Cost and compensation issues in protecting tropical rainforests : case study of Madagascar

A/Voo Z THE WORLD BANK ENVIRONMENT DEPARTMENT AFRICA TECHNICAL DEPARTMENT T**batcal Department Afris. Region bo World 8o COST AND COMPENSATION ISSUES IN PROTECTING TROPICAL RAINFORESTS: CASE STUDY OF MADAGASCAR Randall A. Kramer Narendra Sharma Priya Shyamsundar Mohan Munasinghe January 1994 Environment Working Paper No. 62 This paper has been prepard for internal use. The views and interpretations herein are those of the author(s) and should not be attributed to the World Bank, to its affiliated organizations or to any individual acting on their behalf. ACKNOWLEDGEMENTS AND DISCLAIMER The authors are, respectively, Associate Professor, School of the Environment, Duke University; Principle Economist, World Bank; Visiting Assistant Professor, School of the Environment, Duke University; and Division Chief, World Bank. The authors would like to thank the following for their advice and assistance at various stages in the study: Emmanuel Asibey, Ajit Banerjee, Sofia Bettencourt, Kevin Cleaver, Arne Dalfelt, Jean Doyen, Francois Falloux, Miguel Grut, Claude Heimo, Michael Jacobsen, Walter Lusigi, Simon Rietbergen, Shane Rosenthal, Raymond Rowe, and Shivsharan Someshwa. This work was supported in part by generous grants from the Governments of Norway and Sweden. Departmental Working Papers are not formal publications of the World Bank. They present preliminary and unpolished results of country analysis or research that are circulated to encourage discussion and comment; citation and the use of such a paper should take into account its provisional character. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors and should not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or to the countries they represent. Because of the informality, and to present the results of research with the least possible delay, the typescript has not been prepared in accordance with the procedures appropriate to formal printed texts, and the World Bank accepts no responsibility for errors. ABSTRACT Development projects for protecting rain forests and other ecologically important ecosystems often have considerable impacts on local residents. This study examines the cost to villagers of establishing the Mantadia National Park In the eastern rain forest of Madagascar. The park contains a wide variety of plant and animal species which are threatened by the high rate of deforestation in the country. The park belongs to a system of protect I areas designed to preserve the country's dwindling biodiversity. The villagers surrounding the park are dependent on the forest for a significant part of their livelihood. The forest provides fuelwood, traditional medicines, construction materials and foraged food. In addition, the forest provides opportunities for new land to be used for shifting cultivation. Thus, while the villagers may obtain some long-term benefits from the park, e.g., watershed protection, there is a significant opportunity cost associated with setting aside land that can no longer be exploited for direct use. Two methods are used to estimate the economic impats on the villagers: (1) opportunity cost analysis based on household cash flow models constructed from a socioeconomic survey; and (2) contingent valuation analysis based on direct questioning of villagers about required levels of compensation. The estimated costs to the villagers based on the different methods were quite similar. The results suggest that an annual compensation of approximately $100 per household would be required. Such compensation could be made in .'e form of education, health facilities, alternative Income earning enterprises in the buffer zone, or other development activities. These compensation costs appear to be a significant part of the true cost of Implementing protected area projects and should be built into project design at an early stage. Without adequate compensation and active cooperation of local residents, natural resource management projects are more likely to fail. COST AND COMPENSATION ISSUES IN PROTECTING TROPICAL RAINFORESTS: CASE STUDY OF MADAGASCAR TABLE OF CONTENTS INTIRODUCTION ................................................. I Objectives .................................................. 2 General Approach ............................................. THEORETICAL. RAMEWORK ........................................ 3 A HouseholdProduction Model .................................... 3 W elfare Im plications of E stablishing the N ational P ark . . . . . . . . . . . . . . . . . . . . 6 METHO ODO OGY ................................................. 8 Surv ey Developmet ........................................... Data Ma agemes ............................................. 10 THEPEOPLE AND THEPARK ........................................ II Soci-EEnomcDifferencesAmongteAV g ......................... 11 General Socio-Economic Characteristics ................... . . . . . . . . 11 EconomicActivities .................................. ........ 12 Villager'Percptionof .rests ................................... 13 CASH .OW1ANALYSIS ............................................ 14 Objectives ............................. ................... .. 14 Dam Used ...........*................ :....................... 14 ForestProducts ........................... ......... ... 14 Agricultul Products ...................................... 14 Strategy and Assumptices ........................................ Is Results .................................................... 16 Aggregation ................................................ 18 CONTINGENT VALUAIONNANALYSIS.................................. 19 Ob jet ive .................................................. 19 IssuesoflRelevanceandlStrategy .................................... 19 Theoretical and Econometric Specification for the Dichotomous Choice Analysis ...... 21 Data use ............................................. 21 Results ............................................... 22 The MeanWillingness to Accept .................................... 25 CONCLU1SIONS AND RECOMaNDA1TONS .............................. 26 EmpiricalResults ............................................. 26 ApplicationofMethodology ...................................... 27 Lessonamn ............................................. . 27 Broadermplications ........................................... 28 Costs and Compensation Considerations .......... . ... ........ .*.... 2$ REFERENCES ................................................... 