RECREATIONAL ANGLER LOSSES FROM OFF-ROAD VEHICLE
RESTRICTIONS ON CAPE HATTERAS NATIONAL SEASHORE
Steven J. Dundas*
Oregon State University
Roger H. von Haefen
North Carolina State University
Carol Mansfield
Research Triangle Institute
Working Paper November 2015
Over the past decade, the tradeoffs between preserving biodiversity and restricting
outdoor recreation activities have been part of a contentious policy debate at Cape
Hatteras National Seashore in North Carolina’s Outer Banks. We estimate the welfare
implications and demand responses of shoreline recreational anglers to restrictions on
off-road vehicle (ORV) use designed to protect the nesting sites of endangered species. A
repeated discrete choice model estimates preferences for recreational fishing in the threestate region surrounding the National Seashore. We then evaluate two alternative ORV
policy scenarios to quantify the welfare and demand responses to the restrictions. Our
results suggest the long term economic costs are relatively modest, ranging from
$134,534 to $778,588 (2010 dollars) annually depending on the management strategy and
level of closures in adaptive planning areas. Of the 1.77 million trips predicted for the
three-state region, only 0.3 to 1.5 percent of trips per year are lost, suggesting anglers
adapt by continuing to take trips to affected sites (diminished trips) or by substituting to
non-affected sites (substitute trips). A more encompassing benefit-cost analysis suggests
that the aggregate costs are larger if additional factors are accounted for. Nonetheless, the
upper bound of potential welfare losses found here is still less than a conservative
estimate of benefits associated with preserving habitat for endangered species, implying
the policy passes a benefit-cost test.
Keywords: ORV restrictions, recreation demand, Cape Hatteras, endangered species
JEL Codes: Q26, Q28, Q51, Q57
*
The authors would like to thank Marty Smith, Laura Taylor, Wally Thurman, Zack Brown, Matthew
Godfrey, and Rob Johnston. Additional thanks to participants at the 2015 AAEA Annual Meeting for
helpful comments. Carol Mansfield and Roger von Haefen received support from the National Park Service
(Order No. T2310081036, GS-10F-0283K). The views expressed in this work belong to the authors and do
not reflect the views of the National Park Service.
1
Introduction
Off-road vehicle (ORV) use on public land is a popular, yet contentious, activity in the United
States. Operation of ORVs provides access to remote areas and outdoor recreation such as
fishing, hunting, hiking, camping, and other nature-based activities. Demand for ORV recreation
is growing, with nearly 43 million people participating each year (Cordell et al. 2008). The
annual economic activity associated with ORVs, including gear, vehicle, and trip expenditures, is
estimated to be nearly $66 billion dollars (OIA 2012).
The Bureau of Land Management and the U.S. Forest Service oversee ORV use on federal
lands in the nation’s interior and the National Park Service (NPS) oversees use on the coast.1
These agencies are tasked with management of the inherent conflicts between recreational ORV
access and the preservation of ecologically sensitive areas where ORV recreation occurs.
Opponents of ORVs contend that the vehicles cause irreparable damage to ecosystems and
wildlife habitats, pose safety concerns, and disrupt non-motorized recreation activities for other
individuals. On the other hand, ORV users and advocacy groups clamor for more routes,
increased access, and assert that ORV restrictions are likely to have adverse economic impacts
on communities adjacent to these areas. In the coastal context, the NPS allows over-sand ORV
access in eight coastal park units. The long-standing conflict of note centers on endangered
species habitat preservation (including piping plover and sea turtle nesting sites, among others)
versus unrestricted ORV use in North Carolina’s Outer Banks.
1
The NPS allows ORV use in 12 of 398 park units, with 8 being coastal parks (highlighted in bold below). The 12
units are Big Cypress National Preserve, Gateway, Glen Canyon, Curecanti, and Lake Meredith National
Recreation Areas and Assateague, Cape Cod, Cape Hatteras, Cape Lookout, Fire Island, Gulf Islands, and
Padre Island National Seashores.
2
In this article, we estimate the welfare losses associated with shoreline recreational fishing as
a result of ORV restrictions on Cape Hatteras National Seashore (CAHA). Data for the analysis
come from 2005-2007 Marine Recreational Information Program (MRIP) surveys, which are
collected in two-month waves by the National Oceanic and Atmospheric Administration
(NOAA). A repeated discrete choice model is implemented by sequentially estimating site
choice and participation decisions for anglers in the three-state region surrounding Cape
Hatteras.2 In the model, individuals choose whether, when, and where to engage in coastal
recreational fishing. The model is then used to predict the behavior and welfare effects
associated with time-varying ORV beach closures.
In the context of the ORV rules, a site’s closure for resource protection does not imply a
complete loss of recreation. Individuals have the ability to substitute to other locations unaffected
by the management policy. Alternatively, they can continue to take a trip to an affected site if
they are willing to bear additional costs associated with accessing the site by foot. Therefore, we
employ an econometric model that allows for substitution among alternative sites and change in
participation rates to capture different responses to the ORV policy. By allowing for this
substitution, the welfare estimates produced by the model may be viewed as long-run estimates
once anglers fully adapt to the on-site limits imposed on recreation activity.
The contributions of this research include being the first to estimate the long-run effects of
ORV restrictions on demand for coastal recreation, providing empirical evidence on the
economic costs of ORV restrictions to directly inform the contentious policy debate, and
2
Fishing sites in Virginia, North Carolina, and South Carolina and anglers taking trips to these states from coastal
counties are included in the analysis.
3
advancing the methods associated with recreation demand modeling with MRIP data.3 The
annual angler welfare loss due to ORV restrictions estimated in policy simulations range from
$134,534 to $778,588 (2010 dollars), depending on the rules in place and the extent of closures.
The incremental annual losses from switching from the NPS preferred alternative to the
environmentally preferred alternative (i.e., stricter ORV rules, more resource protection) are also
relatively modest, ranging from $101,767 to $230,419. The effects on total angler trips to sites
within and outside of CAHA are relatively small, with declines of 0.3 to 1.5 percent predicted for
the three-state region. Furthermore, a sensitivity analysis completely closing all CAHA sites
shows a 2 percent decline in regional recreational fishing trips. These results suggest that instead
of shoreline recreational fishing being entirely lost as a result of the ORV rules, anglers
substitute trips to other locations and continue to take diminished trips to affected sites that
remain open for fishing activities (but closed to ORVs).
The results above are subject to five important caveats. First, congestion is a salient concern
at sites on CAHA that remain open to ORV use. Therefore, we conduct a sensitivity analysis
where we incorporate increased levels of congestion into policy simulations. Our results suggest
the annual welfare costs would rise modestly with increasing congestion, ranging from $102,000
to $327,000. Second, the welfare estimates are based on trips originating in coastal counties as
defined in the MRIP. Because approximately 46 percent of CAHA user days are from
individuals residing outside of coastal counties, inclusion of non-coastal trips has the potential to
nearly double the losses reported here. Third, the ORV restrictions are costly to enforce, and
these costs were paid for through increased ORV permit fees. Fourth, other ORV-based
3
Additional ORV restrictions are currently being considered for Cape Lookout National Seashore in North Carolina.
4
recreation is not included in the analysis and anecdotal evidence is used to infer total recreation
losses may be up to two times higher. Using assumptions that allow for an upper bound estimate,
the losses to all recreation activities from the ORV restrictions could reach as high as $4.9
million annually. Lastly, the long-run welfare cost estimates may understate potential short-run
losses before anglers fully adapt to the site closures. A back-of-the-envelope calculation using
data from an auxiliary on-site visitor survey (Mansfield et al. 2010) suggests that demand
responses may be up to 2.4 times larger in magnitude in the first few years after the restrictions
are imposed. However, this upper bound of losses is still relatively small compared to a
conservative estimate of the potential benefits of endangered species protection in coastal North
Carolina derived using results from Whitehead (1993) and Dalrymple et al. (2012).
More broadly, the methods introduced here provide a framework for modeling time-varying
closures of a natural resource with MRIP data. This research estimates welfare losses associated
with closures for wildlife and habitat protection, but the methods could be applied to value
recreational losses associated with oil spills, fish advisories, and beach closures due to dangerous
levels of pollution.
2
Cape Hatteras National Seashore and ORV Policy
Cape Hatteras National Seashore was established as the first national seashore in the United
States in 1953. The barrier island park stretches over 67 miles of North Carolina’s coastline,
three islands (Hatteras, Ocracoke, and Bodie), and covers 24,470 acres (see figure 1).4 CAHA is
a relatively remote portion of the Outer Banks, with primary access available from a single
4
Pea Island National Wildlife Refuge encompasses nearly 6,000 acres on Hatteras Island and has year-round ORV
restrictions in place. It is administered by the U.S. Fish and Wildlife Service.
5
bridge (Herbert C. Bonner Bridge on NC Highway 12) on the north end and ferry service to
Ocracoke Island on the south side. Since 1989, CAHA has experienced annual visitation between
2 and 3 million people.5 These visitors use the islands for a variety of recreation activities,
