Learning from Endangered and Threatened Species Recovery Programs: A Case Study Using U.S. Endangered Species Act Recovery Scores.

Joe Kerkvliet^†

Department of Economics

Oregon State University

Christian Langpap

Department of Economics

Tulane University

ABSTRACT

Threatened and endangered species recovery programs consume increasing resources in the United States and worldwide. Even so, there is increased concern about actual and projected biodiversity losses and in the success of recovery programs in reversing these trends. In this paper, we use a panel data set and ordered probit econometric methods to statistically examine the determinants of the 1990-2002 biennial U.S. Fish and Wildlife Service (FWS) recovery scores for up to 241 vertebrate species listed as threatened or endangered under the Endangered Species Act. We find evidence that species-specific spending is a significant determinant of species’ recovery scores and that increased spending reduces the probability that FWS will classify a species as extinct or declining. The evidence does not support the hypothesis that increased spending will increase the probability that a species is stable or improving. Other FWS management actions have significant and substantive influences on improved recovery scores. These include progress on or completion of a recovery plan and achievement of stated recovery objectives. We find evidence that species achieve better recovery scores if FWS considers them to have high recovery potential and that species whose recovery is judged by FWS to be in conflict with economic activity are more likely to be classified as extinct. We also report a new finding that recovery success varies across FWS administrative regions.

^†Corresponding author: Joe Kerkvliet, Professor, Economics Department, Oregon State University, Corvallis, Oregon, USA, e-mail: joe.kerkvliet@orst.edu.

DRAFT – August 2005. Do not cite without authors’ permission

Keywords: Endangered Species Act, endangered species, recovery plans, U.S. Fish and Wildlife Service, biodiversity, critical habitat.

Title: Learning from Endangered and Threatened Species Recovery Programs: A Case Study of U.S. Endangered Species Act Recovery Scores.

Introduction

Threatened and endangered (T&E) species recovery efforts have arisen worldwide to combat projected losses of biodiversity. In addition to identifying and legally protecting at-risk species, most such efforts provide for threatened and endangered species recovery programs (TESRP) (Foin, et al. 1998; Beissinger and Perrine 2001; Abbit and Scott 2001). These are directed toward increasing the populations of listed species, rather than just limiting their decline.

Improving T&E species’ prospects is costly. Worldwide TESRP expenditures on existing nature reserves alone are an estimated $7 billion annually (James et al. 1999), but Balmford et al. (2002) estimate that “an effective, global reserve program on land and sea” would cost $51 billion annually and even this expansion may fail to provide protection for many species (Rodriques et al. 2004).[1] Recent New Zealand TERSP expenditures are US$20-75 million annually (Cullen, et al 2005; Cullen, et al. forthcoming). From 1991-1999, the U.S. National Marine Fisheries Service spent $2.35 billion on Pacific salmon recovery programs (Baker 1999) and the U.S. Agency for International Development spent $1.27 billion on biodiversity programs in the world’s underdeveloped regions (U.S. Agency for International Development, 2003). Between 1989 and 2003, total species-specific spending on TESRP by state and federal agencies to comply with the Endangered Species Act (ESA) totaled at least $9.5 billion[2] (U.S. Fish and Wildlife Service (FWS) 2003).

Despite these expenditures, the ultimate goal of TESRP under the ESA, the delisting of the targeted species, has been reached in distressingly few cases (Rohlf 1991; Foin et al. 1998; Beissinger and Perrine 2001; Abbitt and Scott 2001). The resulting perceived lack of success has eroded political and popular support for the ESA. For example, a recent news release announcing a U.S. Representative’s sponsorship of proposed legislation to amend the ESA states, “…since inception, only seven species have been recovered out of the 1300 listed under protection of the ESA” (Walden, 2005).

The large expenditures and their apparent lack of success in promoting T&E species recovery, the emerging evidence of the importance of biodiversity in human economic well being (for example, see Balmford et al. 2002; Loomis and White 1996), and the prospect of learning as much as possible about effective TESRP from past and collective experience, justify analyses of TESRP and the determinants of their progress or stasis.

A number of existing analyses have focused on the characteristics of recovery plans for ESA-listed species (Tear et al. 1995; Elphick et al. 2001; Hoekstra et al. 2002), while others have analyzed the determinants of FWS allocations of TESRP resources (Metrick and Weitzman 1996; Metrick and Weitzman 1998; Simon et al. 1995; Dawson and Shogren 2001). While such analyses improve our understanding of forces shaping ESA implementation, they stop short of evaluating whether TESRP work and, if so, how.

Other researchers have estimated correlations between FWS’ management actions and TESRP outcomes (Rachlinski 1997; Beissinger and Perrine 2001; Abbitt and Scott 2001; Boersma et al. 2001; Miller et al. 2002).[3] All of these studies, to our knowledge, use simple bivariate correlation analysis and fail to simultaneously account for the many determinants of species recovery or the effects of species’ characteristics on recovery measures.[4] Taylor et al. (2005) use logistic regressions to examine the effect of time listed, critical habitat designation, and recovery planning on changes in species’ recovery status. Although, in contrast to previous studies, they do simultaneously account for some determinants of recovery, they only include a partial list of controls; most notably, they do not include spending allocations. Hence, although these studies provide useful insights, a comprehensive statistical analysis of the determinants of the ESA’s TESRP has not, to our knowledge, been conducted.

