Short paper #1

PROTECTING SPECIES THROUGH LEGISLATION: THE CASE OF SEA TURTLES

MICHAEL BREI, AGUSTÍN PÉREZ-BARAHONA, AND ERIC STROBL

We evaluate the effectiveness of legislation in reducing the negative impacts of beachfront lighting on sea turtle nesting activity, one of the main threats to the species. To this end we construct a time varying index of ordinance effectiveness across Florida counties and combine this with loggerhead nesting data to create a panel data set covering a twenty-six year period. Our econometric findings show that such legislation can significantly increase nesting activity, where current levels of protection result in an addi- tional 34%. Using our estimates within a calibrated population model we also demonstrate that legis- lation can reduce the time to the animals’ extinction. Finally, when considering estimates of local willingness to pay for sea turtle preservation, we show that alternatively raising sea turtles in captivity under a head start program may be prohibitively expensive.

Key words: species protection, legislation, sea turtles.

JEL codes: Q57, O13, Q15, Q54, Q56.

Estimates suggest that up to 539 species have become extinct in the United States over the past 200 years (Sedjo 2008). Yet while there had been a growing awareness of the extinction threat to a number of prominent species since the turn of the nineteenth century, prior to the Endangered Species Act (ESA) in 1973 no gen- eral protective legislation had been put in place.1 The ESA potentially provides extensive protection for species listed, including protec- tion of critical habitat, implementation of a recovery plan, restrictions on take and trade, authorization to make land purchase or exchanges for important habitat, and federal aid to State and Commonwealth conservation

departments. As a matter of fact, at present over 2,400 species have made it on the ESA’s endangered species list, and current annual expenditure on their conservation are over US$1.5 billion.2 This begs the question as to how effective the ESA has been in terms of facilitating species recovery.

One has to recognize of course that in prac- tice implementation of an effective species recovery plan is not necessarily straightforward since habitats are not always easily defined, threats are often multifaceted, monitoring can be difficult, and implementation can be costly, both more generally and in terms of opportu- nity cost.3,4,5 A representative case in point of these challenges, and the object of our study

Michael Brei, LEM University of Lille. Agustín Pérez-Barahona, THEMA University of Cergy-Pontoise & École Polytechnique. Eric Strobl, University of Bern. Correspondence may be sent to: [email protected]

First published online by Oxford University Press on behalf of the Agricultural and Applied Economics Association.

1 Early calls for wildlife conservation emerged in the 1900s with the near extinction of the bison and the disappearance of the pas- senger pigeon. A number of specific legislation pieces followed, such as the Lacey Act of 1900, the Migratory Bird Conservation Act of 1929, and the Bald Eagle Protection Act of 1940. The first more comprehensive legislation passed was the Endangered Spe- cies Preservation Act of 1966, but while this act enabled the listing of native US animal species as endangered it provided very limited protection (see Roman 2011).

Amer. J. Agr. Econ. 102(1): 300–328; doi:10.1093/ajae/aaz025 © The Author(s) 2019. Published by Oxford University Press on behalf of the Agricultural and Applied Economics Asso- ciation. All rights reserved.

2 USFWS (2015). 3 See, e.g., Boersma et al. (2001), Boyd (2014), and EF (2016).

Kerkvliet and Langpap (2007) find no evidence of increased fund- ing or habitat designation aiding species recovery. However, in a subsequent study, Langpap and Kerkvliet (2012) show that habitat conservation plans do have a significant impact on species recov- ery, as long as these are not multispecies plans.

4 There is also some evidence that landowners preemptively destroy habitat to avoid potential land-use regulations; see, e.g., Lueck and Michael (2003) and Zhang (2004).

5 For instance, Borkovic and Nostbakken (2017) find that in Canada the oil leases that are regulated by species regulation lose 24% in value.

here, is the loggerhead sea turtle (Caretta car- etta). More specifically, sea turtles are threat- ened by a number of factors, including entanglement in fishing gear, poaching and ille- gal trade of eggs, meat, and shells, ocean pollu- tion, and coastal development;6 their population is widely believed to be decreasing at an alarming rate worldwide. A crucial part of the threat of coastal development to logger- heads is the presence of artificial lighting on their nesting beaches, where there has been considerable evidence that shows that artificial nighttime light deters sea turtle adults from nesting and disorients them (e.g., Witherington 1992; Johnson, Bjorndal, and Bolten 1996). Moreover, artificial lighting also increases the mortality rate of sea turtle hatchlings because it interferes with hatchlings’ ability to find their way from their nests on the beach to the sea (see, e.g., Tuxbury and Salmon 2005; Lorne and Salmon 2007). In this regard, Brei, Pérez- Barahona and Strobl (2016) find that coastal light pollution, through its effect on nesting activity, has substantially accelerated the poten- tial extinction of sea turtles in the Caribbean.

In terms of legal protection, loggerheads have been ESA listed and hence protected since 1978, with current annual expenditures of nearly US$9.5 million. Furthermore, in Florida, which hosts 90% of nesting activity in the United States, they enjoy additional protection under the 1995 Florida Marine Tur- tle Protection Act (MTPA).7 These legislative pieces specifically prohibit, among other things, the “take” of loggerhead turtles, where “take” includes their harassment and harm. As confirmed by a ruling of a federal appellate court in 1998, artificial light on beaches during their nesting period falls under this definition of “take” and hence can be viewed as prohib- ited by both the ESA and the MTPA (Barshel et al. 2014). Additionally, the Florida Depart- ment of Environmental Protection (DEP) has enacted rule 62B-55 F.A.C., setting forth a set of guidelines for local government regu- lations that control beachfront lighting to pro- tect nesting females and hatching sea turtles.8

However, importantly, the DEP sea turtle lighting rule does not require local govern- ments to legally adopt the proposed

guidelines. While presently most Florida coastal counties and municipalities have meanwhile adopted some form of beach light- ing ordinances, their ordinances differ widely not only in terms of their legislative details but also in their effectiveness of implementa- tion (Barshel et al. 2014). Moreover, data on sea turtle disorientation collected by the U.S. Fish and Wildlife Service (FWS) since 1987 suggests that nesting turtle disorientation has remained constant.9 This of course begs the question of how effective local legislation really is, particularly since there are other alternative strategies to aid recovery of the loggerhead population, such as captive rearing programs. In this paper we specifically investigate the

effectiveness of sea turtle lighting friendly (STFL) legislation in encouraging loggerhead sea turtle nesting in Florida, what implications this has had for the total Florida sea turtle pop- ulation, and what the monetary benefits of the implemented legislation have been. To this end we, in the spirit of Barshel et al. (2014), build a time-varying county level sea turtle friendly lighting ordinances index that takes the intricacies of the legislative pieces and their implications for sea turtle nesting into account. We combine this with annual local nesting activity and a rich set of controls across a sample of Florida coastal counties to create a twenty-six year panel data set, which we use to econometrically quantify the effectiveness of the ordinances. With this estimate in hand, we employ a calibrated population model for loggerheads to assess the effect of STFL ordi- nances on the evolution of their population over the long term. We find that the regula- tions can significantly delay loggerheads’ extinction. Our framework also suggests that head starting programs, as an alternative pol- icy, may be of very limited effectiveness in terms of reducing extinction and particularly expensive with respect to local estimates of willingness to pay (WTP). There are already a number of studies that

examine the effectiveness of species protec- tion legislation in aiding species recovery.10

The evidence in this regard is rather inconclu- sive. For instance, while Ferraro, McIntosch, and Ospina (2007) show that unless the ESA listing is combined with substantial funds listed

6 See http://www.seeturtles.org/sea-turtles-threats/ (accessed July 10, 2019).

7 See USFWS (2015). 8 Apart from loggerheads, there are four other sea turtle spe-

cies in Florida, although their nesting activity is minimal compared to the former.

9 Personal communication with Robbin Trindell, Florida’s Imperiled Species Management Plan.

10 See Langpap, Kerkvliet, and Shogren (2018) for a compre- hensive review.

Brei, Perez-Barahona, and Strobl Protecting Species through Legislation 301

species are likely to further decline, they also find that listing followed by funding has a pos- itive impact. Similarly, Kerkvliet and Langpap (2007) find no evidence of increased funding or habitat designation aiding species recovery, but a significantly negative impact on the prob- ability that species are classified as declining or extinct. While Gibbs and Currie (2012), in contrast, discover that the number of years listed, years with critical habitat designation, the amount of peer-reviewed scientific infor- mation, and funding did weakly increase recovery, these factors are only able to explain 13%. Also, Langpap and Kerkvliet (2010) dis- cover that spending increases the probability of being classified as stable and improving and decreases the probability of being catego- rized as declining or extinct. Finally, in a follow-up to their earlier study, Langpap and Kerkvliet (2012) discover some evidence that habitat conservation plans have a significant impact on species recovery, as long as these are not multispecies plans. Importantly, how- ever, all of these earlier studies pool data on species as well as the legislation and recovery plans in their analysis. However, not only do species differ widely in the nature of their hab- itat and the threats thereto but, as a perusal of current listed species shows, implemented recovery plans and legislation tend to be very species specific and intricate. The derived results are thus difficult to evaluate in terms of their effectiveness and their implications for policy since they provide only the average effect across species. By examining a single species and legislation in detail, as we do here, one is arguably able to infer much more pre- cise findings and subsequent recommenda- tions for species recovery.11

The remainder of the paper is organized as follows. Section 2 introduces the concept of STFL and the related regulations in Florida. We describe in section 3 the construction of our database, including details regarding the construction of our measure of the effective- ness of STFL ordinances in Florida. Section 4 provides the econometric analysis of the paper. In section 5 we study the effect of STFL ordinances on the loggerheads’ population dynamics, and in section 6 investigate some implications of actual estimates of WTP for

the protection of sea turtles in Florida. Section 7 concludes.

Sea Turtle Friendly Lighting (STFL)

Beaches are key for the survival of sea turtles. More precisely, although sea turtles spend only a very small proportion of their lifetime on beaches, these sites are fundamental to their reproductive phase, since females nest and hatchlings emerge on beaches. In addi- tion, their nesting behavior exhibits natal phi- lopatry, that is, females are likely to only nest on their natal beach.12

Given that sea turtle nesting and the emer- gence of hatchlings occurs almost exclusively at night, artificial beach illumination can dras- tically disturb the normal nesting behavior of adult females and hatchlings (see Raymond 1984; Witherington and Martin 1996; Wither- ington and Frazer 2003, among others). Importantly, adult turtles prefer to nest on unlit beaches. Moreover, nighttime illumina- tion fosters direct human disturbance of the nesting activity, frequently resulting in the abandonment or improper completion of nest- ing.13 Beach lighting interferes as well with female adults’ ability to correctly interpret physical cues that allow them to return to the safety of the sea after nesting. This disorienta- tion problem seems to be even more severe for the hatchlings (e.g., Witherington and Martin 1996), where the unnatural stimuli of the artifi- cial illumination can disrupt hatchlings’ instinctive sea-finding mechanisms, reducing their survival probability due to exhaustion, dehydration, and predation (Bustard 1967; Witherington and Martin 1996).

Preventing nightlight pollution on nesting beaches thus arguably constitutes an impor- tant component of the preservation of sea tur- tles. As noted in the introduction, the Florida DEP explicitly recognizes this and has conse- quently set (nonmandatory) guidelines for local government regulations to control beachfront lighting. The main focus of these guidelines is the adoption of STFL. Moreover,

11 Arguably, our paper also contributes to the literature on the impact of light pollution in general, in that we provide a frame- work with which to examine how legislation can be used to coun- teract its effect, at least in terms of species.

12 Sea turtles could in practice look for alternative nesting sites on neighboring beaches if the original site is no longer suitable (Worth and Smith 1976; Witherington and Martin 1996). How- ever, studies such as Brei et al. (2016) did not find significant empirical evidence of this.

13 Witherington and Martin (1996) also found that sea turtles discard their eggs in the ocean when they do not find an appropri- ate nesting beach.

302 January 2020 Amer. J. Agr. Econ.

it should be noted that the new lighting tech- nologies can additionally minimize the need for human behavior regulation, such as requir- ing residents to close their curtains or to turn off exterior lights during the nesting season.

The general principles of the DEP guide- lines can be summarized as “Keep it Low, Keep it Long, and Keep it Shielded.” “Keep it Low” refers to mounting the light fixtures low in order to minimize light trespass, using as well the lowest lumens output needed. Moreover, since sea turtles are very sensitive to the blue light (short wavelength),14 repla- cing the common blue lamps with long wave- length light sources (greater than 580 nm, that is, amber/red lamps) would significantly reduce sea turtles’ disorientation. This addi- tional principle is frequently known as “Keep it Long.” Finally, it is also advised to fully shield (“Keep it Shielded”) the lamps in order to eliminate point source light using, for instance, full cut-off fixtures (where no light is emitted above a 90� angle).

