INTRODUCTION Geographic Variations in Resale Housing Values Within a Metropolitan Area: An Example from Suburban Phoenix, Arizona Diane Whalley and William J. Lowell-Britt The average cost of single family resale housing displays geographic variation at many scales. Regional variations in the cost of housing are used in the compilation of cost-of-living indices, (Follain and Malpezzi, 1980) and have been well documented (inter alia Butler, 1980; Ozanne and Thibodeau, 1983; Blackley et ai, 1986). Geographic patterns of housing values within a metropolitan area are not as sensitive to the conditions affecting the national market, but are the result of a series of local influences (Bednarz and Wilson, 1982). It is the purpose of this paper to identify the nature of local variations in the price of housing within a metropolitan area. Data are taken from two suburbs of Phoenix, Arizona, and the study area has been divided up into seven identifiable submarkets. A variety of housing characteristics, neighborhood characteristics and different levels of public service provision are combined to produce a distinct spatial pattern of housing values within a relatively confined geographical area. THE STUDY AREA AND DATA Assistant Professor Department of Geography The University of Georgia Athens, GA 30602 Mountain West Research, Inc. Tempe, AZ 85281 Data were collected from a sample of 227 houses listed in the Arizona Multiple Listing Directory for residential Southeast Phoenix, during the first week of January, 1985. The houses sampled were located in the cities of Tempe and Mesa (Figure 1). Tempe and Mesa were selected for analysis because : (1) they are two of the largest (and therefore most diverse) adjacent cities within the Phoenix metropolitan area ; and (2) the property market in each is influenced by its own peculiar characteristics. Arizona State University is located in Tempe, and the exceptionally high demand for rental properties influences the market. In Mesa, the city does not levy its own property tax, an unusual occurrence in North American metropolitan areas. These two cities therefore present the opportunity for a great deal of local variation within the housing market over a relatively restricted geographical area. 31
7. 1. SUN CITY 2. PEORIA 3. GLENDALE A. PARADISE VALLEY 5. SCOTTSDALE 6. TEMPE 7. MESA 8. CHANDLER 9. GILBERT 9. o 1 I " o 2 2 3 I I, 4 4 5 MI " I 6 8 KM STUDY AREA SUBMARKETS MESA 5. 6. 4. 7. 1 2 3 MI I I I I I \ 2 3 4 5 KM Figure 1. The study area and submarkets The study area was divided into a series of seven submarkets, based on housing and neighborhood characteristics as shown in Figure 1. The three south Tempe suburbs (submarkets 3, 4, and 7) were later recombined to preserve a larger sample size, because they did not show much variation from one another. Further, houses priced at over $120,000 were dropped from the analysis because preliminary testing showed that these houses respond to a distinctly different set of variables, and must be considered to represent a submarket of their own, regardless of location. This left 186 houses in the sample. 32
METHODOLOGY The method employed to identify the local influences on housing values was to construct an 'nedonk pr"lce equation, and to solve it using multiple regression. Hedonic equations are based on the assumption that the value of housing, V, is a function of a series of housing attributes, A. such that V = f(a, + a2 + a3 +... an) (Griliches, 1971) The precise combination and effects of the attributes will vary over time and space (Goodman, 1978; Whalley, 1985). The first step in an analysis of local variation in housing values is therefore to identify which attributes are important within each market. Having identified the relevant attributes, the relative importance of each can be determined through the use of multiple regression. Using the estimated market value as the dependent variable, and the important attributes as independent variables, the regression co-efficients will reveal the marginal contribution (in dollars) to housing value, of each of the attributes. IDENTIFICATION OF VARIABLES The Arizona Multiple Listing Directories contain a great deal of information on the internal structural characteristics of resale housing. This can be combined with census data on neighborhood characteristics, and municipal data on schools and locally provided public services. A preliminary regression analysis was conducted to identify which combinations of variables had the greatest impact on property values. These were then used in the final regression equation. COMPONENTS OF HOUSE PRICE Analysis reveals that there are two components to house price: a fixed component or base price, which represents the value of the actual location of the house, and a variable component which changes as a response to the characteristics of the house and neighborhood. Table 1 shows the base prices and Table 2 shows the variable components, derived for this study. The preliminary regression analyses were used to examine the variable