Abstract
The paper concerns the cost of federal policies to protect endangered species in a legal system that gives primary authority over land use decisions to state and local governments. A conceptual model captures the essential features of federal interventions to conserve habitat for listed species, a combination of mitigation and avoidance requirements for greenfield land developments, and considers the effect of preexisting land use regulations. The analysis also demonstrates that the costs of habitat conservation could be reduced if local governments allowed for greater densification of new development in response to federal habitat conservation efforts. (JEL Q24, Q58)
I. Introduction
The Endangered Species Act (ESA) is the broadest and most powerful law aimed at the protection of endangered species and their habitats.1 It is also one of the nation's most controversial environmental laws, often pitting private property and development interests against federal wildlife agencies and environmental groups. The most frequent source of controversy is the perceived cost of the ESA, coupled with the view that the costs of the act are borne disproportionately by a few landowners who happen to have endangered species on their property. To provide some insight into these issues, this paper develops a framework for measuring the economic cost of endangered species regulation of housing development, and applies the framework to the case of California vernal pools, a type of ephemeral wetland that is habitat for a variety of endangered plants and animals (Witham et al. 1998).
The ESA affects private development through its prohibitions on the “take” of endangered species. Section 9 of the ESA defines take broadly to include any action that directly harms or harasses the species, including impairment of habitat in a way that interferes with breeding, feeding, or other essential behavior patterns.2 Section 10 of the ESA allows for exceptions to the take prohibition, most notably in cases where the developer has completed a Habitat Conservation Plan (HCP) that is intended to facilitate recovery of the species. HCPs allow development to proceed and take to occur if the proposed habitat changes are minimized to the “maximum extent practicable,” and if the take will not reduce the likelihood that the species will survive and recover.
The ESA also places significant restrictions on the actions of federal agencies. However, even these restrictions on agency actions can have important consequences for private land development. Section 7 of the ESA, which does not address private land development at all, requires that federal agencies refrain from actions that are likely to harm endangered species or the habitats on which they depend. Such actions cover a broad range of activities carried out by agencies themselves, including military preparedness (Department of Defense), water infrastructure (Department of the Interior or Army Corps of Engineers), timber management and grazing (Forest Service and Bureau of Land Management), and levee maintenance (Army Corps of Engineers). Federal agencies carrying out such projects with the potential to harm endangered species are required to consult with the U.S. Fish and Wildlife Service and the U.S. National Oceanic and Atmospheric Administration Fisheries Service, to ensure that these actions do not cause harm to species or their habitats.
Importantly, the ESA defines federal agency action to include the issuance of permits for projects carried out by private development interests. The most common type of federal permit requiring Section 7 consultation with wildlife agencies is a permit authorizing the discharge of dredge or fill material into waters of the United States. These permits are issued by the Army Corps of Engineers and are authorized under Section 404 of the Clean Water Act. Activities permitted under Section 404 include commercial development, pipeline and electric transmission, renewable energy projects, transportation infrastructure, and, of course, housing developments. Overall, the Corps issues roughly 90,000 discharge permits annually under Section 404 (Congressional Research Service 2012).
Taken together, the provisions of the ESA, particularly Sections 7 and 9, can have a powerful effect on private development projects. Indeed, the ESA has so much potential to affect private property that it has provoked heated debate and, occasionally, alarming newspaper headlines. There have been several examples of housing developments that have been adversely impacted by the requirements of the ESA. One prominent case involved the coastal California gnatcatcher, a small bluegray songbird with habitat in the coastal areas of southern California.
The Fish and Wildlife Service listed the gnatcatcher as threatened in 1993, a decision that imposed land use restrictions across a wide swath of southern California in the midst of one of the nation's greatest real estate booms. Approximately 200,000 acres in southern California were designated as critical habitat for the gnatcatcher.3 Many of the areas declared to be gnatcatcher habitat were highly developable, and several important housing and commercial development projects were stymied as a result of the listing.
