Do non-users value coral reefs?: Economic valuation of conserving Tubbataha Reefs, Philippines

Ecological Economics 102 (2014) 24–32

Contents lists available at ScienceDirect

Ecological Economics journal homepage: www.elsevier.com/locate/ecolecon

Analysis

Do non-users value coral reefs?: Economic valuation of conserving Tubbataha Reefs, Philippines☆ Rodelio F. Subade a,⁎, Herminia A. Francisco b a b

Division of Social Sciences, University of the Philippines Visayas, Miag-ao, Iloilo 5023-A, Philippines Economy and Environment Program for Southeast Asia, Worldfish Center, Penang, Malaysia

a r t i c l e

i n f o

Article history: Received 9 November 2006 Received in revised form 5 March 2014 Accepted 21 March 2014 Available online 15 April 2014 Keywords: Tubbataha Reefs Economic valuation Contingent valuation method Non-use values

a b s t r a c t The main purpose of the study is to determine whether non-use values exist among residents of Quezon City, hundreds of kilometers away from Tubbataha Reefs. The dichotomous choice contingent valuation method (CVM) was employed across 800 randomly selected respondents, 400 of which were personally interviewed (PI) and 400 were asked to accomplish self-administered (SA) questionnaires, 198 of the latter were found useable for the study. Results showed that 46% of all respondents were willing to pay for conservation of the reefs, with bequest motive or concern for future generations as their main reason. The mean WTP ranged from 437 pesos for PI respondents to 233 pesos for SA respondents. These substantial non-use values justify the need for regular government appropriation for conserving Tubbataha Reefs. © 2014 Elsevier B.V. All rights reserved.

1. Introduction In response to the global imperative of conserving biodiversity, the Philippines has joined the world-wide efforts to conserve various habitats in both the terrestrial and marine ecosystems through the enactment of the 1993 National Integrated Protected Area System Act (Republic Act No. 7586) and other legal instruments, which would establish protected areas across the country and arrest biodiversity loss. In 1999, the Haribon Foundation for the Conservation of Nature (Pajaro et al., 1999) tabulated a total of 439 established marine protected areas (MPAs) in the Philippines, but not more than 20% of these were fully implemented MPAs. Funding and the lack of institutional infrastructure have become the major limiting factors for fully implementing MPAs. Various funding agencies have strove to fill-in the gaps for implementation, but the gargantuan lack still persists. Foremost of the fully implemented MPAs is the world-renowned Tubbataha Reefs National Marine Park (TRNMP), a UNESCO World Heritage Site covering 33,200-hectare area in the middle of Sulu Sea. The TRNMP is an environmental resource believed to possess the highest level of marine biodiversity. It is the largest coral reef atoll in the country, well known among fishers in Southern Philippines, and one of the most popular dive sites around the world. TRNMP harbors a rich diversity of marine life equal to or greater than any such area in the world — more than 396 coral species, at least 45 families and 441 species of fish were recorded in the reefs (WWF — Philippines, 2004). ☆ Exchange rate during the time of the study was US$1 = PHP52. ⁎ Corresponding author. Fax: +63 33 338 1534, +63 33 513 7012. E-mail address: [email protected] (R.F. Subade).

http://dx.doi.org/10.1016/j.ecolecon.2014.03.007 0921-8009/© 2014 Elsevier B.V. All rights reserved.

Indeed, the Tubbataha Reefs is a very unique habitat worth preserving and conserving, and can serve as a great national pride and identity for Filipinos throughout the world. 2. Conserving the Tubbataha Reefs The TRNMP has attracted both perennial admiration and, sadly, habitat destruction particularly in the 1980s, when established management and conservation policy for these reefs were still lacking. Surveys showed that in the said period, fishing which often used destructive methods transpired in the reefs. In 1989, observations revealed that living coral cover on the outer flats declined by 24% (Arquiza and White, 1994). Like any resource, the open-access nature of Tubbataha Reefs prior to 1988, brought forth the wanton use and misuse of the resource therein. To arrest the downward trend towards habitat destruction and biodiversity loss, a series of efforts, declaration and legislation from different sectors and the government were undertaken in the 1990s highlighted by the issuance of Proclamation No. 306 on August 1, 1988 declaring it as a national marine park — the Tubbataha Reef National Marine Park (TRNMP) and UNESCO declaration as a UNESCO World Heritage Site on December 1993 (White, Vogt and Arin, 2000). The improving reef quality since 1989 was cited as an indicator of success of the above conservation efforts (White, Vogt and Arin, 2000). However, a major constraining factor in sustaining the conservation efforts is the continuous availability of needed resources, mainly funding. Fig. 1 shows that from an average of approximately US $50,000 yearly budget in 1996–1999, funding peaked to US$327,000 in year 2000, which is the start of the four-year GEF grant. With the

R.F. Subade, H.A. Francisco / Ecological Economics 102 (2014) 24–32

25

350000

CON FEES

300000

IN US DOLLARS

USAID 250000

PACKARD

200000

GEF-UNDP WWF-US

150000

KKP 100000 50000 0 1996

1997

1998

1999

2000

2001

2002

2003

2004

YEARS

Legend: WWF-US = Endangered Seas Campaign of the World-Wide Fund for Nature - U.S.A. KKP = Kabang Kalikasan ng Pilipinas (WWF-Philippines) GEFUNDP= Global Environmental Facility of the United Nations Development Program USAID = United States Agency for International Development PACKARD= Lucille Packard Foundation CON FEES= Conservation fees (from 2000-2003) saved through trust funds Source: Kabang Kalikasan ng Pilipinas (WWF-Philippines) Fig. 1. Funding sources for management and conservation of TRNMP.

grant-dependent nature of funding, conservation financing for 2004 and beyond was uncertain. Despite the effort to collect conservation fees from the users of TRNMP starting year 2000, this does not seem to be a stable, nor an adequate source of conservation funding. Fig. 2 shows that although users' fees have progressively increased up to 4.86 million pesos in 2006, it has remained inadequate as an only source of the annual budget for the conservation of Tubbataha, which stood at 6 million pesos or US$120,000. As per information from the Tubbataha Management Office, the annual budget has been shouldered because there are other partner institutions which provide supplemental or complementary funding of specific projects (like research). Moreover, the actual expense on the salaries of Philippines Coast Guard and Navy personnel, who have been assigned as marine park rangers, have been incorporated in the budget of those agencies. Resource insufficiency, particularly the need for a regular and bigger conservation budget, is highlighted by the huge area to be conserved, managed and patrolled, such that, Taiwanese and Chinese fishing boats were caught poaching in the area in 2000 and 2001, respectively. The evident lack of government commitment to provide consistent and sufficient funding for TRNMP conservation could be seen through the Presidential Task force which was disbanded due to lack of funds, and the limited naval patrols around the marine park. However, in order to sustain the momentum towards conserving and protecting the TRNMP, continuous source(s) of funding are needed.