including shoreline fishing. Recreation visits generate a robust local tourism industry that
supports eight unincorporated villages on the islands (Landry et al. 2015). Access to many of the
park’s prime locations for recreation activities is afforded by ORV use on the beaches
themselves due to limited parking and pedestrian access on the islands. ORV users must
purchase a permit ($120 for annual or $50 for seven-day), access the beach at designated dune
crossovers, and drive seaward of the primary dunes (i.e. driving on dunes is prohibited).
Concomitant with the growth in recreation visits and ORV use, CAHA also experienced a
decline in the population of nesting shorebirds, including the piping plover (Charadrius
melodus). This plover is a protected species under the federal Endangered Species Act and is
designated as threatened in North Carolina. The shorebird’s population along the Atlantic Coast
of the U.S. and Canada was estimated in 2011 at 1,762 nesting pairs (USFW 2011). The plover is
highly vulnerable to disruption from human interaction due to its shallow nests in the sand and
its camouflage that hides both birds and nests from view. The transitional over-wash areas along
the shoreline are prime nesting sites for these plovers and directly overlap with many ORV
routes along the seashore. CAHA also provides critical nesting habitat for two species of sea
turtle: loggerhead (Caretta caretta), and green (Chelonia mydas).6 Loggerhead turtles visiting
5
2011 was the lone exception with an estimated 1.96 million visitors. See:
https://irma.nps.gov/Stats/SSRSReports/Park Specific Reports/Annual Park Recreation Visitation (1904 - Last
Calendar Year)?Park=CAHA
6
Leatherback (Dermochelys coriacea), hawksbill (Eretmochelys imbricata), and Kemp's ridley (Lepidochelys
kempii) sea turtles are also found in the waters off North Carolina but nesting on CAHA is not common.
6
North Carolina beaches are classified as both endangered and threatened, depending on the subpopulation of the turtle while the green turtle populations retain a threatened status. For both
species, North Carolina is typically thought to represent the northern limits of their nesting
grounds.7 Nesting activity is vulnerable to coastal armoring (e.g. jetties and sea walls), nighttime
activity on the beach, and ORV use in nesting areas. The NPS began monitoring sea turtle
nesting sites in 1987 when only 11 were found.
The NPS is tasked with overseeing and managing recreational access and species protection
in CAHA. Their adaptive management plans are under the purview of Executive Orders (e.g.
E.O. 11644 of 1972 and E.O. 11989 of 1977) and federal laws, including the Endangered Species
Act, The Migratory Bird Treaty Act, and the National Environmental Policy Act. On the
recreation side, ORV culture is deeply ingrained with local residents and visitors to the area as
no restrictions on use and access were in place until recently. The Interim Protected Species
Management Strategy was drafted in 2007 to manage ORV use while details for a more
permanent strategy were finalized. Opposition from both sides of the issue was strong to these
proposed restrictions. The NPS was sued by wildlife advocacy groups (The Audubon Society
and Defenders of Wildlife) that claimed the interim rules did not go far enough to protect nesting
sites. A settlement was reached in April 2008 that allowed implementation of a temporary
strategy including larger buffers around sensitive areas and restrictions on nighttime ORV use on
CAHA beaches during the sea turtle nesting season.8
7
This is biologically important, as cooler sands produce more male hatchlings, making NC a critical breeding
ground for the male populations of each species.
8
This restriction appears to be helping as NPS has found an average nesting total in CAHA around 129 annually
from 2008 – 2011.
7
The final management rule aimed at balancing recreation interests with species protection
went into effect on February 15, 2012. The NPS chose between a number of different alternative
plans after a period of public comment (see National Park Service 2010; Mansfield et al. 2011).
The implemented plan was Alternative F, which provides a balanced approach between ORV use
and vehicle-free areas relative to other alternatives considered by NPS. Of the 67 miles of
coastline in CAHA, Alternative F designates 27.9 miles for year-round ORV routes, 12.7 miles
of seasonal routes, and 26.4 vehicle-free miles. Included with this alternative are planned
infrastructure improvements including parking lots at key locations and improvements to an
interdunal road system. Wildlife management areas and village beaches (areas directly adjacent
to a population center) are closed to allow for shorebird breeding activity, typically March to
July. This alternative re-opens these areas earlier and for longer periods of time than other
potential alternatives considered by NPS. Nighttime restrictions on all beaches are in place May
15th to September 15th from one hour after sunset until cleared by patrol in the morning. The map
in figure 2 illustrates the restrictions in place on June 1st, 2015, on Hatteras Island. Alternative D,
or the environmentally preferred alternative, was also considered by the NPS. This plan
protected the environment more so than other alternatives and limited ORV use to designated
year-round routes. No seasonal routes would be established and there would be minimal new
construction of ramps with no new parking areas added. This alternative would have provided
more certainty to visitors as ORV restrictions would apply to larger areas over longer periods of
time. All wildlife management areas and village beaches would be closed at all times to ORV
use and seasonally to pedestrians. In comparison to F, Alternative D would designate 27.2 miles
for year-round ORV routes, 0 miles of seasonal routes, and 40.8 vehicle-free miles. Nighttime
ORV use would be prohibited between 7 PM and 7 AM from May 1st to November 15th to
8
protect nesting turtles. Permit fees would be less under this plan due to the decreased need for
NPS oversight and management.
It is important to note that the CAHA final ORV rule is again being challenged due to a late
addition rider to the 2015 National Defense Authorization Act. Numerous public land provisions
were added to this omnibus legislation, including language forcing a review of the CAHA ORV
restrictions. The Secretary of Interior is now required to investigate the potential for reducing the
size of nesting buffers, opening beaches earlier in the morning during the summer, extending
ORV routes and access points, and modifying the size and location of restricted areas in CAHA.
As of June 16, 2015, the NPS began phasing in changes to wildlife buffers and the establishment
of corridors around closed areas in response to the bill. In August 2015, the NPS held five public
meetings to discuss the potential changes. The NPS plans to address all facets of the new law and
will report back to Congress with alterations to the final rule by December 2015. A revised
management plan will likely be implemented during the summer of 2016. The implications of
this legislation and the existing proposal to extend the ORV restrictions to Cape Lookout
National Seashore make the results of this research relevant to a highly-contested and current
policy debate.
3
Modeling Strategy
Two primary considerations when estimating non-market values for shoreline recreational
fishing under time-varying ORV restrictions are: 1) substitution across alternative sites and time
and 2) changes in participation rates. To address these issues, we use a repeated discrete choice,
random utility maximization (RUM) framework (Morey et al. 1993) to model whether, when,
and where anglers engage in shoreline recreational fishing. In this framework, an individual first
9
chooses whether to participate in coastal fishing and then conditionally chooses which site to
recreate at and when to go. Each individual makes a site choice and a bi-monthly wave choice
when deciding to recreate given the wave-specific ORV restrictions on CAHA. In other words,
the choice each angler faces is which site to choose and which bi-monthly wave to take a trip in a
given year. Figure 3 displays the sequential structure of the recreation decision for this analysis.
In the RUM framework, an individual chooses the alternative that maximizes her utility. The
factors that drive individual decisions can be separated into determinants that are either
observable or unobservable, with the latter treatable as random from a modeling perspective. The
conditional indirect utility of individual i from choosing site j and wave w on choice occasion t is
specified as follows:
Vijwt  U (mit  cijw , X jwt ,  ijwt )
(1)
where mit is income, cijw is travel cost,9 Xjwt is a vector of site characteristics, and  ijwt captures
idiosyncratic, random factors. Conditional on taking a trip, a rational angler selects the site/wave
that generates the highest utility. More precisely, angler i chooses site j in wave w for their
recreation activity if Vijwt  Vikwt ,  k , w . For convenience, we assume utility is linear and
additive in  ijwt (i.e. Vijwt  vijw   ijwt ).
As discussed in a later section, our data come from two separate and independent NOAA
surveys that provide repeated cross-sections of site choice and participation. The data limitations
from the surveys imply that we do not observe both where and how often a given individual
9
Because our income and per mile driving cost data only varies annually, we make the further assumption that
travel costs are equal across waves within a year for each individual/site pair.
10
recreates, and thus we are restricted to econometric models that completely separate these two
dimensions of choice.10 Therefore, we employ a two-level nested logit model (Morey 1999) and
sequentially estimate a conditional site/wave choice model and a participation model. The model
assumes the errors have a generalized extreme value distribution, which allows a common
random component to enter the site-specific errors. This induces correlations in the conditional
indirect utilities of each site and more reasonable substitution patterns. Then, given data in this
analysis on both participation and site choice, sequential estimation generates a complete and
consistent model of recreation behavior.11
The probability of choosing site j and wave w on choice occasion t is:
Pijwt  Pit ( j , w | trip )  Pit (trip )


