The cost of land use regulation versus the value of individual exemption: Oregon Ballot measures 37 and 49.
For a long time, property rights advocates have maintained that the current level of protection for private property rights, particularly rights associated with real property such as land, is inadequate. They argue that although private property owners can sue the government for regulatory takings and demand compensation under the Fifth Amendment to the U.S. Constitution, the process for getting their claims heard in court is too arduous and, even if a case is heard, the conclusion of no taking that is typically the outcome is unfair because the property owners alone are bearing the burden of regulations that generate benefits for all of society (Congressional Budget Office 1998). They also contend that governments tend to overregulate--imposing restrictions beyond the point at which the additional benefits of more regulation are at least as large as the additional costs--because governments rarely bear the costs of regulation.
Property rights advocates see the State of Oregon as an ideal battle ground for changing the current approach to regulatory takings because the state's land use regulation is among the most stringent in the United Slates. The statewide land use planning regime in Oregon places strict limits on the use of resource lands (namely, farmland and forestland) and confines development within designated urban growth boundaries (UGBs). The planning regime contains specific objectives for protecting resource lands, wildlife habitat, and water resources through a number of zoning ordinances and other regulations. In Oregon, local comprehensive land use planning is mandatory.1
In an attempt to protect private property rights from the stringent land use regulations, property rights advocates launched Ballot Measure 37 in Oregon in the 2004 election. Measure 37 provides that the state and local governments must compensate the owner of private real property when a land use regulation reduces its "fair market value" or "remove, modify, or not apply" the regulation. Reflecting controversies surrounding Oregon's land use regulations,2 Measure 37 was passed by 61% to 39% on November 2, 2004. By October 25, 2005, 1,255 claims had been hied with the state, requesting $2.2 billion in compensation and covering 66.000 acres of land (Oregon Department of Land Conservation and Development [ODLCD] 2007). However, to the dismay of property rights advocates. Measure 37 was ruled unconstitutional by a lower court on October 14, 1005v The judgment was appealed to the Oregon State Supreme Court, which upheld Measure 37 in a ruling issued on February 21. 2006. By November 17, 2006, 1,918 more claims had been filed after the ruling, requesting additional $3.9 billion in compensation (ODLCD 2007). In an effort to reverse or modify Measure 37, Oregon voters approved Measure 49 in November 2007 to "ensure that Oregon law provides just compensation for unfair burdens while retaining Oregon's protection for farm and forest land uses and the state's water resources" (ODLCD 2008). Measure 49 essentially modifies Measure 37 by replacing "waivers" of regulations with authorizations to establish a limited number of homesites.
Although Measure 49 has ended another battle of a long war, it does not settle a host of controversies surrounding land use regulations. Central among these controversies is how to determine the "just compensation" for ^unfair burdens." In practice, "just compensations" are often determined through appraisals by calculating the increase in land value when a parcel is exempted from the regulation. However, the increase in the value of a parcel when it is solely exempted from a regulation does not equal the true cost of the regulation. A landowner may benefit from a waiver maybe because the regulation has been applied to other properties. The true cost of the regulation equals the difference between the current value of the properly and the value of the property that would have existed if the regulation had never been imposed in the first place. As the future flows of benefits accruing to the owner of a parcel can be affected by its surrounding land uses (e.g., through agglomeration effects or externalities), the price of a parcel, which reflects the present value of the future flows of benefits, depends in part on the land use in the surrounding parcels. The values of these surrounding parcels are in turn affected by the land use in their surrounding areas. This suggests that it would be necessary to predict the land use patterns and prices that would have existed on the whole landscape when the regulation had never been imposed in the first place.
Because of spatial externalities, the interaction between land markets and regulations is extremely complex. This article presents a simulation model to analyze the interaction. More specifically, we develop a geographic information system (GlS)-based, spatially explicit model to predict land use and property values for a large number of parcels on a landscape near Eugene, Oregon. The model is applied to simulate development patterns and land prices that would have existed if one or more land use regulations had never been imposed. The value of an individual exemption and the cost of land use regulation are calculated for all parcels on the landscape under a number of policy scenarios. The value of an individual exemption is defined as the increase in the value of a parcel when it is solely exempted from a regulation. The cost of regulation is defined as the difference between the current value of the property and the value of the property that would have existed if the regulation had never been imposed in the first place. It is important to note that, because of the data limitations, this study must make some restrictive assumptions to assess the values of land parcels in the absence of land use regulations. In addition, it is likely that the model does not capture the full spectrum of spatial and temporal dynamics of land use changes over a long time period (see the data section for a detailed discussion). Thus, the analysis is exploratory in nature, and the quantitative results must be interpreted with caution.
Our empirical results show that all major zoning regulations in the study region have affected land use and property values both inside and outside the zoned districts. Although the value of a property can never go down when it is solely exempted from a binding regulation, the cost of regulation can be positive or negative. Even when both the cost of regulation and the value of an individual exemption are positive, their magnitude can be quite different. In most cases relevant to Measure 37, the costs of regulation are considerably smaller than the values of individual exemptions. Only the EFU zoning and the UGB designation result in a net increase in total land value in the study region. However, the question whether governments over-regulate--from the perspective of society as a whole--remains unanswered because land use regulations also generate public goods, which are not fully taken into account in this article.
Numerous studies have analyzed the impact of land use regulations on property values (e.g., Beaton and Pollock 1992; Green 1999; Katz and Rosen 1987; Malpezzi, Chun, and Green 1998; Pollakowski and Wachter 1990; Quigley and Rosenthal 2004; Wu and Cho 2007). But few have compared the effects of land use regulations on prices of regulated land versus prices of unregulated land. Henneberry and Barrows (1990) examine the effects of exclusive agricultural zoning in Wisconsin and find that a premium was capitalized in the value of farmland zoned for farm use only compared to farmland with less certain future. Fischel (1990) reviews empirical research on growth controls and concludes that these measures have spillover effects on neighboring municipalities without such restrictions. Parsons (1992) examines the effect of restrictions on residential development on land adjacent to the Chesapeake Bay and finds that the restrictions increased housing prices within the designated area as well as prices near but not in the designated area. Beaton (1991) finds that regulations may affect the price of unregulated lands because of not only the actual imposition of growth controls but also the anticipation of new growth controls on developable land. Schaeffer and Millerick (1991) find that historic district regulations increase land values within the historic district as well as in areas adjacent to the historic districts. Cheshire and Sheppard (2002) develop an empirical approach to evaluate the benefits and costs of land use planning in the context of the Town and Country Planning System of the United Kingdom and find that the net benefit is substantially negative. Walsh (2007) evaluates the equilibrium impact of open space protection and growth control policies on the entire metropolitan landscape and finds that although a growth ring strategy is most effective in reducing total developed acreage in the metropolitan area, this reduction in developed acreage is associated with a large net welfare loss.
Several studies have evaluated the impact of land use regulations on land prices in Oregon (e.g., Cho, Wu, and Boggess 2003; Jun 2004; Kline and Alig 1999; Netusil 2005). For example, Kline and AJig (1999) analyze the efficacy of Oregon's land use program and find that although it has concentrated development within UGBs, its success at reducing development on resource lands remains uncertain. Cho, Wu, and Boggess (2003) analyze the development controls in five western states (including Oregon). They find that regulations reduce the total developed area but increase housing prices, and the magnitude of the effects vary across counties, depending on the stringency of land use regulations. Netusil (2005) finds that environmental zoning in Portland, Oregon, and the related amenities have a positive but small net effect on the value of single-family residential properties. Plantingaa (2004) argues that the reduction in market value resulting from a land use regulation is a fundamentally different concept than the value of an individual exemption to an existing regulation. To our knowledge, no.study has estimated the cost of regulation versus the value of individual exemption (VIE) empirically.
II. OREGON BALLOT MEASURES 37 AND 49
Oregon Ballot Measure 37 was passed on November 2, 2004. By December 5, 2007, 6,857 Measure 37 claims had been filed with the state, requesting a total of $20 billions in compensation. 3,574 final orders had been issued by December 5, 2007; most of the ruled claims had resulted in waivers of regulations, only a small percentage in denial (ODLCD 2008).
Measure 37 was extremely controversial, pitting neighbors against neighbors. Hundreds of court cases relating to Measure 37 were filed; many by claimants, challenging claim decisions; others by neighbors of claimants, challenging claim decisions; a few cases contesting the constitutionality, interpretation, or application of the measure.
On November 6, 2007, Oregon voters approved Measure 49, which modifies Measure 37 to give landowners with Measure 37 claims the right to build homes as compensation for land use restrictions imposed after they acquired their properties (State of Oregon 2007). Claimants may build up to three homes if previously allowed when they acquired their properties, four to ten homes if they can document reductions in property values that justify additional homes, but may not build more than three homes on high-value farmlands, forestlands, and groundwater-restricted lands (State of Oregon 2007). As Measure 37, Measure 49 provides compensation or waiver for new land use regulations. However, Measure 49 defines the category of land use regulations that are eligible for relief more narrowly, to include only those regulations that limit residential uses of property or that restrict farming or forest practices. In addition, under Measure 49. relief is provided only if the owner demonstrates that the new regulations have reduced the value of property. For claims based on regulation of residential uses, claimants are exempted from regulation only to the extent necessary to allow additional residential development of a value comparable to the value lost as a result of the regulation. Measure 49 took effect on December 6, 2007 (State of Oregon 2010).