Several advantages may result from such an analysis (Backhouse et al. 1996; Clark 1996). First, we may find that the current focus on whether delisting occurs or not is shortsighted and incomplete. TESRP may improve T&E species’ prospects or slow the slide to extinction, without achieving the dichotomous delisted status. Marginal improvements, short of delisting, may be discovered and their drivers better understood. Second, the effects of both controllable and uncontrollable determinants of species recovery may be better identified and this knowledge may be used to emphasize effective management tools, direct the redesign or elimination of ineffective tools, and aid in the construction of future TERSP. Third, a comprehensive analysis can improve our understanding of the dynamics of species recovery, including process lags, large-scale patterns, interactions and uncontrollable drivers. Fourth, improved understanding of TESRP may help in prioritizing limited resource expenditures in directions promising the greatest return. Such triage is likely dictated by resource scarcity and it is best to perform it using all available lessons from experience (Clark 1996; Hughey et al. 2003).

In this paper, we conduct such an analysis using the T&E vertebrate species listed under the ESA, described as the “the most comprehensive legislation for preservation of endangered species ever enacted by any nation” (U.S. Supreme Court 1978). Specifically, we use a panel data set and ordered probit econometric methods to statistically examine the determinants of the 1990-2002 biennial FWS recovery scores for up to 241 vertebrate species. These data, while certainly imperfect, are the “best estimate(s) available” (Boersma, et al. 2001, p. 645) and we make the maximum possible use of them by analyzing biennial recovery scores for each species from 1990 or the first reported year until 2002, the last available report. We estimate the marginal contributions made to recovery scores of species-specific spending, other FWS management actions, and species’ characteristics.

In the next section, we provide a general model of species recovery, discuss the available data on FWS’ TESRP outcomes and define the covariates investigated. Section III specifies the econometric model. In section IV, we present the empirical results and discuss their implications. The final section offers a brief summary and conclusion.

II. Modeling the Recovery Process

T&E species recovery can be conceptualized as a production process; outcomes depend on a combination of factors, including inputs or resources allocated to the species, management decisions by FWS and other agencies, the current technology, and characteristics of the species. A model of recovery status can be specified in very general terms as S_it = f(r_it, m_it, c_it), where S_it is the recovery status of species i in period t, r_it measures resources allocated to the species, m_it is a vector of management variables, and c_it is a vector of species characteristics.

Ideally, we would have quantitative measures of S_it, such as population changes, reproductive success, etc. For most listed T&E species, however, these measures are lacking (Boersma et al. 2001). Although not ideal, the only broad-based measure of recovery status available over a time frame long enough to meaningfully discuss species recovery is the population status reported by the FWS[5]. Biennially, FWS classifies each listed species into one of seven categories: Extinct (E), Declining (D), Stable (S), Improving (I), Recovered (R), found only in captivity (C), and Uncertain (U) (U.S. FWS 1990a-2002a). Based on this classification, we measure recovery status as a discrete, ordered variable: STATUS_it = 0 if species i’s status at time t is E, STATUS_it= 1 if status is D, STATUS_it = 2 if S, or STATUS_it = 3 if status is I or R. We exclude status C because only two species are in this class, so the model cannot be estimated if it is included as a separate category. We set STATUS_it = 3 for species in both the I and R categories because only nine species have a status of R. We omit species with uncertain (U) status because of the difficulty of ordering uncertainty among the other categories in a meaningful way[6]. We have constructed a panel data set of recovery status and associated resource allocations, management factors, and species’ characteristics for as many as 241 vertebrates on the ESA’s endangered species list for the years t = 1990, 1992, 1994, 1996, 1998, 2000, 2002.

The focus independent variable in the model is r_it, the resources allocated to conservation of a species. Many researchers have emphasized the importance of adequate funding for ESA recovery efforts (Beissinger and Perrine 2001; Foin et al. 1998; Boersma, et al. 2001; see also Baker 1999). In addition, research suggests that active (and expensive) management is probably the most important determinant of recovery under our control (Foin et al. 1998; Rohlf 1991). Although habitat protection is often emphasized and is likely necessary for species recovery, it is not likely to be enough. For example, Foin et al. (1998) conclude that habitat protection will be insufficient for recovery of 63 percent of 305 listed species and even for the remaining 37 percent “… will not necessarily be cheap or easy. Habitat that is available and suitable may not be protected, dictating an expensive and often complex process of land acquisition. Even if habitat is protected on paper, enforcing protection…may be difficult and costly. … [E]ven if habitat is protected effectively, the species may need to be relocated there, or it may need to be bred to obtain the requisite numbers before reintroduction. …[T]he message is clear: some form of active management, costly in terms of both time and money, will be necessary to recover most threatened and endangered species” (p. 8, emphasis added).