characteristics of a variety of different housing, and to ide [\ti~ which chafacte fistics of the house and the neighborhood had the greatest impact on estimated value. Throughout the study area the size of the house was the single most important determinant of value. On average, every additional square foot of floor space increased the house price by $33, although this figure varied from a high of $44 per square foot in south Mesa to a low of $31 per square foot in northwest Tempe (see Table 2). Over 75% of the variation in market prices can be accounted for by differences in square footage. Other factors which turned out to be important were : The Age of the House In most of the neighborhoods studied, the houses lost value with age. On average, for every year a house aged it declined in value by $512, though there is considerable variation amongst the neighborhoods. Older houses are often in worse physical condition than newer ones. Further, there is a wide range of new housing available in metropolitan Phoenix, and this tends to depress sales of older resale homes. Property Taxes Each $100 of taxation reduced property values by an average of $1,000. However, the property tax represents the extent of payment for local public goods and services. Some properties are overtaxed, given the level of services provided, and others are undertaxed. By examining the tax burden on each property it was possible to estimate that, on average, every $100 of excess taxation per annum, would reduce the property value by $300.' The Presence of a Renter in the Property In each neighborhood, the presence of tenants in a property tended to drive the sales price down. On average, this represented a reduc- 33
No~thv. e s t Te:npe Northeast Tempe South Tempe Mesa South Mesa TABLE 1 Base Prices of Resale Homes, by Location (st:br::arket 1) (submarket 2) (submarkets 3, 4 & 7) (submarket 5) (submarket 6) $24,396 31,746 39,548 31,557 23,389 tion of between $1,000 and $2,000, the exceptions being in South Mesa where the number of houses with renters was insufficient to accurately estimate their effect, and in North Tempe, which contains the area around Arizona State University. In this latter neighborhood the downward effect of tenants on property values was much greater. Since houses which are placed on the rental market are not always in the same condition as owner-occupied housing, this effect may be TABLE 2 Variable Components of House Price Square Age (in Location Additional Presence footage years) by Park $100 in tax of tenant ~:.~'. Tc.iilpe; +31-56 + 3,845 378-1,<;<;':' N.E. Tempe +33-522 895* +1,493-7,328 S. Tempe +37-554 + 606-1,661-1,093 Hesa +33-350 +10,503* - 1,410-1,093 S. t1esa +44 +923 + 3,833-2,281-4,769* *co -efficient not significant at a -.05. One additional unit of each of the following components will increase or decrease the price in dollars, by the amount shown above. 34
a response to housing condition rather than the detrimental effect of a tenant per se. The Average Income in the Neighborhood Many studies have shown that wealthy neighbors will improve the intrinsic value of a property. This study suggests that every $1,000 increase in the average income in the neighborhood will increase property values by $60. Proximity to a City Park The importance of proximity to open space has not typically been shown to significantly affect property values, yet this study reveals that a location within three blocks of a city park will increase the house price by an average of $3,000. This finding is particularly interesting when it is considered that the size of the lot on which a house is located does not significantly affect the estimated sales price. The open space provided by proximity to a park is seemingly more important to home buyers than is the open space provided by a large lot. Together these five variables account for up to 88% of the variation in house prices. The remaining percentage is unpredictable, and is related to matters such as financing arrangements, the need for a quick sale and anomalies such as houses in exceptionally poor condition. The figures presented in Table 2 can be used to estimate the likely sales price of a house. For example, consider a house which is located in south Tempe, has 1200 square feet of floor space, is five years old, is located close to a park, is taxed at $300 per year, and has no tenant. The estimated sales price of such a house should be: $39,548 (base price, south Tempe) +44,400 (1200 sq. ft. x $37 per sq. ft.) - 2,770 (5 years x - 554 per year) + 606 (parkside location) - 4,983 (taxed at $300; - $1,661 per $100) -::-:::-::-:::-::-=-0 (no tenant) $76,801 In northwest Tempe the identical house might be expected to sell for the following amount: $24,396 (base price northwest Tempe) +37,200 (1200 sq. ft. x $31 per sq. ft.) 280 (5 years x - $56 per year) + 3,845 (parkside location) - 1,134 (taxed at $300: - 378 per $100) -:::-::-~~O (no tenant) $64,027 EXCLUDED FACTORS Certain features, such as presence of a swimming pool do not show up as being