Most Measures 37 and 49 claims were related to zoning regulations. EFU zoning, forest zoning, mixed farm/forest zoning, and rural residential zoning account for, respectively, 75%, 12%, 2%, and 10% of the Measure 37 claims filed by October 25, 2005. Eighty-six percent of claims (919 claims) requested land division and dwelling use, and 13% (134 claims) requested dwelling use only (ODLCD 2007).(4)
This study focuses on the effects of zoning ordinances given that almost all Measure 37 claims are related to zoning. Specifically, we examine the following zoning ordinances: EFU zoning, forest zoning, UGB designation, residential density zoning, commercial zoning, and industrial zoning. We examine their effects on land use patterns and land prices when they are used individually and in combinations.
Oregon's land use program is characterized by a clear separation between urban and rural land uses. This is achieved by several instruments, with the UGBs being perhaps the most prominent (or notorious) one. UGBs are lines drawn around urban areas within which all urban development must take place. UGBs thus place an absolute limit on urban development. Land outside the boundary is available only for resource uses unless specifically exempted. However, local governments must include sufficient land within UGBs to meet the requirements for urban land uses, including housing, industry, commerce, recreation, and open space for the next 20 years.
The protection of agricultural land for farm use is effectuated by placing all prime farmland in EFU zones.^ Local governments are also required to inventory, designate, and zone forest lands by limiting uses which can have significant adverse effects on forest land, operations, or practices (ODLCD 2005). Residential zoning in Oregon aims specifically at increasing the population densities in urban areas and encourages construction of affordable housing. Comprehensive land use planning in Oregon also includes designation of land as commercial or industrial zones.
The study area is located in the southern part of Oregon's Willamette Valley. The valley is the state's most heavily urbanized area and is home to about two-thirds of the state's population. About two-thirds of Measure 37 claims came from the valley. The study area (27 square miles, 70 km2) includes urban lands within the Eugene-Springfield metropolitan area as well as rural lands in exurban areas. Table 1 provides a summary of land uses and land values in the study area.
TABLE 1 Land Value and Parcel Area by Land Use Class Land Value Total Acres ($/Acre) Land Mean Standard Mean Standard Sum Use/Cover Deviation Deviation Rural lands 42.181 66,678 5,203 Forest 11.19 19.71 1,766 Agricultural 16.91 30.43 3,003 Rural 2.72 2.57 434 residential Urban residential Low density 180,297 89,104 0.41 0.34 1,281 (0-4 DU/ac) Medium 262,778 103,259 0.20 0.23 1,405 density (4-9 DU/ac) Hioh density 392,090 235,618 0.39 1.43 531 (>9 DU/ac) Commercial 429,056 424,488 0.97 1.81 834 Industrial 253,988 100,584 2.08 4.01 190 Notes: Statistics based on privately owned land. DU = dwelling units. The mean statistics for land value refer to a simple arithmetic mean by parcel of per-acre price of land.
III. THE EMPIRICAL FRAMEWORK
This article focuses on the interplay between spatial externalities and institutional controls in directing land development. Land use is assumed to be driven by changes in land values, which are influenced by spatial externalities of land use (neighborhood effects). Location and the surrounding landscape play a prominent role in our approach. In particular, a zoning regulation may affect the value of a parcel both directly by restricting its use and indirectly by affecting land use in its surrounding areas.6
An exploratory empirical framework is developed to simulate the effects of land use regulations on land values within the study area. The framework has two major components (Figure 1). The first component is a set of land price equations estimated to predict land values for eight land use classes in the study area (low, medium, and high-density residential, commercial, industrial, agricultural, forest, and rural residential). The second component is a simulation model, which predicts land use choice and land value at each parcel on the landscape under alternative regulatory scenarios.
The study area was delineated with the aim of including land in a variety of uses, located in close proximity of human settlements. The land use and land cover map for year 2000 was derived from satellite images, aerial orthophoto maps, and tax lot maps.7 Land use is classified as low-, medium-, and high-density urban residential (0-4, 4-9, >9 dwelling units per acre), commercial (retail and office establishments), industrial (warehouses, assembly plants, R&D), rural residential, agricultural (cropland, pasture-land, wood/nurseries), and forest land. The data sample used for econometric estimation includes privately owned land that is currently in one of the eight land use classes, totaling some 13,000 parcels. This excludes land that is publicly owned or used for other purposes (e.g., roads, water bodies, sand/gravel). The same sample is then used for conducting simulations.
A. The Econometric Model
We estimate a set of land price equations by regressing parcel-level land prices on a vector of socioeconomic, location, and neighborhood characteristics of land parcels. Spatial inter-dependencies between parcels are assumed to take two forms in these land price equations. First, each land price equation is specified as a function of variables summarizing the spatial information. These variables include location characteristics such as the distance of the parcel to certain natural or man-made features (e.g.. the city center and highway) and neighborhood characteristics such as the proportion of land in different uses in the neighborhood. (8) Second, the land price equations arc specified as spatial error models (Anselin 2002) to explicitly deal with the spatial dependency between unobserved variables affecting various land uses. Formally, each land price equation is specified as
(1) y = X[beta] + [mu] and [mu] = [lambda]W[mu] + [epsilon]
where y is a vector of land prices or their log transformation. X is a matrix of observations of explanatory variables or their log transformation, [beta] is the parameter vector, [mu] is a vector of the autoregressive error terms, [lambda] is the spatial autoregressive coefficient, W is the n x n row-standardized spatial weights matrix, and [epsilon] is a vector of error terms, which are independently normally distributed with a mean of zero. Spatial autocorrelation may arise in the land price equations as a consequence of omitted variables. It is likely that parcels located near each other are affected by the same omitted variables, leading to spatial autocorrelation (e.g., Bock-stael 1996). The spatial error specification has been widely used in previous hedonic studies of land values (e.g., Bell and Bockstael 2000; Irwin and Bockstael 2001: Paterson and Bovle 2002; Plantinga, Lubowski, and Stavins 2002; Wu, Adams, and Plantinga 2004).
To estimate the land price equations, it is necessary to define the "neighborhood" for calculating the variables characterizing land use in the neighborhood and for determining the spatial weight matrix W. Specification of neighborhoods has largely been arbitrary in previous studies. One exception is Dale-Johnson and Brzeski (2001) who use semivariograrns to determine the size of neighborhood.9 In this article, kriging was used to estimate a semi-variogram of land values for each of the eight selected land use classes. We concluded that a buffer of 100 m is appropriate for the urban residential parcels. Hence, it is the immediate vicinity that matters most for urban residential lands. In contrast, a wider radius of 500 m is required to capture the spatial dependence among the parcels of the remaining land use classes. The buffer distances were used to calculate two sets of neighborhood variables (one for each buffer distance) for each parcel. Similarly, a corresponding row-standardized spatial weight matrix was generated for each land use class.
The dependent variable used to estimate the hedonic price equations is the assessed "real market value" (RMV) of land, excluding the value of any improvements and built property located on the parcel. The parcel-level RMV data were derived from tax lot maps from the county assessor's office. It is important to emphasize that Oregon is different from many states in that assessors are required by law to estimate RMVs, defined as the amount a parcel would sell for in a transaction between a willing buyer and a willing seller. RMV is separate from the assessed value for computing property taxes, which is referred to as the "maximum assessed value" (MAV) in Oregon. According to Oregon law, assessor must appraise all property at 100% of its market value. A county assessor initially appraises the property using a physical inspection and a comparison of market data from similar properties. During the ensuing tax years, the county assessor updates the value according to trends of similar properties. Assessor estimates of RMV have been found to be quite accurate in Oregon. For example, in a recent study. Grout (2010) compared assessor RMV estimates with sales prices in Portland, Oregon, and concluded that assessor estimates of RMVs provide accurate estimates of sales prices.(10)
The independent variables in each land price equation are selected based on economic theory and an extensive review of hedonic studies on land prices. A detailed description of the specification of the land price equations and the summary statistics of the explanatory variables are presented in Appendix SI, available as Supporting Information.
Median household income (MHINC) and population density in the neighborhood (NPOP-DEN) were derived from the 2000 U.S. Population Census block-level data. Dummy variables were constructed to characterize the parcel with respect to its ocation within the UGB (INJJGB), the type of forest (open forest, mostly oak; closed forest, mostly conifer; and mixed forest), and the soil capability classes (CLNIRR). The USDA Soil Conservation Service classifies the suitability of land for agricultural production using the CLNIRR soil capability index. The index rates soil for non-irrigated agricultural use and ranges from 1 to 8 indicating progressively greater limitations and narrower choices for use.
The distance and neighborhood measures were computed using Python programming language (van Rossum and Drake 2001) and Python scripts for the ArcGIS geoprocessing tools (Environmental Systems Research Institute [ESRI] 2004). The distance measures were calculated as the shortest straight line (Euclidian) distance between the parcel centroid and the nearest edge of a feature of interest, such as the central business district (DIST_CBD), urban growth boundary (DISTJJGB), riverfront (DIS-T_RIVER), lakefront (DISTJLAKE), or major highway (DISTJtfWY).
