Valuing User Preferences for Improvements in Public Nature Trails Around the Sundays River Estuary , Eastern Cape , South Africa

Many valuations have been made of changes to in-estuary attributes but few have been made of out-of-estuary attributes. From a recreation perspective, an important type of out-of-estuary attribute is the availability of public paths by which to access attractive features of the estuary environment. This paper values an improvement in the level of public access in the form of an additional nature trail along the banks of the Sundays River Estuary in the Eastern Cape, but does not compare this value with the costs. By means of choice experiment modelling analyses it is estimated that in 2010 the marginal willingness-to-pay for an investment in a nature trail was R34 per user per annum. In order to determine whether the development of this trail is efficient, this benefit (R34 per user per annum) needs to be compared to the cost of the development, an analysis that remains to be done. However, this find does serve to provide guidance on how much funding could efficiently be allocated to such a development about R1.22 million, assuming a social discount rate of 8.38%.


Introduction
The Sundays River Estuary is a major tourist attraction (Cowley, Childs & Bennett, 2009).Recreational activities that occur at the Sundays River Estuary generate values for various participants, most notably tourism revenues for local businesses.The demands of a growing population, however, have led to increased pressures on the estuary as a recreational outlet.This demand pressure jeopardising the sustainability of these environmental service flows.As Rail Bridge.The east bank to the north of the N2 Bridge is mainly occupied by residential properties.The estuary banks along this stretch are also steep and inaccessible (Cowley et al., 2009).The area to the north of the Mackay Rail Bridge is hardly accessible by road.
In the vicinity of Colchester and Cannonville private jetties have proliferated in an ad-hoc manner along the northern bank of the Sundays River Estuary.Most of these jetties have been constructed on Municipal Public Open Space without authorisation.Although most of the jetties are situated on Municipal land, access is controlled by those who erected them.
In their status quo assessment report, Afri-Coast Engineers recommended that ". . .a continuous strip of green open space be preserved along the river banks (of the Sundays River Estuary) to form an aesthetic nature trail providing a valuable asset to the area for both local residents and tourists" (Afri-Coast Engineers, 2004).The green open space must constitute a sufficiently wide river frontage to allow for safe public access.The engineering company further recommended that ". . .negotiations should be initiated with the private land owners who own private land along the river edges (of the Sundays River Estuary) to investigate a mutually beneficial partnership to conserve this ecologically valuable land" (Afri-Coast Engineers, 2004).Other privately-owned land could be incorporated into conservancies, or bought by the Nelson Mandela Bay Municipality, in order to conserve these areas and to incorporate them into the Nelson Mandela Metropolitan Open Space System (Afri-Coast Engineers, 2004).
The National Water Act of 1998 governs public access to estuaries in South Africa, but is vague on how this public access must be managed and conserved (NWA, 1998).It states: "A person may, subject to this Act-. . .(e) For recreational purposes -(i) use the water or the water surface of a water resource to which that person has lawful access; or (ii ) portage any boat or canoe on any land adjacent to a watercourse. . .".The introduction of a nature trail fronting the banks of the Sundays River Estuary appears to be allowed for in the Act as it would be 'adjacent to the watercourse'.From a social perspective it would improve the quality of the public land fronting the water's edge, and make it more appealing for recreational shore fishing, as well as provide further areas for other recreational activities, such as bird watching or walking.But what would society (the users) be prepared to pay for the development of such a facility?This paper uses the choice experiment (CE) methodology to estimate the recreational user's willingness to pay (WTP) for this development, and by so doing, provides guidance to the estuary management authorities on what level of resources they can efficiently allocate to the path development project.