Resource under-valuation or the failure of either the market or government to capture all the benefits of the natural resource, can lead to its mis-use, misallocation or ruin. Like any natural resource, foregone direct use benefits due to conservation are substantial social costs, which government or communities may not easily give up for biodiversity conservation. In addition, due to lack of information on how the citizens value conservation (non-use values), it might be difficult to justify consistent government budget allocation for conservation (Subade, 2007). Dixon and Sherman (1990) argued that since conservation benefits are only partially accounted for, smaller areas are protected than what is socially desirable. This market failure stems from difficulty to capture conservation benefits, thereby resulting to governments' inadequate budget allocation for management. “It is believed that the increasing environmental awareness by Filipinos over the importance of marine habitats, coral reefs ecosystems in particular, has produced a significant set of beneficiaries (of the marine conservation) who live hundreds of kilometers away from the site where protection or conservation activities take place. Once found that citizens value the conservation of a richly biodiverse habitat like the TRNMP, a strong basis for continuous government and social funding of the conservation efforts lends a strong rationale that cannot be ignored” (Subade, 2007). This paper focuses on how non-users value a distant biodiverse habitat, and determine their willingness to pay (WTP) for its conservation. 3. Methods

Fig. 2. Tubbataha RNMP revenues from diver fees.

Considering the dominantly non-use value and non-market nature of biodiversity conservation, or biodiversity per se, the contingent valuation method (CVM) stands out as the most appropriate economic valuation method used for this study. Mitchell and Carson (1989) expound in depth the various aspects of CVM, which may be employed to estimate values not intimately linked to use — for example, the desire of individuals to pass pristine natural environments on to future generations. They claim that CVM “is potentially capable of directly measuring a broad range of economic benefits for a wide range of goods, including those not yet supplied, in a manner consistent with economic theory” (Mitchell and Carson, 1989, p. 295). Moreover, the National Oceanic

26

R.F. Subade, H.A. Francisco / Ecological Economics 102 (2014) 24–32

and Atmospheric Administration (NOAA) panel of the United States, concluded that, “CVM studies … can produce estimates reliable enough to be the starting point of a judicial process of damage assessment, including lost passive values” (NOAA, 1993, Arrow et al. eds.). Pearce and Moran (1994) explain that interest in CVM has increased because it is the only means available for valuing non-use values and that the estimates obtained from well designed CVM surveys are as good as estimates from other methods. Various articles, and book/manuals have been published, stressing the strength that CVM has attained in estimating non-market values and also showing its practical application in estimating user fees for resource sustainability (Bateman et al., 2002; Padilla et al., 2005). In the Philippines, CVM has been used in determining entrance fees in the Hundred Islands National Park (Padilla et al., 1999), in setting up the two-tiered conservation fees to be paid by divers in Tubbataha National Marine Park (Mejia et al., 2000), and in determining the entrance and dive fees for Puerto Galera and Mabini–Tingloy areas (Padilla et al., 2005). These studies show that CVM results can be influential in effecting policy change towards conservation finance and sustainability.

N

Luzon

Quezon City South China Sea (West Philippine Sea)

Pacific Ocean

Visayas

3.1. Sampling, Survey Design and Implementation The stages involved in conducting a CVM study are: designing and pre-testing the survey, carrying out the main survey, estimating willingness-to-pay (WTP) and/or willingness-to-accept (WTA), bid curve analysis, data aggregation and final assessment (Mitchell and Carson, 1989). Pre-testing can be done thoroughly, particularly when the survey instrument would be used in a self-administered (mailed) format. Boyle et al. (1994), conducted a thorough pre-test of their questionnaire, i.e. up to four successive pre-tests and revisions, prior to the final survey, which solicited people's WTP in order to prevent deaths of migratory waterfowl. A multi-stage stratified sampling was conducted whereby the first stage involved the selection of three cities for the purpose of the bigger survey — Quezon City (a major urban area and the most populous city in the country, whereby various citizens from many parts of the country have migrated and settled); Cebu City (considered as the major city of Southern Philippines) and Puerto Princesa City (the capitol of host province of TRNMP). This paper however, presents the results for Quezon City, which is very far from TRNMP. The second stage was the random selection of barangays per city, while the third stage was the random selection of respondents from the sampling frame provided by the National Statistics Office. For this paper, a representative sample of 800 respondents, were surveyed across 40 barangays (20 per barangay) of Quezon City (Fig. 3). Out of these total respondents, 400 were personally interviewed (PI) for an average of 1 h per respondent, while 400 were asked to accomplish self-administered (SA) questionnaires containing the same information and questions. SA questionnaires were distributed to sampled respondents, followed up and collected between a few hours and up to two days after. Of the total questionnaires, 399 of PI and 198 of SA were found useable for this study, while the rest had items with no responses on key variable (or respondent did not bother to answer it at all). Prior to drafting the questionnaire, focus groups were undertaken to determine people's awareness and opinions concerning biodiversity and economic valuation of such. This also helped determine what people would think about contributing for biodiversity conservation and what would be an acceptable payment mechanism. The verbal protocol technique, as applied in CV (Kramer and Mercer, 1997; Manoka, 2001; Schkade and Payne, 1994) was undertaken as part of the preparatory steps before finalizing the survey questionnaire. It is a “think aloud” technique where the respondent thinks out loud by literally letting his thought speak for himself on a particular CV question (Manoka, 2001). There was no interaction between the interviewer and the

Sulu Sea Mindanao

TRNMP

Celebes Sea (Sulawesi Sea)

Fig. 3. Map of the Philippines showing Quezon City where the CVM survey was conducted and the Tubbataha Reefs National Marine Park (TRNMP), Sulu Sea.