The neighborhood measures characterize the land use in the surrounding area. They measure the percentage of the surrounding landscape that is in low-, medium-, and high-density urban residential (LD, MD, HD), commercial (CO), industrial (IN), rural residential (RR), agricultural (AG), and forest use (FO). The remaining area (OTHER), not accounted for by the above eight categories, includes land in civic use, vacant land, wetlands, grassland, shrub-lands, sand and gravel, roads, and permanent water. The neighborhood is defined as the surrounding landscape within a 100-m or 500-m buffer around the parcel boundary.
We compared the spatial error model with two alternative specifications--spatial lag models and ordinary least squares (OLS) using both the linear and log-linear functional forms. Table 2 reports the values of [R.sup.2] for the estimated models. Although the signs of estimated coefficients are fairly robust to alternative specifications of error structure (OLS vs. spatial error vs. spatial lag), considering the spatial structure of the error terms yields significant improvements in the goodness of lit. For example, the value of [R.sup.2] for the three OLS models of residential land values is about 0.40 on average, but 0.45 for the spatial lag models, and 0.75 for the spatial error models. The spatial error models not only perform better than the spatial lag models in terms of fit, but they are also more amenable to the simulations described in the next section. More specifically, the implementation of the spatial lag models in the simulations would require calculation of an inverse of the spatial weight matrix and an iterative fitting of the predicted values at each time step, which is computationally prohibitive. The log-linear functional form is commonly used to estimate hedonic price equations. One of the major advantages of this specification is that it assures predicted land prices to be nonnegative. For these reasons, we use spatial error models with a log-linear functional form to estimate the hedonic price equations.
TABLE 2 A Comparison of OLS. Spatial Lag, and Spatial Error Models with Linear and Log-linear Specifications by [R.sub.2] In(RMVac) KM Vac OLS Spatial Spatial OLS Spatial Spatial Lag Error Lag Error LD 0.52 0.53 0.80 0.40 0.76 0.79 MD 0.33 0.38 0.73 0.29 0.76 0.78 HD 0.43 0.45 0.71 0.40 0.68 0.75 CO 0.26 0.27 0.36 0.15 0.16 0.17 IN 0.25 0.26 0.29 0.31 0.32 0.33 RR. 0.85 -- 0.23 0.56 -- 0.43 AG, FO
Table 3 presents the coefficients for the spatial error log-linear models of the eight land price equations estimated using maximum likelihood (Anselin 2003). A comparison between observed and predicted land values for all parcels in our sample indicates that the spatial error models generally predict well except for some parcels in rural land. For example, for low-density residential, medium-density residential, and industrial parcels, the difference between observed and predicted land values is less than 5% on 91%, 95%, and 97% of the parcels in the three land use categories, respectively. For high-density residential and commercial parcels, the difference is less than 5% on about 80% of the parcels. The standard errors of regression for the eight land price equations listed in Table 3 are, respectively 0.23, 0.20, 0.37, 0.67, 0.32, 1.27, 1.27, and 1.27, which are 1/62 to 1/8 of the corresponding dependent variables.
TABLE 3 Estimates of the Land Price Functions with Spatial Error Specification Variable LD MD HD CO INTERCEPT 11.f444*** 11.9857*** 12.5417*** 11.4091 *** AREA -0.5914*** -0.3187*** -0.1421*** -0.0416*** MHINC 9.31E-06*** 6.25E-06*** 1.03E-05*** NPOPDEN 0.0029* 0.0032*** -0.0062*** 0.0645*** DIST_HWY J.08E-04*** 7.73E-05*** 3.55E-04*** -5.38E-04*** DIST_RIVER 1.63E-04*** 1.90E-04*** 1.70E-04*** DIST_LAKE -1.94E-04*** -1.57E-04*** -2.48F.-04*** DIST_CBD -8.40E-05*** -3.37E-05*** -8.30E-05*** -1.26E-04** DIST_UGB IN_UGB ZONED_RR %FO 0.0062*** 0.0036** 0.0013 0.0225 %AG 0.0026** -1.43E-04 -0.0779*** 0.0262* %RR -0.0067*"* -0.0012 0.0744 %LD %MD 0.0014** 0.0018*** 0.0018 0.0016 %HD -7.67E-04 -0.0059*** 0.0014 -0.0534*** %CO -8.88E-04 6.60E-04 -0.0031 0.0389*** %IN -0.0095 0.0064* 0.0054 0.0207 %OTHER 9.I5E-04 0.0024*** -0.0024 0.0121 CLNIRR1 CLNIRR2 CLNIRR3 CLNIRR4 OPEN MIXED LAMBDA 0.7714*** 0.8681*** 0.7631 *** 0.7077*** R-sq 0.8025 0.7340 0.7106 0.3641 Obs. 3,102 6.906 1.083 765 Variable IN RR AG FO INTERCEPT 9.7750*** 17.7741** 8.7266** 8.9143** AREA -0.0510* -0.2289*** -0.0096*** -0.0321*** MHINC 1.71 E-05 NPOPDEN -0.0179 -0.3385 -0.0547 -0.0536 DIST_HWY 1.85E-04 -1.03E-04 -4.14E-05 -1.11 E-04 DIST_RIVER 0.0010 DIST_LAKE 2.04E-04 DIST_CBD -3.80E-06 DIST_UGB 4.20E-04 2.25 E-04 5.16E-04* IN_UGB 0.3386 2.0114*** 1.7493*** ZONED_RR 0.3278 0.4635 %FO -0.0294 -0.0909 %AG 0.0575** -0.0420 %RR 0.0852 0.0059 %LD %MD 0.0203 -0.0029 %HD 0.0302 -0.2408 %CO 0.0404* -0.0361 %IN 0.0068 -0.1184 %OTHER 0.0319* -0.0839 CLNIRR1 0.7650 CLNIRR2 0.6525 CLNIRR3 0.5455 CLNIRR4 0.9053 OPEN -0.4818 MIXED -0.0197 LAMBDA -0.7562*** 0.6170*** 0.6170"* 0.6170*** R-sq 0.2999 0.5441 0.5441 0.5441 Obs. 73 392 392 392 Notes: The dependent variable is the natural log of market value of land per acre. The radius of neighborhood is 100 m for the LD, MD, and HD models. A 500 m radius was used for the remaining models. Low-density residential land (%LD) serves as the reference land use category. *. **. ***Statistical sjgnjficance at the 10%, 5%, and 1% levels, respectively.
The first three columns of Table 3 present the estimated coefficients for the low-, medium-, and high-density urban residential equations, respectively. In all three cases, the coefficients of parcel size (AREA) are negative and significant at the 1% level, suggesting that per-acre price of land decreases with parcel size. This is a common finding in the hedonic pricing literature (e.g., Bockstael 1996; Palmquist and Danielson 1989); it is frequently explained by the existence of subdivision costs.11 The coefficient of MHINC is positive and significant at the 1% level, suggesting that there is a land-value premium associated with high-income neighborhoods (e.g., Geoghegan, Wainger, and Bockstael 1997; Irwin 2002; Irwin and Bockstael 2001). The effect of predicted neighborhood population density differs by housing density class: increases in population density have a positive effect on the value of low-and medium-density residential land, but have a negative effect on the value of high-density residential land. The values of residential land, regardless of density, decrease as one moves farther away from the CBD and a lakefront, but increase as one moves farther away from a highway. There is extensive evidence in the hedonic pricing literature that proximity to downtown and amenities increases land values (e.g., Bockstael 1996; Geoghegan, Wainger, and Bockstael 1997; Wu, Adams, and Plantinga 2004).
The neighborhood land use mix variables completely describe the surrounding landscape and sum to 100%. Low-density residential land is used as the reference category and is excluded from the regression to avoid perfect multi-collinearity. The estimated coefficients of the neighborhood measures differ by the housing density class. Overall, the results suggest that higher proportions of medium-density residential land (%MD) and forestland (%FO) in the neighborhood increase the land value. Although the presence of forest land in the neighborhood is frequently found to have a positive effect on land values (e.g.. Bockstael 1996; Irwin 2002), there is mixed evidence as for the effect of residential land (e.g., Geoghegan. Wainger, and Bockstael 1997; Irwin 2002).(12) Our results show that a large proportion of agricultural land in a neighborhood has a positive effect on the value of low-density residential land, but has a negative effect on the value of high-density residential land. This result may reflect that the demand for high-density housing in areas surrounded by agricultural land is low relative to demand for low-density housing.
The fourth and fifth columns of Table 3 show estimates of the commercial and industrial land price equations. In both cases, per-acre price of land decreases with parcel size (AREA). Population density (NPOPDEN) has different impacts on the value of industrial and commercial lands. Locations in neighborhoods with higher population densities are found to be more valuable as commercial sites. This may be because those locations are surrounded by a larger pool of potential customers. Proximity to urban centers and highways has a positive and significant effect on the value of commercial property. It has a statistically insignificant effect on the value of industrial property. Previous studies find that the location and size of a parcel are the most important variables determining the value of industrial property (e.g., Kowalski and Col-well 1986; Lockwood and Rutherford 1996).
The presence of large acreage of agricultural land in the surrounding area is capitalized into the value of commercial and industrial lands. Undeveloped land in the neighborhood not only provides opportunities for future expansion, but also the amenities for employees working in the commercial and industrial facilities. Results for forest land, however, are inconclusive, likely because of the fact that most forest-land in the study area is undevelopable either because of its unfavorable location (located in riparian areas or on river islands) or is zoned as protected forestland. Finally, a high concentration of commercial and industrial land (%CO, %IN) in the neighborhood has a positive effect on commercial and industrial land values, indicating the presence of an agglomeration effect. Granger and Blomquist (1999) found that although agglomeration and scale economies remain the principal determinants of location decisions of manufacturing establishments, amenities also influence manufacturing location in urban areas.
