The Welfare Effects of Restricting Off-Highway Vehicle Access to Public Lands

Paul M. Jakus, John E. Keith, Lu Liu, and Dale Blahna

Off-highway vehicle (ORV) use is a rapidly growing outdoor activity that results in a host ofenvironmental and management problems. Federal agencies have been directed to developtravel management plans to improve recreation experiences, reduce social conflicts, anddiminish environmental impacts of OHVs. We examine the effect of land access restrictionson the welfare of OHV enthusiasts in Utah using Murdock's unobserved heterogeneity randomutility model (Murdock 2006). Our models indicate that changing access to public lands fromfully "open" to "limited" results in relatively small welfare losses, but that prohibiting accessresults in much larger welfare losses.

Key Words: off-highway vehicles, recreational access, unobserved heterogeneity, randomutility model

The use of off-highway vehicles (OHVs) is one ofthe most rapidly growing outdoor activities in theUnited States and in Utah.1 Nationally, participa-tion by residents aged 16 years and older hasgrown from 17 percent in 1999 to just under 20percent in 2007 (Cordell et al. 2008). This meansthat some 44 million people engaged in OHVrecreation in 2007. The Mountain West states as agroup (Arizona, Colorado, Idaho, Montana,Nevada, New Mexico, Utah, and Wyoming) havean OHV participation rate of 28 percent, wellabove the national average. Indeed, Wyoming,Idaho, and Utah all rank in the top five statesbased on OHV participation. This popularity isreflected in the sharp growth of OHV registra-tions, with vehicle registrations in the state of

Paul M. Jakus is Professor, John E. Keith is Professor Emeritus, andLu Liu is a graduate research assistant in the Department of AppliedEconomics at Utah State University in Logan, Utah; Dale Blahna is aResearch Social Scientist with the Pacific Wildland Fire Sciences Lab-oratory of the USDA Forest Service's Pacific Northwest Research Sta-tion in Seattle, Washington.

The authors thank John Harja and Tiffany Pizzulo of the Utah Gov-ernor's Public Lands Policy Coordination Office for their support andadvice concerning this project. The authors also thank Richard Kran-nich and Doug Reiter, Utah State University, for their help in con-structing the survey and managing data collection. The authorsacknowledge the support of the Utah Agricultural Experiment StationProject UTA00052 and USDA Regional Project W2133 for this study.All errors remain with the authors.

Utah growing 233 percent during the 1998-2006period (Burr et al. 2008).

Concomitant with the growth in OHV partici-pation has been a host of management problemsfor stewards of public land. In 2003, then U.S.Forest Service Chief Dale Bosworth declared that"unmanaged recreation," of which OHV use is animportant component, was one of the top fourthreats facing national forests (Bosworth 2003).Chavez and Knap (2004) outline the reasons whyunmanaged recreation by OHV users wasdeclared an ecological threat: OHVs can result inunplanned roads, soil erosion, degradation ofwater quality, destruction of habitat, the spread ofinvasive species, and conflict with nonmotorizedusers, among other problems. Opponents of OHVuse also emphasize safety problems and userconflicts between motorized and nonmotorizedvisitors (Havlick 2002). Proponents argue for thesocial and psychological benefits of the activity,and the economic benefits for recreationists andlocal communities (Blahna 2007).

1 Off-highway vehicles are defined as four-wheel drive vehicles, mo-torcycles, all-terrain vehicles (ATVs), and other specially designedvehicles such as dune buggies and sandrails. We do not includesnowmobiles in this definition, nor' in any of the statistics reportedin the study.

In response to the increasing demand and thepotential problems posed by unmanaged recre-ation, federal agencies have been directed todevelop travel management plans as part of anagency's planning process (Stern et al. 2009). Thepurpose is to improve recreation experiences andreduce social conflicts and environmental impactsby designating specific roads, trails, and areas thatare open to motorized uses. In general, land man-agement agencies such as the U.S. Forest Service(USFS) and the Bureau of Land Management(BLM) are moving away from current policiesthat generally allow OHV access to both road-ways and cross-country travel, with only a fewareas where access is prohibited, to proposedpolicies that have tighter restrictions on cross-country travel and that designate larger amountsof land where OHV access is strictly prohibited(see, for example, USDI BLM 2001, USDAForest Service 2005). Many OHV users objectstrenously to more restricted access to publiclands, and frequently local officials suggest thatreduced OHV use will result in significant reduc-tions in local economic activity. In contrast, manyenvironmental groups indicate that these restric-tions are insufficient to prevent damage.

