Abstract
The Endangered Species Act requires the federal government to designate critical habitat for species listed as threatened or endangered. This provision has proven to be one of its most controversial, as it entails special management and potentially greater regulation. We measure the economic impact of critical-habitat designation by estimating its effect on the market value of vacant land. Using data from 13,000+ transactions that occurred within or near critical habitat for two species in California, we show that it resulted in a large and statistically significant decrease in land value, with the largest decreases occurring within urban growth boundaries. (JEL Q51, Q57)
1. Introduction
The Endangered Species Act (ESA) is the nation’s primary law to protect endangered species and their habitats.1 The ESA prohibits any action, whether undertaken by a federal agency or a private party, that causes the “take” of any listed species, even if the taking is incidental to the activity in question.2 The ESA also requires that federal agencies refrain from actions that would jeopardize the continued existence of any protected species, including the adverse modification or destruction of designated “critical habitat.”
The ESA requires both NOAA3 Fisheries and the U.S. Fish and Wildlife Service (collectively, “the Services”) to designate critical habitat for species once they are listed as threatened or endangered. Critical habitat is defined as specific geographic areas that are essential for the conservation of a threatened or endangered species and that may require special management and protection. Once an area is designated as critical habitat, federal agencies are required to consult with the Services to ensure that actions they carry out, fund, or authorize will not disrupt or “adversely modify” critical habitat.
Critical-habitat designation has proven to be one of the most controversial aspects of the ESA. The federal government maintains that in most situations, its designation of critical habitat has little economic consequence. This stance is repeated in dozens of agency analyses of critical-habitat designation. However, the government’s arguments on this point are theoretical, not empirical. The government notes that critical-habitat designation has no effect in situations where a federal agency is not involved, for example, a landowner undertaking a project on private land that does not involve federal funding or a federal permit. Second, in areas where the species is present, the species already receives protection due to the ESA’s prohibitions on take. Thus, the federal government asserts that critical-habitat designation is potentially costly only in special cases such as land that is not currently occupied by a listed species.
The federal government’s claim is important because critical-habitat designation is one of the few instances where economic costs and benefits are taken into account when determining endangered-species protections. Section 4(b)2 of the ESA requires federal agencies to consider the economic costs of critical-habitat designation when designating critical habitat. The ESA authorizes the Secretary of the Interior to exclude lands from critical habitat where the benefits of designation are less than the costs, provided that exclusion of these lands from critical habitat will not jeopardize the continued existence of the species. Thus, if the government takes an overly narrow view of the consequences and costs of critical-habitat designation, then too much land may be set aside as critical habitat.
As we explain in more detail in Section 2, there are reasons to suspect that the government’s view of the costs of critical habitat is too constrained and that critical-habitat designation may indeed have significant economic consequences to landowners. For instance, the Services openly acknowledge that critical habitat is a signal to other regulators—including state and local land use officials—that the lands and waters designated as critical habitat are essential for the recovery of a listed species and are in need of special protection. Indeed, the signaling effect of critical-habitat designation is routinely listed as one of its main conservation benefits.4 However, for such benefits to exist, other regulators must respond to the signal sent by critical-habitat designation by strengthening their own conservation requirements. In California, the setting of our empirical research, critical-habitat designation explicitly triggers state law provisions that restrict state and local permitting processes.5
This paper and its working paper version provide some of the first econometric estimates of the costs of critical-habitat designation. We isolate the differences in vacant-land prices attributable to critical-habitat designation. Specifically, our analysis uses a difference-in-differences approach to separately measure the impact of critical-habitat designation for two species on prices for vacant land in California: the red-legged frog (RLF) and the Bay checkerspot butterfly (BCB).
The RLF is the largest native frog in the western United States. RLFs have suffered substantial losses in habitat, primarily caused by urban development. In addition, RLF populations have declined due to overexploitation and the introduction of exotic predators. The RLF has a long regulatory history under the ESA, having been listed as threatened in 1996. In 2001, the U.S. Fish and Wildlife Service issued a final rule designating critical habitat for the RLF. This controversial decision immediately met challenges from development interests. These legal challenges found success, and the 2001 designation was vacated in 2003. Finally, in 2006 the Services reissued an extensively revised critical-habitat designation for the RLF.6 The 2006 designation appears to be more permanent and does not currently face any challenges.
The BCB is a species of great interest to biologists, who noticed its rapidly declining population in the early 1980s. Encroachment of human development into BCB habitat has contributed to this decline. Invasive-species proliferation also poses a grave threat to this butterfly (along with many other species of butterflies). The greatest threat to the BCB, however, comes from increasing emissions of nitrogen. Nitrogen from air pollution increases the fertility of naturally low nitrogen serpentine soils, which are the main home of the sources of nectar the butterfly requires.7The increased nitrogen concentrations promote the growth of invasive species on these serpentine soils, which cut off the butterfly’s access to nectar-producing and larval-host plants. The BCB was listed as threatened in 1987. In 2001 the Services designated 23,903 acres in San Mateo and Santa Clara Counties as critical habitat for the BCB.
In a competitive land market, one would expect that the difference in a land parcel’s market value before and after critical-habitat designation would capture the economic impact of critical-habitat designation (on future development). The perfect experiment to measure these effects would entail randomly assigning critical-habitat designation to otherwise identical parcels at random times, and then comparing the difference in values before and after designation, relative to the same value difference for nondesignated parcels. Such an experiment does not exist in the real world because designation is nonrandom, parcels are not identical, assignment often occurs at one time for all parcels, and the market signals the value of any parcel only at the time of a sale—we cannot know the value at every point in time.