The Fish and Wildlife Service itself estimated that protection of the gnatcatcher could result in significant restrictions on housing construction in the region, and result in large economic costs to landowners, developers, and the regional economy. Nonetheless, the ESA contains few mechanisms for adjusting protection of species in the face of large economic costs. Rather, in the landmark Tellico Dam case, the Supreme Court ruled that Congress intended for endangered species to be afforded the highest of priorities when implementing the ESA.4
Despite the rhetoric and controversy that sometimes surrounds federal intervention to protect endangered species, there is surprisingly little economic literature on this subject. Several papers have examined the potentially perverse incentives created by the ESA, pointing out instances in which the ESA's take prohibitions create incentives for landowners to destroy habitat, for example (see Innes 1997; Innes, Polasky, and Tschirhart 1998; Lueck and Michael 2003; List, Margolis, and Osgood 2006). There are fewer papers that directly address the economic costs of the ESA.
One paper that does examine a question closely related to ours is by Auffhammer, Oren, and Sunding (2012), who employ hedonic methods to measure the economic costs of habitat protection under the ESA. The federal government's announcement that certain lands are essential habitat for a listed species may signal to potential buyers that they may incur additional development costs. In a competitive land market, these additional development costs should be capitalized into the market price of land, creating differences in equilibrium sales prices between parcels with and without designated habitat. Looking at land sales in and around areas considered by the government to contain habitat for the redlegged frog and the Bay checkerspot butterfly, two prominent species in California, Auffhammer, Oren, and Sunding show that the presence of endangered species habitat results in a large and statistically significant reduction in the equilibrium price of vacant land.
This paper considers the economic costs of protecting endangered species through the regulation of housing projects. We begin by developing a partial equilibrium model of the costs of federal efforts to protect habitat by controlling the development of new housing. We are interested in the total cost of federal regulations, and also in the variation in cost per acre as a way to gauge the potential for improving the efficiency of federal land use regulation through targeting and differential treatment. A central feature of the analysis is that we examine the effect of federal land use controls in a legal system that reserves primary authority over land use decisions for local governments. That is, we consider the welfare cost of federal regulations when imposed on a backdrop of land and housing markets that are subject to preexisting local regulations.
We apply the conceptual framework to the case of California vernal pools, a type of seasonal wetland that provides habitat to over a dozen endangered plants and animals. Using detailed land use projections at the U.S. Census tract level combined with data on the cost of mitigation and a set of local housing market parameters, we measure the costs of land conservation and other mitigation measures across the vast acreage regulated by the federal government due to the presence of vernal pools. The range of measured costs of habitat conservation is very large, with 80% of total cost in the baseline scenario occurring on roughly 10% of the regulated acreage. This variation reflects the fact that efficiency is not an objective of the federal agencies charged with conserving vernal pools, and suggests that agencies could reprioritize by seeking higher levels of conservation on certain lands, thereby achieving a more efficient outcome.
The paper shows that policy coordination among federal and local agencies can reduce the cost of achieving national policy objectives. For example, by allowing denser development in areas where land conservation is sought, among other measures, cities and counties may help reduce the economic losses from federal land use controls. In the case study, densification of development is shown to reduce the costs of federal regulation by about three-quarters.
II. Model
The efforts of federal wildlife agencies to conserve habitat are akin to a licensing program in that developers can undertake projects that destroy or alter habitat once they have obtained a permit from the relevant agency. Generally, the permitting process can impact housing development projects by altering both costs and output levels (Sunding and Zilberman 2002). In their attempts to conserve land, federal agencies typically specify conservation requirements in terms of both avoidance and mitigation. Avoidance requirements entail leaving some portion of an area proposed for development in an undisturbed condition. Unless other land is made available for development, avoidance requirements result in a net loss of developable land. Mitigation requirements, by contrast, oblige the developer to undertake actions that improve or protect habitat in some other location. Usually, the federal agency specifies a mitigation ratio of acres protected off-site to acres disturbed by development; this ratio is frequently in excess of 1:1 and is affected by factors such as the rarity of the disturbed habitat, uncertainties associated with the ability to produce comparable habitat, and other factors (National Research Council 2001).
Analytically, a mitigation requirement is akin to a tax on greenfield development, typically defined as development occurring on previously undisturbed land. The number of mitigation credits needed equals the area disturbed multiplied by the mitigation ratio. The cost of mitigation is then the number of credits needed multiplied by the price per credit. The credit price is, in turn, determined by market conditions and is the result of negotiations between the project developer and the owner of the mitigation site.
