Abstract
In this paper, we examine whether implicit prices of neighbourhood design features in the housing market vary significantly across traditional, neo-traditional, and conventional suburban neighbourhood types. The set of neighbourhood design features we examine here include neighbourhood development density, street network connectivity, pedestrian access to transit and commercial stores, and land use mixture. Using data from Washington County, Oregon, we first use statistical procedures to identify distinct neighbourhood types. We then employ hedonic price analyses and a series of spatial Chow tests to obtain implicit prices of design attributes for houses in each neighbourhood type. We find that traditional design features such as higher street network connectivity and better pedestrian access to transit and commercial stores are valued more in the traditional and neo-traditional neighbourhoods, and that conventional neighbourhood features such as lower housing density and higher degree of homogeneous land uses are valued more in the suburban neighbourhoods.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
We then asked real estate agents the question whether the boundaries between traditional and neo-traditional, and suburban neighbourhoods are consistent with their perceptions of market areas. The consistency between the statistically determined boundaries and the perceived market areas by real estate agents confirms the reliability of the classification of neighbourhood types.
We removed non-arms-length transactions and prevent coding errors based on the ratio of sale price to assessed value. Transactions that have a sale price that is 60 percent greater than the assessed value or that is less than 60 percent of the assessed value are deleted from the data set. In addition, properties with lots greater than two acres, or age older than 80 years are excluded to maintain a homogeneous pool of transactions. Furthermore, we removed the transactions if their assessed value of the land is less than $1.00 per square foot or the assessed value of the improvements is less than $25.00 per square foot.
The cause of the segmentation, however, will not be identified and is beyond scope of this research.
Results from Box-Cox transformation indicate that a semi-log functional form is appropriate. This is also consistent with Song and Knaap’s (2003) hedonic price study using the same dataset.
To simplify the presentation, we do not include summary statistics by each neighbourhood type here. The information is available from the authors.
The asymptotic Wald statistics distributed as \( \chi^{2} \) with (M − 1)*K (M as the number of submarkets) degrees of freedom. In the most general form, the test for two submarkets can be expressed as a test on the null hypothesis \( H_{0} :g^{\prime} \beta = 0 \), where \( \beta^{\prime} = \left[ {\beta_{1}^{\prime} \beta_{2}^{\prime} } \right] \) is a stacked vector of all regression coefficients and \( g^{\prime} \) is a K by 2K matrix \( \left[ {I_{k} | - I_{k} } \right] \), with \( I_{k} \) as a K by K identity matrix. The corresponding Wald test is of the form: \( W = (g^{\prime} b)\{ {g^{\prime} \left[ {\text{var} (b)} \right]^{ - 1} g}\}^{ - 1} (g^{\prime} b) \) where b are the estimates for the regression coefficients and var(b) is the corresponding (asymptotic) variance matrix.
References
Adair, A. S., Berry, J. N., & MaGreal, W. S. (1996). Hedonic modeling, housing submarkets and residential valuation. Journal of Property Research, 13, 67–83.
Anselin, L. (1990). Spatial dependence and spatial structural instability in applied regression analysis. Journal of Regional Science, 30, 185–207.
Bajic, V. (1985). Housing-market segmentation and demand for housing attributes: Some empirical findings. AREUEA Journal, 13, 58–75.
Bourassa, S. C., Hamelink, F., Hoeesli, M., & MacGregor, B. D. (1999). Defining housing submarkets. Journal of Housing Economics, 8, 160–183.
Bourassa, S. C., Hoesli, M., & Peng, V. S. (2003). Do housing submarkets really matter? Journal of Housing Economics, 12, 12–18.
Chow, G. (1960). Tests of equality between sets of coefficients in row linear regressions. Econometrica, 28, 591–605.
Congress of New Urbanism. (2004). New urban projects on a neighborhood scale in the United States. Ithaca, NY: New Urban News.
