AND SALES 29 4.2.3. SECONDARY SCHOOLS DOWNWARD TREATMENT: RENTS AND SALES 30 4.2.4. UPWARD TREATMENT: RENTS AND SALES 31 4.2.5. PRIMARY SCHOOLS UPWARD TREATMENT: RENTS AND SALES 32 4.2.6. SECONDARY SCHOOLS UPWARD TREATMENT: RENTS AND SALES 33 5. CONCLUSIONS AND DISCUSSION 33 Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 6 5.1. SUMMARY OF MAIN FINDINGS 33 5.2. DISCUSSION: LIMITATIONS AND FUTURE RESEARCH 35 6. APPENDIX 37 6.1. INDIVIDUAL SYNTHETIC CONTROL METHOD 37 6.1.1. INDIVIDUAL SYNTHETIC CONTROL 37 6.2. LONDON-ONLY SCHOOLS ICS ANALYSIS 38 6.3. NO-LONDON SCHOOLS ICS ANALYSIS 39 REFERENCES 40 Tables TABLE 1 TRANSITION PROBABILITIES OF SCHOOL INSPECTION RATINGS IN ENGLAND _____________________________ 18 TABLE 2 COEFFICIENTS OF SCHOOL RATING FOR SALE AND RENTAL PRICES ___________________________________ 23 Figures FIGURE 1 DISTRIBUTION OF SCHOOLS BY ITL LEVEL 2 (LEFT) AND LEVEL 3 (RIGHT) _____________________________ 16 FIGURE 2 DISTRIBUTION OF STUDENTS BY ITL LEVEL 2 (LEFT) AND LEVEL 3 (RIGHT) ____________________________ 16 FIGURE 3 DISTRIBUTION OF AVERAGE NUMBER OF STUDENTS PER SCHOOL BY ITL LEVEL 2 (LEFT) AND LEVEL 3 (RIGHT) ____ 16 FIGURE 4 PROPORTION OF SCHOOLS WITH A RATING ABOVE 2 BY ITL LEVEL 2 (LEFT) AND LEVEL 3 (RIGHT) _____________ 17 FIGURE 5 TRANSITION OF SCHOOL INSPECTION RATINGS AMONG SCHOOLS EXPERIENCING RATING CHANGES ENGLAND _____ 18 FIGURE 6 DISTRIBUTION OF DOWNGRADED SCHOOLS (LEFT) AND UPGRADED SCHOOLS (RIGHT) _____________________ 19 FIGURE 7 INTERPOLATED HOUSING SALES PRICE SURFACES (LEFT) AND RENTAL PRICES SURFACES (RIGHT) IN ENGLAND ____ 20 FIGURE 8 INTERPOLATED HOUSING SALES PRICE SURFACES (TOP) AND RENTAL PRICES SURFACES (BOTTOM) IN LONDON ___ 21 FIGURE 9 BOXPLOT OF SALE PRICES (LEFT) AND RENTAL PRICES (RIGHT) BY SCHOOL RATINGS ______________________ 22 FIGURE 10 TOTAL NUMBER OF SALES LISTINGS BY AREA _______________________________________________ 23 FIGURE 11 TOTAL NUMBER OF RENTAL LISTINGS BY AREA ______________________________________________ 24 FIGURE 12 COMPARISON OF THE TOP THREE AREAS WITH THE LARGEST DECLINES AND INCREASES IN SALE PRICES _______ 25 FIGURE 13 COMPARISON OF THE TOP THREE AREAS WITH THE LARGEST DECLINES AND INCREASES IN RENTAL PRICES _____ 26 FIGURE 14 AVERAGE TREATMENT EFFECTS ON HOUSING PRICES FOLLOWING A SUSTAINED DETERIORATION IN SCHOOL RATING 28 FIGURE 15 AVERAGE TREATMENT EFFECTS ON HOUSING PRICES FOLLOWING A SUSTAINED DETERIORATION IN PRIMARY SCHOOL RATING ____________________________________________________________________________ 29 FIGURE 16 AVERAGE TREATMENT EFFECTS ON HOUSING PRICES FOLLOWING A SUSTAINED DETERIORATION IN SECONDARY SCHOOL RATING ______________________________________________________________________ 30 FIGURE 17 AVERAGE TREATMENT EFFECTS ON HOUSING PRICES FOLLOWING A SUSTAINED IMPROVEMENTS IN SCHOOL RATING 31 FIGURE 18 AVERAGE TREATMENT EFFECTS ON HOUSING PRICES FOLLOWING A SUSTAINED IMPROVEMENTS IN PRIMARY SCHOOL RATING ____________________________________________________________________________ 32 FIGURE 19 AVERAGE TREATMENT EFFECTS ON HOUSING PRICES FOLLOWING A SUSTAINED IMPROVEMENTS IN SECONDARY SCHOOL RATING ______________________________________________________________________ 33 Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 7 Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices Abstract: A large literature documents strong associations between school quality and housing prices, often interpreted as evidence that families make residential choices based on access to better schools. This study examines whether sustained changes in school quality, measured by Ofsted school inspection ratings, lead to measurable adjustments in nearby housing sale and rental prices in England. We combine nationwide Ofsted inspection data with a large panel of residential sales and rental listings from commercial property company Zoopla covering 2016–2024. Housing prices are converted into an inflation-adjusted, district-normalised price-per-bedroom measure and aggregated monthly around each school. Descriptive and multilevel analyses reveal substantial spatial heterogeneity in schools and housing markets, alongside strong persistence in school ratings. Cross-sectionally, a one-grade lower school rating is associated with approximately 4.3% lower sale prices and 1.4% lower rents. To assess causality, we apply the Individual Synthetic Control method within a difference-in-differences framework that accommodates staggered treatment timing and school-specific counterfactuals. Across sales and rental markets, and for both primary and secondary schools, we find no detectable price response following sustained rating upgrades or downgrades. Estimated effects remain close to zero across all contexts. We propose future research to examine broader neighbourhood measures and factors. 