Relatively few studies have estimated an eco-nomic value for OHV recreation, and none havebeen estimated for Utah. Further, in 2008, sixBLM Field Offices in Utah completed andreleased for public comment new Resource Man-agement Plans (RMPs) under which OHV accessto BLM lands was affected. In general, theamount of BLM land classified as "open" to trailand cross-country use was decreased, while theamount of BLM land on which OHV use was lim-ited or prohibited was increased. This studyestimates the welfare implications of proposedaccess restrictions to federally owned land inUtah.2 Using a sample of OHV owners in Utahwho have registered their vehicles, a travel costmodel is developed to link access to public landsto locations where OHV owners choose to recre-ate. Our statistical models are constructed at thecounty level, but defining sites in this mannerraises concerns about unobserved heterogeneity

across sites, so we use Murdock's recent variantof the random utility model to adjust for thisproblem (Murdock 2006). Our models indicatethat changing access to public lands from fully"open" to "limited" or "closed" has negativewelfare impacts on OHV enthusiasts.

Past Studies

While a few authors have estimated the economicvalue of off-highway recreation (see Bowker,Miles, and Randall 1997, Loomis 2006, Englin,Holmes, and Niell 2006, and Silberman andAndereck 2006), only one other study has tackledthe issue of restricted access to public lands(Deisenroth, Loomis, and Bond 2009). Given therapid growth of OHV recreation, it is perhapssurprising to find so few studies measuring theeconomic value of OHV use. Bowker, Miles, andRandall (1997) examined OHV use in Florida;their negative binomial travel cost models foundper-trip economic values between $13 and $66.Silberman and Andereck (2006) used an open-ended contingent valuation question to estimatethe economic value of OHV recreation inArizona. They found enthusiasts' willingness topay (WTP) to be about $54 to $96 per trip,depending upon the type of off-highway vehicleused. Loomis (2006) used the travel cost methodto value recreation in a remote region of north-western Colorado, finding a consumer surplusestimate of $29 per day trip. Englin, Holmes, andNiell (2006) use a demand systems approach tomeasure the economic value of OHV recreation atfour sites in North Carolina. The authors utilizedtwo different assumptions regarding the statisticaldistribution of the model and two assumptionsregarding the appropriate restrictions implied byeconomic theory, for a total of four differentdemand models. The authors argue that the datasuggest the negative binomial distributionalmodel is appropriate; the estimated compensatingvariation for the least restrictive model rangedfrom $27 per trip for the least valued site to$131 per trip for the highest valued site. The morerestrictive model yielded per-trip welfare esti-mates ranging from $587 to $996, depending onthe site visited.

To our knowledge, Deisenroth, Loomis, andBond (2009) provide the only study that tries to

address the problem of access restrictions. Usingan intercept survey of OHV users at three sites inLarimer County, Colorado, the authors use a con-tingent valuation approach to estimate a per daytrip WTP of $78. The payment vehicle for theCVM question was an increase in travel costs forthat day's trip. The authors use this estimate tocalculate a consumer surplus estimate for each ofthe three OHV trails studied, stating that thisallows them to make "educated policy decisions"regarding trail closures. However, the approachused does not directly measure the impacts of trailclosures. The authors' analyses imply that travelcosts are high enough to prevent visits to any site,but do not allow substitution by OHV users tocloser, less expensive sites. This implies that wel-fare losses of trail closures are over-estimated.

Methods

Sampling

A randomly selected group of OHV owners in thestate of Utah was sampled from a list of registeredowners maintained by the state (Burr et al. 2008).Some 181,500 vehicles were registered during thespring of 2007. Eliminating duplicate names(many people own more than one vehicle) yieldeda population of about 113,700 OHV owners, fromwhich some 1,500 names and addresses weredrawn for participation in the mail survey. Thesurvey materials (cover letter, survey, and statemap) were designed in consultation with repre-sentatives of the Utah Governor's Public LandsPolicy Coordination Office and Utah StateUniversity's Institute for Outdoor Recreationand Tourism, which had conducted a numberof OHV surveys in the past. Burr et al. (2008)provides details of the survey on which this studyis based, whereas Fisher, Blahna, and Bahr (2001)provide details of the surveys conducted in 2000and 1994.

