Assessin the Economicacts of Recreation

and TourismDennis B. Propst, Compiler

It is easy to extend the ideas of multi-plicative models and estimated multiplicands tothe calculation of secondary impacts. The USC ofinput-output or multiplier techniques also in-volvc products of cstimatcd parameters. A commoncxamplc is the use of an average wage-to-saltsratio to estimate wage impacts from final demand.

TYPES AND SOURCES OF ERRORS

When a sample of data is used to generatc anestimate of a parameter of the underlying popu-lation thereis usually a sampling error. In thecase whcrc final demand by tourists is the popu-lation parameter, a common estimate is the productof the estimated number of tourists and theirestimated average expenditures. The samplingerror varies from sample to sample and, as thesample size increases and approaches the popu-lation size, the sampling error disappears.Thcrcfore it is convcnicnt to discuss an csti-mator's average behavior over many small samples.The two most common small sample propcrtics arcthe bias and the variance of an estimator. Thebias is the difference bctwccn the true parameterof a population (say, final demand) and the avcr-age value taken by the estimator over many samples.The variance is a measure of the spread of thevalues taken by the estimator around its ownaverage in different samples. These propertifsarc related to the sampling error through theavcragc of its square. That is, the mean squaredsampling error equals the sum of the variance andsquared bias of the estimator. It is convenientto look at the bias and variance scparatcly sincean estimator may be unbiased with a large vari-ance or biased with a small variance. To judgeoverall accuracy, the mean squared error cri-terion (if it can bc calculated) is superior toeither bias or variance criteria alone.

Errors Due to the Model

The errors made in estimating final demand bytourists arc from two general sources: the mul-tiplicative model, which relates the sample datato the population parameter, and the sample dataused in estimation. The accuracy of a modelwhich is the product of scvcral estimated valuesdepends on the properties of the individual csti-mates as well as the indcpcndcncc of the multi-plicands. For cxamplc, when final demand iscalculated as the product of the estimated numberof tourists and the estimated cxpcnditures by theaverage tourist, the bias and variance of eachestimate is compounded in the final demand csti-mate. Both the bias and variance of their prod-uct dcpcnd on the indcpcndcnce of the number oftourists and the level of their cxpenditurcs. Ifthe estimates of the number of tourists and theiraverage expcnditurcs arc unbiased and the twovariables they rcprescnt arc indcpcndent, thertheir product will be unbiased. Othcrwisc, thecstimatc of total tourist cxpcnditures will bcbiased. For example, if weekend tourists spendless per day than weekday tourists and the formeroutnumber the latter, the two variables arc notindepcndcnt and the estimate of final demandbased on the product of means from random sampleswill be biased downward.

The variance of the final demand cstimato ismore complicated; it is not mcrcly the product ofthe variances. An unbiased cstimatc of the vari-ance of a product of independent random variablesis given by the sum of three terms: the varianceof the first variable wcightcd by the squaredsample mean of the second, the variance of thesecond variable weighted by the squared samplemean of the first, and the negative of the prod-uct of the two variances (Goodman 1960). Whenrandom variables are not indcpcndcnt, the vari-ancc of their product bccomcs even more complex.

The same properties are true for multiplicativesupply-related models where, for example, averagereceipts arc multiplied by the average number offirms. When models comprise products of three ormore cstimatcs, the compounding of errors con-tinues.

To illustrate the bias caused by the multi-plicative model, consider the case of an impactstudy of tourism in Wcstcrly, Rhode Island (Tyrrclland others 1982a). In this study a preliminarymodel was used to estimate the wages paid tolocal residents by the 35 hotels in the town.Estimated seasonal hotel capacity was multipliedby an estimated occupancy rate to arrive at totaloccupant days. The latter was multiplied bycstimatcd sales per occupant to arrive at totalsales. This was multiplied by an estimated wage/sales ratio to arrive at total wages. Finallythis was multiplied by the proportion of wagespaid only to Wcstcrly residents by the hotelindustry. By the conclusion of the study, allhotels had been surveyed and the final impact andeach of the intermediate figures were known.These data permitted us to compare the calcu-lations of the preliminary multiplicative modelby using very precise multiplicands (assumed tobc unbiased with zero variance) to the exactintcrmediatc and final values. The only errorsthat would be genzratcd by this process would bethose due to dependence among variables. Thusthe bias caused by the form of the model could bcexamined independently of that caused by thedata. It turned out that the 35 hotels of Wcstcrlydid not constitute a very homogeneous group andthat a significant amount of bias was introducedbccausc WC had assumed that the variables wcrcindcpcndent.

The results shown in table 1 reveal that thefirst two products wcrc both biased downward; thecomputed value for occupant days was 7.2 pcrccntbelow actual and the hotel salts cstimatc was31.6 pcrccnt below actual. This was caused by apositive correlation bctwcen variables: largerhotels had higher occupancy rates and higher roomrates (salts per occupant day). The accumulatedestimate was 36.5 percent below actual from thetwo calculations. The third and fourth productswere both biased upward: the incrcmcntal bias intotal wages was +4.8 percent and the incrcmcntalbias in Westerly wages was +9.0 percent. Thelarger hotels had lowc?r wage/sales ratios andlower Wcstcrly/total wage ratios. The accu-mulated bias dropped to -33.5 pcrccnt by thethird product and to -27.5 pcrccnt by the fourthproduct. This result illustrates the scriousncssof the problem with the multiplicative model.

41

Table l.--Bias in a model of local wages paid by the Westerly Hotelindustry, Rhode Island

Accumulated Incremental Accumulatedcstimatc % bias % biasComputation Actual -

Hotel capacity

x Occupancy rate

= Occupant days

x Sales peroccupant day

= Salts

x Wages/sales

= Total wages

x Westerly wages/total wages

= Westerly wages

202,758

0.7743

169,206 156,996 -7.2 -7.2

$21.01

$5,195,806.

0.0809

$401,064

$3,298,476. -31.6 -36.5

$266,847 +4.8 -33.5

0.9386

$345,424 $250,462 +9.0 -27.5

Even prccisc cstimatcs of the multiplicands didnot overcome the bias caused by the form of themodel which, in this case, was an underestimateof wage impacts of more than 25 percent. If itwas dcsirablc to USC the multiplicative model onewould need to disaggregate the data into homog-zncous groups for which the variables may bc lessdcpcndcnt.

Errors Due to the Data and Their USC

The sources of error in estimating final demandrclatcd to the USC and collection of data can bcclassified as follows:

Errors due to the model

Errors due to the data and their USC

Sampling errors

Nonsampling errors

Survey designInsufficient frameBias in sample sclcctionInadcquatc sample size

Survey executionNonobscrvation biasMcasurcmcnt errors

Data analysisProcessing errorsImproper statistical methodsFaulty intcrprctations

The first category refers to the multiplicativemodel dcscribcd previously. In the second cat?-gory, little can br said about sampling errorsexcept that if it wcrc possible to intcrvicw

every tourist and obtain accurate, relevant in-formation from each, then results would have noerror. Unfortunately, cost and feasibility usu-ally limit surveys to small samples, and samplingerrors will necessarily exist unless entire pop-ulations arc surveyed.

Nonsampling errors, on the other hand, can bccontrolled and minimized to a considerable degrcz.The eight major sources of thcsc errors resultfrom the design and execution of the survey, andthe analysis of the data. Sampling methodologyand recommended strategies for overcoming theeight types of problems arc thoroughly discussedclscwhcrc (Cochran 1977). Howcvcr, it seemsappropriate to comment on these nonsampling er-rors as they rclatc to some? of the unique fca-turcs of rccrcation and tourism surveys.

Insufficient frame.--It is obviously importantto identify the population of rccreationists,tourists, or firms in the industry, but a list orarca description of the population is frequentlyinadequate. It is not a trivial matter, however,to design a sufficient frame. It is usuallyimpossible to list tourist populations because oftheir size. Access to the population of tour-ists, for cxamplc, may bc limited to times whenthey arc participating in recreation or tourismactivities. Since records of all individualparticipants usually arc not kept, counting thesame individual more than once during a season isunavoidable. Thus, attendance records cannot betranslated directly into a frame. In addition,it is not advisable to treat an individual countedtwice as two different individuals. Differentbehavior and ?xpcnditurcs may bc associated withrcpcat visitors and one-time visitors.

Bias in sample sclcction.--Error occurs if thesample is chosen from the frame in such a way

42

that the population is misrcprescnted. Simplerandom sampling or stratified random sampling maybc sufficient to prevent biased sample selectionbut it is often difficult to ensure randomnessacross a season of visitors due to the expense ofinterviewing. Reweighting observations on thebasis of known population characteristics from asecond source can reduce some of this bias.

Inndcquatc sample size.--0ftcn the dcsircdprecision of the cstimatcs cannot be obtainedfrom the number of observations in the sample.Increasing the size of a sample is one of themost commonly discussed methods of reducing cr-rors when designing or conducting a survey. Thisis because the standard deviation of the mean ofthe sample has a simple invcrsc relationship tothe square root of the sample size. This meansthat the cost of sampling is the only reason notto reduce errors in this way. It should be cau-tioncd, however! that most biases cannot be over-come by incrcaslng sample sizes.

To illustrate the relationship bctwccn samplesize and the precision of an estimate, sclcctcdresults from four tourist surveys conducted inRhode Island over the past 2 years arc given intable 2. Each survey had a slightly differentpurpose and was conducted in a slightly differentmanner. All of the survey instruments included aquestion asking for per capita daily food cxpcnd-itures during a leisure trip or a vacation. Thesurveys were of southern Rhode Island beach users(Tyrrcll and others 1982b), Westerly, Rhode Island'shotel guests (Tyrrcll and others 1982a), NewportInternational Sailboat Show (NISS) patrons andthe Newport Yachting Ccntcr's boat manufacturers'

rcndczvous events participants (Tyrrcll 1984).It is convcnicnt to refer to the four as the

beach, hotel, boat show, and boating cvcnt sur-vcys. By using the formula for the standarddeviation of the mean, the four sets of resultswere used to compute sample sizes necessary forthe same rflativc precision of a per capita foodcxpcnditure cstimatc.

The beach survey was conducted by a singleintcrvicwcr who spent 15 to 30 minutes with eachbeach user to complete a multipurpose question-naire. Considerable care was taken to cnsurcrandom sampling. The population of beach userswas cstimatcd to be 64,000, of which 352 wereintervicwcd; the cost per observation of the 272responses that could be used for estimating av-crage food cxpcnditurcs was $10.79 (includingcoding and keypunching).

The hotel survey dcpendcd on volunteer rc-spondents to questionnaires handed out by hotelmanagers. There was no follow-up survey andobservations wcrc not rcwcightcd to compensatefor nonrespondents. The response rate was lowand the results arc bclicvcd to be biased. Thepopulation was estimated to be 25,500; 200 ques-tionnaires were distributed td the hotels; 21useful rcsponscs wcrc reccivcd and the cost perobservation was $3.81.

The boat show survey was conducted during 4days of the NISS by 10 diffcrcnt interviewers whospent 5 to 10 minutes with each patron. Somemcasurcs were taken to cnsurc random sampling.The population was estimated to bc 12,000, ofwhich 492 were interviewed; the cost per obser-vation was $3.00.

The boating event survey was conducted by mailfrom the list of event participants. No follow-up questionnaire was sent, but the response rate

Table Z.--Sample size and precision of four Rhode Island tourist surveys

SurveyBoat Boating

Variable Beach Hotel show event

Population size (no.) 64,000 25,500 12,000 350

Sample size (no.) 352 21 492 126

Cost/observation ($) 10.79 3.81 3.00 2.67

Mean food cxpcnditurcslcapita/day ($1 9.86 16.05 16.69 16.86

Standard error ($) 10.28 9.38 19.38 16.88

Number of observationsrequired for 95% CIof +lO% mean (no.) 417 131 518 7aa_

Total cost for CI ($) 4,500 500 1,550 208

CI = confidence interval.

aBccausc of the small population relative to the sample size, a finitepopulation correction factor was used.

43

was very high. The population was 350; 126 rc-sponded and the cost per observation was $2.67.

It is not possible to compare the averago re-sponses to the food cxpenditurc questions bccauscof the difference in the populations surveyed.Howcvcr, to cxaminc the tradeoffs bctwcen samplesize, precision, and the cost of sampling bydiffcrcnt techniques, the estimated standarderrors can be rclatcd to a +lO percent intervalaround the respective means: A slight modi-fication permits the surveys to be compared onthe basis of the numbcr of observations requiredfor a 95 percent confidence interval (CI) of thatsize. Multiplying this numbcr by the cost perobservation gives an indication of the relativeefficiency of the diffcrcnt surveys in producingan average per capita food expenditure cstimatcwith the same relative precision.