The avoidance requirement reduces the stock of developable land. A great deal of research in urban economics has focused on the question of zoning regulation by local governments and the ways in which zoning and other limitations on the stock of developable land influence the housing market. The vast majority of this literature is in the neoclassical tradition of the Alonso-Muth-Mills model of location choice and urban growth and considers land as an input into the production of new housing. This approach typically assumes that density is variable and can be adjusted by developers in response to land use controls and other market conditions. Accordingly, the neoclassical approach predicts that land prices and density will increase in response to a reduction in the stock of developable land, such as an avoidance requirement.
We are interested in measuring the variation in the cost of regulation across locations, and thus we take a partial equilibrium approach that can be implemented based on particular local conditions. The surplus earned from housing development at a given location is
where λ is the amount of land per house (i.e., the inverse density), p is the price per house, H is the number of housing units, k is the cost of construction, and
is the stock of developable land. When land is subdivided optimally, there is an equality between its extensive and intensive margin values, or
That is, at the margin a unit of land is equally valuable whether used to build an additional house or to expand lot size. The market price of land (i.e., its shadow price, or γ) is then equal to the profit from housing development per unit ofland, and also to homebuyers' marginal valuation of an additional unit of lot size (pλ).
Recently, several economists have shown that while this neoclassical model fits the data well in some of the nation's housing markets, it misses the mark in other markets (Glaeser and Gyourko 2003; Glaeser, Gyourko, and Saks 2005; Sunding and Swoboda 2010). These authors argue that in some housing markets, local regulation creates scarcity by limiting the number of housing units constructed. By creating a shadow value of housing, such regulation drives a wedge between the intensive and extensive margin values of land. This situation has important consequences for development and also for the economic cost of endangered species regulations.
To see how local regulation can affect land development decisions and regulatory costs, consider the case in which a fixed stock of developable land is coupled with a minimum lot size (i.e., an upper bound on density), thus implying a cap on the number ofhousing units built. In this case, the developer's profit is equal to
The equilibrium conditions in this fixed- density model are as follows:
The first condition implies that the price of land is equal to its extensive margin value, or the value of land with a house on it. The second expression implies that when the density constraint is binding and limits the quantity of housing, the price of land exceeds its intensive margin value since it incorporates the shadow value of scarce housing. Note that if the density constraint is not binding, then θ = 0 and the system reduces to the neoclassical model shown above. If the density constraint is not binding, developers will respond to reductions in the stock of developable land by increasing density.
In recent years, there has emerged a sizable empirical literature on the shadow value of housing, and the effect of local land use regulation on the extensive and intensive margin values of land. Glaeser and Gyourko (2003) show that the shadow value of housing is large in many markets, particularly in coastal areas of the United States. Sunding and Swoboda (2010) use locally weighted regression methods to measure the shadow value of housing regulation across jurisdictions in southern California. They find that the magnitude of the shadow value can be quite large, approaching 30% of the market price of housing in some areas.
We now turn to the marginal welfare costs of habitat protection. Avoidance requirements imposed by federal agencies reduce the stock of developable land. With fixed density of development, a reduction in the stock of land causes a proportional reduction in the quantity of new housing. Alternatively, the status quo housing stock can be maintained by increasing the allowable density. We term these alternatives the rationing and densification scenarios, respectively. They bookend the ways in which a local government can respond to federal agency reductions in the stock of developable land. Since our partial equilibrium approach assumes that the demand for housing is perfectly elastic at each location, the social welfare from housing development is the sum of developer profit and landowner rent. Then social welfare is
and the marginal welfare cost of a reduction in the stock of developable land is
Note that in the neoclassical world where density is freely variable, the marginal welfare losses under these two scenarios will be equal. However, if the density constraint happens to bind, then the marginal welfare loss under rationing will exceed the loss under densification. Note that when implemented empirically, our model does not assume that the density constraint is binding, but rather lets the local housing market data indicate whether this is the case.
Recall that federal agency action to conserve species is typically specified in terms of both avoidance and mitigation. That is, development is prohibited on some fraction of the proposed project area and is allowed to proceed on the remainder. In this sense, the federal action has elements of both quantity and price regulation, since the mitigation requirement is akin to a tax on land conversion. Allowing for the possibility that regulation entails a combination of mitigation and avoidance, the per-acre cost of federal regulation is as follows:
where the price of a mitigation credit is ω, α is the percentage development that must be avoided, and ρ is mitigation ratio specified by the agency. Assuming the mix of mitigation and avoidance is primarily a choice based on habitat quality and unaffected by local zoning rules, the marginal cost of conservation through rationing will exceed the cost of conservation when density is variable.