The initial mailing date was June 2007, with allsurvey activities completed by August 2007.Eighty-four surveys were classified as undeliver-able. Following Dillman (2000), five attemptswere made to elicit a response: the initial mailingof the survey, a reminder postcard, a second fullsurvey mailing, a second reminder postcardand, finally, a third mailed survey. Of the 1,416"deliverable" addresses, responses were received

from 600 for a response rate of42.4 percent.3 Therepresentativeness of the sample can be evaluatedby comparing our statistics to those yielded by theNational Survey of Recreation and Enviromnent(NSRE) for the state of Utah. The two samples arenot entirely comparable in that the NSRE is a gen-eral population survey of overall participation invarious recreation pursuits (including OHV recre-ation), whereas our survey is limited to thoseresidents who own and register OHV vehicles.The NSRE sample includes, for example, friendsand family members who participate in OHVrecreation but do not own a vehicle, and thosewho rent OHVs. The reasons for not owning avehicle are many (e.g., a lack of sustained inter-est in OHV recreation, income constraints, etc.)so we did not expect an exact correspondenceacross the two samples.

Table 1 shows that the sample of Utah residentswho own and register their OHVs is older than thegeneral population engaging in OHV recreation.Further, our sample is less ethnically diverse andhas greater household income. Educational ques-tions were not asked in an identical manner acrossthe two surveys: we combined our "some college"and "technical degree" categories into the "somecollege" category of the NSRE. Our sample hasfewer respondents in both the lowest and highesteducational categories. In general, this demo-graphic pattern across the two surveys appears tocorrespond well with our expectations: OHVowners were expected to be older and have largerincomes than the general population of OHVrecreationists.

Primary survey data were supplemented withdata from secondary sources. Our statisticalmodel, based on the cost of travel from one'shome to a destination, required that we identifyboth the origin of the trip (where the rider lefthome) and the destination of the trip (exactlywhere he or she recreated). Origins were identi-fied using the respondent's home zip code.Destinations were elicited by first asking thename ofthe destination (trailhead) for the respon-dent's most recent trip and then asking for thecounty in which this destination is located. This

distinct destinations. Unfortunately, identifying235 choice destinations for a recreation activitysuch as OHV riding presents empirical difficul-ties. The site location is merely a trailhead fromwhich riders depart for recreation, yet the analystmust somehow define site attributes (e.g., milesof trail or acres of public land) for the areaaccessed from a trailhead. How large should thearea around the trailhead be? Is it the mean dailydistance traveled on the OHV (say, a 60-mileradius around the trailhead, or an 11,300-square-mile area) or the median distance (a radius of 40miles, or 5,000 square miles)? Why would a circledefine the area around the trailhead? Would notthe appropriate area and shape of a region differdepending upon the terrain at the site? Analystshave found no clear answers to questions such asthese (see Karou, Smith, and Liu 1995, Lupi andFeather 1998).

In addition to these empirical difficulties, issuesassociated with state and federal land administra-

information was used to develop a list of recre-ation sites that correspond to a set of latitude andlongitude coordinates. Respondents were alsoasked to estimate the number of trips made toeach county in Utah over the previous 12-monthperiod.

Most sites were relatively easy to locate; sitenames provided by respondents were input intothe search procedure in the All Topo CD-ROMmap set for Utah.4 After locating the site, coordi-nates for latitude and longitude were recorded.Other sites, such as the "West Desert," refer to anexpansive geographical region, and there waslittle we could do to identify the trailhead visited.Our final data set consisted of 235 identifiably

tion also affect our site definition decision. Thejurisdiction of one land agency may overlap withthose of many other jurisdictions. Our primaryinterest is the proposed land management policiesof six U.S. BLM Field Offices (FOs). In somecases the overlap of jurisdictions is relativelyclean-the Price BLM FO has responsibility forall of Carbon and Emery counties - but in othercases the overlap is quite messy, as is the casewith the Richfield BLM FO, which has responsi-bility for all BLM land in four counties and aportion of BLM land in two other counties. Fur-ther, none of the BLM FOs coincides withcounties grouped in any of the seven Associationsof Governments (AOGs, the standard planningunit used in Utah) or the nine official travelregions of the state's Office of Tourism. The stateof Utah sponsored our study, and results were tobe aggregated to county combinations specifiedby the sponsor, including aggregation to the AOGlevel. In view of these empirical issues, individ-ual sites were aggregated to the county level(again, see Karou, Smith, and Liu 1995).