The results of this comparison arc that theboating event survey would require the fewestobservations and cost the least to produce a 95pcrccnt CI +lO percent around the mean; the hotelsurvey rankzd second in observations rcquircd andcost; the beach survey ranked third in obscr-vations required, but fourth in cost; and theboat show survey ranked fourth in observationsr-cquircd and third in cost.

The simple interpretation of these results issomewhat misleading since the results of thehotel survey wcrc biased. In fact, from otherdata on Westerly hotel visitors, it was cstimatcdthat the bias in this survey was considerable,overwhelming its small variance in its mcan squarederror (MsE). The other survey cstimatcs werebclicvcd to bc rclativcly free from bias so thattheir MSE's arc the same as their variances.Reranking the survey approaches on the basis oftheir MSE's put the hotel survey last and loftthe others in their same rclativc positions.Because of the size and nature of the bias in thehotel survey estimate, it was cstimatcd that ovena sample of 1,000 hotel visitors would not haveproduced the desired level of precision.

The most successful survey was the one con-ducted at the boating event. Its advantage wasthe small population sampled and enthusiasticcooperation of the boaters. The rcsponsc ratewas 36 pcrccnt. The boat show survey was alsorclativcly successful. Its advantages were thehigh rcsponsc rate because of the intcrvicw ap-proach and the brevity of the survey. The lengthof the questionnaire was the downfall of thebeach survey, which took much time to conduct,cod?, keypunch, and analyze.

Non-observation bias.--A bias results from alack of measurcmcnts for some of the individualsin the selcctcld sample bccausc of failure tolocate thorn or from refusals to answer questionsby those who wcrc located. This was one of theproblems with the hotel survey. It is also aproblem with most mail surveys; the boating cvcntsurvey was an cxccption. A successful strategyin cases known to the author has been to conducta scrics of follow-up rcmindcrs, questionnaires,and telephone calls. Even if the respondent does

not answer all the questions, it is usually pos-siblc to adjust results for biases based on someknowlcdgc of the characteristics of nonrcspondcnts.

Measurement errors.--The diffcrcnce betweenaccurate information and the response to a qucs-tion leads to measurement errors. Such errorsarc commonly caused by a poorly worded questionor the failure of a rcspondont to recall accurateinformation. Careful design and cxtcnsivc testingof a questionnaire is the only solution.

Processing errors.--Thcsc errors occur in cd-iting, coding, and tabulating results.

Improper statistical methods.--If incorrectassumptions about the distributions of variablesare made and the statistical proccdurcs are basedon these assumptions, then there will bc errorsin the data analysis.

Faulty intcrprotations.--Errors in data anal-vsis are made if the results from one sample ofthe population are incorrectly cxtrapolatcd toother samples, or when the meaning of surveyrcsponscs are altered by erroneous induction orthe careless USC of words.

All of these errors arc serious and most can bcavoided by careful attention to details of theproject.

CONCLUSlONS AND RECOMMENDATIONS

This paper has attempted to cxaminc data collec-tion and USE in estimating final demand by tour-ists. The approach has been to rcvicw the impli-cations of the traditional multiplicative modeland the proccdurcs by which the multiplicands arecstimatcd. There has been no attempt to identifyall possible sources of data for this typo ofanalysis, which is done clscwhcrc (Goeldncr 1980;WV Univ. 1981). Furthcrmorc, data availabilityis a problem that has no gcncral solution but onethat must bc solved by local research. The focushcrc has boon on two types of error, bias andvariance, and on the gcncral sources of thcscerrors in traditional rcscarch cfforts.

Four recommendations arc offered:

1. On the choice of a model for final demand.Since tourism is a multigood, multiscrvicc indus-

try, a very complex model is implied. Howcvcr,limitations of data will usually permit only theuse of simple models. If the traditional multi-plicativc model is used, the biases caused bycorrelations between variables should be ac-counted for. The simplest way is to disaggregateand USC a sum of products over the most homoge-neous groups possible.

2. On sample size. The required sample sizecan bc calculated from a desired dcgrcc of prc-cision and a previous cstimatc of the standarderror of a variable. This calculation does not,however, account for the bias which may bc presentin the estimate used. Also, a large sample sizewill not generally ovcrcomc biases in samplingprocedures.

3. On the use of secondary data. Make USC ofall that is available since it is usually veryinexpcnsivc to obtain. Howcvcr, sclcct only datathat can bc rclatcd to final demand by a rca-sonablc and relatively simple model. Also, donot ncglcct the need for measures of variance inthcsc data.

4. On survey dcslgn and execution. Bc asconcise as possible in asking survey questions,and test questionnaires cxtcnsively. The cost ofinformation somctimcs increases more than propor-tionately to the length of a questionnaire. Final-lY> poor survey designs and executions arc themajor causes of biases. It helps to keep themean squared error criterion in mind.

LITERATUKE ClTED

Cochran, William G. Sampling techniques. NewYork: John Wiley & Sons; 1977. 400 pp.

Gocldncr, Charles R. Data sources for traveland tourism research. In: Hawkins, Donald E.;Shafer, Elwood L.; Rov?lstad, James M., cds.Tourism marketing and managcmcnt issues.Washington, DC: George Washington UniversityPress; 1980: 277-290.

Goodman, Leo A. On the exact variance of products.Journal of the American Statistical Association55: 708-713; 1960.

Tyrrell, Timothy J. The economic impact of theboating events at the Newport Yachting Centerin 1982 on the City of Newport, Rhode Island.Marine Technical Report 86. Kingston, RI:University of Rhode Island; 1984. 30 pp.

Tyrrcll, Timothy J.; Emerson, William K.B.;Molzan, David E. The economic impact of tourismon Wcstcrly, Rhode Island. Bull. 433. Kingston,RI: University of Rhode Island, AgriculturalExperiment Station; 1982a.

Tyrrcll, T.J.; Morris, D.M.; Alba, A. The 1981Rhode Island beach USC and value survey: surveydesign and preliminary results. Working Pap.10. Kingston, RI: University of Rhode Island,Department of Resource Economics; 1982b.

West Virginia University. Creating economicgrowth and jobs through travel and tourism.Mcrgantown, WV: West Virginia University,Bureau of Business Rcscarch, Collcgc of Businessand Economics; 1981. 317 pp. (Prepared for theU.S. Department of Commerce.)

45

Computerized Models for Assessing the

Economic Impact of Recreation and

Tourism

Robert C. Bushnell and Matthew Hyle1

INTRODUCTION

Many approaches to national economicmodeling have been taken. When the word"impact" is used, it is generally truethat it is the disaggregate interactionsof the economic process that are beingstudied. Simulations and multiple-equation econometric models sometimesfill this role, but most often it isinput-output analysis that is employed.

At the national level, the Bureau ofEconomic Analysis (BEA) of the U.S.Department of Commerce expends much timeand painstaking effort in establishingthe classification and flows of paymentsthat make up the national input-outputmodel. While a number of small areamodels have been constructed andutilized at the state and local level,the effort involved in compiling thedata is usually more costly than thestudy area can afford. Hence, in recentyears the attempt to develop regional orsmall-area input-output models by usingso-called non-survey methods hasincreased. In the last 10 years, anumber of systems have arisen whichgenerate small area, state, or regionalmodels from the technical coefficientsmatrix of the U.S. National Input-Outputmodel. Three of these models will bediscussed.

RIMS II Modeling System

The first such system is the so-calledRIMS II, constructed and supported bythe Regional Economic Analysis Divisionof the BEA.

RIMS II (Regional Input-OutputModeling System, version II) uses the1972 BEA 496 input-output national modelas the basis for the regionalcoefficients. The coefficients aremodified by the use of the RegionalLocation Quotient (LQ) technique:

the national direct-require-ment-coefficients matrix is

1 Associate Professor and Assistant Pro-fessor, respectively, Department ofFinance and Business Economics,School of Business Administration,Detroit, Michigan 48202.

made region-specific by using4-digit SIC location quotients.According to this mixed-LQapproach, BEA county personalincome data, by place of resi-dence, are used for the calcu-lation of LQ's in the servicesectors, while BEA earningsdata, by place of work, areused for the LQ's in the non-service sectors. The LQ'sare used to estimate theextent to which direct re-quirements are supplied byfirms within the region.(Cartwright and others, 1981)

Simple location quotients are defined bythe following relation:

LQ(i) = E(i,r)/E(*,r)E(i,n)/E(*,n)

(1)

where: E(i,r) = Earnings in the ithindustry in the rthregion,

* = the sum over allindustries, and

n = the sum over allregions.

Hence the concept is the ratio of theproportion of industry i's earnings ofall earnings in region r to the similarproportion of industry i's earnings inthe nation as a whole. An industry inwhich the region specializes will havean LQ greater than 1; an industry inwhich the region does not specializewill have an LQ less than 1.

If

a(i,j,r) is the proportion of thetotal output of the regional indus-try j that is accounted for by thepurchases of inputs from regionalindustry i, and

a(i ,j,n) is the national direct-requirements coefficient,

the relationship is:

a(i,j,r) = LQ'(i) * a(i,j,n) (2)

where: LQ'(i) is LQ(i) if LQ(i) is lessthan 1 or 1 if LQ(i) is greaterthan 1.

This latter condition reflects tile factthat the supplying industry willcertainly not supply more than thedemanding industry requires, even if thesupplying industry is substantiai.

The household row is derived from thenational row by assuming that the valueadded/gross output ratios from thenational model hold for all regions.The household column is derived from the

46

national vector. The national vector isscaled down by multiplication first by(l-T(r)), where T(r) Ls the averageregional tax rate, and then by C(r),where C(r) Ls the national after-taxconsumption rate as measuredby the ratioof National Personal ConsumptionExpenditures to National DisposablePersonal Income. After theseadjustments, each member of the vectoris multiplied by the correspondingLQ(i,r) as was described previously forthe industrial columns.

Since the regional A matrix has nowbeen developed, no estimate of regionaldemand or gross output is needed sincesimply inverting the (I-A) matrLx willprovide the multtpliers. If the Amatrix does not Include the householdsector, the multipliers include thedirect and indirect effects (Type Imultipliers). If the A matrix includesthe household sector, the multLpliersinclude all of the direct, the indirect,and the induced effects (Type IImultipliers).

The RIMS II system also includes some"shortcut" methods where persons wishingto develop overall Impact or earningsmultipliers may do so withoutundertaking the Inversion of the full(I-A) matrix.

The REMI Models

The second model to be considered Lsmaintained by Regional Economic Models,Inc. (REMI). As is the case with RIMS,the REMI model is based on the latestavailable natLona1 Lnput-output modelfurnished by the BEA. It, too, relieson multiplying each of the nationalcoefficients by a factor in order todownscale the multLplLer from nationalimpacts to figures approprLate for thesmaller region.

The REM1 approach, however, Lsdifferent; it uses a concept termed"Regional Purchase Coefficients"(RPC's). The RPC is the proportion of aused commodity purchased by a usingindustry from within its own region.Unlike the LQ's which are applied to theLnverted technical coefficient matrix,the RPC's are applied to the techn.LcalcoefficLents directly, aff_er which thetechnical coefficient matrix is invertedin the normal way. In general, the datarequired to estimate the RPC's directlyare not avaClable, therefore they areestimated by REMI from a regressionequa~.lon.

The basic idea behind the regressLonestimation is that regional purchasesshould be a function of relativedelivered costs where delivered costsare the sum of production and

transportion costs. RelatLve productioncosts should be a function of relativewages, relative other costs, and arelative scale of production andtransportation costs, which is afunction of relative average shipmentdistances for local versus nonlocalpurchases. Average shipment distance 1sposited as being a function of theproportion of shLpper-to-users Ln theregion to shippers-to-users Ln thenation, and the proportlon of land areaIn the region to land area in thenation.

Using this theoretical structure, aregression relation was developed forestimating the log of the RPC for eachof 19 industry groups as a linearfunction of the ratLo of industry per-worker wages in the region to thenatlon, the ratio of industry employmentin the region to the nation, the ratioof Industry national output tonnage toindustry total wages, the locationcoefficient LQ, (as defLned for the RIMSmodel), and the ratio of the land areaof the local area to the land area ofthe nation.

SLnce the independent variables, theRPC's themselves, are not dLrectlymeasureable, it was also necessary toinfer values for some of these in orderto estimate the coeffictents of themodel. This was accomplished by REMIthrough knowledge of the output of eachcommodLty in the local region.

