Valuing Recreational Use of Pakistan’s Wetlands:
Application of a Count Data Model tothe Travel Cost Method
ALI DEHLAVI
World Wide Fund for Nature – Pakistan (WWF-P),
Indus for All Programme,
Karachi, Pakistan
October 2009
South Asian Network for Development and Environmental Economics (SANDEE),
P.O Box: 8975, EPC-1056,
Kumultar, Lalitpur, Nepal
SANDEE Working Paper (Draft submitted to the 19th R&T)
Contains preliminary results of 420 out of 750 observations in the dataset
Co-investigator: Iftikhar Hussain Adil
Table of Contents
Abstract
- Introduction
- Justifying Ecotourism Investment: Answering Economic and Financial Questions
- Hypotheses / Research Questions
- Methodology
- Data and Sampling Strategy
- Summary Statistics of the Major Variables
- Theoretical Ideas Underlying the Study and Methods for Data Analysis
- Preliminary Analysis
Annex I: References
Annex II: Sampling Plan
Annex III: Summary Statistics of the Major Variables
Annex IV: Map of Number of Visitors and Visitors per Capita (28.2.09 – 6.3.09)
Annex V: Key Informant Interviews
Annex VI: Survey Instrument
Abstract
The Global 200 scientifically ranks outstanding terrestrial and aquatic ecosystems in 238 ecoregions worldwide. Keenjhar Lake, Pakistan’s largest freshwater lake and a Ramsar site, falls within the Lower Indus Basin of the Indus Ecoregion, one of the 40 priority ecoregions of the Global 200. Although visits to the lake’s tourism resort can exceed one thousand a day, this represents a fraction of Sindh’s population of 55 million. In such a circumstance, effective sampling frame construction for a travel cost model (TCM) is expensive. Using a site-based sample, the present paper applies Poisson and negative binomial count models to address the integer quality of trip demand andthe endogenously stratified nature of the sample. The trade-off between further improvements in water quality and other competing wants is addressed through measuring the economic value of changes in site quality. Access value estimateswill supply planners with a recreational value of land for comparison with the value of competing uses.Together withexamining the significance of on-site activity choice in determining costs and benefits resulting from an away-trip, the combined analysis aims to set a policy precedent to replace the current use of rules of thumb used to justify investments in ecotourism.
Key words: travel cost method, count model, freshwater lake in Pakistan, welfare measurement at single-site, value of site quality changes, ecotourism
- Introduction
Pakistan’s province of Sindh sits in the Lower Indus Basin of the Indus Ecoregion. From 2007-2012 the World Wide Fund for Nature – Pakistan (WWF-P) is working to conserve key wetlands in this Ecoregion which is also one of the 40 priority ecoregions in the world identified by the Global 200[1]. The Indus for All Programme of the WWF-P is, in particular, responsible for protecting human health and livelihoods along with biodiversity in a delta (Keti Bunder, Thatta District), a desert wetland reservoir complex (Chotiari, Sanghar District), an irrigated forest plantation (Pai, Shaheed Benazirabad District, formerly Nawabshah District), and a freshwater lake (Keenjhar, Thatta District). Species variety is high in the sites. As a whole, the Indus Ecoregion has 329 species of birds, 72 reptile species, 57 species of mammals, five amphibian species, and, even without including brackish water lakes and fresh water species, there are approximately 179 species of freshwater fish fauna (Ghalib and Bhagat, 2002).
The total economic value (TEV) of Keenjhar lake, based on a recent estimate of direct consumptive use value (fishing), indirect use value (residential water supply to 1 million out of Karachi’s population of 15 million), and non-use value (based on a choice experiment administered in Karachi examining willingness to pay for species protection) is in the order of Rs. 9 billion (Dehlavi et al. 2008, forthcoming)[2]. The figure, a net present value estimated over a limitless time horizon for which a sensitivity analysis is presented, is taken from a study jointly conducted by WWF-P and the School of Oriental and African Studies, University of London. The studyestimates the TEVs of four other ecosystems besides Keenjhar’s freshwater ecosystem. In discussing the application of TEV estimatesfor the purpose of modifying Pakistan’s national income accounts, the authors note that tourism – which was omitted in the study’s analysis of Keenjhar – can significantly augment direct use value estimates. They cite a recent study based on the Okavango Delta in Botswana in which Social Accounting Matrix (SAM) based gross national product (GNP) multiplierswhen estimated for tourism were found to be significantly greater than those estimated for household agricultural and natural resource harvesting / processing activities (Turpie et al., 2006).