respondent, except when the interviewer would just intervene on occasions when the respondent stops verbalizing for a few seconds. Pre-testing of the questionnaire was undertaken three times before its finalization. The first pre-test consists of 90 PI respondents, 30 from Muntinlupa, Metro Manila, 30 from Mandaue, Metro Cebu and 30 from a non-sampled barangay of Puerto Princesa City, Palawan. For the first SA pre-test, another 90 respondents, distributed at 30 per same PI sites, where surveyed and requested to accomplish the questionnaire. Second pre-test was undertaken in Calamba, Laguna, consisting of 45 PI respondents and 45 SA respondents. All these pre-tests, utilized the open-ended format of WTP question. A third PI pre-test was also undertaken using the dichotomous choice format of the WTP question. The WTP question used contribution to trust fund as payment vehicle. This was deemed most appropriate payment vehicle considering the present management arrangements for TRNMP. Further details on the survey, questionnaire, and methods are found in Subade (2005). 3.2. WTP Model and Estimation Technique The WTP model, a dichotomous, closed-ended or referendum CVM model, was specified following the work by Hanemann (1984), whereby a representative consumer's indirect utility function V (P, M, Q, S), states that his/her level of utility or satisfaction depends on price P, income M, socio-economic characteristic S or Si, and the quality of the

R.F. Subade, H.A. Francisco / Ecological Economics 102 (2014) 24–32

27

environment Q. The respondent will vote yes if he/she would pay to help conserve the TRNMP at a given price P, and expressed as:

WTP, the mean WTP using Hanermann's formula can be misleading (Whittington, 2003).



 1 0 V M−P; Q ; S N V M–0; Q ; S :

3.3. Empirical Model

ð1Þ

Eq. (1) shows that the respondent will answer or vote yes if the utility he will derive from improving the marine coral reefs of TRNMP (Q1) and paying the price P, is higher than not having improved TRNMP biodiversity (Q0) and paying the price P = 0. If V (P, M, Q, S) is the observable component of the utility, the probability of the respondent saying yes can be expressed as: Prob ðYesÞ ¼ Prob ½V





 1 0 M‐P; Q ; S þ ε1 N V M–P; Q ; S þ

ε0 ð2Þ

where ε1's are unobservable component of the utility. Assuming that the random variable ε1 follows a logistic probability distribution, this can be written as: Prob ðYesÞ ¼

1

1 þ e−Δ

ð3Þ

where −Δ = V (M − P, Q1, S) N V (M – P, Q0, S)thus, the non-use value benefit of the hypothetical market (to improve marine biodiversity at TRNMP via conservation) is defined as:
 1 V M–WTP; Q ; S

N


 0 V M–P; Q ; S :

ð4Þ

Haneman showed that with a linearly specified indirect utility function V (M – P, Q, S), then  Log

 Prob ðYesÞ ¼ α0 þ β1 P 1− Prob ðYesÞ

þ

β2 Q

þ

X

β1 S1 : ð5Þ

Parameters α0 and βi can be estimated parametrically using logistic regression technique which can be done with the use of any of econometric software LIMDEP or SHAZAM. The mean maximum WTP for conservation and improvement of marine biodiversity in TRNMP can be calculated by the formula: Mean maximum WTP ¼


i 1 h α0þβ2 Q þΣβi Si : 1n 1 þ e β1

ð6Þ

Haab and McConnel (2003) explain that when the pattern of responses for discrete choice models as the CVM model specified by Hanemann above is well behaved, the estimated mean willingness to pay will not be specially sensitive to the choice of distribution for the unobserved random component of preferences, or for the functional form of the preference function. This is particularly true when the empirical distribution of “no” responses to WTP question is monotonically increasing as the bid price goes up — i.e. more percentage of respondents per bid price answered no to higher bids. However, there are cases when the distribution can have substantial effect on the estimates of willingness to pay. Because of this sensitivity of WTP for some CV studies, a least-restrictive approach has been developed in estimating WTP, i.e., the Turnbull estimator, a distribution-free estimator. This imposes monotonicity restriction, and has become an appealing alternative WTP estimator to CVM researchers using the dichotomous choice/ referendum format. Moreover, it offers a conservative lower bound on willingness to pay for all non-negative distributions of WTP independent of the true underlying distribution (Haab and McConnel, 2003). In CVM studies where the empirical results show a small portion of respondents with very high WTP, while majority have very low or zero

The empirical model of this study is expressed as, WTP ¼

α0 þ β1 WTPATU þ β2 Q þ Σβi Si þ dm CVMMODE þ Ds SEX þ dintj INTEFFj þ dsek SAEFk

where Si are socio-economic variables (i = 3–5, as follows) such as, INCOME, KNOWIND, HEARD, FMMBIOD, and EDUC; dm is the coefficient of the dummy variable for CVM mode; dintj is coefficient for dummy for interviews, such that j = 1–50; and dsek is the coefficient of dummy variable for survey assistant's effect. The dependent variable is the respondent's reply to WTP question, whether yes or no. The independent variables and their definitions are shown in Table 1a, while their descriptive statistics are in Table 1b. 4. Results and Discussions 4.1. Willingness-to-pay Results Tables 2 and 3 show the distribution of WTP responses across bid prices. The general trend was that, “yes” (or percentage of “yes”) responses decrease as price bid went up. This follows a priori expected economic theory that as price goes higher less would be the quantity demand for the good, or that less will be the willingness to buy by consumers (respondents). The general pattern of lower “yes” (percentage of “yes”) responses as price bid increases is true for PI respondents except for some small spikes in bids of 200, 300 and 1000 pesos; and 100, 150, 500 and 1500 for SA. This small spikes counter the expected increasing monotonicity of the no responses as the offered price goes higher, thus requiring a distribution-free (Turnbull) estimator for computing the WTP in the latter part of this paper. In order to distinguish the components or rationale for positive WTP, those who responded “yes” to the WTP question were also asked to state their reasons or motives for such. The categorization of economic values or motives for willingness to pay in Table 4 is based from the work of Stevens et al. (1994), McConnell (1997) and Manouka (2001). Table 4 confirms the a priori expected outcome that since most (if not all) the respondents are off-site, their direct use values would be small if not nil. The most cited reason for positive WTP was “bequest value or motive”, followed by “existence value”, and by “good cause”. Respondents who indicated non-WTP were asked to check off their reason why. Of the total “no” replies across all sub-samples or survey sites, 52% and 59% for SA and PI, respectively, cited plainly economic reason for non-WTP, i.e. they could not afford to pay or give for the conservation trust fund due to limited income (Table 5). As cited in the pretests and focus group discussions, the survey results confirm that many respondents did not trust the institution(s), which will handle the money for the conservation work (reason number 5). It should be Table 1a Definition of variables used in the logit regression model. Variables

