The trade-off between private lots and public open space in subdivisions at the urban-rural fringe.

In many communities, particularly those on the urban-rural fringe, most housing is located in subdivisions. Increasingly, those developments are subject to "clustering" rules, in which houses must be located on a portion of the total land and the remainder is left as open space. In some communities the zoning law mandates clustering; in others, clustering is recommended but not required. This open space may be held as forests, put into recreation uses, or remain in agriculture as grazing or cropland. Proponents of clustering requirements argue that undeveloped areas convey value, not only to the residents of the subdivisions themselves, but also to the broader community by preserving the aesthetic and rural character of the community and improving environmental quality through habitat protection or water pollution reduction (Arendt 1992). In communities on the urban-rural fringe, clustering residential developments is one approach to maintaining an agricultural base and curbing sprawl. (1)

Open space may provide benefits to subdivision residents, but clustering means that those residents are living in a higher-density setting compared to conventional subdivisions. Although the external benefits from the preserved forest, recreation area, or other kind of open space may be positive, it is unclear whether those benefits offset the loss experienced by smaller lots and higher density. Several studies have examined the value of open space (see McConnell and Walls 2005), but few have focused on subdivision open space and the trade-off with private lot space. This trade-off is the focus of our study. We estimate a hedonic price model with data on subdivision house sales occurring between 1981 and 2001 in a county on the fringe of the Washington, DC, metropolitan area, Calvert County, Maryland. We examine how households value adjacency to open space and more open space in the subdivision as well as how readily they will be willing to trade off those amenities with their own private lot space.

We find that private acreage positively affects prices but so does subdivision open space. And, there is some evidence that subdivision open space does substitute for private lot size. In addition, having a lot that is adjacent to subdivision open space appears to enhance the value of a house, particularly if the open space is not too steeply sloped. However, we find no evidence of willingness to trade off one's own lot size for adjacency to the open space. We use the results of the estimated hedonic model to simulate the effects on prices of jointly increasing open space and reducing average lot size. We find average house prices are slightly lower with the clustering, particularly for lots not adjacent to open space.

Overview of Relevant Literature

McConnell and Walls (2005) review the extensive revealed and stated preference literature on valuing open space. Almost all of the revealed preference studies use a hedonic property value approach. These studies look at a range of different types of open space--natural areas, wetlands, parks, forested lands, and different kinds of farmland, and they have used different measures of open space, including distance to open space and the percentage of open space within a particular radius of a property.

In this brief review, we emphasize studies that focus on the value of open space amenities within and surrounding residential subdivisions. Mohammed (2006) compares the costs and prices of "conservation" subdivisions with conventional subdivisions. He finds conservation subdivisions to have a price premium of 12-16% and also finds development costs to be lower in conservation subdivisions. (2) Lacy (1990) uses a similar methodology. Neither of these studies controls for the many other factors that could explain price differences.

Peiser and Schwann (1993) analyze house prices in a Dallas subdivision that has publicly usable open space between houses and survey residents about their preferences. While the survey findings suggest that households place a high value on open space, the hedonic price analysis shows otherwise. Houses on the open space generally sold at a premium, but the effect was statistically insignificant and much smaller in magnitude than the effect of the size of the private lots themselves.

Thorsnes (2002) uses sales data from three subdivisions in Grand Rapids, Michigan, each bordering permanently preserved forested lands, to estimate hedonic price equations for both building lots and houses for each subdivision. He finds that lots next to the preserves sell at a 19-35% premium over the total lot price, but the benefits are very localized and do not extend even to parcels across the street. Using a random sample of subdivisions in five Maryland counties, Hardie, Lichtenberg, and Nickerson (2006) calculate the average peracre price of developed lots in 255 subdivisions and estimate this price as a function of subdivision characteristics, including: percentage of the subdivision area that must be planted in trees (by the state Forest Conservation Act). They find that the amount of land in forests has a positive effect on prices.