RPC(i,r) = P(i,r) * Q(L,r)D(i,r) (3)

vJhere: Q(i,r) = amount of the commodityi produced in r,

D(i,r) = total use of i in r,and

P(i,r) = proportion of L pro-duced and used in r

The Q's can be measured; the D's areobtained by applying the technicalcoefficients to the Q's and then addingother use as by households, governments,capital expenditures, foreign exports,etc. The limiting factors were the P's.293 P's for 19 commodities could becalculated from the information in theCensus of Transportation. Thus 293 P'swere measured to calculated 293observatLons of the RPC's. An equationto estimate the RPC for any region basedon relative wages, relative employment,the LQ, the weight to wage-bill ratio,and relative land area was calclllated.In addition, 6 of the 19 comodities havenon-zero dummy varLable weights thaL areutilized. This equation is then used tocalculate the RPC for any region for anyoE the 500 commoditLes in the Eull model(for Imore detail, see Stevens andothers, 1980).

47

The IMPLAN Models_-

The third set of models to beconsidered, called IMPLAN, were directedand funded by the USDA Forest Serviceand produced by Engineering EconomicsAssocLates, Inc. Like the other models,this set also depends on a nationalmodel. As part of the effort, however,the Forest Service had the 1972 nationalmodel updated to 1977 for this project.The Forest service is continuing thiseffort and since the 1977 NationalInput-Output model was released by theBEA in 1984, the 1977 model is beLngupdated to 1982 in the same manner.

IMPLAN Ls different from other models,however, because the Forest Servicewished to have a model for every U.S.county (or parrlsh) which wouldaggregate into state models which, inturn, would aggregate into the originalU.S. model. Hence this system producesa flow or a transactions table for eachcounty (or aggregation of counties)whLch Ls then converted into atransaction matrix and then inverted toform the multipliers. As with the othersystems, both direct-and-indirect (TypeI) and direct-and-indirect-and-Lnduced(Type II) multtpliers may be produced.

With this goal, the task ofEnglneerLng Economic AssocLates was tofind justifiable proxies by which tobreak down the components of demand toestimate final demand vectors for:

Personal consumption expendituresGross private domestic investmentForeign exportsInventory changeFederal government expendituresState & local governmentexpenditures

In addition the following other elementsmust be estLmated for each sector foreach county:

Gross domestic outputEmploymentComponents of value added

Employee compensationProperty type LncomeIndirect buslnes taxes

The task Ls complicated by theat the state and county level,

Eact that.most OE

the economic data sets provided bygovernment agencies are character!zed byhavLng certain elements deleted. Thisis due to the legal restrictions on therelease of data gathered by governmentagencies in surveys of indlvldual firms.Therefore, techniques that generateproxy series were employed. Theseproxies could be halanced to knowntotals; for example, for the sum of all

the states at the natLona1 level, andfor the sum of all the countLes in astate.

An advantage of this process is thatthe flow table is generated for thelocal area. It may be inspected andaltered, if desired, before processinginto technical and inverse form. Amodel for any multlcounty area (standardmetropolitan statistical area - SMSA,BEA region, other aggregation) may beconstructed simply by aggregating thecounty data before applying it to thenational coefficient matrix. (Furtherinformation can be obtained fromEngineering EconomLc Associates, Inc.,1700 Solano Avenue, Berkley, CA 94701.)

ASSESSING THE ECONOMIC IMPACTS

All of these models are, Ln theory,equipped to assess the economic impactsof tourism and recreation. In practice,however, each model was designed withdifferent goals in mind so that theappropriateness of a model ~111, inpart, depend on how well the model canmeet the various demands placed on it bythe spec1fLc problem or user. Wh-lle thecriteria for evaluating a model will beshaped by the particular problem that isunder scrutiny, there are five issuesthat should he considered in anyapplrcation of a regional Input-outputsystem: (1) the level of regionaldisaggregatlon, (2) the level ofsectoral disaggregation, (3) thedefinitional basis of the sectors(commodity versus industryclassifications), (4) the relation ofthe direct requLrements matrix to theregion(s) under scrutiny, and (5) themeasurement of final demand. One shouldnote that these five issues arise quitenaturally out of the modeling processand therefore cannot he avoided.Conseqrlently, the following discussionshould not he construed as a criticism,,f any particular model or technique butonly as an aid in the evaluation of amodel's suitability In the use ofmeasuring economic Impacts.

To relate and clarify the Lssues andto give a review of the basic input-out.put relationships, take the followinghypothetical sltuatlon. A family fromWindsor, Ontario, t.akes a week'svacation in the Detroit metropolitanarea. They drive their car through theDetroit-Wtndsor tunnel and stay in ahotel in downtown Detroit. Each daythey drive around the area visitinglandmarks and parks. TheLr budget of$2,000 (American) Ls spent on lodging,gasoline, parking fees, admi.ssion fees,boa: rentals, and food (purchased eitherat restallrants or at grocery stores). Anat,lr.11 quest Len to ask is "What is this

48

family's economic impact on the Detroitmetropolitan area? What will thechanges be in total output and where?"

In theory, an input-output model cantrace the effects of this family'sexpenditures and their repercussionsthroughout the Detroit area's economy byemploying a basic input-output identity.Within an input-output table or model,the total dollar sales (or output) foreach sector must equal the sales to allsectors (including itself) for use asinputs into their production process(intermediate use) and sales to allfinal consumers (final demand). Usingalgebraic notation, this basicdefinition is written as:

q = A q + f (4)

where: q = a vector of sectoral outputin dollars,

A = a matrix relating the inputrequirements per dollar ofoutput, and

f = a vector of sectoral finaldemand in dollars.

Combining similar terms yields asolution imbedded in all input-outputmodels:

q = (I-A)-1 (5)

where: I = the identity matrix.

The importance of equatton (5) is notthe mathematics but that it shows, intheory, that only final demands and thedirect requirements (A) matrix areneeded to measure total productton.(See Mierynk, 1965 or Richardson, 1972for a more detailed explanatlon.) Inpractice, equation (5) shows that iffinal demands are measured correctly,and Lf the direct requirements matrixaccurately portrays the interrela-tionships within the economy, and if thematrix corresponds to the deflnitfonsand conventions used in measuring thefinal demands, then total production canbe measured.

These conditions may seem to be quiteobvious and harmless because each one ofthe input-output models discussed in thefLrst section do give answers to manytypes of questions similar in nature tothe example. Yet, the reliabL1Lty andaccuracy of those answers will depend onhow well the chosen model adapts to thefive generic issues. The first issue(regional disaggregation) provides astraightforward example of the problem.

Regional Dlsaggregation- - -

Ideally, one would hope to have themost disaggregated model possihle in

order to minimize any errors due toaggregation problems. However, manypractical considerations conspire torestrain the manageable level ofdisaggregation. Regardless of thoseconsiderations, the model to be chosenshould, as closely as possible, resemblein its level of regional disaggregationthe requirements of the problem to beexamined. For this example the modelthat has, among other features, input-output relationships for the city ofDetroit (or at least Wayne and OaklandCounties) should give the most reliablemeasurement of the economic impact. Anymodel that has the State of Michiganas its lowest level of disaggregationshould be avoided in this case since itwill require substantial adjustment inorder to yield reasonable estimates ofthe economic impact on the Detroit area.

Sectoral Disaggregation

This issue is very similar to that ofregLona1 disaggregation. Given thespending pattern of the hypotheticaltourist family, the ideal model shouldhave among its different sectors Hotel -Standard Industrial Code (SIC) 7011,Retail gasoline (SIC 5541), Parking lots(SIC 7523), Restaurants (SIC 58), andGrocery stores (SIC 5411).

In practice, the retail trade sector(any SIC of 5000-5800) presents twospecial problems. First, although thereis a wide diversity Ln the types ofretail establishments, most models haveonly a few retail trade sectors (mainlydue to data restrictions). Thisaggregation may impose some measurementbias with the extent of the biasdepending on how differently the varioustypes of retail stores use their variousinputs. Second, within the framework ofinput-output analysis, a retaL storedoes not produce any commodLties butonly a service by acting as a conduitbetween the actual prodicers and finalconsumers. Consequently, any commoditypurchased at a retail establishmentshould be "stripped" of the "service"component and counted in the commodity'sproduction sector. For our example, ifthe Canadian family purchased ahamburger from a restaurant, then thevalue of the restaurant service would besubtracted from the dollar value of thehamburger and then the final demands ofMeat (SIC 2010) would be increased (seethe discussion on the measurement offLna1 demand).

Definitional Basis- -

The third issue is the commodityversus industry definition of a sector.Input-output models can use either acommodity definition, whLch groupsproducts or services with similar SIC

49

codes into a sector, or an industrydefinition, which groups establishmentsinto a sector on the basis of similarprimary product. Establishments canproduce more than one commodity, butonly one commodity can be theestablishment's primary product(typically determined by the product orservice which has the largest dollarvolume). A common example of themulticommodity establishment is thelocal Sears store. This store may beoffering, under one roof, Auto repairservices (SIC 7500), Appliance repairservices (SIC 7600), Optometrist'sservices (SIC 8042), Upholstery cleanLng(SIC 7217), Real Estate brokering (SIC6531 and 6610), Insurance brokering (SIC6400), and Security brokering (SIC 6200)along with its traditional retailoperatlons. If the largest dollarvolume of sales Is in auto repair, thenthis partLcular establishment would becounted under the auto repair industryinstead of the department store (SIC5800) industry. (See ten Raa andothers, 1984, for a discussion ofsecondary products in a broadercontext.)

As a result, a user should be aware ofthe consequences of misapplying thesectoral definitions. If the userwishes to estimate the economic impactcaused by the change in demand for acommodity but is using a model with anindustry definition of a sector, theestimate could be inflated if thatindustry has inputs that are used in theproductLon of other dissimLlarcommodities. From our example, if theCanadian visitors purchased cheese froma grocery store and one placed thatcheese purchase in the cheese industryfinal demand sector, then one will findan increase in the use of milk, enzymes,and sugar since many cheeseestablishments also produce ice cream.

Fortunately, the errors stemming fromthis definitLona1 problem are likely tobe small when estimating the economicimpacts of tourism and recreation. Themultiproduct establishment Ls most.commonly found in the manufacturingsectors while the service sectors (withthe exception of department stores) tendto provide a single service. Because%he tourism and recreation industry islargely composed of the service sectors(or at least in most polLcy questionsf.his is true), this problem may notarise. Iq addition, most of the modelsh.ave a "make" table (which shows thedistribution of commodities that. eachindustry makes for the nation) avaylahle1.0 transform data from one definition toanother. Still, one is bet_ter off beingaware of the pot_ential complicat?ons In

order to assess possible errors in theestimates.

Direct Requirements (A) Matrix

The fourth issue revolves around theapplicability of a model's directrequirements matrix to the regional areaunder question. Every model in thispaper uses the direct requirementsmatrix for the national economy as astarting point (see U.S. Department ofCommerce, 1984 for the latest update).This matrix based on national averagesis then imposed on a region and, ineffect, split into a local matrix (inorder to capture local impacts) and an"import" matrix, but the overallrequirements always equal the controlimposed by the national average.Consequently, the regional input m-lx(regardless of the Lnput's geographicalorigin) for a dollar's worth of asector's output is assumed to equal thenational average for that sector.

The assumption of identLca1 inputrequiremens among different regions maynot be completely desirable, but it iscertainly not unreasonable In theabsence of any additional information.Yet, this assumption results in somemismeasurements, with the extentdepending on how much the regional usediffers from the national average. (Onesuspects that as the region increases insize this difference grows smaller.)For example, the electric utility sectorin the natLona1 matrix combines nuclearpower plants, dams, and coal-firedplants, but a region (especially acounty or group of counties) will useelectricity from only one type of plant.Thus the use of a national average maymisrepresent the economic impact. Onemay argue that in the tourism andrecreation industry, which is dominatedby services, this effect can beneglected because servLce sectorsgenerally use the same technologies.This is open to question, however,because some regions wLth relativelyhigh labor costs may substitute capitalequipment for labor, which should changethe overall input requirements for thosesectors.

If a user has additional Lnformationabout the structure of a region'seconomy, then the estimates of aneconomic impact could be improved ifthat information could be incorporatedinto the model. Consequently, anotherfeature of a prospective model to keepin mind is that model's capability of!ncorporat_ing any additional informatlonabout the target. region's economy. No tonly should the capability for Lncor-poration of new data be present but theprocess should he relatively easy.

50

Measurement of Final Demand

In order for an input-output model toestimate economic impacts, thecategories of final demand (f vector inequation (2)) should correspond asclosely to the sectoral definitions asthose of the direct requirements matrixto ensure a more accurate measurement ofthe economic impact. In general, thecloser the correspondence of the two,the more reliable is the final measure.

As mentioned, a slight technicalproblem occurs with purchases fromretail establishments. The mostdesirable outcome is to have a largeamount of detailed informationconcerning these purchases. Prom theCanadian visitor example this would meanthat, ideally, one would have anaccounting for each meal by item --Monday's breakfast was two eggs, threebowls of corn flakes, etc., and eachretail purchase by item. However, thatdetail is not always available. If itis not, then the model or modeler shouldhave some well-defined method to "break-up" these types of purchases.