Definition

INCOME CVMMODE KNOWIND WTPATU SEX HEARD FMMBIOD EDUC INTEFFi SAEFFi

Respondent's annual income Survey mode; 1 if personal interview, 0 if self-Administered Knowledge index/score Bid price (willingness to pay amount for Tubbataha) Sex of respondent Heard about Tubbataha Familiarity with marine biodiversity (scale is 1 to 10) Education (number of years) Interviewer's effects (where i stands for interviewer number) Survey assistant's effects (where i stands for the asst's number)

28

R.F. Subade, H.A. Francisco / Ecological Economics 102 (2014) 24–32

Table 1b Descriptive statistics of regression variables of Quezon City respondents grouped by CVM mode, 2002. N

Minimum

Maximum

Mean

Std. deviation

Self administered survey AGE EDUC INCOME KNOWIND FMMBIOD HEARD (41.4%) SEX (female = 48.5%)

198 198 198 198 198 198 198

16 1 0 0 1 0 0

78 20 2,500,000 10 10 1 1

38.55 11.65 182,712.29 8.14 4.55 0.41 .48

13.96 2.92 256,189.06 1.59 2.5 0.50 0.50

Personal interview AGE EDUC INCOME KNOWIND FMMBIOD HEARD (16.3%) SEX (female = 60.4%)

399 399 399 39 399 399 399

18 1 12,000 0 1 0 0

95 25 3,600,000 10 10 1 1

40.82 10.82 176,374.77 8.0 3.66 0.163 .60

14.06 3.12 255,488.17 1.57 2.77 0.50 .49

noted here that the survey instrument did not delineate whether the institution(s), which will manage the conservation funds would be government or non-government since the Tubbataha Protected Area Management Board was composed of representatives from both sectors. 4.2. Variables Affecting WTP for TRNMP Conservation Possible effects by particular interviewers and SA survey assistants, in soliciting information from respondents were incorporated in the models through the use of dummy variables. In doing PI CVM surveys, Whittington (2002) mentioned that inconsistent results of CVM could be due to poorly-trained enumerators and the resulting interviewer bias. Though the student-interviewers for this study undergone a thorough training, the interviewer effect dummy variables are aimed at “cleaning” the model of the interviewer bias (Adamowicz, 2003). Desvouges et al. (1996) dealt interviewer bias, i.e. by using dummy variable per interviewer. The sign of INCOME in PI is consistent with a priori expected positive sign, i.e. higher income means a higher probability of saying “yes” to donating to the TRNMP conservation fund. Bid price (WTPATU) significantly determine WTP negatively for both CVM mode (Tables 5A and 5B). The expected a priori negative signs tell us that the higher the bid price, the lower is the respondent's WTP. The coefficient SEX was significant only for PI and has a negative sign, implying that women respondents tended to say “no” to the WTP question. Familiarity with marine biodiversity (FMMBIOD) is a significant determining variable for PI. On the other hand, education (EDUC) positively affected WTP in SA. Only one survey assistant (SAEFFi) was found to be affecting WTP in SA CVM mode. Table 2 Distribution of personal interview responses by Quezon City respondents to willingness to pay for biodiversity conservation of TRNMP, 2002. Bid price

NO

YES

Total

20 50 100 150 200 300 500 1000 1500 2000 Total

7 (18%) 14 (35%) 17 (44%) 25 (62%) 18 (45%) 23 (57%) 27 (67%) 24 (60%) 30 (75%) 32 (80%) 217 (54%)

33 (82%) 26 (65%) 22 (56%) 15 (38%) 22 (55%) 17 (43%) 13 (33%) 16 (40%) 10 (25%) 8 (20%) 183 (46%)

40 40 39 40 40 40 40 40 40 40 399

Note: Values in parentheses are percentage of last-column totals.

Table 3 Distribution of self-administered survey responses by Quezon City respondents to willingness to pay for biodiversity conservation of TRNMP, 2002. Bid Price

NO

YES

Total

20 50 100 150 200 300 500 1000 1500 2000 Total

3 (18%) 3 (20%) 9 (39%) 12 (52%) 9 (56%) 14 (64%) 12 (52%) 15 (79%) 15 (68%) 14 (78%) 106 (54%)

14 (82%) 12 (80%) 14 (61%) 11 (48%) 7 (44%) 8 (36%) 11 (48%) 4 (21%) 7 (32%) 4 (22%) 92 (46%)

17 15 23 23 16 22 23 19 22 18 198

Note: Values in parentheses are percentage of last-column totals.

4.3. Incorporating Respondent Uncertainty Boyle (in Champ et al., 2003) identified the inclusion of uncertainty in CVM models as one of the frontier issues in non-market valuation of environmental resources. As a less-explored approach in CVM, this will be useful in understanding how people answer contingent valuation questions, and why differences exist between CVM and actual transaction estimates. Incorporating uncertainty/certainty would actually qualify those “yes” replies to WTP question, which were actually “no's” because they were not that certain of their “yes”. To incorporate response uncertainty, all responses associated with certainty levels 1 to 7 were recoded as NO = 0, i.e. those who answered “yes” but had these certainty levels (of 1 to 7) actually answered “no”. Separate regression runs were also conducted on the model whereby certainty levels 1 to 9 were recoded as NO = 0, and also for certainty levels 1 to 8 recoded as NO = 0, but these regressions did not produce good results, i.e. very high standard errors amounting to millions or

Table 4 Respondents' most important reason for willingness to pay for conservation of Tubbataha Reefs, by CVM Modes, Quezon City 2002. Most Important Reason for Respondents' WTP

SA

PI

1. I want to preserve Tubbataha Reefs because I visit it (direct use value) 2. I want to preserve Tubbataha Reefs for future generations (bequest value) 3. I take personal pleasure in knowing that Tubbataha Reefs exist (existence value) 4. I would like to contribute because I am concerned about the people who depend upon the goods and services of Tubbataha Reefs (non-paternalistic altruistic motive) 5. I would like to contribute because the goods and services provided by Tubbataha Reefs should be available to others (paternalistic altruistic motive) 6. I do not use Tubbataha Reefs right now, but I am willing to contribute to have the option of visiting or using it in the future (option value) 7. I am contributing because marine plants and animals have the right to exist independent of anyone's use either in the present or the future (existence value) 8. It is a good cause and I enjoy contributing to good causes in general (good cause) 9. It is my moral duty to contribute to preserve Tubbataha Reefs (moral duty) 10. I want to preserve Tubbataha Reefs because I directly consume goods and services such as fish, etc. from it (direct use value) Other reasons

6 (7%) 31 (34%) 7 (8%) 1 (1%)

1 (0.6%) 89 (49%) 13 (7%) 10 (5%)

4 (4%)

11 (6%)

4 (4%)

3 (2%)

9 (10%)

25 (14%)

13 (14%) 5 (5%) 2 (2%) 3 (3%) 7 (8%) 92

14 (8%) 10 (5%) 1 (6%) 1 (0.6%) 4 (2%) 182

No answer Total yes votes

Note: Values in parenthesis are percentage of last row totals. Percentage per column may not add up to 10% because of rounding off.