1. Introduction There is a large body of evidence showing that the quality of the primary education one receives can significantly impact later academic achievements and life opportunities (Chetty et al., 2014; Feinstein, 2000; Hoekstra et al., 2018). Many parents are therefore willing to move to places with better access to high-quality schools, often paying a considerable premium for properties located close to those schools (Gibbons & Machin, 2003). In the long run, this enables wealthier families to reside near better schools while less affluent households are priced out of local housing markets, leading to the geographic sorting of households by socio-economic status. Consistent with Tiebout’s (1956) framework, empirical evidence suggests that housing prices capitalise the quality and quantity of local public goods, including education (Banzhaf & Walsh, 2008; Chay & Greenstone, 2005). Understanding how school quality relates to housing prices is therefore a crucial first step in identifying the mechanisms through which spatial educational inequalities are reproduced (Feng & Lu, 2013). A substantial body of research has examined the causal Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 8 impact of school quality on housing prices (e.g. Gibbons & Machin, 2003; Rosenthal, 2003; Machin, 2011; Gibbons et al., 2013; Jin et al., 2023). Despite these contributions, important empirical challenges remain. A first challenge concerns the measurement of school quality (Machin, 2011). Many studies rely on academic achievement levels as proxies for school quality (e.g. Figlio & Lucas, 2004; Dhar & Ross, 2012). However, student sorting strongly influences observed differences in achievement across schools (Munteanu, 2024), making it difficult to disentangle the effect of school quality from compositional effects. To address this, some studies employ school value-added measures capturing expected academic gains (Brasington & Haurin, 2006; Gibbons et al., 2013). While methodologically appealing, value-added metrics are typically researcher-constructed and not directly observed by households, raising questions about their relevance for parental decision-making (Imberman & Lovenheim, 2016). Moreover, achievement-based indicators capture only one dimension of school quality. This study instead draws on publicly available Ofsted inspection ratings, which provide a broader assessment of school performance encompassing leadership, management, and institutional effectiveness (Ofsted, 2024). A detailed description of these ratings is provided in Subsection 2.1. A second challenge relates to the empirical strategies used to estimate the effect of school quality on housing prices, many of which suffer from endogeneity concerns (Machin, 2011). Early work commonly estimated hedonic pricing models relating housing prices to school quality indicators using regression-based approaches, including parametric and non-parametric controls for unobserved factors (Black & Machin, 2011). However, it is rarely possible to observe and control for all neighbourhood characteristics that jointly influence school quality and housing prices, such as crime, public amenities, or socioeconomic composition, leading to omitted variable bias (Gibbons & Machin, 2003). Instrumental variable approaches have been used to address this issue, but often rely on weak or contestable instruments for school quality or neighbourhood attributes (Black & Machin, 2011). These limitations have motivated a growing literature exploiting quasi-experimental research designs that leverage plausibly exogenous variation in school quality to improve causal identification. Within this tradition, several studies employ regression discontinuity designs based on school catchment area boundaries (Gibbons et al., 2013). This approach exploits price differences between properties located near but on opposite sides of boundaries that grant access to different schools, while holding constant local characteristics that vary smoothly across space. Although boundary discontinuity designs reduce bias from unobserved area-level confounders, they remain vulnerable to household sorting around boundaries and to discontinuities in fiscal or policy variables (Bayer et al., 2007; Kane et al., 2006; Dhar & Ross, 2012; Chan et al., 2020). In addition, these designs often suffer from substantial reductions in sample size, limiting statistical power. Another strand of quasi-experimental research exploits temporal variation in school quality using Difference-in-Differences (DiD) designs. By comparing housing prices for the same or comparable properties before and after changes in school quality, DiD models difference out time-invariant unobserved heterogeneity and attribute price changes to Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 9 changes in school quality per se (Black & Machin, 2011). However, when treatment timing varies across units—as is the case with school inspections—standard two-period DiD estimators may yield ambiguous or biased estimates (Goodman-Bacon, 2021). This study advances the existing literature in two main ways. First, it addresses key data limitations by combining Ofsted inspection ratings with a large commercial property listings dataset from Zoopla, covering both housing sales and rental markets across England. While much of the literature focuses on housing sales alone, rental markets represent an important margin of adjustment for households facing financial constraints (Hussain, 2023), yet remain understudied (Chung, 2015). The nationwide coverage of both datasets further extends existing work that has largely focused on single cities or metropolitan areas (e.g. Gibbons & Machin, 2006; Fack & Grenet, 2010; Chung, 2015). Second, this study extends the DiD literature by employing the Individual Synthetic Control (ISC) method developed by Vagni and Breen (2021). Unlike canonical DiD estimators that assume a common treatment timing, ISC accommodates staggered treatment adoption by constructing unit-specific synthetic counterfactuals with strong pre-treatment fit. This allows estimation of treatment effects at the individual observation level as well as average treatment effects across a large sample of schools experiencing rating changes at different points in time. The remainder of the article is organised as follows. Section 2 describes the institutional context and data. Section 3 outlines the empirical strategy and model specification. Section 4 presents the main results, followed by conclusions in Section 5. 