SUMMARY

The discussion has focused on a fewpotential pitfalls or issues that shouldbe addressed by any researcher usingregional input-output models. Becomingaware of the issues allows the user tomore carefully assess the suitability ofa particular model to the demands of theanalysis. However, these are simplyguidelines and cannot help unless theproblem to be analyzed has been clearlystated in terms that an input-outputmodel can handle. There can be nosubstitute for careful consideration on

the part of the user in structuring theresearch problem. Part of that carefulconsideration should include thelimitations and strengths of the user'sparticular model, not only in light ofthese few guidelines but of the entirestructure of the model.

LITERATURE CITED

Cartwright, Joseph V.; Beemiller,Richard M.; Gustely, Richard D.Regional input-output modelling system.Washington, DC: U.S. Department ofCommerce, Bureau of Economic Analysis,1981. 139 pages.

Mierynk, William H. Elements of input-output economics. New York: RandomHouse, 1965. 157 pages.

Richardson, Harry W. Input-output andregional economics. New York: JohnWiley h Sons, 1972. 294 pages.

Stevens, Benajmin H.; Erlich, David;Bower, James. Estimation of regionalpurchase coefficients and their use inthe construction of non-survey input-output impact models. Reg. Sci. Res.Inst. Discuss. Pap. 119. Amherst, MA:Regional Science Research Institute,1980. 49 pages.

ten Raa, Thijs; Chakraboly, Debesh;Small, Anthony. An alternativetreatment of secondary products ininput-output analysis. Review ofEconomics and Statistics 66(1):88-97,1984.

U.S. Department of Commerce. The input-output structure of the U.S. economy.Survey of Current Business 64(5):42-84,1984.

51

Assessing the Secondary Economic Impacts of

Recreation and Tourism: Work Team

Recommendations

Dennis B. Propst, Dimitris G. Gavrilis,

H. Ken Cordell, and William J. Hansen’

The U.S. Forest Service and U.S. Army Corps ofEngineers are major providers of leisureopportunit ies . The approximately 91 millionacres of public land in the National ForestSystem include more than 25 million acres ofWilderness, 640 miles of Wild and Scenic Rivers,and 24,000 developed and dispersed recreations i t e s . Similar figures for the Corps ofengineers include 11 million acres of land andwater, 442 lakes and other project areas, andover 3,800 recreation areas. The amount ofpublic consumption generated by this enormousfederal supply of recreation opportunities issubstantial : Forest Service - -233 mi l l ionvis i tor days ( f i sca l year 1983) ; Corps o fEngineers--469 million recreation days of annualuse.

Although the Forest Service and the Corps areIntegral parts o f the le isure industry , l i t t leis known about the secondary impacts of thefederal supply on community, regional, andnational economic development. To c lar i fy thatstatement, it is necessary to d ist inguishbetween primary and secondary impacts thatresult from federal investment in providingle isure opportunit ies .

Primary or direct impacts include benefits torecreation users and costs to the providers.These are the measures needed to deriveb e n e f i t - c o s t r a t i o s , which guide investmentd e c i s i o n s . A great deal of research since themid-60’s has been directed toward determiningthe direct benefits of recreation developments.Travel cost and contingent valuation methods arethe two most widely used and recommendedprocedures (Dwyer et al ., 1977; Walsh, 1984).

A similar effort has not been aimed atdeveloping concepts and procedures for examiningsecondary or indirect economic impacts ofsupply ing recreat ional services and fac i l i t ies .Secondary impacts include benefits and costsbeyond those to users and providers. Secondaryimpacts accrue to communlttes, regions, and thenation in the form of income, employment, retailsales , taxes , a n d development of related

1Assistant Professor and Graduate Assistant,

Department of Park and Recreatfon Resources,Michigan State University, E. Lansing, MI 48824:Project Leader, Southeastern For. Exp. Stn.,U.S.D.A., Forest Service, Athens, GA 30602; andEconomist, Waterways Exp. Stn., U.S. ArmyEngineers, Vicksburg, MS 39180.

industries (recreational equipment, information,s e r v i c e , and development industries, such assecond homes, condominiums, and resorts).

Agencies like the Forest Service and Corps ofEngineers require information on secondanyeconomic impacts to make financial allocationd e c i s i o n s . In addition, demonstration of theimportant role that such agencies play in local,reg ional , and national economic developmentwould likely provide more impetus for privateand nonfederal provision of recreationopportunities on or near Corps projects andNational Forests.

STATEMENT OF THE PROBLEM

A system for deriving estimates of thesecondary economic impacts of recreation iscurrently lacking, primarily because availablemethods are costly. Ideal ly , the researcherwould want to generate a unique multiplier foreach economic sector in which first-roundrecreation spending occurs. The methodology toderive unique multipliers exists, but the largedata requirements make this procedure expensiveand time-consuming (Marino and Chappelle, 1978;Leistritz and Murdock, 1981) . The a l ternat iveto collecting a large amount of data over timeis to use input/output (I/O) models developed bygovernment agencies and assume that theirmult ip l iers are accurate . However, exist ing I /Om o d e l s generaJly are not based on sufficientlydetailed breakdowns of the sectors impacted byrecreation and tourism (e.g., marinas,recreational equipment). Thus, t h e r e l i a b i l i t yof such multipliers is unknown (Gartner andHolecek, 1982; Stynes and Holecek, 1982). Torestate the problem, the secondary impactassessment process for other U.S. industries( e . g . , manufacturing) is reasonably clear andwell-developed, but (a) the appropriate economicimpact assessment procedures for recreation areunclear , and (b) the necessary data forconducting such assessments are often missing.

OBJECTIVES

To help solve the problem stated above, thefollowing objectives were pursued:1. To evaluate the state of the art indetermining the secondary economic impacts offederal recreat ion fac i l i t ies and serv ices atl o c a l , regional , and nat ional level ;2. To prepare a detailed economic impactassessment procedure and data collectionmethodology to be employed by the Forest Serviceand Corps of Engineers in determining theimpacts stated in Objective 1.

SCOPE

A range of methods was needed to achieve theseob ject ives . Computer ized l i terature searches ,personal communications, and library researchwere the primary means of achievfng the firstobjective. Objective 2 was achieved throughcontacts with key government agency, university,and industry profess ionals . These contacts were

52

necessary in order to synthesize a tremendousamount of information into recommendations forthe appropriate variables, models, procedures,data sources, and economic sectors to beemployed in the economic impact assessment ofrecreat ion . Some of these contacts were made bytelephone and letters. A majority of theinformation obtained for achieving Objective 2,however, emanated from a technical meeting onassessing the secondary economic impacts ofrecreation and tourism held at Michigan StateUniversity in May of 1984. The goal of thismeeting was to bring together a few keyprofess ionals to ident i fy the best avai labletechnology and data for economic impactassessment of recreation and tourism. Thispaper reports the methodology employed duringthe meeting to meet study objectives as well asthe results of the meeting. The meeting’smethodology was highly successful in transferingtechnology and in identifying importantconsiderations for economic assessment ofrecreat ion . The full report (Propst andGavri l ia , 1984) contains the results o f a l lmethods used to satisfy the two studyo b j e c t i v e s .

PROCEDURES

The technical meeting on assessing thesecondary economic impacts of recreationincluded both presentations by invited speakersand a workshop. A series of steps was followedinorder to se lect the inv i ted speakers . Duringthe fa l l o f 1983, a master l i s t o f regionaleconomics professionals was compiled throughpersonal communications with resource andagricultural economics faculty at numerous U.S.u n i v e r s i t i e s . At the same time, Forest Serviceand Corps of Engineers researchers, planners,and administrators, were contacted to compile alist of issues that both agencies wanted toreso lve . Potential speakers were sent a letterexplaining the purpose of the meeting, listingthe ident i f ied issues , and seeking theiri n t e r e s t . In addition, potential speakers wereasked to indicate from the list of issues thetop three on which they would be willing toprepare a presentation. Potential speakers weretold that the presentations of the invitedspeakers would be published and all travelexpenses paid. A list of invited speakersemerged from this initial wave of correspondence(see Appendix A). The issues covered in papersand formal presentations by the eight invitedspeakers are listed in Table 1. The formalpresentations took approximately 1 day andprovided the necessarworkshop portion of

background for-the

The workshop portion of the meeting lasted 2days. During this time, participating ForestService and Corps of Engineers researchers witheconomics backgrounds became actively involvedin discussions . These participants are alsolisted in Appendix A. For 2 days, all meetingparticipants were divided into work teams offour to five members and asked to complete t’hetasks stated in Table 2. These tasks werewritten to be more specific reiterations of the

Table l.-- General issues covered in formalpresentations during the “Technical Meeting onAssessing Secondary Economic Impacts ofRecreation and Tourism,” Michigan StateUniversity, 14-16 May, 1984.

1. What is the state of the art in developingmultiplier for assessing the secondary economicimpacts of recreation and tourism?

2. What methods besides I/O analysis areavailable for assessing the secondary economicimpacts of recreation and tourism?

3. How should regions be defined and sectorsdisaggregated in existing I/O models to accountfor the secondary economic impacts of recreationand tourism?

4. What are the data requirements andappropriate sources of data for assessing thesecondary economic impacts of recreation andtourism?

5. What are the strengths and weaknesses ofusing I/O analysis to assess the secondaryeconomic impacts of recreation and tourism?

6. What computerized models for assessing thesecondary economic impacts of recreation andtourism are currently available and what aretheir strengths and weaknesses?

7 . What are the data requirements and sourcesof data for measuring the economic impacts ofthe Forest Service and the Corps of Engineerssupply on leisure/tourism fidustries, such asrecreational equipment, second homes,recreat ional vehic les , and boat inq?

issues covered the previous day by the invitedspeakers. Work teams were arranged so as tocontain both university and agencyrepresentat ion . All teams worked on the sametasks , one at a time, after being given thefo l lowing instruct ions :

“There is a specific environment or mood wewould like to create in each group in order tobe most efficient in satisfying meetingo b j e c t i v e s . This mood can result if you keepthe following points in mind:

a.

b .

C .

we have a series of specific problems tosol ve ;all of you have ideas for how to solvethese problems;the goal for the work teams is notnecessar i ly to reacn a consensus but todiscover new ways to solve theseproblems; how can this be done?

carefully listen to what others have tosay:feel free to respond in an open,spontaneous way (the aim is to have anexciting exchange of ideas;discuss ideas, do not debate thembecause we want to encourage divergent

53

points of view and keep ideas flowing,not cut them off;

d. the purpose of the work teams is notnecessarily to change anyone’s mind;

e. everyone has useful ideas andinformation--we’re here to combine allthis into new ideas.

Thus, the environment just described is aproblem solving mode. In this mode, consensusis not necessary. I will converge all of yourideas and recommendations after you leavehere. Then, I will mail what I converge fromyour recommendations to you for comment. Thisdivergence and convergence of ideas willbecome part of the proceedings.

So, what you do in these work teams is notthe end but only the beginning of the workthat needs to be done in the economic impactassessment of recreation and tourism.”

After they discussed a task, work teams wereg i v e n 1 to 1 l/2 hours to deve lop so lut ions .GrOUQS were asked to des ignate someone to j o tdown their recommendations on large sheets ofpaper and someone to be the spokesperson. Atthe end of the allotted time, the spokespersonof each team (four teams in all) took 5 to 10minutes to present the team’s recommendations tothe entire audience. A question and discussionperiod followed each team’s presentation of thes ix task so lut ions . The large sheets of papercontaining the recommendations remained postedaround the room for the duration of the meeting.

The same process was followed for each of thesix tasks, but the composition of the teams waschanged after each task to give each participantthe opportunity to interact with all others.All work team recommendations and discussionswere taped. Synthesis of the material containedin the tapes, teams notes (from the large sheetsof paper) and reviews of this synthesis bymeeting participants provide the results thatfo l low.

The method utilized in bringing together agrOUQ o f profess ionals and e l i c i t ing so lut ionsto specific problems was a creative problems o l v i n g QrOCeSS fashioned af ter Nol ler et a l .(1976, 1981) and Hare (1982).

RESULTS

Table 2 contains fu l l descr ipt ions o f the s ixtasks the work teams were asked to complete.The tasks are restated in abbreviated form inthis section along with specific recommendationsof the work teams for accomplishing the tasks.

Task 1: Short Cut Methods

Work teams were asked to describe “short cut”methods (methods other than I/O analysis) whichthe Forest Service and Corps of Engineers coulduse to obtain reasonable estimates of theeconomic impacts of recreation. The four work

Table 2.-- Tasks completed by work teams in the“Technical Meeting on Assessing SecondaryEconomic Impacts of Recreation and Tourism,”Michigan State University, 14-16 May, 1984.