Patterson and Boyle (2002) focus on the value of a view. Using a hedonic estimation, they find that the visibility of forested areas has higher value than visibility of agricultural land. Kearney (2006) in a survey of residents of subdivisions with different configurations and different amounts and types of open space, also finds that the views from a home, and particularly views of forested areas, are strongly valued by residents.

There are also studies by urban planners who look at the environmental merits of clustering rules and conservation subdivisions. Berke et al. (2003) find that there are potential watershed protection benefits, and Kaplan, Austin, and Kaplan (2004) find that "natural features" of clustered subdivisions appear to be important to residents.

Data

Calvert County is located in southern Maryland, on the western shore of the Chesapeake Bay. It has 101 miles of shoreline, along the Chesapeake Bay and the Patuxent River to the east. The county has historically had an agricultural economy consisting of small villages and rural lands, but in the past twenty years, population growth has been high, and it has increasingly become part of the broad Baltimore-Washington metropolitan area.

Most of the housing growth in recent years has been in low-density suburban subdivisions in the residential and rural areas of the county. Figure 1 shows average lot sizes within the county during different time periods. (3) Although the average gross lot size, calculated as total subdivision acreage divided by the number of houses, has remained relatively high and constant over time, the average lot size net of open space has declined. This provides some indication of the extent to which clustering has been increasing in the county in recent years. Gross lot size trended up in the late 1990s due to the downzoning that occurred, but actual house lots continued to fall in size slightly, reflecting more open space in subdivisions.

In this study we limit the sample to subdivisions that had at least ten house sales over the study period, 1981-2001. This allows us to include 3,386 individual house sales within eighty-nine subdivisions. Table 1 provides summary statistics for house and subdivision level variables included in the model. The mean lot size is 1.5 acres and subdivisions are, on average, 134 acres, with a little over 20% of their land under easement as protected open space. The degree of clustering varies considerably over the sample; however, sixteen of the eighty-nine subdivisions have minimal open space (less than one acre), and twenty subdivisions have over 40% of their acreage in open space.

[FIGURE 1 OMITTED]

Econometric Model and Results

We assume households choose housing characteristics, location, and open space amenities so as to maximize utility. Under this assumption and a housing market in equilibrium, we can use the hedonic price model (see Rosen 1974) to examine consumer behavior with regard to housing choices. The hedonic price function can be specified as P = f(l, C, S, T, O), where P is the price of the property; l represents the lot size; C is a vector of structural characteristics (age, number of bathrooms, square footage, etc.) associated with the house; S is a vector of subdivision characteristics other than open space amenities; T represents a vector of accessibility measures; and O is a vector of open space attributes.

Since evidence from the literature suggests that the value of open space amenities to residents may vary by proximity or the type of open space (e.g., number of trees, usability, steepness), in our model we include three subdivision open space variables: open space acreage, a dummy for whether a house is adjacent to subdivision open space, and the percentage of subdivision open space that is in steep slopes. We also include interaction variables, which we discuss more below, as well as surrounding land-use variables, including adjacency to preserved agricultural land.

Several econometric issues must be addressed in estimating the hedonic model. In a study focusing on farmland and forested lands, Irwin (2002) discusses the problem of possible endogeneity of these kinds of open space. Since we are focusing on the amount of open space inside the subdivision in this study, it is unlikely that endogeneity is a serious concern. Moreover, all subdivisions since 1993 have had minimum open space requirements. We also assume households know and accept county rules, which require that subdivision open space has a permanent easement barring future development.

We consider the issue of unobserved spatial correlation in the error term, a common source of inefficiency and inappropriate covariance estimates in spatial models. To partially address this problem, we have included dummy variables for the thirty-one census block groups in our sample to account for any unexplained effect of different neighborhoods on prices. (4) In addition, we tested for spatial autocorrelation and rejected the null hypothesis that it is not present, even with block group dummies. (5) Therefore, to account for spatial correlation caused by misspecification of the regression function (e.g., omitted variables), we specify the error term with a standard firs[order AR process. The errors are assumed to depend on the weighted average of the errors of "neighboring" houses, which we define to be houses that are in the same or adjacent subdivisions. (6) The results are shown in table 2. We tested the sensitivity of our results by estimating the model with an alternative weighting scheme where all houses within one mile of each other were considered neighbors. (7) This latter specification yielded coefficients of similar magnitude and significance as the results presented in table 2.