R.F. Subade, H.A. Francisco / Ecological Economics 102 (2014) 24–32 Table 5 Most important reason for respondents' non-willingness to pay for conservation of Tubbataha Reefs, by CVM modes, Quezon City 2002 (scenario rejectors not excluded, all yes responses considered as certain). Respondents' reasons for non-willingness to pay

SA

1. I cannot afford to pay/I have no spare income

55 (52%) 2. As being far from the place I feel paying anything is irrelevant to me 3 (3%) 3. I do not believe paying will solve the problem 8 (8%) 4. I feel this improvement will take place without my contribution 13 (12%) 5. I do not trust the institutions who will handle the money for this 10 conservation work (9%) 6. It should be the government's responsibility since it has money 1 from tax (1%) Other reasons 7 (7%) No answer(s)/reply 9 (8%) Total respondents who were not WTP 106

PI 128 (59%) 11 (5%) 5 (2%) 15 (7%) 36 (17%) 4 (2%) 10 (5%) 8 (4%) 217

Note: Values in parenthesis are percentages of column totals.

even non-estimation of the regression models with the corresponding uncertainty assumption. Thus, this study was constrained to adopt certainty levels 8 to 10 as “Yes”, while those certainty levels 0 to 7 were counted as “No”. As discussed above, scenario rejectors (or protest votes) refer to those “no” replies to the WTP question but which are actually nonzeros due to other reasons for declining the offered price for WTP. They could actually be rejecting the scenario, which describes the hypothetical provision of the good (the conservation of TRNMP), and thus they can be removed from the regression analysis. To come up with a more suitable regression model it was deemed necessary to look at how the inclusion/exclusion of the scenario rejectors would affect the results. Moreover, it was also deemed necessary to examine how the simultaneous inclusion/exclusion of uncertainty in the model would affect the results. Table 5A Logit regressions results for Quezon City self-administered survey respondents, 2002 (n = 198). Variable

Constant INCOME AGE KNOWIND WTPATU SEX HEARD FMMBIOD EDUC SAEFF1 SAEFF2 SAEFEF4

Scenario rejectors excluded

Scenario rejectors included

Certainty model

Original model

Certainty model

Original model

SR out CERT in

SR out CERT out

SR in CERT in

SR in CERT out

−5.0137 (−2.5510) −0.00000030 (−0.3930) −0.00122 (−0.0800) 0.0922 (0.6410) −0.00292 (−4.1640) 0.0932 (0.2270) 0.5255 (1.4780) 0.1278 (1.4230) 0.2006 (2.2690) 1.9413 (2.1110) 0.7222 (0.8220) 1.7521 (1.4760)

−4.2514 (−2.5070) 0.00000104 (1.0870) −0.00969 (−0.6880) 0.10027 (0.8150) −0.00124 (−4.0330) 0.00826 (0.0220) 0.3493 (1.1440) −0.0229 (−0.2900) 0.2280 (2.9430) 2.3630 (2.9430) 1.5408 (1.9130) 1.5335 (1.4620)

−0.2399 (−0.9940) 0.000000013 (0.1090) −0.00142 (−0.6550) 0.0177 (0.9170) −0.000235 (−5.2560) −0.00598 (−0.1010) 0.0583 (1.2150) 0.01609 (1.3090) 0.0266 (2.4120) 0.2244 (1.8740) 0.0622 (0.5370) 0.2179 (1.3560)

0.3453 (−1.2510) 0.00000018 (1.3040) 0.00209 (−0.8450) 0.0242 (1.0930) 0.000223 (1.0930) 0.00696 (−0.1030) 0.0709 (1.2930) 0.000168 (−0.0120) 0.0413 (3.2810) 0.3680 (2.6890) 0.2335 (1.7650) 0.3395 (1.8500)

Note: Figures in parentheses are standard errors.

29

Thus, for the logit regressions and WTP estimations, the analyses can be classified into four models per CVM modes: 1. Scenario rejectors are not excluded, and uncertainty/certainty is not included in the model (SR in CERT out). 2. Scenario rejectors are not excluded, and uncertainty/certainty is included in the model (SR in CERT in). 3. Scenario rejectors are excluded, and uncertainty/certainty is not included in the model (SR out, CERT out). 4. Scenario rejectors are not excluded, and uncertainty/certainty included in the model (SR out CERT in). In all four models, the estimated coefficients for bid/price (WTPATU) were all consistent with a priori economic theory, i.e. negatively signed, and were all significant. This means that as bid price increases, there was less willingness to buy the given good (biodiversity conservation of TRNMP). Furthermore, except for two instance (SA SR out CERT in, and SA SR in and CERT in) the coefficient for INCOME was found to be positive, consistent with a priori economic theory for a normal good. This means that as income increases, the demand for the good increases, or the WTP for the good/services increases. Since there were four models (SR in CERT out, SR in CERT in, SR out CERT out, and SR out CERT in) it was important to select a preferred (logit) model per mode, based on which mean WTP could be computed. Based on the four possible logit models per sub-sample, the preferred/ selected model to be used for covariate analysis, and WTP estimation is the fourth model above, where scenario rejectors are excluded and that certainty/uncertainty is included in the model (SR out CERT in). It is the authors' opinion that this model provides a reliable and more realistic WTP estimate. The “no” replies of scenario rejectors, otherwise called protests, are actually non-zero “no's”, as such it does not make sense to include them in the analysis and taking them as zeroes. If these cases/observations with non-zero no's are included in the regression, it only means recognizing them as actual zero, and so the resulting WTP would be smaller. Moreover, the “yes” replies need to be qualified since many of those were actually given by respondents whose certainty of their reply is less than 8, and thus the “yes” virtually became a “zero”. Ideally the adjustment could have considered certainty replies below 10 as just like no, but the regression estimates generated were not good, same as in considering 8 to 0 as NO reply. 4.4. Mean Willingness to Pay Estimates Using coefficients based on regression results, and the corresponding mean of the variables, the mean WTP was computed following the formula of Hanemann. For example, for the SA CVM mode (see Table 5A, second column) h
i 1 f‐0:5:0137–0:000000302 ð182712:29Þþ… ::g ln 1 þ e −0:00292 ¼ PHP261:37:

Mean WTP ¼

Due to the long equation for logit model estimates, only the constant's parameter estimate, and the product of coefficient for income and the mean income, are written above, however it indicates (as implied by “…..”) that products of other variables' coefficients and the corresponding means are to be added to the subscript of e. The other mean WTP values as shown in Table 6 are computed in the same manner using the respective coefficients and means of variables used. Note that Table 6 also shows the mean WTP for the four different models. Although the selected/preferred model is SR out and CERT in, the computed WTP for the other three models per CVM mode is shown. As to incorporating uncertainty, i.e. qualifying the “yes” of respondents if indeed they are “yes” (referring to the 2nd and 4th columns of Table 6 and comparing the corresponding WTP value to that of the 3rd and 5th columns), it can be noted that when the certainty of

30

R.F. Subade, H.A. Francisco / Ecological Economics 102 (2014) 24–32

Table 5B Logit regressions results for Quezon City personally interviewed respondents, 2002 (n = 399). Variable

Constant INCOME AGE KNOWIND WTPATU SEX HEARD FMMBIOD EDUC INTEF1 INTEF2 INTEF3 INTEF4 INTEF7 INTEF8 INTEF9 INTEF35 INTEF36

Scenario rejectors excluded

Scenario rejectors included

Certainty model

certainty model

Original model

Original model

SR out CERT in

SR out CERT out

SR in CERT in

SR in CERT out

−0.7288 (−0.666) 0.00000168 (2.192) −0.01890 (−1.892) 0.05093 (0.595) −0.00106 (−4.385) −0.5506 (−1.994) 0.4442 (1.535) 0.09348 (1.921) 0.04503 (0.918) −0.5825 (−0.899) 0.1822 (0.319) −0.06555 (−0.113) −0.6194 (−0.951) −0.2514 (−0.447) −0.7236 (−1.170) 0.9597 (1.440) 0.5998 (1.106) 0.7686 (1.386)

−0.4566 (−0.418) 0.00000253 (2.375) −0.0233 (−2.387) 0.1003 (1.186) −0.00122 (−5.671) −0.5418 (−1.901) 0.3302 (1.039) 0.07486 (1.504) 0.08535 (1.726) −0.02531 (−0.040) 0.3420 (0.572) −0.1429 (−0.248) −1.3558 (−2.064) 0.8896 (1.513) −0.3196 (−0.532) 0.7103 (1.044) 0.2498 (0.441) 1.0759 (1.793)

−1.2990 (−1.234) 0.00000184 2.539 −0.0152 (−1.636) 0.0971 (1.167) −0.000973 (−4.203) −0.3045 (−1.188) 0.3740 (1.414) 0.07698 (1.680) 0.0166 (0.355) −0.5141 (−0.852) 0.1129 (0.213) −0.04105 (−0.075) −0.90069 (−1.489) −0.2276 (−0.433) −0.6372 (−1.090) 0.4144 (0.698) 0.7991 (1.566) 0.4717 (0.941)

−1.0436 (−1.059) 0.00000230 (2.675) −0.0164 (−1.917) 0.1426 (1.825) −0.00102 (−5.228) −0.2136 (−0.875) 0.20009 (0.756) 0.0489 (1.119) 0.03498 (0.795) −0.00747 (−0.014) 0.2496 (0.487) −0.04361 (−0.086) −1.5067 (−2.595) 0.7061 (1.419) −0.1846 (−0.353) 0.0919 0.162 0.5715 (1.143) 0.5952 (1.210)

Note: Figures in parentheses are standard errors.

respondents is considered (CERT in), mean WTP is estimated as the lower amount. In some cases, like the SA when scenario rejectors are excluded, mean WTP was reduced to almost a fifth of the mean WTP estimate (from PHP1010 to PHP261). Moreover, if scenario rejectors or “protest votes” are not remove from the regression (or if they are included in the analysis), which means counting their responses as zeroes

Table 6 Willingness to pay measures by average respondent, per site and CVM mode, according to assumption on scenario rejectors and respondent uncertainty, 2002. Scenario rejectors excluded

Scenario rejectors included

Incorporates

Uncertainty not

Incorporates

Uncertainty not

Uncertainty

Incorporated

Uncertainty

Incorporated

(SR out CERT in)

(SR out CERT out)

(SR in CERT in)

(SR in CERT out)

1026.12 780.00 339 182 (53%)

485.78 363.17 399 117 (29%)

874.34 638.63 399 182 (46%)

1010.99 754.99 166 92 (55%)

209.29 189.18 198 51 (26%)

829.94 618.21 198 92 (46%)

Quezon City personal interview Mean WTP 562.37 Tumbull WTP 437.28 Sample size 339 Yes response 117 (35%) to WTP Quezon City self-administered Mean WTP 261.37 Tumbull WTP 233.00 Sample size 166 Yes response 51 (31%) to WTP