In spite of the likely necessity of spending funds for recovery, previous research has found that FWS spending decisions do not necessarily follow the guidelines established under the ESA. Metrick and Weitzman (1996, 1998) and Simon et al. (1995) find that FWS spending choices are determined more by visceral than by scientific characteristics. Cash (2001) finds that funding decisions are driven by political variables. Dawson and Shogren (2001) find that spending is insensitive to yearly variations in time-variant factors (e.g. endangerment levels, economic conflict), but that time-invariant factors (e.g. historical use, cultural value, size, charisma) do matter. These results raise the question of whether spending, driven apparently by non-scientific factors, has a positive effect on species recovery. In fact, several observers have called for evidence that funding is effective in promoting species recovery (Baker 1999, Bean 1991; Clark 1996).

Indeed, at first glance it is not clear from the data that spending effectively promotes recovery. Table I shows the twenty species receiving the most federal and state spending and the changes in their STATUS between 1990, or the year of initial listing if earlier than 1990, and 2002. These 20 species represented 5.5 percent of listed species, yet commanded 55 percent of total spending up to 2002. Although some of these select 20 are doing noticeably better (red cockaded woodpecker, Colorado pike minnow, three salmon populations) others show little or no change, or even deterioration.

We measure species-specific SPENDING_it as total funds (in millions of U.S. $) expended by federal and state governmental agencies on TESRP, as reported by the FWS (U.S. FWS 1989-2002b). Agencies are required to report all “reasonably identifiable expenditures” that can be traced back to specific species[7]. This includes species-specific funding for items such as refuges, land acquisition, law enforcement, research, surveys, listing, captive breeding, reintroduction, recovery, and consultation. Opportunity costs such as the value of unsold energy resources or minerals, timber, or water are generally not included[8]. Thus, these reported expenditures understate the true resource costs of T&E species recovery. Nevertheless, as Metrick and Weitzman (1998) argue, they provide “the most direct and least noisy measure of preservation attention.” Given that these expenditures are reported on a yearly basis, but our data on recovery status is available only every two years, we construct SPENDING_it by adding the expenditures for the two-year period ending in year t (e.g. SPENDING_i1990 includes spending for 1989 and 1990) and all amounts to U.S. 2002 dollars using the Consumer Price Index.

The recovery status of a T&E species also depends on management actions taken by the FWS and other public and private agents to promote its recovery, including critical habitat designation, recovery plan preparation, and achievement of various recovery objectives. Section 4 of the ESA originally required that critical habitat be designated for each species listed. However, 1982 amendments require designation only to the “maximum extent prudent and determinable” and allow consideration of economic impacts. There is often a lag between a species’ listing and the corresponding designation of critical habitat. Such implementation considerations may impair the effectiveness of this management tool (Smith et al. 1993, Tear et al. 1995). To account for the effect of critical habitat designation, we include TIME HABITAT_it‡0, defined as the length of time, in days, that species i has had critical habitat designated at time t.

Section 4 of the ESA also requires the FWS to develop recovery plans for listed species, unless such plans would “not promote the conservation of the species.” Recovery plans identify management tasks and research needs, and define measurable criteria to determine when recovery objectives have been attained. Some species are exempt from preparation of a recovery plan if a state plan is used instead or the FWS considers the species probably extinct. To account for the effect of a recovery plan, we include two dummy variables: SOME PLAN_it = 1 if some level of recovery plan preparation has taken place by time t, but the plan is not yet complete (e.g. a draft or revision), SOME PLAN_it = 0 otherwise; and FINAL PLAN_it = 1 if a final plan has been approved by the regional director by time t, FINAL PLAN_it = 0 otherwise (the third possible category, no plan preparation by time t, is the reference category).

Achievement of recovery objectives reflects the success of FWS’ and others’ management activities (Abbitt and Scott 2001). To distinguish between species recovery plans and recovery objectives, one can think of the elaboration of a recovery plan as more of a formal requirement, whereas recovery objectives are specific goals set for “on the ground” management of a species. Furthermore, preparing a recovery plan is not a necessary condition for achieving recovery objectives. In 1996, for example, the recovery plan for the least bell’s vireo (Vireo bellii pusillus), listed in 1986, was incomplete, but between half and three quarters of its recovery objectives had been achieved (FWS 1996a). We define three dummy variables[9]: RECOVERY2_it = 1 if 25%-50% of recovery objectives for species i are achieved by time t and RECOVERY2_it = 0 otherwise, RECOVERY3_it = 1 if 50%-75% of recovery objectives are achieved, and RECOVERY4_it = 1 if 75%-100% of objectives are achieved (species with 0%-25% of objectives achieved are the reference category).