A county-level "aggregate site" was createdusing a weighted average of the latitude and lon-gitude coordinates for all destinations within thatcounty. Weights were defined by the number ofpeople visiting each site, such that the most heav-ily visited site in a county received the greatestweight, while the least visited site received thesmallest weight. As a final step in the process,travel distances were measured from the center ofeach origin zip code to the geographic coordinatesfor each of the twenty-nine aggregate county-level sites using the USDA computer programZIPFIP. The cost of travel to these sites was cal-culated by multiplying by a constant per mile costof vehicle operation, 20.1 cents per mile, the AAAestimated variable cost of operating a sport utilityvehicle in the year 2006.

The most important attribute for our purposesis the amount of area available for OHV activities.The State of Utah Automated Geographic Refer-ence Center provides relevant GeographicInformation System (GIS) data for the entire state.These GIS databases were used to constructmeasures of current land use by county. The keyGIS categories for the purposes of this study arethe total amount of land in a county, the totalamount of public land in a county, and the amount

of public land on which OHV use is limited orprohibited. While the first two categories wereeasily determined from GIS data, we did not haveaccess to an exact measure of public lands fromwhich OHVs are currently prohibited. Public

lands used for military purposes, designatedwilderness areas, and wilderness study areas haveOHV access prohibited, and we were able to cal-culate the amount of land within these categories.Similarly, six BLM FOs have published draftresource management plans (RMPs) that reportacreage and provide maps on which OHV use iscurrently prohibited or limited. Table 2 showsacreage for the Kanab, Moab, Monticello, Price,Richfield, and Vernal FOs.5

Each BLM RMP reported land managementacreage for the entire Field Office. The RMPs alsoincluded maps of land on which OHV use is per-mitted, limited, or prohibited under current andproposed management alternatives. Because allsix BLM Field Offices encompass more than onecounty, it was necessary to convert this informa-tion to correspond to our designation of recreationsites (counties). The published maps were digi-tized to allow calculation of acreage in the open,limited, and closed categories under the currentand preferred management alternatives. Ourmeasure of "closed" public lands in a givencounty includes not only closed BLM land, butalso land administered by the military, designatedwilderness, and wilderness study areas, all ofwhich are off-limits to OHVs. Unfortunately, wedid not have access to the amount of acreageclosed to OHV use by other state and federalmanagement agencies (e.g., the USFS or theNational Park Service), or for BLM Field Officesthat have not yet developed an RMP. This meanswe are undercounting the amount of public landon which access is prohibited or limited undercurrent management conditions." Other attributessuch as the presence of a well-known sand dunedestination (Kane and Juab counties) or red

rock country (the broad swath of counties acrosssouthern and eastern Utah) may also be important.These influences were captured using dummyvariables for the presence of major dune or redrock destinations within a county.

A key attribute that is unavailable for thisanalysis is the miles of trail that can be accessedby OHV users. The numbers of miles of "A50"road (four-wheel drive) in each county can beobtained from census data, but these data do notcorrespond very closely with trail data publishedin the draft BLM RMPs. Further, the miles of trailare likely to be correlated with the amount ofacreage in a county, causing collinearity problemsin the statistical model. We have chosen to treatthis as an important unobserved site attribute,a decision that drives our choice of statisticalmodel.

A Random Utility Model with UnobservedHeterogeneity Across Sites

Our basic approach is to use the random utilitymodel to measure OHV visitation patterns withinUtah. Our hypothesis is that if a federal agencyrestricts OHV access to public lands, OHV users

are more likely to choose alternative sites forrecreation, i.e., to recreate in other counties. Therandom utility model (RUM) version of the travelcost model allows the analyst to estimate theimpact of changing access policies on use pat-terns. The RUM is a probabilistic modelingapproach, in which the demand for a given recre-ation site is measured through the probability thatthe site will be visited (Morey 1999). Sites withmore desirable characteristics (for OHVs thesecharacteristics could be low travel cost, abundantpublic lands, and the opportunity to travel throughspectacular red rock country) will be chosenwith greater frequency relative to sites with lessdesirable characteristics. The theoretical basis forthe model is that the recreationist will comparethe utility (satisfaction) associated with one site j,Uj, to the utility of visiting an alternative site k,Uk. The recreationist will choose the site thatyields the most satisfaction, choosing site j if

Put simply, a person will choose to go where heor she derives the most satisfaction, relative to allavailable choices.