Following the DEP guidelines, local govern- ments establish STFL ordinances in order to induce appropriate lighting conditions for tur- tle nesting. In general, STFL ordinances con- sider two main dimensions. On the one hand, they set the appropriate photic habitat condi- tions of county/municipality’s nesting beaches. On the other hand, they include instruments to ensure the implementation of these condi- tions. Based on these two dimensions, we quantify the strength of local ordinances regarding the regulation of beach light pollu- tion. We provide a detailed description of this measurement in section 3.2.

Data and Summary Statistics

Loggerhead Nesting Data

There are currently two main loggerhead sea turtle nest-count surveys in Florida, namely, the Statewide program and the Index pro- gram.15 While the Statewide program is intended to be as complete as possible in terms of geographic coverage, it has not been consis- tent over time, adding beaches, changing boundaries, and changing the survey dates. In contrast, while not geographically exhaus-

tive, the Index program has been constant in effort and coverage over time. More specifi- cally, trained observers count and record nest- ing activity daily from the 15th of May to 31st of August, which represents most of the log- gerhead nesting season. These data are used as the main source for statistical assessments of temporal trends in loggerhead nesting in Florida (Witherington et al. 2009). For our analysis here we hence rely on the nesting data from the beaches covered under the Index program for estimation purposes but use the Statewide program beaches to make predic- tions regarding the population implications of the sea turtle lighting legislation on nesting activity for the wider Florida. The total num- ber of nesting beaches for Florida in 2014 are depicted in figure 1.16 Of these 214 beaches, 33 are covered under the Index program, pre- sented in red, and the rest under the Statewide program. We use the annual Index nesting data for

the period from 1989, its onset, until 2014, for the twenty-five years for which we have non- missing data.17 Summary statistics of the num- ber of nest counts (i.e., number of successful nests)18 of loggerheads for the Index beaches for our sample period are given in table 1. As can be seen, the average annual number of nest counts per beach is 1,468, although with considerable variability. Biologists point out the importance of pre-

serving nesting activity and, in particular, the number of nests for the protection of sea tur- tles. In this paper we consider a standardized survey covering the main nesting sites in Flor- ida, during a fairly long time span. This is in line with the recommended conditions for a database to study sea turtle populations (see, e.g., Eckert et al. 1999). Due to the significant monitoring required with the collection of the data, such surveys concentrate on the number of nests. In this respect our survey systemati- cally provides data about the number of suc- cessful nests. However, STFL ordinances might have improved other features of nesting activity, such as hatchlings disorientation. As noted in the introduction, nightlights can inter- ference with hatchlings’ ability to find their

14 See, e.g., Witherington (1992). 15 For a detail discussion about these surveys, see Witherington

et al. (2009).

16 The shapefiles for the nesting beaches were obtained from the Florida Fish and Wildlife Conservation Commission.

17 For the remaining three beaches, namely, Siesta Key, Egmont Key, and Dry Tortugas, there were only a few data points and hence we dropped these.

18 A successful nest is defined as a nest that results in successful oviposition (SWOT 2011).

Brei, Perez-Barahona, and Strobl Protecting Species through Legislation 303

way from their nests on the beach to the sea, thus increasing their mortality rate (see Peters and Verhoeven [1994], for a case study of two nesting sites on the Turkish Mediterranean coast). Unfortunately, the information avail- able is not comprehensive enough to examine this additional dimension of nightlight regula- tion. The population implications of our esti- mates (see section 5) should thus be interpreted as a lower bound of the effect of STFL ordinances.

STFL Ordinance Measure

While the DEP enacted the legislative rule “Model Lightning Ordinance for Marine Tur- tle Protection” in 1993, it did not make it man- datory for local governments to adopt the model. Rather sea turtle friendly lighting has been regulated at the county and/or munici- pality level only. In order to identify all

ordinances relevant to sea turtle friendly light- ing we started with the list of ordinances com- piled by the Florida Fish and Wild Conservation Commission (FWC)19 but com- pleted these with an exhaustive search of county and municipality level legislation.20

This resulted in a total of ninety-two coastal ordinances on sea turtles nesting activity, including their implementation date. The first of these was adopted in 1986.

Important for our analysis, following the STFL principles outlined in section 2, Barshel et al. (2014) introduced a method to measure the strength of local ordinances regarding the regulation of beach light pollution. It is based on an approach called Content Analysis, which aims at systematically quantifying the information included in texts such as media messages or legal documents (e.g., Krippendorff 2013). To this end the authors evaluate to what extent an ordinance sets appropriate beach lighting conditions for turtle nesting on two fronts. First, they assess the photic habitat conditions provided by the legislation using seventeen statements termed the STFL Principles Component. Second, they consider whether the ordinance includes appropriate legal devices in order to ensure the implementation of these conditions in terms of a further nine statements, labeled Implementation Component. For comprehen- sive lists of both sets of statements see Appen- dix A. For instance, the statement “Exterior artificial light for existing development must be long wavelength (i.e., 580 nm or greater)” (item 5 of STFL Principles Component) objec- tively defines the wavelength for exterior illu- mination, while the statement “Is a provision made for a compliance inspection during the nesting season?” (item 1 of Implementation Component) requires the provision of appro- priate means to monitor the installation of this favorable type of beach lighting.

To rate each ordinance in terms of its effec- tiveness of ensuring sea turtle friendly lighting we follow the approach by Barshel et al. (2014) closely. More precisely, an ordinance is evaluated on two specific scales. Regarding the Principles Component statements, each of the seventeen listed statements is compared to statements in the ordinance. We then assign, for each listed statement, a scale that

Table 1. Descriptive Statistics

Variable Mean St.Dev. Min. Max.

NL 5.48 10.72 −0.11 66.11 NESTS 1468 3231 0 23712 SCORE 16 23 0 83 ROOMS [d = 100m]

48 253 0 2490

ROOMS [d = 200m]

212 927 0 7787

INCOME/CAP 40 12 15 80 NOURISHMENT 2 24 0 829 STORMS 0.63 1.52 0 12 DEMOCRATS 0.63 1.52 0 12

Note: NL ≡ intensity of nightlights in 2014; NEST ≡ number of nests; SCORE ≡ legislation score; ROOMS ≡ number of rooms within d meters of the shoreline; INCOME/CAP ≡ county income per capita (2014 US dollars); NOURISHMENT ≡ average annual volume (cubic yards) of sand placed on nesting beaches; STORMS ≡ number of storms that affected a beach in a given year; DEMOCRATS ≡ share of registered democratic voters.

Figure 1. Nesting beaches surveyed under the index program

19 http://myfwc.com/conservation/you-conserve/lighting/ ordinances/ (accessed July 10, 2019).

20 County and Municipality legislation can be found on library. municode.com.

304 January 2020 Amer. J. Agr. Econ.

takes four possible values: 0 ≡ concept not mentioned in the ordinance; 1 ≡ concept vaguely addressed in the ordinance; 2 ≡ con- cept addressed in the ordinance but in a less stringent manner than what is required by the statement (i.e., wording provides loop- holes); and 3 ≡ concept addressed in the ordi- nance with the same strength as in the statement. With respect to the Implementa- tion statements, we do a similar comparison of the ordinance to each of the 9 corresponding statements. This time the scale assigns 1 if the concept is addressed in the ordinance or 0 oth- erwise. For each group of statements the scores were summed and normalized to have a maximum value of 50, and then these two sums were added to provide a total score that ranges between 0 and 100 for each ordi- nance.21 For each of the nesting beaches, we determined which municipality and county they were located in and assigned to them the highest score of the two for each year, tak- ing changes over time into account in doing so. In cases where beaches crossed county or municipality borders we weighted the scores according to the length of the beach located within them.

Figures 2 and 3 depicts the distribution of the legislation score corresponding to each nesting beach at the end of our sampling period, 2014. Accordingly, there is clearly sub- stantial spatial heterogeneity with regard to sea turtle friendly legislation. From table 1 one can see that the average ordinance effec- tive proxy, SCORE, in the Index beaches is somewhat smaller (16) than the total sample (42), but both have considerable variation. The maximum score observed over our sam- ple period is 83.

Verification of the ordinance score proxy. Given that the creation of our ordinance score variable is based to some extent on a subjec- tive evaluation of the legislation involved, it is important to verify that it actually captures what it is supposed to, namely, the variation in the effectiveness in ensuring STFL during nesting season. To this end Anderson et al. (2013) identified a number of beaches in Flor- ida with and without lighting ordinances and compared lighting intensity in months during the nesting season to the months outside of the nesting season using nighttime luminosity

data derived from satellite imagery.22 Their analysis, employing difference in means tests, suggested that almost all beaches with ordi- nances in their sample seemed to be in compli- ance with the legislation in the sense that nightlight intensity was significantly lower dur- ing nesting season. We extend their approach here to allow for ordinance effectiveness, rather than just incidence, and econometri- cally test any differences across beaches using the following specification:

ð1Þ NLimt = α + ηSCOREimt + βSCOREimt × MAYNOVi + γMAYNOV

+ ζX + πm + λt + μi + eimt,

where NL is the intensity of nightlights on beach i in month m of year t, SCORE is the nesting friendly legislation score, MAYNOV

Figure 3. Distribution of SCORE

Figure 2. Score of nesting beaches

21 We provide, as an illustration, a specific detailed example in Appendix A.

22 The available information about the ordinances does not allow one to distinguish between the induced changes only imple- mented during the nesting season and the ones that remain beyond May to November.

Brei, Perez-Barahona, and Strobl Protecting Species through Legislation 305

is an indicator variable capturing the joint effect of the May through November months, that is, the official nesting season when ordi- nances are generally in effect, X is a vector of control variables, π is a vector of monthly indi- cator variables controlling for monthly differ- ences in nighttime brightness independent of lighting ordinance effectiveness, λ is a vector of yearly indicator variables capturing yearly effects common to all beaches, μ is a set of beach specific indicator variables capturing time invariant beach specific factors, and e is an error term. Given the likely serial and spa- tial correlation of the data we calculate Dris- coll and Kraay (1998) standard errors for equation (1). To proxy local nightlight intensity, NL, we,

as in Anderson et al. (2013), use the Visible Infrared Imaging Radiometer Suite (VIIRS) monthly nightlight imagery collected and pro- cessed from the Suomi National Polar- Orbiting Partnership (SNPP) satellite since April 2012. These processed data provide measures of nightlight intensity across the globe at roughly 1h30 in the morning at a res- olution of roughly 500 m.23 We use images over the three-year period 2013–2015 in order to have three complete annual cycles. To cap- ture the monthly nighttime brightness on all nesting beaches, both Index and non-Index, we took the nesting beach polylines in figure 1, created 250 m buffers around these, extracted the nightlight grid cells from the monthly VIIRS, and averaged these for each month for each beach. As can be seen from table 1, the average (pseudo) radiance value, given in watts per steradian per square meter, varies considerably across beaches. We also depict the evolution of the average monthly nesting beach nightlight intensity over our sample period in figure 4, along with the difference in this between beaches with and without STFL lighting ordinances. Accordingly, there is clearly seasonal variation in nighttime brightness of the beaches, although relative to the standard deviation shown in table 1, this is not particularly large. More specifically, the brightest period is around March, after which the level of brightness falls until around September when it noticeably rises again. Arguably there are potentially two factors underlying these trends. First, this cyclical

pattern is certainly due in part to the nature of the tourist season in Florida. For instance, in the subtropical south of Florida the peak season roughly spans from mid-December until mid-April. In terms of estimating equa- tion (1), one should note that the monthly indi- cators π should control for this seasonality across years. A second reason is that sea turtle nesting friendly lighting legislation during nesting season existent on some beaches may also play a role. This is, however, not discern- ible by comparing the trends in beaches with and those without any legislation since the red line in figure 4 depicting the difference in nightlight intensity between legislation and nonlegislation beaches is positive, suggesting that the extent of legislation, rather than the incidence, may instead play a greater role.

In terms of estimating equation (1) we first start off with modeling μ as a beach specific random effect, and include, in addition to monthly and year dummies, the number of hotel rooms in 2013, average annual number of storms since 1989, average annual beach nourishment since 1989, and average income per capita in 2013 as controls in X.24 As can be seen in table 2, SCORE only has an effect on beach nightlight intensity during the official nesting season, May to November. More spe- cifically, beaches with greater STFL legislation display lower light pollution during this period. This effect is, as shown in the second column, only linear.

We next experiment with modeling μ as a time invariant fixed effect. One should note that in this regard, that while there were

Figure 4. Monthly nightlight intensity on nesting beaches

23 The fact that light intensity is only measured at 1h30 in the morning possibly means that using it as a proxy may entail missing some of the variation in nightlights if some areas reduce nightlights after a certain hour regardless of legislation.