The last three columns of Table 3 show the estimated coefficients of the rural residential, agricultural, and forest land price equations. As expected, location within the urban growth boundary (IN_UGB) significantly increases the value of agricultural and forest parcels, reflecting the growth premiums and option values associated with development potential. In all the three equations, the coefficients of parcel size, distance to highway, and population density are negative. These results are consistent with previous studies of rural lands (e.g., Bockstael 1996: Hardie and Parks 1997).
B. The Simulation Model
The second major component of the empirical framework is a simulation model, which predicts land use and land prices that would have existed at each parcel if some or all land use regulations had never been imposed. The sample of 13,000 privately owned parcels is used for conducting the simulations. The remaining land within the study area, which is publicly owned or is used for other purposes (e.g., roads, water bodies, sand/gravel) stays invariant throughout the simulations. The simulations are conducted in a sequential manner both spatially and temporally. We first discuss the spatial dimension, then turn to the temporal dimension, and finally define a set of indicators to help us assess the land use changes.
The Spatial Dimension. There are two alternative approaches to simulate land use on a landscape. One approach assigns land use to all parcels in the landscape at the same time. However, because land value of a parcel can be affected by land uses on surrounding parcels, which may in turn be affected by their surrounding land use, this approach would require selecting land uses on a matrix of 13,000 parcels simultaneously, which is computationally prohibitive. An alternative approach is to assign land use to a single parcel at a time, considering the surrounding land, and proceed from parcel to parcel in a pre-defined order, as has been done in this article.
It has long been argued that agglomeration economies play a key role in land development. In the classic urban economics models, parcels close to city centers are developed first because transportation costs are lower to the city center. Thus, in this article, land parcels are processed according to their distance to the CBD. Land use choice at the parcel located nearest to the CBD is considered first. The spatially variant neighborhood measures (including %LD, %MD, %HD, %CO, %IN, %AG, %FO) are computed based on the land uses in the neighborhood at the time the parcel is processed.(13) The values of the parcel when put to different uses are then estimated using the land price equations. The land use that yields the highest value is selected for the parcel unless constrained by a land use regulation. This process continues on a parcel-by-parcel basis, outwards from the city center, until the last (most distant) parcel is processed. If the neighborhood includes a parcel whose land use has not yet been simulated, optimal land use from the previous period is assumed for that parcel (further details are discussed below). The most important feature of this approach is that each parcel's optimal land use is chosen while taking into account the surrounding land use, because the measures of surrounding land use are updated before simulating each individual parcel.
The Temporal Dimension. To simulate development patterns and land prices in year 2000 that would have existed on the landscape if a regulation had never been imposed, it is necessary to trace the evolution of land use patterns over lime because land use choice is both spatially and temporally variant. Land use choice is also temporally variant because current land use patterns are shaped by previous land use decisions because of land use irreversibility and spatial externalities.
This study simulates land use choice in 1960, 1970, 1980/1990, and 2000 on each parcel in the study region based on historical income and population data and urban boundaries. Simulations are conducted for every 10 years because of data and computational constraints.(14) Multiple rounds of simulations are conducted to give the simulation exercise a distinct temporal dimension. Comparing land use patterns and values in the last round (2000) with and without a regulation allows us to assess the impact of the regulation.(15)
Land use changes are simulated subject to a set of irreversibility constraints. The constraints prohibit conversions of land to lower-intensity use, such as conversion of developed land to resource use (agriculture or forest), conversion of urban land to rural land, or conversion of medium-density to low-density housing. Land in high-density residential, industrial, or commercial use is only convertible within these three uses.
Before the first round of simulation can begin, an "initial" land use must be first assigned to each parcel in the GIS-based landscape. In this article, the landscape is "initialized" in the following way: The delineation of the ca. 1950 city limit is derived from Loy et al. (1976) and Institute for a Sustainable Environment (1999a, 1999b). Medium-density residential use is attributed to areas in the historic downtown south of the Willamette River. The remaining areas within the 1950 city limits are attributed to low-density residential use. Land capability index is used to attribute agricultural (CLNIR = 1 to 4) and forest use (CLNIR > 5) in rural areas.
Placing the simulations in the historical context is useful and necessary because significant demographic and institutional changes occurred in the study region during the period. The total metropolitan population quadrupled in the study area during the post-World War II period. In 1948, Lane County, the location of the study area, began zoning and requiring building permits. A series of land use laws resulted in the establishment of the Oregon's statewide land use planning system in 1973. And finally, the comprehensive land use plan for Lane county and the city of Eugene was developed in 1977 (Helm 1984; Knaap and Nelson 1992: Jackson and Kimerling 2003).
C. Assessment Indicators
The impacts of a zoning regulation are evaluated by comparing land prices that would have existed when the regulation had never been imposed (the baseline) with those when the zoning regulation had been in place. Two baselines are considered: First, the no-regulation baseline in which land use allocation is not constrained by any land use regulation. Second, the all-but-the-selected-regulation baseline in which all zoning regulations are imposed except the land use regulation that is being evaluated. Land use regulations are typically designed to work in concert with each other in a complementary fashion.(16) A challenge in evaluating the relative impact of land use regulations is to decouple the effect of a given regulation from the impact of the other regulations. Land use regulations may be endogenously determined with land use changes. This article, however, does not focus on the causes of land use regulations; it simply evaluates what would happen to land use patterns and prices if some of the regulations had never been imposed. The cost of a selected regulation (CR) to a property is evaluated relative to the two baselines as
(2) [[CR.sub.i].sup.l]=[[V.sub.i].sup.no regulation] - [[V.sub.i].sup.regulation z]
(3) [[CR.sub.i].sup.2]=[[V.sub.i].sup.all zoing except z] - [[V.sub.i].sup.all zoing]
The cost of regulation for parcel i is calculated as the difference between the land value in the no-regulation baseline [[V.sub.i].sup.no regulation] and the land value under a given zoning regulation [[V.sub.i].sup.regulation z] (Equation ). Alternatively, it is calculated as the difference between the land value in the all-but-the-selected-regulation baseline [[V.sub.i].sup.all zoing except z] and the land value under the existing land use regulations [[V.sub.i].sup.all zoing] (Equation ). On the basis of historical records of land use regulations, zoning regulations affecting rural land uses are imposed at the third round of simulation (1980) (17) and onwards. The remaining regulations are imposed from the start of the simulation (I960).(18)
Similarly, the VIE with respect to a selected regulation z is evaluated relative to the two baselines as
(4) [[VIE.sub.i].sup.l] = [[VIE.sub.i].sup.only i exempted from z] - [[VIE.sub.i].sup.regulation z]
(5) [[VIE.sub.i].sup.2] = [[VIE.sub.i].sup.only i exempted from z] - [[VIE.sub.i].all zoing]
Under the no-regulation baseline, the value of an individual exemption for parcel i is calculated as the difference between the value if only parcel i is exempted from the regutlation [[V.sub.i].sup.only i exempted from z] and the land value under a given zoning regulation [[V.sub.i].sup. regutlation z].(Equation ) Alternatively, under the ail-but-the-selected-regulation baseline, the value of an individual exemption is calculated as the difference between the value if only parcel i is exempted from z (all other regulations still in place) [[V.sub.i].sup.only i exempted from z only] and the land value under all the existing land use regulations [[V.sub.i].sup.all zoing] (Equation ).
To implement the simulations, a computer program was written in Python programming language and executed in Python Win (Hammond 2001; van Rossum and Drake 2001). The program integrates ArcGIS geoprocessing tools (ESRI 2004) with the capacity to access and store the generated data in a geodatabase.
D. Model Limitations, Validation and Sensitivity Analysis
One limitation of the framework is related to the prediction of past land use using the estimated land price equations. The potential of the estimated hedonic coefficients as reasonably accurate measures of people's relative preferences (e.g., for amenities vs. accessibility) decreases with time because people's preferences may change over time. Also, if zoning reduces the total supply of a certain type of land, say, land for residential development, it would raise the price of land zoned for residential development, and the hedonic price equation parameters for residential land would be dependent on the regulatory regime. However, it is important to point out that although Oregon's restrictive land use planning regime confines development within designated UGBs, Oregon law also requires fast-growing cities, cities with populations over 25,000, and metropolitan service districts to include enough buildable land for the next 20 years of residential growth within their UGBs. Perhaps because the law ensures the sufficient supply of land for residential development, previous studies have found that zoning ordinances in Oregon tend to affect the location of development, but not the total area of development (Wu and Cho 2007). Because our hedonic price equations include variables that capture the effect of surrounding land use on a parcel's value, zoning should not significantly directly affect hedonic equation parameters, although it may affect a parcel's value by changing its surrounding land use. In addition, the estimated land price equations may not reflect well the demand for land development in the early stages of simulation even if they include household income and population density in those years as explanatory variables. This is because the amount of land development in a region is also influenced by local infrastructure and economic geography, which are not modeled explicitly in this study.