Results: Preferences for Lot Size and Subdivision Open Space Amenities

Households have a consistent preference for larger lots, ceteris paribus. We calculate the marginal willingness to pay for additional private acreage and subdivision open space by the partial derivatives of the price function with respect to each attribute, evaluated at the mean values of the relevant interaction variables. The elasticity estimates at the bottom of table 2 summarize the results of that calculation. We find that a 10% increase in private lot size is associated with an approximately 0.6% increase in house price, ceteris paribus. This suggests that for an average priced house in 2004 (about $300,000), an increase in lot size from 1 to 1.5 acres would increase price by about $9,000 (year 2000 dollars). The magnitude of this estimate is robust across various specifications of the model, including one with subdivision fixed effects.

The amount of open space in the subdivision, given subdivision size, is also statistically significant, and its effect on house prices is positive, but small. A 10% increase in subdivision open space leads to a 0.1% increase in average house price, ceteris paribus. This result was also robust to alternative specifications of the model. This suggests that increasing open space acreage from 20 to 30 acres would increase sales price by 0.5 percent, or $1,500 per house (evaluated at an average of $300,000), ceteris paribus. Of course, all houses in the subdivision would be affected if open space increased.

The significant, negative interaction term between the amount of open space and own lot size provides evidence that residents will trade off their own lot size for the amount of open space in the subdivision. Adjacency to subdivision open space also has a positive effect on house prices, but the magnitude of the effect depends on how much of the open space is in steep slopes. The greater the percentage of open space that is steep, the smaller the impact that adjacency has on house prices. (8)

Perhaps our most surprising result is that we find households unwilling to trade off their own lot size to be adjacent to open space. One explanation for this may be a result from the literature that suggests that proximity to open space is less valuable than having a view of forested or undeveloped areas (e.g., Patterson and Boyle 2002).

Results: Other Variables

Most of the other explanatory variables in the model are significant and of the expected sign. All of the variables describing house characteristics and variables measuring proximity to commuting routes are significant at the 99% or 95 % level. The northern edge of the county marks the closest point in the county to urban centers of Washington, DC, and Baltimore; moving the average house one mile farther south reduces house price by a little more than 1%. Locating farther from the major highway in the county, Route 2/4, also significantly reduces sales price. Larger and newer subdivisions tend to have slightly higher-priced houses.

Some of the other amenities and surrounding land uses are important in explaining house prices while others are not. Being on the water is highly valuable: sales prices of waterfront houses (on the Patuxent River or Chesapeake Bay) are found to be 30% higher than prices of other houses. However, being adjacent to parkland, privately owned preserved farmland, or the open space area of another subdivision does not significantly affect housing prices. (9)

Simulating the Effect of Clustered Subdivisions on House Prices

We can illustrate the overall effects of changes in subdivision configuration by a simple simulation. We start with a representative subdivision in our sample: 134 acres in size, with about 30 acres of open space and an average lot size of 1.5 acres. Holding total subdivision size and the number of lots constant, doubling the amount of open space to about sixty acres would require average lot size to fall to 1.1 acres. Based on the results in table 2, we find that such an increase in clustering (from about 22% to 44%) would decrease the average house price by 1.2% (for a house not adjacent to open space). The loss in value from the smaller lot size dominates any increased value from more subdivision open space. The additional clustering may also increase the probability of a house being adjacent to the open space area, however, and this adds some value. For houses on lots that become adjacent to subdivision open space as a result of the increased clustering, we find the change in sale price is minimal, decreasing by only 0.3%.

Conclusions

Our results suggest why we may not see many clustered subdivisions on the urban--rural fringe without government regulations requiring such clustering. Households appear to strongly value their own private lots. While we do find in our analysis that households also value having more open space in their subdivisions, or having a lot that is adjacent to subdivision open space, they do not value these amenities nearly as much as a larger lot. Thus, reducing private acreage to provide more public subdivision open space tends to lead to overall reductions in house prices, all else equal.