Policy receptiveness both in terms of accepting the validity of valuation study results and applying them has accelerated in the past five years. At the federal level, a 2007 Planning Commission “Capacity Building Project” included “Environmental Economic Valuation / Green Accounting” among its six project components. There is strong government acceptance of the World Bank’s 2006 “Pakistan Strategic Country Environmental Assessment” study and its TEV estimates of both ecosystems (rangelands, forests, soil salinity and erosion) and environmental parameters (water, urban air, airborne lead and indoor air) which place the cost of environmental degradation at 6 per cent of gross domestic product(World Bank, 2006). Also, between 2009-2010, similar to the United States’ National Oceanic and Atmospheric Administration’s contingent valuation guidelines (Federal Register, 1993), the Ministry of Environment and the National Forest Programme (NFP) Facility, funded by the Food and Agriculture Organization of the United Nations, is preparing best practice guidelines for all valuation techniques as applied to the forestry sector[3]. At the provincial level,Planning and Development Department (P&DD), the Government of Sindh, has considered using valuation study estimates to inform budgeting in their planning cycles. It has however first requested that a valuation study of the marine fisheries sector be conducted and measurable progress demonstrated through, among others, the periodic compilation of raw data series by the Statistical Division / Federal Bureau of Statistics, as needed (WWF-P, 2009).
The aim of this paper is toset a precedent to replace the current use of rules of thumb used to justify investments in ecotourism. It proceeds by estimating access values using a travel cost model (TCM), thereby supplying planners witha recreational value of land for comparison with the value of competing uses. Economic values associated with changes in site quality are also quantified in order to assist Keenjhar’s tourism operation managers to evaluate trade-offs between further improvements in water quality and other competing wants.They will also be helped bythe paper’s examinationofsourcesof on-site activity choice to better target their marketing and determine their pricing.A site-based sample was selected owing to resource constraints so that the TCM applies Poisson and negative binomial count models to address the integer quality of trip demand and the endogenously stratified nature of the sample.
- Justifying Ecotourism Investment: Answering Economic and Financial Questions
Pakistan’s largest freshwater lake (14,000 ha), Keenjhar, also a wildlife sanctuary and Ramsar site, hosts a resort managed by the Sindh Tourism Development Corporation (STDC).It is located in Sindh province’s Thatta district, about 120km from Karachi (estimated 2008 population of c. 15 million). The lake is 24 km long and 6 km wide, with a maximum depth of 8 meters. The lake is fed by the Kalri Bagar feeder canal which enters at the northwest corners, and by small seasonal streams entering on the western and northern shores. The lake’s only outlet is through the Jam branch canal at the southeast corner of the lake. Towards the southwestern corner of the lake the Karachi Water Sewerage Board (KWSB) has its own set up to regulate the outlet of the lake for agricultural uses but also supply of water to Karachi, among others.
The STDC, a public limited company, was established in 1990. Its Fiscal Year 2004-2005 grant-in-aid was PKR 2.5 mn. According to STDC, the government’s inability to provide it sufficient funds is to blame for its “financial crisis”, its inability to run as a stand-alone corporation, and it has recently recommended a one-time PKR 20 mn grant ( Priorities placed on government spending are frequently reassessed with rapid transfers in the provincial administration. One of the areas of expenditure called into question is overnight accommodation facilities, of which there are 7 in Thatta, Dadu, Mithi and Larkana districts. Pressure to justify money received by STDC for these and other recreational services will only mount as per capita incomes increase.