To mitigate these limitations, two constraints are imposed in the simulation process. One constraint limits urban development within the urban boundary existing in a particular year. However, using boundaries existing in time when zoning regulations were in place to simulate land development that would have existed if no regulations were imposed would be questionable. As a compromise, urban boundaries in 1960, 1970, and 1980 were imposed in the simulation of land use patterns in those years (when land use regulations were minimal or latent), but such a constraint was not imposed in the simulation of land use patterns in the absence of land use regulations in 1990 and 2000. This decision reflects that there were few regulations that limited land use choices before 1977 when a comprehensive land use plan for the city of Eugene was developed.
The other constraint imposed in the simulation process involves placing a cap on the total acreages of commercial, industrial, high-and medium-density residential development, and rural residential use. The cap equals the total acreage of these land uses in our sample. There is empirical evidence that zoning regulations only affect the location of land development, but not the total acreage of land development (see, e.g., Wu and Cho 2007). Admittedly, even with these mitigation measures, the analysis is still exploratory in nature, and the quantitative results must be interpreted with caution.
Another limitation of this analysis is related to treating parcel boundaries as fixed. It is likely that the boundaries in 2000 have been shaped by the historical patterns in zoning (and hence may be endogenous). Unfortunately, there is currently no easy way around this problem.
As a partial validation of the empirical framework, we determine whether the model can make accurate in-sample predictions. The framework was used to predict land use choices at each parcel in a scenario in which all zoning regulations are imposed. This scenario comes closest to replicating the real-world conditions. Table 4 presents the actual and the simulated land acreages for the eight land use classes in 2000. The results suggest that the modeling framework performs well in allocating land among major uses (resource and rural uses, urban residential uses, and commercial and industrial uses); the subtotals for these three groups are close to 100% (column 4 in Table 4). However, performance of the model by land use class varies; the model performs well in allocating agricultural, rural residential, commercial, and industrial land (95%-119% of actual acreage), but it over predicts low-density residential acreage (195% of actual acreage) and under predicts forestland, and medium-and high-density residential acreage (39%-67% of actual acreage). The mismatch in allocating land among the urban residential classes may be a result of noncompliance with zoning regulations (e.g., there is a lot medium-density development in areas zoned for low-density development). It could also be driven by forces outside the study region such as population and income growth or large-scale land use regulations on land supply in areas surrounding the region.
TABLE 4 Model Validation: Simulated versus Actual Land Acreages Simulated Actual Land Use Class Acres Acres Ratio Forest 1,219 1,806 0.67 Agricultural 3,642 3,057 1.19 Rural residential 415 436 0.95 Subtotal 5,276 5,299 1.00 Low-density residential 2,506 1,282 1.95 Medium-density residential 549 1,405 0.39 High-density residential 251 559 0.45 Subtotal 3,306 3,246 1.02 Commercial 816 862 0.95 Industrial 201 190 1.06 Subtotal 1,017 1,052 0.97 Notes: The simulated acreage refers to the scenario when all land use regulations are imposed.
Sensitivity analysis was conducted to test the robustness of the results. A number of alternative designs have been tested by dropping the constraints or by adjusting the level of control (e.g., by imposing the urban boundary only in early stages of simulations or by imposing acreage caps on fewer land uses). The qualitative results are robust to the alternative designs. For example, in all cases, the estimated values of individual exemptions from resource protection zoning are generally much larger than the estimated costs of regulation. However, the quantitative results do change under the different designs. For example, simulated test runs suggest that, as expected, a higher degree of control slows the progression of urban development and yields a more gradual development pattern. At the same time, more control over urban development lessens the importance and the impact of resource protection zoning. Even in the absence of zoning, more control over urban development materializes in larger acreage of undeveloped land. Consequently, if zoning is imposed, the estimated cost of such regulation goes up with more control. If acreage caps are not used, high-intensity uses (HD, CO, IN) dominate the landscape in the no-regulation scenario. Such land use pattern may not correspond to the real-world situation. Absence of acreage caps also impacts the magnitude of the estimated land price changes.
Although results reported in this article are based on simulating parcels in the order of their distance to the CBD, other "ordering rules" could be imagined. For example, an alternative is to process parcels according to their distances to major highways and other transportation corridors. Adopting this approach yields a different no-regulation baseline. However, the estimated aggregate costs of regulation are very similar to those presented in this article. The major differences are that, compared to the no-regulation baseline, forest and industrial zonings yield net benefits overall. Similarly, residential zoning yields a net benefit compared to the all-but-the-selected-regulation baseline. The estimated costs of regulation versus values of individual exemptions suggest essentially the same findings as reported in this article.19
IV. IMPACTS OF LAND USE REGULATIONS
Table 5 presents the aggregate impact of selected land use regulations relative to the no-regulation baseline. (20) The results are decomposed by zoned and unzoned lands and by winners and losers. For farmland zoning, 65% of the parcels located within the EFU zones lose value. These losses represent the direct effect of farmland zoning. At the same time, about 36% of the parcels located outside the EFU zones gain value, and 55% of them are not affected. Hence, although EFU zoning has a negative direct effect on land values within the zones, it has a positive indirect effect on land values outside the zones via neighborhood effects. For example, low-density residential development at the urban-rural fringe may benefit from farmland preservation through open-space provision. Exclusive farmland zoning may increase commercial and industrial land values because it reduces the supply of land for commercial and industrial development. Results for forest-land zoning and UGB designation are similar to those of farmland zoning.
TABLE 5 Estimates of the Impacts of Land Use Regulations. Relative to the No-Regulation Baseline Number of Parcels Change in Land Value Total Total Gain Loss Zoning Gainers Losers Unaffected ($1,000) ($1,000) Farmland zoning Zoned land 1 400 212 84 27.453 Un/oned land 4.339 1.172 6.674 284.589 3.743 Foresiland zoning Zoned land 0 43 1 0 3.285 Un/oned 1.94 43 12.723 760 32 land UGB designation Zoned land 1 567 255 84 38.394 Un/oned 4.219 1.147 6.587 286.808 3.701 kind Farm, forest, and UCiB together Zoned land 1 586 241 84 40.148 Un/oned 4.284 1,135 6.524 288.184 3.682 land Residential density zoning Zoned land 1.568 7.903 1,215 8.928 510.927 Un/oned 927 623 426 205.429 253.381 land Commercia zoning Zoned land 326 446 0 110.450 153.912 Unzoned 1.326 5.525 5.361 6.312 417.399 land Industrial zoning Zoned land 1 94 7 6.455 45.508 Un/oned 2.751 3.702 6.446 292.178 450.221 land All zoning Zoned land 1.614 9.315 1.361 1110.000 753.053 Unzoned 338 1 0 40.223 5 land Gain/Loss Zoning Ratio Farmland 9.13 zoning Zoned land Un/oned land Foresiland 0.23 zoning Zoned land Un/oned land UGB 6.82 designation Zoned land Un/oned kind Farm, 6.58 forest, and UCiB together Zoned land Un/oned land Residential 0.28 density zoning Zoned land Un/oned land Commercia zoning 0.20 Zoned land Unzoned land Industrial 0.60 zoning Zoned land Un/oned land All zoning 0.12 Zoned land Unzoned land
Results for residential density zoning suggest that 15% of the parcels gain and 74% of the parcels lose land value inside the zoned district. Outside the zone, 47% of the parcels gain, 32% lose, and 21% unaffected, with total gains amounting to 81% of total losses. The negative impact on land values inside the zoned district is mostly caused by reductions in housing density resulting from the regulation. The impact on land outside of the zoned district is caused by relocation of unregulated land uses. For example, parcels that could have been developed for commercial and industrial uses lose value, whereas parcels that would otherwise not be developed for commercial and industrial uses gain value. On the balance, the total loss is greater than the total gain.
In contrast to farm and forest zonings, which generally reduce the value of land located inside the zoned areas hut increase the value of land outside the zones, commercial and industrial zonings reduce land values both inside and outside of the zones. This is because commercial and industrial zonings restrict land use both inside and outside the zoned areas. Outside the zoned areas, no commercial and industrial development is allowed. In contrast, outside the exclusive farm and forest use zones, both farm and forest uses are allowed, although other land uses are not allowed inside the zoned districts. Compared to the no-regulation scenario, commercial zoning creates multiple smaller commercial zones in contrast to a more concentrated pattern of commercial development in the no-regulation scenario, leading to a reduction in the agglomeration effect. Therefore, although total acreage remains largely unchanged, the regulation yields a net loss in land value overall.
Table 6 presents estimates of the aggregate impacts of selected land use regulations relative to the all-but-the-selected-regulation baseline. Overall, the results in Table 6 are qualitatively similar to those in Table 5. However, compared with the no-regulation scenario, the magnitude of the estimated changes in land values is smaller. This is because zoning regulations often work in concert; imposing a regulation on top of others will be less restrictive than imposing the regulation alone.