We control for a number of biological characteristics that may affect a species recovery status. Species recovery (or decline) is partly random, particularly in the case of small populations[10] (Gilpin and Soulé 1986) and some species may be more susceptible to population vicissitudes than others. The likelihood of extinction due to random events probably depends on population size, longevity, reproductive rate, growth rate variability, and population density (Shaffer 1981, Goodman 1987, Pimm et al. 1988). Data on most of these variables is unobtainable for most U.S. vertebrates. However, longevity and growth rate are highly correlated with body size, which is readily available (Cash et al. 1998). The precise relationship remains controversial (Pimm et al. 1988, Tracy and George 1992, Bennet and Owens 1997), but we rely on Johst and Brandl’s (1997) apparently unifying result of a U-shaped relationship between body size and the likelihood of extinction. We include BODYLENGTH and BODYLENGTH² as explanatory variables. Additionally, we include dummy variables to account for taxonomic differences: MAMMAL_i, BIRD_i, REPTILE_i, and AMPHIBIAN_i are set equal to one if species i corresponds to that taxonomic group (fish is the benchmark group). We also control for the distinctiveness of the species by including the variable DISTINCT_i = 1 if the species is monotypic (the only species in its genus) or belongs to a small genus (2–5 species), and DISTINCT_i = 0 otherwise. We control for the degree of endangerment of the species by including ENDANGERED_it, a dummy variable equal to one if species i is listed as endangered, rather than threatened, at time t. Finally, we control for a number of additional factors that may affect a species’ recovery status. Species that are in conflict with economic activity may be less likely to recover because FWS is less willing to take strong action or because opposition to TESRP is greater, so we include the dummy variable CONFLICT_it = 1 if species i is considered by the FWS to be in conflict with economic activity at time t. The FWS also classifies species according to their recovery potential, so we include HIGH POTENTIAL_it = 1 if the species is classified as having high recovery potential. To control for location-related features that could affect recovery, such as regional administrative differences or geographic or climatic factors, we include dummies for the FWS’s administrative regions (REGION1_i – REGION6_i, with region 7 as the reference).[11] Generally, the data are provided biennially by FWS (U.S. FWS 1990a-2002a). Biennial means for all the variables included in our model are provided in Table II.

III. Econometric Model

Given that our dependent variable, STATUS_it, is a discrete, ordered variable, the correct method of estimation is an ordered probit. Let be the true measure of recovery status:= β'x_it +α_i + ε_it, where x_it is a vector containing the variables described above and a constant, β is a vector of parameters to be estimated, ε_it ~ N[0,1] is an error term, and α_i is a an unknown parameter capturing unobserved species-specific effects. We cannot observe ; instead, we observe STATUS_it = j if μ_j_-1 < ≤ μ_j, j = 0, 1, 2, 3, where μ₀ = 0, μ₃= + ∞, and the remaining μ_j are parameters to be estimated. The corresponding probability that species i is in category j at time t is:

Prob[STATUS_it = j] = Prob[μ_j_-1 < β'x_it +α_i + ε_it ≤ μ_j]

= Φ(μ_j – (β'x_it +α_i)) – Φ(μ_j_-1 – (β'x_it + α_i))

where Φ(∙) is the c.d.f. for the standard normal distribution.

Estimation requires an assumption about the species-specific effect, α_i. The econometrics literature suggests that there is no consistent estimator of β for fixed-effects probit models, and hence the standard assumption is that α_i is an unobserved random variable, distributed normally with mean 0 and variance. A conditional maximum likelihood approach is used to estimate the parameters β, μ_j, and. Because the α_i are not observed, it is necessary to integrate them out to get the joint distribution of (STATUS_i1990, …, STATUS_i2002) conditional on x_it. Given the normal distribution of α_i, for each i this is given by

f(STATUS_i1990, …, STATUS_i2002|x_i; β, μ_j,) =

where is defined in and is the standard normal p.d.f. Taking the log gives the conditional log likelihood function for each i. The resulting log-likelihood function for the entire sample is then maximized with respect to β, μ_j, and (Wooldridge 2002, Hsiao 2003).

We estimate three specifications of the index function, equation , which differ in the way the focus variable, spending, is measured. In some cases, recovery actions are not likely to have immediate effects. For example, captive breeding, re-introducing individuals into unoccupied historical habitat, public education, and/or eradication of invasive species are unlikely to immediately affect T&E species populations. Hence, we specify spending in various lagged forms. Model 1 specifies spending as the amount spent in the immediate time period, Model 2 specifies it as the amount spent in the previous time period, and Model 3 specifies it as the amount spent two biennial periods prior to the immediate time period. The parameter estimates from Models 1-3 appear in Table III.

Ordered probit parameter estimates are somewhat difficult to interpret since the algebraic signs of the estimated coefficients unambiguously correspond to the signs of the marginal effects only for STATUS_it = 0 (opposite the sign of the coefficient) or STATUS_it = 3 (same sign as the coefficient) (see Long 1997 for details). The correspondence between the statistical significance of the coefficients and of the marginal effects is also ambiguous. We approximate this using the delta method (see Greene 2000 for details). Table III contains estimated marginal effects and indicators of statistical significance for each explanatory variable. These results are discussed in the next section.