e
W
 vijw /  
J
 e
w1 j 1
vijw / 
 1
W J
 W J vijw /  
v / 
v /  
e ijw   e ijw 
  e

 w1 j 1

 w1 j 1




W
J
W
J


vijw /  
vijw /  
vi 0
vi 0
e    e
e   e


 w1 j 1

 w1 j 1

(2)
where  is the dissimilarity coefficient, bounded by theory between 0 and 1, and W and J are the
total number of waves and sites. The probability of not taking a trip is then:
Pi 0 wt 
evi 0
 W J vijw /  
e    e

 w1 j 1


(3)
vi 0
10
The data do not allow estimation of a random coefficient or latent class model.
The large datasets used here suggest that efficiency loss relative to full-information maximum likelihood
estimation is relatively small.
11
11
where j = 0 denotes the non-participation alternative. All parameters of this model can be
estimated by first estimating the site/wave choice model and then conditionally estimating the
participation model using standard logit estimation techniques (Ben-Akiva and Lerman 1985).
3.1
Empirical Implementation of the Model
Models are estimated using data from 2005 to 2007 to avoid the potential confounds associated
with nighttime driving restrictions that were implemented in April 2008. Estimation of the model
proceeds in three steps. Step one utilizes the MRIP intercept data to estimate a conditional logit
site/wave choice model with a full set of alternative specific constants (ASCs) separately by year
(2005 – 2007). The conditional indirect utility (i.e. equation 1) from individual i visiting site j in
wave w on choice occasion t is defined as follows:
Vijwt   cijw   jw   ijwt
(4)
where we employ the common assumption of a constant marginal utility of income. The site
choice and wave probabilities can then be written:
Pit ( j , w | trip) 
e
W
  ciwj  jw  / 
J
 e
 ciwk  kw  / 
(5)
w1 k 1
where  jw is an ASC specific to each site/wave pair. The ASCs are designed to capture site and
wave specific characteristics that vary across years (i.e. catch rates, species composition) but are
common across sites during a given wave. Although this approach does not explicitly include
site characteristics, the ASCs should control for them to the extent that they do not vary across
individuals within each wave. Note the first stage estimation does not permit separate
12
identification of the dissimilarity coefficient (λ) from (  ,  jw ) . Rather, the ratios  /  and  jw / 
are estimated in steps 1 and 2. As described below, we lay out a strategy for estimating λ
separately in step 3, which in turn allows us to back out  and  jw .
This first step generates consistent estimates for the normalized travel cost coefficients (β/λ).
However, since the MRIP survey samples at most 40 percent of fishing sites in each year/wave,
this estimation does not generate consistent estimates for the ASCs because ASCs are not
identified for unsampled sites. Therefore, the second step in estimation involves calibration to
recover estimates for all ASCs. Fortunately, NOAA maintains site registry files with data on
fishing pressure (or expected visitors in a normal 8-hour period) that we utilize to construct
aggregate trip estimates for every site/wave/year combination. The MRIP intercept data is then
used to estimate the share of trips originating from both coastal and noncoastal counties in order
to adjust the aggregate estimates to reflect predicted localized (i.e. coastal) recreation trips. These
estimates are then converted to shares of total trips and Berry’s (1994) contraction mapping is
used to recover calibrated estimates of the ASCs (see Appendix A for further discussion of this
calibration procedure). The adjusted ASCs are then used to generate the inclusive value index:
 W J   /  cikw  k  
IVi  ln   e

 w1 k 1

(6)
where  /  is estimated in step one and  is estimated the second step. The inclusive value, or
log-sum, term IVi can loosely be interpreted as the expected utility of a trip (Hausman et al.
1995).
13
The third and final step estimates a standard discrete choice logit model of participation as a
function of inclusive values, demographics, and year fixed effects. The utility function associated
with non-participation is defined as:
Vi 0t   0  IVi  + Y  Di   i 0t
(7)
where Y and 𝑫𝑖 are vectors of year dummies and demographics. The recovered parameter
estimate on IVi is the dissimilarity coefficient    , which can be multiplied by the first-stage
normalized travel cost coefficients to recover a consistent estimate of β.
4
Data
MRIP data from NOAA’s National Marine Fisheries Service used for this analysis include the
point-of-access Angler Intercept Survey and the Coastal Household Telephone Survey. Table 1
provides a concise summary of all variables used in each stage of the analysis. For the intercept
data, observations are restricted to individuals fishing at shoreline sites in Virginia, North
Carolina, and South Carolina from 2005 – 2007. The timeframe is restricted to these years to be
representative of the region before any new ORV restrictions were implemented. The data are
compiled in two-month intervals, resulting in six waves per year.12 This survey collects data
from interviewed individuals on their catch and mode of fishing. The primary variable of interest
for this work is the zip code of residence for each survey respondent. This information allows
estimation of travel costs from their home to the fishing site where they were intercepted for the
survey. Note that each individual is interviewed only once so the resulting dataset is a repeated
12
Waves: 1 = Jan/Feb, 2 =Mar/Apr, 3=May/Jun, 4 = Jul/Aug, 5 = Sep/Aug, & 6 = Nov/Dec.
14
cross-section. Intercept sites along the Atlantic coast of Virginia, North Carolina, and South
Carolina (344) and origin zip codes of intercepted anglers taking trips to these sites (4,657) are
geocoded for inclusion in the analysis. The restriction to shoreline fishing yield a sample size of
17,559 intercepts across 18 waves. These observations are used to estimate the conditional
site/wave choice model in step one described in the previous section.
Consistent estimation of our nested logit model requires an adjustment for the stratified
survey design. The intercept survey is stratified by site, state, mode, year and wave, implying
that where anglers fish is correlated with their likelihood of inclusion in the sample. To address
this choice-based sampling attribute of the data, we employ MRIP’s recently introduced
sampling weights. In 2012, the NMFS introduced design-based sampling weights and variance
adjustments to the intercept data and applied these weights to the survey results dating back to
2004. By construction, these weights produce unbiased estimates of angler effort and reflect the
proper proportion of trips from coastal and non-coastal origins (Breidt et al. 2012; Lovell and
Carter 2014). Use of these weights to quasi-randomize the sample obviates the need for
econometric solutions to endogenous stratification (e.g. Hindsley et al. 2011).
Calculation of the travel costs for each observation involves several steps. First, PC*Miler
calculates the one-way driving distance (dist), travel time (time), and tolls (toll) from an origin
zip code centroid to all potential shoreline fishing sites in a given choice set. Next, additional
data are obtained on average fleet fuel economy (fe) from the U.S. Department of Transportation,
gas prices by state (gas) from the US Energy Information Administration, automobile per-mile
operation costs (cpm) from AAA, and zip code level income from the U.S. Census Bureau. Costs
that can be shared by all persons on a given trip (e.g. tolls, gas, and mileage) are divided by the
15
average number of individuals in each party from our sample ( n ).13 The opportunity cost of time
(oppc) is determined using the common assumption of 1/3 of the wage rate (Cesario 1976).
Round trip travel costs (in 2010 dollars) for individual i to site j at time t are calculated as:
cijwt
  distij *(cpmt )  