24 Unfortunately we do not have monthly data for these controls.

306 January 2020 Amer. J. Agr. Econ.

considerable changes of legislation over the sample period of our nesting data, there were no changes over the three year period of the nightlight data, 2013–2016, so that SCORE itself is actually time invariant for these years. Its effect is thus absorbed by the beach fixed effects. The results of estimating equation (1) are depicted in the third and fourth column of table 2. Accordingly, one can decisively reject the null hypotheses that time invariant beach specific effects have no influence on differences in nightlight intensity across beaches. Again, we find a negative and statistically significant coeffi- cient on the SCORE × MAY NOV interaction term, but no non-linear effect. Taking the coef- ficient at face value, the estimate suggests that a one point increase in SCORE reduces night- light intensity on a nesting beach by 0.1% on the average lit beach. If we take our average (max) legislatively protected nesting beach rela- tive to those that have no legislation in place, then the relative reduction in nightlight intensity on the average beach would be 5.2 (12)%. Overall, while our lack of other monthly time varying controls does not allow us to decisively conclude that the relationship between night- light light intensity and lighting ordinances is causal, it is certainly suggestive of this and gives us some confidence that our proxy is a reason- able measure of ordinance effectiveness.

Other Determinants of Nesting Activity

Given the likely nonrandom nature of the location of sea turtle lighting friendly

legislation, it is of course of crucial importance to ensure that in our empirical estimation we capture all determinants of sea turtle nesting that may be correlated with the implementa- tion of the regulatory framework across time and space. Our strategy to this end was to extensively survey both the economic and noneconomic literature for factors that have been shown to affect or are correlated with nesting, and create indicators of these.

Hotels. Florida is well known for its attractive beaches. In this regard, guest accommodation near the beachfront is often particularly val- ued, and unsurprisingly Florida has seen a surge in hotel construction over the last few decades. However, hotels constructed close to the waterfront may, apart from causing lighting at night, also have other adverse impacts on sea turtle nesting activity that one would want to control for. For one they may cause coastal squeeze and thus reduce the nat- ural habitat of sea turtles.25 Additionally, the inherent tourist activity surrounding hotels may act as a direct disturbance to sea turtle nesting (see Davenport and Davenport 2006). Our data source for hotels within prox- imity of a beach is the SRT Share hotels data- base. More specifically, the SRT lists all hotels, existing and closed, since 1987, including the

Table 2. Nightlight Regression

(1) (2) (3) (4)

SCORE 0.1075 0.0034 (0.0940) (0.2985)

SCORE × MAYNOV −0.0105** 0.0084 −0.0105** 0.0084 (0.0041) (0.0123) (0.0042) (0.0125)

MAYNOV −0.1278 −0.3018 −0.0606 −0.2345 (0.2722) (0.2924) (0.2307) (0.2548)

SCORE2 0.0015 (0.0041)

SCORE2 × MAYNOV −0.0003 −0.0003 (0.0002) (0.0002)

Model RE RE FE FE Observations 1188 1188 1188 1188 (within) R2 0.087 0.089 0.063 0.065 Fðμi 6¼ 0Þ … … 2130*** 2124*** Note: (a) Estimator: Fixed Effects (FE) Linear Estimator; (b) All time invariant factors are purged from equation (1) via the FE estimator; (c) Driscol and Kraay (1998) standard errors in parentheses; (d) (within) R2 is the percentage of within beach explained variation; (e) ***, **, and * indicate 1%, 5%, and 10% significance levels, respectively; (f) Fðμi 6¼ 0Þ is an F-test of the beach specific effects μ being jointly equal to zero.

25 See, e.g., Mazaris, Matsinos, and Pantis (2009) for an analysis of coastal squeeze on sea turtles.

Brei, Perez-Barahona, and Strobl Protecting Species through Legislation 307

exact starting year, number of rooms,26 and latitude and longitude of location, among other things. It thus allows us to create a time varying measure of hotel capacity for each nesting beach within a chosen threshold of proximity. We as a benchmark consider hotels within 100 meters of the shoreline but also experiment with setting this threshold further away. As can be seen from table 1, across our nesting beach sample there is, on average, a hotel room capacity of about forty-eight rooms but with considerable variation.

Income per capita. The wealth of the local com- munity of a beach may also play a role in nest- ing activity in that wealthier communities are more likely to be environmentally friendly and hence supportive of sea turtle friendly lighting. For instance, Jin et al. (2010) showed that income played an important positive role for the valuation of sea turtle conservation in Asia. To capture this aspect we use the most disaggregated measure of local income avail- able for Florida, that is, county level local per- sonal income from the BEA Regional Economics Accounts. We normalize this series by county population and deflate it to be in thousands of 2014 US dollars. Each nesting beach is assigned the income per capita series of the county that it is located in. For those bea- ches that stretch across more than one county we use a beach length weighted average of the county level deflated per capita data. Table 1 shows that the average income in the counties containing the sample nesting beaches is about $US 40,000, with some beaches located in counties with an income double this value.

Beach nourishment. Beach nourishment, the human replacement of lost sand on beaches, which is a common practice in Florida, may also have an effect on sea turtle nesting. For instance, Rumbold, Davis, and Perretta (2001) find that in the first season after a beach nourishment pro- ject was implemented on Palm Beach (Florida) the number of loggerhead nests fell significantly, with a smaller effect in the second season. To take account of beach nourishment we construct annual time series of beach nourishment

projects using information from the Strategic Beach Management Plan (SBMP) reports. More specifically, the SBMP reports, one for each of Florida’s seven regions,27 serves as an inventory of Florida’s strategic beach manage- ment areas located on the Atlantic Ocean, Gulf of Mexico, and Straits of Florida. Importantly, these reports contain for each beach a detailed account of beach nourishment projects over time, including the year, the volume of sand used, and the location of the segment(s) of beach treated in terms of the nearest starting and nearest ending beach range monument.28

We allocated projects to the nesting beaches if their segments fell within the stretch of a beach. In the case where the segment(s) was not completely contained within a nesting beach, we allocated the proportion of volume equiva- lent to the proportion of segment of the project intersected by the beach. Table 1 shows that the average annual volume of sand placed on nesting beaches is about 2 million cubic yards.

Storms. Another factor potentially affecting sea turtle nesting are storms causing beach erosion. In this regard, Houtan and Bass (2007) used monitoring surveys over a ten- year period from the Dry Tortugas National Park to show that the incidence of tropical cyclones decreased the number of loggerheads substantially. In order to capture the potential role of storms we use the SBMP reports which list for each region all known damaging storms as well as the beaches affected. To take account of their effect, we simply use the total count of storms, both tropical storms and east- erlies, that affected a beach in a given year. The summary statistics in table 1 reveal that on average a beach is affected by roughly one storm per year, but that some have been sub- stantially more affected than others.

Econometric Analysis

Nesting Activity Regression

In order to investigate the impact of sea turtle friendly lighting legislation on loggerhead nesting activity we estimate the following:

26 For a few hotels there was no information with regard to the number of rooms. For these we estimated the number of rooms by using the mean number of rooms of hotels by type of operation (chain management, franchise, or independent), type of location (airport, interstate, resort, small metro/town, suburban, or urban), type of price (budget, economy, luxury, midprice, or upscale), class (economy, luxury, midscale, upper midscale, upper upscale, or upscale), and whether it had a restaurant.

27 These regions are the Northeast Atlantic Coast, Central Atlantic Coast, Southeast Atlantic Coast, Florida Keys, Southwest Gulf Coast, Big Bend Gulf Coast, and Panhandle Gulf Coast.

28 The range monuments are FDEP beach reference points, spaced offshore approximately 1,000 feet apart, and are typically known as “R” stations.

308 January 2020 Amer. J. Agr. Econ.

ð2Þ NESTSit = α + Xn j = 1

βjSCORE j it + λCit

+ πt + μi + eit,

where NESTS are the number of loggerhead nests found on Index beach i in year t, and SCORE is our legislation scoring variable, possibly including higher order terms. The vector π are a set of yearly indicator variables capturing year specific shocks affecting all nesting beaches, μ are time invariant beach specific effects, and e is the error term.

The dependent variable in equation (2), NESTS, is by nature a count variable and hence standard linear regression methods are not appropriate. We thus instead use a count variable estimation model. In this regard, the two most common choices are the Poisson and Negative Binomial model, where the lat- ter is preferred if there is over-dispersion. As table 1 shows, this is indeed true in our case, and we therefore employ a negative binomial count model. In those specifications where we take account of the beach specific time invariant effects, μ, we run the fixed effects version of this model (see Cameron and Tri- vedi 2013).29

Our aim is to identify the causal effect of SCORE on nesting activity. Clearly, given that there a number of environmental and eco- nomic factors that are likely to affect NESTS but are also correlated with SCORE, simply regressing NESTS on SCORE is likely to fail to do so. Incorporating beach fixed effects μ will take account of any potentially problem- atic time invariant unobservables, such as unobserved beach characteristics, that are beneficial to nesting but also may attract tour- ist activity. To also control for beach specific time varying factors that may affect both, our vector C consists of our other control vari- ables, which, as outlined earlier, are the known relevant determinants of nesting as identified from the existing literature. More specifically, we include the number of hotel rooms, ROOMS, county level income per capita, INCOME / CAP, the volume of sand placed under beach nourishment projects, NOURISHMENT, and the number of storms,

STORMS. Arguably, with this rich set of con- trols we are likely to be capturing all time vary- ing beach specific factors relevant to sea turtle nesting activity that may also be related to the extent of sea friendly lighting legislation at a beach. One may also worry about possible simulta-

neity of the passing of ordinances, or changes therein, and nesting activity. In particular, if nest counts at a beach indicate a potential fall in nesting activity for that season a ordinance may be passed to counteract such a decrease. There are a number of reasons why in reality this is unlikely to be a concern. First, nest counts are only collated at the end of the sea- son. Second, passing an ordinance can be a tedious and long process.30 For instance, at the municipality level the typical procedure involves drafting a proposal that needs to be submitted to the city council, which may then consult on the content through a special com- mittee. Often there is also a public hearing, which may result in further proposed changes to the legislation. Finally, after potentially sev- eral revisions and repetitions of this proce- dure, the ordinance must be voted on by the council and possibly the city major, and then takes effect at a specified later date. This whole process is likely to take some time and almost certainly would not be completed within a nesting season.31

Regression Results

The results of estimating variations of equa- tion (2) are provided in table 3. In the first col- umn we ran a negative binomial regression of NESTS on SCORE without controlling for any covariates apart from yearly indicator var- iables. As can be seen, the coefficient on the legislative variable is positive and significant, indicating that sea turtle protective legislation has acted to increase nesting activity of logger- heads. When we subsequently allowed for beach specific time invariant differences by using the fixed negative binomial estimator in the second column, the coefficient was reduced by nearly 90%. This indicates that

29 One should note that the fixed effects in Cameron and Tri- vedi (2013) estimator are conditional fixed effects and thus not strictly equivalent to the inclusion of panel unit dummies, which would produce inconsistent estimates; see Allison and Waterman (2002). For convenience sake we nevertheless simply refer to these as fixed effects.

30 See http://www.statescape.com/resources/local/ordinance- process.aspx (accessed July 10, 2019).

31 To further check this we have also investigated whether there is autocorrelation in 2. More specifically, we implemented Wool- dridge’s (2002) test using logged NESTS (since the test is designed for a linear panel model). The resultant test statistic of 2.163 was insignificant, and thus we could not reject the null hypothesis of no autocorrelation in 2.