1. Describe and provide reference to othermethods besides I/O analysis for assessing theeconomic impacts of recreation and tourism. Arethere one or two “quick and dirty” methods thatForest Service and Corps personnel could use toobtain a fairly reasonable estimate of suchimpacts?

2. Recommend appropriate ways for the ForestService and Corps of Engineers to define regionsand disaggregate sectors in I/O models toaccount for economic jmpacts of recreation andtourism. That is, list the sectors impacted andrecommend ways to separate them from commonlyused sectors. Also, describe the problemsassociated with measuring these impacts at thelocal vs . s tate vs . regional vs . nat ionall e v e l s . Recommend areas for future research.

3. Indicate the means (research, funding,administrative changes, etc.) by which theForest Service economic impact model, IMPLAN,may be modified to account for recreation andtourism impacts. Describe the cost/accuracytrade-offs of making such modifications.Recommend other I/O models that may be modifiedin this fashion.

4. Provide a list of variables that should beassessed and questions that should be added tonationwide federal estate recreation surveys(mailback and personal on-site interviews)relative to economic impacts. The goal here isto create consistency in data collection andanalyses that federal agencies routinely performto evaluate the economic impacts of recreationand tourism.

5. Describe the role of the prfvate sector inproviding data tha would satisfy the goals ofthe Forest Service and Corps of Engineers indetermining the economic impacts of recreationand tourism. To perform this task, you shouldanswer the following questions: Are data fromthe private sector necessary? What types ofdata? What strategies should be followed toobtain such data given that some of it isproprietary in nature?

6. Articulate the changes that need to be madeto the data collection and analysis proceduresof the Bureau of Economic Analysis to accountfor the economic impacts of recreation andtourjsm. Also, indicate the pros and cons of a“Census of Tourism ,” similar to the Census ofManufacturing or Agriculture.

54

teams described six such methods: responsec o e f f i c i e n t s , minimum requirements, use ofexist ing mult ip l iers , brainstorming, Delphiprocess , and s imilar s i tes .

Response Coefficienct Method

Response coe f f i c ients (RC’s) are the same astradit ional mult ip l iers except that RC’s arereported in d i f ferent units o f analys is to easeinterpretation (personal communications withAdam Rose, West Virginia University, 1984). Forexample , a new industry in a region may generate500 new jobs in direct employment in thatindustry. An I/O analysis may reveal anemployment multiplier of ‘2 indicating that thetotal employment impact (direct, indirect andinduced) of the new industry is 1,000 new jobs(500 x 2). Response coefficients merelytransform the 1,000 jobs figure into number ofjobs per $ expended on an activity. Thus, if1,000 new jobs were generated and $1 million ofgoods were produced, the RC would be 1,000jobs/$ million or 1 job/$l,OOO o u t p u t . N o t h i n ghas been done to the multiplier; the only changeoccurs in the manner in which the employmentimpact is reported. The RC transformation may beapplied to Type I, I I , o r I I I m u l t i p l i e r s .

The big advantage of RC’s over multipliers isre lat ive ease o f interpretat ion . Sincemult ip l ier are , in e f fect , part ia l der ivat ives ,they are sometimes ambiguous to interpret andprovide the opportunity for misleadingconclus ions . This is because it is notnecessarily true that sectors with highmultipliers have the highest impacts in aregion. For example, Rose et al. (1981), usingI/O analysis, derived multipliers to determinewhich alternative solar energy technology wouldhave the greatest employment impact on the Cityof Los Angeles. The employment multiplier forsolar energy was much higher than that ofweatherization. However, standardizingemployment impacts by translating them into RC’s(number of jobs created per million dollarsspent) revealed just the opposite finding: morejobs created by weatherization than by solarenergy. The authors explain this discrepancy bynoting that traditional employment multipliersfor solar energy are high partially becausesolar energy is expensive to produce and thusrequires more production than weatherization.However, the respending effects ofweatherization generate more employment than theproduct ion e f fec ts o f so lar energy .

A general mathematical expression of how tocal.culate regional impacts us ing RC’s is:

Total regional impact = total expenditures X RCfor income or employment; where RC = thed i r e c t , i n d i r e c t , and induced effect peramount spent in dollars.

The conventional multiplier is defined as totale f fects (d irect , indirect , and induced)throughout an economy divided by direct effectsin a given sector or the proportion by whicht o t a l e f f e c t s e x c e e d d i r e c t e f f e c t s . B yd e f i n i t i o n , t h e n , a large multiplier may result

because of a small denominator (direct effects).In other words, the multiplier may represent alarge multiple of a small base. Furthermore,the relationship between total effects anddirect effects will vary greatly among sectors,meaning that there is no standard for comparisonof mult ip l iers by sector . The RC is simply thenumerator of the conventional multipliere q u a t i o n ( t o t a l e f f e c t s ) . The RC permits astandard for comparison across sectors, removesthe ambiguity in mu1 tipliers, and maintains thebasic meaning of the multiplier concept.

Archer (1977) also provides evidence for andformulates the response coefficient concept.Instead of the term “response coefficient ,”however, Archer uses “normal multipliers” notingthat multipliers expressed as partialderivatives are valueless as planning toolswithout additional information which relatesendongenous income (or employment) to units ofexogenous spending.

The advantage of RC’s over traditionalmultipliers has already been discussed. Thereare also two major drawbacks to the RC method:(a) total expenditures must be collected asprimary data or taken from secondary sources,and (b) RC’s must be calculated by a centralresearch unit with access to a computer and I/Omodel. Both drawbacks also apply to I/Oanalysis in general . Overal l , the RC method isnot a separate impact assessment procedure atall, but a useful way of reporting impactsderived by traditional procedures. In l ight o fi ts abi l i ty to avoid mis leading results , thecalculation of RC’s may be worth the minimalextra e f for t required .

Minimum Requirements Method

Under this method, the analyst determines theminimum level of all services (not iustrecreat ion) for prototypica l areas or count iesneeded to sustain a local economv (res ident~populat ion) . That is , the analyst determinesthe economic base of an area. Any economicactivity above this minimum level would beattr ibuted to bas ic income (e .g . , expendituresby nonresidents). In this manner, an economicbase multiplier may be established (seeLeistritz and Murdock, 1981; Bendavid-Val, 1983;and Propst and Gavrilis, 1984 for furtherdiscussions of the derivation of economic basem u l t i p l i e r s ) .

A variant of this approach would involve a twostep procedure. First, plot certain economicindicators such as income or sales tax overtime. Second, compare the sales tax collectedin a month (March, say) when tourism is low withthose in a month during the peak tourist season.

Multiplier “Given” Method

In this method, expenditures by recreationistsmust be determined, but previously computedmult ip l iers are accepted . This method is bestexplained by two relationships:

1. Total Area Economic Impact = Mu1 tiplier(given) X Total Direct Expenditures;

55

2. Total Direct Expenditures = Expenditures perRecreation Visitor Day (EXP/RVD) X TotalRecreation Visitor Days (RVD) perA c t i v i t y .

Measures of participation other than RVD’s maybe equally valid. However, a method forseparating resident and nonresident expendituresper RVD must be devised because nonresidentpurchases represnet new money to the regionwhile resident purchases do not. I f necessary ,the var iable , EXP/RVD, may later be adjust byusing participation and supply (quantity andqual i ty o f fac i l i t ies ) as independent var iablesin regression analyses.

One word of caution is needed here. If notalready avai lable (usual ly the case) ,expenditure data mist be collected directly fromr e c r e a t i o n i s t s . This task is not for theunski l led or the fe int o f heart . Thus, in manycases , the mu1 tiplier “given” method may onlygive the appearance of being a short cutprocedure. Yet, the point is well taken thatinstead of spending a great deal of effort ondeveloping new mu1 tipliers, the planner shouldbe gathering quality expenditure data andenumerating the costs and direct benefits offuture developments.

“Brainstorming”

In th is procedure , experts , user groups ( i . e . ,the recreationists or tourists themselves), andbusiness leaders are assembled and asked toestimate participation, spending, and leakages.We do not mean to imply, however, that all ofthese groups should be assembled at one place atone time. This may not be feasible. Rather,these groups (and individuaJs in some cases) mayhave to be contacted at their convenience overan extended period.

Del.phi Process

The Delphi technique is a means of creting aconsensus of opinion concerning future likelyevents from the insights of experts rather thanfrom a theoretical body of knowledge. Moellerand Schafer (1983) fully describe the Delphitechnique, the steps in carrying out thetechnique, and the applications in recreation.

In terms of economic impacts, the Delphiprocess would involve having a group of expertspredic t the mult ip l ier e f fec ts o f current orfuture tourism and recreation developments in anarea. Moeller and Schafer state that the Delphitechnique can provide general estimates where noother techniques are available or appropriate.However, they warn that the process may requiremore effort (time and money) than the analystmight in i t ia l ly expect . Thus, it may not be ashort cut method in all cases.

Similar Sites

When economic impacts must be computed for acertain site. it would be extremely useful toknow the results of computations for simil,ar

s i tes e lsewhere . If another site weresuf f ic ient ly s imi lar , l i t t le or no addit ionalcomputation might be required. Unfortunately,few high quality analyses have been completed atpresent, but as experience is gained andanalyses are completed, it is recommended thatthey be catalogued for future use. This catalogwould contain surrogate multipliers and spendingprof i les with fu l l descr ipt ions o f the s i tecondit ions . The development of such a catalogwould be a major undertaking, but once done, itwould make future analyses quite easy.

Task 2: Defining Regions and DisaggregatingSectors

The teams of meeting participants were nextasked to make recommendations concerning how todefine regions and disaggregate sectors in I/Omodels to account for the economic impacts ofrecreation and tourism.

Defining Regions

Al l teams fe l t that , in i t ia l ly , the region o finterest should be the unit of the decisionmaker (members of Congress, governors, statel e g i s l a t o r s , etc.) or determined by the specificproblem being addressed. After these in i t ia lconsiderat ions , subreeions should be defined asspat ia l economic units ( count ies , SMSA’s, BEAu n i t s , etc.) according to the following generalscheme of increasing regional size:

1. Individual s i tes - -physical attr ibutes ( lake ,f o r e s t , e tc . ) where recreat ional act iv i t ies anddirect economic impacts occur.

2. Recreation focal area (trade area)--one ormore counties (SMSA’s, etc.) surrounding thesite or facility development which may beconsidered a “local” impact zone; likely to bethe source of most direct recreation employment.

3. Travel corridors--from the consumerresidence area to the site and define locationof impacts along the travel route.

4. Substate or mult is tate reg ions- -port ions o fseveral states or large group of countiessurrounding the site where both direct andindirect impacts occur; may also be defined asthe site’s market area by inspection ofo f v is i tat ion data .

5. Consumer residence areas--origins of ther e c r e a t i o n i s t s .

6 . Extended region-- national in scope; thesource of all goods imported into any of theabove 5 regions; capi ta l input to recreat ion ata given site likely to extend over the entirenat ion.

Once the market area (no. 4 above) isestablished, the internal boundaries may bedelineated by further analysis of population andvis i tat ion data . This hierarchy of regions isnot intended to resul t in concentr ic c i rc lesaround individual s i tes .

56

Maki (1985) and Stevens and Rose (1985)describe these regions and define their datarequirements in more detail. In general, anyattempts at regional delineation and aggregationmust consider the additivity problem. That is,there are sometimes differences between impactsderived from summing over numerous small areasversus an overall large area impact (the wholemay not be the sum of its parts). This isprimarily a methodological problem which may beovercome by clearly defining export expendituresas being outside the region of which thecounties surrounding the site are a part.

Sector Disaggregation

To discuss this topic properly, an importantd i s t i n c t i o n must be made between intermediateand final purchases. Final purchases are thesectors in which consumer expenditures occur( e . g . , tourist spending on food and beverages,angler purchase o f f i shing bait ) . In theaccounting system of an I/O table, finalpurchases are enumerated. in the final demandvector . Intermediate purchases occur when firmswithin sectors that produce recreation goods andservices buy from or sell to each other (e.g.,canoe manufacturer’s sales to a canoe livery).Intermediate purchases are represented in theinterindustry matrix of an I/O table.