 2*  (distij / fet )* gast  / n  oppcit * timeij 
 



  tollij

(8)
From the phone survey, information is collected by county-stratified random-digit-dial
(RDD) from coastal households from 2005 to 2007 on the frequency of fishing trips in the
preceding two months.14 The data include the number of anglers who have taken trips and the
number of trips taken by each angler in the previous two months. For this study, the phone
survey pulls from coastal counties in the three-state region surrounding the Outer Banks of North
Carolina with six-digit phone exchanges as the spatial unit of analysis. Data from 4,928 phone
exchanges are included. Survey records for individuals who actively participated in coastal
recreation fishing are used to characterize participation in the model. However, the nonparticipating households are only identified at the county level – a less refined spatial scale then
the fishing households. Since the phone survey was RDD within each county, we disaggregate
the county level non-fishing data to the phone exchange level to facilitate analysis in the
following way. Each relevant six-digit phone exchange is assigned a population weighted
proportion of the count of non-fishing individuals in the county where the exchange in located.
For example, assume a county with three phone exchanges, each with a population of 1,000
13
The intercept data contains a variable for the number of people per fishing party. The average is 2.73.
Another reason for choosing the 2005-2007 timespan for this analysis is to limit the potential impacts on
representativeness of RDD surveys due to growing cell phone use and a decreased willingness to participate in
phone surveys in recent years.
14
16
people. If the RDD survey contacted 30 non-anglers in the county, randomization implies we can
assign 10 non-anglers to each phone exchange in that given two-month period.
As noted earlier, the third step of model estimation includes demographics and year dummies
as covariates affecting the participation decision. Zip code level demographic data on average
household income, race, sex, population density, and education are gathered from U.S. Census
Bureau’s American Community Survey. Zip codes are linked to six-digit phone exchanges using
a proprietary data set from Melissa Data.15 A population-weighted average of zip code
demographic data is assigned to each phone exchange.
5
Results
First stage estimation of the conditional site/wave choice model yields travel cost coefficients
that are robust and negative as expected across all years (see panel A of table 2). These estimates
are highly significant as evidenced by the large t-statistics. As a sensitivity check, we also
estimated a separate model (not reported here) with wave and site specific weather variables (i.e.,
temperature and precipitation) included. The magnitude and precision of the travel cost estimates
did not change, and the sign and significance of the weather variables fluctuated substantially
from year to year. We suspect the relative uniformity in weather conditions across the three-state
region and the inclusion of site/wave constants absorbed most of the variation in weather from
year to year and wave to wave, leaving little meaningful variation to be captured by our weather
variables.
15
Dataset description available here: http://www.melissadata.com/reference-data/fonedata.htm
17
Two sets of results for the logit participation model are reported in panel B of table 2. Yearspecific inclusive values are included in Model 1 yet this specification does not produce
significantly different results from the preferred model (Model 2).16 For each phone exchange,
increases in the percent of the male population and percent of individuals with bachelor’s
degrees (or higher) are estimated to increase participation rates. Higher population densities and
average household income are estimated to decrease participation in shoreline recreational
fishing.
The estimate for the dissimilarity coefficient in the preferred model is 0.04, which falls
within the 0-1 interval and thus is consistent with RUM theory (Herriges and Kling 1997).
However, using the formula reported in Haab and McConnell (2002),17 the estimate implies a
relatively large per trip value of approximately $342. We suspect that this finding is driven by
the imprecise nature of the trip origin information in the phone survey data. In particular, the
phone data only includes the respondent’s phone exchange, not a more geographically precise
origin such as a zip code. As a result, measurement error is introduced into the inclusive values
which in turn likely generates attenuation bias with the estimated dissimilarity coefficient.18, 19
16
Largely for computational reasons, we do not correct the second-stage standard errors to account for the fact that
generated inclusive values are constructed with travel cost parameters that are econometrically estimated in the first
stage. Because these parameters are tightly estimated, however, we do not suspect that doing so would imply
substantially different inference.
17
This estimate is constructed using the formula -1/ for a value of trip, where  is estimated by taking the product
of the mean value of the first-stage normalized travel cost estimate (-0.073) and the estimated dissimilarity
coefficient (0.04). See Haab and McConnell (2002) for a derivation of this result. This approximation should be
relatively good given our large choice set application.
18
Observations where phone exchanges did not match the reported state of residence were removed from the
analysis to minimize the attenuation bias.
19
Moreover, the fact that the ASCs that feed into the inclusive values are calibrated with fishing pressure data in the
site registry and not precisely estimated with choice data may introduce additional measurement error.
18
In order to correct for this in our simulations, we calibrate the dissimilarity coefficient using
a $30 per trip value, which is consistent with empirical results reported in Moeltner and
Rosenberger’s (2014) and Johnston and Moeltner’s (2014) recent meta-analyses.20 As shown by
Haab and McConnell (2002), the value of a fishing trip in a repeated RUM framework is
approximately -1/β, where β is the travel cost coefficient. From our first stage trip allocation
model, we have an estimate of β/λ. Therefore, if we assume a $30 value of a trip, we can infer
the value of λ. As reported below, our welfare results are robust to not incorporating this
calibration and instead using the estimated dissimilarity coefficient. However, the demand
responses are quite different, as the predicted number of lost, substituted, or diminished trips
vary across the models with calibrated and uncalibrated dissimilarity coefficients.
5.1
Policy Simulations
The policy simulations explore how demand for shoreline recreational fishing is impacted under
two different ORV management scenarios for CAHA. As described in a previous section,
Alternative F is the NPS preferred alternative that was eventually implemented in 2012 and
Alternative D is the environmentally preferred alternative. There are 16 MRIP survey sites in
CAHA that would be impacted by either strategy. Tables 3 and 4 display the different
management strategies at these sites under Alternatives F and D. The letter ‘O’ indicates that the
20
Moeltner and Rosenberger (2014) report the average WTP/day for a saltwater fishing trip in the Northeast is
$39.39 (2010 dollars) from five relevant valuation studies. In Johnston and Moeltner (2014), the authors show that
the mean Hicksian WTP/day from 14 different studies for saltwater fishing of big-game species is approximately
$33.06. They also report the average WTP/day for small-game saltwater fishing across 13 studies as $21.33.20 Since
neither meta-analysis contains a directly equivalent value for this research (i.e. all shoreline fishing in the three-state
region), the average of the meta-analysis WTP/day means from the two studies (~$30) is used here as an
approximation of the value of a trip.
19
site is open during that particular wave with no impact from the proposed management plan. ‘X’
indicates that ORV use is restricted and the site is likely only accessible by ORV, rendering the
site effectively closed for modeling purposes. ‘XP’ denotes that ORV use restrictions are in place
but the site is still open to pedestrian access. Lastly, ‘A’ indicates adaptive planning areas where
a site could either have no impact or be closed for resource protection if nesting activity is
present.
The policy scenarios for site/wave choice combinations ‘O’ and ‘X’ are straightforward to
model – sites remain open (‘O’) or are closed (‘X’). Modeling the other two possibilities
necessitates estimating a range of possibilities (i.e. an upper and lower bound) due to the
specifics of the management strategies. For `XP’ sites, access on foot is not restricted, but
individuals would have to walk considerable distance carrying their gear to reach the fishing site
(0.5 – 2 miles one way) from the closest available parking lot. In some cases, the distance to
cover on foot would be excessive, so the site would essentially be closed (i.e. upper bound of
potential damages). Alternatively, if a site is reasonably accessible by foot, two hours of roundtrip travel time are added to account for additional opportunity cost of time needed to access the
site. This represents the lower bound of potential damages under `XP’ restrictions. The range of
options for the adaptive strategy ‘A’ is more direct. The upper bound is modeled as if the site is
closed (i.e. strategy `X’) and the lower bound as if the site is open and unaffected (i.e. strategy
`O’). These modeling assumptions are summarized in table 5.
As noted earlier, nighttime driving restrictions were imposed in 2008 so the primary policy
simulation analysis uses 2005 – 2007 as the baseline before any ORV restrictions were imposed.
ASCs and inclusive values (equation 5) are estimated under each alternative strategy. The change
20
in WTP for coastal shoreline recreational fishing for individual i under each policy scenario is
estimated using the following equation (Haab and McConnell 2002):
WTPi  
ln e
 
Ti
 vi 0
 exp( IVi1 ) 

  ln e
 vi 0
 exp( IVi 0 ) 