The satisfaction derived from any site j is afunction of the cost to gain access to the site (the"travel cost") as well as other attributes of the site.For any site j, let TCj be the travel cost to the site,Oj be the measure of public land open for ATVrecreation at the site, Lj be the measure of publicland with limited ATV access at the site, and Cjbe the measure of public land closed to ATVrecreation at the site. Dj and Rj, are zero-onedummy variables indicating the presence of sanddunes and red rock at site j, respectively. Further,multiple trailheads were combined into a singleaggregate destination so the analyst must includea variable, Sj , measuring the number of trailheadswithin the aggregate. Whereas all factors influ-encing site choice are known by the recreationist,some may remain unknown to the analyst, thusintroducing random error, Ej, into the choiceproblem. Again, the recreationist will choose tovisit the site yielding the greatest utility, choosingto visit site j rather than site k if,

If the errors are assumed to be additive andindependently and identically distributed accord-ing to a type I extreme value distribution, theprobability that a person will choose site j over allother K-1 alternative sites is given by,

The model is made operational by specifyingthe form of the U (.) function; for example, acommon specification is linear,

where a is an intercept term, B is the travel costparameter, the y vector consists of parameters forthe vector of site attributes Aj, and the parameteron the site aggregation term is fixed equal to one.7

7 The site attribute vector A includes the variables OJ. Lj. Cj. Dj, andR]. Setting the parameter of the site aggregation variable follows thestandard approach in dealing with aggregated sites (Lupi and Feather1998).

The parameters can be estimated via the methodof maximum likelihood using equation (1) asthe basis for the likelihood function. Economictheory indicates that we should observe a negativesign for B and positive signs for elements of thevector y if the site attributes are desirable; nega-tive if the attributes are undesirable.

Other factors may also influence the site choiceof an individual recreationist, but the standardRUM model cannot capture all possible attributesof a site; Murdock (2006) demonstrates thatwelfare measures are biased downward (upward)for sites with desirable (undesirable) yet unob-served attributes. For example, we do not have agood measure for miles of off-road trails, yet weknow this to be an important determinant of sitechoice. Failure to account for the unobservedattributes leads to biased standard errors that tendto overstate the precision of the estimates ofa standard RUM model. Murdock proposes asimple two-stage estimation procedure that miti-gates these problems. At the first stage, onesimply estimates a standard RUM model usingsite-specific constants, attributes that vary acrossindividuals and sites, such as the travel cost vari-able, and our site aggregation adjustment,

The basic intuition is that the site specific con-stants, aj, capture all observed and unobservedattributes that vary across sites. Thus, measure-ment error is removed and an unbiased parameterfor travel cost is estimated with precision. Siteattributes are not included in the first stagebecause they cannot be identified separately fromthe vector of constants. At the second stage, oneruns a simple OLS regression of the constants onsite attributes, yielding the taste parameters for theobserved characteristics of the sites. Using theattributes in our example, the OLS regressionappears as,

where n is a constant and cj represents unobservedattributes. In a very real sense, one can think ofthe Murdock approach as being akin to a varyingparameters approach.

Results

The travel cost model presented below focuseson the trip-making behavior of 534 owners ofOHVs who supplied complete data on trips toUtah destinations.8 Of this group, the averageOHV owner took 9.3 trips in Utah with their OHVduring the 12 months preceding the survey. Tripswere dispersed across the state, with the mostpopular sites for the most recent trips being inUtah, Juab, Tooele, and Washington counties,with these four counties accounting for nearly 25percent of in-state trips during the l2-monthperiod. Some 76 percent of OHV owners reportedspending most of their riding time traveling alongestablished roads and trails, although a sizableminority (24 percent) reported spending more rid-ing time off established trails.

Travel Cost Modeling

Our random utility models for OHV trips totwenty-nine Utah counties appear in Table 3.Columns (2) and (3) present coefficients andt-statistics for the standard RUM, whereascolumns (4) and (5) show the first stage RUM ofthe Murdock approach. In the standard model, allthe parameters follow expectations. The TravelCost parameter is negative, implying that, all elsebeing equal, closer sites are preferred to siteslocated farther away. The greater the Proportionof Open Acreage in a county, the more likely ATVenthusiasts are to choose the site. Similarly, thegreater the Proportion of Limited Acreage, themore likely that the site will be chosen. Oneshould note the magnitude of these last twoparameters: open acreage is preferred to limitedacreage. In contrast to open and limited acreageat a site, as the Proportion of Closed Acreagegrows, the site is less likely to be chosen. Thepresence of Red Rock at a site also increases thelikelihood that the site will be visited, whereas theDunes measure was insignificant.

