Open space, forest conservation, and urban sprawl in Maryland suburban subdivisions.

Rapidly urbanizing jurisdictions face substantial challenges in maintaining the provision of public goods like preservation of open space and other scenic amenities. Numerous empirical studies have demonstrated that the presence of open space increases residential property values, suggesting that households place a positive value on this public good (Cheshire and Sheppard 1995; Geoghegan, Wainger, and Bockstael 1997; Tyrvainen and Mettinen 2000; Geoghegan 2002; Irwin 2002; Thorsnes 2002; Geoghegan, Lynch, and Bucholtz 2003; Wu, Adams, and Plantinga 2004; Hardie, Lichtenberg, and Nickerson 2007). Ensuring adequate provision of these amenities is one common justification for land-use regulations like zoning as well as voluntary preservation programs such as easement purchases (Bockstael and Irwin 2000).

But some open space preservation measures might induce developers to reduce the number of lots within subdivisions. Population increases would then result in more extensive development and thus contribute to urban sprawl. Theoretical analyses using closed and semi-closed city models show how minimum lot size zoning results in more extensive, lower-density development and an equilibrium urban boundary extending farther into rural areas (Moss 1977; Pasha 1996). McConnell, Walls, and Kopits (2006) provide empirical confirmation of these theoretical results: Their econometric analysis of Calvert County, Maryland shows that zoning regulations reduce density significantly. Irwin and Bockstael (2004) present econometric evidence indicating that preservation of open space can promote development of nearby land. Wu and Plantinga (2003) use simulations based on an open city model to show that public provision of open space can result in low-density, non-contiguous ("leapfrog") development.

This article examines the effects of two regulations on the average size and number of lots in suburban residential subdivisions: minimum lot size zoning and forest planting requirements under Maryland's Forest Conservation Act (FCA). We present a conceptual model of a developer's decisions regarding average lot size and the provision of forested and non-forested open space in the presence of these two regulations. We then examine the effects of these regulations empirically using data from subdivisions developed in the Baltimore-Washington suburbs during the mid-1990s.

The Maryland Forest Conservation Act

Concerned over rapid losses of forested land from development during the preceding three decades, Maryland enacted the FCA in 1991. The Act has been described in more detail elsewhere (see for example Galvin, Wilson, and Honeczy 2000; Hardie, Lichtenberg, and Nickerson 2007; Lichtenberg, Tra, and Hardie 2007). Briefly, the FCA applies to any project involving grading on 40,000 or more square feet (slightly less than an acre). Under the Act, developers must obtain approval for a forest conservation plan, specifying: the total amount and location of forested area retained; protective measures for stand edges and specimen trees; and measures that will protect retained forested areas permanently (e.g., covenants or easements incorporated into land deeds). The FCA also specifies minimum amounts of forested area to be provided. The FCA is administered by county planning agencies as part of the overall development permit approval process.

A Model of Land Allocation within a Residential Subdivision

Our conceptual framework builds on the model of a subdivision developer presented by Hardie, Lichtenberg, and Nickerson (2007). This model considers the problem of a land developer subdividing a parcel of fixed size L into n identical lots of size s and forested and non-forested open space, z and a, respectively. Forested and non-forested open space provide amenities f(z, [phi]s, [z.sup.o]) and h(a, [a.sup.o]), where [phi] denotes the share of forested area incorporated into building lots and [z.sup.o] and [a.sup.o] denote forested and non-forested open space nearby but outside of the subdivision. Households are assumed to be identical with willingness to pay per unit of developed land given by the bid rent function

R(s, f(z, [z.sup.o]), h(a, [a.sup.o]), y, T, g, u)

= y - T - x(s, f(z, [phi]s, [z.sup.o]), h(a, [a.sup.o]), g, u)/S. (1)

Here y denotes household income, T commuting cost, x a composite of all other purchased commodities, g other public good amenities (e.g., school quality), and u the equilibrium level of utility in the metropolitan area.