Conceptual framework
The CE method is a choice modelling technique that is used to measure user preferences for goods where goods are described in terms of attributes and levels (Hanley, Koop, Alvarez-Farizo, Wright & Nevin, 2001).Respondents are presented with a number of different options made up of differing attributes and levels and asked to choose their most preferred.The theoretical background to the CE method dates back to Lancaster's model of consumer choice (Lancaster, 1966) and random utility theory (Thurstone, 1927).It is argued that consumers enjoy utility not from the good as a whole, but rather from the attributes or characteristics that comprise this good (Lancaster, 1966).Random utility theory is the basis for analysing respondent behaviour within the context of a CE.It is based on the assumption that individuals make their choices based on the good's observable attributes, together with a random element.Manski (1977) identified four sources of uncertainty contributing to the unobserved or random element of utility: effects of unobserved alternative attributes, effects of unobserved consumer characteristics (or taste variations), measurement errors, and the use of imperfect proxy (or instrumental) variables.Indirect utility for individual i in choice scenario q can be defined as: where, x iq is a vector of observed attributes, and β is a vector of the parameters to be estimated (Hanley et al., 2001).This indirect utility function comprises one observable component of utility, V iq , and a random component, ε iq .It is assumed that individuals make choices in order to maximise their utility.An individual (i) will select alternative q if and only if (iff) U iq is greater than the utility derived from any other alternative facing person i in the choice set C. This probability is formulated by: The utility of each alternative (q and j) is divided into two parts, namely an observable component and a random element.Equation 2 is rewritten to include this information: Choice can be predicted by estimating the probability of individual (i) ranking alternative q higher than any other alternative j in the set of choices available (Louviere, Hensher & Swait, 2000;Nam Do & Bennett, 2007).In order to derive this probability expression, an assumption needs to be made with regards to the distribution of the error term (Hanley et al., 2001).The error term is typically assumed to be independently and identically distributed (IID) with an extreme-value type 1 distribution (EV1): This error term distribution implies that the probability of any option q being chosen can be expressed in terms of a logistic distribution (McFadden, 1974).This logistic distribution is known as the conditional logit (CL) model: where there are a different alternatives in choice set C. The scale parameter, µ, is typically assumed to be one (Ben-Akiva & Lerman, 1985).This model is restrictive in terms of its underlying assumptions.In addition to the IID assumption of the error term, it assumes that random components do not exhibit serial correlation, i.e. they are independent of irrelevant alternatives (IIA), that utility parameters are set, and that homogeneity of preferences across respondents is maintained (Bhat, Eluru & Copperman, 2000).In order to relax the IIA assumption, Bhat (1995) proposed a heteroscedastic extreme-value (HEV) model which allows different scale parameters across alternatives.The variances of the error terms are allowed to differ across all options.However the error terms of one alternative must have a scale parameter that is normalised to one for identification purposes (Bhat et al. 2000).The CL model can be adapted to allow for variance of the scale parameter (λ) as shown in Equation 6: The HEV specification models the probability that an individual (i) will choose the qth alternative in a choice set (C), but relaxes the assumption of independence among the random components.Substituting z in place of (ε i /λ i ), the HEV specification of the choice estimation model is: The CL and HEV models do not allow variations to an attribute across individuals (Bhat et al. 2000).Unobserved respondent characteristics can, however, affect responses to the choice questions.If these unobserved characteristics are ignored, biased and inconsistent choice probability estimates could result (Bhat et al. 2000) These weaknesses are addressed in the random parameter logit (RPL) model, after making it superior to the CL model in terms of fit and overall welfare estimation (Carlsson, Frykblom & Liljenstolpe, 2003;Morey & Rossman, 2003;Birol, Karousakis & Koundouri, 2006).The RPL model generalises the CL model by allowing the coefficients of observed variables to vary randomly over individuals rather than being fixed (Hynes & Hanley, 2005).It achieves this by dividing the error component of the utility function in Equation 1 into two parts, so permitting the possibility that information relevant to making a choice, that is unobserved, induces correlation across alternatives in a choice situation (Hynes & Hanley, 2005).One component of the error term is correlated over alternatives.Another follows the IID assumption across alternatives and individuals (Goibov, Schmitz, Bauer & Ahmed, 2012): where X iq is a vector of identifiable attributes of alternatives within a choice set, β is a vector of parameters unobserved for each individual that varies ran-domly over individuals based on their tastes, η i denotes a vector of standard deviation parameters representing individuals tastes relative to the average tastes in the population of individuals, and ?iq denotes the random portion of utility (Hynes & Hanley, 2005;Goibov et al., 2012).The probability of individual i choosing alternative q in choice set C is (Louviere et al., 2000): where α qi is a fixed or random alternative specific constant (ASC) with j = 1,....,J alternatives and i = 1,....,I individuals; and α j = 0, θ q z i is a vector of non-random parameters responsible for individual characteristics, β ji is a parameter vector that is randomly distributed across individuals; z i is a vector of individual-specific characteristics, income, x ji is a vector of individual-specific and alternative-specific attributes, and µ i is the individual-specific random disturbance of unobserved heterogeneity (Train, 1998).
In order to estimate this model, it is often assumed that the parameters β are either normally, triangularly, uniformly or log-normally distributed over the population of individuals (Bhat et al., 2000;Bhat, 2001).Normally distributed parameters, means and standard deviations of coefficients can determine to what extent respondents place positive or negative values on a change in an environmental attribute (Train, 2003).