Turning now to the two-stage model thatcontrols for unobserved heterogeneity acrosssites, the first stage is reported in columns (4) and

(5) of Table 3. Here we see that the Travel Costparameter is negative, as expected. The site-specific constants follow a pattern consistent withexpectations. The site-specific constants for Utah,Juab, and Tooele counties are relatively small, butthese sites are located either in or immediatelyadjacent to the four most populous counties of theWasatch Front (home to 59.5 percent of Utah'sOHV owners) and have relatively small travelcosts. The largest site-specific constants are forcounties that are home to Utah's spectacular redrock country (Grand, Washington, San Juan) or ina county that provides the primary trailheads of ahugely popular complex of high-elevation trailslocated in the central portion of the state (Piute).

Our second-stage model uses the site-specificconstants appearing in Table 3 (along with theimplied "zero" for the county omitted at the RUMstage, Weber county) on the left-hand side of anOLS regression against observable site attributes.The second-stage models (Table 4) perform quitewell.9 Again, our key variables concern theproportion of acreage in a county that is open forcross-country travel, the proportion limited to trailtravel only, and the proportion closed to OHVusers. In Specification # 1, the coefficient on theProportion of Open Acreage is positive andstatistically significant, implying that as theamount of open acreage in a county increases, thesite is more likely to be selected as a place to visit.The coefficient on the Proportion of LimitedAcreage is also positive and statisticallysignificant. This is consistent with data indicatingthat about 75 percent of riders in our samplepreferred to recreate on trails as opposed to cross-country travel.10 The coefficient on Proportion ofClosed Acreage is negative and statisticallysignificant, implying that as the amount of"closed" acreage in a county increases, the site isless likely to be selected as a place to visit.Finally, the effect of Dunes in a county is not asignificant determinant of site choices, whereasthe presence of Red Rock country is a positive andsignificant factor in determining where OHVrecreationists choose to visit.

consumer surpus of OHV recreation. The estimatecoming from the standard model is a bit higherthan the estimates generated using contingentvaluation, whereas the estimate from the two-stage model is a bit lower. The estimate ofSilberman and Andereck (2006) for their pooledsample of all types of off-road vehicles (similar toour pooling of all vehicle types) was $68, whereasthe estimate in Deisenroth, Loomis, and Bond(2009) was $78 per day trip (also pooled acrossall vehicle types). For those estimates derivedfrom the travel cost methodology, both estimatesare greater than that estimated by Loomis (2006)for a remote site in Colorado ($29), or for three ofthe four North Carolina sites in Englin, Holmes,and Niell (2006).

The welfare effects of restricting access topublic land can be calculated by using thepublished BLM RMP maps to measure how muchland is moving from one access category toanother. Examining the magnitude of thecoefficients for both the standard and the two-stage models allows one to see that moving landfrom "open" to "limited" will cause a relativelysmall loss in consumer welfare, but moving from"open" or "limited" to "closed" will cause a largeloss in welfare. Conversations with BLM officialsindicated that they were well aware of this, andthe six BLM RMPs tended to move public landclassified as "open" in recent years more to the

Our second specification (Table 4) controls forthe size of the county by including the naturallogarithm of the county size as an explanatoryvariable. Nearly all the variables retain the samesign, magnitude, and level of significance as inSpecification #1, with the exception being theProportion of Closed Acreage, which showsreduced magnitude and statisical significance.Overall, this specification was not as strong as thefirst.