Intermediate purchases.--For intermediatepurchases, meeting participants agreed that theexisting level of aggregation in RIMS, thenational 500-sector I/O model, was appropriatefor a l l but the reta i l , wholesale , and services e c t o r s . For example , there is already adetailed breakdown of manufacturing at the4-dig i t SIC level . Since capital expendituresfor recreation or tourism go into manufacturing,suf f i c ient d isaggregat ion exists . Such is notthe case for the retail, wholesale, and services e c t o r s . For example , marinas do not have aseparate 4-digit code and are completelydominated and subsumed by the boating dealerss e c t o r . Certain manufacturing sectors havetheir problems as well. This is espec ia l ly truefor boat bui ld ing ( i .e . , smal l boats) which ishidden within the ship bui ld ing sector . Yet ,boat building and marinas are important elementsin the recreation/tourism industry and havedi f ferent input structures that ship bui ld ingand boat dealers per se. An example of theaggregation problem in the service sector iscommercial amusements. This sector is so highlyaggregated that it contains everything frombowling a l leys to ski l i f ts . In recreation andtourism, retailing and services are majorcomponents of the economic activity of manylocal areas . Thus, being wrong in these sectorscan create more errors in multiplier developmentthan would be the case for large metropolitanareas or other areas with diverse economies.

REIS, the national I/O model developed by theRegional Science Research Institute (Stevens eta l . , 1975) overcomes some of these aggregationproblems by providing 34 wholesale and 40 retailsectors . The 40 reta i l sectors inc lude RV’s(recreation vehicles like motor homes andcamping trailers) and most of the categories

that appear in the Census of Retail Trade.However, REIS does not solve the aggregationproblems in the service sector.

In light of the above discussion, workteams recommended that existing I/O categoriesbe used except for wholesale, retail, andserv ice sec tors . These sectors should bedisaggregated further into 2-digit SIC sectors,perhaps using REIS as a starting point.

Final Purchases.--One recommendation was to usethe 84 consumer expenditure categories from theNational Income and Product Accounts (NIPA),differentiating between local and nonlocalexpenditures for each sector for each recreationa c t i v i t y . These 84 categories would become thesectors in the final demand vector. Under thisrecommendation, the NIPA categories would alsoserve as the basis for gross private capitalformation, government expenditures, and exports.For gross private capital formation(construction of new facilities), one would needto distinguish between private,recreation-related construction and otherconstruct ion . For federal , s tate , and loca lgovernment spending, it would be necessary tod i f ferent iate between recreat ion-re latedspending (both construction, and operations andmaintenance activities) and spending for otherpurposes. For exports, expenditures o f v is i torsfrom outside the region of concern would have tobe separated from the expenditures of otherv i s i t o r s . The 84 consumer expenditurecategor ies f rom NIPA can also be used totransform direct expenditures into I/Ocategor ies . Such a transformation becomes amovement from purchaser to producer prices.

There are other ways of transforming oneexpenditure system to another. One way is tosurvey visitors to obtain expenditureinformation and then transform the expendituresinto I/O categories through the use of theSurvey of Current Business “CommodityComposition of Personal ConsumptionExpenditures .” This is the procedure currentlybeing followed by the Forest Service’s IMPLANsystem.

Both the NIPA and the Survey of CurrentBusiness approaches call for the collection ofexpenditure data d irect ly f rom v is i tors . Analternative to primary data collection would beto pay someone to identify and publish an indexof ex is t ing sources o f v is i tor expenditure data .The point is that there are databases andpublications not widely circulated that containexpenditure information necessary to estimatethe economic impacts of recreation and tourism.Assembly of these sources might sometimespreclude the need to collect primary data andwould be an important contribution.Nonetheless, noncomparability of many databaseswould likely be so troublesome that only generalexpenditure profiles could be published in suchan index. In terms of accuracy, primary datacollection holds a strong advantage overprocedures involving secondary data.

57

Task 3: Variables to be Assessed

To complete this task, work teams provided a listof variables that should be measured in nationwidesurveys of the economic impacts of recreation andtourism. There was general agreement among workteams on the variables that should be assessed.Differences were based on ways of categorizing ororganizing the variables. One way to organize thevariables is to divide them into those that may beasked of an entire sample of visitors and those thatmay be asked of a subsample on-site or at home aftera tr ip :

1. General variables to be assessed of entiresample (necessary for visitor segmentationpurposes):

- -or ig in and dest inat ion- -purpose o f t r ip--type of accomodations where staying overnight- - length o f s tay--mode of transportation--phone number and address to recontact

(recontact ing cr i t i ca l to obta ining accurateassessment of trip home expenses)

- -day tr ip vs . mult iday, s ingle dest inat ion tr ipv s . mult iday, mult ip le dest inat ion tr ip

--number in party and party composition (family,f r i e n d s , e t c . )

--equipment type (because, for example, thosewith RV’s may have different expenditure patternsthan those in family auto)

--demographics--some expenditure data according to distance

from site (most useful would be food and beverage,lodging, fees and charges, gasoline, equipment):exercise caution with equipment expenditures becausesome equipment purchases would be made regardless ofex istence o f a part icular s i te

2. Specific expenditure data to collect from asubsample o f v is i tors ( co l lect according to d istancefrom s i te ) ; list not intended to be comprehensive(may opt to use some subset of the 84 NIPAexpendi ture categor ies ) :

--public accomodations--eating and dining out- - g r o c e r i e s- - l i q u o r s t o r e s- -gaso l ine and re lated serv ices- - inc idental sport ing goods (bai t , c lothing ,

e t c . )- - car rental- -boat rental--equipment rental- -publ ic transportat ion- -personal serv ices- - p r o f e s s i o n a l s e r v i c e s- - h o s p i t a l s e r v i c e s- - f inance serv ices--camping fees- - l i c e n s e s- -out f i t ters and guides--marinas--mdvi es--amusements- -o thers

Depending on the objectives of a particular study,there are other ways of classifying these variables.For example , agencies may want to measure all theabove variables in a given sample. Subsampling fordetailed expenditure data is meant to minimizesurvey cost and respondent burden; its applicabilitydepends on the goal of the survey. The primary goalof a nationwide expenditure survey might be todevelop a general model from a sample of visitors atd i f f e r e n t s i t e s . In that case, the samples wouldtoo small to determine spending patterns at anygiven site. Through a relatively small increase ine f f o r t , the national sample could be segmented bygeographic region and other variables as listed in(1) above. The national spending patterns could thenbe applied to any site in the U.S. given someknowledge o f that s i te ’ s v is i tat ion character is t i cs( n u m b e r s o f v i s i t o r s , o r i g i n , a c t i v i t i e s , e t c . ) .Thus, a fairly large sample to obtain the datal isted in (1 ) above plus a re lat ively smal lsubsample to obtain the detailed expenditureinformation listed in (2) would meet the goal ofestablishing a nationwide recreation expendituredatabase.

The use to which the survey data will be put mustbe clearly specified before a methodology or surveyinstrument can be properly developed. For example ,do the potential users want measures of a few keyvariables from a large sample in order to reducesampling errors or do they want detailed expendituredata from a relatively small sample? The more detailthat is required of respondents, the greater thelikl ihood of increased sampling error.

Due to time constraints, discussion ofmethodolog ica l deta i ls ( i . e . , spec i f i c wording o fsurvey i terns, sampling procedures) was superficial .Nonetheless, an important point for considerationwas that listing the variables should notnecessar i ly be the f i rst s tep in co l lect ing qual i tyexpenditure data. Instead, the f i rst s teps shouldbe the specification of goals as stated above andthe development of a data collection methodology.This methodology will then point out the keyvariables to be measured and specific survey itemsw i l l f o l l o w .

There was a divergence of opinion as to the mostappropriate methodology to employ. Recommendationsincluded the following:

1. Personal, on-s i te interv iews to increaseaccuracy by reducing recall -problems.

2. Pay people to keep an expenditure diary oftheir trip as is done in states like Massachusetts.

3. Have respondents keep a log of theirexpenditures dur ing a l l t r ips for 1 year .

4. Conduct mailback surveys especially for thepurpose of obtaining estimates of trip h o m eexpenses.

Since there was no consensus concerning the mostappropriate method, the suggestion was made toemploy a variety of methods and allow the results soderived to serve as checks of reliability andv a l i d i t y .

58

There was consensus on two important points:

1 . Federal agenc ies , s tates , and pr ivate interestsshould work together (pool resources) to develop amethodology and collect quality expenditure data ona national basis .

7_. A proper database will attract researchers todo the needed analyses because quality databases ofthis nature are difficult and expensive to obtain.

Task 4: Yodif ying IMPLAN--_

Work teams were asked to recommend modificationsin the Forest Service economic impact model, IMPLAN,to account explicitly for recreation and tourismimpacts. The general recommendation was to taiJorIMPLAN to meet recreation and tourism needs.Specific ways to perform such tailoring follow.

One recommendation was to encourage the Bureau ofEconomic Analys is (BEA) to collect data moreappropriate to recreation and tourism therehy makingthe national I/O model more realistic in terms ofthis industry . Since IMPLAN is a subset of thenat iona!. model , necessary improvements in IMPLANwoul d f 011 ow sui t . The results presented under Task6 to follow provide further detail on this point.

Another recommendation was to develop or improvecertain intermediate purchase sectors in IMPLANrelative to recreation and tourism. Touristexpenditures currently are not well represented inthe sectoring of IMPLAN. The key sectors related tothe forest and grazing industries have already beeni d e n t i f i e d . The same could be done for recreationby specifying the appropriate retailing and servicesectors (see also previous discussion under Task 2 -d isaggregat ion o f sectors ) . Much o f th isspec i f i cat ion o f recreat ion sectors could he doneimmediately. Other tasks, such as placing thehoating industry in the model correctly, could takemuch 1 anger .

In terms of final demand modifications, it isagain necessary to differentiate expendituresspec i f i c to recreat ion and tour ism, inc ludingprivate capital formation and governmentexpenditures. In other words, retail trade andservices should he di saggregated in the final demandsectors . 4s a start ing point , th is d isaggregat ionmight he based on NIPA categories, which are closerto consumer spending categories than those currentlyin IMPLAN. 4n a l ternat ive for d isaggregat ion is toestabl ish standard tourist expenditure vectors on atotal purchase basis (i .e., for now, do not worry.about where purchases are made or by whom butestablish standard vectors on a per person per dayb a s i s b y a c t i v i t y ) . The next step would be toregionalize the tour is t vectors . IMPLAN current1 yallows this without additional work hy usingimp1 ici t regional response coefficients generated hythe suppl y-demand pool ing approach. Possibleimprovements would he to adjust these imp1 ici tresponse coefficients by regional experts or byregress ion est imates o f these coe f f i c ients us ingaddit ional exploratory var iables .

The purpose of disaggregating final demand sectorsis to transform final demand categories into

intermediate purchase categories (usual.ly SICcodes ) . The NIPA approach would require respondentsto allocate their purchases into categories that arealready very similar to many I/O sectors but may nothe specific to recreation and tourism. The touristvector approach would require respondents to statehow much they spent in various categories specificto recreation and tourism. This latter approach hasthe advantage of couching expenditures in termsrelative to the respondent, not the I/O model. Theanalyst would still be required to transformexpenditures, v ia NIPA or o ther categor ies , into I /Os e c t o r s . These transformations could he devel.opedbased on several case studies employing theprocedures recommended in Task 3. Once theexpenditure vectors and transformations ares p e c i f i e d , it would not he necessary to co1 lect newexpenditure data for every situation. Instead, onecould predict spending based on data previouslycollected and gather new data only on visitor daysof use by act iv i ty . Whichever approach is used, itwi l l s t i l l be necessary to d i f ferent iate the reg ionof impact according to resident versus nonresidentspending (i .e., have separate vectors for residentsand nonresidents.

Once the sectors are disaggregated or specified,IMPLAN’s output relative to recreation can andshould be simplified. That is, the full model may bereduced to just those sectors impacted byrecreat ion . This i s espec ia l ly important for theIMPLAN user because confusion with irrelevantsectors i s avo ided .

One of the most serious gaps in the currentcapability of IMPLAN relative to recreation is inthe payments sector. That is, there is nothing nowin IMPLAN to specify the location of employees, orowners of capital. Overal 1, this problem is relatedto the lack of adequate data on income generatedversus income retained in region (i .e., in thepayments sector). This problem is important inrecreation and tourism because of the seasonality ofemployment and husiness ownership . For exampl e,how much do college student employees spend in anarea? How much do they save for, say, tuition spente lsewhere? This i s a cr i t i ca l i ssue hecause theinduced portion of the income multiplier comes fromincome respent in the region. 1Jy overestimatingincome retained in a region, the income multiplieris prohahly biased upward. IJsual methods ofadjusting for income generated versus incomeretained in a region (e.g., residence adjustmentsfrom BEA, commutation data from the Census) areprobably inadequate due to the transient seasonalemployees in recreation and tourism sectors.

Task 5: Private Sector Data

This task required the work teams to describe thepotent ia l ro le o f the pr ivate sector in prov id ingdata useful in determining the economic impacts ofrecreation and tourism. Work teams suggested types,sources , and means of obtaining such data at thelocal ( individual f i rm) , s tate , and nat ional levels .An initial question raised was whether data from theprivate sector were even necessary. The response wasthat these data were useful at least as a supplementand method of cross-checking puhl ic expendituredata. Also , a closer working relationship with the

59

private sector may reveal more efficient economicimpact assessment methods than are in current usagein the puhl ic sector.