Our estimation strategy differs slightly between the two species that we study. For the RLF, we assemble statewide data on vacant-land transactions in California between 1993 and 2008. We then overlay the map of sold parcels with the boundaries of critical-habitat-designated land for the 2001 and the 2006 designations. Because we observe transactions on designated and nondesignated parcels for both of these designations—both before and after the designation dates—we can econometrically identify the effect of critical-habitat designation on land values, after controlling for parcels’ observable differences.
For the BCB we assembled data from vacant-land transactions in the two regulated California counties between 1988 and 2007 (Santa Clara County and San Mateo County). This sample is indeed smaller, but it has a key advantage in that we have identified a reasonable instrument for critical-habitat-designation status for the BCB: an ecological feature known as serpentine soil.8
To preview the results, we find that for the parcels in our sample, critical-habitat designation decreased land values by 48% for the RLF and at least 78% for the BCB. These results suggest critical-habitat designation can carry significant economic costs, contrary to frequent assertions of U.S. federal wildlife agencies.
2. Habitat Protection and the ESA
Federal legislation has recognized the importance of conserving habitat for endangered species since the beginnings of legislation protecting endangered species. The predecessor to the current ESA, the Endangered Species Preservation Act of 1966,9 stated that “it is . . . the policy of Congress that the Secretary of Interior, the Secretary of Agriculture, and the Secretary of Defense . . . shall preserve the habitats of such threatened species on lands under their jurisdiction.” Following this vein, the current ESA focuses on conserving endangered species “and the ecosystems on which they depend”—a clear effort to link the conservation of species with the conservation of habitat (National Research Council 1995). The first consideration for the listing of a species as threatened or endangered under the ESA is whether the species faces “present or threatened destruction, modification or curtailment of its habitat or range.” Section 4 of the ESA also requires the designation of a species’ “critical habitat” concurrently with the listing of a species, unless earlier listing is “essential to the conservation” of the species (or unless the designation of critical habitat is not “prudent” or “determinable”).The ESA requires that critical-habitat designations must, “to the maximum extent practicable,” be accompanied by “a brief description and evaluation of those activities (whether private or public) which, in the opinion of the Secretary, if undertaken may adversely modify such habitat, or may be affected by such designation.” Designation of critical habitat occurs “on the basis of the best scientific data available and after taking into consideration the economic impact, and any other relevant impact, of specifying any particular area as critical habitat.” Importantly, Section 4 of the ESA provides for the exclusion of areas from critical habitat if it is determined that “the benefits of such exclusion outweigh the benefits of specifying such area,” unless failure to designate the area “will result in the extinction of the species concerned.” The primary benefit of exclusion is avoided regulatory costs (U.S. Department of the Interior 2008).
Interagency consultation (i.e., consultation between the Services and other federal agencies) is one mechanism by which the ESA’s critical-habitat provisions affect private landowners. The ESA requires federal agencies to consult with the Services whenever activities that the federal agencies permit may affect a listed species. In particular, Section 7 requires federal agencies to ensure that their permitting of a project will not result in the “destruction or adverse modification” of designated critical habitat. For example, the Army Corps of Engineers must consult with the U.S. Fish and Wildlife Service before issuing a discharge permit under Section 404 of the Clean Water Act if the project takes place on designated critical habitat.
There are three main ways in which the existence of critical habitat can result in additional project costs and reduced profits for developers: (1) critical habitat may result in additional consultations; (2) the outcome of consultations may require additional conservation; and (3) critical habitat may cause other regulatory bodies—especially at the state and local levels—to impose additional requirements/restrictions. Critical-habitat designation may trigger additional consultations that would not occur absent the designation. This circumstance is especially likely to occur when critical habitat is designated on areas that are not currently occupied by the species. The administrative costs and required conservation efforts associated with these additional consultations are incremental impacts of critical-habitat designation.
Critical-habitat designation may also increase the costs of Section 7 consultation by triggering more stringent conservation requirements. This outcome is particularly likely in light of the Ninth Circuit Court of Appeals’ decision in Gifford Pinchot Task Force v. U.S. Fish and Wildlife Service, which held that conservation requirements imposed by the Services in areas of critical habitat should facilitate the recovery of the species, and not merely satisfy the prohibitions against take.10 That is, in areas that are occupied by the species but not designated as critical habitat, the Services use a lower “jeopardy” standard in the consultation process. In the presence of critical habitat, however, the Services must require conservation actions that will facilitate the recovery of the species—a higher standard than merely avoiding jeopardy.
Importantly, critical-habitat designation can result in incremental impacts even when there is no federal nexus. For instance, critical-habitat designation may provide new information to a state or local government about the sensitive ecological nature of a geographic region, triggering additional economic impacts under other state or local law. The California Environmental Quality Act (CEQA), for example, requires that lead agencies (i.e., public agencies responsible for project approval) must consider the environmental effects of proposed projects that are considered discretionary in nature and not “categorically or statutorily exempt.” Thus, critical-habitat designation may trigger further CEQA-related requirements. These impacts are most likely in areas where the critical-habitat designation provides clearer information on the importance of particular areas as habitat for a listed species. In addition, applicants who are categorically exempt from preparing an Environmental Impact Report may lose their exemption following the designation of critical habitat (Economics and Planning Systems, Inc. 2004). Beyond these technical legal requirements, critical-habitat designation may signal to local regulators that certain lands have special biological significance and are therefore deserving of higher levels of protection. The U.S. Fish and Wildlife Service recognizes this signaling effect, noting that “designating critical habitat also helps focus the conservation efforts of other conservation partners such as state and local governments, nongovernmental organizations and individuals.”11
3. Related Literature
The literature concerning the economics of the ESA focuses on the incentive effects of the ESA’s prohibition on harming listed species. In one strand of the literature, economists argue that the ESA discourages the creation and maintenance of species habitat on private land. Section 9 of the act makes it illegal for a private landowner to engage in activities that could harm an endangered species—including habitat modification—without first obtaining a federal permit. Such regulations can reduce private land values and antagonize private landowners who might otherwise cooperate with conservation efforts. Simply put, Section 9 turns endangered species into economic liabilities for property owners. Consequently, landowners have been known to preemptively destroy or degrade potential habitat on their land in order to avoid the act’s habitat-based requirements. Importantly, it is not illegal to modify land that might eventually become endangered-species habitat, nor are landowners required to take affirmative steps to maintain endangered-species habitat beyond refraining from actions that “harm” the endangered species. There are a variety of approaches to study the impact of critical habitat designation, which are discussed by Plantinga, Helvoigt, and Walker (2014).