30 COST AND COMPENSATION ISSUES IN PROTECTING TROPICAL RAINFORESTS: CASE STUDY OF MADAGASCAR INTRODUCTION Tropical countries in Africa want to give more attention to better management and protection of intact rainforests. Preservation of tropical rainforests, either as national parks, wilderness, or protected reserves, has significant implications in terms of the three main objectives of sustainable development: social, economic, and environmental. Protecting forests gives rise to benefits in terms of conservation of biodiversity and maintenance of environmental services. But there are also some negative impacts, as well, particularly the social cost borne by many people who depend on these forests for their livelihoods. People living adjacent to rainforests depend on these forests for their basic needs, including food, fuelwood, fodder, and small timber. They also use these forests for subsistence farming through shifting cultivation. Often these traditional rights are lost when large areas of tropical rainforests are protected or converted to other uses. Finally, there may be sociocultural costs because protected forest areas to which access is denied may have religious and cultural significance to local people. Development projects have often failed to take into account the opportunity costs of people with traditional rights to forests where large forest areas are protected or converted to other land use activities, such as agriculture or grazing for livestock animals. The failure to adequately compensate or involve people in the establishment and management of protected areas has resulted in poor performance of many projects dealing with reserves and natural parks. In many instances, these parks and reserve areas are vulnerable to open access problems from local populations. This study analyses the economic and social impacts of establishing a national park in Madagascar on village households living adjacent to tropical rainforests in the Andasibe region. The first section of the report provides an overview of Madagascar and the national park and defines the objectives of the study. Methodology and the theoretical foundation are presented in subsequent sections. The study then examines the socioeconomic characteristics of the protected areas based on a field survey. The cash flow and contingent valuation analyses are presented next, followed by a discussion of the conclusions and recommendations. A Case *uy of Madagascar Worldwide, up to 20 million hectares of forest area are lost each year, of which a major portion are tropical moist forests containing the bulk of global biodiversity (Sharma 1992). Madagascar has been recognized as one of the most biologically rich countries in the world (World Bank 1988). One of the twelve currently identified "megadiversity" countries (McNeely et al. 1990), Madagascar is biologically unique because of its high rates of endemism. The country, however, is very poor economically and is fast losing its phenomenal natural heritage to satisfy important economic needs. Recognizing that steps need to be taken to conserve Madagascar's unique natural heritage, the Malagasy Government has developed plans for establishing a system of parks and protected areas (World Bank 1988). The Mantadia National Park in the Andasibe region of eastern Madagascar is one of the new parks being established. The park extends over an area of 9,875 hectares, representative of the eastern rain forests. Steep terrain, dense undergrowth and a mixture of a number of different species of trees and plants characterize the land and forests in this region, and therefore in the park as well. Most .2- Importantly, the park contains the habitat for the Indri, the largest known extant lemur in Madagascar (Harcourt and Thornbacl 1990). This study examines the economic impact on the surrounding villages of Mantadia Park. Although there are no human settlements within the park boundaries, several villages lie in close proximity. While most of the forest land in the area is under government ownership, there traditionally has been limited enforcement of property rights. As a result, forest resources have been subject to open access. In the few years since the park was established, there has been some prosecution of those who have encroached inside the park boundaries for resource exploitation. The villages surrounding the park depend on the forests within and immediately around the park for forest products and agriculture (Dufresne and Cloutier 1989). For example, from the forests, fuel wood is collected on a regular basis, a wide variety of fish and animals are foraged for consumption, and a number of different types of grass are harvested and used for assorted purposes. Furthermore, forest plats and herbs also serve as important sources of medicine for the villagers. The method of cultivation practiced in eastern Madagascar serves as a principal mechanism of deforestation. For several hundred years, the eastern Malagasy have practiced a form of shifting cultivation known as "Tavy," but this land use practice has become less sustainable as population pressure has shortened the fallow period. About 200,000 hectares of land in Madagascar are cleared every year through this practice (Andriamampianina 1985). While shifting cultivation plays a decisive role in the deforestation of eastern Madagascar, it is also the only means of livelihood known to many of the inhabitants of this region (Oxby 1985). Thus, establishment of the Mantadia National Park impacts both on the villagers' opportunity to collect forest products and on the practice of their principal form of cultivation. Objectives The creation of a park generates a variety of benfits and costs to society. Based on the initial reconnaissance of the area and examination of the World Bank's appraisal report, it was concluded that the local benefits were limited but the costs were significant due to the considerable reliance of the local people on the forest resources for subsistence needs. This research seeks to estimate the change in the welfare of local residents due to the loss of access to lands in the Mantadia National Park and to assess how costs ard compensation issues need to be taken into account in project design to promote more efficient manwgement of protected areas. General Approach This objective is assessed through two different methods: (a) estimation of the opportunity cost to local residents of establishing the park; and (b) direct estimation of welfare loss using the contingent valuation technique. The opportunity cost approach uses a standard economic analysis of market values to determine the net economic benefits associated with the alternative uses of one or more resources. In this study, the relevant opportunity costs are those associated with the traditional land-use practices of peoplo living near the park. By determining recent land use in and around the park and by projecting future land-use changes in the absence of the park, the cost to villagers of losing the opportunity to exploit the park area for agricultural or forest products was estimated. -3- The second method employed the contingent valuation method (CVM), uses survey techniques to establish the value of goods and servces which are not exchanged in markets and which, therefore, do not have any prices associated with them (Mitchell and Carson 1989). Within this framework, demand for non-market goods is established by first describing a simulated market to the respondents, and then asking them directly to reveal their preferences in terms of some common denominator. The advantage of using the CVM is that it is suitable for valuing a wide range of non-priced environmental goods and services (Kramer, Healy and Mendelsohn 1992). It is a technique that has been applied infrequently in the developing countries, although its use is growing (Munasinghe 1993). In this study, CVM was implemented by asking villagers what would compensate them for the loss of access to lands within the park. THEORETICAL FRAMEWORK A Household