TABLE 6 Estimates of the Impacts of Land Use Regulations, Relative to the All-but-the-Selected-Regulation Baseline Number of Parcels Change in Land Value Total Total Gain Loss Zoning Gainers Losers Unaffected ($1,000) ($1,000) Farmland zoning Zoned land 0 5 623 0 982 Unzoned 79 123 12.185 128 246 land Forestland zoning Zoned land 0 18 27 0 1.494 EJMoaed 0 6 12.964 0 5.441 land UGB designation Zoned land 0 72 674 0 16.545 Unzoned 63 50 12.156 99 769 land Farm, foresl. and UGB together Zoned land 17 566 168 500 64.858 Unzoned 535 373 11.356 42.801 7.059 land Residential density zoning Zoned land 1.294 7.662 1.796 11.371 190.006 Un/.oned 741 405 862 107.225 13.736 land Commercial zoning Zoned land 583 161 40 95.460 7.167 Unzoned 1.081 2.098 9.052 18.397 7.035 land Industrial zoning Zoned land 14 109 16 3.317 13,025 Unzoned 611 134 12.131 15.966 116 land Gain/Loss Zoning Ratio Farmland 0.10 zoning Zoned land Unzoned land Forestland 0.00 zoning Zoned land EJMoaed land UGB 0.01 designation Zoned land Unzoned land Farm, 0.60 foresl. and UGB together Zoned land Unzoned land Residential 0.58 density zoning Zoned land Un/.oned land Commercial 8.02 zoning Zoned land Unzoned land Industrial 1.47 zoning Zoned land Unzoned land
For commercial zoning, we found a net increase in the value of land located outside as well as inside the zoned district. This is a rather counterintuitive result. Commercial zoning increases the value of land inside the zone because, compared to the all-but-commercial-zoning baseline, it concentrates otherwise scattered commercial establishments within the designated commercial zones, thus enhancing the values of neighboring commercial land (the agglomeration effect). The cost advantage from the agglomeration effect dominates the cost of displacement.
Overall, our results suggest that zoning not only affects the value of land within the zoned area but also the value of land outside of the zone. With the exception of commercial zoning, lands inside the zoned areas tend to lose value, whereas lands outside the zoned areas tend to gain value. Thus, benefits of zoning largely accrue to landowners whose land use choices are not restricted.
The simulated gain/loss ratios for all zoning regulations always improve over years. This suggests that there is a premium associated with land use adjustments in response to changes in the surrounding land use over time. For example, zoning to protect farmland not only excludes any development within the zone; it may also have an indirect effect on the land outside the zoned district. Over time, the presence of open-space amenities may alter the structure of residential neighborhoods. Thus, a zoning regulation not only affects land values, but also affects the spatial pattern of land development both inside and outside of the designated zones.
V. THE VALUE OF INDIVIDUAL EXEMPTION
Table 7 reports estimates of the VIE and the cost of land use regulation (CR) relative to the no-regulation baseline. Only those parcels whose values are reduced by a regulation are used in the estimation. Parcels which grandfathered an exemption from a land use regulation were not used in the estimations.
TABLE 7 The Value of Individual Exemption (VIE) versus the Cost of Regulation (CR), Relative to the No-Regulation Baseline Change in Average Land Value ($/Acre) Change in Total Value (Million $) Change in Average Land Change in Total Value ($/Acre) Value (Million $) Zoned Land Unzoncd Land Zoned Land Zoning VIE CR VIE CR VIE Farmland 104,855 32,614 -133,140 87.99 zoning Forestland 4,962 4,630 -1,552 3.52 zoning UGB 88.717 33,413 -156.116 101.72 designation Farm, 58,273 22,075 -156.067 105.76 forest, and UGB together Residential 165,807 185.805 20,161 447.97 density zoning Commercial 18,153 54,771 109,877 170,621 14.41 zoning Industrial 39,034 323,499 141,856 61.277 4.71 zoning All zoning 219,076 122,149 74,394 -182,754 1.153.32 Unzoned Land Zoning CR VIE CR Farmland 27.37 -280.85 zoning Forestland 3.28 -0.73 zoning UGB 38.31 -283.11 designation Farm, 40.06 -284.50 forest, and UGB together Resident!a] 502.00 47.95 density zoning Commercial 43.46 264.73 411.09 zoning Industrial 39.05 365.86 158.04 zoning All zoning 643.05 3.20 -40.22 Notes: The reported values are average changes in land values ($/acre) and changes in total land value of the study area (million $).
The results indicate that the value of an individual exemption from farmland zoning is much larger than the cost of the regulation. Results for the UGB designation suggest a similar pattern. These results suggest that landowners seeking exemption under Measure 37 tend to overstate the cost of regulation because an estimate obtained using the standard appraisal methods yields the value of an individual exemption rather than the cost of regulation. The estimates of cost of regulation are negative for unzoned land located outside the zoned district suggesting that the three resource protection zoning regulations increase the value of unzoned lands. Hence, although the resource protection policies may reduce the value of the regulated lands, these policies increase the value of unregulated lands.
Residential density zoning is found to reduce the total value of the affected parcels inside the zone. However, once residential density zoning is instituted, an exemption is valued at less than the cost of the regulation. The cost of this regulation is higher because, relative to the no-regulation baseline, it tends to displace commercial and high-density residential uses in favor of low-density housing. However, once the regulation is in place, and commercial and high-density development is forced to other locations, the value of an individual exemption will be reduced because the potential for agglomeration effects in commercial use is reduced.
Relative to the no-regulation baseline, commercial zoning changes the location of commercial land use by creating several smaller commercial areas. Landowners inside a commercial zone suffer a loss. However, once commercial zoning is instituted, an exemption may not remove all the losses. This occurs because commercial use outside the commercial zones is not allowed and the potential gains from agglomeration are eliminated. An exemption to use land for commercial purposes commands a premium only when the neighbors do commerce as well.
Compared to the no-regulation baseline, the estimated values of individual exemption are also lower than the cost of regulation for industrial zoning within the zoned area. In a situation when the total acreage of unregulated uses remains largely unchanged, this phenomenon can be attributed to the agglomeration effect and the displacement effect (i.e., a reduction in land value due to relocation to an inferior location). Although the cost of a regulation in an urban area may be large, once the regulation is in place and changes in the spatial pattern of land use have materialized, the temptation to ask for an exemption is reduced. The value of an individual exemption goes down precisely because of the changes in the surrounding land use that have occurred as a result of the regulation.
Table 8 shows estimates of the VIE and the cost of regulation relative to the all-but-the-selected-regulation baseline. When farmland or forestland zoning is imposed on top of an enforced UGB, the value of an individual exemption is about the same as the cost of regulation. This is because dropping any one of these regulations would not change the spatial pattern of development. Land use is still subject to other regulations. For example, an exemption from farmland zoning carries no premium above the cost of regulation when the land is located outside an enforced UGB. It is only when these regulations are taken together that an exemption carries a premium. The farmland and forestland zonings and the UGB together impose a cost to landowners located inside the zoned areas. Outside the zone, the regulations actually increase land values. These findings are consistent with those reported in Table 7.
TABLE 8 The Value of Individual Exemption (VIE) versus the Cost of Regulation (CR). Relative to the All-but-ihe-Selected-Reguiation Baseline Change in Average Land Value Change in ($/Acre) Total Value (Million $) Zoned Land Unzoned Land Zoned Land VIE CR VIE CR VIE CR Farmland 70.055 70.145 862 0.98 0.98 zoning Fnreslland 2.258 2.258 91.317 1.49 1.49 zoning UGB 86.546 81.211 5.090 15.43 16.55 designation Farm, 46.231 29.485 -56.599 96.13 64.36 forest, and UGB together Residcnlial 134.210 68.925 -77.401 332.93 178.64 density zoning Commercial 64.693 -123.546 74.394 -8.863 1.93 88.29 zoning Industrial 55.327 27.238 0 -33.375 8.52 9.71 zoning Unzoned Land VIE CR Farmland 0.12 zoning Fnreslland 5.44 zoning UGB 0.67 designation Farm, -35.74 forest, and UGB together Residcnlial 93.49 density zoning Commercial 3.20 -11.36 zoning Industrial 3.20 -15.85 zoning Notes; The reported values arc average changes in land values (S/acre) and changes in total land value of the sludy area (million $;).
When all other regulations are in place, commercial zoning increases the total land value both inside and outside of the commercial zone. This suggests that when potential for agglomeration effects exists, a well-planned land use zoning can enhance agglomeration advantages and increase the land values. However, some landowners would still benefit from an exemption from commercial zoning. The costs of residential and industrial zonings are much smaller than the corresponding costs in Table 7. The difference arises because imposing a regulation on top of all other regulations is less costly than imposing the regulation alone given that land uses are already restricted by other regulations.
In the absence of land use regulation, individual landowners typically have little incentives to take into account the spatial externalities of their land use. This is where government intervention in land markets is warranted. Zoning ordinances and other forms of land use regulations aim to correct inefficient land use patterns by promoting provision of positive externalities and by limiting negative externalities. However, private property rights advocates contend that governments tend to overregulate because governments rarely bear the costs of regulation.
Our exploratory analysis suggests that governments indeed overregulate from the perspective of landowners. Although some land gained value and some lost value, the aggregate loss of land value is several times larger than the gain under the six regulations considered in this study. Only the HFU zoning and the UGB designation result in a net increase in total land value in the study region. Even for these two regulations, hundreds of parcels lose land value. However, this does not necessarily mean that the governments overregulate from the perspective of society because land use regulations may also generate public goods such as flood control, wildlife habitat, and water quality protection.
Compared with the no-regulation baseline, zoning ordinances regulating residential, commercial, and industrial development impose a much larger cost to landowners (as a whole) than the zoning ordinances aimed at protecting resource lands. However, an individual exemption from residential, commercial, and industrial zoning regulations does not carry a premium beyond the cost of regulation. In fact, the average value of an individual exemption is smaller than the average cost of those regulations because an exemption would result in a loss of the agglomeration effect. In contrast, the value of an individual exemption is much larger than the cost of regulation for farmland zoning and the UGB restrictions because these regulations increase the value of developed land.