Brief literature review
The literature that exists on the valuation of wetlands, estuaries and rivers is largely focussed on valuing specific attributes of interest within each site.Studies that make use of the CE methodology to value these attributes include Opaluch, Grigalunas, Diamantides, Mazzotta and Johnston (1999), Heberling 2009), and Beharry-Borg and Scarpa (2010).In South Africa, there have only been a few studies reported that make use of the CE technique to value environmental attributes.Most notably, it was applied in order to value improvements in freshwater inflows into the Kruger National Park catchment areas (Turpie & Joubert, 2004), and the Bushmans River Estuary (Oliver, 2010).
The value of improved public path access has not been addressed by many empirical studies.Kline and Swallow (1998) estimated WTP values for public access to Gooseberry Island, Massachusetts, USA.They did this through the use of the contingent valuation method (CVM).A study by Dyack, Rolfe, Harvey, O'Connell and Abel (2007) estimated the impact of different public access situations on recreational value in the Murray River Basin, Australia.This study made use of both revealed and stated preference techniques to estimate non-market recreational values.A contingent behaviour survey was applied to ascertain how recreational users would adjust their use of the site in question if public access conditions changed.The conditions could change due to changes in management and/or the provision of additional tracks and facilities.

CE format and attributes
Interviews were conducted with experts as well as members of the Sundays River Ratepayers Association.These interviews identified the most important issues facing recreational users of the Sundays River Estuary and specified them in terms of attributes with differing levels.One of the key problems identified related to the lack of adequate public access to the estuary.Based on these discussions, and following the steps outlined by Hasler, Lundhede, Martinsen, Neye and Schou (2005), a questionnaire was developed and pre-tested through a pilot survey.Following this pilot survey, some editorial changes were made to the questionnaire in order to improve its clarity and reduce the overall cognitive burden on the respondent.The CE section of the edited questionnaire included three management attributes, namely 'Physical size of fish stocks caught', 'Boat congestion' and 'More public access?'These three attributes had two qualitative levels each.The qualitative attributes were used because respondents related better to these -they were less cognitively demanding (see also Hasler et al. 2005).The fourth attribute represented the payment vehicle and was defined as an annual environmental levy added on to the existing boat license fee structure.This cost variable was expressed by four different Rand values in the CE.It was found to be the most understandable and least controversial option out of those discussed in the focus groups.Attributes selected for this study and their corresponding levels are presented in TABLE 1 below.
The two alternatives presented to the recreational users of the estuary were different combinations of these three management attributes and their levels with a cost value attached.For the purposes of this study, a status quo or 'opt out' alternative was not included.The reason for this was twofold.It was difficult to define a status quo option as some of the current recreational uses pertaining to the estuary can be defined as illegal, for example, bag and size limits in the fishery are currently not being adhered to.In addition, it was thought unnecessary to include a status quo (opt out) alternative when the purpose of the study was to guide policy-making (Hasler et al., 2005).
An introductory section was provided at the beginning of the CE questions to familiarise the respondent with the different management attributes and their levels.Information on the CE payment was also presented so that the respondents were aware of the payment vehicle, as well as the need to consider the constraints on the household's budget.The assumptions with respect to the payment were (1) that the costs of implementing the policy alternatives would be covered by the recreational users of the estuary, and (2) that all users would contribute equally to the implementation of the scenarios by means of a fixed annual sum per household.