Production Model Household production models have been used by economists to study a number of issues pertaining to farm households (Singh, Squire and Strauss 1986). These include labor supply, transportation, intra-household decision making, etc. (Gronau 1977). The same framework can be used to study the impact of the establishment of a national park on Malagasy villagers, in terms of their demand for forest and agricultural products and their land use practices (Shyamsundar 1993). Households are assumed to consume a vector of market goods X., a vector of subsistence goods X,. and a vector of minor forest products Xf. Two of these, X, and &,, are both produced and consumed by the household. Furthermore, households obtain utility from the consumption of leisure X1. The household utility function is given by: U = U(X,X,X,X) The household uses the surplus from the production of the agricultural goods and the minor forest products to buy market goods. It therefore faces the following income constraint. p, X. = po(A-X) + p/F-X) - wl, - p&k + I - crc]" where X. a vector of market goods purchased by the household A a vector of quantities of the different agricultural goods produced by the household F a vector of quantities of forest products harvested or processed by the household p. a vector of prices of market goods ps a vector of prices of agricultural goods Pf a vector of prices of forest products l, a hired labor used in the production of agricultural goods k a vector of capital goods used in the production of agricultural output w the market wage Pk a vector of prices of capital goods I - other income available to the household -4- cyT = costs of clearing primary feasts for swidden agriculture Thus, the right-hand side of the above expression represents household net surplus. It is the income available to the household after consuming a portion of its produce, and purchasing the inputs used in home production. This income is used to purchase a vector of market goods X., which are available at market prices pm. A second constraint the household faces is a technical constraint relating to its agricultural production function. The village households around the park practice swidden agriculture. Every year they clear a portion of a hillside and cultivate rice and other crops on it. These cleared lands are mainly lands under secondary forests, but can also be lands under primary forests. Usually this land is left in fallow the following year when the household moves on to another hill slope. The production constraint is:' A = 41,T,,,T) cla(Z >0 ' ,kT 8a a Agricultural output is therefore a function of total labor used 1, a vector of capital inputs k, land under primary forests cleared for agriculture, T,, and land under secondary forests cleared for agriculture, r,. The quality of land under primary forests is assumed to be better than that of lands under secondary forests. One reason for this condition to hold would be a greater agricultural use land under secondary forests. The difference in land quality accounts for the higher marginal productivity attributed to lands under primary forests (za/aTbOa/8r,). Total labor used in agriculture, 1, refers to own labor 1,, plus hired labor I,,. I = 1. + lt. The capital goods k refer to pesticides and various tools used in agricultural production. A third constraint the household faces is a technical constraint representing the production or harvesting of forest products. These resources include wildlife, grasses of various sorts, medicinal plants, and wood for construction, furniture, and fuel. F = f(1,, T, T',) 'Some of the agricultural products are jointly produced. However, the assumption of non-joint production does not hamper the empirical analysis to be undertaken, and is therefore retained. -5- Forest products collected are a function of labor and land under different types of tree cover.' Tf, refers to lands which are under primary forest cover from which the household obtains forest products. T', represents the total amount land controlled by the household which lies in fallow and is not under cultivation. This land, as mentioned earlier, is characterized by shrubby secondary forests, and the household obtains some of the required forest products from these forests. A total labor effort lr, is used to produce F. Land under secondary forest cover, T,, and land unde- primary forest cover T., together represent the total amount of land, T, to which the household has access. Total land, T, is characterized by two differing property regimes. T, is privately controlled, while T, is an open access resource. However, there is no trade or excange of either types of land. Land is therefore not priced. The different land constraints under which the household operates are specified here. T, = Tr, + T's Tp = TA, + TI, and, T = Ts + Tp Finally, the agricultural household is constrained by the total amount of time it has: H = X, + + 1, H refers to totai hours available, which are used either in leisure, X, or labor (ld +l.). The household's optimization problem can be re-written as follows: Max U(X,,Xa,XX,) subject to the following constraints: pa& + paX. + pf/ f + wX = pA + pfF - w(1 + lof) + wH + I - aicT* $T, T, rI The first constraint equates the households consumption expenditures to its full income. The right hand side of this expression combines the household's money income, and production and labor constraints to construct the household's full income. The full income of the household includes the total value of its labor resources and home-produced goods, and reflects the net income retained by the household after expending money and labor resources on home production. The second and third constraints, as presented earlier, specify the land-use restrictions under which the household operates. 'Some capital inputs are also used. Many of these are the same tools used in agriculture, and are treated as fixed factors in considering decisions on forest resources. Hired labor is assumed to be negligible. There are some cases, however, where mutual help is used in felling large trees for construction. -6- The LaSrangian for the maximization problem is: L U(X.,X,,XX,) x..xA4. AA0.*..r4r S(pa(1,k,T,,T;') + p9l11J/T/) - w(I + 14) - pAk + wH + - p -p - p f - wX] + A(T - T d - T/f) + 6(1,- T,d - T f) where X is the Lagranglan multiplier associated with the household's full income constraint and A and 0 are the multipliers associated with the household's land constraints. It can be thought of as the shadow price of lands under secondary forests and 0 as the shadow price of land under primary forests (Shyamsundar 1993). Welfare Implications of Establishlng the National Park The establishment of the park will result in a change in the total area the household has access to, both for agricultural purposes and for obtaining forest products. Initially, i.e. without the park, the average household is assumed to have access to T hectares of land. With the establishment of the park, the household loses access to some of the land it previously had access to, and retains access to only T' hectares of land (T' s. T). Thus, with the establishment of the park, the household loses access to AT hectares (AT?aO) of land. Some of this land is land under primary forest cover, and the rest is land under