2. Institutional context and data 2.1. Primary education in England Education is compulsory for all children aged between 5 and 16 years in England. Pupils typically begin primary school at the age of 4 or 5 and transfer to secondary school at around 10 or 11 years old. Following Gibbons & Machin (2003), this study focuses on both primary and secondary schools for multiple reasons. First, primary school performance is a central focus of choice by parents seeking to enhance their children’s future life chances (Gibbons & Machin, 2003). Evidence suggests that educational attainments in the early years strongly influence later academic achievements and economic success (Chetty et al., 2014; Feinstein, 2000; Hoekstra et al., 2018). Moreover, high-quality primary education also enhances access to selective secondary schools in the British context (Gibbons & Machin, 2003). Given the high costs of moving (i.e. social and economic), parents are likely to make a single, long-term locational decision when their first child enters the education system (ibid.). Finally, secondary schools typically admit students over greater distance and from different Local Authorities (LAs) as opposed to primary schools (Gibbons et al., 2013). Therefore, the effects of secondary school quality on local housing prices are likely to be weaker and less observable. Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 16 4.1.1. Number of schools and students Figure 1 Distribution of schools by ITL level 2 (left) and level 3 (right) Figure 2 Distribution of students by ITL level 2 (left) and level 3 (right) Figure 3 Distribution of average number of students per school by ITL level 2 (left) and level 3 (right) Figure 1 - Figure 3 demonstrate the spatial distribution of total numbers of schools, enrolled students, and average numbers of students per school by ITL level 2 and 3 in England. Regarding the total numbers of schools, clusters with higher values appear in Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 17 parts of Midlands (Derbyshire and Nottinghamshire, and West Midlands) and selected regions in the South East England at ITL level 2. The patterns are more fragmented when observed at ITL level 3, with high-value regions scattered mostly across North East England, Midlands, and South East England. Durham tops with a total number of 134 primary community schools, followed by Birmingham (117) and North Yorkshire (114). Regarding the total number of students, similar clusters of high values can be observed in parts of Midlands and South East England at ITL 2 level, with West Midlands leading with over 145,000 students, followed by Greater Manchester with over 118,000 students, Derbyshire and Nottinghamshire with over 107,000 students, and a number of districts in London with around 110,000 students (East Inner London, West and Northwest Outer London and East and Northwest Outer London). As of ITL level 3, similar clusters stand out with Birmingham leading with over 48,000 students, followed by Leeds with over 33,000 students, districts in Hertfordshire with around 30,000 students, and districts in Greater Manchester Region and in Tyneside and Durham with around 27,000 students. The spatial patterns shift markedly when examining the average school size, measured as total numbers of students divided by total numbers of schools. At the ITL level 2, only a few districts exhibit notably high values. Specifically, districts in the Greater London Region are distinctive with exceptionally high student-to-school ratio. The East and North East Outer London area tops the list with 528 students per school, followed by West and North West Outer London (471), South Outer London (431), and East Inner London (410). Outside London, no district exceeds an average of 400 students per school, with the West Midlands representing the upper bound at 396. 4.1.2. Distribution of schools with different ratings Figure 4 Proportion of schools with a rating above 2 by ITL level 2 (left) and level 3 (right) Figure 4 shows the distribution of proportion of schools with ‘satisfying’ ratings in England, measured as the number of schools with a rating of 1 or 2 divided by the total number of schools. Here we can see significant concentration in South East England and in North West England. At the ITL level 2, districts in London and Essex show remarkably high values ranging from 91% to 96%. Cumbria up in the North also shows a high value of 93%. At the ITL level3, several districts achieve a proportion of 100%, including Camden, Westminster Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 18 and City of London, Bromley in London, Essex Thames Gateway, Warrington, Blackburn with Darwen, and Westmorland and Furness. 