IV. Results and Implications

The empirical results show remarkable consistency across the three models[12]. For the focus variable, SPENDING, the estimated coefficient is always positive and statistically significant. The estimated marginal effects suggest that an additional one million dollars in species-specific spending decreases the probability that a species is classified as extinct or declining (Status = 0 or Status = 1). The predicted changes in probability are quite small, less than one percent for STATUS = 0 and 1-1.5 percent for STATUS = 1, but statistically different from zero (a £ 0.05). In contrast, the estimated marginal effects of an additional million dollars show increases in the probability that a species is stable (STATUS = 2) or improving (STATUS = 3), but the estimates are very small and not statistically different from zero at conventional confidence levels. Overall the results suggest that ESA-related spending is more effective in preventing deterioration than in promoting improvements in recovery status.

The results from all three models also suggest the TESRP outcomes are not affected by the existence or duration of designated critical habitat. For all three models, the estimated coefficients for TIME HABITAT and the estimated marginal effects are extremely small and never statistically significant[13]. This finding is consistent with Belovsky et al.’s (1994) contention that habitat loss is generally not decisive in extinction when populations are already very small, and with the empirical results of Abbitt and Scott (2001), who find that the percentage of a species’ range included in recovery management activities is not a good predictor of recovery status. Moreover, Hoekstra et al. (2002) find few differences between recovery plans for species with and without critical habitat designations. However, this result does not agree with those of Taylor et al. (2005), who find that recovery trends are positively correlated with the duration of critical habitat designation, or Rachlinski (1997), who finds that species with critical habitat are less likely to be declining and more likely to be stable[14].

The coefficients for SOME PLAN are positive and statistically significant in all three models, while the coefficient for FINAL PLAN is positive and significant in Models 2-3. Using Models 2-3, the statistically significant marginal effects suggest that progress on recovery plans has a negative influence on the probability that a species is declining, with progress or completion of a plan decreasing the probability of declining status by 27-30 percent. SOME PLAN and FINAL PLAN also result in an estimated increase of 16-20 percent in the probability of improving status. Recovery plans have been strongly criticized. Tear et al. (1995) argue most plans fail to provide biological data essential for recovery decisions, including species abundance, demographics, and dynamics. Smith et al. (1993) argue that plans are not binding agreements, that often they are not implemented, and that they sacrifice good biology for economic considerations or fail to address political realities. The empirical results here suggest that, in spite of their shortcomings, the information and actions produced by recovery plans make positive contributions to assessed outcomes of TESRP.

Recovery objective achievement appears to strongly affect the outcomes. The estimated coefficients for RECOVERY 2, RECOVERY 3, and RECOVERY 4 are positive and statistically significant in all models. These results are consistent with Abbitt and Scott’s (2001) findings for a smaller sample. The relatively large marginal effects are statistically significant in many cases. For example, compared to a species with less than 25 percent of its recovery goals achieved (the reference), a species with 75 to 100 percent of recovery goals achieved is 38-39 percent less likely to be declining and 36-50 percent more likely to be improving. It is notable that the marginal effects show a monotonically increasing pattern for Models 1-3. That is, the higher the proportion of recovery goals achieved, the lower the probability that a species will be classified as declining and the higher the probability it will be classified as improving. About half of the marginal effects of recovery objectives are statistically different from zero ("† 0).

We find a negative, statistically significant estimate for the CONFLICT coefficient, while the CONFLICT marginal effect is significant only for STATUS = 0. This suggests that species whose recovery is in conflict with economic activity are about one percent more likely to be scored as extinct. Also, we find that species listed by FWS as having high recovery potential (HIGH POTENTIAL = 1) have greater probabilities of favorable recovery outcomes. In some cases, the statistically significant marginal effects are relatively large. A species with HIGH POTENTIAL = 1 is 7-10 percent more likely to be classified as improving and 14-20 percent less likely to be classified as declining. The coefficient estimate for DISTINCT is not statistically different from zero in either model, but some of the estimated marginal effects are statistically significant. Distinct species appear more likely to be scored as extinct or declining and less likely to be scored as stable, although the effect on the probability of improvement is not statistically different from zero. The ESA requires FWS to give priority to species with either CONFLICT, DISTINCT, or HIGH POTENTIAL equal to unity. Simon et al. (1995) find that CONFLICT and HIGH POTENTIAL are positively associated with FWS expenditures, while they find no statistically significant correlation for DISTINCT.

The estimated effects of FWS regions suggest intriguing questions. Consistent across Models 1-3, the statistically significant marginal effects for REGION 4 suggest that species managed by Region 4 (Southeast) have a 15-17 percent smaller chance of being classified as declining and 9-11 percent larger chance of being classified as improving. Conversely, we estimate a species managed by Region 1 has nearly a one percent greater chance of being extinct and a 4-5 percent small chance of being classified as stable. Whether these effects are due to uncontrollable variables, such as climatic conditions, or differences in the success of these regions in avoiding some of the organizational pitfalls of TESRP discussed in Clark et al. (1994) is a potentially important area for future research.

Turning to the biological variables, the results suggest that mammals, amphibians, birds, and reptiles are less likely to have favorable recovery outcomes compared to the reference category, fish. In many cases the marginal effects are statistically significant, with these taxa being more likely to be scored as extinct or declining and less likely to be scored as stable or increasing. Consistent with Johst and Brandl’s (1997) results, statistically significant estimates of the effects of BODY LENGTH and its square suggest a positive, but diminishing relationship between species size and recovery. The estimates for ENDANGERED are positive for the three models, but never statistically significant.

