Background
What is the Travel Cost Method (TCM)?
The TCM is a means of determining value figures for things which are generally not bought and sold, and therefore fall outside of the market’s pricing system. The non-market assets which it is most often applied to are ‘recreational resources which necessitate significant expenditure for their enjoyment.’[9, p. 93] This means that the TCM is often used to assess the value of parks, lakes, and similar public areas which host a good deal of recreational activity, and which are far enough away from many people to require users to drive or fly to the site.
The basic premise of the TCM is that, although the actual value of the recreational experience does not have a price tag, the costs incurred by individuals in travelling to the site can be used as surrogate prices. The weak complementarity of the goods required for travel to the site makes it possible estimate a demand curve for the site, and from it, a measure of the sites’ consumer surplus can be found. It is important to note that the consumer surplus figure is a measure of the user value of the site only, and does not necessarily measure the site’s environmental or intrinsic value.
Where does it come from?
The idea for the TCM is attributed to Harold Hotelling, who proposed the basic notion of the method to a park service director in a 1947 letter. It was not put into practice extensively until the late 1960’s, and has only reached a more refined state in relatively recent years. Jack Clawson and Marion Knetsch are widely regarded as two of the most important figures in the early development of the TCM.
Philosophically, the TCM falls into the general category of neo-classical welfare economics, which assumes that individuals maximize their utility subject to certain constraints. This has implications for both the technical exposition of the method and for the premises upon which it is built. Debate about the merits or demerits of alternative economic theories exceed the scope of this project, but it is important to be conscious of the model’s theoretical foundations.
Who uses it?
Essentially, anyone interested in studying the value of recreation may have reason to apply the TCM. As noted earlier, only certain types of sites make good candidates for travel cost analysis, so its applicability is bounded. But there are legislative and public support-oriented reasons in both the United States and the EC which promote cost-benefit analysis to determine the financial merit of a project, and so the impetus to find good valuation techniques is strong. In the US, much of the TCM’s application has involved water-based recreation, partly because water resources make good subjects of travel cost analysis, and partly because federal agencies involved in the management of water-based sites like the Army Corps of Engineers have been particularly enthusiastic about the use of the TCM.
What are its alternatives?
The search to find means of estimating the value of non-market goods has yielded two categories of approaches: direct and indirect. The TCM is one of the primary members of the indirect category, all of which utilize the complementarity of market and non-market goods, and study people’s behavior to determine their preferences. The direct methods, on the other hand, attempt to guage value by asking people directly for their ideas of the worth of an asset. Contingent Valuation, which is the TCM’s primary alternative, is based on surveying people about how much they would be willing to pay to preserve (or create) a non-market asset. The criticisms of Contingent Valuation center on the issue of whether people give accurate responses to such hypothetical questions. It is quite popular, however, and CV studies have the advantage of allowing respondents to aggregate the various types of benefits of a site (non-user as well as user value.) An associated disadvantage is the question of whether the public is sufficiently educated to properly estimate some categories of non-user value: ecological value, for example.
Summary
How does someone do a travel cost analysis?
The first step involved with the TCM is the creation of a trip generating function. In an actual travel cost study, this stage could not take place before a certain amount of thought and research concerning the goals and form of the study, and a significant amount of data collection work. We will touch on some of those issues later, but focus here on the exposition of the model.
The purpose of the trip generating function is to provide a model of site use. There are two types of functions: one based on an individual model, the other zonal. The type of function determines the dependent variable, which is either the number of trips made by inhabitants of a given geographical zone, or it is the number of trips made by individuals. In either case, the independent (or explanatory) variables describe the costs of travel, including time and on-site costs along with gas, etc. There is no fixed formula for the other independent variables to include. Ideally, everything which affects the dependent variable is included, without harming the statistical integrity of the function. In practice, both statistical problems and theoretical issues, as well as the demands of data-collection, mean that we have to settle for less-than ideal functions. Other independent variables which may be included, depending on the study, are: zonal populations, socioeconomic characteristics of study participants, information concerning substitute sites, environmental quality indicators, etc. Fuller treatment of each of these is included under Particular Concerns and Independent Variables, below.