The land developer's goal is to maximize the rent generated by the subdivision

V [equivalent to] R(x)ns(1 + [gamma]) - cz - ka - Q(L) (2)

where [gamma] is the amount of land per building lot needed for roads, sidewalks, and other infrastructure, assumed fixed; c is the unit cost of afforestation; k is the unit cost of developing other open space; and Q(L) is the acquisition cost of the parcel, that is, the price of raw land prior to subdivision.

Development is subject to several constraints. First, development is constrained by the total area of the subdivision

ns(1 + [gamma]) + z + a = L. (3)

Second, zoning imposes a restriction on minimum lot size

s [greater than or equal to] [sigma]. (4)

Third, the FCA requires that the developer provide a minimum amount of forested area, which can consist of forested open space z or forested area incorporated into building lots [phi]ns

z + [phi]ns [greater than or equal to] [zeta]. (5)

Developers in the Maryland suburbs typically purchase entire farms for subdivision; hence, we assume that the constraint on total land availability (3) is always binding. If both regulatory constraints are binding, the developer's problem can be concentrated in the choice of forested and non-forested open space (z, a). The necessary conditions characterizing these choices are

[partial derivative]R/[partial derivative]f[[partial derivative]f/[partial derivative]z - [partial derivative]f/[partial derivative][phi] [sigma](1 + [gamma])(L + [gamma])(L - [zeta] - a)/[(L - z - a).sup.2]] x L - z - a/1 + [gamma] - R/1 + [gamma] - c [less than or equal to] 0 (6)

[[partial derivative]R/[partial derivative]h [partial derivative]h/[partial derivative]a + [partial derivative]R/[partial derivative]f [partial derivative]f/[partial derivative][phi] [sigma](1 + [gamma])([zeta] - z) /[(L - z - a).sup.2]] x L - z - a/1 + [gamma] - R/1 + [gamma] - k [less than or equal to] 0. (7)

With an interior solution, the choice of forested open space equates the increased value of building lots due to amenities provided by forested open space [partial derivative]R/[partial derivative]f [partial derivative]f/[partial derivative]z L - z - a/1 + [gamma] with the opportunity cost of land diverted from building lots R/1 + [gamma] plus the cost of developing forested open space c adjusted for any change in the value of building lots due to the substitution of forested open space for permanent forested open space incorporated into building lots [partial derivative]R/[partial derivative]f [partial derivative]f/[partial derivative][phi] [sigma](L - [zeta] - a)/(L - z - a). The choice of non-forested open space similarly equates the increased value of building lots due to amenities provided by non-forested open space [partial derivative]R/[partial derivative]h [partial derivative]h/[partial derivative]a L - z - a/1 + [gamma] with the cost of developing that open space k plus the opportunity cost of land diverted from building lots R/1 + [gamma] adjusted for any change in the value of building lots due to the substitution of non-forested open space for permanent forested open space incorporated into building lots [partial derivative]R/[partial derivative]f [partial derivative]f/[partial derivative][phi] [sigma]([zeta] - z)/(L - z - a).

Assuming that the FCA is met entirely using forested open space and ignoring infrastructure requirements ([phi] = [gamma] = 0), Hardie, Lichtenberg, and Nickerson (2007) show that an increase in minimum lot size [sigma] decreases the average value of land in the subdivision (Rns/L), while an increase in the FCA forestation requirement [zeta] increases the average value of land in the subdivision if willingness to pay for forested open space exceeds the opportunity cost of land. Assuming in addition that utility is Cobb-Douglas, Lichtenberg, Tra, and Hardie (2007) show that an increase in minimum lot size [sigma] decreases the amount of land devoted to total open space in the subdivision z + a, while a one-unit increase in the FCA forestation requirement [zeta] decreases non-forested open space and thus increases land devoted to total open space in the subdivision by an amount less than one.