Brei, Perez-Barahona, and Strobl Protecting Species through Legislation 309

there are important time invariant differences across beaches that are positively correlated with both legislation and nesting activity and not taking account of these would severely upwardly bias the estimated effect of legisla- tion on nesting. We thus for further estima- tions proceeded to rely on the fixed effects negative binomial model. In the third column of table 3 we include our

vector of controls, C, as outlined above. As can be seen, their inclusion only slightly changes the coefficient on SCORE after controlling for beach time invariant effects. In terms of their role in affecting nesting activity, we find that more hotel rooms located near the beach lowers the number of loggerhead nests, whereas beaches located in richer counties tend to have more nesting activ- ity. In contrast, neither beach nourishment pro- jects nor the number of storms have significantly affected loggerhead nesting. To investigate whether hotels further than 100 meters from the shoreline might still have an impact on nesting we increased the threshold d to 200 meters, where the results of including this alternative proxy are depicted in the fourth column. However, the lower threshold renders the coefficient on ROOMS insignificant. We also explored whether the fact that we have not found any significant role for beach nourishment pro- jects and storms is because we were assuming only a contemporaneous effect by including their lagged values in the fifth and sixth columns. However, as can be seen, this does not change our conclusion regarding their lack of impor- tance in discouraging loggerheads to nest on a beach. Since remigration, when it occurs, is about every two years,32 we also included up to three lags of dependent variable as additional controls in the seventh column, but this also does not change the significance on SCORE. Thus far we have assumed that an increase in

our legislative score can only have a linear impact on nesting activity. Feasibly at higher levels of effectiveness additional refinement of the code may have less of an impact than where there are few legislative provisions to ensure sea turtle friendly lighting. In the final column of table 3 we hence included the squared value of SCORE as an additional control. Accord- ingly, while the linear term remains significant, the quantitative size of the coefficient increases nearly fourfold. At the same time, however, the squared value of SCORE is significantly nega- tive. Taken together, this suggests that there is

an inverted u-shaped relationship between SCORE and NESTS, so that a higher legisla- tive score will encourage nesting activity but at a decreasing rate.33

Robustness Checks

We argued earlier that since we included prox- ies for all possible other determinants of nest- ing, as found in the existing literature, our specification in equation (2) is unlikely to suffer from omitted variable bias. Moreover, given the lengthy process of passing legislation there is also unlikely to be a simultaneity issue. To further verify this we additionally experimen- ted with instrumenting SCORE. More specifi- cally, a plausible instrument might be the political composition of the area around a beach, under the argument that local voters may pressure local legislators to pass more STFL ordinances, or less, according to their political affiliation, but that this composition is unlikely to have any direct effect on sea turtle nesting. To proxy such local political composi- tion we use data on voter registration by county and by party available from the Florida State Department and construct from this the share of democratic party registered voters related to each beach. Since these data are only avail- able from 1995 onward, we first reran our spec- ification in equation (2) allowing for a linear effect of SCORE on this reduced sample. As can be seen from the first column in 4, while the coefficient is somewhat higher than in the full sample, it remains statistically significant. Since equation (2) is a nonlinear model, stan- dard 2SLS methods would be inappropriate in order to instrument SCORE. Rather we, as suggested by Woolridge (2005), use a control function approach where the endogenous vari- able is regressed on the instrument and all other control variables and the predicted error term is then included in the second stage as an additional variable. One should note in this regard that for the first stage the share of dem- ocratic registered voters was a highly significant (negative) predictor of STFL,34 with a t-statistic

32 See Bjorndal, Meylan, and Turner (1983).

33 One may want to note that, although not reported here, we also experimented with further higher order terms but these turned out to be insignificant.

34 The negative effect may at first sight appear to be counterin- tuitive. However, Florida has traditionally had bipartisan support for environmental legislation. As a matter of fact, Jones (2015) found that in Florida republicans were no less likely to support environmental legislation than democrats. Similarly, Wang (2011) found that the share of republican voters did not determine the amount of environmental funding in Florida counties.

310 January 2020 Amer. J. Agr. Econ.

T ab

le 3.

N es ts R eg

re ss io n

(1 )

(2 )

(3 )

(4 )

(5 )

(6 )

(7 )

(8 )

S C O R E

0. 04 27 0* *

0. 00 47 8*

* 0. 00 48 1*

* 0. 00 47 9*

* 0. 00 48 2* *

0. 00 47

8* *

0. 00 89 2* **

0. 02 04 0*

* (0 .0 05 24 )

(0 .0 00 99 )

(0 .0 01 01 )

(0 .0 01 11 )

(0 .0 01 01 )

(0 .0 01 01 )

(0 .0 01 21 )

(0 .0 03 74 )

S C O R E 2

− 0. 00 02 9* *

(6 .6 2e -0 5)

R O O M S [d

= 10 0m

] − 0. 00 22 3*

* − 0. 00 22 3*

* 0. 00 22 2*

* − 0. 00 22 9* **

− 0. 00 20 7* *

(0 .0 00 70 5)

(0 .0 00 69 5)

(0 .0 00 70 7)

(0 .0 00 62 )

(0 .0 00 74 1)

IN C O M E / C A P

0. 00 88 8*

0. 00 88 5*

0. 00 90 6* *

0. 00 86 9*

0. 00 98 2

0. 00 66 7

(0 .0 03 54 )

(0 .0 03 58 )

(0 .0 03 51 )

(0 .0 03

56 )

(0 .0 03 70 )

(0 .0 03 51 )

N O U R IS H M E N T

− 0. 00 23 3

− 0. 00 28 8

− 0. 00 24 0

− 0. 00 23 4

− 0. 00 21 3

− 0. 00 21 7

(0 .0 02 02 )

(0 .0 02 01 )

(0 .0 02 04 )

(0 .0 02 02 )

(0 .0 01 71 )

(0 .0 01 99 )

S T O R M S

− 0. 00 33 7

− 0. 00 32 9

− 0. 00 15 3

− 0. 00 09

9 0. 00 57 8

0. 00 51 9

(0 .0 12 2)

(0 .0 12 2)

(0 .0 11 9)

(0 .0 11 9)

(0 .0 10 8)

0. 01 05

R O O M S [d

= 20 0 m ]

− 1. 89 e- 05

(3 .7 0e -0 5)

N O U R IS H M E N T (t – 1)

− 0. 00 04 65

(0 .0 01 73 )

S T O R M S (t – 1)

− 0. 00 60 6

(0 .0 11 4)

N E S T S (t – 1)

0. 00 00 2* **

(9 .7 5e -0 6)

N E S T S (t – 2)

0. 00 00 2

(0 .0 00 01 )

N E S T S (t – 3)

0. 00 00 2

(9 .9 3e -0 6)

O b se rv at io n s

78 0

78 0

78 0

78 0

78 0

78 0

69 0

78 0

B ea ch es

33 33

33 33

33 33

33 33

L o g li k el ih o o d

− 59 62

− 42 21

− 42 14

− 42 16

− 42 14

− 42 14

− 36 32

− 42 05

χ2 -t es t

66 .3 9* **

34 2. 96

** *

35 6. 66

** *

35 3. 01 ** *

36 6. 66 ** *

36 5. 67

** *

61 1. 19

38 9. 96 ** *

N o te :( a)

R o b u st st an

d ar d er ro rs in

p ar en

th es es ;( b ) **

*, ** ,a n d * in d ic at e 1%

,5 %

,a n d 10

% si gn

ifi ca n tl ev

el s; (c ) D ep

en d en

tv ar ia b le is n u m b er

o f lo gg

er h ea

d n es ts ;( d ) A ll re gr es si o n s in cl u d e ye

ar ly in d ic at o r va

ri ab

le s; (e ) co lu m n 1 is a st an

d ar d

n eg

at iv e b in o m ia l es ti m at o r, w h il e co lu m n s 2 th ro u gh

8 in cl u d e ti m e in va

ri an

t b ea

ch sp ec ifi c ef fe ct s.

Brei, Perez-Barahona, and Strobl Protecting Species through Legislation 311

of 8.51, and thus a relevant instrument. Impor- tantly, instrumenting for SCORE in equation (2) only marginally changes its coefficient, as shown in the second column. As a matter of fact, a z-test (statistic of 0.32) does not suggest that one should reject the null hypothesis that instrumented and noninstrumented coefficients are the same. In the third and fourth columns we show similarly for the 1995 to 2013 sample the specification allowing for a nonlinear effect of SCORE noninstrumented and instrumented, respectively.35 Again, the coefficients on SCORE and SCORE2 hardly change when we instrument for them, where this lack of differ- ence is further confirmed by a z-test.36

Next we explored whether it is the incidence of having legislation, rather than the extent of legislation, that affects loggerhead nesting activity. To this end we created a dummy var- iable, FIRST, that takes on the value of one when SCORE > 0 and zero otherwise, and include it as well as its interactions with SCORE and SCORE2 in our benchmark spec- ification. As can be seen, from the fifth column in table 4, it is the extent of the legislation that is driving the positive impact on nesting. We also experimented with using a piecewise poly- nomial rather than a quadratic term to capture nonlinearity in the legislation-nesting relation- ship, with knots chosen at 25, 50, and 100. The estimated coefficients on these, depicted in the sixth column, similarly that SCORE has a decreasing impact on nesting, especially once it passes a value of 50. One may recall that we constructed SCORE

as a simple sum of points given to the Principal Components and Implementation statements. In other words, we assume that both aspects have a similar and cumulative impact on nest- ing. To further investigate this we split SCORE into these two factors, PSCORE and ISCORE, respectively, and included these as well as their interaction term in equation (2). As can be seen from the final column in table 4, ISCORE has a 85% higher per point impact than PSCORE, but as the interaction term suggests, the two factors act partially as substi- tutes. This may explain the inverted u-shaped relationship we find when we use our total leg- islation score.

Marginal Effect

We can now use our coefficients to determine the quantitative impact of ordinance effective- ness on loggerhead nesting. One should note in this regard that there are two sources of non-linearity as a result of our econometric specification: the nonlinear nature of the nega- tive binomial model and the nonlinear effect of SCORE. Using the estimated coefficients from column 7 of Table 3 and setting the value of all other co-variates at their mean, we depict the implied marginal impact of SCORE across its range along with 95% confidence intervals in figure 5. As can be seen, moving from a beach not covered by any sea turtle lighting related legislation to one with a score of 1 increases the number of nests by 9. This mar- ginal effect increases as one moves across the range of possible values to reach a peak at 36 to then start falling. At the end of the spec- trum, increasing the score from 99 to 100 results in an additional four nests.

In figure 6 we plot the full range of cumula- tive effects by summing the marginal effects across the range of possible values of SCORE. Accordingly, the additional nests to be gained for a beach with no legislation in place of one with the highest possible score of a 100 is 1,119 nests. Comparing this to the mean, this suggest that the average beach would experi- ence a 76% increase to go from no legislation to adopting an ordinance with maximum effec- tiveness. If we consider the mean score of all nesting beaches in Florida, that is, 42, then our results suggest that on average an addi- tional 502 nests annually on a beach can be attributed to the current legislation implemen- ted in Florida, that is, about 34%. One may also note that the turning point of inverted u- shape relationship is beyond the maximum score of 100.

Sea Turtle Population Dynamics

In the previous section we quantified the effect of STFL ordinances on nesting activity. How- ever, importantly in terms of species preserva- tion, impacts at the nesting stage will also feed into the population dynamics of turtles. We thus now quantify the effect of coastal ordi- nances on sea turtles by means of introducing our estimates into a dynamic model of their population. In this regard it is well-known that the life cycle of sea turtles is composed of a

35 This consisted of instrumenting SCORE and SCORE2 with the share of demographic voters and its value squared. The F- statistics on these predictors for the first stage of the levels and squared equations were 44.53 and 29.51, respectively.

36 The z-statistic was 0.70 and 0.59 for SCORE and SCORE2, respectively.

312 January 2020 Amer. J. Agr. Econ.

series of development stages (see, e.g., Heppell, Snover, and Crowder 2003). We thus employ the stage-structured popula- tion model of Crouse, Crowder, and Caswell (1987) and Crowder et al. (1994), in which the individual females are grouped by stage. Each stage is characterized by its annual reproduction and survival rates, as well as by the number of years that a turtle stays in that stage.

Population Model

Our model consists of five stages of sea turtle development: (1) eggs/hatchlings, (2) small juveniles, (3) large juveniles, (4) subadults, and (5) adults. We define the stage distribution vector at time t ≥ 0 as

ð3Þ xt � ðx1t,x2t,x3t,x4t,x5tÞ,

Table 4. Robustness Checks

(1) (2) (3) (4) (5) (6) (7)

SCORE 0.01161*** 0.01089*** 0.04453*** 0.03621*** (0.00143) (0.00176) (0.00518) (0.01074)

SCORE2 −0.0060*** −0.00048** (0.0009) (0.00019)

FIRST −0.40874 (0.24571)

SCORE*FIRST 0.04211*** (0.01305)

SCORE2*FIRST −0.00053*** (0.00016)

SCORE25 0.37751*** (0.06373)

SCORE50 0.42175*** (0.08889)

SCORE100 0.13164** (0.05526)

PSCORE 0.01406*** (0.00291)

ISCORE 0.02622*** (0.00672)

PSCORE*ISCORE −0.00114*** (0.00026)

Observations 600 600 600 600 780 780 780 Beaches 33 33 33 33 33 33 33 Log Likelihood/R2 −3168 −3157 −3146 −3142 −4203 −4201 −4202 χ2/F-test 427.83*** 476.69*** 500.05*** 521.15*** 398.19*** 415.41*** 396.07***

Note: (a) Robust standard errors in parentheses; (b) ***, **, and * indicate 1%, 5%, and 10% significant levels; (c) Dependent variable is number of loggerhead nests; (d) All regressions include yearly indicator variables; (e) Column (1) is a standard negative binomial estimator, while columns (2) through (7) include time invariant beach specific effects.