The types of private sector data most needed ineconomic impact assessment of recreation and tourismare:1. total sa les2. employment3. payroll4. tour ist c l iente le sa les as a percentage o f tota l

sal es5. percent of business purchases locally versus

outside an area6. tourist expenditures (e.g., in private campground

stores )7. industry inventories (e.g., when and what are the

s izes o f boat inventor ies? )

In regards to the last data type, the point wasraised that industry inventories are often estimatedin I/O t a b l e s , a practice that may lead toinaccurate resul ts. During downward cycl es In anindustry, inventories can accumulate and cushion theresponse of increases in an activity. For exampl e,an increase in boating production may be misleadingif there is no accounting for inventories. That is,a 15% increase in hoating activity may result in a5-10X increase in production because of accumulationof inventory. The difference may be insignificanti f pro jec t ions are for a re lat ive ly short per iod o ftime (5 years, say).

Important sources of private sector data include:

1 . Industry assoc iat ions represent ing RV’s, skiing,hoats, marinas, sports equipment, lodging, sportf i s h i n g , and so on; most of these possess visitorand capital expenditure data.

3. The American Recreation Coalition (perhaps as alead into the various industry associations), theTravel. and Tourism Research Association’s NationalData Center, the U.S. Travel Data Center, chambersof commerce , ut i l i t ies , transportat ion agencies ,American Automobile Association -- all may at leastprovide some purchaser characteristic data;

3. A new Bureau of Lahor Statistics quarterly surveywi l l inc lude a sect ion on le isure /recreat ionpurchases, hut it is uncertain when these data willhe avai lable .

4. Special industry studies (may be proprietary innature) .

5. Consul ting firms that conduct market surveys.

6. Sal es Management Magazine’s annual survey ofbuying power.

7. New York Stock Exchange (NYSE) shareowner survey.The usefulness of the NYSE data wou3d he to trackwhere profits go and to include an incomedistribution analysis in IMPLAN (who wins anti wholoses within and across regions). That is, whatincome is generated within versus what flows out ofa region? Is incomeincome in the hands of a few or spreadout among many? Often much of the income that issT,enerated in a region flows away and is therefore no

longer a benefit. The NYSE shareowner surveyprovides data on which to estimate the origin sectorand recipient income class for one portion ofproprietary income -- dividends and payments. Thedistribution of the other major incomecomponent--wages and salaries--can he ohtained hyref in ing “manpower requirements matrices” publishedby the Il.‘?. Bureau of Labor Statistics and Bureau ofEconomic Analysis (see Rose, et al ., 1982). Theincome mu1 tiplier in IMPLAN is for total income.This may be an inaccurate indicator of well-being ina region because wages generated may remain whiledividends and royalties may leak out. Thus, there isoften the need to disaggregate income intoappropr iate categor ies be fore a mult ip l ier i sappl ied . In the example where wages remain but allother forms of income flow out of a region, theproper analysis would he to apply the incomemultiplier to household income alone.

In order to obtain private sector data, twoconsiderations are mandatory. F i r s t , there must hean assurance of confidentiality. Second, mutualbenefits must he identified (i.e., what are theadvantages to individual firms?). Because of thesetwo important considerations, a nongovernmental dataco l lector (univers i ty or consult ing f i rm) wasrecommended.

Task 6: Changes Needed in BEA System

In the f inal task, work teams recommended changesneeded in data collection and analysis procedures ofthe Bureau of Economic Analysis (BEA). Team membersident i f ied f ive needed modi f i cat ions .

F i r s t , it was recommended that tourism andrecreat ion pro fess ionals , not BEA staf f ,disaggregate the tourism/recreation industry intofiner sectors as per the suggestions made in Tasks 2and 4. One of these earl.ier suggestions was tod e v e l o p 2-digit c lass i f i cat ions for services andretail trade instead of the current combinations ofcategor ies . For example , efforts should be aimedat : (I) aggregating the no longer appropriatemanufactur ing sectors ; (2) identifying newmanufactur ing sectors (genet ics , robotics, etc.) ;(3) disaggregating some manufacturing sectors (e .g . ,hoat manufacturing from ship building); and (4)disaggregating (eliminating the noise) theamusements sector into major amusement categories,such as major amusement centers, marinas, ski areas,gal f courses, t enni s compl exes , and f i tness centers .

Second, it was recommended that the BEA presentemployment data in full-time equivalents (FTE’s)instead of the current practice of mixing full-timeand part-time employment. This change is especiallycr i t i ca l to the tour ism industry hecause of the highdegree of part-time employment in many sectors.

Thi rd there is need for more consistency indefini tfon of sectors and employment categoriesamong the BEA system, County Business Patterns, andother national models. The current BEA I/O tahl e isi.nconsi stent with County Business Patterns becausedi f ferent ru les are used to categor ize certa inhusinesses and these rules <are not clearlya r t i c u l a t e d . Furthermore, other national models(e.g., REIS) else d i f f e r e n t c l a s s i f i c a t i o n r u l e s t h a n

6060

either County Business Patterns or the BEA. Forexample, County Business Patterns uses unemploymentinsurance figures to develop employment data. As ar e s u l t , many small firms not covered by unemploymentinsurance are omitted in the analysis. REIS, us ingdifferent rules and data, is able to include verysmall firms in its analysis; however, these rulesand data are not clearly specified.

Fourth, the BEA should conduct a periodic “Censusof Spec ia l Serv ices” adjusting for seasonalfluctuations in such industries as tourism andrecreat ion . This census is needed because thecurrent Census of Services is for selected servicesonly , not special ones like tourism. Such a census,especially done in conjunction with transportationand manufacturing censuses, would overcome many ofthe problems associated with estimating the economicimpacts of recreation and tourism.

The fifth recommendation concerns betterinteragency cooperation. The BEA should berepresented on tour ist assoc iat ion stat is t i ca lcommittees and research branches of recreationmanagement agencies.

SUMMARY AND CONCLUSIONS

The results represent a convergence of sometimesdiverse opinions from regional economics expertsregarding six key issues which any agency mustface when involved in assessing the economicimpacts of recreation and tourism. Work teamparticipants were encouraged to present divergentpoints of view on the assigned tasks. Beforepreparing this paper’s resul t, we wrote a draft ofwork team recommendations and mailed it toparticipants for further comment. We thenintegrated the participants’ comments with anextensive literature review concurrentlyperformed under contract with the U.S. ForestService and Corps of Engineers (Propst andG a v r i l i s , 1984). Our paper, therefore, is theproduct of this synthesis of professionalexperience and literature.

There are several important conclusions to bedrawn from the results presented herein. F i r s t ,“short cut” methods (methods other than I/Oanalysis) are appropriate only when theexpertise or resources (i .e. , computer andaccess to an I/O model ) for performing an T/Oanalysis are not availahle. These so-calledshort cut methods are likely not “short” in thesense of time or money. Four of the six methodsdiscussed (response coefficient , minimumrequi rements, mu1 tiplier given, and similarsites) require the expertise of a regionaleconomist and/or the collection and analysis ofprimary or secondary visitor expenditure data.The two remaining methods (hrainstorming andDelphi) require a relatively large investment intime in contacting and ohtaining the appropriateinformation from individuals and groups. We donot mean that these methods should never heempl oyed. We merely wish to point out that theterm “short cut” may be mi sl eadi ng.

4 second conclusion is that I/O analysisrepresents the most rigorous, accurate method OF

economic impact assessment. Thi s concl us i onemanates from the long-term experience of themeeting participants with I/O analysis, theavailability of I/O models, and the existence ofcomputers capable of handling the data andmat hemat i cal requi rements of such models .Yowever, the rigor of I/O analysis can hecome animportant drawback. That is , a great deal ofexperience and training is required beforeanalysts can understand how to perform an I/Oanalys is and interpret i ts results proper ly .The jargon that comes as baggage with any bodyof knowledge is particularly voluminous andconfusing . Thus, to the uninit iated , I /Oanalysis may appear to he a black box withvolumes of data entering one side and results(usual ly mu1 ipliers) exiting the other. Thiscommunication prohlem may be overcome to someextent by empJoying one or more of thealternative methods discussed above. Thesealternative methods may be less rigorous thanI/O analysis but more readily comprehended bydecisionmakers untrained in quantitativeeconomic analysis. These alternative methodsmay al so serve as a useful check on the resultsof an I/O analysis.

Despite their complexity, I/O analysis givesthe most complete picture of the sophisticatedinteractions in an economy and they accuratelyprovide much of the information being requestedby decisionmakers (impacts on income, johs,e t c . ) . However, a third conclusion to be drawnfrom the results is that the Forest Service’sIMPLAN, or any other I/O model, requires certainkey modifications to estimate precisely andaccurately the economic impacts of recreationand t ouri sm. The f i rst modi f i cat ion required isthe disaggregation of the retail and who1 esal etrade and services sectors into categories thataccurately reflect the sellers and the producersof recreation goods and services. Second, thepayments sector should he modified to reflectthe amount of income generated versus the amountretained in a region. Thi rd, instead ofe x p r e s s i n g muJ tipliers as partial derivatives(the usual procedure), analysts should expressthem in units that relate endogenous income oremployment to exogenous spending, such as totalemployment generated per $1 ,000 of lodgingexpenditures in a region. The result isotherwise known as a response coefficient (Rosee t . al ., 1981) or normal muJ tip1 ier (Archer,1977). The fourth modification is to develop amatrix of transformation indices that convertf inal purchases (v is i tor expenditures ) intoproducer prices. Either National Income andProduct Accounts or Survey of Current Businessdata may be used to develop this matrix.Visitor expenditure data may be obtaineddirectly from consumers or from a compendium ofresults from previous studies. Once vi si torexpenditure profiles and a transformation matrixare spec i f ied , it will no longer he necessary tocollect new spending data for every site ors i t u a t i o n . F i f t h , for any economic impactmet hod, not just I/O analys is , res ident spendingwithin a region must clear1 y be separated fromnonresident spending wi thin the same region.Such separation requires careful de1 ineation ofregional houndaries according to study objectives.

61

Regardless of the method chosen, perhaps theoverriding concern of all meeting participantswas the estahl ishment of a rel iahle and validvisitor expenditure database from a nationwidesample. Input/output specialists can make theaforementioned structural changes in IMPLAN orother models relative1 y easily. However, theresul ts obtained (e.g., mulitipliers) w i l l o n l ybe as accurate as the the expenditure data enteringthe models. Currently, this high qual i tyexpenditure database is lacking. Itsdevelopment should proceed in the followingsequence : (1) determine objectives and uses ofdata, (2 ) estahl ish detai led methodology , (3)specify variables to be measured, (4) specifythe measurement instrument and items, (5)pretest the procedure and instrument, and (6)revise methodology and instrument. Agreement onmethodology is lacking at this time due to theabsence of data comparing the rel iahility andval idity of such methods as personal interviews,mai 1 back surveys, and expendi ture diaries.Because Because opinions on appropriate methods vary, weconclude that, wherever possihl e, a variety ofprocedures should he used on subsamples of thepopulation and results compared and madeavailable to the academic community forcriticial rev iew. The work teams developed ani n i t i a l l i s t o f v a r i a b l e s ( b o t h v i s i t o rsegmentation and expenditure variables) forconsiderat ion .

llsing the ahove model for devel aping a qual i tyexpenditure data hase, the U.S. Forest Service,Corps of Engineers, and National Park Servicehave launched a nationwide effort at collectingsuch data. The Public Area Recreation VisitorSurvey (PARVS) will he conducted in 1985 athundreds of federal resource agency and statepark sites across the U.S. A comhi nat i onon-site and mailhack survey, the PARVS has as am a j o r objective the co l lec t ion o f deta i led tr ipand annual expenditures for recreation andtourism. The end product will be the on1 ynational expenditure data hase of its kind. Astrong recommendation made by meetingparticipants was an interagency cooperativee f fort (poo l ing o f resources ) to estahl ish theheretofore missing national expendi turedatahase. PARVS is such an effort. However,there is no implied continuity to the PARVS.That is there is no guarantee that the s,amed a t a wili he collected 5 or 10 years f rom now.Therefore, another recommendation of the meetingparticipants is that the Bureau of EconomicAnalysis establish a periodic “Census of SpecialServices” with recreation and tourism as one ofthe special services high1 ighted.