Day, B. (2003). Submarket identification in property markets: A hedonic housing price model for Glasgow. Norwich: The Centre for Social and Economic Research on the Global Environment, School of Environmental Sciences, University of East Anglia. Working paper.
Duany, A., & Plater-Zyberk, E. (1992). The second coming of the American small town. Plan Canada, Winter, 6–13.
Goodman, A., & Thibodeau, T. G. (1998). Housing market segmentation. Journal of Housing Economics, 7, 121–143.
Grigsby, W., Baratz, M., Galster, G., & Maclennan, D. (1987). The dynamics of neighbourhood change and decline. Progress in Planning, 28(1), 1–76.
Harsman, B., & Quigley, J. M. (1995). The spatial segregation of ethnic and demographic groups: Comparative evidence from Stockholm and San Francisco. Journal of Urban Economics, 37, 1–16.
Leishman, C. (2001). House building and product differentiation: An hedonic price approach. Journal of Housing and the Built Environment, 16, 131–152.
Maclennan, D., & Tu, Y. (1996). Economic perspectives on the structure of local housing systems. Housing Studies, 11, 387–406.
Malizia, E., & Exline S. (2003). Consumer preferences for residential development alternatives. The Center for Urban and Regional Studies, the University of North Carolina at Chapel Hill. Working paper. Available Online: http://curs.unc.edu/wkgpapers.html.
Myers, D., Pitkin, J., & Park, J. (2002). Estimation of housing needs amidst population growth and change. Housing Policy Debate, 13(3), 567–596.
Palm, R. (1978). Spatial segmentation of the urban housing market. Economic Geography, 54, 210–221.
Rothenberg, J., Galster, G. C., Butler, R. V., & Pitkin, J. R. (1991). The maze of urban housing markets. Chicago: University of Chicago Press.
Schnare, A. B. (1980). Trends in residential segregation by race: 1969–1970. Journal of Urban Economics, 7, 293–301.
Song, Y. (2005). Smart growth and urban development pattern: A comparative study. International Regional Science Review, 28(2), 239–265.
Song, Y., & Knaap, G. J. (2003). New urbanism and housing values: A disaggregate assessment. Journal of Urban Economics, 54, 218–238.
Song, Y., & Knaap, G. J. (2004). Measuring urban form: Is Portland winning the war on sprawl? Journal of American Planning Association, 70(2), 210–225.
Song, Y., & Knaap, G. J. (2007). Quantitative classification of neighbourhoods: The neighbourhoods of new single-family homes in the Portland Metropolitan Area. Journal of Urban Design, 12(1), 1–24.
Author information
Authors and Affiliations
Corresponding author
Appendix: Classifying neighbourhood types
Appendix: Classifying neighbourhood types
We have identified four different urban neighbourhood types: urban core neighbourhoods, middle ring suburbs, outer ring suburbs, and neo-traditional greenfields. We combine two approaches—statistical and a priori procedures—to identify these neighbourhood types.
To define different neighbourhood types using statistical methods, we use GIS data from Portland Metro’s Regional Land Information System (RLIS). These data include (1) parcel based property (taxlot) data such as: year the structure was built, land use type, lot size, and floor space; (2) street network centrelines; (3) major transit stations and lines; (4) parks, open space and other recreational land uses; (5) tree canopy; (6) political and planning boundaries, such as county and city boundaries and urban growth boundaries; and (7) aerial photographs.