A potential concern with the results presented here is that the SPENDING_it variable may be endogenous in Model 1. We control for this possibility in three ways. First, note that we estimate Models 2-3 with lagged values of SPENDING, which is therefore exogenous in these models, and the results are qualitatively and quantitatively very similar to those from Model 1. Second, we conducted a Hausman endogeneity test by estimating a spending model, obtaining predicted values for spending, and including the difference between SPENDING_it and its predicted value as an additional regressor in the recovery model. We cannot reject the null hypothesis of exogeneity at a 5% confidence level. Third, we used the predicted values of SPENDING_it to implement an instrumental variables approach and estimated the recovery model using these predicted values, and the results are very similar to those from Model 1 as well. Finally, Miller et al. (2002) report that up to 75 percent of spending decisions are line items in appropriations legislation from Congress; unless the FWS is uncharacteristically influential in Congress, this suggests that spending decisions are mostly coming from outside the agency determining the STATUS of species.

V. Summary

We find evidence that species-specific spending is a significant determinant of species’ recovery scores and that increased spending reduces the probability that FWS will classify a species as extinct or declining. The evidence does not support the hypothesis that increased spending will increase the probability that a species is stable or improving. Other FWS management actions have significant and substantive influences on improved recovery scores. These include progress on or completion of a recovery plan and achievement of stated recovery objectives. We find evidence that species whose recovery is in conflict with economic activity are more likely to be classified as extinct and species with high recovery potential have 14 percent less chance of being classified as declining and 7 percent greater chance of being classified as improving. We also report a new finding that recovery success varies across FWS administrative regions. This result may be due to differences in organizational/management methods or in ecological factors.

Our results suggest that evaluating investments in TESRP programs, specifically in this case those initiated by mandate of the ESA, only by their success in delisting T&E species or improving their recovery status may in fact be shortsighted. Even though funds spent on T&E species may in general not lead to full recovery (and delisting), they seem to avoid further decline and eventual extinction. Hence, decreases in funding for TESRP like those experienced by the FWS in recent years (for instance, the projected budget for 2006 decreases funding for TESRP by nearly 10% relative to 2005 (Defenders of Wildlife 2005)) may well result in further declines in population status, and possibly extinction, for many species. Furthermore, our results indicate that the full effect of investments in T&E species probably occurs over a period of at least six years (recall that each period in our data encompasses two years) rather than exclusively in the current period; these lags should be considered when evaluating the outcomes of TESRP investments. Finally, our results suggest that it may be beneficial to emphasize recovery planning and completion of recovery objectives over other management tools that may not be as effective in promoting species recovery.

The recovery model developed in this paper and the results derived from it are a first step in developing a framework to evaluate investments in TERSP programs. This framework could be expanded to examine the effects of other aspects of FWS management of T&E species, such as Habitat Conservation Plans, or to evaluate different criteria to allocate scarce resources among a growing list of T&E species. We leave these questions for future research.
Literature Cited

R. J. F. Abbitt and J. M. Scott, Examining differences between recovered and declining endangered species, Conservation Biology 15(5), 1274-1284 (2001).

G. Backhouse, T. Clark, and R. Wallace, Reviewing recovery programs for endangered species: some considerations and recommendations, in S. Stevens and S. Maxwell (Eds.), Back from the Brink: Refining the Threatened Species Recovery Process, Surrey Beatty and Sons, New South Wales (1996).

B. Baker, Spending on the Endangered Species Act-to much or not enough, Bioscience, 49(4), 279 (1999).

A. Balmford, A. Bruner, P. Cooper, R. Costanza, S. Farber, R.E. Green, M. Jenkins, P. Jefferiss, V. Jassamy, J. Madden, K. Munro, N. Myers, S. Naeem, J. Paavola, M. Rayment, S. Rosendo, J. Roughgarden, K. Trumper, R. K. Turner, Economic reasons for conserving biodiversity, Science, 297 (5583), 950-953 (2002).

Bean, M. J. “Issues and controversies in the forthcoming reauthorization battle.” Endangered Species Update 9 (1, 2), 1-4 (1991).

S. R. Beissinger and J. D. Perrine, Extinction, recovery, and the Endangered Species Act, in J. Shogren and J. Tschirhart (Eds.), Protecting Endangered Species in the United States: Biological Needs, Political Realities, and Economic Choices, Cambridge University Press, Cambridge (2001).

G. E. Belovsky, J. A. Bissonette, R. D. Dueser, T. C. Edwards, Jr., C. M. Luecke, M.E. Ritchie, J. B. Slade, and F. H. Wagner, Management of small populations: concepts affecting the recovery of endangered species, Wildlife Society Bulletin 22, 307-316 (1994).

P. M. Bennett and I. P. F. Owens, Variation in extinction risk among birds: chance or evolutionary predisposition? Proceedings of the Royal Society of London, Biology 264, 401-408 (1997).