secondary forest cover. T T '=AT = AT, + AT, The loss In accessible land has a direct impact on household production, and thereby, on household income and consumption. The change in production and consumption behavior in turn affects household welfare. The loss to household welfare as a result of the park can be estimated in a number of different ways.' The welfare loss can be estimated indirectly by estimating the opportunity cost borne by the household because of the establishment of the park. Opportunity costs are defined as "the benefit foregone by using a scarce resource for one purpose instead of for its next best alternative use" (Gittinger 1982). As long as the household is using optimal and positive quantities of land in its production activities, a decline in land availability results in a decline in household benefits because of a decrease in household profits, 11(T). The cost the household bears equals the change in profits as a result of the change in access to land. This change in household profits is given by: II(T) is the maximum net profit obtained by the household without the park. I(T') is the maximum net profits obtained with the park, with optimal adjustments made in production. The cost the household bears as a result of the park is the difference in the profits obtained, or alternatively, the change in the 'If a market for land existed, the loss in welfare could be fairly simply approximated by considering land prices. However, because of the lack of trade in land, the value of land and thus the welfare loss because of loss of access to land, is established by considering the returns to land holdings. ,7, Tn() lT(7) - n(T') = (p,a(l',k,T , *) + ptjI,T,-T*,T-Ts*) - w(+*) -p) - full income of the household as a result of the park. The discounted value of the change in profits as a result of the park, gives the net present value of the costs of establishing the park to the agricultural household. The not prsent value of costs, OC, is therefore given by: OCm t (1+1)' where i refers to the discount rate. The chaige in costs, or equivalently, the change in profits approach yields an approximation of the welfare loss borne by the household as a result of the park. This loss is empirically established in the section on cash flow analysis. An alternate method for establishing the welfare loss to the household as a result of the park, is by considering the household's expenditure function. The expenditure function is used in standard welfare economics to link income and utility through the indirect utility function. It is the minimum expenditure required to obtain a given level of utility (Varian 1984). In the case of the Mantadia National Park, the expenditure function is defined in the following fashion: e(pw, ps. P. w, U*, T) - min p.X + pX, + pfX, + Wx subject to, U(X., k. X. X) W Uo where Uo is the household's indirect utility function, i.e. Uo - V(pa, p,, pt, w, T, Y*) Uo is the maximum obtainable utility without the park, given prices, land availability, and the production constraints. The expenditure function e(.) is therefore the minimum amount of income required by the household to obtain the level of utility it would without the park. A change in hnd controlled from T to T# because of the establishment of the park, results in a change in the maximum utility obtained by the household, and therefore a change in the minimum expenditure required to obtain the previous level of utility Uo. The welfare change attributable to a quantity change in a fixed public good is the difference in the minimum expenditure required to retain the level of utility achievable prior to a change (Mitchell and Caron 1989). This welfare change is referred to as the Hicksian compensating surplus. In the case of the Mantadia National Park, the Hicksian compensating surplus is given by: e(p., pa. pf. w, UO, T) - e(p, p., p, w, U9, T') The indirect utility function is the maximum utility achievable given prices and income. It is the value of the maximand after the consumer has solved her utility maximization problem (Varian 1984). -8- Since the initial 'without park' utility level equals the maximum utility obtainable, given full income, duality between the expenditure and the indirect utility function implies that the initial expenditure function equals the optimal 'without park' full income, Y*, i.e. e(pm, p., pt, w, UO, T) =Y Therefore the annual welfare loss as a result of a change in T is: e(pm, Pg. Pf, UO, T') - Y* where T' ! T The total welfare loss is therefore given by: e(p,,P,,,p,U,,Tj - Y (1 +iY where i again refers to the discount rate. The change in welfare as a result of the change in T is estimated using contingent valuation and econometric techniques in the section on contingent valuation analysis. METHODOLOGY Survey Developmenet' The first stage of the study was a pilot study of villages living near the Mantadia National Park. This pilot study was conducted in the fall of 1990 by SA.FA.PI., an agricultural non-government organization which frequently conducts socio-economic surveys. This pilot study provided profiles of the villages' characteristics and economic activities. These profiles were used in the initial design of the survey instrument. In the second stage, the initial survey instrument for obtaining village level informaticu was developed. This instrument was based on theoretical considerations and prior understanding of ru-al households in the Mantadia region. The initial draft of the questionnaire was developed before going to the field, and the questionnaire was translated into French and Malagasy. The third stage of survey instrument development was comprised of expert group and focus group discussions of the survey. The basic questionnaire was reviewed and reformulated in Antananarivo, the capital of Madagascar, after a focus group was held with the survey team provided by SA.FA.FI. After reviewing and critiquing the survey, the interviewers received initial training on survey implementation. The survey instrument was also reviewed at this stage by several foreign and Malagasy experts. 5The expenditure function is obtained by inverting the indirect utility function, and solving for income (Varian 1984). The authors acknowledge the considerable assistance of Paul Ferraro, graduate research assistant at Duke, in organizing many aspects of the data collection. -9- A focus group consisting of 15 Andasibe residents was next held to obtain some feedback from local people on several aspects of the survey, especially the contingent valuation questions. Focus groups are increasingly recognized as a critical component in the development of successful survey instruments (Desvouges and Smith 1988). In the fourth stage, the survey instrument was revised again based on the focus group discussions, and a pre-test of the revised survey instrument was conducted in the village of Farahevitra, 5 kilometers away from Andasibe. Twenty-five households (7 percent of the sample) were interviewed. This village itself was not included in the final survey. The pre-test motivated several additional changes in the questionnaire, and was especially critical in the final formulation of the CV question. As a result of the pre-test, the CV format was changed from a willingness-to-pay to a willingness-to-accept format.1 Furthermore, the