(9)
where  is the travel cost parameter, Ti is the number of choice occasions, IVi 0 is the inclusive
value for individual i in the baseline period and IVi1 is inclusive value under the policy scenario.
In equation (9), the differences in inclusive value terms from the baseline to each policy scenario
drive the differences in WTP.
The policy simulations yield three primary results, which are reported in tables 6 and 7.
Standard errors for the welfare and demand response estimates are generated with a parametric
bootstrap (Krinsky and Robb 1986) with 50 draws taken from the estimated parameter vector and
covariance matrix. All predictions for welfare and demand responses are highly significant, as
suggested by the 95% confidence intervals reported in the table. The primary result is the
relatively modest damage estimates predicted under both management alternatives. The total
range of welfare losses is $134,534 (Alternative F, lower bound) to $778,588 (Alternative D,
upper bound) annually. The annual welfare loss under the most restrictive scenario (i.e. upper
bound) of the current management plan (Alternative F) is estimated to be approximately
$548,169 per year. Second, the projected incremental cost associated with moving from the
current plan to the environmentally preferred Alternative D ranges from $101,767 to $230,419.
These estimates represent the cost of moving from the current plan to one with more protections
in place for threatened wildlife.
Third, long-run demand responses to the policies predicted in the simulations are also
relatively modest. These results support the notion that, over time, anglers are adapting to the
21
ORV restrictions and choosing alternative sites for shoreline recreational fishing activities. For
each year, the model is utilized to estimate of the number of affected trips (i.e., trips impacted by
the ORV restrictions), lost trips (trips that occur in the baseline scenario but not with ORV
restrictions in place), and diminished trips (trips to impacted sites that continue to occur with
ORV restrictions but generate less utility to anglers) as a result of the ORV restrictions.
Substitute trips, or trips that were shifted to non-impacted sites after the ORV rules were
implemented, are then calculated using the following identity:
Affected Trips = Lost Trips + Diminished Trips + Substitute Trips
(10)
In sum, approximately 53,354 trips are affected under Alternative D and 38,313 under
Alternative F. The upper bound on the number of lost trips for Alternative F is 1 percent of total
trips in the three-state region, or approximately 18,125 trips. Switching from F to D at the upper
bound only increases the decline to 1.5 percent, or an additional loss of 7,630 trips. Of the total
affected trips, 71 to 76 percent are diminished trips in the lower bound simulations. By
construction, there are zero diminished trips at the upper bound for each scenario as all sites that
have the potential to be closed are closed. At the lower bound, substitute trips represent
approximately 12 to 15 percent of affected trips, with the aggregate numbers being relatively
similar to the numbers of lost trips. When diminished trips are not possible in the upper bound
scenarios, substitute and lost trips each account for roughly half of the affected trips. If adaptive
management areas remain accessible (i.e. lower bound), diminished trips account for a majority
of the affected trips with lost trips only accounting for less than 15 percent and 12 percent for
Alternatives D and F, respectively.
For comparison purposes, the above simulations are also estimated using the empirically
estimated values for the dissimilarity coefficient and constant term reported in table 2. These
22
results are presented in table B.1 in the appendix. Despite the large difference in the implied
value of a trip (i.e. $30 calibrated; $342 uncalibrated), the welfare results are remarkably similar.
The average increase in total welfare losses in the upper bound and lower bound scenarios are
4.7 percent and 1.2 percent respectively. These numbers are likely similar because anglers
behave and adapt in different ways when the value of a trip is substantial higher. In the
uncalibrated case, a lost trip is more costly in welfare terms so anglers take significantly more
substitute trips. For example, under Alternative F at the upper bound, anglers at the calibrated
trip value choose to take approximately 16,500 less trips than anglers at the high trip value
because of the ORV rule. The simulation then predicts those same anglers will increase their
substitute trips by the same amount, offsetting the decrease in lost trips.
5.2
Additional Costs
This section identifies five limitations associated with the main results presented in the previous
section that imply additional costs that should be incorporated into a full benefit-cost analysis of
CAHA ORV restrictions. These limitations center around the following five issues: 1)
congestion; 2) non-local angler recreation; 3) other ORV-based recreation (i.e. surfing); 4)
enforcement costs; and 5) short-run policy impacts.
5.2.1
Congestion Effects
Given that ORV use is being restricted at some CAHA fishing sites and anglers are potentially
substituting to other sites, increases in congestion at CAHA sites that remain open is a salient
concern. Welfare estimates presented above may be biased downward as they ignore any lost
angler utility from recreating at more congested sites. To assess the potential magnitude of such
23
biases, we conduct policy simulations that add congestion costs to the travel costs of visiting all
CAHA sites that remain open under each ORV restriction scenario. Our congestion cost estimate
comes from a study by Hindsley et al. (2007), who estimate that visitors experience a $6.40 loss
(2010 dollars) for a 10 percent increase in congestion at beaches along the Outer Banks.
Results from these simulations are presented in panel B of table 6 and in table 7. 21 At the
upper bound, congestion has the potential to increase annual long-run welfare losses in
Alternatives F and D by approximately $185,179 and $103,010, respectively. At the lower
bound, the potential increases in welfare losses are $326,026 (F) and $292,394 (D). Lower bound
estimates are larger here due to the fact that more sites remain open and therefore susceptible to
congestion than under the upper bound policy simulations. Although the percentage increases in
cost might appear large, the overall magnitude of the welfare losses remains quite modest. In
terms of the demand response, our preferred model predicts that lost and substitute trips will
increase as a response to congestion. For instance, under Alternative F, the number of lost trips
increases by approximately 10,800 (lower bound) and 6,100 (upper bound). Substitute trips
increase by 9,400 (lower bound) and 3,100 (upper bound). Again, these are relatively large
percentage increases but not large in absolute magnitude.
5.2.3
Local and Non-local Recreation
The lost consumer surplus estimates above only apply to local anglers residing in coastal
counties as defined by the MRIP survey. Over the time period studied (2005-2007), MRIP
intercept data indicate that anglers who live in these areas are responsible for approximately 54
21
Table B.2 in the Appendix presents the congestion results for the uncalibrated model.
24
percent of the user days to sites in the Outer Banks of North Carolina. This implies that 46
percent of the user days are from anglers residing in non-coastal counties that are not represented
in the above welfare analysis. Comparison of coastal and non-coastal angler user days in the
weighted MRIP intercept data reveal that number of hours fished, catch rates, number of fish
landed, and mode of fishing is relatively similar between these groups at varying spatial scales
(three-state region, North Carolina, and Dare County, NC). These comparisons are reported in
table 8. Based on these estimates, we conclude that it is plausible to assume an equal value of
user days for both local and non-local anglers. Based on this assumption, total welfare losses
have the potential to be roughly double the results reported above.
5.2.3
Other ORV-based Recreation
Shoreline fishing is not the only recreation activity pursued by ORV users on CAHA. For
example, access to popular surf breaks (e.g. “The Hook” just south of Cape Point) is also limited
under the NPS management plan and these closures are likely to have additional welfare
implications for other recreation. A NPS-sponsored user survey on CAHA conducted in 2009 –
2010 (Mansfield et al. 2010) found that shoreline fishing represents a large share of recreation
activities impacted by the ORV rule, as 38 percent of respondents reported “beach fishing” as an
activity they had taken part in during their current trip. “Swimming, sun-bathing, or enjoying the
beach,” “bird-watching,” and “surfing” are other recreation activities garnering relatively high
percentages in the survey. However, for our purposes, this estimate is of limited value because
respondents include both ORV and non-ORV users of CAHA. Nonetheless, our sense is that
anglers remain the primary recreation group impacted by the ORV restrictions as evidenced by:
1) the avid opposition to the rules by fishing clubs (Williams 2012), and 2) anglers representing
25
the large majority of individuals present to voice opinions at recent NPS public meetings to
discuss modifications to the current ORV rules. Additionally, anecdotal evidence from our
discussions with NPS staff, ORV users, and journalists suggest that a plausible assumption for an
upper bound on the percentage of ORV users that are not anglers is about 50 percent. Therefore,
we again double the welfare effects to account for the upper bound of other potential ORV-based
recreation losses.
5.2.4
Enforcement Costs
The new rules on CAHA for ORV use require purchasing a permit to drive on the beach. An
annual pass can be purchased for $120 and a seven-day pass for $50. Since 2012, the Park
Services has sold approximately 10,000 annual passes and 20,000 daily passes each year. This