Figure 5. Marginal effect of legislation Figure 6. Cumulative effect of legislation

Brei, Perez-Barahona, and Strobl Protecting Species through Legislation 313

where xit is the number of female sea turtles in stage i at time t for i = 1,…,5. By means of a five-stage Leslie matrix L, we obtain the popu- lation distribution at time t + 1 as

ð4Þ x0t + 1 = Lx0t,

where x0 denotes the transpose of vector x. Hence, starting from a given initial stage distri- bution x0, we get the evolution of the popula- tion by recursively applying equation (4). In order to compute the matrix L, we need

data about the duration (di), the annual fecun- dity rate (Fi), and the yearly survival rate (σi) of turtles in each stage i (details reported in Appendix B). Crowder et al. (1994) provide this information for loggerheads in Florida (see Appendix C). With respect to the initial distribution x0, we use available current esti- mates of the number of adult loggerhead females in Florida. With the share of turtles in each stage, we can then establish the num- ber of individuals per stage and, consequently, the initial vector x0. As in Crowder et al. (1994), we suppose that

the distribution of individuals among stages is stable. We show in Appendix C that one can reasonably assume this when the population is near to its long-run equilibrium. In this regard, provided that λ1 is the dominant eigen- value of L (see first row of table 11 in Appen- dix D), the model predicts that the percentage of female loggerheads in each stage (table 12 in Appendix D) will be 21.72 (stage 1), 67.65 (stage 2), 9.76 (stage 3), 0.61 (stage 4), and 0.26 (stage 5). Moreover, since jλij < 1 for i = 1,…,5, sea turtles characterized by the model parameters will face the risk of extinc- tion in the long-run.37 Indeed, as noted in the introduction, loggerheads in Florida are cur- rently listed as threatened according to the US Endangered Species Act (Witherington et al. 2009). Richards et al. (2011) estimate that the pop-

ulation of adult loggerhead females in the western North Atlantic is about 38,334 indi- viduals, most of them stemming from Florida (the largest subpopulation). Considering this value together with the percentages of individ- uals per stage stated above, we can calibrate x0 (see last column of table 12 in Appendix D). Note that in our simulations we set x10 = 3,528,180 following the estimated

production of loggerhead hatchlings per year in Florida reported by Brost et al. (2015). This number is fairly close to our estimate of 3,159,771 loggerhead hatchlings.38

With our calibrated model in hand, we can give numerical projections of the population of female loggerheads in Florida (per stage and in total) starting from the initial distribu- tion x0 until a time horizon t = T > 0. In partic- ular, by considering a large enough T, we can quantify the years to extinction. In the context of our model, this can be defined as the num- ber of years for less than one individual to remain.39 We first consider the benchmark scenario of no ordinances, which is described by the L matrix corresponding to table 10 in Appendix C. We then compare it with the population dynamics under different strength levels of STFL ordinances.

In order to quantify the ordinances’ effect on the population dynamics we incorporate the estimated effect of the SCORE variable into the population model. Since our earlier results show how legislation raises logger- heads’ nesting activity, the implementation of STFL ordinances will increase the annual fecundity rate Fi in the matrix L. In order to modify this parameter, we compute the increase in hatchlings per year due to the adoption of ordinances at a strength level S. Let us denote by NESTavg the average number of nests per beach. If the ordinances are set at a strength level S, the estimated cumulative marginal effect of the legislation on nesting activity, CSCORES, will be as depicted in fig- ure 6. The average percentage point increase in nests τ(S) is then given by

ð5Þ τðSÞ = CSCORES=r NESTavg + CSCORES=r

100,

where r represents the remigration interval (in years) of loggerheads. One should observe that we are working at the individual sea turtle level. Hence, since each turtle does not nest every year, we adjust the accumulative effect of the ordinances by dividing CSCORES by the remigration interval r. In this regard Bjorndal, Meylan, and Turner (1983) show evidence of a remigration interval of

37 Notice that, for jλij < 1 for i = 1,…,5, the number of individ- uals asymptotically converges to zero.

38 As a robustness check, we also did simulations considering x10 = 3,159,771. The corresponding results remain essentially unchanged.

39 As it is standard in the sea turtles’ biology literature, the pop- ulation models focus on female dynamics. Therefore, extinction occurs when the last female disappears.

314 January 2020 Amer. J. Agr. Econ.

loggerheads in Florida of two years, and Phil- lips et al. (2014) observe that this remigration interval has not changed over time in Florida. We thus consider r = 2 for our simulations. Moreover, with no empirical evidence avail- able, we simply assume that the percentage increase in nests will result in the same per- centage increase in eggs per female sea turtle. As observed in section 3.1, our proxy for nest- ing activity of loggerheads in Florida is the number of successful nests. Without system- atic data about emergence success,40 we can- not be more specific about the effect of STFL ordinances on the annual fecundity rate. Hence, similarly to Brei, Pérez-Barahona, and Strobl (2016), we modify the annual fecundity as ~Fi = ½1 + τðSÞ=100�Fi considering τ(S) in equation (5).

Out of Sample Prediction

Our econometric analysis is based only on the Index Beaches in Florida since surveying efforts for these have been consistent over time. However, we would like to use our results to make predictions for the total Flor- ida loggerhead population, that is, also for nesting activity in non-Index beaches. It is thus important to demonstrate that the estimated legislation-nesting relationship is representa- tive for these beaches as well. To this end we have access to nesting data from the Statewide program for the years 2008 to 2014, under which both Index and non-Index were sur- veyed, but surveying efforts across time were not necessarily consistent. Sample statistics by beach category are shown in table 5. As can be seen, while the number of hotel rooms within 100 m of a beach, income per capita, and storm activity are not statistically different between the two groups of beaches, there are considerable differences in terms of nightlight intensity, legislation score, beach nourish- ment, and number of nests.41

To examine whether the differences in the mean characteristics across beach groups also translates into a different relationship between legislation and nesting activity, we reestimated equation (2) using the Statewide program nesting data, but separately for Index

and non-Index beaches in table 6. As can be seen, for both samples there is a significant inverted u-shaped relationship between legis- lation and nesting activity. If we compare the coefficients on the Index beach sample to those in table 3 one finds that they are good bit larger. This might be in part due to the dif- ference in monitoring technique between the two surveys, as well as the more recent sample period of the statewide program.42 However, more importantly, there appears to be not too much difference in the coefficients on SCORE and SCORE2 between the two sam- ple of beaches. As a matter of fact, a z-test of the difference in coefficients across the two samples was 0.0721 and 0.0002 for SCORE and SCORE2, respectively. One can thus con- clude that, despite the differences in some of the mean characteristics, the base impact of sea turtle friendly legislation on nesting activ- ity is similar on Index and non-Index beaches, and we thus can confidently use our results from the last column in table 6 to predict log- gerhead population dynamics for the entirety of Florida.

Dynamic Results

Starting from the calibrated initial population x0, we provide projections of the population of female loggerheads in Florida by recur- sively applying equation (4). We focus on three scenarios, considering different SCORE levels of the ordinances. The scenario repre- sented by the L matrix directly computed from table 12 in Appendix C corresponds to the sit- uation of “no ordinances” in place, while the other scenarios consist of introducing various levels of STFL regulations. As noted before, we will assume that the

ordinances modify the annual fecundity rate by equation (5) and set NESTavg at the aver- age observed of our nesting data, namely 1,468 nests (see table 1). We first consider the case corresponding to the current level of ordi- nances in the nesting beaches in Florida, called “ordinances” scenario. In our model this situa- tion is characterized by the average score S = 42 of all Florida nesting beaches as is evi- dent from table 1. This corresponds to a cumu- lative effect of CSCORE42 = 502 nests (see

40 In the biology literature emergence success refers to the number of hatchlings that reach the beach surface.

41 The larger nesting activity in Index beaches is not surprising, since these are in part chosen to capture much of the nesting activ- ity in Florida.

42 As a matter of fact, in reducing the sample of the Index beach survey data that we used earlier to the same 2008 to 2014 period, the estimated coefficients from (2) on SCORE and SCORE2 were 0.045 and −0.0005, respectively.

Brei, Perez-Barahona, and Strobl Protecting Species through Legislation 315

figure 6) and implies an increase of the annual fecundity of about 15%. Finally, we study the prospect of setting the ordinances at the maxi- mum effectiviness. Under this scenario (“ordi- nances max.”) S = 100 across beaches and the corresponding cumulative effect is CSCORE100 = 1,019 nests, inducing the rate of annual fecundity to increase by 26%. We depict the evolution of the Florida log-

gerhead population for our three different sce- narios in figure 7. Under the scenario of no ordinances jλij < 1 for i = 1,…,5, and thus, as noted before, our model predicts that logger- heads will eventually become extinct. We find that this outcome is not reversed even when coastal areas adopt nesting friendly regulation at their maximum strength level (see rows 2 and 3 of table 11 in Appendix D). However, implementing the ordinances allows the popula- tion to increase along the transition due to the rise in fecundity. We plot in Appendix E the

population difference between scenarios as a percentage of the initial population of logger- heads x0 (figure 8). The increase of loggerheads’ fecundity makes this difference particularly important for hatchlings (figure 9).

A direct implication of implementing STFL legislation is that loggerheads extinction will be delayed. We provide the exact years to extinction in table 7 (scenarios “no TTA”). As can be seen, current level of ordinances increases years to extinction by 22 (10%), while implementing maximum effective legis- lation across Florida’s nesting beaches will add a further nineteen years.

One can reasonable expect that adopting a STFL legislation with a certain strength level will take some time. In order to incorporate this into our analysis we first estimated how long it takes on average for the legislation score to change by an additional unit, once it is positive for a beach. To this end we took all beaches once they had a positive score and regressed this score on a (annual) time trend. This gave us a coefficient of 2.7, mean- ing that the score rose (on average) 2.7 points per year. Therefore, for a county with no ordi- nances in place, it would take about thirty- seven years to reach the maximum score of 100 and about fifteen years to get to the observed average score of 42.

The time needed to adopt the legislation might inhibit the success of the ordinances to delay loggerheads’ extinction. In order to eval- uate this possibility we incorporate legislation timing into our population model. More pre- cisely, starting from the initial level we increase the legislation score at the annual rate estimated above, computing the correspond- ing Leslie matrix L in every step. We report the implications of this exercise for time to extinction in table 7, scenarios TTA. The

Table 6. Regression by Beach Type

Index Beaches

Non-Index Beaches

(1) (2)

SCORE 0.06008*** 0.05273*** (0.00524) (0.00986)

SCORE2 −0.00072*** −0.00072*** (0.00027) (0.00016)

Observations 164 796 Beaches 33 161 Log likelihood

−686.5 −2413.7

χ2-test 252.3*** 536.7***

Note: (a) Robust standard errors in parentheses; (b) ***, **, and * indicate 1%, 5%, and 10% significant levels; (c) Dependent variable is number of loggerhead nests; (d) All regressions include yearly indicator variables; (e) column 1 is a standard negative binomial estimator, while columns 2 through 8 include time invariant beach specific effects.

Table 5. Index vs. Non-Index Beach Comparison

Sample Index Beaches Non-Index Beaches

Difference Variable Mean St.Dev. Mean St.Dev. t-stat

NL 5.48 10.71 8.62 10.55 13.11*** NESTS 1295.62 2422.36 176.25 505.65 11.95*** SCORE 34.29 21.79 40.21 21.26 3.18** ROOMS [d=100m] 92.16 432.45 77.25 213.18 0.63 INCOME/CAP 50.44 12.52 50.64 13.82 0.17 NOURISHMENT 1.91 11.16 0.322 4.51 3.02*** STORMS 0.68 1.04 0.73 1.09 0.56

Note: NL ≡ intensity of nightlights; NEST ≡ number of nests; SCORE ≡ legislation score; ROOMS ≡ number of rooms within d meters of the shoreline; INCOME/CAP ≡ county income per capita (’000s); NOURISHMENT ≡ average annual volume (cubic yards) of sand placed on nesting beaches; STORMS ≡ number of storms that affected a beach.

316 January 2020 Amer. J. Agr. Econ.

results suggest that our conclusions about STFL ordinances and time to extinction are robust to allowing time for adoption of the leg- islation. Considering the estimated pace of adoption, the time to extinction in the scenario “ordinances” remains unchanged. It slightly worsens (by three years, implying a relative change of 17% instead of 18%) for the case of “ordinances max.” due to a much longer transition phase (thirty-seven instead of fif- teen years).