Parameter estimates from econometric studies using data from a random sample of suburban single-family residential subdivisions in the Washington-Baltimore corridor bear out predictions derived from these theoretical models. The average value of land in these subdivisions was decreasing in zoned minimum lot size and increasing in the FCA forestation requirement (Hardie, Lichtenberg, and Nickerson 2007). Total open space was decreasing in zoned minimum lot size, while a one-acre increase in the FCA forestation requirement increased total open space by an amount significantly less than one, confirming the prediction that FCA forest planting requirements crowd out other forms of open space (Lichtenberg, Tra, and Hardie, 2007). Open space nearby but outside a subdivision had no effect on either the average value of land or total open space within the subdivision, indicating that the benefits of open space are largely internalized within subdivisions.

Data

We investigate the effects of these regulations on average lot size empirically using these same data, which are described in detail in Hardie, Lichtenberg, and Nickerson (2007) and Lichtenberg, Tra, and Hardie (2007). Briefly, the dataset comprises a random sample of single-family residential subdivisions approved for development during 1991-1997, in five Maryland counties (Charles, Carroll, Howard, Montgomery, and Prince Georges) in the Baltimore-Washington corridor. It includes information on: the size of developed lots; forest planting requirements under the FCA; zoning requirements; the availability of public water and sewer services; total subdivision size; geographical attributes of the subdivision, such as areas of floodplain and wetlands and linear stream frontage; commuting distances from Washington and Baltimore; the amounts of land surrounding the subdivision in farms, residential use, parks and recreational facilities, and undeveloped forest and brush; and similar information. Information on geographic features of each subdivision, subdivision size, the physical utilization of space within the subdivision (including the number and sizes of building lots and total area in open space), forest conservation plans (including FCA forest planting requirements), and the availability of public sewer service were obtained from county planning agency files. Maryland Property View county databases were used to obtain commuting (road) distance to the nearest central business district (Washington, DC or Baltimore) and the area within a given distance of the centroid of each subdivision in farmland, parks, and recreational facilities, and undeveloped forest and brush (combined). County zoning documents were used to determine minimum lot sizes corresponding to zoning codes obtained from the Property View data. Subdivisions regulated under transferable development rights (Montgomery County) or planned use development zoning (Prince Georges and Charles Counties) were excluded from the analysis, resulting in a usable sample of 229 subdivisions. Descriptive statistics can be found in Lichtenberg, Tra, and Hardie (2007).

Econometric Specification and Estimation

The dependent variables in our econometric models are the (1) average size and (2) number of building lots in each subdivision. Following the conceptual framework, we assume that average lot size is a function of: regulatory restrictions (zoned minimum lot size, the area planted to forest as required by the FCA, and whether the subdivision is exempt from the FCA); the size of the subdivision; geographic features of the subdivision that may limit the way space can be used (area of floodplain and wetland, linear stream frontage); land uses outside but nearby the subdivision (areas of farmland, parks, and forest/brush in a half-mile radius ring outside the subdivision); and the county in which the subdivision is located (a proxy for other regulatory restrictions).

We considered three specification issues: non-linearity, structural differences between subdivisions with and without access to public sewer service, and potential endogeneity of zoning.

We report the results of a linear model (table 1) because the coefficients lend themselves to more intuitive explanations. A log-linear model gave similar results.

The effects of zoning and FCA forest planting requirements may differ according to whether a subdivision has access to public sewer service. In areas without public sewers, public health regulations specifying the amount of land needed for septic systems may supersede minimum lot size zoning and thus also change the implicit cost of incorporating FCA forest plantings into building lots. Likelihood ratio tests rejected the hypothesis of no differences between subdivisions with and without public sewer access in both the linear and log-linear models, so we estimated separate models for these two classes of subdivisions.

It may be appropriate to treat zoning as endogenous because land-use regulations are frequently revised in response to economic pressures (see for example Wallace 1988; McMillan and McDonald 1991; Munneke 2005). Features of the zoning regulations in the counties we consider give further grounds for potential endogeneity. Howard County offers an explicit formula that trades lot size for open space, while other counties set different open space requirements for townhouses and detached homes. However, Hausman tests indicated no correlation between unobserved factors influencing zoned minimum lot size and both average lot size and the number of lots, so we estimated both equations treating zoning as exogenous.