represents an additional cost to recreators of approximately $2.2 million. These fees are used
primarily to cover enforcement costs within CAHA but also to maintain and build parking lots
and access points for ORV users. Although new parking lots and access points are likely to
benefit ORV users, we conservatively assume all of the fees are social costs that generate no
benefits to ORV users.
5.2.5
Short-run and Long-run Impacts
Lastly, it is important to note that these results are based on empirical estimates identified off of
angler decisions at distinct points in time, not changes in those choices over time. This implies
the estimated model reflects long-run substitution patterns. This has the potential to mask
potentially larger short-run impacts before anglers can re-optimize their choices after the
implementation of the ORV restrictions. Intuitively, these short-run impacts are likely to last
26
only 1 to 2 years, because most anglers have relatively short planning horizons for recreational
fishing trips.
To provide a perspective on the potential magnitude of short-run welfare loss, we again use
data from the auxiliary on-site user survey (Mansfield et al. 2010) to estimate these impacts.
Unfortunately, the on-site survey data does not provide a direct match to the localized recreation
scope of the demand analysis and does not differentiate between ORV and non-ORV recreation.
Therefore, the following estimates should be interpreted cautiously. For residents of the Outer
Banks, none of respondents report that they would be unlikely to take a trip under Alternative D
and would instead substitute to other locations. Responses to the plan that was eventually
implemented (Alternative F) were not evaluated in the survey but it is reasonable to assume no
impact since F is less restrictive than D. This implies the additional short-run welfare loss
associated with lost trips for local residents is minimal. For out-of-town visitors (that is, anyone
not a resident of the Outer Banks), the survey results suggest that under Alternative F, 6.3
percent of visitors would have been unlikely to take a trip in the current year to CAHA due to
the ORV restrictions. Of the 6.3 percent not taking a trip to CAHA, 3 percent were likely to take
substitute trips and 3.3 percent were likely to not take a trip at all. This percentage of lost trips in
the short run found in the survey is approximately 2.4 times higher than the long-run estimates in
the demand analysis with congestion. We therefore assume that lost trips may increase in the
short-run by a factor of 2.4. Assuming trips and welfare are perfectly correlated, short run
welfare losses are likely to be 2.4 times higher than the long-run estimates in the initial years of
the restrictions.
In sum, all five issues discussed above imply that additional costs arise from ORV
restrictions that are missing from our initial cost estimates. As an illustration of their potential
27
magnitude, let us consider our upper-bound cost estimates under the current management plan,
Alternative F. Our model estimates a welfare loss to local anglers of approximately $548,169 per
year. Including congestion increases this loss to $733,348 and adding welfare losses associated
with non-local anglers may raise the loss as high as $1.36 million annually. Then, if this number
is scaled up to include all other ORV-based recreation, total long-run welfare losses have the
potential to reach $2.7 million annually. Adding in the enforcement costs brings the total to
approximately $4.9 million. Given that estimate, short-run losses (with congestion, non-local
anglers, enforcement costs and all ORV-based recreation) then have potential to be as high $11.8
million in the first year the rules are implemented. This back-of-the-envelope exercise is
summarized in table 9. It is important to note that this is an upper bound calculation and is likely
overstating the potential increases in welfare loss compared to the model estimates.
5.3
Discussion of Potential Benefits
To put the welfare loss estimates from the analysis and the back-of-the-envelope calculation
above into perspective, it is important to consider the benefits for wildlife preservation that are
motivating the ORV restrictions. Economic estimates for such non-market goods are typically
generated with stated preference survey methods and the most relevant for the analysis here is
Whitehead (1993). This study presents results from a contingent valuation survey of wildlife
preservation programs in coastal North Carolina. Whitehead (1993) estimates the annualized
household WTP (1991 dollars) for loggerhead sea turtle preservation of $10.98 and for all nongame endangered and threatened species in coastal North Carolina (including piping plovers and
28
loggerhead sea turtles, among others) of $14.74.22 Aggregating these estimates to all North
Carolina households (approximately 3.7 million) and inflating them into 2010 dollars produces
aggregate benefits of $65 million to $87.3 million annually.23 According to N.C. Wildlife
Resources Commission’s seaturtle.org (September 30, 2015) website, approximately 20 percent
of North Carolina loggerhead nests occur in CAHA. Multiplying the aggregate benefit estimates
by one-fifth to bring the spatial scale of the benefits in line with the costs yields annual benefits
of $13 million. Thus, the long-run benefit cost ratio is 2.6 when comparing the potential upper
bound of costs and a conservative estimate of the benefits. The short-run benefit cost ratio is also
positive (albeit marginally so) even after multiplying the long-run welfare losses by 2.4.
It should be recognized that our analysis does not include all the costs and benefits associated
with ORV restrictions in CAHA. First, we do not include lost profits to local businesses (i.e.
shops, hotels and restaurants) from a reduction in visitors. Given the vocal opposition to ORV
restrictions by local businessmen, these losses very well may be substantial. But from a strict
benefit-cost perspective, any lost profits associated with diminished economic activity in CAHA
should be offset by increased profits at non-CAHA businesses to the degree that visitors respond
to the ORV restrictions by spending their dollars elsewhere. Ultimately, the overall effect on
economic activity is an empirical question worthy of further research, but our sense is that the
net impact on the North Carolina (or, more broadly, the Southeastern) economy is minimal.
The contingent valuation questions were asked as follows: “Suppose that a $A contribution from each North
Carolina household each year would be needed to support and fund the loggerhead sea turtle program (nongame
wildlife management program). Would you be willing to contribute $A each year to the 'Loggerhead Sea Turtle
Preservation Trust Fund' ( 'Coastal Nongame Wildlife Preservation Trust Fund') in order to support the loggerhead
sea turtle program (nongame wildlife management program)?" where A = {1,5,10,25,50,100}.
23
Whitehead’s (1993) estimates may be viewed as conservative when compared to other studies of WTP for
endangered species protection. Dalrymple et al. (2012) find household WTP for non-game endangered species in
North Carolina of up to $98.80 per year with a preferred tax mechanism. Stanley (2005) finds household WTP for
local endangered species in Orange County, CA between $50 and $60 annually.
22
29
Second, our analysis does not take into account benefits accruing to non-ORV users who derive
more satisfaction from CAHA trips with fewer ORVs. Our informal discussions with several
CAHA visitors suggest these benefits are real, and research by Magee (2008) suggest they may
be substantial. Finally, we do not include in our benefit calculations the WTP for endangered
species protection in CAHA of residents outside of North Carolina. Accounting for these
benefits would surely make the benefit-cost ratio more favorable to the policy.
6
Conclusion
This article examines the impact of endangered species protection and ORV restrictions on
coastal recreation. Our results suggests that the welfare losses to shoreline recreational anglers
associated with ORV restriction on CAHA are relatively modest compared to the likely benefits.
These estimate range from $134,534 to $778,588 annually depending on the management
strategy and level of closures in adaptive planning areas. The results also suggest that switching
from the current management scheme to an alternative with greater focus on preservation of
threatened species would have a marginal impact on welfare. Demand responses to the policies
are relatively small and suggest that anglers adapt in the long-run to ORV restrictions. It is
important to note that the welfare losses estimated here only provide guidance on a portion of the
economic impacts of the ORV restriction on CAHA. When we account for congestion, non-local
recreation, enforcement, other ORV-based recreation impacts, and short-run impacts, estimated
benefits still outweigh costs.
In addition to the policy-relevant welfare results, our research also provides a methodological
framework for estimating welfare effects of time-varying closures using the MRIP data – a large,
publically available dataset. The methods here could be applied to value welfare losses related to
30
any type of closure, including oil spills, fish advisories (e.g. high dioxin levels in catfish and carp
in Albemarle Sound, NC), and beach closures due to dangerous levels of pollution (e.g. sewage
drainage issues after rain event in Southern California).
Our results also come at a time when an additional federal statute, the 2015 National Defense
Authorization Act, has forced the NPS to again revisit ORV management on CAHA. The
policies have been highly contested from both sides over the past decade and this new legislation
continues that debate. When ultimately resolved, the final outcome in CAHA is likely to have
far-reaching impacts on shoreline management strategies in other NPS-administrated areas, such
as Cape Lookout (NC) and Padre Island (TX) National Seashores. Overall, the political
uncertainty surrounding this issue has potential to increase total welfare losses in the near term as