WTP for the Protection of Sea Turtles

A number of studies have been conducted in the United States and in Florida in order to measure the public WTP for the protection of sea turtles. One group of studies focuses on WTP at the household level. For instance, con- sidering North Carolina, Whitehead (1992) estimates an average WTP to prevent logger- heads from becoming extinct of about $US 54.72 annually per household.43 Assuming a similar figure for Florida then would imply that Florida households as a whole would be willing to pay $US 399.48 million per year.44

Allowing for the presence of an alternative preservation program for threatened or endangered species in general (i.e., not only sea turtles), Whitehead (1993) identified a lower WTP for the protection of loggerheads. Accordingly, each household was willing to pay $US 18.09 per year, implying a total

WTP of $US 132 million per year in Florida. Wallmo and Lew (2012) examine the WTP of households for the entire United States for a ten-year recovering program of loggerheads, and their results imply a total yearly WTP in Florida of $US 693.7 million. Finally, specifi- cally for Florida, Hamed (2013) estimates the WTP for a five-year protection program of sea turtles’ nesting habitat as ranging from $US 22.48 to 29.80 per household per year, suggesting a total WTP of between $US 164 and 217.5 million per year.45

Oceana (2008) follows a different approach by focusing on scuba divers’ WTP for the greater likelihood of seeing a sea turtle during a dive and discovers this to be about $US 33.50. Since the average number of dives per year of a scuba diver is five, the WTP per diver per year would be $US 167.42. In considering what these figures would mean for Florida, one should note that according to the Diving Equipment and Marketing Association (DEMA) there are between 2.7 and 3.5 mil- lion of “active” scuba divers in the United States, 46 while other sources provide a more conservative estimate of about 1.2 million (see, e.g., http://undercurrent.org). DEMA also provides data about the number of new diving certificates each year per state, where Florida’s share of these is 9.57%. Using these figures suggests that the number of active scuba divers in Florida is between 258,390 and 334,495, or, more conservatively, around 114,840. The WTP provided by Oceana (2008) then implies that yearly WTP in Florida would be between $US 43.28 and 56 million, with a conservative estimate of $US 19.21 mil- lion. We summarize our WTP calculations in the first two columns of table 8. We now investigate the implications of the

WTP estimates above in terms of increasing the time to extinction of loggerheads in Flor- ida. Ideally we would like to do so by compar- ing them to the cost of STF legislation, that is, relative to the monetary expenses of introduc- ing, implementing, and enforcing the legisla- tion, as well as the possible subsequent losses in tourism activity due to lower lighting on beaches. However, unfortunately, such cost

0

5

10

15 N

u m

b e

r o

f tu

rt le

s (m

ill io

n s)

0 5 10 15 20 25 30 35 40 45 50 55 60 Years

No ordinances Ordinances Ordinances max.

Figure 7. Total population

43 All figures of WTP are expressed in 2016 US prices. 44 As stated by the US Department of Commence (https://www.

census.gov/quickfacts/FL, accessed July 10, 2019), the population of Florida in 2016 is of 20,612,439 residents, comprising 7,300,494 households (average of 2011–2015).

45 Hamed (2013) mainly concentrates on sea level rise threat and on two different towns in Florida (coastal vs. inland locations): Cocoa Beach and Oviedo.

46 DEMA is a major nonprofit international organization (www.dema.org) to promote recreational scuba diving and snor- keling industry. It is the main available data source for the number of scuba divers certificates, collecting information provided by the three main certification agencies (PADI, SDI, and SSI).

Brei, Perez-Barahona, and Strobl Protecting Species through Legislation 317

figures are not available. We thus take a more indirect approach. In particular, we instead consider how many sea turtles could be financed from the WTP through a head- starting program, where loggerheads are first raised in captivity and then released into the wild. Inserting the implied number of sea tur- tles into our population model will then allow us to explicitly find the years of reduction due to a head starting program financed by the WTP. We can use this figure as benchmark against the years of reduction implied by the current level of STFL ordinances. As in Jin et al. (2010) and Brei, Pérez-

Barahona, and Strobl (2016), we concentrate on a five-year head starting programs. One should note that such head starting programs for turtles are currently implemented in the United States, despite concerns about their effectiveness.47 As determined from a per- sonal communication with Benjamin Higgins, NOAA Federal/National Marine Fisheries Service Galveston Laboratory, we note that turtles are typically raised in Florida/Gulf of Mexico until they are about two years old in order to benefit from the use of turtle excluder devices (TED), mandatory in shrimp trawls since 1987 (see Duffy 2016b).48 We thus focus on the rearing costs of two -year-old logger- heads, and in line with information provided by Nicholas Blume, Florida Atlantic Univer- sity (FAU), assume this to be US$836 per two year-old.49 Alternatively, we use the fig- ure of about US$2000 per loggerhead as sug- gested by Benjamin Higgins from the NOAA facility in Galveston Laboratory.50 The implied numbers of released turtles per year are reported in the last two columns of table 8.

We next insert the implied number of logger- heads bred (two -year-old small juveniles) from the captivity-rearing program into our popula- tion model (stage 2, “small juveniles”) during five consecutive years., where table 13 in Appendix F provides the corresponding num- ber of extinction years avoided. According to our simulations, even under the highest WTP scenario,51 such a head starting program would only delay loggerheads’ extinction by a maxi- mum of two years. This constitutes only 3% of the reduction in time to extinction implied by current levels of STFL legislation. One should note that this rather sobering outcome is consis- tent with the general wisdom of conservation- ists that sea turtle farming is expensive and of questionable ecological viability. In particular, head starting programs are generally consid- ered as a rather controversial sea turtles’ pro- tection policy (see, e.g., Ross 1999; Bell et al. 2005; Webb, Manolis, and Gray 2008; Arena, Warwick, and Steedman 2014). Apart enabling research leading to a better understanding the biology of sea turtles, it is not clear that sea tur- tle farming can be beneficial for the conserva- tion of populations in the wild (Ross 1999). With this in mind, there is a tendency to suspect that STFL ordinances may be the more viable policy for loggerhead preservation, in particu- lar since there is evidence that its costs may be limited (see, e.g., Witherington and Martin 1996, 2000). Nevertheless, any more precise conclusions in this regard can only be drawn once exact costs of STFL implementation is available.

Concluding Remarks

In this paper we investigated the effectiveness of using legislation to protect endangered

Table 7. Time to Extinction

Ordinances Ordinances Max.

No Ordinances No TTA TTA No TTA TTA

Years to extinction 223 245 245 264 261 Relative change (%) … 9.87 9.87 18.38 17.04

Note: TTA stands for “time to adopt”. Scenarios TTA incorporates the effect of legislation timing.

47 See, e.g., Shaver et al. (2016); Tetzlaff, Sperry, and DeGre- gorio (2019); and Rees et al. (2016).

48 TEDs are escape hatches, which allows captured turtles to escape nets before drowning (see, e.g., Crowder et al. 1994; Hep- pell, Snover, and Crowder 2003).

49 $US 418 per turtle and year according to the FAU Marine Laboratory in 2014.

50 This amount was also used in a legal case against sea turtle eggs smugglers in Southern California (see Duffy 2016a, 2016b).

51 This is the sum of the WTP of households implied by the esti- mates of Wallmo and Lew (2012) and that of scuba divers as calcu- lated from the results of Oceana (2008), leading to a total WTP in Florida per year of $US 749.7 million.

318 January 2020 Amer. J. Agr. Econ.

species by examining to what extent lighting ordinances have limited the negative impact of light pollution on sea turtles in Florida. To this end we constructed an index of ordinance effectiveness across counties and combined this with loggerhead nesting data and a set of rich controls to create a panel data set cover- ing twenty-six years. Our econometric findings showed that legislation can significantly increase nesting activity, where current legisla- tion results in an additional 34% increase in nests. Using our estimates within a calibrated population model we also demonstrated that the current legislation aids sea turtles by extending the number of years to their extinc- tion by 22. The findings here thus arguably suggest that carefully crafted species protec- tion legislation can potentially be an effective means of wildlife management.

It is important to point out that, as it stands, the level of protection with regard to beach front lighting appears to be still far from what it could be, particularly since sea turtles are believed to be an important point of attraction for tourists in Florida.52 More precisely, on average, counties in Florida have implemen- ted legislation that is 42% effective, according to our constructed index. Whether the current situation is economically efficient will of course depend on a comparison of its benefits and costs. With regard to the latter, since sea turtle lights generally project less luminosity than regular lighting, the loss in tourism due to the potential drop in beach safety at night during nesting season, or at least the

perception thereof, should ideally be quanti- fied (Witherington and Martin 1996). There is also the cost of substituting regular beach front lighting with more sea-turtle friendly lighting that we were unable to consider due to a lack of data (see Ernest 2002). In this regard it should, however, be noted that while sea turtle friendly lighting itself may not be inexpensive, it is more efficient and thus more energy-saving than regular illumination and could, as noted by Witherington and Martin (2000) and Barshel et al. (2014), result in potentially significant reductions in the elec- tricity bill of housing facilities and hotels.53

While we could not determine the monetary cost of current sea turtle friendly lighting legis- lation, or its further expansion, we were able to estimate the cost of reducing loggerheads’ time to extinction via alternatively raising them in captivity to be released in the wild. Considering a range of WTP estimates derived from the existing literature in this regard, we show that under such an alternative policy, however, the Florida public would at best only finance 3% of the current legislation’s effec- tiveness. This suggests that STFL ordinances may be the cheaper alternative. Nevertheless, further data and analysis are required to draw any clear conclusions in this regard.

Table 8. Yearly WTP in Florida

Per Household Total No. Turtles Bred No. Turtles Bred (US$) (US$ millions) (FAU) (NOAA)

Whitehead (1992) 54.72 399.48 464,445 199,740 Whitehead (1993) 18.09 132 153,466 66,000 Wallmo and Lew (2012) 95.02 693.7 806,512 346,850 Hamed (2013) 22.48–29.80 164–217.5 190,670–252,871 82,000–108,750 Oceana (2008) 33.5 43.28–56 50,318–65,107 21,640–28,000 Oceana (2008)(*) 33.5 19.21 22,334 9,605

Note: (a) All figures in terms of 2016 US prices; (b) We consider five-year equivalent protection programs; (c) WTP per scuba diver in Oceana (2008); (d). (*) Considering the conservative estimate of the number of active scuba divers; (e) Turtles bred corresponds to two year-old loggerhead small juveniles, provided the rearing costs per turtle based on the estimates of FAU and NOAA labs.

52 Stokes and Lowe (2013) state that, in 2011, about 11.5 million tourists visited Gulf states for wildlife viewing, spending US$ 6.5 billion. According to Carr et al. (2016) over 500,000 tourists visit annually the coastal communities in the Southeast of the US to watch sea turtles. Moreover, there are twenty-three centers in Florida organizing sea turtles walks, which attract about 10,000 vis- itors per year.

53 Ernest (2002) reports that the Florida Power and Light Com- pany estimated for 2001 that, per light, replacing non-cutoff with cutoff lighting cost US$674, repositioning head to redirect light US$300, reducing the wattage US$675, lowering the mounting height US$300, installing amber filter lens US$280, and turning the lights off for nesting season US$0. However, Barshel et al. (2014) observe that the general principles “Keep it Low, Keep it Long, and Keep it Shielded” imply significant savings on outdoor electricity bills. They point out, as an example, that La Playa Con- dominiums in Satellite Beach (Florida) realized a nearly 70% cost savings after installing STFL.

Brei, Perez-Barahona, and Strobl Protecting Species through Legislation 319

Acknowledgments

The authors are thankful to Nicholas Blume, Benjamin Higgins, Robbin Trindell, and Jean- ette Wyneken, for their comments and help. They would also like to express our gratitude to the participants at WCERE 2018, Gothen- burg (Sweden), SURED 2018, Ascona (Switzerland), SALISES 2018, Montego Bay (Jamaica), and Séminaire Cournot, BETA, University of Strasbourg (France), for useful discussion. The Associate Editor of this Journal and three anonymous referees provided sugges- tions that helped to make substantial improve- ments to the paper. Agustín Pérez-Barahona acknowledges financial support from the Chaire Développement Durable (Ecole Polytechnique – EDF) and the Labex MME-DII.

References

Allison, P., and R. Waterman. 2002. Fixed Effects Negative Binomial Regression Models. In Sociological Methodology, ed. Ross M. Stolzenberg. Oxford: Basil Blackwell.

Anderson, S., S. Nuernberge, K. Yamamoto, and P. Sutton. 2013. Evaluating the Com- pliance of Sea Turtle Light Ordinances in Florida Using Remote Sensing. Geogra- phy Compass 7 (12): 867–78.

Arena, P.C., C. Warwick, and C. Steedman. 2014. Welfare and Environmental Impli- cations of Farmed Sea Turtles. Journal of Agricultural and Environmental Ethics 27 (2): 309–30.

Barshel, N., R. Bruce, C. Grimm, D. Haggitt, B. Lichter, and J. McCray. 2014. Sea Tur- tle Friendly Lighting. A Model Ordinance for Local Governments. Gainesville, FL: Levin College of Law, Conservation Clinic, University of Florida.

Bell, C., J. Parsons, T. Austin, A. Broderick, G. Ebanks-Petrie, and B. Godley. 2005. Some of Them Came Home: The Cayman Turtle Farm Headstarting Project for the Green Turtle Chelonia Mydas. Oryx 19 (2): 137–48.

Bjorndal, K.A., A.B. Meylan, and B.J. Turner. 1983. Sea Turtles Nesting at Melbourne Beach, Florida, I: Size, Growth and Reproductive Biology. Biological Conser- vation 26 (1): 65–77.