instead of non-zero “no's”, WTP would be slightly underestimated. This is because there would be more zeros (no's) which pulled down the mean WTP compared to when these non-zero no's are excluded from the analysis. For example, with certainty incorporated in the model (CERT in), retaining scenario rejectors reduces mean WTP from PHP562 to PHP486 such as in PI. In summary, qualifying which of the “yes” replies are actually “no's” or incorporating uncertainty in the model (CERT in) results to lower (or more conservative) mean WTP. Moreover, retaining scenario rejectors will mean lower estimates of mean WTP. Since this study was conducted years ago, several CVM studies that included respondent uncertainty have been undertaken. Incorporating uncertainty in CVM studies allowed more reliable estimates for WTP (Moore et al., 2010), and mitigated hypothetical bias (Ready et al., 2010). Table 6 also shows that SA mean WTP estimates are much lower compared to their PI estimates. Several CVM researchers explained these higher estimates of personal interview CVM survey as compliance bias, i.e. respondents in the PI are somehow driven by their desire to please the interviewer. This can be an opined explanation to the difference, but the data gathered does not provide clear proof on this. It can also be opined that, as Whittington explains (2003), providing time to think to respondents, as in the case of the SA respondents gave them advantage of expressing a more real WTP since between the time of receiving the questionnaire to be answered and the time of returning it during collection, they would be able to consult family members as well as compute his/her family expenses and budget. Moreover, although PI and SA CVM modes should have been conducted to provide the same level of information to respondents, it could not be denied that PI provided opportunity for respondents to clarify/verify information. This could have been one reason why PI respondents gave higher WTP. Furthermore, recent CVM studies using dichotomous choice format, and which generated non-monotonically increasing WTP distribution results across price bids could have substantial effects on the estimates of (mean) WTP, thereby making them unreliable. However, to have a conservative estimate, a lower bound estimate of mean WTP can be derived using the Turnbull approach. Details of this are discussed in Haab and McConnel (2003). Using their formula and computing for each CVM mode and under each of the four models (assumption scenarios) resulted the corresponding Turnbull WTP as shown in Table 6. It should be noted that, as Haab and McConnel explains the Turnbull estimates are lower or more conservative.

4.5. Estimating the Social Benefits of TRNMP Conservation from WTP Results Tables 2 to 3 show non-monotonically increasing WTP distribution— i.e. as the offered price goes higher, the percentage of no replies per total of respondents (in the given price) did not consistently go higher. Thus, the Turnbull estimator is more fit to determine for social WTP. Additionally, taking the path of conservative WTP estimation, the Turnbull WTP has been proven to provide the conservative estimate. Using the SA Turnbull WTP, the total social WTP or social benefits of conserving TRNMP for Quezon City amounted to 112 million pesos, a big amount which can cover the annual costs of the present level of conservation efforts in TRNMP, i.e. PHP 10 million (Dygico, personal communication, 2004). On the other hand, if social WTP is computed based on the PI Turnbull estimate, the amount will be almost double, i.e. social WTP or benefits of PHP210 million in Quezon City (Table 7). Despite the big discrepancy between the two estimates, these amounts are much conservative compared to very huge preliminary estimates using the Hanemann's mean WTP estimates. Moreover, these estimates may be better than those of benefit transfer considering the accuracy issues and various steps needed (Ghermandi and Nunes, 2013).

R.F. Subade, H.A. Francisco / Ecological Economics 102 (2014) 24–32 Table 7 Estimation of social WTP or social benefits of conserving TRNMP based on Turnbull willingness-to-pay, 2002. Study site/city Number of households Turnbull WTP (2000 Census) SA PI

Quezon City

Social WTP SA

PI

(a) × (b)

(a) × (c)

(a)

(b)

(c)

480,624

233

437 111,985,392 210,167,263

The preceding empirical results found by this study affirmed the existence of non-use values for a highly biodiverse resource, as earlier hypothesized by the study. In particular, both SA and PI CVM modes confirm that Filipinos who live several hundreds of kilometers away from the TRNMP expressed substantial economic valuation of conserving such threatened environmental resource. 5. Conclusions, Policy Implications and Further Research This study provides empirical evidence on non-use values in the developing country context. In particular the respondents hold positive non-use values for biodiversity conservation of TRNMP. Across CVM modes, those who responded yes to the dichotomous choice WTP question towards contributing to a trust fund for biodiversity conservation of TRNMP amounted to 46% of 597 valid observations. Both PI and SA respondents had 46% of the respondents who replied yes to the WTP question. The lower than 50% “yes” replies is just consistent with the findings of other studies on non-use values (Kramer and Mercer, 1997; Seenprechawong, 2001). The main motives or reasons for WTP were dominantly non-use values. Among non-use motives, bequest value/motive (concern for future generations) was ranked highest, cited by 49% of total respondents. PI respondents had higher probability of agreeing to the WTP decision, compared to SA respondents. This was translated to higher mean WTP per year, which may be due to compliance bias, or due to the fact that SA respondents have time to think. Using the more conservative Turnbull WTP for the SA CVM, estimated mean WTP (per year) will be P233 pesos. In contrast, the Turnbull mean WTP for the PI CVM was P 437. The social WTP, which may be viewed as part of social benefit of conserving biodiversity in TRNMP was computed by multiplying the total number of households in Quezon City with the mean WTP. The total social WTP using the SA Turnbull WTP would amount to 112 million pesos (or 2.15 million U.S. dollars) per year. In contrast social WTP using the PI CVM Turnbull WTP would amount to 210 million pesos (4 million US dollars). Either of these estimates (112 million or 210 million) will dwarf the required cost of conserving TRNMP (presently 10 million pesos to cover the core costs). Thus, tapping a portion of non-use values provides potential source of conservation funding for TRNMP. Though the survey was conducted a decade ago, it is believed that citizens' non-use values for Tubbataha have remained, and may have possibly been increased by the increased awareness. Moreover, given the size of the estimated aggregate WTP, there could be political pressure to allocate more of general revenue to this protected area or protected areas in general. Various mechanisms or options, through corresponding policy can be employed in appropriating or capture of the non-use values (or just a portion of it) for conserving the TRNMP. Among these are: tax attached to property value, tax attached to utility bill, and a voluntary contribution. Another possibility is extracting some portion of the non-use values through the yearly residence tax. Just collecting a small 5 to 10 pesos surcharge will already mean a lot of money (46 to 92 million pesos) even if only 10% of the 92 million population would pay such residence tax surcharge. Though the limitation of the present study is the loose specification or non-specificity of the payment vehicle, this provides some leeway as to various ways by which the WTP may be