At present, STDC does not employ valuation or similar advanced quantitative techniques in their planning or pricing of accommodation and recreational activities. Models of recreational demand can be put to a number of uses. Of particular interest in this case is their use to answer economic questions (e.g., measuring the welfare derived from the reserve), to supply policy recommendations arising from public perceptions of polluted bathing water, and to answer financial questions (e.g., responsiveness to cost components of that have bearing on revenue per unit of on-site paying activities). While the latter is a pertinent issue for reserve managers attempting to maximize revenues, relatively few recreational models have examined effective targeting of marketing and pricing efforts within a rangeof on-site paying activities (e.g., type of lodging to promote andoptimal length of stay). Day (2000) and Adhikari et al. (2006) estimate TCMs that incorporate just such dimensions specifically to address planning at local levels in Kwazulu-Natal Province and Louisiana State, respectively. The application in Natal tests both a multinomial and a 3-level nested multinomial logit model (NMLM) to compute welfare estimates for continued access to different game reserves. The Louisiana study treats total recreational time as two separate variables, finding that time spent at the site has no significant impact on the probability of choosing a specific activity from a choice set while travel to the site was significant and had a negative sign for all activities.
As regards policy recommendations arising from public perceptions of polluted bathing water, few recreational demand models have based such analyses on time-series data (Haab and McConnell, 2002). The present study uses cross-section data so that changes in quality variables over time are not observed; however,averting behavior of frequent visitors is observed for given time periods in the year based on scientific recordings of site-quality corroborated by key informants. For example, recordings from monitoring stations at the lake (station 2 in particular which corresponds to Helaya village in the immediate vicinity of the 2km STDC resort) provide interesting information we have used to link environmental quality variables in our study (see Figure 1 below). In particular, we note that for recordings carried out in 2005, both transparency and dissolved oxygen are higher for October as compared to June-July (Korai et al., 2008a). Findings borne of analyses of bathing water quality or water quality generally are not without significant policy relevance. One notable economic and epidemiological contingent valuation (CV) valuation survey undertaken at Lowestoft and Great Yarmouth established that the British public was prepared to pay above the clean-up cost of attaining European Community standards at all of Britain’s beaches (GBP 9 billion in 1995) (Georgiou et al., 1996).Willingness to pay was in fact similar to the costs of attaining more stringent requirements outlined in a House of Lords Committee report that considered the issue in February of 1996 and called for just such an assessment of the benefits of cleaning up Britain’s coastal waters.
Source: Korai et al, 2008b
As mentioned above, the policy relevance of findings relating to water quality in general, as opposed to bathing water quality alone, is also important and this is indeed the case for Keenjhar Lake. Karachi is dependent for as much as 80% of its total water supply for both commercial and residential user categories. In 2007 1 million residential users were connected to Karachi’s reticulation system. Estimates place the dependent population on the lake at 50,000 persons, mainly inhabitants of the surrounding twelve large and twenty small villages (WWF, 2006). Nevertheless, there are several sources of pollution at Keenjhar Lake. For the past 10 years at least, tanneries situated in Hyderabad district have been draining effluents into the lake. Recreationists as well as owners of commercial vehicles engaged in lake construction work, for example, wash their vehicles on a daily basis, contributing thereby to grease for which a major source is also motorized fishing boats. A major source of pollution is also the effluent load expelled by the upstream industrial areas of Nooriabad and Kotri. Units at Kotri dispose of their wastes in the Kalri Bagar feeder which is Keenjhar’s feeding source. Upstream industrial effluents are transported during the Monsoonal period of June-July into the lake at which time eutriphication is also highest. Furthermore, water levels are typically low during the month of October just before water is allowed into the lake for the aforementioned multiple uses.
This paper aims further to answer the economic question of what is the welfare derived from STDC’s reserve. As noted earlier, a welfare estimate based on recreational use of the lake will serve on the one hand to augment Keenjhar’s existing total economic value (TEV) estimate of Rs. 9 bn (Dehlavi et al. 2008, forthcoming). On the other hand, the welfare analysis itself is expected to aid management decisions. That is, for every observation in the data set, the paper examines the question of whether the opportunity to visit other destinations en route to Keenjhar is truly essential to visitors’ decision to journey to Keenjhar. By examining the compensating variation for a large number of visitors from the sample who made such incidental visits it is possible to compute expected compensating variation – i.e., the amount of money that when given to households after closure of the STDC reserve would return them exactly to their original level of utility. This reveals in general whether and to what extent a welfare loss is expected from a particular recreational alternative; the magnitude of welfare loss if any to those arriving to Keenjhar from different districts and towns; and, whether, in view of road networks and the relatively long or short distances covered, certain sections of Sindh’s populations are likely to value Keenjhar highly.
