Estimation of Hedonic Single-Family House Price Function Considering Neighborhood Effect Variables

Chihiro ShimizuSustainability 2014 · DOI

TL;DR

This paper tackles the persistent problem of "omitted variable bias" in hedonic housing price models by explicitly incorporating neighborhood-level environmental and socio-economic factors. By comparing models with and without these "neighborhood effect variables," the study demonstrates that their exclusion leads to significant, non-negligible biases in price estimations, highlighting the importance of a more holistic approach to housing valuation.

What Problem Does This Paper Address?

Hedonic pricing models are a standard tool for understanding how the price of a good, like a house, is determined by its individual characteristics. However, a major challenge in these models is "omitted variable bias." This occurs when important factors influencing price are not included in the model, leading to inaccurate estimations of the coefficients for the variables that are included. In the context of housing, while building attributes and general location are often considered, local environmental conditions (like building density, land use patterns) and the characteristics of the people living in the neighborhood (like income levels) have been difficult to incorporate. These "neighborhood effects" can significantly influence property values, and their omission can distort the perceived impact of other, more easily measurable, factors. This paper aims to quantify the extent of this omitted variable bias in the Tokyo housing market by systematically adding these previously neglected neighborhood variables.

Methods

The study builds a series of hedonic models to estimate single-family house prices in Tokyo.

  • Model-1 (Base Model): This model includes standard property characteristics like ground area, floor space, age of building, proximity to transportation, and basic locational dummies (ward, railway line).
  • Model-2 (Neighborhood Effects - Basic): This model expands on Model-1 by incorporating more readily available neighborhood variables, such as floor area ratio and zoning dummies, which are often found in urban planning documents.
  • Model-3 (Neighborhood Effects - Comprehensive): This is the most advanced model, incorporating detailed neighborhood characteristics derived from GIS data (average building area, building density, land use conditions) and household characteristics (office worker ratio, average household building area) from census data. It also includes geographical coordinates (latitude and longitude) to capture any remaining spatial effects.

The core of the methodology lies in comparing the estimated coefficients of key variables across these models. By observing how these coefficients change as more neighborhood-specific variables are introduced, the researchers can identify and quantify the omitted variable bias.

Datasets

The primary dataset used is from the "Residential Information Weekly" magazine published by RECRUIT Co., which provides transaction data for single-family houses and land in the 23 wards of Tokyo between January and December 2000. This dataset includes information on property characteristics and prices.

Additional data sources include:

  • Geographic Information System (GIS) data: Specifically, the "Survey on Land Use in Tokyo" (fiscal 2001) provided detailed building data at a fine-grained (500m x 500m mesh) level.
  • National Census data: Used to derive household characteristics like the ratio of office workers and average household building area.

Key Insights and Ideas

  • Neighborhoods Matter, A Lot: The study provides strong empirical evidence that neighborhood-level environmental and socio-economic factors have a significant and measurable impact on house prices, beyond just the property's intrinsic characteristics.
  • Omitted Variable Bias is Real and Substantial: Failing to account for these neighborhood effects leads to distorted estimates for more traditional housing attributes. This means our understanding of what truly drives housing prices can be skewed if we don't consider the broader context.
  • Data Granularity is Key: The paper highlights the challenge and importance of obtaining fine-grained data for neighborhood analysis. While some neighborhood variables are easily accessible (like floor area ratio), others require more effort (GIS data, census data) but yield more robust results.
  • Beyond Physical Attributes: House prices are not solely determined by physical characteristics. Socio-economic factors and the quality of the local environment play a crucial role, aligning with established economic theories of hedonic pricing.

Key Results

  • Significant Coefficient Shifts: The study demonstrates that as more comprehensive neighborhood effect variables are introduced (moving from Model-1 to Model-3), the estimated coefficients for several key variables, such as "Travel Time to CBD," "Ground Area," and "Front Road Width," shift noticeably and often fall outside the confidence intervals of the simpler models.
  • Quantifiable Bias: Table 4 provides a direct comparison of estimated parameters across the models, showing that the inclusion of neighborhood variables leads to changes in coefficients that are statistically significant, indicating the presence of omitted variable bias. For instance, the "Travel Time to CBD" coefficient becomes less negative as neighborhood effects are accounted for, suggesting that proximity to amenities and better local environments can offset the disutility of distance from the city center.
  • Improved Model Fit: While not the primary focus, the adjusted R-squared values show a slight improvement from Model-1 (0.644) to Model-3 (0.662), indicating that the inclusion of neighborhood variables enhances the overall explanatory power of the hedonic model.

Limitations and Future Work

  • Data Availability Challenges: The authors acknowledge the difficulty in obtaining detailed local environmental information and individual buyer characteristics, which is a primary reason these variables have been underutilized.
  • Temporal Specificity: The study uses data from a specific year (2000) in Tokyo. While the findings are likely generalizable, the exact magnitude of effects might vary across different time periods and urban contexts.
  • Future Work: The paper suggests that the application of hedonic models for evaluating the effects of "Green buildings" is an active area of research in Japan, implying that further work could explore the integration of environmental quality metrics more directly into hedonic models. The study also implicitly points towards the need for more sophisticated data collection and integration methods to capture these complex neighborhood effects.

Summary updated

Abstract

In the formulation of hedonic models, in addition to locational factors and building structures which affect the house prices, the generation of the omitted variable bias is thought to occur in cases when local environmental variables and the individual characteristics of house buyers are not taken into consideration. However, since it is difficult to obtain local environmental information in a small neighborhood unit and to observe individual characteristics of house buyers, these variables have not been sufficiently considered in previous studies. We demonstrated that non-negligible levels of omitted variable bias are generated if these variables are not considered.
BibTeX
@inproceedings{shimizu2014estimation,
  author = {Chihiro Shimizu},
  title = {{Estimation of Hedonic Single-Family House Price Function Considering Neighborhood Effect Variables}},
  booktitle = {Sustainability},
  year = {2014},
  doi = {10.3390/su6052946},
}
Figure