important in the determination of market value. This is because square footage "masks" the effects of many other factors. For example, very few 900 square foot houses have pools, yet most 2,500 square foot houses in Tempe and Mesa do have pools; however, any price difference between the two houses is automatically ascribed to the difference in size. Other variables, such as the number of rooms, are dominated by the effects of size. A large house with a few rooms is valued more than a small house with several rooms. Other factors which are known to affect house prices, such as the quality of schools or the crime rate in the neighborhood do not vary much in this sample and are therefore included as part of the base price which is listed in Table 1. Some factors were simply not important. For example, electrical power in this study area is provided by two separate companies, one of which charges considerably higher rates than the other. However, the utility district in which the house was located did not have a significant effect on the price of the house. Lot size is another example of a variable which did not significantly affect property values. PRICE SUBMARKETS The price of houses at the high end of the housing market does not respond to the same characteristics as less expensive housing. The price of housing worth more than $120,000 does not change in response to changes in square footage. Different variables are important in determining the price of more expensive housing. In this kind of housing, 35
the basic requirements for space and living amenities have been met, and the value is driven by location and custom design characteristics. THE "AVERAGE" HOUSE The data for this study were taken from a sample of 227 houses listed in the Multiple Listing Service Directory. Of these, 186 were priced at $120,000 or under. These formed the basis for the analysis. The "average" house in the sample was priced at $73,706, had 1400 square feet of floor space, averaged 14 years old and carried a tax burden of $358. CONSUMER PREFERENCES This type of analysis is able to suggest which characteristics of housing are most valued by the house-buying public. Most variations can be explained by location and square footage, however, some other characteristics such as proximity to a park or the age of the house have also been shown to be important. This analysis holds promise to reveal consumer preferences for other housing, neighborhood or public service characteristics. The technique can therefore be used to explore other geographic variations in housing market responses as a response to changing consumer preferences. CONCLUSION This study has presented the results of an hedonic price analysis of housing in two adjacent metropolitan Phoenix suburbs. The analysis has revealed strong locational submarkets within the study area. For example, on average a location in south Tempe has been shown to be worth $15,000 more than a location in northwest Tempe, and a parkside location is worth $3,000 more than a non- parkside location. These are significant geographical influences on the price of housing at the intra-urban scale. For the average home buyer, these mesoscale variations in price are a stronger influence on the residential choice decision, than are the broader regional and national variations. This paper has demonstrated that these local geographic variations are an important ~omponent of a metropolitan housing market. NOTE 'This figure cannot be derived from a simple linear regres sion. because property values are used to determine taxation levels. which in turn influence property values. This estimate was derived from a more complex analysis of the same data set (Whalley. in preparation). REFERENCES Arizona Reg ional Multiple Listing Service. 1985. Residential Southeast Directory. Vol. 1. Bednarz. R.. and D. Wilson. 1982. Spatial Patterns of Property Value in a Small SMSA. Geographical Perspectives 50:50-57. Blackley. D. M. J. R. Follain and H. Lee. 1986. An Evaluation of Hedonic Price Indices for Thirty-four Large SMSAs. AREUEA Journal 14: 179-205. Butler. R. V.. 1980. Cross-Sectional Variation in the Hedonic Relationship for Urban Housing Markets. Journal of Regional Science 20:439-453. Follain. J. Roo and S. Malpezzi. 1980. Estimates of Housing Inflation for 39 SMSAs: An Alternative to the Consumer Price Index. Annals of Regional Science 14:41-56. Goodman. A. Hedonic Prices. 1978. Price Indices and Housing Markets. Journal of Urban Economics 5:471-484. Grlliches. Z. 1971. Hedonic Price Indices Revisited. in Z. Griliches. ed. Price Indices and Housing Markets. Harvard University Press: Cambridge. Ozanne. L.. and T. Thibodeau. 1983. Explaining Metropolitan Housing Price Differences. Journal of Urban Economics 13:51-66. Whalley. D. 1985. Hedonic Price Functions and Progressive Neighborhood Improvement: A Theoretical Exploration. Mathematical Social Sciences 10:275-279. Whalley. Doo A Multiequation Method for the Estimation of Property Tax Capitalization in Urban Housing Values. (in preparation). 36