House price impacts of construction quality and level of maintenance on a regional housing market: Evidence from King County, Washington

ABSTRACT

Accurately estimating the monetary value of the construction quality and the level of maintenance of a house is a significant gap in house price research. Such estimates would be valuable for various actors in the housing market. Homebuyers would benefit from this information during the bidding process. Homeowners would benefit from it during renovation—particularly when financing their homes with a home equity line of credit—and when deciding on the need for renovation before selling their homes. Property value assessors would be able to provide better estimates of property values, and developers might make better decisions regarding the quality of new construction projects. Finally, lending institutions and public-sector housing agencies would benefit when providing loans or grants for housing rehabilitation and renovation.This research employs spatial econometric regression modeling and finds that construction quality and level of maintenance significantly affect house prices. For example, a medium-quality house sells for approximately 25% more than a low-quality house, and a well-maintained house sells for approximately 5% more than a house that is not well-maintained. The research also reports on the interaction between the quality and maintenance variables and how these variables vary with house size and the number of bedrooms.

KEYWORDS:

• House quality
• hedonic regression
• house maintenance
• house prices
• spatial econometrics

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. Since the house quality and the level of maintenance may be correlated with other neighborhood- and jurisdiction-level characteristics, a well-specified model is critical for accurate coefficient estimates.

2. The more traditional statistical techniques for outlier identification did not work for this data set perhaps because of errors in the County Assessors data. For example, for the existing housing dataset, the following data outlier identification method was attempted. First, the obviously erroneous data were removed, which included observations with a value of zero for any one of the following variables: sales price, number of bedrooms, number of bathrooms, the size of the house, and the size of the lot. Next, the regression model was run, and the standardized residual values were obtained. Thereafter, an observation was removed if the standardized residual value was less than ‒3 or more than 3. Still, several variables had very low or very high values, indicating data errors. For example, the minimums for lot size and house size were six square feet and one square foot, respectively, whereas the maximums were 6.3 million square feet and 12,560 square feet, respectively. Removing the observations with standardized residuals outside the ‒2 and 2 range did not solve the data error problems. For example, the minimum lot and house sizes were 24 square feet and one square foot, respectively, and the maximum lot and house sizes were 5.8 million square feet and 12,560 square feet, respectively.

3. The building grades are defined as follows (King County, 2016): Grades 1–3: Falls short of minimum building standards. Normally cabin or inferior structure. Grade 4: Generally older, low-quality construction. Does not meet code. Grade 5: Low construction costs and workmanship. Small, simple design. Grade 6: Lowest grade currently meeting building code. Low-quality materials and simple designs. Grade 7: Average grade of construction and design. Commonly seen in plats and older sub-divisions. Grade 8: Just above average in construction and design. Usually better materials in both the exterior and interior finish work. Grade 9: Better architectural design with extra interior and exterior design and quality. Grade 10: Homes of this quality generally have high-quality features. Finish work is better and more design quality is seen in the floor plans. Generally have a larger square footage. Grade 11: Custom design and higher quality finish work with added amenities of solid woods, bathroom fixtures and more luxurious options. Grade 12: Custom design and excellent builders. All materials are of the highest quality and all conveniences are present. Grade 13: Generally custom designed and built. Mansion level. Large amount of highest quality cabinet work, wood trim, marble, entry ways etc. conveniences are present.

4. Two chi-square tests, one for the jurisdiction in which the house is located and the quality of the house, and the other for the jurisdiction in which the house is located and the level of maintenance of the house, are statistically significant, indicating that quality and level of maintenance vary by jurisdiction.

5. For a model that includes an interaction variable, it is common for the interaction variable to be collinear with one or both of the interacted variables.

6. The positive sign for the property crime variable may be due to over-reporting of such crimes in neighborhoods with higher house prices than in neighborhoods with lower house prices. The negative sign for the variable measuring the number of bedrooms may indicate that, after controlling for the total living space of a house, an increase in the number of bedrooms reduces bedroom size. Moreover, the published literature disagrees on the effect of the number of bedrooms on house prices. Zietz, Zietz, and Sirmans (2007) note that of the 40 studies reviewed by Sirmans, MacPherson, and Zietz (2005), 21 studies find a positive impact of the number of bedrooms on house prices, nine studies find a negative impact, and the remaining 10 find no impact. A bivariate regression analysis shows an intuitive negative relationship between the age of the house and the house price. The Pearson correlation test also shows a negative correlation (correlation coefficient, ρ = −0.092). However, the relationship between the age of the house and its price turns positive upon inclusion of any other structural attribute of the house (for example, the house size, the lot size, the number of bedrooms, and the number of bathrooms). A bivariate analysis shows an intuitive positive relationship between house prices and accessibility by public transportation to non-retail employment. Furthermore, the relationship remains positive if we do not include the quarter dummies in the models. The change in sign upon the inclusion of quarter dummies may indicate a high sales volume of high-priced houses far from employment during certain years covered by the study. Finally, strong multicollinearity is not suspected for the variables discussed above because their VIF is less than five, the variables retain statistical significance when other variables are omitted, and the models’ adjusted R² values increase with inclusion of these variables, such as the quarter dummies.

7. For spatial lag models, the parameter estimates do not provide the magnitude of the impact of an independent variable on the dependent variable, except when the value of ρ is small—which is for all the three spatial lag models estimated in this paper. Specifically, the value of ρ is 0.01, 0.026, and 0.018 for Model 1.4b, Model 2.1, and Model 2.3, respectively. We thank an anonymous referee for this suggestion.

8. The 95% confidence intervals for these two coefficients overlap.

Additional information

Funding

This work was supported by the College of Social Sciences, San Jose State University [CoSS 2016–2017 Summer Salary Award].

Notes on contributors

Shishir Mathur

Shishir Mathur is Associate Dean of Research in the College of Social Sciences and a Professor of Urban and Regional Planning at San Jose State University. His research interests include urban and real estate economics, affordable housing policy, international development, infrastructure and development finance, growth management, transportation planning, and spatial econometrics.