One of the most important questions we wanted to address in this study is whether households would be willing to trade off the size of their own lot for open space in the subdivision. Clustering subdivision development is being viewed as a way to reduce the development footprint and preserve open space in fringe communities. Our findings suggest that there is some small willingness to trade off lot size for more subdivision open space. One caveat to our findings is that they may be specific to the community we were examining--one on the urban-rural fringe with very large average lot sizes and a great deal of surrounding open space and farmland. It is possible that households in these areas value their large lots and also have adequate substitutes for subdivision open space.

Our analysis only attempts to measure the effects of subdivision open space on property values within the subdivision. The external benefits of subdivision open space, such as aesthetic values, and ecological and environmental benefits, may accrue to the larger community. These benefits will not be capitalized into subdivision property values, and to the extent they are important, suggest additional reasons why the private market may underprovide open space and government intervention may be necessary.

Land-Use Policy Experiments at the Rural-Urban Interface (Lori Lynch, University of Maryland, Organizer)

References

Arendt, R. 1992. "'Open Space' Zoning: What It Is and Why It Works." Planning Commissioners Journal 5 (July), pp. 1-9, http:// www.plannersweb.com/articles/are015.html.

Berke, R, J. McDonald, N. White, M. Holmes, K. Oury, and R. Ryznar. 2003. "Greening Development to Protect Watersheds." Journal of the American Planning Association 69:397-413.

Daniels, T.L. 1997. "Where Does Cluster Zoning Fit in Farmland Protection?" Journal of the American Planning Association 63:129-37.

Hardie, I., E. Lichtenberg, and C. Nickerson. 2006. "Regulation, Open Space, and the Value of Land Undergoing Residential Subdivision." University of Maryland.

Irwin, E.G. 2002. "The Effects of Open Space on Residential Property Values." Land Economics 78:465-80.

Kaplan, R., M.E. Austin, and S. Kaplan. 2004. "Open Space Communities: Resident Perceptions, Nature Benefits, and Problems with Terminology." Journal of the American Planning Association 70:300-12.

Kearney, A.R. 2006. "Residential Development Patterns and Neighborhood Satisfaction, Impacts of Density and Nearly Nature." Environment and Behavior 38:112-39.

Lacy, J. 1990. "An Examination of Market Appreciation for Clustered Housing with Permanent Open Space," University of Massachusetts Center for Rural Massachusetts report, Amherst, MA.

McConnell, V., and M. Walls. 2005. The Value of Open Space: Evidence from Studies of NonMarket Benefits. Washington DC: Resources for the Future, January.

Mohamed, R. 2006. "The Economics of Conservation Subdivisions: Price Premiums, Improvement Costs, and Absorption Rates." Urban Affairs Review 4:376-99.

Patterson, R., and K. Boyle. 2002. "Out of Sight, Out of Mind: Using GIS to Incorporate Visibility in Hedonic Property Value Models." Land Economics 78:417-25.

Peiser, R.B., and G.M. Schwann. 1993. "The Private Value of Public Open Space within Subdivisions." Journal of Architectural and Planning Research 10:91-104.

Rosen, S. 1974. "Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition." Journal of Political Economy 82:34-55.

Thorsnes, R 2002. "The Value of a Suburban Forest Preserve: Estimates from Sales of Vacant Residential Building Lots." Land Economics 78:426-41.

(1) Daniels (1997) takes a critical view of the use of clustering for preserving farmland on the urban fringe, arguing that it may be incompatible with an agricultural economy.

(2) The term "conservation" subdivision usually implies something broader than clustering. See Mohammed (2006) for a description of a conservation subdivision.

(3) The time periods are chosen to reflect zoning and other changes in the county.