anglers cannot fully adapt to shifting regulations. If a policy can be finalized, our research
suggests that long-run adaptation has the potential to minimize social costs.
References
Ben-Akiva, M. and S.R. Lerman. 1985. Discrete Choice Analysis. Cambridge, MA: The
MIT Press.
Berry, S.T. 1994. “Estimating discrete-choice models of product differentiation.” RAND
Journal of Economics 25:242-262.
Breidt, F. J., H.-L. Lai, J. D. Opsomer, and D. A. Van Voorhees. 2012. A Report of the
MRIP Sampling and Estimation Project: Improved Estimation Methods for the Access Point
Angler Intercept Survey Component of the Marine Recreational Fishery Statistics Survey.
Silver Spring, MD: NOAA National Marine Fisheries Service, January.
Cesario, F.J. 1976. “The Value of Time in Recreation Benefit Studies.” Land Economics
52:32-41.
Cordell, H.K., C.J. Betz., G.T. Green, and B. Stephens. 2008. Off-highway Vehicle
31
Recreation in the United States and its Regions and States: A National Report from the
National Survey on Recreation and the Environment. Washington DC: U.S. Forest Service,
February.
Dalrymple, C.J., M.N. Peterson, D.T. Cobb, E.O. Sills, H.D. Bondell, and D.J.
Dalrymple. 2012. “Estimating Public Willingness to Fund Nongame Conservation Through
State Tax Initiatives.” Wildlife Society Bulletin 36:483-491.
Haab, T.C. and K.E. McConnell. 2002. Valuing Environmental and Natural Resources.
Northampton, MA: Edward Elger Publishing.
Hausman, J.A., G.K. Leonard and D. McFadden. 1995. “A Utility-Consistent, Combined
Discrete Choice and Count Data Model Assessing Recreational Use Losses Due to Natural
Resource Damage.” Journal of Public Economics 56:1-30.
Herriges, J.A. and C.L. Kling. 1997. “The Performance of Nested Logit Models when
Welfare Estimation is the Goal.” American Journal of Agricultural Economics 79:792-802.
Hindsley, P. C.E. Landry, O. Bin, and H. Vogelsong. 2007. “Site Congestion in
Recreation Choice Models: A Generated Regressors Approach to Beach Site Selection.”
Working paper, East Carolina University.
Hindsley, P., C.E. Landry, and B. Gentner. 2011. “Addressing onsite sampling in
recreation site choice models.” Journal of Environmental Economics and Management
62:95-110.
Johnston, R.J., and K. Moletner. 2014. “Meta-modeling and Benefit Transfer: The
Empirical Relevance of Source-consistency in Welfare Measures.” Environmental and
Resource Economics 59:337-361.
Krinsky, I. and A.L. Robb. 1986. “On Approximating the Statistical Properties of
Elasticities.” Review of Economics and Statistics 68:715-719.
Landry, C.E., A.R. Lewis, H. Liu, and H. Vogelsong. 2015. “Economic Value and Impact
of Visitation to Cape Hatteras National Seashore: Addressing Onsite Sampling. Working
paper, East Carolina University.
Lovell, S.J. and D.W. Carter. 2014. “The use of sampling weights in regression models of
recreation fishing-site choices.” Fishery Bulletin 112:243-252.
Magee, L.E. 2008. “Closing Motor Vehicle Beach Access in the Mid-Atlantic:
Implications for Social Welfare,” MS thesis, University of Delaware.
Mansfield, C., R. Loomis, Evans, B. and B. Munoz. 2010. Visitor Intercept Survey: Offroad Vehicle Management, Cape Hatteras National Seashore. Denver: National Park
Service, Final Report, October.
32
Mansfield, C., R. Loomis, and F. Braun. 2011. Benefit-cost Analysis of Proposed ORV
Use Regulation in Cape Hatteras National Seashore. Denver: National Park Service, Final
Report, January.
Moeltner, K., and R.S. Rosenberger. 2014. “Cross-context Benefit Transfer: a Bayesian
Search for Information Pools.” American Journal of Agricultural Economics 96:469-488.
Morey, E.R., R.D. Rowe, and M. Watson. 1993. “A Repeated Nested-logit Model of
Atlantic Salmon Fishing.” American Journal of Agricultural Economics 75:578-592.
Morey, E.R. 1999. “Two RUMs Uncloaked: Nested-logit Models of Site Choice and
Nested-logit Models of Participation and Site Choice.” In: Herriges, J.A., Kling, C.L., eds.
Valuing recreation and the environment: revealed preference methods in theory and
practice. Cheltenham, UK: Edward Elgar Publishing, Chapter 4.
National Park Service. 2010. Cape Hatteras National Seashore Off-road Vehicle
Management Plan: Final Environmental Impact Assessment. Washington DC: November.
OIA (Outdoor Industry Association). 2012. The Outdoor Recreation Economy: Technical
Report on Findings and Methods. Accessed Nov. 16, 2013. Available:
http://www.outdoorindustry.org/pdf/OIA-RecreationEconomyReport2012-TechnicalReport.pdf
Stanley, D. L. 2005. “Local Perception of Public Goods: Recent Assessments of
Willingness-to-Pay for Endangered Species.” Contemporary Economic Policy 23:165–179.
USFW (U.S. Fish and Wildlife Service). 2011. Abundance and Productivity Estimates –
2011 Update, Atlantic Coast Piping Plover Population. Hadley, MA.
Whitehead, J.C. 1993. “Economic Values for Coastal and Marine Wildlife: Specification,
Validity, and Valuation Issues.” Marine Resource Economics 8:119-132.
Williams, T. 2012. “The Battle Over a North Carolina Beach Continues.” Audubon,
September-October.
33
Figure 1: Study Area and Intercept Sites Impacted by Policy Changes
Note: Coastal counties included in participation model are highlighted in grey. Cape Hatteras National Seashore is highlighted in light green and the red
dots indicate the location of 16 MRIP intercept sites potentially impacted by the NPS management policies.
34
Figure 2: Policy Implementation on Hatteras Island: June 1, 2015
Source: Image from National Park Service website - http://www.nps.gov/caha/planyourvisit/off-roadvehicle-use.htm
35
Site 6,
Wave 1
Site 5,
Wave 2
Site 4,
Wave 3
Site 3,
Wave 5
Angler = Trip
Second Stage
Site/Wave Choice
Individual Choice
Occasion
First Stage
Participation
Non-Angler = No Trip
Figure 3: Decision tree for two-level nested logit model
36
Site 2,
Wave 4
Site 1,
Wave 3
Table 1: Definition of Variables
Description
Variables Entering Site Choice Model
Site-Specific
Travel Cost (TC)
From zip code of origin to each intercept site
Alternative Specific
Constants (ASCs)
ASCs for each site choice captures all site
characteristics that are the same across
individuals, both observed and unobserved
Variables Entering Participation Model
Mean
Min
Max
$70,777
5,650
0.70
0.49
$20,884
0.11
0.04
0.32
$345,497
130,031
1
0.84
0.30
0
0.91
Demographics
Income
Population Density
White
Male
Education
No Trip ASC
Inclusive Value
Average annual real household income
People per square mile
% of population that is white
% of population that is male
% of population completing bachelor’s degree
or higher
ASC for no trip alternative
Expected utility for the alternative choice
Note: Demographic variables are at the phone exchange level and represent a population weighted average of the zip code demographic data contained
in each exchange. Source: Demographic data obtained from U.S. Census American Community Survey.
37
Table 2: Model Estimates
Panel A. Site Choice Model
Travel Cost
2005
2006
2007
Panel B. Participation Model
Constant
Dissimilarity Coefficient
2005
2006
2007
Demographics
Average HH Income
Percent White
Percent Bachelor’s Degree
Percent Male
Population Density
Year Fixed Effects
2006
2007
Observations
Model Fit (Log pseudo-likelihood)
Parameter
T-statistic
-0.073***
-0.072***
-0.075***
Model 1
Parameter
-7.486***
0.039***
0.043***
0.039***
-10.74
-9.20
-12.15
Std. Err.
0.601
0.004
0.004
0.004
Model 2 (Preferred)
Parameter
Std. Err.
-7.491***
0.601
0.040***
0.004
-
-4.1e-06*
-1.8e-06
1.618***
7.093***
-0.0001***
2.1e-06
2.3e-06
0.338
1.111
0.000
-4.1e-06*
-1.8e-06
1.618***
7.094***
-0.0001***
-0.0003
0.0157
0.006
0.015
25,122
-6.188e+08
2.1e-06
2.3e-06
0.338
1.111
0.000
0.024***
0.009
0.006
0.012
25,122
-6.189e+08
Notes: For the participation model, Model 1 is estimated with year specific dissimilarity coefficients as a sensitivity to the
preferred model. Both models are estimated conservatively with robust standard errors clustered by phone exchange.
*** Significant at the 1 percent level. * Significant at the 10 percent level.
38
Table 3: Off-Road Vehicle Restriction Policy Scenarios: Alternative D
Fishing Site
Island
Wave 1 Wave 2 Wave 3 Wave 4 Wave 5 Wave 6
Oregon Inlet (North)
Bodie
X
A
A
A
A
X
Rodanthe Fishing Pier
Hatteras
XP
XP
O
O
XP
XP
Beach Access Ramp 20
Hatteras
XP
XP
O
O
XP
XP
Beach Access Ramp 23
Hatteras
O
O
O
O
O
O
Beach Access Ramp 27
Hatteras
XP
XP
XP
XP
XP
XP
Beach Access Ramp 30
Hatteras
A
A
A
A
A
A
Beach Access Ramp 34
Hatteras
XP
XP
XP
XP
XP
XP
Avon Fishing Pier
Hatteras
XP
XP
O
O
XP
XP
Beach Access Ramp 38
Hatteras
O
O
O
O
O
O
Buxton Beach
Hatteras
XP
XP
O
O
XP
XP
Cape Point
Hatteras
X
A
A
A
A
X
Beach Access Ramp 49
Hatteras
O
O
O
O
O
O
Frisco Pier
Hatteras
XP
XP
O
O
XP
XP
Hatteras Inlet
Hatteras
X
A
A
A
A
X
Hatteras Inlet Beach
Ocracoke Inlet & Beach
Ocracoke
Ocracoke
X
X
A
A
A
A
A
A
A
A
X
X
Notes: O = Open, no impact. X = Closed: ORV restrictions and need ORV for access. XP = ORV restrictions but
pedestrian access. A = Adaptive management with closures – could be O or X.
39
Table 4: Off-Road Vehicle Restriction Policy Scenarios: Alternative F
Fishing Site
Island
Wave 1 Wave 2 Wave 3 Wave 4 Wave 5 Wave 6
Oregon Inlet (North)
Bodie
O
A
A
A
O
O
Rodanthe Fishing Pier
Hatteras
XP
XP
O
O
XP
XP
Beach Access Ramp 20
Hatteras
XP
XP
O
O
XP
XP
Beach Access Ramp 23
Hatteras
O
O
O
O
O
O
Beach Access Ramp 27
Hatteras
XP
XP
XP
XP
XP
XP
Beach Access Ramp 30
Hatteras
O
O
O
O
O
O
Beach Access Ramp 34
Hatteras
O
A
A
A
O
O
Avon Fishing Pier
Hatteras
XP
XP
O
O
XP
XP
Beach Access Ramp 38
Hatteras
O
O
O
O
O
O
Buxton Beach
Hatteras
XP
XP
O
O
XP
XP
Cape Point
Hatteras
O
A
A
A
O
O
Beach Access Ramp 49
Hatteras
O
O
O
O
O
O
Frisco Pier
Hatteras
XP
XP
O
O
XP
XP
Hatteras Inlet
Hatteras
X
A
A
A
A
X
X
O
A
A
A
A
A
A
A
A
X
O
Hatteras Inlet Beach
Ocracoke
Ocracoke Inlet & Beach Ocracoke
Notes: O = Open, no impact. X = Closed: ORV restrictions and need ORV for access. XP = ORV restrictions but
pedestrian access. A = Adaptive management with closures – could be O or X.
40
Table 5: Definition of Upper and Lower Bounds for Policy Simulations
Policy Scenarios
Lower Bound
Upper Bound
O
Site Open to ORVs
Site Open to ORVs
X
Site Closed to ORV
& Pedestrian Access
Site Closed to ORV
& Pedestrian Access