4.1.3. School ratings change Table 1 Transition probabilities of school inspection ratings in England Before the inspection 1 2 3 4 After the inspection 1 99.92% 0.03% 0.04% 0 2 0.07% 99.95% 1.46% 0.06% 3 0.01% 0.01% 98.47% 0.06% 4 0.01% 0.01% 0.03% 99.87% Figure 5 Transition of school inspection ratings among schools experiencing rating changes England Table 1 shows the transition probabilities of four school inspection ratings in England. The results reveal a strong tendency of schools to retain their original ratings. Specifically, for schools initially rated from 1 to 4, the probabilities of retaining the same ratings after an inspection are overwhelmingly high - 99.92%, 99.95%, 98.47% and 99.87%, respectively. Figure 5 presents a Sankey diagram showing the transition of school inspection ratings for schools experiencing rating changes in England. Each block represents schools with a particular original rating, and the labelled flows indicate the times when a school changes its rating following an inspection. Among all combinations involving a change in rating, the one that occurs for the most time is the transition from 3 to 2 (449), which translates to a probability of 1.46% for schools initially rated at 3. None of the remaining combinations occurs for more than 100 times, with the transition from 2 to 1 following with 92 times (or a probability of 0.03% for schools rated at 2), the transition from 2 to 4 with 46 times (or Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 19 a probability of 0.01%), the transition from 2 to 3 with 40 times (or a probability of 0.01%), the transition from 1 to 2 with 29 ties (or a probability of 0.07%). The other combinations are extremely rare, occurring no more than 10 times during the observation period. The distribution of the transition probabilities highlights a key limitation of the employed approach in this study. That is, the small number of schools that experience a change in rating constrains the sample size of the treatment group in the DiD analysis. Figure 6 Distribution of downgraded schools (left) and upgraded schools (right) Figure 6 shows the spatial distribution of schools that have been downgraded and upgraded in England. Most downgrades involve a one-rating decline as illustrated by light grey points. They appear broadly dispersed across England but also exhibit a certain degree of clustering in London, West Midlands, and Manchester – regions with comparatively high concentration of schools. The more severe two-rating downgrades are very rare and are sporadically scattered in northern and southern England. Most upgrades involve an increase of one or two ratings as illustrated by light blue points and medium blue points, respectively. They are broadly dispersed across England, and again, with slight clustering in London, West Midlands, and Manchester. The more remarkable upgrade of three ratings, shown as dark blue points, are very rare and exhibit no discernible spatial pattern. Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 20 4.1.4. Distribution of sale and rental prices Figure 7 Interpolated housing sales price surfaces (left) and rental prices surfaces (right) in England Figure 7 shows the spatial distribution of housing sales and rental prices in England. The values are interpolated using inverse distance weighting (IDW) to generate continuous surfaces that represent the underlying spatial variation in housing sales and rental prices. The figure reveals substantial spatial variation in housing prices, especially between major urban centres and other areas, and between South East England and the rest of the country. In terms of sales prices, values generally exceed £300,000 per bedroom across much of South East England, with London exhibiting the most pronounced concentration of high prices, reaching more than £600,000 per bedroom at its peak. Additional price spikes are evident in several major cities, including Liverpool, Manchester, Sheffield, Birmingham, Cambridge, Oxford, and Bristol. A broadly similar pattern emerges for rental prices. In most of South East England, average rents exceed £300 per bedroom per month, with some London districts recording values above £500. Outside London, several notable clusters of higher rental prices occur in and around major cities, including Cheshire East, Lincolnshire, Birmingham, Cambridge, Oxford, and Gloucestershire. Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 21 Figure 8 Interpolated housing sales price surfaces (top) and rental prices surfaces (bottom) in London Figure 8 zooms in to London and displays the spatial variation in housing prices across boroughs in London. Overall, the distribution exhibits a clear monocentric pattern, with prices rising sharply in areas closer to the city center. Particularly, Camden, Westminster and the City of London, Kensington and Chelsea and Hammersmith and Fulham are characterized by the highest housing prices in London, with sale prices commonly exceeding 500,000 pounds per bedroom and rental prices surpassing 350 pounds per bedroom per month. Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 22 4.1.5. Correlation between school ratings and housing prices Figure 9 Boxplot of sale prices (left) and rental prices (right) by school ratings Figure 9 2 shows very similar associations between school ratings and housing prices in terms of sales and rentals. First, it appears that housing prices are generally positively correlated with school ratings. Median sale and rental prices associated with schools rated 1 are noticeably higher than those associated with schools rated 2, 3, or 4, although the difference between ratings 3 and 4 is less pronounced. Second, while lower-quantile sale and rental prices vary little across school ratings, higher-rated schools tend to be associated with substantially higher upper-quantile prices. This pattern suggests that the premium for school quality is most evident in the upper segments of housing markets. Moreover, we acknowledge that listings in the same districts or regions might experience similar effects, leading to correlated errors that complicate the direct observation of correlation between housing prices and school ratings. Such clustering effects of listings might be associated with a broad range of attributes such as the abundance of local job opportunities and crime levels, which can be hard to quantify based on existing data. To address this issue and further explore the association, we employ multi-level modelling which accommodates such within-group correlations by allowing for group-specific random intercepts. We construct two-level random intercept models where observations of listings are nested within ITL level 3 regions. The model is written as: 𝑌 𝑖𝑗 = 𝛽0𝑗 + 𝛽 ∗ 𝑅𝑖𝑗 + 𝜀𝑖𝑗 𝛽0𝑗 = 𝛾0+ 𝑢0𝑗 Where 𝑌 𝑖𝑗 represents the log-transformed housing prices for listing 𝑖 in region 𝑗. 