(4) We tested a number of different specifications, including a subdivision fixed effects model. While that specification has the advantage of controlling for all unobserved subdivision characteristics, it prevents us from estimating effects of specific subdivision level variables, including the total amount of subdivision open space. Coefficients on house-specific variables are very similar to the results with block group fixed effects.

(5) The Moran I statistic is found to be 11.570.

(6) Letting u denote the vector of error terms, [u.sub.i], i = 1 to N, in our model, we assume [u.sub.i] = [rho][W.sub.i]u + [[epsilon].sub.i], where [rho] is the spatial autocorrelation parameter to be estimated, [W.sub.i] is the ith row of the weighting matrix, W, and [[epsilon].sub.i] is the component of the error term made up of independently and identically distributed (iid) random variables. The weighting matrix, W, selects the "neighbors" so that W = {[w.sub.ij]}, where [w.sub.ij] = 1/[n.sub.i] (where [n.sub.i] = number of "neighbors" of house i) if i and j are within the same or adjacent subdivisions; because a subdivision is not viewed as its own neighbor, [w.sub.ii] = 0 for all i.

(7) With the one mile weighting scheme, the Moran I test statistic is 12.895.

(8) Our steepness variable cannot measure the slope of the open space adjacent to particular houses, but the more open space in the subdivision in steep slopes, the higher the probability that any house adjacent to open space will be adjacent to steep open space.

(9) Adjacency to another subdivision's open space becomes significant at the 10% level if subdivision fixed effects are included in the model.

Elizabeth Kopits is Economist at the U.S. EPA National Center for Environmental Economics. Virginia McConnell is Professor of Economics at University of Maryland and Senior Fellow at Resources for the Future (RFF). Margaret Walls is Senior Fellow at RFF.

The views expressed in this article are those of the authors and do not necessarily represent those of the U.S. EPA. No official Agency endorsement should be inferred. The helpful comments of Soren Anderson are greatly appreciated.

This article was presented in a principal paper session at the AAEA annual meeting (Portland, OR, July 2007). The articles in these sessions are not subjected to the journal's standard refereeing process. Table 1. Summary Statistics for House and Subdivision Data, Calvert County, MD Variable Mean SD House variables (3,386 house sales)

House sale price (a) 221,749.30 74,000.48

Lot size (acres) 1.511 1.586

Sale year 1,994.422 5.087 Structural characteristics:

House size (square feet) 2,075.047 801.959

House age (years) 5.821 10.026

Dwelling grade (dummy) (b) 0.177 0.382

Number of full bathrooms 1.999 0.512

Number of half bathrooms 0.583 0.498

Fireplace (dummy) 0.618 0.486

Townhouse 0.008 0.091 Open space and surrounding land uses (dummy variables):

Adjacent to subdivision

open space area 0.249 0.433

Adjacent to another

subdivision open space 0.024 0.153

Adjacent to preserved

farmland, parkland 0.014 0.118

Adjacent to undeveloped,

unpreserved land 0.142 0.351

Adjacent to water 0.016 0.126 Subdivision variables (89 subdivisions)

Size of subdivision (acres) 133.615 104.853

Size of subdivision open

space area (acres) 28.951 46.316

Pct.open space in steep

slopes (>15% slope) 0.373 0.347

Subdivision recording year 1,983.675 10.218 Accessibility/location:

Distance to the Northern

border (meters) 18,965.17 13,351.70

Access to Town Center

(gravity index) 10.732 83.375

Distance to Route 2/4

(meters) 2,419.191 1,799.510

RCD Rural zoning district

(dummy) 0.685 0.467

FCD Rural zoning district

(dummy) 0.079 0.271

Residential zoning district

(dummy) 0.213 0.412

Town Center (dummy) 0.022 0.149 Variable Range House variables (3,386 house sales)

House sale price (a) 12,642-939,180

Lot size (acres) 0.034-30.41

Sale year 1,981-2,001 Structural characteristics:

House size (square feet) 576-6,575

House age (years) 0-186

Dwelling grade (dummy) (b) 0-1

Number of full bathrooms 1-5

Number of half bathrooms 0-2

Fireplace (dummy) 0-1

Townhouse 0-1 Open space and surrounding land uses (dummy variables):