Recently, Quigley and Swoboda (2007) developed a theoretical treatment of the effects of habitat conservation efforts on the housing market, particularly locational choices. Their findings confirm that habitat conservation can have a significant effect on welfare in urban areas, and that habitat conservation can transfer large amounts of wealth among landowners, housing consumers, and the rest of society. Their paper, however, is predicated on the standard neoclassical model as described above and does not consider the effects of local regulation. In fact, their analysis assumes that density and other housing characteristics can vary in response to land conservation efforts. Further, they assess only the welfare effects of avoidance, and not the mitigation requirements described above.
One way in which Quigley and Swoboda's approach is more general than this one is that they consider both open and closed cities, whereas this paper uses an open city model. The results are consistent in that Quigley and Swoboda note that in the open city version of their model, the marginal welfare cost of the avoidance requirement (or a reduction in the amount of developable land at a particular location) is equal to the price of land at that location. Where the approaches differ is that our analysis recognizes that local regulation can drive that price of land well above consumers' marginal valuation of lot size, implying that local regulation can drive the costs of federal land use controls well above levels envisioned by the standard neoclassical model.
III. Case Study: California Vernal Pools
This section of the paper presents a case study of the costs of federal land use controls: protection of vernal pool habitat in California. Vernal pools are ephemeral wetlands that occur primarily in Mediterranean climates. Their name derives from the fact that they often reach their maximum depth in the spring.
A large number of endangered species occur in vernal pool areas (Witham et al. 1998). For example, the San Diego mesa mint, a highly endangered plant, is found exclusively in vernal pools in the San Diego area. Many of the amphibians that breed only in vernal pools spend most of their lives in the uplands within hundreds of feet of the vernal pool. Eggs are laid in the vernal pool, then the juveniles leave the pool two or three months later, not to return until the following spring to breed. Some other species, notably the fairy shrimp, lay eggs that hatch when rains replenish the pool.
Due to the presence of listed species, and to the fact that even seasonal wetlands may be classified as jurisdictional waters of the United States, development in areas containing vernal pools is highly regulated by the federal government under both the ESA and the Clean Water Act. From an economic point of view, conservation of vernal pools is an ideal case in which to explore the impacts offederal land use controls, since they are scattered across numerous local government jurisdictions, and many vernal pool complexes lie squarely in the path of projected development.
Federal Agency Conservation Requirements
Project developers must obtain federal agency permission before commencing construction on sites that contain vernal pools. Because vernal pools are a type of wetland, they fall under the jurisdiction of the Army Corps of Engineers, which is tasked under Section 404 of the Clean Water Act with the issuance of permits to discharge fill material into wetlands and other waters of the United States. Because vernal pools contain a number of endangered plant and animal species, prior to issuing a permit the Corps must consult with the Fish and Wildlife Service under Section 7 of the ESA dealing with interagency cooperation.
California vernal pools are classified by the Fish and Wildlife Service as falling into one of two categories: Group A and Group B (USFWS 2005). Each type of vernal pool habitat is subject to different federal conservation requirements. Group A habitat has a higher frequency of occurrence and is relatively easy to provide via conservation banks. Group B habitat, by contrast, supports a greater number of endangered species and is unlikely to be successfully created in conservation banks. The Fish and Wildlife Service's recovery planning document for vernal pools reports that Group A habitat is not subject to any avoidance requirement, but that developers must provide compensatory mitigation at a ratio of 2:1 (USFWS 2005). That is, for every acre of vernal pools eliminated by development, the permit applicant must provide 2 acres of created or restored vernal pool habitat. Projects may fulfill the mitigation requirement by purchasing credits from a conservation bank, purchasing suitable habitat and managing that habitat in perpetuity, or dedicating land already owned by the project applicant and having suitable vernal pool habitat.
Federal conservation requirements for Group B habitat are much stricter. The Service's recovery plan indicates that avoidance should occur on 85.7% of the project site within vernal pool critical habitat (a 6:1 avoidance requirement), thus allowing development to occur on only 14.3% of the project site. In addition, the developer is required to mitigate disturbance of vernal pool habitat at the rate of 3:1 for each acre of vernal pools filled (USFWS 2005).