𝑅𝑖𝑗 represents the rating of the school that listing 𝑖 in region 𝑗 is assigned to, with estimated regression coefficients of 𝛽1. 𝜀𝑖𝑗 is the error term which is assumed to have a Gaussian distribution with a mean value of zero. The intercept 𝛽0𝑗 can be further decomposed into two terms, where 𝛾0 is the fixed mean and 𝑢0𝑗 is the random residual error term at regional level. We confirm that multi-level modelling is necessary via results of the null model. 2 Note that the prices employed here are the original prices rather then the deflated and standardized prices. This is because multilevel modelling already takes into account the regional differences in housing prices. Therefore, there is no need to double adjust the prices to offset the regional differences. Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 23 The results show that a one-level downgrade of the assigned school is associated with an average 4.29% (calculated as exp(-4.387e-02)-1) decrease of sale prices and 1.37% decrease of rental prices (see Table 1) Table 2 Coefficients of school rating for sale and rental prices Sale price Rental price School rating -4.387e-02*** -1.377e-02*** 4.1.6. Total Listings per Area (Sales) Figure 10 Total number of sales listings by area Figure 10 displays the total number of sales listings by area, ordered from highest to lowest. The distribution is highly uneven: a small number of areas account for very large volumes of listings—well above 200,000—while most areas have progressively fewer. Toward the right-hand side of the chart, many areas record fewer than 50,000 listings. This pattern reflects the substantial heterogeneity in local housing markets: some areas have deep, highly active sales markets, whereas others are relatively small or quieter. The long tail on the right-hand side suggests that any estimation strategy relying on local market activity must account for these differences in market size. Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 24 4.1.7. Total Listings per Area (Rents) Figure 11 Total number of rental listings by area Figure 11 shows the total number of rental listings by area, sorted from most to least active. The distribution is even more skewed than in the sales market: a handful of areas exceed half a million rental listings, while most others fall rapidly toward much lower volumes. The long right-hand tail indicates that rental market depth varies considerably across locations, with some areas exhibiting very large, liquid rental markets and many others operating at far smaller scales. This imbalance highlights the structural heterogeneity in London’s rental market and underscores the importance of normalising and contextualising local housing indicators before comparing outcomes across schools. Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 25 4.1.8. Most Delines and most Increases (Sales) Figure 12 Comparison of the top three areas with the largest declines and increases in sale prices Figure 12 compares the three areas with the largest real declines in per-bedroom sale prices and the three with the strongest increases over the 2016–2023 period. The contrast is stark. The fastest-decreasing areas—principally central London locations such as WC and EC—show a clear downward trend, with average real prices falling steadily year after year. South West London follows the same pattern, though starting from a lower baseline. By contrast, the increasing areas—Jersey, Torbay, and Hereford—show modest but consistent growth, with prices edging upwards over the period. The increases are gradual rather than dramatic, but they stand in clear opposition to the broad decline observed in central London. Overall, the figure illustrates a widening divergence in regional housing market performance: central London has experienced sustained real price falls, while a small set of peripheral or non-mainland areas have continued to see incremental gains. Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 32 4.2.5. Primary Schools Upward Treatment: Rents and Sales Figure 18 Average treatment effects on housing prices following a sustained improvements in primary school rating Figure 18 shows the results for primary schools experiencing a sustained improvement in rating. As with the aggregate analysis, both rental (295 treated schools) and sales (279 treated schools) markets remain essentially flat before and after treatment. The estimated effects fluctuate mildly around zero, and the confidence intervals consistently span zero throughout the window. There is no visible shift in either market following the improvement. In short, upgrades in primary school quality do not appear to generate a measurable price response in nearby rents or sales. Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 33 4.2.6. Secondary Schools Upward Treatment: Rents and Sales Figure 19 Average treatment effects on housing prices following a sustained improvements in secondary school rating Figure 19 reports the results for secondary schools following a sustained improvement in rating. As elsewhere, both rents (81 treated schools) and sales (75 treated schools) show no meaningful departure from zero in either the preor post-treatment periods. There is no indication of a systematic upward shift after the rating improves. 