To classify neighbourhood types for all neighbourhoods, which are defined by 186 census blockgroups, we compute 16 urban form measures developed by Song and Knaap (2004) to characterize neighbourhood design. All the calculations were computed using ARCInfo and ArcView with data from Metro’s RLIS. Definitions of these measures are shown in Table A1. To measure street network design, we include the number of different types of street nodes, the size of street blocks, the lengths of cul-de-sacs, and the distance between points of access into the neighbourhood. To capture house characteristics, we include the size of lot and dwelling. To quantify neighbourhood density, we offer two measures: single-family residential (SFR) dwelling unit density and population density. To compute the level of mixed land uses in a neighbourhood, we keep track of the amount of commercial, public, multi-family residential (MFR), and light industrial land uses. To estimate pedestrian walkability, we include the percentage of single-family homes that are within one-quarter mile network distance (suggested by Duany and Plater-Zyberk 1992) of commercial uses and bus stops. Finally, to approximate the amount of open space in a neighbourhood, we include the acres of public parks and the area with tree canopy. Previous studies have proved that this set of variables measuring neighbourhood design can capture meaningful differences between different neighbourhoods (Song and Knaap 2003, 2004).
We then factor the above computed 16 measures of neighbourhood design into fewer dimensions, and then use cluster analysis to classify neighbourhood types statistically. We use factor analysis, a technique for data reduction, to help us understand the dimensional structure of our variables. Some of the 16 measures of neighbourhood design are highly correlated: for example, the distribution of cul-de-sacs is highly correlated with the distribution of large blocks. We therefore condense these variables into a smaller set of variables to remove the correlation in the data. Seven dimensions (factors) are extracted. The results are presented in Table A2. The variables in Table A2 are listed in the order of the size of their factor loadings sequentially for each factor. The extracted factors reproduce about 75% of the total variation. The last row of Table A2 presents the percent of the total variation accounted for by each factor. Principal component analysis for extraction and Varimax with Kaiser Normalization as rotation method in the factor analysis are used since this combination explained the most variation in the data. Varimax is used to maximize the variance of the squared loadings. Table A2 shows that the first factor reflects the dimension Street Connectivity. Factor loadings indicate that shorter cul-de-sacs and higher internal and external connectivity contribute to a smaller value of factor 1. The second factor includes Pedestrian Accessibility variables: higher pedestrian access to bus stops and smaller blocks contribute to a larger value of factor 2. The third factor includes Density variables: more multi-family residential uses within the neighbourhood and higher population and single-family residential dwelling unit density contribute to a larger value of factor 3. The fourth factor reflects the level of Land Use Mix: more industrial and more public land uses lead to a larger value of factor 4. The fifth factor relates to House Characteristics and shows that larger lots and houses contribute to a larger value of factor 5. The sixth factor reflects the level of Commercial Uses: higher pedestrian access to commercial stores and more commercial land uses contributes to a smaller value of factor 6. The last factor relates to the Natural Environment: more area of tree canopy and more parks contribute to a smaller value of factor 7.
Next, we perform cluster analysis, a method of combining observations into groups based on their similarity within a set of predetermined characteristics, to group neighbourhoods into neighbourhood types. K-means cluster analysis is performed on the seven factor scores derived from the previous step in such a way that each neighbourhood type is internally as similar as possible but externally dissimilar to other neighbourhood types. The best clustering solution, based on the interpretability of the results and associated cluster statistics, is found to be a six-cluster solution. The values of the cluster centroids, which indicate the performance of each neighbourhood type on each dimension (factor), are presented in Table A3. Six neighbourhood types emerge from the analysis and they are: rural neighbourhoods with dispersed homes, rural neighbourhoods with clustered properties, urban core neighbourhoods, middle ring suburbs, outer ring suburbs, and neo-traditional greenfields. We only focus on four urban neighbourhood types in the hedonic price regressions since we are only interested in examining housing product differentiation in an urban housing market.
We then verify the boundaries of urban core, middle ring suburbs, outer ring suburbs, and neo-traditional greenfields that resulted from the statistical method by inquiring among local real estate agents. The consistency between the statistically determined boundaries and the perceived market areas by real estate agents confirms the reliability of the classification of neighbourhood types.
Rights and permissions
About this article
Cite this article
Song, Y., Quercia, R.G. How are neighbourhood design features valued across different neighbourhood types?. J Hous and the Built Environ 23, 297–316 (2008). https://doi.org/10.1007/s10901-008-9122-0
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10901-008-9122-0