These conservation requirements reflect the universalist approach frequently taken by federal agencies with respect to environmental regulation (Sunstein 1997; Ackerman and Stewart 1985) in that they are not affected by the conditions of the underlying markets for land and housing. Rather, whatever conditioning does occur is due to biological differences in the types of vernal pool at issue, with the rarity of Group B habitat leading to more severe land use restrictions and greater on-site avoidance requirements.
Conservation bank prices are used to estimate the mitigation costs associated with federal agency action. Prices were obtained from Service personnel who routinely survey conservation banks. The largest prevalence of existing banks is in the Sacramento region, where each vernal pool conservation credit costs roughly $200,000 per acre (in 2010, seasonal wetland credits in the San Francisco Bay Area reached a market price of $780,000 per acre, reflecting both the importance of mitigation requirements and the scarcity of housing in northern California). In addition, the Service estimates that the average cost of a mitigation credit is $135,000 in Placer County and $105,000 elsewhere in the study region.
The geographic extent of vernal pools is also well defined. On August 6, 2003, the Fish and Wildlife Service proposed critical habitat for 15 vernal pool species on the threatened or endangered lists.5 Critical habitat is defined as areas that are in need of special management to ensure conservation (including recovery) of the species. The vernal pool designation is notable in several respects, first for its size. Geographically, biologically, and economically, the designation cuts across a broad swath of the California landscape. The Service designated a total of 1.2 million acres— over 1% of the entire state—as far south as Ventura County and as far north as Modoc County on the Oregon border. The analysis that follows calculates the cost of the federal conservation requirements for vernal pools within the area of critical habitat.
Market Price of Land
As defined in the previous section, the extensive margin value of land is the difference between the sales price of housing and the cost of building it, per unit of land. Building costs include the costs of both construction and development. Construction costs include expenditures on labor and materials, while the costs of development include architecture, grading, utilities, provision of common space, and local fees such as utility hookup charges. This section describes how the extensive margin value of land is calculated for each acre in the study area.
Data on the market prices of new homes were obtained from DataQuick Information Systems,6 which maintains a database of new single-family home sales in the study area. Based on information gathered from county recorders and assessors, the database provides a rich set of house descriptors, including assessor's parcel number, home size, lot size, number of stories, number of bedrooms, number of bathrooms, build year, sale price, and sale date for all transactions dating back to 1997. It is worth emphasizing that our dataset is for newly constructed homes, thus avoiding problems of overaggregation that frequently plague hedonic analyses of this type. Each observation is also spatially referenced, so the data can be aggregated to any level using a geographic information system (GIS). Table 1 presents descriptive statistics. The database exists for 120 of the 158 census tracts comprising the federally designated critical habitat. The remaining tracts were excluded from the analysis.
Because California home prices roughly tripled between 1997 and 2007, the nominal sale prices reported by DataQuick are not directly comparable across time. The prices were inflated to real dollars using the Office of Federal Housing Enterprise Oversight's home price index. This index provides quarterly data on price inflation for detached, single-family dwellings by metropolitan statistical area (MSA). For observations that did not lie in any MSA, they were matched to the closest.
Data on the cost of residential construction were obtained from Marshall and Swift,7 which publishes a quarterly guide to building cost per square foot indexed by region, construction quality (average, good, very good, or excellent), and home size. New homes were assumed to be one story, stud-framed with stucco siding, and of average construction quality, which is typical for newly constructed tract homes.
Housing Data
Descriptive statistics for the extensive margin value are presented in Table 2. The first two columns present the means and standard errors of the mean. Means range from $25 to $165 per square foot. The remaining four columns display observations per county as well as the 25th, 50th, and 75th percentiles.
Intensive Margin Values of Land
Intensive margin values of land are calculated using a hedonic regression whereby home value, characterized by observed selling price, is decomposed into an additive bundle of housing, geographic, and demographic attributes:
where L is a vector of neighborhood and geographic control variables, plus an intercept. Spatial variables included the following: distance to the nearest incorporated city, distance to the nearest MSA, county and city fixed effects, and elevation and the square of elevation above sea level. The demographic controls were block group data in the 2000 census and consisted of population density (persons per square mile), median age, percent population growth over the period 2000-2004, percent white, percent black, percent Hispanic, percent Asian, and percent of housing that was renter occupied. The coefficient β1 is then the intensive margin value of land as defined in the previous section. The specification was estimated on observations grouped by MSA so as to model the composition of markets for new housing and developable land, which frequently span county lines. Geospatial datasets often exhibit spatially autocorrelated error terms, leading to a violation of the assumption of homoskedastic disturbances (Anselin 1988). To account for this possibility, we employ robust regression using iteratively reweighted least squares to estimate the parameters of our model (Hampel et al. 2006).