Innes (1997) constructs an economic model of government takings in which some property owners develop their land earlier than others. Because it is efficient for the government to “take” undeveloped land before developed land, uncompensated takings incentivize landowners to develop their property early—reducing the risk of government appropriation. Innes shows that appropriately compensating landowners for taken property—or “equal treatment” toward owners of developed and undeveloped land—will (1) counter the incentive for overdevelopment and (2) restore efficiency. However, when the government responds to political pressure, a judicial compensation requirement will often lead to less government land use regulation than is efficient. In contrast, a judicial equal-treatment requirement can elicit efficient development and regulatory decisions.
Innes, Polasky, and Tschirhart (1999) surveyed private economic incentives for species preservation created by alternative property rights and compensation regimes. The authors show that the possibility of compensation affects investments in land and the landowners’ willingness to collect and impart information about their land’s preservation value. Innes, Polasky, and Tschirhart also address the government’s incentives and discuss how the deadweight loss from compensation influences the design of property rights and how the government’s susceptibility to interest group pressure can cause inefficient preservation.
In the past, there was little more than economic theory and anecdotal accounts upon which to assess the effect of the ESA on private land. Now, however, empirical data on several important species can demonstrate how the act itself influences landowners’ behavior—and species conservation—on private land.
A 2003 study by Lueck and Michael asked whether private landowners engaged in preemptive habitat destruction when the presence of endangered red-cockaded woodpeckers placed landowners at risk of federal regulation and, consequently, at risk of losing their timber investments. Providing habitat for a single woodpecker colony could cost up to $200,000 in foregone timber harvests. The landowners facing the greatest risk of restrictions were also most likely to harvest their forestlands prematurely and reduce the length of their timber-harvesting rotations. This behavior is consistent with landowners seeking to avoid potential losses. The ultimate consequences of this behavior were potentially significant in that these harvests resulted in losses of several thousand acres of woodpecker habitat—a major loss for a species dependent upon private land for its survival.
In a second study involving the red-cockaded woodpecker, Zhang (2004) similarly finds that “regulatory uncertainty and lack of positive economic incentives alter landowner timber harvesting behavior and hinder endangered species conservation on private lands” (p. 151). Zhang also concludes that “a landowner is 25% more likely to cut forests when he or she knows or perceives that a red-cockaded woodpecker cluster is within a mile of the land than otherwise” (p. 160).
Simply listing a species could discourage private landowners from participating in conservation efforts, according to a 2003 study by Brook, Zint, and De Young. Surveys of private landowners within the animal’s range found that as landowners became aware that their land contained habitat for Preble’s meadow jumping mouse, some became less likely to support conservation efforts. In addition, landowners would refuse to give biologists permission to conduct research on their land to assess mouse populations out of fear of the consequences of such a discovery. This revelation is especially troubling, as accurate data on species’ populations and their habitats are essential for successful conservation efforts.
One economic study that does consider the impact of critical-habitat designation is the 2006 paper by List, Margolis, and Osgood. Keeping to the theme of preemption, the authors find that species listing can accelerate the development of potential habitat, as landowners seek to preempt the ESA’s land use restrictions. Specifically, land proposed for critical-habitat designation for the endangered cactus ferruginous pygmy owl was, on average, developed one year earlier than equivalent parcels that were not identified as critical habitat.
Greenstone and Gayer (2008) surveyed the use of and potential for quasi-experimental and experimental techniques in environmental economics. Specifically, Greenstone and Gayer illustrate quasi-experimental methods by assessing the validity of a quasi-experiment that aims to estimate the impact of the ESA on property markets in North Carolina. The authors consider neighborhood housing-market characteristics and assess whether they are balanced in census tracts with and without protected species. As the authors note, the main limitation of the paper is that by focusing on the housing market, the analysis does not reveal much about the effect of development restrictions on undeveloped parcels, where habitat conservation is likely to have the largest effect.
Nelson et al. (2017), who cite the working paper version of our paper, examine the impacts of the ESA on land cover change. They find no significant evidence that the rate of land cover change on agricultural and developed land changed in areas designated as critical habitat between 1992 and 2011. Their controls are areas subject to ESA regulation, but which are not designated as critical habitat.
The present paper contributes to the literature by addressing the economic costs of critical-habitat designation. For a variety of reasons, described in Section 2, critical-habitat designation may lower future development profits and reduce land values as a result. Among these reasons is that critical-habitat designation signals to state and local land use authorities that designated lands are in need of special protection to accomplish the recovery of listed species. Critical-habitat designation may also signal to potential buyers that there may be additional costs and delays associated with future development of the land, reducing designated land’s market price. In the next section, we detail the data collected to test the impact of critical-habitat designation on land value.