The trip generating function is statistically determined through multiple regression once the various issues involving the dependent and independent variables have been addressed and the labor-intensive work of data collection has been performed.
Once the trip generating function has been determined, it can be used to define a demand curve for the site. This is done by considering what impact price increases would have on aggregate demand, and tracing out a curve through this process. A fuller explanation of this process can be found in Clawson and Knetsch.[7] Once the demand curve has been established, finally, it is only a short conversion to finding an estimate of consumer surplus, which is the area below the demand curve and above the current price line. Mathematical, rather than geometric, definitions of consumer surplus can be found in some of the other resources listed below, but they are just different ways of representing the same ideas.
What are some particular concerns?
The formulation of the trip generating function gives rise to a number of issues of concern. They range from the type of function to use to the treatment of specific independent variables or types of travellers. A summary of many of these issues follows directly, and under ‘Treatment of Independent Variables.’
Function type: As mentioned earlier, the function can follow an individual or zonal model. Much of the literature on the subject presumes the adoption of a zonal model, which means that recreationists are classified according to their zone of origin, according to some natural breakdown of the area surrounding the site. In order to have a reasonable sample size, the zones of origin may need to be quite small, in which case travel costs may not behave well in the statistical model. Another problem with the zonal model is that some of the independent variables may need to be aggregated across the zones’ inhabitants, which may increase the error or decrease the likelihood of finding significant features. Ultimately, this point raises the question about whether a zonal model is compatible with an individual ultility-maximization approach. The individual model seems more compatible with this approach, but necessitates a more labor-intensive data collection process, as all relevant data must be collected from each site visitor. This has the potential to cause problems as well.
Different types of travellers: The TCM is best suited to sites which draw only day-trip visitors. But many recreation sites are visited by people who are on holiday for an extended time period, or who stop in to the site without necessarily making the trip exclusively for the purpose of visiting it. In either case, the traveller poses difficulties for the would-be TCM modeller; including all of the travel costs of the latter visitor seems inappropriate, while including only the local travel costs of the former is equally dubious. But the TCM does not necessarily have sufficient flexibility to treat such individuals differently. These types of recreationists violate the two foundational assumptions of the TCM–that the travel costs which people incur to visit a site are a good proxy for the amount they value the use of the site, and that the travel costs can be calculated quite simply. This problem is often solved by discarding holiday-makers and other non-traditional visitors from the sample, which may well bias many estimates downwards. Two other solutions have been proposed. One is to include a trip-type variable among the explanatory variables. This would be a dummy variable that could demonstrate if someone was taking part in an extended trip, a day trip, or just stopping in as part of a multi-purpose trip. The other solution, specifically formulated to deal with the case of holiday-makers, is a model proposed by Bell and Leeworthy. [3] The authors create a new model as an alternative to the standard TCM, which uses the number of days of recreation as the dependent variable, and handles fixed trip costs in addition to daily on-site costs. For a more in-depth look at this alternative, see the article and a response to it, by Shaw. [12]
Time: Perhaps no other single issue is so vexing for the TCM modeller as the treatment of time costs. The original conception of the TCM envisioned the costs of travel (for gas, for example) as being a sufficient explanation of people’s willingness-to-pay for a recreational site. But the most cursory of observations indicates that decisions about recreation are based in large measure not just on the costs of travel, but the availability of time. This suggests that one common approach to dealing with time costs–to simply omit them–is inappropriate. One other common response is to value time at a fixed percentage of the wage rate. This raises the question of whether only travel time should be counted, or whether on-site time should be included as well. These are the prominent means of assessing time costs. The feasibility of other options depend upon the flexibility of the model used. There is a considerable debate about the value of time in reference to recreation studies, and it exceeds the scope of this analysis to address all of the issues. But for more information, see Cesario [6].
There are also some technical concerns relating to the formulation of the demand curve, and the resulting consumer surplus figure. These center on the difference between Marshallian and Hicksian curves. Another question is the correct form for the trip generating function (linear, log, etc.) For analysis of these subjects, see McConnell [11].