Adjacent to subdivision

open space area 0-1

Adjacent to another

subdivision open space 0-1

Adjacent to preserved

farmland, parkland 0-1

Adjacent to undeveloped,

unpreserved land 0-2

Adjacent to water 0-1 Subdivision variables (89 subdivisions)

Size of subdivision (acres) 16.61-589.59

Size of subdivision open

space area (acres) 0.250-295.130

Pct.open space in steep

slopes (>15% slope) 0-1

Subdivision recording year 1,928-1,999 Accessibility/location:

Distance to the Northern

border (meters) 955.72-49,912.3

Access to Town Center

(gravity index) 0-768.431

Distance to Route 2/4

(meters) 167.87-7,633.7

RCD Rural zoning district

(dummy) 0-1

FCD Rural zoning district

(dummy) 0-1

Residential zoning district

(dummy) 0-1

Town Center (dummy) 0-1 (a) Year 2000 dollars. (b) Dwelling grade equals 1 if house is categorized as low to fair quality, 0 otherwise. Table 2. Results of the Spatial Error Model, with Block Group Fixed Effects (Dependent Variable is the natural log of house sale price) Variable Description Coefficient (t-Stat) 1 Own lot size (acres, logged) 0.078 *** (10.423)

Variables related to subdivision

open space 2 Subdivision open space

(acres, logged) 0.010 ** (2.279) 3 Percent of open space acres

in steep slopes -0.024 (-1.140) 4 Subdivision open space (var 2)

* pct steep (var 3) -0.003 (-0.410) 5 Subdivision open space (var 2)

* own lot size (var 1) -0.007 *** (-2.715) 6 Adjacent to own subdivision

open space (dummy) 0.029 ** (2.181) 7 Adjacent to own open space

(var 6) * pct steep (var 3) -0.059 ** (-2.327) 8 Adjacent to own open space

(var 6) * lot size (var 1) 0.016 (1.512)

Other adjacency variables 9 Adjacent to another subdivision's

open space area 0.010 (0.582) 10 Adjacent to water 0.300 *** (12.805) 11 Adjacent to undeveloped,

unpreserved land -0.006 (0.741) 12 Adjacent to preserved farmland

or parkland -0.012 (0.532)

House characteristics 13 House size (square ft, logged) 0.280 *** (23.042) 14 Age of house -0.002 *** (-5.896) 15 Dwelling grade -0.090 *** (-6.845) 16 Number of full baths 0.073 *** (10.159) 17 Number of half baths 0.039 *** (5.437) 18 Fireplace (dummy) 0.037 *** (5.922) 19 Townhouse (dummy) -0.113 ** (-2.435)

Accessibility variables 20 Distance to northern border

(meters, logged) -0.129 *** (-4.361) 21 Distance to Route 2/4

(meters, logged) -0.026 ** (-2.533) 22 Accessibility to Town Centers 0.000 (0.245)

Other subdivision variables 23 Subdivision size (acres, logged) 0.026 *** (2.751) 24 Year subdivision was recorded 0.002 *** (75.502) 25 Subdivision in Farm Community District 0.011 (0.473) 26 Subdivision in Residential zone -0.025 (1.241) 27 Subdivision in Town Center 0.037 (0.168) 28 Constant 4.792 (14.909) 29 Spatial autocorrelation parameter, [rho] 0.358 (41.269) [R.sup.2] 0.7795 Elasticity of Sales Price Marginal Effect with Respect to: Eval. at Variable

Means (t-Scat in

Parentheses) Own lot size (a) 0.055 *** (7.15) Subdivision space acreage 0.006 * (1.75) Adjacency to own subdivision open space 0.014 * (1.68) (a) Marginal effect for interior lot; for lot adjacent to open space, marginal effect is 0.070. Note: Coefficients on sale year and census block group dummy variables are available upon request. Triple asterisk signifies significance at 99% level; double asterisk at 95% single asterisk at 90%.