31

collected from people. Also, this hold prospects for actual collection of or a portion of WTP. Moreover, given the magnitude of the difference between social WTP and the costs of running the park, the nonspecificity of payment vehicle may be of less significance. Acknowledgments This study was undertaken through a dissertation research grant from the Economy and Environment Program for Southeast Asia (EEPSEA). The first author is grateful to the technical advice of Dr. Vic Adamowicz and Dr. Dale Whittington. The dissertation study was conducted during the PhD studies of the first author, with scholarship support from the Southeast Asian Center for Graduate Studies and Research in Agriculture (SEARCA). Ana Liza A. Subade did the technical editing of this article. References Adamowicz, V., 2003. Personal Communication. Arquiza, Y., White, A., 1994. Tales from Tubbataha: Natural History, Resource Use and Conservation of the Tubbataha Reefs, Palawan, Philippines. Rainee Trading and Publishing, Inc., Mandaluyong City. Arrow, K., Solow, R., Leamer, E., Portney, P., Randear, R., Schuman, H., 1993. Report of the National Oceanic and Atmospheric Administration (NOAA) on the contingent valuation method. Fed. Regist. 58 (10), 4602–4614. Bateman, I.J., Carson, R.T., Day, B., HAnneman, M., Hanley, N., Hett, T., Jones-Lee, M., Loomes, G., Mourato, S., 2002. Economic Valuation with Stated Preference Techniques: A Manual. Edward Elgar, Cheltenham. Boyle, K.J., Desvouges, W.H., Johnson, F.R., Dunford, R.W., Hudson, S.P., 1994. An investigation of part-whole biases in contingent valuation studies. J. Environ. Econ. Manag. 27 (1), 64–83. Champ, P.A., Boyle, K.J., Brown, T.C., 2003. A primer on nonmarket valuation. In: Bateman, I.J. (Ed.), The Economics of Non-Market Goods and Resources. Kluwer Academic Publishers. Desvouges, W.H., Hudson, S.P., Ruby, M.C., 1996. Evaluating CV performance: separating the light from the heat. In: Bjornstad, D.J., Kahn, J.R. (Eds.), The Contingent Valuation of Environmental Resources: Methodological Issues and Research Needs. Edward Elgar, Cheltenham. Dixon, J., Sherman, P., 1990. Economics of Protected Areas: A New Look at Benefits and Costs. Earthscan Publication Ltd., London. Dygico, M., 2004. Personal Communications. Ghermandi, A., Nunes, P.A.L.D., 2013. A global map of coastal recreational values: results from explicit meta-analysis. Ecol. Econ. 86, 1–15. Haab, T.C., McConnel, K.E., 2003. Valuing Environmental and Natural Resources: The Econometrics of Non-market Valuation. Edward Elgar Publishers, Cheltenham, UK. Hanemann, W.M., 1984. Welfare evaluations in contingent valuation experiments with discrete responses. Am. J. Agric. Econ. 66, 332–341. Kramer, R.A., Mercer, D.E., 1997. Valuing a global environmental good: U.S. residents' willingness to pay to protect tropical rain forest. Land Econ. 73 (2), 196–210. Manoka, B., 2001. Existence value: a re-appraisal and cross-cultural comparison. Research Report. The Economy and Environment Program for Southeast Asia (EEPSEA), Singapore. McConnell, K.E., 1997. Does altruism undermine existence value? J. Environ. Econ. Manag. 32, 22–37. Mejia, M.N., Dygico, M., Spergel, B., Subade, R.F., 2000. Paying for marine conservation through sustainable financing: the Tubbataha experience. Paper Presented in the International Coral Reefs Symposium Held in Bali, Indonesia. Mitchell, R., Carson, R., 1989. Using surveys to Value Public Goods: the Contingent Valuation Method. Resources for the Future, Washington D.C. Moore, R.C., Bishop, R.C., Provencher, B., Champ, P.A., 2010. Accounting for respondent uncertainty to improve willingness-to-pay estimates. Can. J. Agric. Econ. 58, 381–401. Padilla, J.E., Rosales, R., Predo, C., Baun, S., Amponin, A., Meñado, A., Blastique, T., Cabrera, S.J., Buen, R., 1999. A Report on the survey of tourists and resorts at Hundred Islands National Park. ENRAP IV Technical Paper. Department of Environment and Natural Resources. Philippine Economic-Environmental and Natural Resources Accounting System. Padilla, J.E., Ansula, A.D., Tolosa, M.O., 2005. Getting Users to Pay for Conservation: A Guide to Site-based Sustainable User Fee Schemes. WWF_Philippines. Pajaro, M., Olano, F., San Juan, B., Nozawa, C.M., 1999. Inventory of marine protected areas in the Philippines. In: Uychiaoco, A., Schoppe, S., Aliño, P., Hermes, R. (Eds.), Marine Protected Areas in the Philippines: Proceedings of the Workshop on Marine Protected Areas in the Philippines. PCAMRD, Los Baños. Pearce, D., Moran, D., 1994. The Economic Value of Biodiversity. Earthscan Publications Ltd, London (172 p). Ready, R.C., Champ, P.A., Lawton, J.L., 2010. Using respondent uncertainty to mitigate hypothetical bias in a stated choice experiment. Land Econ. 86 (2), 363–381. Republic Act 7586. National Integrated Protected Area system Act. Congress of the Philippines. Schkade, D.A., Payne, J.W., 1994. How people respond to contingent valuation questions: a verbal protocol analysis of willingness to pay for an environmental regulation. J. Environ. Econ. Manag. 26, 88–109.

32

R.F. Subade, H.A. Francisco / Ecological Economics 102 (2014) 24–32

Seenprechawong, U., 2001. An economic analysis of coral reefs in the Andaman sea of Thailand. Research Report. No. 2001-RR7. Economy and Environment Program for Southeast Asia, Singapore. Stevens, T.H., More, T.A., Glass, R.J., 1994. Interpretation and temporal stability of CV bids for wildlife existence: a panel study. Land Econ. 70 (3), 355–363. Subade, R.F., 2005. Valuing biodiversity conservation in a world heritage site: citizens' non-use values for Tubbataha Reefs National Marine Park, Philippines. Research Report No 2005-RR4. Economy and Environment Program for Southeast Asia, Singapore. Subade, R.F., 2007. Mechanisms to capture economic values of marine biodiversity: the case of Tubbataha Reefs UNESCO World Heritage Site. Philippines. Mar. Policy J. 31 (2), 135–142.

White, A.T., Vogt, H.P., Arin, T., 2000. Philippine coral reefs under threat: the economic losses caused by reef destruction. Mar. Pollut. Bull. 40 (7), 598–605. Whittington, D., 2002. Improving the performance of contingent valuation studies in developing countries. Environ. Resour. Econ. 22, 232–367. Whittington, D., 2003. Personal Communication. WWF — Philippines, 2004. Tubbataha Reef National Marine Park and World Heritage Site. State of Conservation Report.