codes, units of measurement and instructions used in the questionnaire were also revised. In this stage, there was also further training of the interviewers. Based on the pre-test and understanding of village level issues, a second questionnaire for village leaders was developed to shorten the household survey and to control respondent fatigue. This questionnaire obtained information on (a) village level parameters such as prices of forest products which were common to all households, and (b) locationally and culturally sensitive issues such as shifting cultivation which household respondents might not be inclined to answer. In the final stage, five teams administered the final versions of both questionnaires to the pre- selected sets of villages. The English versions of the final household and village leader surveys are found in Shyamsundar, Kramer and Sharma (1993). The sample was stratified by population location and income. Villages were selected in regions east and south of the park since this where most of populations in the vicinity of the park were located. Villages located within an approximate radius of 7 kilometers from the park boundaries were chosen since these were most likely to have used park areas in the past. Three sets of villages were identified. A total of 17 villages were surveyed. The survey teams were instructed to survey approximately 80 households in each identified set of villages. Where villages contained more than 80 households, the choice of households was based on income distribution within the village so that all income levels were included. The survey was administered over a pericd of eight days between July 13, 1991 and July 22, 1991. The survey was administered entirely by the SA.FA.F. team, who were accompanied by some members of the Duke research team. To increase the villagers' willingness to participate in the survey, a health team was organized to accompany the survey team. Consisting of doctors and nurses, the health 'Andasibe is a town 3 kilometers from Mantadia National Park which was the base for conducting field research. 'In the pre-test, a number of villagers indicated that they were willing to sacrifice labor time to walk an additional distance in order to obtain forest products or for swidden cultivation, if this meant the park would be established. This seemed to indicate that the villagers were willing to pay to protect the forests. However, from further discussion it became apparent that the villagers' responses did not stem fron, perceived non-use values but from a sense of coercion from local authorities who had made several arrests for incursions into the park area. - 10* team provided basic medical consultations and medicines to the villagers who have very little access to health services. The survey team obtained data from 351 households and from 11 village leaders. The surveyed population consisted of 1,598 people. The total number of villages, households and population surveyed are given in Table 1. Table 1: Characterisdcs of Surveyed Population Total number of villages 17 Total number of sub-groups into which villages 5 were divided Total number of households surveyed 351 Total population surveyed 1598 Number of village leaders surveyed 11 Data Maagement Data management after the completion survey required considerable effort and time. Data were collected and entered into Lotus 1-2-3 computer files on a total of over 85 basic variables. The basic variables on which data were gathered include age and education of respondent, land area cropped, crops grown, forest products extracted, labor days in agriculture etc. However, because of the variability (in terms of different units of measure specified) in some of the basic variables the total number of variables in the data base exceeds 300.' Units of a number of different goods were standardized to be comparable. Some additional data on prices, exchange rates, labor used in agriculture and yield variability were obtained from secondary sources. Conversion factors (for standardizing units) and data sources are found in Shyamsundar, Kramer, and Sharma (1993). Changes in land use patterns and deforestation within the park were estimated for this study by combining aerial photographs, Landsat satellite images, and topographic maps of the region containing the park (Green 1993). Once the data were recorded, they were modified to suit different computer programs. Five different computer programs were used in analyzing the data: SAS, Statgraphics, @ Risk, Lotus, and LIMDEP. LOTUS was used for Inputting the data from the surveys and for constructing the spreadsheets used for the cash flow data. SAS was used to compute summary statistics of the data. Statgraphics was used to fit probability distributions for the risk analysis. 0 RISK, an add-on program for LOTUS, was used to conduct the risk analysis. LIMDEP was used to estimate the logit regressions for analyzing 0 A major portion of the data entry was done by Cynthia Gill, Denise Noble, and Colin Holler. * 11 - responses to the CV questions. Unfortunately, no single package is available to carry out all the analyse conducted for this report. THE PEOPLE AND THE PARK Soclo-Economic Differences Among the Villages The villages surrounding the Mantadia National Park can be classified into three sub-sets, based on their socio-economic characteristics. The villages' geographic variation, to a large extent, contributes to differences in socio-economic conditions. The first sub-set of villages lies to the north-east of the park. These are remote villages characterized by limited commercial exchange, subsistence agriculture, and very little access to the outside world. Their main link with the outside world is a train station several kilometers away where the villagers market their produce when they require money. The second sub-set of villages lies to the south-east of the Mantedia National Park. These are isolated villages similar to the north-eastern villages and are hidden in the valleys which border the park. Like the north-eastern villages, they are traditional societies made up mainly of a single ethnic group, the Betsimasaraka. The primary traditional occupation of the Betsimasaraka is shifting cultivation or swidden agriculture. These villages have existed in the eastern valleys of Mantadia and have used its resources for several decades. The third sub-set of villages which lies immediately to the south-west of the park are different from the other two sub-sets in that many of its inhabitants are employed in two graphite mines which operate along the western border of the park (GPF 1991). Although many of the men hold a regular job with the mining companies, households continue to undertake swidden agriculture. This is done during vacations, with the help of family members, and by hiring labor with the money they make as mine employees. The graphite mines date back to French colonial times, and since that period, villagers have lived a life which is a combination of industrial employment and itinerant agriculture. General Socio-Economic Characteristics The village household survey covered a total of 17 villages lying to the east and south of the Mantadia region. The total population covered by the household survey was 1,598, indicating that the average household size in