P.D. Boersma, P. Kareiva, W.Fagen, J.A.Clark, and J. Hoeskstra, How good are endangered species recovery plans? BioScience, 51(8): 643-650 (2001).

D. Cash, J. R. DeShazo, A. Metrick, S. Shapiro, T. Schatzki, and M. Weitzman. Database on the Economics and Management of Endangered Species (DEMES). Harvard University: Department of Economics (1998).

D. Cash, Beyond cute and fuzzy: science and politics in the U.S. Endangered Species Act, in J. Shogren and J. Tschirhart (Eds.), Protecting Endangered Species in the United States: Biological Needs, Political Realities, and Economic Choices, Cambridge University Press, Cambridge (2001).

T.W. Clark, R.P. Reading, A.L. Clarke (Eds.), Endangered species recovery: finding the lessons, improving the process, Island Press, Washington, DC (1994).

T.W. Clark, Appraising threatened species recovery efforts: practical recommendations, in S. Stevens and S. Maxwell (Eds.), Back from the Brink: Refining the Threatened Species Recovery Process, Surrey Beatty and Sons, New South Wales (1996).

R. Cullen, G.A. Fairburn, and K.F.D. Hughey, Measuring the productivity of threatened species programs, Ecological Economics, 30(1), 53-66 (2001).

R. Cullen, E. Moran, and K.F.D. Hughey, Measuring the success and cost effectiveness of New Zealand multi-species projects to the conservation of threatened species, Ecological Economics, 53, 311-323 (2005).

R. Cullen, K.F.D. Hughey, G.A. Fairburn, and E. Moran, Economic analysis to aid nature conservation decision making, Oryx, 39(3), forthcoming.

D. Dawson and J. F. Shogren, An update on priorities and expenditures under the Endangered Species Act, Land Economics 77(4): 527-532 (2001).

Defenders of Wildlife, The Bush budget: sacrificing tomorrow for today. Press release. http://www.defenders.org/releases/pr2005/facts05.html, accessed August 18, 2005.

C. S. Elphick, J.M. Reed, and J.M. Bonta, Correlates of population recovery goals in endangered birds, Conservation Biology, 15(5), 1285-1291 (2001).

T.C. Foin, S.P.D. Riley, A.L. Pawley, D.R. Ayres, Improving recovery planning for threatened and endangered species, Bioscience, 48(3), 177-184 (1998).

M.E. Gilpin and M. E. Soulé, Minimum viable populations: Processes of species extinction, in M.E. Soulé, (Ed.), Conservation Biology. The Science of Scarcity and Diversity, Sinauer Associates, Inc., Sunderland (1986).

D. Goodman “The Demography of Chance Extinction”. In Viable Populations for Conservation. M. E. Soulé, (Ed), Cambridge: Cambridge University Press (1987).

W.H. Greene. Econometric Analysis (4^th Edition), Prentice Hall: Upper Saddle River, NJ (2000).

L. Hatch, M. Uriarte, D. Fink, L. Aldrich-Wolfe, R. Allen, C. Webb, K. Zamudio, A. Power, Juristiction over endangered species habitat: the impacts of people and property on recovery planning, Ecological Applications, 12(3), 690-700 (2002).

J.M. Hoekstra, W.F. Fagan, and J.E. Bradley, A critical role for critical habitat in the recovery planning process? Not yet. Ecological Applications, 12:701-707 (2002).

K.D. Hughey, R. Cullen, and E. Moran, Integrating economics into priority setting and evaluation in conservation management. Conservation Biology, 17(1), 93-103 (2003).

C. Hsiao. Analysis of Panel Data (2^nd Edition). Econometric Society Monographs. Cambridge: Cambridge University Press (2003).

J.A. Jackson, The red-cockaded woodpecker recovery program, in T.W. Clark, R.P. Reading, A.L. Clarke (Eds.), Endangered Species Recovery: Finding the Lessons, Improving the Process, Island Press, Washington, DC (1994).

A.N. James, K.J. Gaston, and A. Balmford, Balancing the earth’s accounts, Nature. 40 (23), 323-324 (1999).

K. Johst and R. Brandl. “Body size and extinction risk in a stochastic environment.” Oikos 78(3), 612-617 (1997).

J. S. Long, Regression Models for Categorical and Limited Dependent Variables, Sage Publications, Thousand Oaks (1997).

J. B. Loomis and D. White, Economics benefits of rare and endangered species, Ecological Economics, 18, 197-206 (1996).

A. Metrick and M. L. Weitzman, Patterns of behavior in endangered species preservation, Land Economics 72(1), 1-16 (1996).

A. Metrick and M. L. Weitzman, Conflicts and choices in biodiversity preservation, Journal of Economic Perspectives 12(3), 21-34 (1998).

J.K. Miller, J. M. Scott, C.R. Miller, L.P. Waits, The Endangered Species Act: Dollars and sense? Bioscience 52(2), 163-169 (2002).

S.L. Pimm, H. L. Jones, and J. Diamond, On the risk of extinction, American Naturalist 132(6), 757-785 (1988).