Boersma, P., P. Kareiva, W. Fagan, J. Clark, and J. Hoekstra. 2001. How Good Are

Endangered Species Recovery Plans. Bio- science 51: 643–9.

Borkovic, B., and L. Nostbakken. 2017. The Cost of Endangered Species Protection: Evidence from Auctions for Natural Resources. Journal of Environmental Economics and Management 81: 174–92.

Boyd, J. 2014. Best Available Science and Imperiled Species Conservation: Chal- lenges, Opportunities, and Partnerships. Resources for the Future Discussion Paper, 14–44.

Brei, M., A. Pérez-Barahona, and E. Strobl. 2016. Environmental Pollution and Biodiversity: Light Pollution and Sea Tur- tles in the Caribbean. Journal of Environ- mental Economics and Management 77: 96–116.

Brost, B., B. Witherington, A. Meylan, E. Leone, L. Ehrhart, and D. Bagley. 2015. Sea Turtle Hatchling Production from Florida (USA) Beaches, and 2002–2012, with Recommendations for Analyzing Hatching Success. Endangered Species Research 27 (1): 53–68.

Bustard, H. 1967. Mechanism of Nocturnal Emergence from the Nest in Green Turtle Hatchlings. Nature 244–317.

Cameron, A., and P. Trivedi. 2013. Regression Analysis of Count Data. New York: Cam- bridge University Press.

Carr, B., S. Emmert, P. Mustain, and L. Snyder. 2016. TEDs for All Trawls: A Net Positive for Fishermen and Sea Tur- tles, Oceana Reports, May 2016.

Crouse, D., L. Crowder, and H. Caswell. 1987. A Stage-based Population Model for Log- gerhead Sea Turtles and Implications for Conservation. Ecology 68 (5): 1412–23.

Crowder, L., D. Crouse, S. Heppell, and T. Martin. 1994. Predicting the impact of tur- tle excluder devices on loggerhead sea tur- tle population. Ecological Applications 4 (3): 437–45.

Davenport, J. and J.L. Davenport. 2006. The Impact of Tourism and Personal Leisure Transport on Coastal Environments: A Review. Esturine, Coastal and Shelf Sci- ence 67 (1–2): 280–92.

Driscoll, J., and A. Kraay. 1998. Consistent Covariance Matrix Estimation with Spa- tially Dependent Panel Data. Review of Economics and Statistics 80 (4): 549–60.

Duffy, L. 2016a. United States’ Sentencing Memorandum, Case 15-cr-2867-JLS. Government’s Sentencing Memorandum, 15-cr-2867-JLS, 1–15.

320 January 2020 Amer. J. Agr. Econ.

. 2016b. United States’ Sentencing Memorandum, case 15-cr-2867-JLS, sup- plemental sentencing, Government’s Sen- tencing Memorandum, 15-cr-2867-JLS, supplemental sentencing, 1–28.

Eckert, K., K. Bjorndal, F. Abreu-Grobois, and M. Donnelly. 1999. Research and Management Techniques for the Conser- vation of Sea Turtles. IUCN/SSC Marine Turtle Specialist Group Publication No. 4.

Ernest, R. 2002. Coastal Roadway Lighting Manual. Jensen Beach, FL: Ecological Associates, Inc.

EF. 2016. Do Habitat Conservation Plans Deserve Wider Implementation. The Environmental Form May/June: 48–57.

Ferraro, P.J., C. McIntosh, and M. Ospina. 2007. The Effectiveness of the US Endan- gered Species Act: An Econometric Anal- ysis Using Matching Methods. J Environ Econ Manage 54 (3): 245–61.

Frazer, N. 1983. “Demography and Life His- tory Evolution of the Atlantic Logger- head Sea Turtle, Caretta caretta”, Dissertation. Athens, Georgia, USA: Uni- versity of Georgia.

Gibbs, K., and D. Currie. 2012. Protecting Endangered Species: Do the Main Legis- lative Tools Work? PLoS One 7 (5): e35730.

Hamed, A. 2013. “Economic Valuation of Florida Sea Turtles in Face of Sea Level Rise”, Electronic Thesis and Dissertations, University of Central Florida, Paper 2750.

Heppell, S., M. Snover, and L. Crowder. 2003. Sea Turtle Population Ecology. In The Biology of Sea Turtles, ed. P. Lutz, J. Musick, J. Wyneken. Boca Raton, FL: CRC Marine Science Series.

Houtan, K., and O. Bass. 2007. Stormy Oceans Are Associated with Declines in Sea Tur- tle Hatching. Current Biology 17: 590–1.

Jin, J., A. Indab, O. Nabangchang, T. Thuy, D. Harder, and R.F. Subade. 2010. Valuing Marine Turtle Conservation: A Cross- Country Study in Asian Cities. Ecological Economics 69 (10): 2020–26.

Johnson, S., K. Bjorndal, and A. Bolten. 1996. Effects of Organized Turtle Watches on Loggerhead (Caretta caretta) Nesting Behaviour and Hatchling Production in Florida. Conservation Biology 10 (2): 570–7.

Jones, M. 2015. Environmentalism and Envi- ronmental Constitutional Ballot Initiatives in Florida: The Elements of Support for Amendment One in 2014 in the Context

of Current Environmental Attitudes. Uni- versity of Florida Master’s Thesis.

Kerkvliet, J., C. Langpap. 2007. Learning from Endangered and Threatened Species Recovery Programs: A Case Study Using U.S. Endangered Species Act Recovery Scores. Ecological Economics 63 (2-3): 499–510.

Krippendorff, K. 2013. Content Analysis: An Introduction to Its Methodology, 3rd ed. Thousand Oaks, CA: SAGE Publications, Inc.

Langpap, C., and J. Kerkvliet. 2012. Endan- gered Species Conservation on Private Land: Assessing the Effectiveness of Hab- itat Conservation Plans. Journal of Envi- ronmental Economics and Management 64 (1): 1–15.

Langpap, C., J. Kerkvliet, and J. Shogren. 2018. The Economics of the US Endan- gered Species Act: A Review of Recent Developments. Review of Environmental Economics and Policy 12 (1): 69–91.

Lorne, K., and K. Salmon. 2007. Effects of Exposure to Artificial Lighting on Orien- tation of Hatchling Sea Turtles on the Beach and in the Ocean. Endangered Spe- cies Research 3: 23–30.

Lueck, D., and J. Michael. 2003. Preemptive Habitat Destruction under the Endan- gered Species Act. Journal of Law and Economics 46 (1): 27–60.

Mazaris, A., G. Matsinos, and J. Pantis. 2009. Evaluating the Impacts of Coastal Squeeze on Sea Turtle Nesting. Ocean and Coastal Management 52 (2): 139–45.

Oceana. 2008. Sea the Value: Quantifying the Value of Marine Life to Divers, Oceana Reports, August 2008. Available at: http://oceana.org/reports/sea-value- quantifying-value-marine-life-divers

Peters, A., and K. Verhoeven. 1994. Impact of Artificial Lighting on the Seaward Orienta- tion of Hatchling Loggerhead Turtles. Journal of Herpetology 28 (1): 112–14.

Phillips, K., K. Mansfield, D. Die, and D. Addison. 2014. Survival and Remigration Probabilities for Loggerhead Turtles (Car- etta caretta) Nesting in the Eastern Gulf of Mexico. Marine Biology 161 (4): 863–70.

Raymond, P. 1984. Sea Turtle Hatchling Dis- orientation and Artificial Beachfront Lighting. Washington DC, USA: Center for Environmental Education.

Rees, A.F., J. Alfaro-Shigueto, P.C.R. Barata, K.A. Bjorndal, A.B. Bolten, J. Bourjea, and A.C. Broderick, et al. 2016. Are We

Brei, Perez-Barahona, and Strobl Protecting Species through Legislation 321

Working towards Global Research Priori- ties for Management and Conservation of Sea Turtles? Endangered Species Research 31: 337–82.

Richards, P., S. Epperly, S. Heppell, R. King, C. Sasso, F. Moncada, G. Nodarse, D. Shaver, Y. Medina, and J. Zurita. 2011. Sea Turtle Population Estimates Incorpo- rating Uncertainty: A New Approach Applied to Western North Atlantic Log- gerheads Caretta caretta. Endangered Spe- cies Research 15 (2): 151–8.

Roman, J. 2011. Listed: Dispatches from America’s Endangered Species Act. Boston: Harvard University Press.

Ross, and J. 1999. Ranching and Captive Breeding Sea Turtles: Evaluation as a Conservation Strategy. In Research and Management Techniques for the Conser- vation of Sea Turtles, ed. K. Eckert, K. Bjorndal, F. Abreu-Grobois, M. Donnelly. IUCN/SSC Marine Turtle Specialist Group Publication No. 4, 1999.

Rumbold, D., P. Davis, and C. Perretta. 2001. Estimating the Effect of Beach Nourish- ment on Caretta caretta (Loggerhead Sea Turtle) Nesting. Restoration Ecology 9 (3): 304–10.

Sedjo, R. 2008. Perspectives on Sustainable Resources in America. Washington DC: Resources for the Future, Routledge.

Shaver, D.J., C. Rubio, J. Shelby Walker, J. George, A.F. Amos, K. Reich, and T. Shearer. 2016. Kemp’s Ridley Sea Turtle (Lepidochelys kempii) Nesting on the Texas Coast: Geographic, Temporal, and Demographic Trends through 2014. Gulf of Mexico Science 33 (2): 4.

Stokes, S., and M. Lowe. 2013. Wildlife Tour- ism and the Gulf Coast Economy. Durham, NC: Datu Research, LLC.

SWOT. 2011. The State of the World’s Sea Turtles (SWOT) Minimum Data Stan- dards for Nesting Beach Monitoring, SWOT Scientific Advisory Board Hand- book, version 1.0, page 26.

Tetzlaff, S.J., J.H. Sperry, and B.A. DeGregorio. 2019. Tradeoffs with Growth and Behavior for Captive Box Turtles Head-Started with Environmental Enrichment. Diversity 11 (3): 40.

Tuxbury, S., and M. Salmon. 2005. Competi- tive Interactions between Artificial Light- ing and Natural Cues during Sea Finding by Hatchling Marine Turtles. Biological Conservation 121 (2): 311–6.

USFWS. 2015. Federal and State Endangered and Threatened Species Expenditures: Fis- cal Year 2015. Washington DC: U.S. Fish and Wildlife Service.

Wallmo, K., and D. Lew. 2012. Public Willing- ness to Pay for Recovering and Downlist- ing Threatened and Endangered Marine Species. Conservation Biology 26 (5): 830–9.

Wang, X. 2011. Exploring Trends, Sources, and Causes of Environmental Funding: A Study of Florida Counties. Journal of Environmental Management 92 (11): 2930–8.

Webb, G., S. Manolis, and M. Gray. 2008. Cap- tive Breeding and Marketing of Turtles, Rural Industries Research and Develop- ment Corporation (RIRDC) Publication, Australian Government, 08/012.

Whitehead, J. 1992. Ex ante Willingness to Pay with Supply and Demand Uncer- tainty: Implications for Valuing a Sea Tur- tle Protection Program. Applied Economics 24 (9): 981–8.

. 1993. Total Economic Values for Coastal and Marine Wildlife: Specifica- tion, Validity, and Valuation Issues. Marine Resource Economics 8 (2): 119–32.

Witherington, B. 1992. Behavioral Responses of Nesting Sea Turtles to Artificial Light- ing. Herpetologica 48: 31–9.

Witherington, B., and N. Frazer. 2003. Social and Economic Aspects of Sea Turtle Con- servation. In The Biology of Sea Turtles, ed. P. Lutz, J. Musick, J. Wyneken, Boca Raton, FL: CRC Marine Science Series.

Witherington, B., P. Kubilis, B. Brost, and A. Meylan. 2009. Decreasing Annual Nest Counts in a Globally Important Logger- head Sea Turtle Population. Ecological Applications 19 (1): 30–54.

Witherington, B., R. Martin. 1996. Understand- ing, Assessing, and Resolving Light-Pollu- tion Problems on Sea Turtle Nesting Beaches, Florida Marine Research Insti- tute Technical Report, TR-2.

. 2000. Understanding, Assessing, and Resolving Light-Pollution Problems on Sea Turtle Nesting Beaches, 2nd ed., Flor- ida Marine Research Institute Technical Report, TR-2, 2nd edition, revised 2000.

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

. 2015. Control Function Methods in Applied Econometrics. Journal of Human Resources 50: 420–45.

322 January 2020 Amer. J. Agr. Econ.

Worth, D., and J. Smith. 1976. Marine Turtle Nesting on Hutchinson Island, Florida, in 1973, Vol. 18, 1–17. Florida Marine Research Publication.