this region is 4.6 persons (see Table 2). Fifty percent of the surveyed population was female. Most of the villages do not have access to any medical facilities, running tap water, or electricity. The village children in general suffer from malnourishment. Malaria, chest congestion-related illnesses, and venereal diseases are other significant health problems affecting this population. In general, most of the villages surveyed either had or were within 4-5 kilometers from primary school facilities. However, the survey indicated the average number of years of education per person to be only 2.35 years. - 12 - Table 2: Socio-economic Characteristics of the Surveyed Households Variable Range Mean Number of household members (N=351) 1 - 13 4.55 Age (years) (N=335) .5 - 100 17 Proportion younger than 10 years of age -- 0.37 (N = 593)1 Education (years) (N= 1542) 0 - 14 2.35 The average annual per capita income in Madagascar in 1988 was $190 (World Bank 1989). The households covered in the survey generally have income levels below this average. The survey provides estimates on several indicators of wealth. Approximately 95 percent of the households surveyed control some amount of land, which they cultivate. However, the average amount of land controlled is only 1.89 hectares per household. Other proxies for wealth relate to ownership of watches, radios and oil lamps. In the sample surveyed, 37 percent of the households own a watch, while 33 percent own radios, and 97 percent of the households have a kerosene lamp to light their huts. Economic Activities The dominant economic activity in regions around Mantadia is swidden agriculture. Seventy nine percent of the households surveyed said that they would add to existing land for cultivation. Ninety-nine percent of these respondents acknowledged that they planned to cut forests to add to their land. The average household expected to clear 1.7 hectares of land in the coming year to undertake swidden agriculture. Rice is the primary crop. The average household produces 487 kilograms of rice per year which is worth about US $12810 (see Table 3). Other crops grown are maize, beans, manioc, sweet potato, taro, sugar cane, ginger, banana, and coffee. In general, households which depend solely on agriculture for subsistence undergo a period of famine every year. The famine period can extend for more than six months, during which households do not have sufficient rice to feed themselves. Most of the other crops grown act as rice substitutes during this period. Livestock activity in the study area is primarily restricted to cattle which is a symbol of prestige and used mainly for ceremonial purposes. 11he price of rice used is the farm gate price of 500 fing/kg and the exchange rate used is US $1 fig 1,900. . 13 - Table 3: Value of Agricultural and Forestry Activities Activity Number of Total annual Mean annual value per observations value for all household ($US) villages ($US) Rice 351 $44,928 $128 Fuelwood 316 $13,289 $38 Crayfish 19 $220 $12 Crab 110 $402 $3.7 Tenreck 21 $125 $6 Frog 11 $71 $6.5 The villagers are dependent on the forests around their villages for a number of different purposes. As expected, data on forest products suggest that fuel wood is the most important forest product collected (Table 3). Annually, the average household collects about 6164 kilograms of fuelwood. The total value of the firewood collected is $13,289 or about $38 per household per year." Ranking next to fuelwood in terms f frequency of harvest are fish, crayfish, crab, wood for other purposes, frogs, crabs and shrimp, medicinal plants, honey, and a variety of grasses. Grasses are used as material for constructing houses, for weaving mats, and for making clothing and hats. Survey data also indicate that labor, such as gathering and harvesting forest products, is shared by men, women and children. For example, while women harvest most of the crustacean species, men mainly harvest other fish and wildlife, and wood, and collect honey and medicinal plants. Furthermore, women tend to harvest two types of grass, Harefo and Herana, while male members of the family harvest another type of grass, Vakoana. Children participate in the gathering of all of the forest products, and adults and children equally share the task of frog harvesting. Villagers' Perceptions of Forests In order to assess local views about conservation and to set the stage for the contingent valuation exercise, the questionnaire included a number of questions which specifically probed villagers' perceptions of forests. Although, 65 percent agreed that floods occur less frequently with forests, 40 percent indicated that forests did not help soil protection. Interestingly, 91 percent of the respondents agreed that primary forests were more "pleasurable" because of the plants and animals they contain, than secondary forests. This suggests that recreation may be an important forest use. Another surprising result was that 77 percent of the villagers interviewed indicated that preserving forests in order to protect ancestral graves was not very important. This is interesting because local folklore (or tradition) suggested that one of the most important uses of the forests to the Malagasy is providing this resource as a resting "The price of fuel wood used is fmg 11.67 per kilogram. . 14- place for their ancestors.12 Finally, 68 percent of the villagers indicated that it is advantageous to clear forests as a form of post management. CASH FLOW ANALYSIS Objectives In this section, the opportunity costs of establishing the Mantadia National Park to the local populations are determined by considering: (a) The income generated from forests by the locals in terms of forest products (both timber and non-timber) harvested. (b) The income generated by locals from swidden agriculture from the lands on which the forests lie. (c) The extent to which the area included in the national park includes forests which were used by the villagers. (d) The extent to which villagers continue to have access to the area within the national park for harvesting of forest products and for swidden agriculture. Data Used Forest Products The income generated by villagers from forest product harvesting is established by using primary data obtained in the village household survey. The data used include: (a) Annual quantity of different products harvested. (b) Annual frequency of forest products collected. (c) Distribution of harvesters by gender and within household categories (for example, children vs. adults). (d) Distribution of households in terms of where the products were harvested (i.e. if harvested from the primary forests or secondary forests). Price data on forest products and data on labor expended were obtained from a number of different sources including interviews with villagers, shop owners, household and village leaders, and from published reports. Agricultural Products Agricultural income from using the park land is established by considering data obtained from the village household survey on: (a) Area under crops. (b) Annual quantity of different crop yields obtained from subsistence farming. (c) Annual costs incurred from use of different capital goods for agricultural purposes. 