4. Data
For our analysis of the economic impact of critical-habitat designation, we assembled data on vacant-land transactions in the state of California between 1993 and 2008, using the state dataset compiled by DataQuick. These sales data include the parcel number, lot size, sale price, sale date, zoning, and centroid coordinates (latitude and longitude). We drop multiproperty sales, as each parcel is assigned the price of the whole transaction in the data. We match additional characteristics to the dataset of vacant-land sales. We obtained geographic information system data from the U.S. Fish and Wildlife Service (2008) on the location of the RLF critical habitats for both the 2001 and 2006 designations, as well as for the 2001 designation of critical habitat for the BCB.12 These boundaries determine whether a sold parcel was located in an area (eventually) designated as critical habitat under either the 2001 or the 2006 rule. In addition, we calculate the distance from each parcel’s centroid to (1) the nearest highway and (2) the nearest city boundary. Finally, we match parcels to their historical monthly temperature and rainfall data using historical weather data from PRISM (PRISM Climate Group 2008).
Table 1 provides the summary statistics for the RLF dataset. We have a universe of 7,777 vacant-land transactions within our sample frame. Table 1 provides summary statistics by (1) designation assignment (i.e., treated with designation or control) and (2) timing (i.e., before or after the designation), for both the 2001 and 2006 critical-habitat-designation policies. The first column lists the average transaction price per acre, which varies greatly from $91,000 per acre (in designated parcels after the 2006 designation) to $1.2 million per acre (in nondesignated parcels following the 2006 designation). The next column lists average lot size transacted, which ranges from 4.95 acres to 16.04 acres. Columns (6) and (7) give average historical precipitation and maximum temperature, followed by columns (8) and (9), which give distances to the nearest highway and city, respectively. Column (10) lists the number of transactions within each group. Finally, columns (11) and (12) illustrate the first and last dates, respectively, that we observe for each group. For instance, the first row contains summaries for vacant-land sales that occurred before the 2001 designation for parcels that were not deemed critical habitat during the 2001 designation. We observe 3,051 of these sales, beginning in January of 1993 and continuing until March of 2001 (when the 2001 designation occurred). We then observe 2,707 sales of properties not designated critical habitat by the 2001 designation, following the 2001 designation (the second row of Table 1). We observe 387 sales in the 2001-designation territory before the designation and 420 sales in the territory following the designation (rows 3 and 4, respectively). For the 2006 designation, we observe 2,742 control sales before the designation (row 5) and 1,018 sales after the designation. Finally, for the 2006 designation, we observe 83 treated sales before designation and 34 treated sales after designation, which is admittedly a small number, yet all the data available to us currently. On average, before the 2001 RLF designation, (eventual) RLF-designated parcels are larger and less valuable (on a peracre basis) than nondesignated sold parcels. For the 2001 designation, the designated parcels are closer to cities and slightly farther from highways. However, for the 2006 designation, designated parcels are slightly farther from highways and farther from cities.
Red-legged Frog. Summaries of Observable Variables for Land Sales: Pre- and Postpolicy Levels Summaries (Means. Standard Deviations, and Counts) Calculated for Policy by Treatment Status by Period Groups
For our analysis of the BCB we obtained an additional dataset that contains all vacant-land transactions for Santa Clara and San Mateo Counties from the county assessor’s offices for both counties (San Mateo County Planning Office 2008; Santa Clara County Planning Office 2008). San Mateo County is the coastal county directly south of San Francisco, making up most of the peninsula. It is one of the 20 most affluent counties in the United States and is home to a population of approximately 700,000 people. Santa Clara County, home of the Silicon Valley, is directly adjacent (to the south) and is home to 1.6 million people. For both counties, we have assembled the universe of vacant-land transactions between 1988 and 2007 from DataQuick (2008). These data also include the parcel number, lot size, sale price, sale date, zoning, and the centroid’s coordinates.
The Santa Clara and San Mateo County planning departments provided spatial data on the locations of each county’s urban growth boundaries. Both counties regulate development types and densities in order to confine urban development within their respective urban growth boundaries. We used data from U.S. Geological Survey13 to locate wetlands and slopes for the geographic area of analysis. The location of farmland types came from the State of California Department of Conservation Farmland Mapping and Monitoring Program (California Department of Conservation 2008). For our analyses, we chose the locations of the three top-rated farmland types (as determined by their soil quality and irrigation status): prime farmland, state-recognized farmland, and unique farmland. Finally, we acquired the location of serpentine soil ecosystems in the two counties from the private consulting firm Jones and Stokes Associates.14
For this BCB sample, we have 3,433 sales of undeveloped parcels. Figure 115 maps these sales, in conjunction with the local critical-habitat zones and the urban growth boundaries. Table 2 provides a summary of our data for the BCB, with treatment determined by the 2001 designation of critical habitat for the butterfly. The format of Table 2 mirrors the formatting of the RLF summary table (Table 1). Because our analysis of the BCB focuses on the two counties of Santa Clara and San Mateo, the numbers of posttreatment sales are even smaller for the BCB than the RLF. As with the RLF, the BCB designation tended to affect larger, less valuable (price per acre) properties. In these two counties, the average sale included 2.4 acres of land for approximately $655,000. Wetlands and floodplains make up small shares of the land sold, and 69% of sales occurred within the counties’ urban growth boundaries. As we discuss below, serpentine soils are highly predictive of BCB critical-habitat designation in Santa Clara and San Mateo Counties.
Locations of Bay Checkerspot Butterfly (BCB) and Red-legged Frog (RLF) Critical Habitats and Vacant-Land Transactions in San Mateo and Santa Clara Counties
Sources: Data from U.S. Fish and Wildlife Service (2008, 2018) and Santa Clara County Planning Office (2018); authors.