Treatment of independent variables
The most basic models of the TCM include only a few explanatory variables–mostly, in keeping with the original conception of the model, involving travel cost. But an effort to more thoroughly model the individual decision process has led to the inclusion of an array of different variables. Before entering into some discussion about particular variables, and how they are incorporated, this is a good opportunity to discuss the practicality of attempting to capture every factor affecting demand in a trip generating function. First, as Clawson and Knetsch mention, it may not be appropriate to include factors which are fixed for the period of the study. Second, a number of factors may be directly correlated so that inclusion of one suffices to represent a number of them. And last, it is important to bear in mind the process of data collection involved in actually implementing the TCM. So while it may seem tempting to include a vast variety of possible factors, practical limitations on data availability must be considered. And even when it is feasible to collect data on a wide variety of factors, the modeler must choose between comprehensiveness and the risk of insignificant coefficients, particularly when the sample size is small, as it tends to be in zonal travel cost studies. In short, an elegant trip generating function retains a careful balance between thoroughness, statistical integrity, and realism in the face of data requirements.
That said, some independent variables which are not completely straightforward include:
Costs: Many of the of concerns relating to costs have been discussed above, in conjunction with both time and the traveller type dilemma. Some remaining questions are how to translate distance to travel costs (i.e. what mileage cost to use) and what on-site costs to include. Note that the zonal model generally does not have the flexibility to include costs which vary between individuals, because it assumes that everyone from a given zone has approximately the same costs, based on travel from the center of the zone.
Socioeconomic characteristics: The socioeconomic characteristics which are considered will depend on the type of function. In the individual model, any information collected on the survey and thought relevant to recreation decisions might be included. This could include not only income, but also education level, race, age, etc. The zonal model could include these characteristics as well, but based on averages for the region collected through census data. The inclusion of income data, Bockstael and McConnell point out, may prove insignificant, as income may differentiate participants from non-participants, but not occasional users from frequent users. Particularly in the case of studies of day trips, the variable costs of travel may be relatively insignificant for those who have passed the entry barriers–car ownership, boat ownership, specialized recreational equipment ownership. Studies which sample only users may not, therefore, find income to be significant
Substitute Sites: The importance of incorporating information concerning substitute sites into the trip generating function is quite straightforward; if residents of one area have close access to a high number of substitute sites, while residents of another region do not, the demand for the site in question will be affected. Without some inclusion of cost figures for substitutes, the decision process is not likely to be accurately modelled. But the treatment of substitute sites appears to be omitted in the most basic models. In some cases this may be warranted. For example, among some samples recreation sites may all be a comparable distance. But the omission in many studies appears to be more a function of the difficulty associated with accounting for substitute sites than the result of a decision that the figures will be insignificant. The literature listed below does not contain any substantial discussion of how substitute sites might be adequately and simply represented. In the zonal model, it would be possible to select a set of substitute sites and determine their distance from each zone. In the individual model, the survey could include questions about the distance to substitute sites. But further complications arise. Other sites may have different entry costs; people may perceive the opportunity cost of time differently if the route to a site is more or less scenic; there is a question of whether to include environmental information on substitute sites as well. In addition, studies involving long distance travellers could become mired in an endless series of questions and calculations involving substitute sites. At some point, assumptions need to be made to limit the complexity of the treatment of substitutes. But this assumption should not, perhaps, be that substitute sites are insignificant, which appears to be one common choice.
Environmental quality: This topic encompasses a wide of factors which might differentiate one site from another. In addition to the range of issues which might often be thought of as belonging to the environmental quality category, the various types of infrastructure related to the site and the site’s congestion level could be considered part of environmental quality characteristics.
It is important to include some representation of environmental quality if it plays an important role in individuals’ decisions about what site to visit. But without having some comparison available for other sites, it may not be clear what role the data is playing in a single study. This, generally, is another issue which does not appear to receive a great deal of attention. In standard TCM analyses, environmental quality will be a consideration only if it influences a person’s decision to visit one site over another, and therefore it would appear silly to focus on environmental quality and not substitute sites. The most extensive treatment of site quality in the litereature below is in the hedonic travel cost study by Englin and Mendelsohn.(See the description of hedonic TCM below.) As their aim is to determine the value of particular qualities, it’s natural that they should treat each site’s environmental conditions quite thoroughly. They include a barrage of measures, including: each trail’s forest characteristics, the amount of old growth, the presence of clear cuts, the grade and maintenance level of the trail, the amount of rock and ice along the trail, the presence of campsites, the amount of dirt road to each trailhead, the facilities at the trailhead, etc.