121n fact there is a Malagasy term which refers to the forests as the "robe of the ancestors" (Olson 1984). Price data on agricultural products ace farm gate prices, and have been obtained by interviewing various people operating within the village economy. Data on labor expended in agricultural production have been obtained from a zomprehensive study undertaken on the costs of agricultural production (MPASSA 1987). Strategy and Assumptions Three cash flows have been developed based on the geographical groupings of villages which were surveyed and an estimate of the populations which lost access to lands and forest products as a result of the park. The three cash flows refer to sets of villages that lie north-east, south-east, and south-west of the park. In each of these cash flows, the net loss from the establishment of the park to the average household in the sub-group of villages being considered is estimated. The first cash flow presents the opportunity costs of establishing the park to the average household in the fokontony of Vohibazaha. Fokontony refers to an individual village or a set of villages belonging to a single administrative unit. The Vohibazaha village (Group 1) lies north-east of the park, and its inhabitants previously had access to land currently within the park boundaries. The second cash flow refers to a group of villages which fall under the fokontony of Volove. This geographical sub-set (Group 2) of villages lies on the south-eastern border of the park. The third cash flow uses data mainly from respondents in the villages south-west of the park. The responses from nearby surveyed villages have been used to provide an approximation of the opportunity costs of establishing the park to the villagers (Group 3) who occupy regi.s immediately south-west of the park.'3 Each of the cash flows is treated identically in terms of assumptions made unless otherwise specified. The primary data used to establish the opportunity costs to each sub-set of villages differ across each of the three cash flows, but the secondary data used (agricultural labor) are common to all of them. Parameter values used, for example, the discount rate, and the foreign exchange rate, are also common across all the cash flows. Each cash flow consists of two sets of estimations: (a) 1991 (single year) estimates of the net cash flow to the average village household from the harvesting of forest products and the net cash flow accruing as a result of swidden agriculture. (b) The final cash flow which extends from 1991 to 2010. This cash flow is based on the initial year's flow of befits and costs, but also incorporates certain adjustments in benefits and costs over time (e.g., changes due to population growth and increasing 3The population impacted by the park in the south-west live in the fokontony of Falierana aW Andsifahatelo. The sample did not include these two sets of villages, but included other nearby villages which were used to proxy the responses of the missing villages. Therefore, inasmuch as the village economy differs in Falierana and Andasifahatelo from the nearby villages surveyed, the opportunity cost estimates for the population south-west of the park should be treated as an approximation. - 16 - deforestation in park areas; see Shyamsundar, Kramer, Sharma (1993) for additional details). Each of the three cash flows measures the economic benefits from the forests within the park to the locals if they continued to have access to the park. (This is the "without park" scenario.) The regulations under which the park has been formulated indicate that the villagers are expected not to use the area within the park for swidden agriculture or forest product harvesting (the "with park" scenario). The following analysis is based on this stringent assumption. "Cash in" refers to the flow of benefits from the land which the average household uses for swidden agriculture and for obtaining forest products, and "cash out" refers to the costs incurred by the average household in undertaking the same activities. It is assumed that village households are not able to substitute other forest land for the park and do not intensify their production activities on their other land. These are reasonable assumptions since there is little orimary forest land outside of the park boundaries and there are virtually no purchased inputs used in agriculture which could substitute for land in the production function for agriculture. The cash flows, therefore, estimate the value of land to the average household assuming a "without park" scenario. This value is equivalent to the opportunity cost of establishing the park to the average household. Therefore the net present value figures in each of the three cash flows represents the present value of the opportunity costs from the establishment of the Mantadia National Park to the average village household in each of the three different village groups. Monte Carlo Simulation was used to examine the effects of fluctuations in key variables on the cash flows. The detailed cash flows are found in Shyamsundar, Kramer, Sharma (1993). That report also details the different data used, the sources of these data, and the logic underlying several of the calculations. Results I The following table presents the net cash flow in terms of U.S. dollars for forest products and agricultural products for the year 1991, and the net present value of the opportunity costs of establishing the park to the average village household in each of the three groups of villages. In all three geographical locations, the average household bears losses in use values from the establishment of the park. Cash flow analysis indicates that the net present value of opportunity costs per household from establishing the Mantadia National Park is, $489 for Group 1, $848 for Group 2, and $975 for Group 3. The net annual income from forest products ranges from $12 to $95. The net annual agricultural income from the park ranges from $10 to $36. It is to be noted that these amounts are not equal to thp total annual real income of the average household under consideration. Only a part of the total agricultural and forest product *related income comes from the park itself. Geographical variations exist in the opportunity costs of establishing the park. The NPV of costs per household differs within the eastern region (Groups I and 2), and between the east and the south (Group 3). In the eastern region, the net returns from agriculture exceed the annual net returns from forest product extraction. However, this is not true for villages south of the park. There are several reasons for the higher relative value of forest products in the south. Areas south of the park are relatively commercialized. There exists a better market for forest products in this region, and this has resulted in higher prices for forest products, and a higher net cash flow from the use of forest products. Furthermore, again as result of greater commercialization, labor wage rates tend to be higher in regions south of the park. The use of capital inputs also tends to be higher. This results in lower net incomes from agriculture. Finally, areas south of the park grow less of the higher priced commercial crops of ginger and coffee, relative to areas east .17- of the park. This again results In a further decline in Group 3 agricultural profits relative to agricultural profits