When the all-but-the-selected-regulation scenario is used as a baseline, the value of an individual exemption from farmland zoning, forest zoning, or the UGB designation is about the same as the cost of the corresponding regulation. This is because imposing one of these regulations on top of all other regulations would not significantly change the spatial pattern of land use. Conversely, lifting one of the regulations at a time while keeping all other regulations in place would simply remove the cost of the regulation. In contrast, an individual exemption from the residential, commercial, and industrial zoning regulations carries a premium beyond the cost of the regulation.
The empirical framework was used to estimate the difference between the VIEs and the cost of land use regulations to illuminate the controversy surrounding Oregon's Measures 37 and 49. The measures are likely to have national ramifications because property rights advocates may sponsor similar measures in other states. Our empirical results show that all major zoning regulations in the study area have affected land use and property values both inside and outside the zoned districts. Although the value of a property can never go down when it is exempted from a binding regulation, the cost of regulation to a landowner can be positive or negative, depending on the location of the parcel. Even when the cost of regulation and the value of an individual exemption are both positive, their magnitude can be quite different. For most of the regulations contested in Oregon's Measure 37 claims, the cost of regulation to a landowner is considerably smaller than the value of an individual exemption.
This study develops a simulation model that provides an illustration of the interactions between spatial effects of land use and institutional controls. In addition to the limitations discussed in the modeling section, this analysis also does not consider the impact of large-scale land use restrictions on land supply outside of the study region--a factor that is likely to alter equilibrium land prices. Therefore, the estimates presented here are only exploratory and cannot be used for assessment of specific policies. To do so would require placing the analysis into a general equilibrium framework that covers the entire land market.
(1.) For more on Oregon's land use planning regimes, see Nelson (1992) and Knaap and Nelson (1992).
(2.) Land use regulation is particularly controversial in Oregon. According to a 2005 survey on land use issues, two in three Oregonians said that growth management has made the stale a better place to live (Oregon Land Use Statewide Survey 2005). Most respondents were concerned about the environment and favored public planning over market-based decisions. According to the poll. Oregonians want to protect land for future needs rather than develop it now. Yet, they also recognize a fundamental value in property rights. Two-thirds of respondents firmly believed in property rights protection and most valued individual rights more than responsibility to the community. The survey illustrates the appeal of land use regulation among Oregon residents. It also demonstrates the controversy surrounding it. Oregon residents want urban sprawl controlled but they also want to use their land as they see lit.
(3.) See http://www.0reg0n.g0v/I,CD/docs/measure37/maephersorL.opinion.pdf for the opinion issued by the Marion County Circuit Court in MacPherson v. Department of Administrative Services that holds that Measure 37 is unconstitutional.
(4.) The break down information is unavailable for claims filed alter October 25. 2005.
(5.) Goal 3 requires each county to adopt exclusive farm use ones by using US DA Soil Conservation Service land capability classes. West of the Cascade Mountains EFU zones include agricultural lands of classes I-IV; east of the Cascades they include classes I-VI (ODLCD 2005).
(6.) Several studies have focused on the mechanism of spatial interaction. For example. Irwin and Bockstael (2002) develop a model of land conversion that incorporates local spillover effects among spatially distributed agents. They find that fragmented patterns of land development in the rural-urban fringe could be explained by the negative externalities generated by the surrounding lands. Likewise, open-space designation may affect property values both directly and indirectly; it affects property values directly by reducing the total supply of developable land and indirectly by making certain areas more attractive, thereby changing the spatial patterns of demand within a given metropolitan area (Wu. Adams, and Plantinga 2004).
(7.) We are grateful to John Bolte. Pat Berger, Michael Guzy, and Frank Miller at Oregon State University for assistance with the land use/cover geospatial dataset.
(8.) For example, Wu. Adams, and Plantinga (2004) use location and amenity variables such as distance to CBD. public park, river, lake, wetland, proximity of the nearest industrial and commercial property, and distance to nearest public transportation. Irwin and Bockstael (2004) use neighborhood land-use variables representing the percentage of low-, medium-, and high-density residential land, commercial and industrial land, undeveloped land, and open space, Geoghegan. Wainger. and Bockstael (1997) use landscape pattern indices adopted from the ecological literature to measure diversity or fragmentation of the surrounding landscape.
(9.) A scmivariogram is a plot of semivariance values against the lag distance and is frequently used in geostatistics to describe the spatial correlation of observations. The distance at which the curve levels off is called range. Any two pairs of locations separated by distances closer than the range are spatially autocorrelated, whereas locations farther apart than the range are independent of each other. Consequently, the radius of the neighborhood is set equal to the range of the estimated semivariogram.
(10.) At the beginning of this research project, we polled county assessors concerning their assessment methodologies. We got the following response from the Lane County: "In all of the land class categories. Lane County uses a combination of methods. We value land based on its highest and best use as if vacant; all known influence factors are taken into consideration when a new land value is established."
(11.) One exception is Lin and Evans (2000), who find that land price per unit increases with lot size.
(12.) Instead of using shares of alternative land uses in the neighborhood, an alternative approach is to construct indexes to describe land use patterns in the neighborhood. For example, Geoghegan. Wainger. and Bockstael (1997) constructed a diversity index to measure the extent to which the landscape is dominated by a few or many land uses.
(13.) Variable %OTHER stays invariant because it measures land that is not subject to simulation.
(14.) Household income data are collected every ten years by the U.S. Census Bureau.
(15.) We start the simulations from I960 for several considerations. First, although Lane County started zoning ordinances in 1948, a Comprehensive Land Use Plan for Lane County was approved only in 1977. Second, land use patterns become increasingly complex over lime. The spatial structure of land use within the study area was relatively simple before 1960, with only a fraction of land allocated to development and the rest being in agriculture or forestry. The difficulty of reconstructing a landscape and the margin of error increases as we progress with time, as development becomes the dominant feature of the landscape. Consequently, setting the starting time earlier rather than later increases accuracy of capturing the initial landscape. However, setting the starting time earlier involves more simulations, which may lead to a large margin of error at 2000. As a balance, we start the simulations from 1960.
(16.) For example, Oregon's farmland preservation policies work as a package. Exclusive farm use zones preserve farmland for farming; UGBs limit urban sprawl; and exurban districts accommodate the demand for rural residential development. These and other land use policies work toward a common goal of protecting resource lands (Nelson 1992).
(17.) The Comprehensive Land Use Plan for Lane County was approved in 1977.
(18.) Lane County, (he location of the study area, began zoning and requiring building permits in 1948.
(19.) Yet another alternative would be to model land development in a "cherry-on-the-cake" manner. In this case, locations of development would be selected from among all available parcels, and land use would be allocated to the highest yielding parcels first. The downside of this approach is that it requires significantly more calculations; to complete one round of simulation of the entire landscape, the current approach needs to evaluate a single parcel only once, hence the number of calculations equals j, where j are the various land uses and N is the total number of parcels. The alternative approach would require evaluation of all parcels simultaneously, which is computationally prohibitive.
(20.) As discussed in Section III, the models generally predict well, particularly in terms of the number of parcels gaining or losing values. However, it is computationally prohibitive to calculate the standard errors for results in Tables 5 and 6 because they are a summary of results on hundreds of thousands of parcels.
Anselin, L. "Under the Hood. Issues in the Specification and Interpretation of Spatial Regression Models." Agricultural Economics, 27(3), 2002, 247-67.
--. GeoDa 0.9 User's Guide. Urbana-Champaign, IL: Spatial Analysis Laboratory. University of Illinois. 2003.
Beaton, P. W. "The Impact of Regional Land Use Controls on Property Values: The Case of the New Jersey Pmekmds."'Land Economics, 67(2), 1991, 172.
Beaton. P. W., and M. Pollock. ''Economic Impact of Growth Management Policies Surrounding the Chesapeake Bay." Land Economics, 68, 1992, 434.
Bell, K. P., and N. E. Bockslael. "Applying the Generalized-Moments Estimation Approach to Spatial Problems Involving Microlevel Data." Review of Economics and Statistics, 82(1), 2000, 72-82.
Bockstael, N. E. "Modeling Economics and Ecology: The Importance of a Spatial Perspective." American Journal of Agricultural Economics, 78, 1996, 1168-80.
Cheshire, P., and S. Sheppard. "The Welfare Economics of Land Use Planning." Journal of Urban Economics. 52. 2002, 242-69.
Cho, S. H.. J. J. Wu, and W. G. Boggess. "Measuring Interactions among Urbanization, Land Use, and Public Finance." American Journal of Agricultural Economics, 85(4), 2003, 988-99.
Congressional Budset Office (CBO). "Regulatory Takings and Proposals for Change." 1998.
Dale-Johnson, D., and W. J. Brzeski. 2001. "Spatial Regression Analysis of Commercial Land Price Gradients." A paper presented at the Asian Real Estate Society Sixth Annual Conference at Keio University, Mita Campus, Tokyo, Japan, July 31 to August 2, 2001.
Environmental Systems Research Institute (ESRl). What Is ArcGIS'l Rcdlands, CA: Environmental Systems Research Institute, 2004.