5. Conclusions and discussion 5.1. Summary of main findings A first conclusion from our descriptive analysis is that England’s school and housing landscapes are highly uneven and spatially clustered, but in different ways. The descriptive Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 34 analysis reveals strong geographic concentration in both the education system and housing markets, though the patterns do not fully overlap. Schools and students are densely clustered in parts of the Midlands, the North West, and London, with large urban areas such as Birmingham, Leeds, and Greater Manchester dominating in absolute numbers. London stands out less for the number of schools than for their size: average student–school ratios are markedly higher than elsewhere, reflecting dense populations and larger institutions. Housing markets display an even sharper spatial gradient. Sale and rental prices are substantially higher in South East England, especially London, where a clear monocentric pattern emerges, with prices rising steeply toward the city centre. Outside London, pockets of high prices appear around major cities and prosperous university towns, but the national picture is one of stark regional inequality. These patterns underscore the heterogeneity of local contexts in which schools operate and households make housing decisions. Here we also note that: − School size varies more sharply than school counts, with London uniquely characterised by very large schools. − Housing prices show stronger regional divergence than school distributions, dominated by London and the South East. − Local housing markets differ enormously in scale, particularly in rentals, necessitating methods that account for uneven market depth. A second main finding of our descriptive analysis is that school ratings are strongly persistent, and changes—while spatially dispersed—are rare events. Inspection ratings in England exhibit remarkable stability. Transition probabilities show that over 98% of schools retain their existing rating across inspections, with meaningful changes occurring in only a small minority of cases. Both upgrades and downgrades are predominantly one-rating movements, while larger shifts are extremely rare. Spatially, these changes are broadly dispersed but tend to cluster in areas with high school density, such as London and major metropolitan regions. This persistence limits the number of treated observations available for causal analysis and highlights that rating changes represent relatively uncommon shocks rather than frequent signals. Here we also note that: − Rating changes are dominated by stability rather than mobility. − Large rating swings are exceptionally rare. − Clustering of changes reflects school density rather than regional dynamics. A third and main conclusion is that despite cross-sectional correlations, sustained changes in school quality do not produce detectable causal effects on local housing prices. While descriptive and multilevel models indicate a positive association between higher school ratings and higher housing prices—particularly in the upper segments of the market—the Individual Synthetic Control analysis finds no evidence of causal effects following sustained rating changes. Across sales and rental markets, and for both primary and secondary schools, estimated treatment effects remain close to zero before and after both school rating upgrades and downgrades. Confidence intervals consistently span zero, suggesting that any response of housing prices to changes in school quality is either very Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 35 small or dominated by broader market forces. These findings imply that school quality is capitalised into housing prices slowly, indirectly, or through long-standing reputations rather than discrete inspection outcomes. Here we also note that: − Cross-sectional price premiums do not translate into shortor medium-term causal effects. − Results are consistent across market segments and school types. − Housing responses to school quality appear muted relative to wider regional and market dynamics. 5.2. Discussion: Limitations and Future Research The analysis has been constructed with care: extensive cleaning, conservative treatment definitions, and a synthetic control design aimed at producing credible comparisons. Across all specifications, we do not observe a measurable shift in local housing prices following sustained changes in school quality. The results are consistent, but they are not entirely conclusive. Several limitations are worth stating: First, our outcome data come from a single vendor (Zoopla). While Zoopla provides broad coverage, no single platform captures the entire housing market and after main market leader RightMove, Zoopla has been estimated to capture around 30% of the market (Best Agent, 2025). Listing practices differ across regions and agents, and some areas may be under-represented. If particular segments of the market react more strongly to school quality changes, such as off-platform transactions or properties handled by agents who list less frequently, those effects would not be visible here. Second, the data cleaning decisions that make the analysis tractable inevitably come at a cost. Trimming extreme values, discarding incomplete listings, deflating prices, and converting everything into z-scores all help stabilise the estimators, but they also remove a great deal of idiosyncratic variation. Some of that variation may contain real behavioural responses, especially small or short-run or highly localised changes, that