Welfare Impacts of Access Restrictions

The model presented in Tables 3 and 4 can be usedto estimate baseline measures of per-tripcompensating variation and changes in welfaredue to changes in land access. Evaluated at themean travel cost and the current level of siteattributes in each of the twenty-nine destinationcounties, the per-trip compensating variationusing the standard model is $80.41, whereas theestimate arising from the two-stage model is$52.12.11 Both estimates are at a midpointbetween previous published estimates of the

"limited" category rather than the "closed"category. 12 The proposed BLM RMPs areconcentrated in eastern and central Utah, such thatthe effect of the new restricted access will be toshift some visitation from eastern and centralUtah to western and northern Utah. The netchange in compensating variation due to theaccess restrictions from the standard model is88 cents per trip, or about $8.20 per season if thenumber of trips taken by the average OHV ownerstays constant under the new restrictions. UnderMurdock's two-stage approach, the estimates area loss of $1.14 per trip if the restrictions are putinto effect, for a seasonal loss of $10.60 if thenumber of trips remains constant. Using theresults from the two-stage approach, andassuming our sample is representative of OHVowners for the state as a whole, welfare losseswill total $1.21 million. Should recreationistsdecide to stay home or recreate at sites outside ofthe state, surplus losses will be larger.

The welfare losses estimated by the two modelsindicate that some degree of unobserved heter-ogeneity was present in the data. The BLMrestrictions were primarily located in the southand east portions of the state, home to the inter-nationally known OHV destination of Moab, aswell as to millions of acres of spectacular red rockcountry. The fact that greater welfare losses weremeasured with the unobserved heterogeneitymodel than with the standard RUM supports ourhypothesis that counties experiencing increasedaccess restrictions had positive attributes thatwere not fully captured using the information andmeasurements available to us.

Conclusions

As noted in the introduction, OHV managementis a complex issue with social, ecological, andeconomic aspects. While there are many potentialmanagement actions to address the issue-repairing and maintaining existing roads, buildingmore roads and trails in ecologically resistantareas, adding mass transit options in national andstate parks, and zoning conflicting uses-mostagencies focus first on OHV road closures and

OHV use restrictions (Havlick 2002). There is adebate in the literature about the social andecological value of this approach (Blahna 2007),but there is only one study to date that evaluatesthe economic tradeoffs involved with userestriction policies (Deisenroth, Loomis, andBond 2009). Our study adds to this relativelysmall but critical body of literature.

We have estimated a version of the randomutility model that accounts for unobservedheterogeneity across sites, a situation that is likelyto hold in our modeling. Our statistical models arerelatively robust, indicating that open acreage ismost highly valued by OHV users. The modelindicates that limiting the use of motorizedvehicles to trails only has a relatively smallimpact on consumer welfare, but the completeloss of access to public land has the potential tocause relatively large welfare losses. Our findingshave implications for management of OHVs: therelatively small welfare losses associated withrestricting OHV travel to existing trails androadways suggest that agencies can assure accessfor OHV enthusiasts while simultaneouslysatisfying mandates for resource protection.

A wide variety of federal and state agencieshave responsibility for the management of publicland. In recent years, such agencies have beenmoving toward land management methods thatacknowledge the complex interactions within anecosystem (Grumbine 1994). Such ecosystemmanagement approaches require that agenciesintegrate ecological, social, and economic factorsin management decisions. While agencies cancollect information regarding ecological impacts,visitor preference, and social conflicts relativelyeasily, there is often a dearth of informationregarding economic values when making non-commodity management decisions. Thus,acheiving management goals is difficult becauseeconomic factors are rarely given as much weightas biological or social factors in decisions relatedto recreation access and use, aesthetics, and othernonmarket values of public lands. This studyshows the relative economic implications of userestrictions for the state of Utah and provides anapproach that could be adapted by other stateand federal agencies as they develop travelmanagement plans.

Blahna, DJ. 2007. "Introduction: Recreation Management."In L.E. Kruger, R. Mazza, K. Lawrence, eds., Proceed-ings: National Recreation Workshop on RecreationResearch and Management. Gen. Tech. Rep. PNW-GTR-698. Portland, OR: U.S.D.A. Forest Service, PacificNorthwest Research Station.

Bosworth D. 2003. "We Need a New National Debate."Speech to the Izaak Walton Leagne, 81st Annual Conven-tion, Pierre, South Dakota, July 17. Available at http://www.fs.fed.us/news/2 003/ speeches/07 /bosworth. shtml(accessed January 2010).

Bowker, J.M., M.P.Miles, and EJ. Randall. 1997. "A DemandAnalysis of Off-Road Motorized Recreation." In Expand-ing Marketing Horizons into the 21st Century, the Pro-ceedings of Association of Marketing Theory and Practice,pp. 387-391.

