The proximity of high-quality schools significantly influences housing choices, leading to high price premiums in desirable school districts globally. This phenomenon, known as education capitalization, exacerbates socioeconomic inequalities, particularly for lower-income families. While existing literature primarily focuses on the linear relationship between housing sales prices and school academic performance, this study investigates the impact of entrance opportunity—a crucial aspect of school quality—on both housing sales and rental prices in the context of Shenzhen, China's new point-based school enrollment policy. This policy prioritizes homeowners over renters in school admissions, creating a disparity in access to quality education. The study hypothesizes that the relative importance of entrance opportunity and academic performance in determining housing values will differ between sales and rental markets, with tenants potentially prioritizing entrance opportunity more than academic performance compared to homeowners.

Existing literature on education capitalization largely centers on the Tiebout hypothesis, suggesting residents sort themselves into neighborhoods based on preferences and affordability for local public goods, including education. Studies consistently demonstrate the capitalization of school quality into housing sales prices, but often overlook rental markets and the multidimensionality of school quality. Most research relies on standardized test scores as the primary measure of school quality, neglecting other important factors such as parental perception, school reputation, and admission policies. The lack of comparative analysis between housing sales and rental markets is particularly noticeable in the Chinese context, where the recent point-based school enrollment policy differentiates access based on housing tenure, impacting both rental and sales prices. Existing studies on this disparity in China have not conducted a comprehensive comparison of capitalization effects on both markets, and have not fully accounted for the multidimensional nature of school quality.

This study employs a comprehensive methodology involving data collection, multidimensional school quality measurement, machine learning model development, and spatial analysis. Data were collected from the Shenzhen Education Bureau, school websites, social media, and the China Real Estate Index System (CREIS). School quality was measured across four dimensions: entrance opportunity (measured by the school's acceptance cut-off point), academic performance (percentage of students admitted to top high schools), parental impression (unofficial school rankings), and policy signal (official school grades and name signals). Three machine learning algorithms—random forest regression (RFR), gradient boosting regression trees (GBRT), and extreme gradient boosting (XGBoost)—were employed to model the relationship between housing prices (sales and rental) and multidimensional school quality, controlling for other housing attributes (structural, locational, and neighborhood characteristics). The XGBoost model, exhibiting superior performance based on R², RMSE, and MAE, was selected for further analysis. Relative importance analysis determined the influence of each school indicator on housing prices, while partial dependence plots visualized nonlinear relationships. Spatial analysis using bivariate local Moran's I identified spatial clusters of school quality and housing prices.

The XGBoost model outperformed linear regression and other MLAs in predicting both housing sales and rental prices. Relative importance analysis revealed that academic performance was the most influential factor for housing sales prices across all price segments, particularly the percentage of secondary school students admitted to top high schools. In contrast, entrance opportunity (the acceptance cut-off point) was the most significant factor for housing rental prices, especially in the lower-price segment. This highlights the different priorities of homeowners and renters, reflecting the stratified nature of school-based housing choices influenced by the point-based enrollment policy. Nonlinear relationships were found between key school indicators and housing prices, revealing threshold and saturation effects. Homebuyers were willing to pay higher premiums for increases in academic performance only above a certain threshold, while the impact of entrance opportunity on rental prices was consistently positive across the entire range. Spatial analysis indicated significant spatial clustering of high housing sales prices with high academic performance in certain districts, and a less pronounced but still present clustering of high rental prices with high entrance thresholds. Further analysis showed that schools with the name "education group" had relatively higher price premiums.

The findings address the research questions by demonstrating the differing capitalization effects of school quality on housing sales and rental prices, influenced by the point-based school enrollment policy. The prioritization of academic performance by homeowners reflects a focus on educational outcomes, while the emphasis on entrance opportunity by renters highlights the importance of access to quality education under the policy's limitations. The nonlinear relationships indicate that improvements in school quality only yield significant price premiums within certain ranges, suggesting potential policy interventions to target specific quality improvements. The significant spatial clustering underlines the uneven distribution of school quality and its impact on housing markets, underscoring the need for equitable urban planning and resource allocation. The results show how distinct housing markets prioritize different school quality indicators reflecting the complex interplay of policy, market forces, and parental choices.

This study offers a nuanced understanding of the education capitalization effect, highlighting the differentiated importance of entrance opportunity and academic performance for homeowners and renters. The nonlinear relationships and spatial patterns reveal the complexities of school choice and its impact on housing markets. Policy implications include targeted improvements to school quality based on the identified threshold and saturation effects, delinking housing ownership from school access, and enhancing vocational education to reduce the pressure of the highly competitive senior high school entrance examination. The study's methodological framework can be applied to other urban contexts globally to understand similar interactions between school policies, housing markets, and socioeconomic inequalities.

While this study provides valuable insights, several limitations should be considered. First, while the study explores the effects of policy-related school indicators, further research using quasi-natural experiments could better isolate the precise impacts of the school admission policy and education groups on housing prices. Second, further research is needed to fully understand the complex interplay between various school quality indicators and their combined effects on housing prices. Finally, the study is focused on Shenzhen, China, and the generalizability to other contexts may need further validation.