Table 3 displays point estimates, standard errors, and regression diagnostics. The least expensive land values in the dataset are located in the Modesto MSA, at $3.03 per square foot. At more than five times that amount, the Oakland MSA is the most expensive in the study area.
Development Activity in Areas Containing Vernal Pools
The total cost of vernal pool conservation is the product of the cost of conservation per acre multiplied by the number of acres that will be developed. Determining the amount of new housing that will be developed within areas of vernal pool habitat presents several challenges. First, a suitable time frame must be selected, one that is short enough to give the model strong explaining power, yet long enough to fully capture the effects of the regulation. Second, a suitable and precise spatial frame must be selected, due to the highly localized effects of habitat conservation. Third, predicting the location of development requires modeling the urban growth process, a notoriously difficult problem.
Extensive Margin Values of Land
Intensive Margin Values of Land
This paper examines housing growth that is forecasted to occur over a 20-year time frame. This study period coincides with the planning horizon of the California state-mandated jurisdictional General Plans and population and employment projections by regional associations of governments. It is the longest time span for which the demographic and economic forecasts needed for this study are reliable and available.
This study reaches below the regional level in order to illustrate the location-specific, heterogeneous nature of the effects of federal land use regulation. Census tracts are used as the unit of analysis. This choice is motivated by both the nature of the problem and data availability. The census tract is a standard level of aggregation in socioeconomic research and is the finest level of distinction at which the above data are published. This is important in light of our finding that local, even neighborhood-level characteristics are responsible for a high degree of heterogeneity in the effects of habitat conservation. For example, a county-level analysis may not be sensitive enough to discern any noticeable effect even though the effects are large on a smaller scale.
The primary sources for estimates of future housing and population growth were the study area's federally designated Metropolitan Planning Organizations (MPOs). Typically created by county governments, these forecasts are the preferred source for growth estimates because they are created using detailed knowledge about local growth trends and characteristics, potentially resulting in more accurate estimates than those obtained with mathematical forecasting techniques. The organizations that created the estimates used in this analysis are the Association of Bay Area Governments, the Sacramento Area Council of Governments, the Southern California Association of Governments, and the Association of Monterey Bay Area Governments.
If these forecasts were not available for a given area, for example because the MPO does not publish these data out to 2025 or at the tract level, mathematical forecasts created for a CalTrans planning survey by UCLA's Institute of Transportation Studies were used instead (Crane et al. 2002).
Identifying the housing and population deltas also requires estimates of present-day conditions. These were obtained from Applied Geographic Solutions,8 a company that offers yearly updates to the latest available decennial census housing estimates based on the U.S. Census Bureau's American Community Survey, change-of-address records, Federal Emergency Management Agency registrations, U.S. Postal Service delivery statistics, Internal Revenue statistics, and credit-reporting databases.
Finally, growth estimates must be modified to account for infill development, a common practice in California, where many cities are already built out. Infill development entails redeveloping low-density, typically single-family, dwellings into high-density apartment blocks, condominiums, or shared dwellings. Landis and Reilly (2003) completed a study of projected infill development by county over the next 50 years. Growth estimates were adjusted according to their numbers to avoid overstating greenfield development.
Since the designated federal critical habitat does not delineate the precise location of vernal pools, but rather areas that are likely to contain them, the model was adjusted probabilistically to account for the likelihood of a federal permitting requirement. The expected loss due to vernal pool conservation, E(L), was set equal to the expectation over the potential states C (presence of vernal pools triggering federal conservation requirements) and NC (no conservation), with E(L|NC)≡ 0. Thus, E(L) = p(C)E(L | C). Determining p(C), the probability of the presence of vernal pools thereby triggering the imposition of conservation requirements, requires knowledge on the distribution of pools within critical habitat. Vernal pool density was calculated using a study by Holland (1998), who performed a complete visual survey in twenty Central Valley counties for vernal pools using aerial photography. His results were used to determine average vernal pool density by county. In counties he did not survey, density was assumed to equal the mean across surveyed counties, 6%.