Designing the CE choice sets
Overall, four attributes were defined.Three of these attributes had two levels each, and one had four levels.A full factorial design (2x2x2x4 = 32) was generated using SPSS, yielding 32 different treatment combinations or alternatives.These alternatives were randomly allocated to 32 different questionnaires.Each questionnaire contained four choice sets, and within each choice set, the respondent had to make a choice (trade-off) between two alternatives.Two alternatives were adopted in this case as more than two per choice set can result in 'respondent fatigue' (Bateman, Carson, Day, Hanemann, Hanley, Hett, Jones-Lee, Loomes, Mourato, Ozdemiroglu, Pearce, Sugden & Swanson, 2002).A sample choice set is provided in FIGURE 3, in which Option B was selected.

Sample design and data collection
The population of interest with respect to the Sundays River Estuary was all users and potential users of the recreational services provided by each estuary.The sample frame, however, could not be compiled, as this population does not reveal itself until it visits the estuary.As it was impossible to identify a sample frame, the closest to this objective was knowledge of the sample population and use of this knowledge to sample select.This form of non-list sampling can be used when the target population refers to visitors to a beach, or in this case, an estuary (Bateman et al., 2002;Dillman, Smyth & Christian, 2009).It entails onsite sampling, and is known as an intercept survey (Bateman et al., 2002).Two approaches to the determination of sample size in choice modelling exercises are often proposed (Hensher, Rose & Greene, 2005): the use of probability sampling and rule of thumb.Probability sampling is very often abandoned due to practical considerations (budget and time constraints).A 'rule of thumb' approach was used to calculate the minimum sample size required to estimate a model of choice using unlabelled experiments and design attributes onlya sample of 50 respondents each exposed to 16 choice sets is thought to be capable of yielding meaningful and significant results (Bennett & Adamowicz, 2001).This translates into a sample of 200 respondents if they are offered 4 choice sets each.
The face-to-face interview method is the most common approach to use when valuing recreational sites (Lee & Han, 2002).This personal interview method was adopted for this study.Although costly, it affords the interviewer the best opportunity to encourage the respondents to cooperate with the survey.The interviewer is also given an opportunity to explain complex information and valuation scenarios to the respondent -which is very important in the CE setting (Mitchell & Carson, 1989).The Sundays River Estuary questionnaire was administered on-site by four trained interviewers during August, 2010.Interviewers approached every n th potential respondent and asked them if they would be willing to spend approximately 15 minutes filling in the questionnaire.In total, 175 completed questionnaires were collected, a number below the recommended sample size, but still considered adequate in order to estimate 'robust' models (Hensher et al., 2005).

Socio-economic characteristics
Selected socio-economic results revealed that most of the visitors surveyed came from areas less than 50km away from the estuary, were male, over the age of 35, had a matric qualification with university exemption, and earned an average annual income of R184 000.Of these respondents, most came from Port Elizabeth (59%).Permanent residents of the estuary, living in Colchester and Cannonville, accounted for approximately 21% of the sample.

CE results
Three different choice model specifications were estimated for the CE: a CL model, an HEV model and an RPL model.The LIMDEP NLOGIT Version 4.0 programme was used in all these estimations.TABLE 2 reports the estimation results of the three model specifications.
The coefficients in all three models have the correct signs, a priori, and are significantly different from zero at the 95% confidence level (at the least).The cost attribute parameter is as expected i.e. significant and has a negative sign in all three models estimated.The probability that an alternative would be chosen was reduced: the lower the physical size of the fish stock, the higher the amount of boat congestion, the lower the amount of public access available, and the higher the environmental quality levy.
In the case of the CL model, the significant coefficients can be interpreted by estimating their odds ratios.This is done by calculating the antilog -the value of 10 to the power of the coefficient's value -of the various coefficients.Odds interpretation indicates how an increase (decrease) in an attribute's level would result in a change in the probability of choosing an option which includes this increase (decrease).With respect to the attribute of interest, i.e. public access, an increase in public access will result in an increase in the probability of a respondent choosing this option by 2%.The explanatory power of the CL model is measured by the Pseudo R 2 .At 22% this is a good fit for CE-type studies -Louviere et al. (2000) suggested that anything between 0.2 and 0.4 can be considered very good.
In order to relax the IIA assumption inherent in the CL specification, an HEV model is estimated.The HEV specification relaxes the assumption of independence among the random components (Bhat, 1995).The odds interpre-tation reveals that an increase in public access will result in an increase in the probability of a respondent choosing this option by 3%.
The CL and HEV models do not allow sensitivity variations to an attribute across individuals (Bhat et al. 2000).Unobserved respondent characteristics can, however, affect responses to the choice questions.To address this problem, the RPL model was estimated.Consistent with recommendations made by Hasler et al. (2005), all parameters except the cost parameter were treated as random and assumed to be normally distributed.Cost was specified as fixed, and not randomly distributed, because in this case, the distribution of the marginal WTP for an attribute is simply the distribution of that attribute's coefficient.Results reveal that the estimated RPL model shows statistically insignificant attribute coefficients.This indicates statistically similar preferences for attributes across respondents, implying a largely homogenous recreational user group of mostly boat users.