Results

The econometric models for both classes of subdivisions fit the data quite well. The estimated coefficients confirm that zoning influences both average lot size and the number of lots in these subdivisions while FCA forest planting requirements influence average lot size but not the number of lots.

The coefficients of zoned minimum lot size are statistically significantly greater than zero in the average lot size equations; they are less than zero in the number of lots equations (significantly so in subdivisions with public sewer access). Interestingly, the coefficients of zoned minimum lot size are not significantly different from one in either average lot size equation (i.e., a one-acre increase in zoned minimum lot size is associated with a one-acre increase in average lot size). These results suggests that in the absence of minimum lot size zoning developers would provide a larger number of smaller lots, confirming theoretical predictions and empirical evidence from nearby Calvert County obtained by McConnell, Walls, and Kopits (2006) that zoning promotes urban sprawl by reducing density. In subdivisions with access to public sewer service the coefficient of zoned minimum lot size is greater than one in the average lot size equation and quite large in the number of lots equation, suggesting that zoning is highly restrictive in these closer-in areas. It is less than one in the average lot size equation and much smaller in magnitude (and not significantly different from zero) in the number of lots equation in subdivisions without access to public sewer service, possibly because septic system requirements are binding determinants of lot size and siting in these areas.

The coefficient of the FCA forest planting requirement is positive in both equations in both kinds of subdivisions. It is significantly different from zero in the average lot size equation but not in the number of lots equation in subdivisions with access to public sewer service. It is not significantly different from zero in either equation in subdivisions without public sewer access. It is quite small in magnitude in all cases. A one-acre increase in forest planting required under the FCA is associated with increases of a hundredth of an acre in the average size of lots and a quarter of a lot in subdivisions with access to public sewer service. It is associated with increases of a fiftieth of an acre in the average size of lots and less than a hundredth of a lot in subdivisions without public sewer access. These results suggest that developers respond to FCA forest planting requirements in part by incorporating permanent forested acreage into building lots, at least in closer-in subdivisions with public sewer access, but that they do so only to a very limited extent.

These econometric results indicate that FCA planting requirements increase average lot sizes and leave the number of lots per subdivision unchanged, thereby increasing total land in building lots ns. Lichtenberg, Tra, and Hardie (2007) find that FCA planting requirements also increase total open space z + a. The land availability constraint (2) suggests that land in building lots and open space can both increase only if the amount of land allocated to roads, sidewalks, and other forms of infrastructure is reduced. In other words, these results suggest that FCA planting requirements induce developers to lay out building lots and open space to economize on land allocated to infrastructure. It thus appears that even though these forest planting requirements increase average lot size they do not necessarily exacerbate sprawl since they do not appear to increase the amount of land utilized to accommodate any given level of population growth and hence do not push the urban boundary farther into the countryside than would be the case in their absence.

Concluding Remarks

Jurisdictions in rapidly urbanizing areas often enact policies to preserve open space and similar scenic amenities in the face of rising demand triggered by increasing scarcity of those amenities. But some of those policies may induce developers to reduce the number of housing lots within subdivisions, so that population increases result in more extensive development. It is possible, in other words, that open space preservation policies contribute to urban sprawl.

We present a conceptual framework of the choices facing a developer subdividing a parcel of fixed size. We use that conceptual framework to specify econometric models of average lot size and the number of lots per subdivision as functions of: minimum lot size zoning; forest planting requirements under the Maryland FCA; subdivision size; geographic features of the subdivision; subdivision location; and land use surrounding the subdivision. Our empirical analysis uses data from suburban subdivisions in the Baltimore-Washington suburbs. The estimated coefficient of minimum lot size zoning is significantly greater than zero (and not significantly different from one) in the average lot size equation and negative in the number of lots equation, implying that this form of regulation contributes to sprawl by reducing density and thereby confirming findings in the theoretical literature and empirical findings from nearby Calvert County, Maryland. The estimated coefficient of forested planting requirements under the FCA is positive in the average lot size equation, indicating that these requirements increase the average size of lots in suburban subdivisions, but also positive (albeit not significantly different from zero) in the number of lots equation. Thus, forest planting requirements increase average lot sizes but not the number of lots per subdivision, suggesting that they induce developers to economize on space allocated to roads, sidewalks, and other forms of infrastructure. Because forest planting requirements do not necessarily increase the number of lots per subdivision, they do not increase the amount of land needed to accommodate any given level of population growth and hence they do not contribute to sprawl. Thus, there may not be a conflict between forest preservation goals and prevention of urban sprawl.