It is now necessary to allocate this growth within the census tract. This is an important leap over assuming growth will occur uniformly across the landscape. The hydrographic and edaphic features of vernal pools may cause land to be unsuitable for development, preventing conserved habitat from interfering with planned development and resulting in no added cost. Conversely, conserved habitat may occupy the last portions of undeveloped land within a tract, meaning future development will be shoehorned toward vernal pools. These scenarios illustrate the need for more precise growth allocation.
This allocation was performed using the California Urban and Biodiversity Analysis (CURBA) model developed by Landis and Reilly (2003). CURBA is a statistical model that incorporates both spatial and nonspatial data to project urban growth in California. Its explanatory variables include demand variables pertaining to job accessibility and income growth; location-specific variables such as freeway proximity, whether the land is classified as farmland, and whether it lies in a floodplain; neighborhood variables; and regulatory variables, such as whether a location is in an incorporated city.
CURBA analyses the state of California by dividing it into a matrix of one-hectare grid cells. It outputs a probabilistic score that a given cell will be converted from undeveloped to developed in the next 20 years. Let G be the total amount of projected greenfield development, defined above, and define the CURBA prediction function 
mapping each cell to its respective probability of development. The analysis assumes the following identity holds:
Thus, the sum of probability scores within each census tract, scaled by a fixed multiplier, is identically equal to the total projected greenfield development for that tract. Now solve for λ, and let the sets HAand HBbe those cells that fall in group A and B habitat, respectively. Then the expected development in group A habitat is given by
with GB defined similarly.
Welfare Costs by County (Thousands of Dollars)
IV. Results
Results of the analysis are shown Table 4. Federal agency efforts to conserve vernal pools result in approximately $890 million in lost welfare within the study area under the rationing scenario, and $240 million if developers are permitted to increase density in response to federal action. A total of $119 million in costs is due to mitigation requirements, and the remainder is due to the requirement that project developers avoid portions of the project site.
The county-level summaries mask a considerable heterogeneity in impacts across the study area. When costs are aggregated by census tract, the finest level available at this stage in the analysis, the results are striking. Figure 1 shows overlaid Lorenz curves for the two scenarios. The x-axis contains the cumulative percentage of critical habitat contained in each census tract, and the y-axis contains the cumulative percentage of cost. In the densification scenario, nearly 70% of the cost of federal agency action stems from roughly 20% of the affected land area. Results are even more varied in the rationing scenario, where 80% of the costs relate to protection of 10% of the habitat.
Carrying this logic a step further, we examine individual examples of census tracts in the study area to demonstrate that the most expensive protected habitat may also be the least desirable from a conservation perspective, at it usually surrounds urban fringes. Table 5 displays results for the top five most expensive census tracts in the study. The single most expensive tract, in Sacramento County, has impacts totaling $423 million under rationing ($62 million if densification is allowed), which is almost half the total cost of the proposed regulation. Figure 2 shows a map of this tract, along with satellite imagery taken in 2005. Development and grading preparations are readily visible, and urban expansion appears primed to target the area set aside as critical habitat, which is marked with slash lines. This location presents an excellent example of how a spatially disaggregated approach can be used to identify specific areas with high economic costs.
Lorenz Curve Plotting the Percentage of Habitat Conserved against the Percentage of the Total Cost of Vernal Pool Conservation
Top Five Census Tracts in Terms of Total Welfare Costs (Thousands of Dollars)
One of the least impacted census tracts in the study is 06013304000, in Contra Costa County. Satellite images of the area, again from 2005, are shown in Figure 3, with critical habitat overlaid with slash lines. This example illustrates how terrain informs the results of the model. Undeveloped land in the western portion of the tract costs little to designate since it borders farmland and is hilly, both of which make development more difficult.
An additional metric that can be used to analyze the effects of federal land conservation efforts is implicit cost per acre of habitat conserved. Table 6 displays the same results as Table 5, except that tracts have been sorted on this basis. The census tract with the highest cost per acre, in Solano County, results in $3.9 million in losses ($757,000 in the densification scenario) for 28.5 acres of designated, group B habitat. The cost of the designation in this tract approaches $135,276 per acre of vernal pools. A map of this tract is shown in Figure 4. The only undeveloped parcel in the census tract is that which was conserved; all remaining land has been densely urbanized. Not only does this result in large producer gains from constructing additional housing, but it may also be of only marginal ecological value for preserving the listed species due to the close proximity of urban development.