Estimation of WTP values
Measures of welfare can be calculated from the estimated coefficients in the form of marginal WTP values (Goibov et al., 2012).More generally, the marginal rates of substitution between the different attributes can be calculated from the ratios of the coefficients.The marginal WTP value for public access calculated was for a change from limited recreational appeal to an improvement in the recreational appeal of the estuary through the introduction of a path access along the estuary banks.TABLE 3 reports the implicit prices, or marginal WTP, for each of the Sundays River Estuary's recreational attributes estimated using the Delta method (Wald procedure) in LIMDEP NLOGIT Version 4.0 (Greene, 2007).This procedure automates the process of estimating standard errors for non-linear functions, such as marginal rates of substitution (Suh, 2001).As parameter coefficients are stochastic in nature, confidence intervals are also estimated.These prices are based on the RPL model coefficients.
The WTP for more public access in the form of a path access or nature trail along the banks of the Sundays River Estuary was R34.00 per user per annum.

Conclusion
The Sundays River Estuary is a common property resource controlled by the state.It is not controlled by a marine protected area, a closed area or a national park but by the relevant municipality i.e. the Sundays River Municipality.Laws and regulations that relate specifically to this estuary are often breached and poorly enforced (Cowley et al., 2009).The absence of effective administration results in a situation whereby many of the advantages attributable to state management are lost, inducing problems of open access at the Sundays River, including high recreational demand.This recreational demand puts pressure on existing public facilities and the estuary space.Improving access to the various attractions of the estuary is one way to increase its accommodative capacity.
The current lack of public access to the Sundays River Estuary holds back its overall recreational appeal as a tourist destination.
In an assessment report by Afri-Coast Engineers (2004), it was recommended that the development of a nature trail or path access along the river banks be investigated.The introduction of a nature trail fronting the banks of the Sundays River Estuary would be an attractive complementary investment for both local residents and tourists.This investment would improve the recreational appeal of the estuary's banks and open up further areas for other recreational activities, such as bird watching and walking (Afri-Coast Engineers, 2004).To determine whether this project is efficient, however, information is required on the main group of current users' WTP, in the form of additional boat license fees, for the project, and costs involved for the projects implementation.The application of the CE method revealed that the marginal WTP for an investment in a nature trail was estimated to be R34 per user per annum.Although users are willing to pay R34 per annum to implement such a project, it cannot be determined whether this investment is efficient as project cost information was not collected.It is recommended that a cost-benefit analysis be conducted on the feasibility of this project at a later date.However, this finding does serve to provide guidance on how much funding could be allocated to such a development.
Total WTP for this development can be calculated by multiplying the number of registered boat owners, by the WTP per user per annum, which is R34.Given that in 2010 there were 900 boat users registered at the private Pearson Park slipway and about 2100 non-registered users (3000 in total), the present value (PV) of this project is: Annual total W T P Social Discount rate (10) Assuming a social discount rate of 8.38% (the average 10 years and over bond yield in 2010) it would have been efficient to spend about R1.22 million on this development in 2010 (South African Reserve Bank (SARB), 2013).