simply disappear once we enforce strict comparability. Third, the synthetic control framework itself demands long and stable pre-treatment periods. Many schools do not have rating histories that meet these requirements, and many more operate in markets with too little data to support credible matching. These exclusions improve internal validity but they narrow the window through which we observe the world. Schools that experience faster or more irregular changes, where behavioural responses might plausibly be stronger, are absent from the analysis. Fourth, housing markets are driven by a host of forces that sit entirely outside this study: interest rates, tax changes, planning decisions, regeneration projects, and local labourmarket shocks. We do not model any of these explicitly. If these forces move strongly, they can easily obscure an effect of school quality even if one exists. Finally, the chosen metric (z-scored price per bedroom) captures one channel of adjustment but ignores others. Families may respond to school quality in ways that do not show up in this measure: the timing of moves, changes in preferred tenure, willingness to Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 36 stretch budgets, shifts in neighbourhood preferences, or even staying put and adjusting other aspects of household life. These behaviours never enter the dataset. In sum, the absence of detected effects should not be taken as evidence that no effects exist. It tells us only that within this dataset, using this design, and under these necessary constraints, we do not find a systematic price response. The underlying behavioural mechanisms may still be present; they may simply lie in parts of the housing market, or channels of adjustment, that this study cannot observe. A general substantive limitation is that we include only school ratings and house or rental factors as the decisive factors in our models. Parents tend to choose schools based on multiple factors, giving priority not only to school quality and housing affordability but also to safety and space. Our results, however, do concur with previous research that has also shown a correlation between school quality and house prices, precisely showing that the causal effect is often smaller than cross-sectional correlations suggest (Kane et al., 2006). Evidence from natural experiments suggest that clearer information about school quality can shift prices and search behaviour, particularly when the information is new or removes uncertainty (Hussain, 2023). Safety, neighbourhood stability and child-friendly amenities are also reported as top priorities for young families (Luo, 2025), which could be topics to explore in future research. Space - including extra bedrooms, private outdoor space - are also key as is affordability. Parents may trade school access and low-crime neighbourhoods against higher prices, meaning that households might sort along income lines into different schooling or amenity bundles. Theoretical work on school-quality uncertainty explains these trade-offs — better or less uncertain school quality steepens bid-rent gradients and is reflected in prices, while uncertainty or weak assignment rules flatten them (Turnbull et al., 2018) Families often accept longer commutes or pay premiums to secure good schools and safer neighbourhoods; space and garden access matter more for young children. Links and associations with these other factors can be explored in further detail, using many indicators from the Mapineq Link database (mapineq.org). Also, the observed crosssectional link between good schools and high prices does not by itself prove that discrete inspection changes will move prices quickly — capitalization can be gradual, reputationdriven, or constrained by other market forces. Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 37 6. Appendix 6.1. Individual Synthetic Control method The Synthetic Controls Method (SCM) is ideal for studying how changes in school ratings affect housing markets because it builds a credible “what if nothing had changed” comparison for treated areas. In the UK context, where school inspections and ratings shift at specific points in time, SCM can isolate the impact of those rating changes on property values by comparing each affected area to a weighted mix of similar, unaffected ones. Using Zoopla’s rent and sales listing data, this method makes it possible to estimate how prices respond to changes in school quality without assuming that trends were identical across all areas. It captures both observed and unobserved factors—like local demand, amenities, or neighbourhood trends—that evolve over time, making it far more robust than simple before-and-after or difference-in-differences approaches. The Synthetic Controls Method (SCM) builds a time-consistent counterfactual for a treated unit when no direct comparison exists. It was first proposed for natural experiments as an alternative to matching estimator (Abadie & Gardeazabal, 2003). Later, Abadie et al. (2010) expanded it for broader use — such as policy evaluations and large interventions — mainly focusing on aggregated units (like regions or countries). The SCM finds the optimal weights for each control unit j∈J as follows: 6.1.1. Individual Synthetic Control SCM was originally designed for large groups, but little work has adapted it to cases with multiple treated units l∈{1,2,…,L}. Vagni and Breen (2021) apply it at the micro level to estimate the average treatment effect (ATT) as: where τ^lt represents the estimated effect from Eq. 2 for each treated case. The weights for each j∈J are recalculated for each treated unit l. Abadie and L’Hour (2021) proposed a variation that adds a penalty term λ to favor control units more similar to each treated one: Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 38 This penalization ensures a unique solution when multiple treated units exist, emphasizing donors most similar to each treatment case. 6.2. London-only schools ICS Analysis We reran the Individual Synthetic Control analysis restricting the sample to schools located inside London boundaries. The London-only estimates—for both rents and sales, and for upand down-ward rating changes—show no systematic post-treatment deviations from their synthetic controls. Confidence bands overlap zero throughout the event window, so we find no detectable price response within London alone. Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 39 6.3. No-London schools ICS Analysis We also ran the same procedure on the complement set (schools outside London). The results mirror the full-sample findings: point estimates remain close to zero and bootstrap intervals cover zero across months. There is no consistent evidence of a housing-price response to sustained rating changes in the non-London sample. Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 40 References Abadie, A., Diamond, A., & Hainmueller, J. (2010). Synthetic control methods for comparative case studies: Estimating the effect of California’s tobacco control program. Journal of the American Statistical Association, 105(490), 493–505. Abadie, A., & Gardeazabal, J. (2003). The economic costs of conflict: A case study of the Basque Country. American Economic Review, 93(1), 113–132. Abadie, A., & L’hour, J. (2021). A penalized synthetic control estimator for disaggregated data. Journal of the American Statistical Association, 116(536), 1817–1834. Banzhaf, H. S., & Walsh, R. P. (2008). Do people vote with their feet? An empirical test of Tiebout’s mechanism. American Economic Review, 98(3), 843–863. Bayer, P., Ferreira, F., & McMillan, R. (2007). A Unified Framework for Measuring Preferences for Schools and Neighborhoods. Journal of Political Economy, 115(4), 588–638. https://doi.org/10.1086/522381 Best Agent. (2025). The 2025 Battle for Listings between Rightmove, Zoopla and OTM. https://bestagent.co.uk/the-2025-battle-for-listings-between-rightmove-zooplaand-otm/ Black, S. E., & Machin, S. (2011). Chapter 10—Housing Valuations of School Performance. In E. A. Hanushek, S. Machin, & L. Woessmann (Eds), Handbook of the Economics of Education (Vol. 3, pp. 485–519). Elsevier. https://www.sciencedirect.com/science/article/pii/B9780444534293000107 Brasington, D., & Haurin, D. R. (2006). Educational Outcomes and House Values: A Test of the value added Approach. Journal of Regional Science, 46(2), 245–268. https://doi.org/10.1111/j.0022-4146.2006.00440.x Chan, J., Fang, X., Wang, Z., Zai, X., & Zhang, Q. (2020). Valuing primary schools in urban China. Journal of Urban Economics, 115, 103183. https://doi.org/10.1016/j.jue.2019.103183 Chetty, R., Friedman, J. N., & Rockoff, J. E. (2014). Measuring the impacts of teachers II: Teacher value-added and student outcomes in adulthood. American Economic Review, 104(9), 2633–2679. Chung, I. H. (2015). School choice, housing prices, and residential sorting: Empirical evidence from inter-and intra-district choice. Regional Science and Urban Economics, 52, 39–49. https://doi.org/10.1016/j.regsciurbeco.2015.01.004 Report: Causal Effects of Changes in Ofsted School Ratings on Local Housing Prices 41 Dhar, P., & Ross, S. L. (2012). School district quality and property values: Examining differences along school district boundaries. Journal of Urban Economics, 71(1), 18–25. https://doi.org/10.1016/j.jue.2011.08.003 Fack, G., & Grenet, J. (2010). When do better schools raise housing prices? Evidence from Paris public and private schools. Journal of Public Economics, 94(1), 59–77. https://doi.org/10.1016/j.jpubeco.2009.10.009 Feinstein, L. (2000). The relative economic importance of academic, psychological and behavioural attributes developed on childhood (Issue 443). London School of Economics and Political Science. Centre for Economic …. Feng, H., & Lu, M. (2013). School quality and housing prices: Empirical evidence from a natural experiment in Shanghai, China. Journal of Housing Economics, 22(4), 291–307. https://doi.org/10.1016/j.jhe.2013.10.003 Figlio, D. N., & Lucas, M. E. (2004). What’s in a Grade? School Report Cards and the Housing Market. American Economic Review, 94(3), 591–604. https://doi.org/10.1257/0002828041464489 Gibbons, S., & Machin, S. (2003). Valuing English primary schools. Journal of Urban Economics, 53(2), 197–219. https://doi.org/10.1016/S0094-1190(02)00516-8 Gibbons, S., & Machin, S. (2006). Paying for Primary Schools: Admission Constraints, School Popularity or Congestion? The Economic Journal, 116(510), C77–C92. https://doi.org/10.1111/j.1468-0297.2006.01077.x Gibbons, S., Machin, S., & Silva, O. (2013). Valuing school quality using boundary discontinuities. Journal of Urban Economics, 75, 15–28. https://doi.org/10.1016/j.jue.2012.11.001 Goodman-Bacon, A. (2021). Difference-in-differences with variation in treatment timing. Journal of Econometrics, 225(2), 254–277. https://doi.org/10.1016/j.jeconom.2021.03.014 Hoekstra, M., Mouganie, P., & Wang, Y. (2018). Peer quality and the academic benefits to attending better schools. Journal of Labor Economics, 36(4), 841–884. Huang, B., He, X., Xu, L., & Zhu, Y. (2020). Elite school designation and housing pricesquasi-experimental evidence from Beijing, China✰. Journal of Housing Economics, 50, 101730. https://doi.org/10.1016/j.jhe.2020.101730