J. J. Rachlinski, Noah by the numbers: an empirical evaluation of the Endangered Species Act, Cornell Law Review 82, 356-389 (1997).

A.S. Rodriques, S.A. Andelman, M.I. Bakarr, L. Boitani, T.M. Brooks, R.M. Cowling, L.D.C. Fishpool, G. A. B. da Fonseca, K.J. Gaston, M.Hoffman, J.S. Long, P.A. Marquet, J.D. Pilgram, R.A. Pressey, J. Schipper, W. Sechrest, S. N. Steward, L.G. Underhill, R.W. Waller, M.E.J. Watts, X. Yan, Effectiveness of the global protected areas network in representing species diversity, Nature, 428(8), 640-643.

D. J. Rohlf, Six biological reasons why the Endangered Species Act doesn’t work – and

what to do about it, Conservation Biology 5(3), 273-282 (1991).

M.L. Shaffer, Minimum population sizes for species conservation, BioScience 31(2), 131-134 (1981).

B. M. Simon, C. S. Leff, and H. Doerksen, Allocating scarce resources for endangered species recovery, Journal of Policy Analysis and Management 14(3): 415-432 (1995).

A. A. Smith, M. A. Moote, and C. R. Schwalbe, The Endangered Species Act at twenty: an analytical survey of federal endangered species protection, Natural Resources Journal 33, 1027-1075 (1993).

M. F. J. Taylor, K. F. Suckling, and J. R. Rachlinski, The effectiveness of the Endangered Species Act: a quantitative analysis, BioScience 55(4): 360-367 (2005).

T. H. Tear, J. M. Scott, P. H. Hayward, and B. Griffith, Recovery plans and the Endangered Species Act: Are criticisms supported by data?, Conservation Biology 9 (1), 182-195 (1995).

R. C. Tracy and T. L. George, On the determinants of extinction, American Naturalist 139, 102-122 (1992).

U.S. Agency for International Development. Biodiversity Conservation: A Report on USAID’s Biodiversity Programs in Fiscal Year 2002. Washington, DC (August 2003 [jk1] ).

U.S. Fish and Wildlife Service. Report to Congress on the Recovery Program for Threatened and Endangered Species. Years 1990 - 2002. Washington, DC (1990a, 1992a, 1994a, 1996a, 1998a, 2000a, 2002a).

U.S. Fish and Wildlife Service. Federal and State Endangered and Threatened Species Expenditures. Fiscal Years 1989-2003. Washington, DC (1989, 1990b, 1991, 1992b, 1993, 1994b, 1995, 1996b, 1997, 1998b, 1999, 2000b, 2001, 2002b, 2003).

U.S. Supreme Court, Tennessee Valley Authority v. Hill, 437 U.S. 153, 180 (1978).

G. Walden. Walden cosponsors bipartisan critical habitat legislation to update ESA. Press Release. http://www.house.gov/apps/list/press/or02_walden/pr_050315_CriticalHabitat.html, accessed June 23, 2005.

J.M. Wooldridge, Econometric Analysis of Cross Section and Panel Data. Cambridge: The MIT Press (2002).

TABLE I
TWENTY SPECIES WITH HIGHEST SPENDING
SPECIES	TOTAL SPENDING TO 2002 (millions of US$)	PERCENT OF TOTAL SPENDING	CHANGE IN STATUS BETWEEN 1990 or INITIAL LISTING AND 2002	YEAR OF INITIAL LISTING
Salmon, Chinook, Snake R. spring/summer run	201.94	7.16	Uncertain-Increasing	1992
Salmon, Chinook, Snake R. fall run	156.54	5.55	Uncertain-Increasing	1992
Salmon, sockeye	153.34	5.44	Uncertain-Uncertain	1992
Eagle, bald	121.50	4.31	Increasing-Increasing	1967
Owl, northern spotted	104.15	3.69	Declining-Declining	1990
Salmon, coho, OR, CA	103.26	3.66	Uncertain-Uncertain	1997
Woodpecker, red-cockaded	86.39	3.06	Declining-Increasing	1970
Tortoise, desert (Mojave)	76.36	2.71	Declining-Uncertain	1980
Salmon, Chinook, Sacramento R. winter run	60.02	2.13	Uncertain-Increasing	1990
Steelhead, Snake R.	58.90	2.09	Declining-Uncertain	1998
Manatee, West Indian (Florida)	57.85	2.05	Declining-Uncertain	1972
Salmon, Chum	53.50	1.90	Stable-Increasing	1999
Murrelet, marbled	46.05	1.63	Uncertain-Declining	1992
Bear, Grizzly or brown	44.04	1.56	Stable-Stable	1967
Steelhead, middle Columbia R.	41.13	1.46	Declining-Increasing	1998
Turtle, loggerhead sea	39.14	1.39	Declining-Stable	1978
Trout, Bull	38.28	1.36	Uncertain-Stable	1998
Salmon, Chinook, Columbia R.	38.02	1.35	Declining-Declining	1999
Sucker, razorback	37.50	1.33	Declining-Uncertain <