Bay Checkerspot Butterfly. Summaries of Observable Variables for Land Sales: Pre- and Postpolicy Levels Summaries (Means. Standard Deviations, and Counts) Calculated for Policy by Treatment Status by Period Groups
5. Empirical Model
In order to identify the effect of critical-habitat designation on land values, we adopt an approach that is similar in principle to a difference-in-differences estimation strategy. We do not observe parcels being sold many times in critical-habitat and non-critical-habitat areas. A broad panel dataset on repeated sales would allow us to control for time-invariant differences in unobservables across parcels via a fixed effects strategy. Since we do not observe a sufficient number of repeat sales, we use the individual sale of parcel i in month t as our unit of observation. At each point in time t, however, multiple parcels are being sold simultaneously. A simple hedonic equation for parcel i can therefore be written as
[1]
where pit refers to the per-acre transaction price of parcel i in month t, and zit denotes a vector of exogenous observable characteristics of parcel i in month t. The coefficient vector β refers to the marginal effect of a one-unit change in the zit on pit. The term εit references a stochastic disturbance to the per-acre price of parcel i in month t.
Motivated by a difference-in-differences framework, we augment the basic specification in equation [1] as
[2]
where RLFi references a vector of dummy variables indicating whether the parcel is on land that at any point in time received designation as critical habitat under a given rule.16 The term Post-RLFt denotes a vector of dummy variables indicating whether month t occurs prior to (0) or after (1) implementation of a given rule. The interaction of these two indicators (RLFi × Post-RLFt) takes the value 1 if the parcel’s land had already received designation as critical habitat at the time of the transaction. We construct these indicator variables separately for the 2001 and 2006 rules. The two-element coefficient vector δ provides the main effect of interest. If critical-habitat designation significantly and negatively affects land values, then we should expect δ < 0 (and δ is statistically significantly different from zero). The term φt refers to month-of-sample17 fixed effects, which account for any unobservable shocks to land prices common to all parcels in a given month, for example, business cycle effects and inflation. Lastly, ηj corresponds to geographic fixed effects. In our RLF analysis, the ηj are fixed effects for each three-digit zip code18; in our butterfly regression, we use census-tract fixed effects. These fixed effects account for unobservable, time-invariant differences in housing prices across three-digit zip codes or census tracts.
The model in equation [2] rests upon two key identifying assumptions. The first of these identifying assumptions requires that we observe transactions pre- and postdesignation for each species within the critical-habitat boundaries in the study counties. For our RLF sample, Table 1 shows that there is indeed good coverage for the 2001 designation: 420 postdesignation sales on critical-habitat land. For the 2006 designation, we have to rely on a smaller sample of 34 postdesignation sales. The differences in observable characteristics for the residential parcels are shown in Table 1.
This observation leads us to the second key identifying assumption:
In words, this assumption requires that the right-hand-side variables are orthogonal to the stochastic disturbance. This assumption would be violated if regulators take into account the value of land when drawing the boundaries of critical habitat. However, the Services’ stated belief that critical-habitat designation has few economic implications suggests that land values play little role in determining which lands are included in critical habitat. Moreover, Sunding and Terhorst (2014) demonstrate that for a group of important endangered species in California—namely, those existing in vernal pool complexes—the U.S. Fish and Wildlife Service designated a significant amount of critical habitat right up to the urban edge—literally across the street from previously developed subdivisions and commercial centers.
To investigate the plausibility of the parallel-trends assumption underlying the difference-in-differences design, Figure 219 plots the (logged) per-acre sales price over time for properties within critical-habitat areas (dots) and outside of critical-habitat areas (crosses) for the 2001 and 2006 RLF designations (top and bottom panel), respectively. While the identifying assumption requires that the treated and control groups follow similar paths in the absence of treatment, we can examine this assumption’s feasibility by comparing the extent to which the predesignation trends are parallel (Angrist and Pischke 2009). The predesignation trends for the 2001 policy, depicted in the left panel of Figure 2, are nearly parallel,20 and the 2006 pretrends (Figure 2, right panel) are quite parallel.21 Figure 2 also foreshadows our main results: the overturned 2001 RLF designation had no detectable effect on designated property values, while the 2006 RLF designation significantly reduced designated property values.
Land-Sale Prices by Red-legged Frog Designation Status. 2001 Announcement (1993–2005) Log(Price per acre). Raw Data
For the BCB, we explicitly address potential endogeneity by exploiting characteristics of the species itself as a reasonable instrument for critical-habitat designation. As discussed in the introduction to this paper, we use serpentine soils as an instrument for critical-habitat designation. This instrument strongly predicts critical-habitat designation for the BCB, as the butterfly biologically relies upon serpentine soils for habitat. Further, serpentine soils are an exogenous and time-invariant characteristic of the landscape. For this instrument, we calculate a measure of serpentine soils for each parcel and then use this measure as an instrumental variable for treatment (critical-habitat designation for the BCB) in a two-stage least squares (2SLS) estimation framework. This identification strategy allows us to determine whether bias in our ordinary least squares (OLS) butterfly regression is likely attenuated to zero. Unfortunately, we lack a good instrument for the RLF’s critical-habit designation. However, we can compare our OLS estimates for the butterfly to the 2SLS estimates for the same species. If the coefficient on the policy variable moves significantly toward zero, we would be concerned that any estimated nonzero impact may be due to endogeneity.
To examine the potential underlying endogeneity in a difference-in-differences analysis of the BCB’s critical-habitat designation, Figure 322 illustrates the predesignation trends for properties eventually designated as critical habitat (dots) and for nondesignated properties (crosses). The two groups’ predesignation trends in Figure 3 appear quite parallel, further strengthening the credibility of the BCB analysis. In addition, the pre- and postdesignation trends in Figure 3 preview our BCB results: critical-habitat designation for the BCB significantly and substantially reduced designated properties’ values.