It would be excessive and futile to include this amount of information in a non-comparative travel cost survey, but they at least demonstrate that it is possible to deal with quite a complex data set. For other purposes, the relevent indicators hinge on the type of site. For beaches, for example, the cleanliness of the water and the amount of litter on the beach are major variables. For the Grand Canyon, haze and pollution is a factor. For fishing sites, the stock of fish is significant. And so on. Regardless of the site’s indicator types, there is a question about how to apply the information. Either nonbiased indicators can be used, or, in the individual model, people can be questioned about their perceptions of the site’s environmental quality.
The related congestion issue receives somewhat more attention in the literature listed, but not to any definitive end. Clearly an individual may choose a less congested site over one which is thought to be too crowded, but the presence of a large number of other people may not be interpreted as congestion, depending on the site and the person’s perspective. Furthermore, the popularity of a site may be an encouragement for some visitors. Again, there is a question about whether to measure congestion by unbiased indicators or individuals’ perceptions. Congestion also poses the difficulty of varying both across sites and across time periods, and in his critique of Bell and Leeworthy, Shaw notes that an average of congestion levels during a year may do little to explain individual behavior during peak or off-peak periods. Shaw quotes work by Smith, who says that ‘without specific details on the time patterns of use by individual recreationists and the associated facility conditions, we cannot leap to conclusions on the effect of congestion on benefit estimation with the travel cost model.’ The treatment of congestion, therefore, needs to be considered carefully.
Weather is another factor which falls under this broad category, and it, like congestion, varies among visits and is subject to a certain level of personal opinion. In order to capture some representation of the influence of weather, it may not be necessary to include a complex description. McConnell (1977) notes that inclusion of precipitation information in addition to temperature data did not yield significant results, as temperature and precipitation are quite closely linked. In addition, the complexity of the data required is an important consideration here. An attempt to very thoroughly model all possible weather factors over the period of the study would be a daunting undertaking in and of itself. So a means of simply capturing the essence of the impact of weather on people’s choices would be useful. An appropriate means of accomplishing this will depend on the model in question–the trip length being considered, the means of modelling the individual decision process, etc.
Explanation
What do the products of a travel cost analysis tell us?
Ultimately, the purpose of a travel cost study is to arrive at an estimate of the site’s consumer surplus. Viewed uncritically, this is the value of the site to society during the period of time of the study. But there are a number of important things to bear in mind. The first, as noted earlier, is that the value derived is recreational value only, and not environmental or intrinsic value. Secondly, the transformation from a fixed time period value to a value over time would require the choice of a discount rate, which tends to be a hairy issue in environmental economics. And, lastly, the literature on the topic in general is somewhat ambiguous about the reliability of the measure found through the TCM. Ultimately, questions are raised about the nature of user value figures, their reliability and comparability.
Is there only one kind of travel cost study?
No. The information above has dealt with a basic single-site analysis, but similar principles can be applied to multiple sites as a composite good, or, in a slightly varied form, to environmental characteristics of a selection of sites. The latter study type is sufficiently different from the standard TCM to warrant its own name: Hedonic Travel Cost Analysis. Essentially, hedonic analysis evaluates the worth of particular site features, concluding with an estimate of the marginal value of increased environmental quality, for example, by studying the extra travel expenses which people are willing to absorb in order to find certain characteristics. In theory, this is a useful means of evaluating site management decisions. It is performed by including a variety of site characteristic values along with travel costs in a travel cost function, and then determining the marginal cost of each characteristic, although there are methodological variations.