References

Bockstael, N.E., and E.G. Irwin. 2000. "Economics and the Land Use-Environment Link." In T. Tietenberg and H. Folmer, eds. International Yearbook of Environmental and Resource Economics 2000/2001. Cheltenham, UK: Edward Elgar.

Cheshire, P., and S. Sheppard. 1995. "On the Price of Land and the Value of Amenities." Economica 6:247-67.

Galvin, M.E, B. Wilson, and M. Honeczy. 2000. "Maryland's Forest Conservation Act: A Process for Urban Greenspace Protection During the Development Process." Journal of Arboriculture 26:275-80.

Geoghegan, J. 2002. "The Value of Open Spaces in Residential Land Use." Land Use Policy 19:9198.

Geoghegan, J., L. Lynch, and S. Bucholtz. 2003. "Capitalization of Open Spaces into Housing Values and the Residential Property Tax Revenue Impacts of Agricultural Easement Programs." Agricultural and Resource Economics Review 32:33-45.

Geoghegan, J., L.A. Wainger, and N.E. Bockstael. 1997. "Spatial Landscape Indices in a Hedonic Framework: An Ecological Economics Analysis Using GIS." Ecological Economics 23:25164.

Hardie, I., E. Lichtenberg, and C.J. Nickerson. 2007. "Regulation, Open Space, and the Value of Land Undergoing Residential Subdivision." Land Economics" 83: 455-71.

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

Irwin, E.G., and N.E. Bockstael. 2004. "Land Use Externalities, Open Space Preservation, and Urban Sprawl." Regional Science and Urban Economics 34:705-25.

Lichtenberg, E., C. Tra, and I. Hardie. 2007. "Land Use Regulation and the Provision of Open Space in Suburban Residential Subdivisions." Journal of Environmental Economics and Management 54: 199-213.

McConnell, V., M. Walls, and E. Kopits. 2006. "Zoning, TDRs and the Density of Development." Journal of Urban Economics 59:44057.

McMillan, D.P., and J.F. McDonald. 1991. "A Simultaneous Equations Model of Zoning and Land Values." Regional Science and Urban Economics 21:14-27.

Moss, W.G. 1977. "Large Lot Zoning, Property Taxes and Metropolitan Area." Journal of Urban Economics 4:408-27.

Munneke, H.J. 2005. "Dynamics of Urban Zoning Structure: An Empirical Investigation of Zoning Change." Journal of Urban Economics 58:455-73.

Pasha, H.A. 1996. "Suburban Minimum Lot Size Zoning and Spatial Equilibrium." Journal of Urban Economics 40:1-12.

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

Tyrvainen, L., and A. Miettinen. 2000. "Property Prices and Urban Forest Amenities." Journal of Environmental Economics and Management 39:205-23.

Wallace, N.E. 1988. "The Market Effects of Zoning Undeveloped Land: Does Zoning Follow the Market?" Journal of Urban Economics 23:307-26.

Wu, J., R.M. Adams, and A.J. Plantinga. 2004. "Amenities in an Urban Equilibrium Model: Residential Development in Portland, Oregon." Land Economics 80:19-32.

Wu, J., and A.J. Plantinga. 2003. "The Influence of Public Open Space on Urban Spatial Structure." Journal of Environmental Economics and Management 46:288-309.

Erik Lichtenberg is Professor and Ian Hardie Professor Emeritus in the Department of Agricultural and Resource Economics, University of Maryland College Park.