Zhang, D. 2004. Endangered Species and Tim- ber Harvesting: The Case of Red-cock- aded Woodpeckers. Economic Inquiry 42 (1): 150–65.

Appendixes

A Legislation Statements

A.1 List of Statements

Barshel et al. (2014) identify seventeen state- ments, called “Sea Turtle Friendly Lighting Principles Component,” that define favorable conditions for sea turtle nesting. The set “Imple- mentation Component” considers nine state- ments in order to measures to what extent the ordinance ensures such nesting conditions. We provide below a detailed list of these statements.

Sea turtle friendly lighting principles component (17 items) 1. Exterior artificial light for existing devel-

opment must be low lumens. 2. Exterior artificial light for existing devel-

opment must be full cut-off (e.g., no light emitted above 90� angle).

3. Exterior artificial light for existing devel- opment must be downward directed.

4. Exterior artificial light for existing devel- opment must not be visible from the beach.

5. Exterior artificial light for existing devel- opment must be long wavelength (i.e., 580 nm or greater).

6. Exterior artificial light for existing devel- opment must be shielded.

7. Exterior artificial light for new develop- ment must be low lumens.

8. Exterior artificial light for new develop- ment must be full cut-off (e.g., no light emitted above 90� angle).

9. Exterior artificial light for new develop- ment must be downward directed.

10. Exterior artificial light for new develop- ment must not be visible from the beach.

11. Exterior artificial light for new develop- ment must be long wavelength (i.e., 580 nm or greater).

12. Exterior artificial light for new develop- ment must be shielded.

13. Artificial light shall not be visible (e.g., directly/indirectly/cumulatively) from the beach.

14. Areas seaward of the frontal dune are not to be directly illuminated.

15. Areas seaward of the frontal dune are not to be indirectly illuminated.

16. Areas seaward of the frontal dune are not to be cumulatively illuminated.

17. The building of campfires or bonfires shall be prohibited during the nesting season.

Implementation component (9 items) 1. Is a provision made for a compliance

inspection during the nesting season? 2. Does the ordinance provide for a pre-

enforcement warning? 3. Does the ordinance provide for a notice of

violation? 4. The ordinance establishes civil penalties for

noncompliance. 5. The ordinance establishes criminal penal-

ties for noncompliance. 6. Are the enforcement penalties incorpo-

rated into the ordinance by reference? 7. Will each day of any such violation consti-

tute a separate and distinct offense? 8. Does the ordinance provide for the educa-

tion of the general public? 9. Does the ordinance provide for the educa-

tion of the affected public (e.g., those sub- mitting an application for construction)?

A.2 Legislation Score: Specific Example

As a matter of illustration of how we con- structed the legislation scores for municipali- ties and counties, we use the case of Escambia County, FL. According to the list of ordinances provided by the FWC, the refer- ence for Escambia County’s legislation is Ord. No. 2013-28, § 2, 7-11-2013. We first study the Principles Component

statements. As specified above, they comprise seventeen statements (items). We compare each statement with the statements in the ordi- nance, assigning a value {0, 1, 2, 3} to each item in the actual ordinance. This scale represents the degree of compliance of the legislation with the content of each item. If the concept is not addressed in the ordinance, we then assign the value 0. If the concept is vaguely addressed in the ordinance, the item gets the value 1. If the concept is addressed but in a less stringent manner than what is required by the

Brei, Perez-Barahona, and Strobl Protecting Species through Legislation 323

statement, the item gets a value 2. We assign 3, the maximum value, if the concept is fully addressed in the ordinance. Let us consider, for instance, items 2 and 8 in

the list above (Principles Component state- ments): item (2) “Exterior artificial light for existing development must be full cut-off (e.g., no light emitted above 90� angle)” and item (8) “Exterior artificial light for new devel- opment must be full cut-off (e.g., no light emit- ted above 90� angle)”. Regarding to those components, the legislation of the county gets the maximum score (a value of 3) for items 2 and 8 because the ordinance explicitly requires “full cut-off fixtures” for both existing and new developments. The situation is differ- ent for items 5 and 11: item (5) “Exterior arti- ficial light for existing development must be long wavelength (i.e., 580 nm or greater)” and item (11) “Exterior artificial light for exist- ing development must be long wavelength (i.e., 580 nm or greater)”. The legislation gets a score of 2 for each of those items; it mentions the requirement of long wavelength lights (for both existing and new developments) but “580 nm or greater” is to vaguely stated. We provide the score of all seventeen items

of Principles Component in the second row of table 9. The sum of those scores for Escambia County gives 34 out of 51. After normalizing to 50, Escambia County gets an aggregate score of 33 for the Principles Component. Regarding the Implementation component

(nine items), we do a similar comparison but the scale is 1 if the concept is addressed in the ordinance or 0 otherwise. For example, item 1 refers to “Is a provision made for a compliance inspection during the nesting sea- son?” Escambia County gets a 1 because a provision for compliance is explicitly men- tioned. However, regarding item 8: “Does the ordinance provide for the education of the general public?”, the county gets a 0 because its legislation neither mentions nor specifies that the county shall conduct a community education effort to support sea turtles against nightlights.

All the scores for the nine items of the Imple- mentation component are summarized in the fourth row of table 9. Escambia County receives a score of 2 out of 9, which represents 11 out of 50 for the Implementation component. Finally, considering Principles and Implementation components together, the total legislation score for Escambia County is 33 + 11 = 44.

B Stage-Structured Population Model

In this appendix we compute the Leslie matrix L (also known as stage class population matrix) of our model. Moreover, we provide several properties that will be useful to cali- brate the model and to describe the dynamics of the population.

Let Pi denote the percentage of females in stage i that survive but remain in that stage, Gi be the percentage of females in stage i that sur- vive and progress to the next stage, and Fi be the number of hatchlings per year produced by a sea turtle in stage i (i.e., annual fecundity). Therefore, the number of hatchlings produced by each stage class at time t + 1 is given by

ðA:1Þ x1t + 1 = F1x1t + F2x2t + F3x3t + F4x4t + F5x5t:

Moreover, the number of females present in the subsequent stage j, for j = 2,…,5, is

ðA:2Þ xjt + 1 = Gj−1xj−1t + Pjxjt:

Taking equations (A.1) and (A.2), the matrix L in equation (4) is

L �

F1 F2 F3 F4 F5 G1 P2 0 0 0 0 G2 P3 0 0 0 0 G3 P4 0 0 0 0 G4 P5

2 66664

3 77775 :

Table 9. Score, Escambia County

Item 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

Principles C. 2 3 1 3 2 3 2 3 1 3 2 3 3 1 1 1 0 Aggregated score 33 Implementation C. 1 0 1 0 0 0 0 0 0 Aggregated score 11 Total score 44

Note: All aggregated scores normalized to 50.

324 January 2020 Amer. J. Agr. Econ.

The fecundity rates Fi are typically directly obtained from actual data on sea turtles (for the loggerheads, see table 10 in Appendix C). However, Gi and Pi need to be calculated from information about the duration (di) and the yearly survival rate (σi) of each stage i. Let us first determine the percentage of sea turtles from stage i that grow into stage i + 1 as

γi =

ð1−σiÞσdi −1i 1−σdii

if σi 6¼ 1 1 di

if σi = 1:

8 >>< >>:

ðA:3Þ

Consequently, the percentage of turtles in stage i that remain in that stage is 1 – γi. We can then obtain Gi and Pi as

ðA:4Þ Gi = γiσi,

ðA:5Þ Pi = ð1−γiÞσi:

The solution of equation (4) for all t ≥ 0 is x0t =

P5 i = 1civ

0 λ1 λti, where vλi denotes the eigen-

vector corresponding to the eigenvalue λi of L, and ci are constants determined by the ini- tial stage population distribution. Considering the specific characteristics of loggerheads in Florida, we will show later on that jλ1j > jλjj for j = 2,…,5. Then, for a large t the solution takes the form

ðA:6Þ x0t ≈c1v0λ1λ t 1,

where λ1 is frequently called the dominant eigenvalue. Therefore, the long-run propor- tion of the population in stage i is given by

ðA:7Þ ξi = vλ1i

P5 k = 1

vλ1k

,

where vλ1k is the kth coordinate of the eigen- vector vλ1 (for the loggerheads, see table 11 in Appendix D).

C Stage-Based Life Table

In their stage-structured population model for females, Crowder et al. (1994) set a five-stage life history parameters for loggerheads. This is based on the stage-based life table of Crouse, Crowder and Caswell (1987), which is built on data from Frazer (1983).

Table 10. Loggerhead Sea Turtles

Stage Description

Stage Duration

in Years (di)

Annual Survival Rate (σi)

Annual Fecundity

(Fi)

1 Eggs/ Hatchlings

1 0.6747 0

2 Small juveniles

7 0.75 0

3 Large juveniles

8 0.6758 0

4 Subadults 6 0.7425 0 5 Adults >32 0.8091 76.5

Table 12. Long-run Stage Distribution, No Ordinances

vλ1 ðvλ1 kÞ Stage Description Distribution (%) x0 ðxk0Þ 0.3028 1 Eggs/hatchlings 21.72 3159771 0.9432 2 Small juveniles 67.65 9842492 0.1360 3 Large juveniles 9.76 1419475 0.0084 4 Subadults 0.61 88162 0.0037 5 Adults 0.26 38334

Table 11. Eigenvalues

λ1 λ2 λ3 λ4 λ5

No ordinances 0.9281 0.7318 + 0.2037i 0.7318-0.2037i 0.4744 0.0059 Ordinances 0.9344 0.7328 + 0.2110i 0.7328-0.2110i 0.4651 0.0069 Ordinances max. 0.9388 0.7336 + 0.2161i 0.7336-0.2161i 0.4585 0.0076

Brei, Perez-Barahona, and Strobl Protecting Species through Legislation 325

D Matrix L Properties

Taking the information in table 10, we can compute the eigenvalues of L for the logger- heads in Florida. We also provide these values

for the modified L when we consider the effect of coastal ordinances on the annual fecundity.

Since λ1 is the dominant eigenvalue, we cal- culate the stable stage distribution (A.7) by considering the corresponding eigenvector vλ1.

0

2

4

6

8

10

12

14

P e

r ce

n t

0 10 20 30 40 50 60 70 80 90 100 Years

Ordinances vs. no ordinances Ordinances max. vs. no ordinances Ordinances max. vs. ordinances

Figure 8. Population difference, total (% of initial population). (a) Ordinances max. vs. no ordinances. (b) Ordinances vs. no ordinances. (c) Ordinances max. vs. ordinances

E Dynamic Results: Population Differences

326 January 2020 Amer. J. Agr. Econ.

F WTP: Extinction Years

0

5

10

15

20

25

P e r

ce n t

0 10 20 30 40 50 60 70 80

Years

Hatchlings Small juveniles Large juveniles Subadults Adults

0

5

10

15

20

25

P e r

ce n t

0 10 20 30 40 50 60 70 80

Years

Hatchlings Small juveniles Large juveniles Subadults Adults

0

5

10

15

20

25

P e r

ce n t

0 10 20 30 40 50 60 70 80

Years

Hatchlings Small juveniles Large juveniles Subadults Adults

(a) (b)

(c)

Figure 9. Population difference, per stage (% of stage initial population)

Brei, Perez-Barahona, and Strobl Protecting Species through Legislation 327

Table 13. Number of Extinction Years Avoided

Substitute Complement Complement Max.

FAU NOAA FAU NOAA FAU NOAA

Whitehead (1992) < 1 < 1 1 1 1 1 Whitehead (1993) < 1 < 1 1 1 < 1 < 1 Wallmo and Lew (2012) 1 < 1 2 1 2 1 Hamed (2013)l < 1 < 1 1 1 1 < 1 Hamed (2013)u < 1 < 1 1 1 1 < 1 Oceana (2008)l < 1 < 1 < 1 < 1 < 1 < 1 Oceana (2008)u < 1 < 1 1 < 1 < 1 < 1 Oceana (2008)(*) < 1 < 1 < 1 < 1 < 1 < 1 Whitehead (1992)(**) < 1 < 1 1 1 1 1 Whitehead (1993)(**) < 1 < 1 1 1 1 < 1 Wallmo and Lew (2012)(**) 1 < 1 2 1 2 1 Hamed (2013)l(**) < 1 < 1 1 1 1 < 1 Hamed (2013)u(**) < 1 < 1 1 1 1 < 1

Note: (a) Substitute: captivity-rearing as substitute for current STFL ordinances; (b) Complement: captivity-rearing as complement to current STFL ordinances; (c) Complement max: captivity-rearing as complement to max. STFL ordinances; (d) subindex l for WTP lower bound; (e) subindex u for WTP upper bound; (f). (*) Considering the conservative estimate of the number of active scuba divers; (g). (**) WTP of households + WTP of scuba divers in Oceana (2008).

328 January 2020 Amer. J. Agr. Econ.