A fifth conclusion from the results presentedherein is related to the usefulness andavni lahi l i ty o f data f rom the pr ivate sector .Such data are needed to check some of theestimates derived from public data (Census,PARVS, etc.) and to fill in some large gaps inpublic datahases (e.g., percentage o f tour is tsales in selected firms, inventory data, andnonproprietary i “come generated and retained i i nn

a region). Furthermore, such data such data mdstmdst he he

collected in a collected in a highlyhighly professional manner with professional manner with

cxtrempcxtremp cat-pcat-p given given tntn confidentiality confidentiality ,3;14,3;14

henefits to private interests of making thisinformation avai lable .

A final conclusion concerns the method forobtaining the results-- the small workteam/creative problem solving format. The methodseems to hold much promise for technologytransfer . We hrought together a group ofregional economics professional s representingwe1 1 over a hundred years of training andexperience and were able to apply their talentsto a specific problem area in anonconfrontational manner. Numerousinvestigations of the economic impacts ofspecific recreation and tourism events anddevelopments have been conducted. Never hef ore,however, however, has there been a concerted effort ataccurately assessing the secondary economicimpacts on a nationwide basis. The approach usedtransferred vital technology to federal resourcemanagement agencies, providing the basis forcreating a national expenditure database forrecreation and tourism. At-1 east two of theseagencies have plans to transfer this technologyone step further to their field planningo f f i c e s . The method also allowed those withless experience in economic impact assessment tolearn much from those with more. Thus, as atraining too l , the method was also successful.

There are always improvements that can he madein any methodological approach. We feel onlyminor improvements are needed in the protocoland format of the work team portion of themeeting. However, we feel we should have heenmore diligent in obtaining comments on the workteam recommendations from the participants afterthe meeting ended. ln sum, we recommend themethod followed herein as an efficient approachfor transferring knowledge from one field toanot her.

LITERATURE CITED

Archer, B.M. Tourism mu1 tip1 iers: thes t a t e - o f - t h e - a r t . Bangor Occasional Papers inEconomics No. 1 1 .1 1 . Bangor, Wal es: University ofWales Press; 1977. 44 p p .p p .

Bendavi d-Val , A. Regional and 1 ocal economicanalysis for practitioners. New York: Praeger;1983. 195 p p .p p .

Dwyer, .J.F., J.R. Kel 1 1 y, and M .n. Bowes. Improvedprocedures for va luat ion o f the contribution ofrecreat ion to nati 0naI economic devel opmcnt .

Qesearch Report No. 178.178. Urbana-Champaign,Urbana-Champaign, TL.: TL.:

'Jniv.'Jniv. of Illinois of Illinois WaterWater Resources Resources Cer?tcr:Cer?tcr: 1977.1977.

718718 pp. pp.

Cnrtner,Cnrtner, V.C.V.C. and D.F. Holecek. The economic impactof a short-term tourism industry e x p o s i t i o n .e x p o s i t i o n .

QescarchQescarch Report Report 436.436. East East I_ans'ng,I_ans'ng, MI: Michigan MI: Michigan

State Univ. State Univ. Agric . Exp. Stn. ; 1982. R pp. pp.

Yare,Yare, P .A. Creat iv i ty in small in small groilps.groilps. Beverly Beverly

Hills, CA:Hills, CA: SageSage Publications;Publications; 1982. 199 pp. 1982. 199 pp.

62

L e i s t r i t z , F.L. and S.H. Murdock. The socioeconomicimpact of resource development: methods forassessment. Boulder, CO: Westview Press; 1981.289 pp.

Maki , W.R. Measuring supply side economic impactsof tourism and recreation industries. In:Assessing the economic impacts of recreation andtourism. 1984 May 14-16; East Lansing, MI:Michigan State Univ. Published in this report;1985.

Marino, M.L. and D.E. Chappelle. Lodging andrestaurant establishment spending patterns innorthwest lower Michigan. Research Report 346.East Lansing, MI: Michigan State Univ. Agric.Exp. Stn.; 1978. 32 pp.

Moeller, G.H. and E.L. Shafer. The use and misuseof Del phi forecasting. In : Lieber , S .R. ;Fesenmaier, D.R., eds. Recreation planning andmanagement. State College, PA: VenturePubl ishing; 1983:96-104:

Nol ler , R .B. , S .J . Parnes ,Creative actionbook. NewSons ; 1976. 399 pp.

Koberg, D. and J. Bagnall.

and A.M. Biondi.York: Charles Scrihner’s

The universaltrave ler . Revised Ed. Los Altos, CA: WilliamKaufmann, Inc.; 1981. 128 pp.

Propst, D.B. and D.G. Gavrilis. Evaluation ofmethods and models for determining the economicimpacts of public recreation services andf a c i l i t i e s : final report. East Lansing, MI:Michigan State University Department of Park andRecreation Resources; 1984. 77 pp.

Rose , A . , D. Kolk, andand urban employmentc i ty o f Los Angel.es.1981.

M. Brady. Energy developmentcreat ion : the case of theE n e r g y 6(1(I): 1041-1052;

, B . N a k a y a m a a n d B . S t e v e n s . M o d e r n e n e r g yregion development and income distribution.Journal of Environmental Economics and Management9(2): 149-164; 1982.

Stevens , B.H., G.L. Treyz, D.J. Ehrlich, and J .R.Bower. State input-output models fortransportation impact analysis. Discussion PaperSeries No. 128. Amherst, MA: Regional ScienceResearch Institute; 1981. 23 pp.

, and A. Rose. Regional input-outputmethods for tourism impact analysis. In:Assessing the economic impacts of recreation andtourism. 1984 May 14-16; East Lansing, MI:Michigan State Univ. Published in thisreport; 1985.

Stynes, D.J. and D.F. Holecek. Michigan Great Lakesrecreat ional boat ing : a synthesis of currentinformation. Michigan Sea Grant ReportMICHU-SG-82-203. Ann Arbor, MI: Michigan SeaGrant Publications; 1982. 86pp.

Wal sh, R.G. The intelligent consumers and managersguide to recreation economic decisions.Preliminary bound draft for classroom use. Ft.Collins, CO: Colorado State Univ. Department ofAgricultural and Natural Resource Economics;1984. 609 pp.

63

APPENDIXAPPENDIX

Participants in the conference and workshop onParticipants in the conference and workshop on"Assessing the Economic Impacts of Recreation and"Assessing the Economic Impacts of Recreation andTourism,"Tourism," May 14-16, 1984.May 14-16, 1984. Michigan State Univer-Michigan State Univer-sity, East Lansing.sity, East Lansing.

Dr. Greg Dr. Greg AlwardAlwardLand Management Planning UnitLand Management Planning UnitUSDA Forest SeriveUSDA Forest SeriveFort Collins, Fort Collins, COCO 80526 80526

Dr. Robert C. BushnellDr. Robert C. BushnellDep. of Finance Dep. of Finance RR Business Economics Business EconomicsWayne State UniversityWayne State UniversityDetroit, Detroit, MI 48202

Dr. Daniel E. ChappelleDr. Daniel E. ChappelleDep. of Resource Development Dep. of Resource Development RR Forestry ForestryMichigan State UniversityMichigan State UniversityEast Lansing, MI 48824East Lansing, MI 48824

Dr. Dr. H.H. Ken Cordell Ken CordellForest Sciences LaboratoryForest Sciences LaboratoryUSDA Forest ServiceUSDA Forest ServiceAthens, GA 30602Athens, GA 30602

Mr. Dimitris Mr. Dimitris G.G. Gavrilis GavrilisDep. of Park Dep. of Park && Recreation Resources Recreation ResourcesMichigan State UniversityMichigan State UniversityEast Lansing, East Lansing, MI 48824

Mr. William J. HansenMr. William J. HansenWaterways Experiment StationWaterways Experiment StationCorps of EngineersCorps of EngineersVicksburg, MS 39180Vicksburg, MS 39180

Dr. Donald HolecekDr. Donald HolecekDep. of Park and Recreation ResourcesDep. of Park and Recreation ResourcesMichigan State UniversityMichigan State UniversityEast Lansing, MI 48824East Lansing, MI 48824

Dr. Matthew HyleDr. Matthew HyleDep. of Finance Dep. of Finance RR Business Economics Business EconomicsWayne State UniversityWayne State UniversityDetroit, Detroit, MI 48202

Dr. Jay A. LeitchDr. Jay A. LeitchDep.Dep. of Agricultural Economicsof Agricultural EconomicsNorth Dakota State UniversityNorth Dakota State UniversityFargo, ND Fargo, ND 581035581035

Dr. Wilhur R. MakiDr. Wilhur R. MakiDep. of Agriculture Dep. of Agriculture RR Applied Economics Applied EconomicsUniversity of MinnesotaUniversity of MinnesotaSt. Paul,St. Paul, MN MN 551118551118

Dr. Charles Palmer (deceased)Dr. Charles Palmer (deceased)Resources Planning AssessmentResources Planning AssessmentUSDA Forest ServiceUSDA Forest ServiceDenver, CODenver, CO

Dr. Dennis B. PropstDr. Dennis B. PropstDep.Dep. of Park of Park RR Recreation Resources Recreation ResourcesMichigan State UniversityMichigan State UniversityEast Lansing, East Lansing, MI 48824

Dr. Adam RoseDr. Adam RoseCollege of Mineral and Energy ResourcesCollege of Mineral and Energy ResourcesWest Virginia UniversityWest Virginia UniversityMorgantown, WV 26506Morgantown, WV 26506

Dr. William A. Dr. William A. SchafferSchafferDep. of Industrial ManagementDep. of Industrial ManagementGeorgia Institute of TechnologyGeorgia Institute of TechnologyAtlanta, GA 30332Atlanta, GA 30332

Mr. Eric SivertsMr. Eric SivertsLand Management Planning UnitLand Management Planning UnitUSDA Forest ServiceUSDA Forest ServiceFort Collins, CO 80526Fort Collins, CO 80526

Dr. David SnepengerDr. David SnepengerWaterways Experiment StationWaterways Experiment StationCorps of EngineersCorps of EngineersVicksburg, MSVicksburg, MS 3918039180

Dr. Benjamin StevensDr. Benjamin StevensRegional Science Research InstituteRegional Science Research InstituteP. 0. Box 3735P. 0. Box 3735Peace Dale, RI 02883Peace Dale, RI 02883

Dr. Daniel StynesDr. Daniel StynesDep. of Park and Recreation ResourcesDep. of Park and Recreation ResourcesMichigan State UniversityMichigan State UniversityEast Lansing, MI 48824East Lansing, MI 48824

Ms. Nancy TessaroMs. Nancy TessaroNatural Resources Management BranchNatural Resources Management BranchChief of EngineersChief of EngineersDepartment of the ArmyDepartment of the ArmyCorps of EngineersCorps of EngineersWashington, DC 20314Washington, DC 20314

Dr. Timothy J. TyrrellDr. Timothy J. TyrrellDep. of Resource EconomicsDep. of Resource EconomicsUniversity of Rhode IslandUniversity of Rhode IslandKingston, RI 02881Kingston, RI 02881

6464

Propst, Dennis Propst, Dennis B.,B., compiler compiler Propst, Dennis Propst, Dennis H.,H., compiler compiler

Assessing the economic impacts of recreation andAssessing the economic impacts of recreation and Assessing the economic impacts of recreation andAssessing the economic impacts of recreation andtourism. Conference and Workshop; 1984 May 14-16;tourism. Conference and Workshop; 1984 May 14-16; tourism. Conference and Workshop; 1984 May 14-16;tourism. Conference and Workshop; 1984 May 14-16;East Lansing,East Lansing, MI. Asheville, NC: U.S. DepartmentMI. Asheville, NC: U.S. Department East Lansing,East Lansing, MI. Asheville, NC: U.S. DepartmentMI. Asheville, NC: U.S. Departmentof Agriculture,of Agriculture, Forest Service, Southeastern Forest Service, Southeastern for- of Agriculture, Forest Service, Southeastern For-of Agriculture, Forest Service, Southeastern For-est Experiment Station; 1985.est Experiment Station; 1985. est Experiment Station; 1985.est Experiment Station; 1985. 6464 PP=PP=

A collection of eight papers that explore and assessA collection of eight papers that explore and assess A collection of eight papers that explore and assessA collection of eight papers that explore and assessthe best available technology to evaluate the the best available technology to evaluate the econo-econo- the best available technology to evaluate the econo-the best available technology to evaluate the econo-mic impact on recreation and tourism.mic impact on recreation and tourism. Research Research stra-stra- mic impact on recreation and tourism.mic impact on recreation and tourism. Research stra-Research stra-tegies for meeting methodological and data needs aretegies for meeting methodological and data needs are tegies for meeting methodological tegies for meeting methodological andand data needs are data needs arerecommended.recommended. recommended.recommended.

KEYWORDS:KEYWORDS: Technology transfer,Technology transfer, problem-solving problem-solving pro-pro- Technology transfer, problem-solving Technology transfer, problem-solving pro-pro-