Funding for this project was provided by the National Center for Smart Growth Small Grants Program, the USDA National Research Initiative Competitive Grants Program, and the Maryland Agricultural Experiment Station. Some data were provided from EPA STAR Grant R-82801201. We thank Lori Lynch, Jackie Geoghegan, Bruce Gardner, Anna Alberini, and participants of the AREUEA 2007 mid-year meeting for their comments and suggestions.

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. Estimated Parameters of the Average Lot Size and Number of Lots Models

Subdivisions with Public

Sewer Access

Average Number

Lots Size of Lot Variable Intercept 0.08639 54.65036 **

(0.05319) (11.08031) Subdivision exempt from -0.02513 -15.22887

FCA (yes = 1) (0.04181) (8.71092) Forested acres required by 0.01129 ** 0.26113

FCA (0.00369) (0.76947) Zoned minimum lot size 1.13259 ** -78.64303 ** (acres) (0.06485) (13.51071) Total site acreage 0.000436 0.88913 **

(0.000997) (0.20780) Acres of floodplain -0.00723 -2.37453

in subdivision (0.00578) (1.20472) Acres of wetland 0.017278 * -0.50431

in subdivision (0.00796) (1.65846) Linear feet of stream -0.0001 ** 0.01761 **

in subdivision (0.000025) (0.00524) Percentage of land within a 0.00362 * 0.27224

half mile (0.00149) (0.31123)

in farmland Percentage of land within a 0.00106 -0.34276

half mile in parks, public (0.00277) (0.57774)

spaces, etc. Percentage of land within a -0.0013 0.01271

half mile in forest, brush, (0.00110) (0.22861)

or undeveloped Commuting distance to 0.000775 -1.09250 *

nearest CBD (road miles) (0.00212) (0.44138) Subdivision located -0.13551 63.25595 **

in Charles County (0.09984) (20.79842) Subdivision located -0.24474 * -13.32909

in Carroll County (0.10026) (2.88585) Subdivision located (0.07560) -8.85609

in Howard County (0.04227) (8.80669) Subdivision located -0.13117 ** 5.33535

in Montgomery County (0.04628) (9.64200) [R.sup.2] 0.8194 0.4931 Number of observations 163 163

Subdivisions without

Public Sewer Access

Average Number

Lots Size of Lot Variable Intercept (2.80318) 13.97038

(2.18670) (8.03264) Subdivision exempt from 4.03485 ** -7.45351 *

FCA (yes = 1) (0.99072) (3.63933) Forested acres required by 0.02153 0.00794

FCA (0.02617) (0.09614) Zoned minimum lot size 0.65777 * -1.32597 (acres) (0.31124) (1.14332) Total site acreage -0.00327 0.23831 **

(0.01135) (0.04169) Acres of floodplain -0.00253 -0.21721 **

in subdivision (0.01873) (0.06879) Acres of wetland -0.03700 0.33103

in subdivision (0.05150) (0.18917) Linear feet of stream -0.00006 -0.00022

in subdivision (0.000154) (0.000566) Percentage of land within a 0.01524 -0.01114

half mile (0.02436) (0.08949)

in farmland Percentage of land within a 1.06904 4.13133

half mile in parks, public (0.83823) (3.07917)

spaces, etc. Percentage of land within a 0.05457 -0.12712

half mile in forest, brush, (0.03242) (0.11911)

or undeveloped Commuting distance to 0.03091 -0.02236

nearest CBD (road miles) (0.02680) (0.08945) Subdivision located -0.75091 -2.43527

in Charles County (1.66999) (6.13456) Subdivision located 0.85297 (0.66985)

in Carroll County (1.25878) (4.62401) Subdivision located 0.68730 (6.01609)

in Howard County (1.67620) (6.1574) Subdivision located 0.09892 (6.01609)

in Montgomery County (1.30738) (6.15739) [R.sup.2] 0.4974 0.8181 Number of observations 66 66 Note: Standard errors are shown in parentheses. Double asterisk denotes significantly different from zero at a l % significance level and single asterisk denotes significantly different from zero at a 59% significance level.