Burr,S.W., J.w. Smith, D. Reiter, P.M. Jakus, and J. Keith.2008. "Recreational Off-Highway Vehicle Use on PublicLands Within Utah." Institute for Outdoor Recreation andTourism, Utah State University, April. Available at http://www.governor.utah.gov/publiclandsIPLPCOStudies/OHVFinal Report 8-1-08.pdf (accessed September 13, 2008).

Chavez, D., and N. Knap. 2004. "Management Problems ofand Strategies for Off-Highway Vehicle Management:National Forests in California." Available at http://www.fs.fed. us/psw/topicsirecreation/studies/values _ohv_mgt.shtml (accessed January 8, 2010).

Cordell, H.K., C.J. Betz, G.T. Green, and B. Stephens. 2008."Off-Highway Vehicle Recreation in the United States andits Regions and States: An Update National Report fromthe National Survey on Recreation and the Environment(NSRE)." Available at http://warnell.forestry.uga.edu/nrrt/ NSRE/IRJSReclIrisRec Irpt.pdf (accessed January 8,2010).

Deisenroth, D.D., J.B. Loomis, and C.A. Bond. 2009. "Non-Market Valuation of Off-Highway Vehicle Recreationin Larimer County, Colorado: Implications of TrailClosures." Journal of Environmental Management,90(11): 3490-3497.

Dillman, D. 2000. Mail and Internet Surveys: The TailoredDesign Method (2nd ed.). New York: John Wiley & Sons,Inc.

Englin, J., T. Holmes, and R. Niell. 2006. "AlternativeModels of Recreational Off-Highway Vehicle SiteDemand." Environmental and Resource Economics 35(4):327-338.

Fisher, A.L., D.J. Blahna, and R. Bahr. 2001. "Off-HighwayVehicle Uses and Owner Preferences (Revised)." UtahState University Institute for Outdoor Recreation andTourism Professional Report IORT PR2001-02.

Greene, W.H. 2008. Econometric Analysis (6th ed). UpperSaddle River, New Jersey: Prentice-Hall.

Grumbine, R.E. 1994. "What is Ecosystem Management?"Conservation Biology 8(1): 27-38.

Havlick, D.G. 2002. No Place Distant: Roads and MotorizedRecreation on America s Public Lands. Washington, DC:Island Press.

Karou, Y, V.K. Smith, and J.L. Liu. 1995. "Using RandomUtility Models to Estimate the Recreational Value ofEstuarine Resources." American Journal of AgriculturalEconomics 77(1): 141-151.

Loomis, J. 2006. "Estimating Recreation Use, Expenditures,and Economic Benefits at Little Snake River ResourceArea Using Visitor Data and Travel Cost Method."Working paper, Department of Agricultural and ResourceEconomics, Colorado State University, Ft. Collins. A copyof this paper is on file with the authors.

Lupi, F., and P.M. Feather. 1998. "Using Partial Site Aggrega-tion to Reduce Bias in Random Utility Travel CostModels." Water Resources Research 34(12): 3595-3603.

Morey. E.R. 1999. "Two Rums Uncloaked: Nested-LogitModels of Site Choice and Nested-Logit Models ofParticipation and Site Choice." In J.A. Herriges and C.L.Kling, eds., Valuing Recreation and the Environment.Northampton, MA: Edward Elgar.

Murdock, J. 2006. "Handling Unobserved Site Characteristicsin Random Utility Models of Recreation Demand."Journal of Environmental Economics and Management51(1): 1-25.

Silberman, J., and K.L. Andereck. 2006. "The Economic Valueof Off-highway Vehicle Recreation." Journal of LeisureResearch 38(2): 208-223.

Stern, M.J., D.J. Blahna, L.K. Cerveny, and M.J. Mortimer.2009. "Visions of Success and Achievement in Recreation-Related USDA Forest Service NEPA Processes."Environmental Impact Assessment Review 29(4): 220-228.

U.S. Department of the Interior Bureau of Land Management.2001. "National Management Strategy for MotorizedOff-Highway Vehicle Use on Public Lands." Availableat http://www.blm.gov/ohv/OHV_NL.pdf (accessedJanuary 8, 2010).

USDA Forest Service. 2005. 36 CFR Parts 212, 251,216, and295, Travel Management; Designated Routes and Areasfor Motor Vehicle Use; Final Rule. Available athttp://www.fs.fed.us./recreation/programs/ohv/final.pdf(accessed January 8, 2010).