As confirmed by gate count conducted during a reconnaissance survey (28.2.09 – 6.3.09), visits to the lake’s tourism resort can exceed one thousand a day. However, this still only represents a fraction of Sindh’s population of 55 million. In such a circumstance, effective sampling frame construction for a TCM is expensive.Due to resource constraints, in our case a site-based sample is used so that number of visits, the dependent variable in our regression analysis, takes on non-negative integer values, or, counts. Consideration of aggregate data in country-level TCMs and inclusion of zero-demanders in count models has nevertheless been considered (Hellerstein, 1991).Count models with truncated samples – i.e., when only those visiting the site are sampled – must make use of appropriate functional form but also be observant of the effects of functional form choice and truncation on consumer surplus estimates (Ozuna et al., 1993). The class of permissible functions depends on the distribution assumed and the Poisson regression model (PRM) is commonly used in recreational demand models (von Haefen and Phaneuf, 2003). The PRM is however subject to misspecification owing to its implicit restriction on the number of counts – the variance is constrained to equal the mean – so that researchers typically use more general specifications, usually the negative binomial (NB) model (Greene, 2005). Aside from truncation, our TCM model addresses overdispersion – i.e., when variance is greater than the mean – and endogenous stratification – i.e., the case where the sample average number of trips is expected to be higher than the population mean due to the on-site interviewing process being inherently likely to intercept avid visitors. This study is unique in at least two senses. Firstly, to our knowledge, this is the first TCM being applied to freshwater in Pakistan. Secondly, survey instrument design partitioned travel costs into within-city travel to a common point of departure and travel onwards to Keenjhar for those using rented / shared transport. This considerable design effort and risk of respondent fatigue is outweighed by the fact that travel distances are non-trivial within Karachi Districtfrom where this category of transport usersoriginate and is dominantbut also because visits from Karachi can represent up to 80% of total visits to Keenjhar. This design has the potential for replicability and improvement in South Asia where travel from urban centers to recreational resortsusing shared transport is certainly common.
- Hypotheses / Research Questions to be Answered
The paper applies a TCM to Keenjhar Lake to attempt to answer the following specific research questions:
- What is the average annual access value of the site?The analysis should generate estimates of annual net flows of social values to enable comparison to alternative land uses; and, enable prediction of changed visitation in response to simulated changes in entrance fees.
- Are visitors responsive to changes in on-site water quality? The analysis should indicate if changed environmental quality at given time periods in the year leads to fewer recreational trips as well as circumscription of swimming / wading, boating, rubber tube and swimming costume rental. A welfare calculation based on compensating variation associated with assumed changes in water quality is attempted.
- To what extent is the choice of on-site activity (both paying and non-paying) determined by visitors’ household characteristics? The analysis will regress household characteristics on selection of on-site activities.
The relevance of the first research question (above) is to answer the economic question of what is the welfare derived from STDC’s reserve. This can be added to existing TEV estimates destined to permit planners to decide whether to allocate land to wilderness or development. Even if the value may not be applied in a benefit cost analysis in the context of Keenjhar since it already benefits from a protected status as a Ramsar site, the exercise will be useful for its replicability within the Indus Ecoregion and elsewhere in Pakistan. Simulated increases in entrance fees would aid management decisions relating to optimizing revenue in order to move STDC out of its present financial crisis and towards becoming a stand-alone, self-sustaining corporation. Further, by examining the compensating variation associated with incidental visits en route to Keenjhar it is possible to compute expected compensating variation – i.e., the amount of money that when given to households after closure of the STDC reserve would return them exactly to their original level of utility. This would reveal whether and to what extent a welfare loss is expected from particular recreational alternatives; the magnitude of welfare loss of visitors from different districts and towns; and, which specific sections of Sindh’s populations are likely to value Keenjhar highly.