Land-Sale Prices by Bay Checkerspot Butterfly Designation Status, 2001 Announcement (1988–2007) Log(Price per acre), Raw Data
6. Estimation Results
Because there is no a priori guidance on the specific functional form of how our right- and lefthand-side variables enter equation [2], we use a Box-Cox transformation to select between a log-log, log-linear, or linear-linear specification. The Box-Cox parameter was most consistent with a log-linear specification.23
Red-legged Frog
Table 3 contains the results from estimating equation [2] for the RLF using the 26 counties that did or did not have any parcels sold in critical-habitat land before or after designation. Column (1) of Table 3 displays the coefficient estimates based on a simple OLS regression with the natural logarithm of the transaction price per acre as the left-hand-side variable—and using only vacant land classified as residential. In this basic specification we do not control for any fixed effects. The coefficients all have the expected signs. Parcels that are larger, hotter, wetter, and farther from highways or cities collect lower per-acre prices.
Red-legged Frog, Main Regression Results: All-Counties Sample
Column (2) adds to column (1) the policy variables for the RLF. The RLF2001 coefficient indicates that land affected by the 2001 designation prior to designation had a lower per-acre sales value. The coefficient of interest—the coefficient on Post2001 × RLF2001— is positive but insufficiently precise to rule out no effect. For the so-far-uncontested 2006 designation, the RLF2006 coefficient is positive and significant, indicating that land affected by the 2006 designation had a higher per-acre value prior to designation. The coefficient of interest for 2006, Post2006 × RLF2006, is negative and statistically different from zero at the 1% level. The point estimate of the coefficient is –0.748, which indicates a 53% drop in the per-acre price due to the 2006 designation. It is important to keep in mind that this effect is estimated off a fairly small sample—though the sign and magnitude are consistent with economic expectations.
Cross-sectional regressions—like the regressions in columns (1) and (2) of Table 3— can suffer from substantial omitted-variables bias. While we cannot control for fixed effects at the parcel level, we include three-digit zip code fixed effects (e.g., “945” if the parcel sold was in zip code 94595) in the regressions that produce columns (3)–(7) of Table 3. The only coefficient that changes considerably between columns (2) and (3) is RLF2006, which triples in magnitude (remaining positive and significant). This jump is not surprising given the number of additional fixed effects column (3) adds to column (2). The coefficient of interest is slightly larger in magnitude (–0.764), again indicating a large decrease in properties’ values when designated as critical habitat for the RLF in 2006. Column (4) adds month-of-sample fixed effects, and the coefficient of interest moves slightly toward zero, yet is still statistically significantly different from zero and economically significant. The estimate in column (4) indicates a 48% drop in per-acre sales value.
Column (5) of Table 3 estimates model 2 using nonresidential parcels, instead of the residential-parcel focus of columns (1)–(4). The regression specification underlying column (5) is identical to that of column (4), except column (5) adds a set of land use indicators (e.g., agricultural, commercial, industrial). The coefficients of interest for this non-residential sample do not differ statistically from zero for either year’s designation.
Column (6) returns to the residential sample and considers only parcels that sold two or more times during our sample period. We do not have enough multiple sales on critical-habitat land after designation to conduct a parcel fixed effect estimation, but this multiple-sales sample allows us to provide some evidence for whether the results are sensitive to looking at a balanced set of sales. The point estimate for this multiple-sale residential sample is very similar to those of columns (2)– (4), implying a 51% decrease in sales value. Finally, column (7) of Appendix Table A3 repeats the specification of column (5) using repeat sales for nonresidential properties. Column (7) bears no substantive changes relative to column (5).
Appendix Table A3 limits the sample to counties that experienced a designation, which reduces the sample size by about 15%. Despite this smaller control group, the point estimates in Appendix Table A3 are almost identical to the estimates in Table 3. Our best estimate— drawing upon the results from column (4) of Table 3—is that the RLF critical-habitat designation decreased residential properties’ values by approximately 48%. This estimate is valid for our sample of sales, and one should be careful when applying these estimates to other locations and/or species. As discussed above, we do not have a good instrument for the designation of RLF critical habitat.
Bay Checkerspot Butterfly
We now turn to the analysis for the BCB’s critical habitat. As discussed above, we constructed a dataset for two California counties substantially affected by critical-habitat designation: San Mateo and Santa Clara Counties.
Table 4 contains the estimation results for vacant-land sales from these two counties. In each regression (column) of Table 4 the outcome (dependent) variable is the natural logarithm of price per acre. Columns (1)–(3) estimate the parameters via OLS, and column (4) employs the 2SLS strategy discussed above. In the basic specification of column (1), we control for census tract and month-of-sample fixed effects, as well as lot size. The coefficient on lot size is almost identical to its value for the statewide regressions of Table 3 and Appendix Table A3. Column (2) augments the specification by adding the BCB policy variables. The coefficient of interest (Post2001 × BCB2002) is statistically significant and strongly negative, which is consistent with the previous analysis. Column (1)’s point estimate of -1.56 is interpreted as a 79% loss in per-acre value (price) due to critical-habitat designation for the BCB. The fact that the BCB effect is so much larger than the effect for the RLF is not surprising, due to the BCB’s considerably more stringent conservation requirements.
Bay Checkerspot Butterfly: Main Regression Results; Dependent Variable: Log(Price per acre), San Mateo and Santa Clara Counties
Column (3) of Table 4 further augments this specification by adding several observable land characteristics and 2001 RLF policy variables. Higher sales prices associate with parcels outside of floodplains, outside of wetlands, with shallow slope, and within urban growth boundaries. The point estimate for the policy variable of interest is essentially unchanged and implies a 78% lower per-acre sales price. The coefficients on the 2001 RLF designation are statistically insignificant, in keeping with our statewide results.
