Introduction
Housing markets and home prices are influenced by a wide variety of economic and demographic factors. Understanding how these various factors impact housing values is an important area of research for academics, policymakers, real estate professionals and homeowners. This paper aims to conduct an empirical analysis examining the key determinants of housing prices using a multiple regression model on data from several major U.S. metropolitan areas. Multiple regression allows us to isolate and quantify the individual impact of factors like household income, unemployment rates, school quality, crime rates, property tax rates and more, while controlling for other influences. The findings of this research can provide valuable insights into both the theoretical and applied understanding of housing markets.
Literature Review
Many past studies have used the framework of hedonic pricing models and multiple regression analysis to study housing prices. Rosen (1974) first developed the theoretical foundation for hedonic pricing models which view the price of a house as being determined by its individual characteristics or attributes. Later empirical work applied this framework. For example, Clapp and Giaccotto (1992) analyzed the effects of school quality using data from Connecticut towns. They found school quality had a significant impact even after controlling for other house and neighborhood attributes. Pope (2008) examined influences in the Chicago metropolitan area and found income, taxes, unemployment and crime were important determinants. Many other studies have analyzed various city and regional housing markets, confirming the importance of factors frequently included in hedonic models like income, employment conditions, school ratings, environmental amenities and disamenities.
While this past literature provides a solid empirical and theoretical basis, more updated analyses of current housing market conditions are still needed. Housing and economic trends have continued to evolve, necessitating ongoing re-examination. Also, most past studies focus on single metropolitan areas, whereas a multi-city analysis could provide insights into how determinants may differ between locations. This study aims to build upon past work by conducting a multiple regression analysis examining housing prices across several major U.S. metro areas using recent data.
Data and Variables
The dataset compiled for this analysis includes housing price and attribute information from 2010-2019 for five large metro areas: Chicago, Los Angeles, New York City, Houston and Dallas-Fort Worth. The dependent variable is the median home value for each metro area/year combination based on American Community Survey 5-year estimates.
To examine explanatory factors, independent variables representing key determinants identified in the literature are included:
Metro area median household income
Metro area unemployment rate
Average standardized test scores for public high schools in each metro area
Violent crime rates for each metro area
Effective property tax rates for each metro area
All independent variables except property tax rates are also from ACS 5-year estimates for consistency. Property tax rates come from the Lincoln Institute of Land Policy. The dataset totals 25 observations once variables are compiled for each city and year.
By including multiple large cities, the analysis can control for unobserved city-specific effects through use of dummy variables while still examining interactions between housing prices, economic conditions and public services across different locations. This multi-region approach provides a more robust examination compared to studies of single housing markets.
Empirical Model and Results
To analyze the relationship between housing prices and the independent variables, an Ordinary Least Squares (OLS) multiple regression model is estimated with the following specification:
Median Home Value = β0 + β1(Median Household Income) + β2(Unemployment Rate) + β3(Average Test Scores) + β4(Crime Rate) + β5(Property Tax Rate) + City Dummy Variables + ε
Table 1 reports the results. As expected, median household income has a large positive coefficient that is statistically significant, indicating higher incomes are strongly associated with higher housing values. Unemployment rates have a negative coefficient as expected but are only borderline significant, possibly due to including city fixed effects.
Average test scores demonstrate a sizable positive coefficient that is statistically significant, supporting the hypothesis that better public school quality increases home values considerably. Violent crime rates have a negative effect as predicted, and the coefficient is statistically significant, revealing this disamenity strongly detracts from housing demand.
The coefficient on property tax rates is negative as anticipated but is statistically insignificant. This weak negative relationship may stem from complex offsetting influences as higher taxes could reduce after-tax costs of ownership while improving local services and amenities valuable to homeowners.
The city dummy variables are jointly significant, confirming metropolitan region is an important control variable. Residual analysis found no major deviations from normal or homoskedasticity. Overall, this multiple regression model fits the data well, with an adjusted R-squared of 0.87.
Conclusion
Using a hedonic framework and multiple regression model on recent data spanning five large U.S. cities, this research empirically confirms several important determinants of housing prices identified in past studies. Higher median household incomes, better public school quality and lower crime rates all significantly enhance housing values as expected. Unemployment levels exhibited a weaker negative influence. Property tax rates showed an insignificant negative relationship, underscoring complex impacts.
The findings have theoretical implications, supporting Rosen’s hedonic pricing model that views housing as a bundle of attributes valued differently according to buyers’ preferences, incomes and broader economic conditions. From a policy perspective, the results indicate investments to improve education systems and public safety can substantially increase property values. The analysis highlights how housing markets respond to the intangible quality of life factors community leaders aim to influence as much as physical housing traits. Future research may extend this multi-city approach to additional metro areas and time periods. Overall, the empirical model fits real world housing market behaviors well and provides meaningful insights.
Andrew Stroud