One example of this type of work can be found in Englin andMendelsohn [8], who apply the hedonic travel cost method to analyze the demand for certain site characteristics (such as forest type, facilities available, the presence of views, etc.) on hiking trails in Washington State USFS wilderness areas. After determining the form of the model to use, the authors apply data from site visits to a number of trails and determine marginal social value for various site characteristics. In this way the authors determine, for example, that the marginal social value for a mile of old growth forest on the Pasayten trail is USD 1254. Clearly results such as these, if they are reliable, could be important tools for site management. Englin andMendelsohn, however, perform their survey with the knowledge that the hedonic TCM has been heavily criticized recently. The primary basis of criticism seems to be that the method has a tendency to produce ‘incorrect signs’–that is, negative values for characteristics which would have been thought desirable–implying that the individual results of hedonic surveys cannot be trusted.
Discussion
Is the TCM a perfect system?
No. The complexities alluded to above should give a reasonable indication of the weaknesses of the TCM. A quick summary of some of its more significant problems would include the question of what consumer surplus measures, the correct approach to dealing with time, the treatment of different types of travellers, and the correct specification of the trip generating function. Despite these reasonably formidable questions, however, there is energetic work attempting to make progress on these and related issues. The TCM has not been given up for a lost cause. Rather–whether because of its own merits or a lack or alternatives–it is one of the most significant tools we have for the estimation of user value for non-market assets.
Where can I look for more information?
A cursory search for good on-line resources has proven reasonably unsuccessful, but two organization which have been instrumental in the development and/or application of the TCM are the USFS Southern Research Station– http://www.rtp.srs.fs.fed.us/econ/econhome.htm —and Resources for the Future — http://www.rff.org
Following is a list of print resources which are pertinent to the topics discussed above. It is far from complete, but is a good start:
[1]Adamowicz, W, and T Graham-Tomasi. ‘Revealed Preference Tests of Nonmarket Goods Valuation Methods,’ in Journal of Environmental Economics and Management. 20: 29-45 (1991)
[2]Adamowicz, W, J Louviere, and M Williams. ‘Combining Revealed and Stated Preference Methods for Valuing Environmental Amenities,’ in Journal of Environmental Economics and Management. 26: 271-292 (1994)
[3]Bell, Frederick, and Vernon Leeworthy. ‘Recreational Demand by Tourists for Saltwater Beach Days,’ in
Journal of Environmental Economics and Management. 18: 189-205 (1990) Back to the text
[4]Bockstael, N, K McConnell, and I Strand. ‘Recreation,’ in Measuring the Demand for Environmental Quality,
John Braden and Charles Kolstad, eds. Elsevier: Amsterdam, 1991.
[5]Burt, Oscar and Durward Brewer. ‘Estimation of Net Social Benefits From Outdoor Recreation,’ in
Econometrica. 39: 813-827 (1971)
[6]Cesario. ‘Value of Time in Recreation Benefit Studies,’ in Land Economics. 52: 32-41 (1976) Back to the text
[7]Clawson, Marion and Jack Knetsch. Economics of Outdoor Recreation. Johns Hopkins University Press,
Baltimore: 1966. Back to the text
[8]Englin, Jeffrey andRobert Mendelsohn. ‘A Hedonic Travel Cost Analysis for Valuation of Multipla
Components of Site Quality,’ in Journal of Environmental Economics and Management. 21: 275-290
(1991) Back to the text
[9]Hanley, Nick and Clive Spash. Cost Benefit Analysis and the Environment. Edward Elgar Publishing, England: 1995. Back to the text
[10]McConnell, Kenneth. ‘Congestion and Willingness to Pay: A Study of Beach Use,’ in Land Economics.
53:185-195 (1977)
[11]McConnell, Kenneth. ‘The Economics of Outdoor Recreation,’ in Handbook of Natural Resource and Energy Economics, vol. II. Kneese and Sweeney, eds. Elsevier Publishers: Amsterdam, 1985. Back to the text
[12]Shaw, Douglass. ‘Recreational Demand by Tourists for Saltwater Beach Days: Comment,’ in Journal of
Environmental Economics and Management. 20: 284-289 (1991) Back to the text
[13]Smith, Kerry and Yoshiaki Kaoru. ‘Signals or Noise? Explaining the Variation in Recreation Benefit
Estimates,’ in American Journal of Agricultural Economics. May 1990: 419-433.