The estimates so far have not accounted for the potential endogeneity of the BCB’s critical-habitat designation. Column (4) instruments for the critical-habitat-designation variables using (1) the serpentine-lands indicator variable and (2) interaction between the serpentine-lands indicator and the postdesignation time dummy. We argue that these instruments are plausibly exogenous to land values, conditional on observables and the remaining included dummies. The F-statistic from the first-stage regression via a linear probability model indicates that the serpentine-soils indicator correlates with the critical-habitat designation indicator at the 1% level. Column (4) reports the coefficient estimate from the second stage of this 2SLS regression, which indicates that the estimated effect is still statistically significant and larger in magnitude than the estimated effect of columns (1)–(3), consistent with prior expectations. This result suggests that designating vacant land as critical habitat substantially reduces the value of the land from the market’s perspective.
Robustness
Because difference-in-differences designs can be sensitive to specification, we now examine our results’ robustness to specification (Imbens and Wooldridge 2009). Imbens and Wooldridge recommend that researchers examine their models’ and results’ sensitivities to specification by comparing normalized differences of observable covariates (between treatment and control). Imbens and Rubin (2015) suggest that a normalized difference greater than 0.25 (in absolute value) may indicate linear-regression methods that are sensitive to specification. We compute the normalized differences for each of our treatment/control pairs—the 2001 RLF, the 2006 RLF, and the 2001 BCB—and for each of their covariates. Appendix Tables A1 and A2 contain these normalized differences for the RLF and BCB, respectively. For the RLF, the largest (absolute) normalized difference is 0.426, for city distance in the 2001 designation. For the 2006 RLF designation, precipitation, temperature, and highway distance each exceed 0.25, ranging from 0.344 to 0.393 (in absolute value).24 For the BCB, the largest (absolute) normalized difference comes from the slope covariate.25
Imbens and Wooldridge note that normalized differences that exceed 0.25 may signal settings that are sensitive to specification. Thus, for each of the covariates that considerably exceed 0.25 (the distance and climate variables for the RLF and the slope variable for the BCB), we examine our results’ sensitivity to (1) these covariates’ specification and (2) our fixed effect specification.
To consider our RLF results’ robustness to specification, we focus on column (4) of Appendix Table A3, which estimates the effect of the RLF critical-habitat designation on residential land values, using only treated counties. Column (1) of Appendix Table A4 repeats this specification but drops all covariates (precipitation, temperature, distance to city, and distance to highway), and column (2) adds the covariates (replicating the original result). Adding the covariates drops the coefficient of interest (Post2006 × RLF2006) from -0.96 to -0.65. Column (3) replaces the three-digit zip code fixed effect with a city fixed effect, and the point estimate (-0.55) does not change substantively in magnitude or significance. Across a wide range of specifications for the distance variables, in columns (3)–(8), the coefficient of interest remains stable in magnitude and in precision. Specifically, column (4) specifies a quadratic polynomial for the distance variables, including a linear interaction between city distance and highway distance, and column (5) allows a quartic polynomial for the distance variables. Column (6) applies cubic splines, column (7) uses a semiparametric binned specification, and column (8) applies the same semiparametric specification but also includes city fixed effects. Appendix Table A5 repeats this robustness exercise for the climate variables in the RLF analysis. Across the 13 specifications of the two distance variables, the climatic variables, and the geographic fixed effects,26 the estimated effect remains quite stable—ranging between -0.66 and -0.52—and significant.27
Appendix Table A7 considers the degree to which our OLS and 2SLS results for the BCB depend upon the functional form of slope.28 Column (1) reproduces the original OLS result displayed in column (3) of Table 4. Column (2) allows slope to enter as a quartic polynomial, and column (3) specifies slope through a cubic spline. Throughout these three specifications, the estimated coefficient hardly changes, varying from -1.53 to -1.44. The 2SLS results in columns (4)–(6) display a similar stability, varying from -6.77 to -6.24.29 Thus, while the normalized difference in slope signals a potential for our results’ sensitivity to specification, we observe no evidence of this sensitivity.30
In summary, for both species that we consider, critical-habitat designation yields statistically and economically significant losses in land value within the transactions in our sample. These results demonstrate robustness across a number of covariate and fixed effect specifications. For the RLF’s designation, our results suggest a 47% loss in property value. For the BCB’s designation, our results imply a loss in value ranging from 78% (OLS) to a complete loss in land value (2SLS).
7. Conclusions
Despite the controversies surrounding the Endangered Species Act, there is surprisingly little empirical research on its impacts. Habitat protection is central to the aims of the ESA, and the act requires the federal government to designate lands as critical habitat for protected species soon after they are listed as threatened or endangered. Because critical-habitat designation can limit future development and increase costs, the price of designated land should adjust downward to capitalize the loss of future development profits.
This paper contributes to the literature on the economics of endangered-species regulation by measuring the impact of critical-habitat designation for two listed species on land prices in the market for vacant land in California. Using a quasi-difference-in-differences approach that controls for exogenous differences in parcels, we show that critical-habitat designation resulted in large and statistically significant decreases in the value of parcels sold. We opt for this approach because the perfect experiment—randomly assigned critical-habitat designation across a large number of parcels—does not (and will not) exist. In this paper, we measure the impact of critical-habitat designation on land values that includes selection of parcels into treatment.
Critical-habitat designation is the only instance in which the Endangered Species Act requires economic analysis. Indeed, the ESA prohibits the government from considering economic factors when making other decisions under the act, notably, including when listing species as threatened or endangered. This paper provides some of the first direct, market-based evidence that critical-habitat designation under the ESA imposes significant economic costs on landowners. The results also illustrate how federal land use controls interact with local regulations that, in turn, shape patterns of development. This finding suggests that the costs of habitat protection may be reduced by coordinating federal and local land use interventions.
More broadly, the results of this paper suggest that current interpretations/implementations of the ESA likely omit important economic costs generated by critical-habitat designation.