Census Tract 06067008701
Census Tract 06013304000
Top Five Census Tracts in Terms of Cost per Acre (Thousands of Dollars)
Census Tract 06095252502
Conclusions
The paper considers the economic costs of federal land use regulation to protect habitat for endangered species in a legal system that gives primary regulatory authority over land use to state and local governments. The conceptual model of housing development shows how local regulations can influence the impacts of federal interventions, and develops expressions for welfare losses. The model also incorporates basic elements of federal permits, distinguishing between avoidance and mitigation requirements, which can have quite different welfare implications.
The paper applies this framework to federal efforts to conserve vernal pools in California. Considering both avoidance and offsite mitigation requirements, we show that the potential welfare costs of protecting vernal pools can run to $890 million in present value terms. This cost is unevenly distributed across the landscape, with over 80% of the total welfare cost associated with protecting roughly 10% of the existing vernal pool habitat in the rationing scenario.
One way in which federal regulators could tailor their habitat protection efforts to local land market conditions is by altering the mix of avoidance and mitigation requirements according to local market conditions. With respect to vernal pools, the analysis has shown how the presence of group B habitat, and the consequent requirement to avoid areas that would have otherwise been developed, causes costs to increase by a factor of roughly three over the mitigation expenditures that would have been required otherwise. There are other reasons why mitigation is desirable as well. It streamlines the permitting process by affording developers a ready-made stock of habitat, obviating the costly and time-consuming process of creating it on a project-by-project basis. This would also seem to generate efficiency gains—although the effects of time delay have not been considered in this paper, they are judged to be considerable since surplus gains from housing are large, and development requires an outlay of fixed assets that cannot be shifted as permitting proceeds (Mayer and Somerville 2000).
Another major finding of the paper is that the welfare cost of federal agency action depends heavily on the nature of local regulation, and how local governments respond to federal intervention. When the housing supply is limited by preexisting local regulation, the welfare cost of additional federal regulation can be large. Conversely, local governments can reduce the cost of federal action by relaxing or altering zoning and other regulations. We consider the case in which local governments accommodate federal protection efforts by allowing increases in the density of development. If federal regulation reduces the stock of developable land by imposing avoidance requirements, then losses per acre are equal to the extensive margin value of land, which incorporates the associated shadow price of new housing. If density is allowed to adjust to maintain the original number of new housing units, then the per-acre loss from the avoidance requirement is equal to the intensive margin value of land. This result follows from the fact that when the density constraint binds from below, homebuyers are forced to consume too much land in equilibrium. Relaxing this constraint is an important way to accommodate Congress's mandate for the federal government to conserve habitat for endangered species.
Acknowledgments
This research was funded by the Giannini Foundation, the U.S. Department of Housing & Urban Development and the U.S. Department of the Interior. We would like to thank seminar participants at UCLA, UCSB, Berkeley, Ohio State, Wisconsin, VPI, Maryland and Arizona. We would also like to acknowledge helpful conversations with John Quigley, Jeff Zabel, JunJie Wu, Andrew Plantinga, Jennifer Baxter, Robert Unsworth, Chip Patterson, Ted Maillett, Richard Adams, David Zilberman, Tom Davidoff, and Steve Raphael.
Footnotes
The authors are, respectively, professor, Department of Agricultural and Resource Economics; and Ph.D. candidate, Department of Statistics, University of California, Berkeley.
↵1 16 U.S.C. §1531 et seq. (1973).
↵2 50 C.F.R. §17.3. See Babbitt v. Sweet Home Chapter of Communities for a Great Oregon, 515 U.S. 687 (1995).
↵3 Federal RegisterVol. 72, No. 243, Wednesday, Decemer 19, 2007, at 72010-72213.
↵4 437 U.S. at 169.
↵5 Federal Register Vol. 68, No. 151, Wednesday, August 6, 2003, at 46684-46732.
↵6 See www.dataquick.com.
↵7 See www.marshallswift.com.
↵8 See www.appliedgeographic.com.