XP
Site Open to Pedestrian
Access Only; Add 2 Hours of
Travel Time 1
Site Closed to ORV
& Inaccessible to Pedestrians
A
Site Open to ORVs
Site Closed to ORV
& Pedestrian Access
Notes: 1 – This represents additional cost of accessing a fishing site on foot instead of with an ORV.
41
Table 6: Long-Run WTP (in thousands) for Counterfactual Policy Scenarios
Alternative F
Alternative D
Upper Bound
Lower Bound
Upper Bound
Estimate
95% CI
Estimate
95% CI
Estimate
Lower Bound
95% CI
Estimate
95% CI
Panel A. Baseline
Aggregate
-$548
-$691
-$448
-$135
-$172
-$107
-$779
-$976
-$640
-$236
-$296
-$191
Year-Specific
2005
2006
2007
-$460
-$647
-$538
-$489
-$693
-$576
-$445
-$627
-$518
-$110
-$161
-$133
-$117
-$172
-$142
-$106
-$155
-$128
-$656
-$913
-$766
-$699
-$979
-$821
-$635
-$888
-$739
-$196
-$277
-$236
-$208
-$296
-$253
-$188
-$267
-$228
-$733
-$916
-$600
-$462
-$572
-$377
-$881
-$1,100
-$723
-$529
-$654
-$433
-$616
-$859
-$725
-$655
-$921
-$778
-$597
-$832
-$698
-$387
-$537
-$461
-$411
-$575
-$495
-$375
-$518
-$445
-$741
-$1,033
-$870
-$790
-$1,111
-$934
-$720
-$999
-$838
-$443
-$614
-$529
-$472
-$658
-$567
-$429
-$593
-$510
Panel B. With Congestion
Aggregate
Year-Specific
2005
2006
2007
Notes: Models are calibrated to impose the value of a trip of $30 and assumes $6.40 in congestion costs to sites that remain open under the ORV
restrictions. Simulation estimates are for WTP (in 2010 dollars) for residents of coastal counties covered by the MRIP survey. The upper bound of
welfare is estimated given the most restrictive possibilities on ORV rules. The lower bound of welfare is estimated given the most relaxed possibilities
on ORV rules. Confidence intervals are estimated using a parametric bootstrap (Krinsky and Robb 1986).
42
Table 7: Long-Run Demand Responses for Counterfactual Policy Scenarios
Alternative F
Upper Bound
Estimate
Alternative D
Upper Bound
Lower Bound
95% CI
Estimate
95% CI
Lower Bound
Estimate
95% CI
Estimate
95% CI
Panel A. Affected Trips
Baseline
38,313
31,054
46,403
-
-
-
53,354
43,555 64,153
-
-
-
Congestion
70,618
57,669
84,853
-
-
-
70,618
57,669 84,853
-
-
-
18,125
14,624
21,988
4,442
3,495
5,466
25,755
20,889 31,045
7,805
6,234
9,420
Congestion
24,260
Panel C: Substitute Trips
19,590
29,221
15,259
12,304
18,206
29,171
23,616 35,112
17,480
14,106
20,821
Panel B. Lost Trips
Baseline
Baseline
20,187
16,422
24,415
4,685
3,742
5,752
27,598
22,666 33,107
7,768
6,331
9,337
Congestion
23,271
19,339
27,814
14,099
11,743
16,761
28,695
23,839 34,321
16,330
13,594
19,394
18,735
27,945
29,186
41,259
23,821
33,627
35,285
50,060
0
12,752
10,215 15,585
37,781
36,807
31,008
29,988
45,588
44,851
Panel D. Diminished Trips
Baseline
Congestion
0
23,088
Notes: Models are calibrated to impose the value of a trip of $30 and assume $6.40 in congestion costs to sites that remain open under the ORV
restrictions. The upper bound of trips affected is estimated given the most restrictive possibilities on ORV rules. The lower bound of trips affected is
estimated given the most relaxed possibilities on ORV rules. Confidence intervals are estimated using a parametric bootstrap (Krinsky and Robb 1986).
43
Table 8: Comparison of Local and Non-Local Anglers at Three Spatial Scales
Trip Location
Coastal County Origin
Hours Fished
Catch (Binary)
Fish by Individual
Mode of Fishing
Beach
Pier
Observations
Non-Coastal County Origin
Hours Fished
Catch (Binary)
Fish by Individual
Mode of Fishing
Beach
Pier
Observations
Three-State Region
Mean
North Carolina Dare County
Mean
Mean
3.79
0.20
6.87
3.69
0.17
7.02
3.99
0.15
7.17
0.41
0.53
10,298,584
0.54
0.42
6,883,154
0.51
0.38
3,762,994
3.73
0.14
7.17
3.69
0.14
7.22
3.82
0.11
7.42
0.35
0.59
7,931,929
0.49
0.47
5,231,832
0.57
0.44
3,353,770
Source: Authors calculations from MRIP survey data.
44
Table 9: Limitations of the Welfare Analysis and Potential Implications
Limitations
Impacts
Upper Bound of Loss
Alternative F (Upper Bound) Model Estimate
Aggregate Impact
$548,169
$548,169
Congestion
Simulations adding congestion to open sites
($6.40/trip from Hindsley et al. 2007)
+ $185,179
$733,348
Non-local Recreation
46 % of user days from outside coastal counties
+ $624,704
$1,358,052
Other ORV-based
Recreation
Anecdotal evidence that upper bound of non-angler
ORV use is approximately 50%
+ $1,358,052
$2,716,104
Access & Enforcement
Costs
Approximately 30,000 permits sold with 1/3 being
annual passes ($120 ea.) and 2/3 seven day passes
($50 ea.)
+ $2,200,000
$4,916,104 1
Short-run
Mansfield et al. (2010) – current year demand
response 2.4 times higher than long-run demand
response predicted in model
+ $6,882,546
$11,798,650 2
Notes: 1 – This number represents the upper bound of costs after anglers have re-optimized in response to the ORV rules.
2
– This number represents the upper bound of potential costs in the short-run (1 to 2 years) before anglers re-optimize to the ORV rules.
45
APPENDIX
Appendix A: Site Registry Data
This appendix describes the steps used to transform the trip frequency information
contained in the MRIP site registries into aggregate trip estimates for each of the 344
shoreline fishing sites in the three-state region (NC, SC, and VA). Every two months,
NOAA updates its master list of public access shoreline fishing sites, or site registry.
Each site has weekday and weekend trip frequency or “fishing pressure” estimates
associated with it, which are also updated bimonthly. These estimates represent NOAA’s
best estimate of the number of trips occurring at a site in a normal 8-hour period, and this
information informs whether and how intensely to sample at each site in each
wave. Bimonthly updates are based on feedback from NOAA field staff as well as
auxiliary sources (e.g., published newspaper reports about pier closures).
To construct estimates of aggregate trips for each MRIP site/wave combination, we
employ the following steps. First, weekday and weekend trip estimates are constructed
for each site and wave from the contemporaneous site registry. We assume that the
average fishing day is 16 hours at manmade sites (e.g., piers) which are generally lighted
and 12 hours for all other sites. These daily estimates are then aggregated to the
bimonthly period. Finally, a regression-based adjustment is made to these estimates to
account for the fact that not all trips at a site originate from coastal counties. Data from
the MRIP intercept survey is used for this task. Specifically, intercepted respondents
report their home zip code, which allows us to determine if they live in a costal or
46
noncoastal county. For every sampled site, the share of trips originating from coastal
counties can be constructed, and because sampling is by design simple random sampling
at the site level, this constructed share is an unbiased estimate of the population share at
that site. A weighted linear regression is then used to predict the share of coastal trips as
a function of observable site characteristics. The weights employed in the regression
analysis are inversely proportional to the intensity of sampling at the sites. These
predicted shares are then combined with the bimonthly trip estimates to generate total trip
predictions from coastal counties for all 344 MRIP sites.
47
Appendix B: Uncalibrated Model Results
Table B.1: Long-Run WTP (in thousands) for Counterfactual Policy Scenarios (Uncalibrated)
Alternative F
Alternative D
Upper Bound
Lower Bound
Upper Bound
Panel A. Baseline
Lower Bound
Estimate
95% CI
Estimate
95% CI
Estimate
95% CI
Estimate
95% CI
Aggregate
-$569
-$718 -$464
-$136
-$173 -$108
-$822
-$1,032 -$672
-$240 -$300
-$194
Year-Specific
2005
2006
2007
-$475
-$673
-$560
-$509 -$461
-$724 -$653
-$602 -$542
-$110
-$162
-$134
-$118 -$106
-$175 -$157
-$144 -$130
-$688
-$966
-$810
-$737
-$666
-$1,040 -$938
-$872
-$786
-$198 -$212
-$281 -$302
-$240 -$258
-$190
-$271
-$232
-$772
-$966 -$625
-$476
-$590 -$387
-$938
-$1,174 -$761
-$547 -$678
-$445
-$644
-$906
-$765
-$686 -$621
-$970 -$879
-$824 -$733
-$397
-$554
-$476
-$423 -$382
-$592 -$535
-$513 -$457
-$783
-$1,102
-$929
-$834
-$755
-$1,180 -$1,068
-$1,001 -$890
-$457 -$486
-$636 -$681
-$549 -$591
-$439
-$615
-$526
Panel B. With Congestion
Aggregate
Year-Specific
2005
2006
2007
Notes: Models allow uncalibrated value of a trip ($342) and assumes $6.40 in congestion costs to sites that remain open under the ORV restrictions.
Simulation estimates are for WTP (in 2010 dollars) for residents of coastal counties covered by the MRIP survey. The upper bound of welfare is
estimated given the most restrictive possibilities on ORV rules. The lower bound of welfare is estimated given the most relaxed possibilities on ORV
rules. Confidence intervals are estimated using a parametric bootstrap (Krinsky and Robb 1986).
48
Table B.2: Long-Run Demand Responses for Counterfactual Policy Scenarios (Uncalibrated)
Alternative F
Upper Bound
Estimate
Alternative D
Upper Bound
Lower Bound
95% CI
Estimate
95% CI
Estimate
Lower Bound
95% CI
Estimate
95% CI
Panel A. Affected Trips
Baseline
Congestion
38,313
70,618
31,054 46,403
57,669 84,853
-
1,079
1,462
387
1,359
-
-
53,354
70,618
43,555 64,153
57,669 84,853
-
-
-
685
1,563
450
1,036
930
2,115
Panel B. Lost Trips
Baseline
1,627
Congestion 2,205
Panel C: Substitute
Trips
Baseline
36,687
Congestion 42,789
2,240
3,018
251
900
537
1,841
2,347
2,680
1,562
1,778
3,217
3,679
29,682 44,402
35,342 51,288
8,553 6,588
25,370 21,070
10,116
30,119
51,007
53,597
41,615 61,637
44,030 65,112
13,765
29,513
11,206
24,485
16,523
34,990
20,510 31,175
29,673 24,157
43,889 35,492
35,904
53,360
0
14,541
11,438 17,894
38,904
39,542
31,809
31,911
46,985
48,294
Panel D. Diminished Trips
Baseline
0
Congestion 25,623
Notes: Models allow uncalibrated value of a trip ($342) and assumes $6.40 in congestion costs to sites that remain open under the ORV restrictions.
Simulation estimates are for WTP (in 2010 dollars) for residents of coastal counties covered by the MRIP survey. The upper bound of welfare is
estimated given the most restrictive possibilities on ORV rules. The lower bound of welfare is estimated given the most relaxed possibilities on ORV
rules. Confidence intervals are estimated using a parametric bootstrap (Krinsky and Robb 1986).
49