Fischel, W. A. Do Growth Controls Matter? A Review of Empirical Literature on the Effectiveness and Efficiency of Local Government Land Use Regulation. Cambridge. MA: Lincoln Institute of Land Policy, 1990.
Geoghegan, J., L. Wainger, and N. E. Bockstael. "Spatial Landscape Indices in a Hedonic Framework: An Ecological Economics Analysis Using GIS." Ecological Economics, 23(3), 1997/251-64.
Granger, M. D., and G. C. Blomquht. "Evaluating the Influence of Amenities on the Location of Manufacturing Establishments in Urban Areas." Urban Studies, 36(11), 1999, 859-73.
Green, R. K. "Land Use Regulation and the Price of Housing in a Suburban Wisconsin County." Journal of Housing Economics, 8(2), 1999, 144-59.
Grout, C. A. "Land Use Policy and Property Values." Dissertation, Department of Agricultural and Resource Economics, Oregon State University, 2010.
Hammond, M. Pythonwin-Python IDE and GUI Framework for Windows. 2001.
Hardie, L W., and P. J. Parks. "Land Use with Heterogeneous Land Quality: An Application of an Area Base Model." American Journal of Agricultural Economics, 79(May), 1997.299-310.
Helm, M. Eugene, Oregon: A Guide 1984-85 Edition. Eugene, OR: Rainy Day Press, 1984.
Henneberry. D. M., and R. L. Barrows. "Capitalization of Exclusive Agricultural Zoning into Farmland Prices." Land Economics, 66(3), 1990, 249-58.
Institute for a Sustainable Environment, University of Oregon (ISE). "1930 Population by Minor Civil Divisions. Spatial Dataset." Eugene, OR, March 1999a. Accessed June 10, 2005. http://oregonslate.edu/Dept/pnvv-erc
--. "1970 Population by Census County Divisions. Spatial Dataset." Eugene, OR. March 1999b. Accessed June 10, 2005. http://oregonstate.edu/Dept/pnw-erc
Irwin, E. G. "The Effects of Open Space on Residential Property Values." Land Economics, 78(4), 2002, 465-80.
Irwin, E. G., and N. E. Bockstael. "The Problem of Identifying Land Use Spillovers: Measuring the Effects of Open Space on Residential Properly Values." American Journal of Agricultural Economics, 83(3). 2001. 698-704.
--. "Interacting Agents, Spatial Externalities and the Evolution of Residential Land Use Patterns." Journal of Economic Geography, 2, 2002, 31-54.
--. "Land Use Externalities, Open Space Preservation, and Urban Sprawl." Regional Science and Urban Economics, 34. 2004, 705-25.
Jackson, P. L.. and A. J. Kimerling. eds. Atlas of the Pacific Northwest. 9th ed. Corvallis. OR.: Oregon State University Press, 2003.
Jim. M. J. "The Effects of Portland's Urban Growth Boundary on Urban Development Patterns and Commuting." Urban Studies, 41(7), 2004. 1333-48.
Katz. L., and K. T. Rosen. "The Interjurisdictional Effects of Growth Controls on Housing Prices." Journal of Law and Economics, 30, 1987. 149.
Kline. J., and R. Alig. "Does Land Use Planning Slow the Conversion of Forest and Agricultural Land?" Growth and Change..30(1) 1999,3-22.
Knaap. G., and A. C. Nelson. The Regulated Landscape: Lessons on State Land Use Planning from Oregon. Cambridge. MA: Lincoln Institute of Land Policy, 1992.
Kowalski. J. G.. and P. F. Cohwell. "Market versus Assessed Values of Industrial Land." Journal of the American Real Estate and Urban Economies Association. 14(2), 1986, 361-73.
Lin. T.-C. and A. W. Evans. "The Relationship between the Price of Land and Size of Plot When Plots Are Small." Land Economics, 76(3). 2000. 386-94.
Lockwood, L. J., and R. C. Rutherford. "Determinants of Industrial Properly Value." Real Estate Economics. 24(2), 1996,257-72.
Loy. W. G., S. Allan, C P. Pattom and R. D. Plank. Atlas of Oregon. Eugene. OR: University of Oregon, 1976.
Malpczzi. S., G, H. Chun, and R. K. Green. "New Ptaee-to-Place Housing Price Indexes for U.S. Metropolitan Areas and their Determinants: An Application of Urban Indicators." Real Estate Economics. 26(2), 1998, 235-74.
Nelson, A. C. "Preserving Prime Farmland in the Face of Urbanization: Lessons from Oregon." Journal of the American Planning Association, 58(4), 1992, 467-88.
Netusil, N. R. "The Effect of Environmental Zoning and Amenities on Property Values: Portland. Oregon." Land Economics. 81(2). 2005. 227-46.
Oregon Department of Land Conservation and Development (ODLCD). "Statewide Planning Goals and Guidelines:' 2005. Accessed April 28, 2005. http://www.lcd.state.or.us/LCD/goals.shtml
--. "Measure 37." 2007. Accessed March 25.
--. "Measure 49 Guide." 2008. Accessed August
4. 2008. http://www.oregon.gov/LCD/MEASURF49/docs/general/m49_guide.pdf
Oregon Land Use Statewide Survey. A report for Oregon Business Association and Institute of Portland Metropolitan Studies, Portland State University. Portland. OR: CFM Research, March 2005.
Palmquist, R. B., and L. E. Danielson. "A Hedonic Study of the Effects of Erosion Control and Drainage on Farmland Values." American Journal of Agricultural Economics. 71(1), 1989, 55-62.
Parsons. G. R. "The Effects of Coastal Land Use Restrictions on Housing Prices: A Repeat Sale Analysis." Journal of Environmental Economics and Management, 22. 1992. 25.
Paterson. R. W., and K. J. Boyle. "Out of Sight. Out of Mind? Using GIS to Incorporate Visibility in Hedonic Property Value Models." Loud Economics. 78(3). 2002, 417-25.
Plantinga, A. J. Measuring Compensation under Measure 37: An Economist's Perspective. Corvallis, OR: Department of Agricultural and Resource Economics, Oregon Stale University, 2004.
Plantinga. A. J.. R. N. Lubowski. and R. N. Stavins. "The Effects of Potential Land Development on Agricultural Land Prices." Journal of Urban Economics. 52. 2002, 561-81.
Pollakowski. H. O., and S. M. Wachier. "The Effects of Land Use Constraints on Housing Prices." Land Economics, 6601. 1990.315-24.
Quigley. J. M., and L. A. Rosenthal. "The Effects of Land Use Regulation on the Price of Housing: What Do We Know? What Can We Learn?" Berkeley Program on Housing and Urban Policy. Working Paper No. W04-002. 2004.
Schaeffer, P. V.. and C. A. Millerick. The Impact of Historic District Designation on Property Values: An Empirical Study." Economic Development Quarterly, 5.1991.301.
State of Oregon. Voters' Pamphlet: Oregon Special Election November 6, 2007. Salem. OR: State of Oregon. 2007.
--. "Measure 49 Guide." Department of Land Conservation and Development, Salem, OR. 2010. Accessed September 2. 2010. http://www.oregon.gov/LCD/MEASURE49/docs/general/m49"guide.pdf
Van Rossum. G., and F. L. Drake, Jr., eds. Python Reference Manual. Virginia: PythonLabs. 2001.
Walsh. R. "Endogenous Open Space Amenities in a Locational Equilibrium." Journal of Urban Economics, 61. 2007.319-44.
Wu. J. J., R. M. Adams, and A. J. Plantinga. "Amenities in an Urban Equilibrium Model: Residential Development in Portland. Oregon." Land Economics. 80(1), 2004. 19-32.
Wu, J. J., and S. H. Cho. "The Effect of Local Land Use Regulations on Urban Development in the Western United Slates." Regional Science and Urban Economics, 37. 2007. 69-86.
Additional Supporting Information may be found in the online version of this article:
Appendix S1. The Land Price Models.
Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.
* The views expressed here are the authors' own and do not necessarily reflect those of the OECD or its member countries. The senior authorship is shared.
Hascic: Economist, OECD Environment Directorate, 2 rue Andre Pascal, 75775 Paris Cedex 16, France. E-mail Ivan.Hascic@oecd.org
Wu: E. N. Castle Professor, Department of Agricultural and Resource Economics, Oregon State University, 213 Ballard Ext. Hall, CorvaUis, OR 97331; University Fellow. Resources for the Future. Washington DC; Chang Jiang Visiting Professor. Renmin University of China. Phone 541-737-3060, Fax 541-737-2563, E-mail JunJie.Wu@oregonstate.edu
CBD: Central Business District
DU: Dwelling Unit
EFU: Exclusive Farm Use
GIS: Geographic Information System
MAV: Maximum Assessed Value
MHINC: Median Household Income
NPOPDEN: Population Density in the Neighborhood
OLS: Ordinary Least Squares
RMV: Real Market Value
UGB: Urban Growth Boundary
VIE: Value of Individual Exemption
doi: 10.1111/j.1465-7287.201 1.00259.x
|Printer friendly Cite/link Email Feedback|
|Author:||Hascic, Ivan; Wu, Junjie|
|Publication:||Contemporary Economic Policy|
|Date:||Apr 1, 2012|
|Previous Article:||Active labor market programs and regional mobility of labor: evidence from the Swedish recession, 1994-1995.|
|Next Article:||The property value premium of a place of worship.|