Acknowledgments
The authors would like to thank seminar participants at the University of Arizona, PERC, UCLA, UCSB, and UC Berkeley. We acknowledge helpful conversations with Andrew Plantinga, Robert Innes, and Dean Lueck. All errors in this manuscript are solely ours.
Footnotes
↵1 16 U.S.C. 1531-1544, 87 Stat. 884.
↵2 “Take” is broadly defined as any action causing harm to a listed species.
↵3 “NOAA” refers to the National Oceanic and Atmospheric Administration.
↵4 81 F.R. 7415.
↵5 There are other, more technical, reasons why critical habitat can result in reduced profits from future development and reduced land values as a result. The Services recognize these mechanisms, and we outline these mechanisms in Section 2.
↵6 66 F.R. 14626; 71 F.R. 19244; 73 F.R. 53492.
↵7 73 F.R. 50420.
↵8 Critical-habitat-designation status for the BCB (among other species) has often been argued to be endogenous. A valid instrument in this setting thus helps assuage concerns about bias from omitted variables.
↵9 P.L. 89-669.
↵10 Gifford Pinchot Task Force v. U.S. Fish and Wildlife Serv., 378 F.3d 1059, 1069 (9th Cir. 2004), amended, 387F.3d 968 (9th Cir. 2004).
↵11 81 F.R. 7415.
↵12 See U.S. Fish and Wildlife Service, Sacramento Office, https://www.fws.gov/sacramento/es/Critical-Habitat/Bay-Checkerspot-Butterfly/Current/ and https://www.fws.gov/sacramento/es/Critical-Habitat/CA-Red-Legged-Frog/Current/,_accessed January 22, 2008, and October 8, 2018.
↵13 U.S. Geological Survey, “National Wetlands Research Center Spatial Data and Metadata Server” available at http://sdms.cr.usgs.gov/.
↵14 Jones and Stokes Associates, “Serpentine Soils Layer,” May 30, 2008.
↵15 “BCB Critical Habitat” depicts the final area of the U.S. Fish and Wildlife Service’s designation of the critical habitat for the BCB. Similarly, “RLF Critical Habitat” illustrates the coverage area of the U.S. Fish and Wildlife’s designated critical habitat for the RLF. The white boundaries delineate counties’ borders.
↵16 In our case, the relevant rules are the 2001 and 2006 critical-habitat designations.
↵17 Also called month-year.
↵18 As we discuss below, our results are robust to the choice of geographic fixed effect.
↵19 Prior to the 2001 (top panel) and 2006 (bottom panel) RLF land designations (left of the dotted vertical line), sales prices for the eventually protected land (dots), and unprotected land (crosses) traveled similar trajectories (trends). Following the announcement of the 2001/2006 policy, land values (shown here as price per acre, logged) may have briefly declined (top panel) or followed similar trajectories (bottom panel) before returning to trends similar to those in the prepolicy period (top panel). The solid lines plot the linear trends for the two groups of properties, before and after the policy.
↵20 If anything, 2001-designated properties were gaining on nondesignated properties prior to 2001, which would bias our study against finding a negative effect of critical-habitat designation.
↵21 Appendix Figures A1–A3 repeat this exercise for both species with Winsorized data.
↵22 Prior to the 2001 BCB land designation (left of the dotted vertical line), sales prices for the eventually protected land (dots) and unprotected land (crosses) traveled similar trajectories (trends). Following the announcement of the 2001 policy, land values (shown here as price per acre, logged) may have precipitously declined for protected (critical-habitat-designated) properties. The solid lines plot the linear trends for the two groups of properties, before and after the policy.
↵23 In addition, we drop parcels of less than 1,000 square feet from the estimation, as these are too small to build a free-standing structure.
↵24 For the 2001 RLF designation, lot size (acres) also exceeds the 0.25 benchmark, with a normalized difference of 0.254.
↵25 Appendix Table A2 also shows a normalized difference greater than 0.25 for the serpentine variable. However, rather than being problematic, this “significant” difference implies a strong first stage in our 2SLS strategy. Lot size also narrowly exceeds 0.25, with a normalized difference of 0.294.
↵26 Ignoring the specification in column (1) of Appendix Table A4, which excludes all covariates.
↵27 Appendix Table A6 continues to analyze the RLF results’ robustness to fixed effect specification and potential heterogeneity. Across these specifications and subsamples, the estimated treatment effect remains fairly stable. In particular, the columns of Appendix Table A6 step through (1) zip code and month fixed effects; (2) zip code and city-by-quarter fixed effects; (3) zip code, city-by-quarter, and month fixed effects; (4) city, month, and zip-code-by-quarter fixed effects; and (5) city-by-month fixed effects. Columns (6a) and (6b) then split the sample into the two largest treated two-digit zip codes.
↵28 Recall that slope is the covariate for the BCB, whose normalized difference considerably exceeds 0.25.
↵29 Across analogous functional specifications of slope.
↵30 Matching-based estimators present a slightly different empirical strategy, but still centered on the assumption of conditional independence (Angrist and Pischke 2009). Appendix Table A8 compares the previous results with matching-based estimators and doubly robust estimators (regressions controlling for propensity score and enforcing overlap). The estimates of the effect of critical-habitat designation are generally of similar magnitudes, and their confidence intervals do not reject the previous point estimates.
In this issue
Jump to section
Related Articles
Cited By...
- Impact of Land Subsidence on Housing Sale Values: Evidence from the San Joaquin Valley, California
- The Trade-Off between Wildlife Conservation and Renewable Energy: Evidence from Eagles and Wind Turbines
- Distributional Impacts of Conservation on Land Prices: Evidence from Natura 2000
- Estimating the Impact of Critical-Habitat Designation on the Values of Developed and Undeveloped Parcels