for Groups I and 2. Table 4: Results from the Cash Flow Analysis Group Ne cash flow from forest Net cash flow from agricuaral NPV of opponwly carts per products per household products per household household (1991 $f (1991 $f (1991 $) 1991 * 2000 I - 10% Group 1 12.44 25.44 488.89 (n.e.) ______________ Group 2 34.67 35.76 847.89 Group 3 94.53 10.40 975.04 (s.w.) I __ A real exchange rate of US$1 * fmg 1111.8 (which is obtained from a nominal exchange rate of US$1 fmg 1900) has been used for the cash flow analysis. Using the real exchange rate and the real discoun rate (I.e. Inflation adjusted rates) allows us to assess the opportunity costs in real terms. The results of the risk analysis from the Monte Carlo simulations are presented in Table S. The Table presents the mean value, minimum value, maximum value, range, standard deviation, and probability estimates of the not present value of opportunity costs for each sub-group. Table 5: Risk Analysis of the Cash Flows Gro Mean NPV MA ma Std De". Psoability of obainig a Leative NPV Group 1 543.43 -22 4095 468.5 .4% Group 2 931.8 171 5711.7 656 0% Group 3 1001.24 725.6 2574.8 240.3 0% Incorporating risk into the analysis allows us to consider how variations in key factors might affect the cash flow NPV estimates. Risk analysis reinforces the significance of the opportunity costs. The mean NPV values estimated after taking certain risks into account are somewhat higher than the point estimates obtained from the cash flow analysis due to the effects of nonsymmetric probability distributions. The mean NPV estimates range from $543 in Group 1 to $1001 in Group 3. Risk analysis not only provides mean estimates, but also the entire distribution of the NPV estimates. The not present value of opportunity costs can lie within a wide range of values; the NPV estimates of costs range from - 18 - a low - $22 in Group 1 to a high of $5,712 in Group 2. However, there is a high probability that the opportunity costs are positive. While there is a 0.4 percent chance that the NPV of costs for Group I is negative (i.e. a 0.4 percent chance that these villagers will not have to bear any significant costs), for the other two groups there is a zero probability alat the NPV of costs are negative. Aggregation The issue of population aggregation must also be considered when determining the total opportunity costs to local villagers of establishing the Mantadia National Park. The three cash flows developed correspond to the people in the north-eastern fokontony of Vohibazaha, the south-eastern fokontony of Volove, and the south-western fokontony of Falierana and Andasifahatelo who are immediately impacted by the establishment of the park. Lands which these people previously used currently lie within the park boundaries and cash flow analysis has demonstrated a high probability that the costs villagers from these three areas will have to bear are positive. Furthermore, in the periphery of the northern border of the park is another set of villages which fall within the fokontony of Andonaka. The population of this village currently does not have any lands in the park, as evidenced by aerial photographs, but it is likely that they would use lands which now lie within the park in the very near future. Therefore, in aggregating the opportunity costs of establishing the Mantadia National Park, the villagers of Andonaka are also included within affected population. Population figures have been obtained from the 1992 census data for all of the relevant villages. These population figures along with some estimates of the aggregate NPV of costs are presented in Table 6. Table 6: Aggregation of the Net Present Value of Opportunity Costs (Foregone Income) Fokentony (group of Socioeconomic Measures villages) Names of fokontony Population NPV of opp. NPV of opp. Aggregate NPV Annualized costs per costs per of opp. costs ($) opp. costs* per household ($) capita ($) household ($) Vohibazaha (n.e) 521 488.89 115.57 60.217 57.42 Andonaka (n.e) 554 488.89 115.57 64,029 57.42 Volove (s.C.) 775 847.89 188.84 146,354 99.59 Falierana and 1554 975.04 190.11 295,437 114.53 Andasifahatelo (s.w.) All villages 3404 729.75 166.28 566,038 91.11 *Annualized Opp. Costs = NPV*[ i(1+i)A (l+i)a-1 Cash flow analysis indicates that the average NPV of opportunity costs per household are $489 in Vohibazaha. $848 in Volove, and $975 in Andsifahatelo and Falierana. Furthermore, assuming that the costs born by the populace of Andonaka are identical to those borne by the people of Vohibazaha, -19- a per household value of costs of $489 is obtained for households in Andonaka. Assuming that people outside of these areas do not use the park area,"' the aggregate net present value of the opportunity costs of establishing the Mantadia National Park to the local people Is $566,038. The estimated annual opportunity costs borne by the average household is $91. This amounts to a total annual loss of $66,487. CONTINGENT VALUATION ANALYSIS Objectives The objective of the contingent valuation exercise is to estimate directly the welfare loss accruing to the households. Issues of Relevance and Strategy In order to assess the value of lands within the park, a number of issues needed to be clarified prior to the survey. The following issues were addressed in the pre-test to the survey and in the focus group sessions. (a) What are the actual property rights prevalent in this region, i.e. who holds the rights to the use of the park area? What are the perceived property rights in this region, i.e. who believes that they hold rights to the land within the park? (b) Given that this is mainly a non-monetary economy, what is the right numeraire to be used in obtaining monetary measures of indirect use and non-use values? (c) Given that this is a subsistence economy with limited market transactions, what is the most appropriate elicitation method to use? What is the best manner in which the contingent valuation (CV) questions could be framed to obtain valid responses? (d) How much information should be provided prior to the CV question? How best could the people be motivated to think about the indirect and non-use benefits derived from the forests? The property rights question becomes critical in deciding between a willingness-to-pay (WTP) versus a willingness-to-accept (WTA) format when posing the CV question. A considerable amount of discussion in the literature has focused on the appropriateness of each format (Carson 1991). It can be shown that the difference between WTP and WTA can differ based on income substitution effects (Hanemann 1991). The survey implemented in this study used a willingness-to-accept format. The pre-test conducted suggested that while property rights over forested land are held by the state, the people in this region have been using forest resources for a long time, and they perceive that they have traditional rights to the land. "People in other regions are unlikely, at least in the near future, to use lands within the park. This is mainly because they are land locked by villages which surround them. It is therefore only the villages which currently border the park which can and will expand (if permitted) and thereby will bear the costs of loss of access. -20- Willingness-to-accept seemed not only the most appropriate format to use, but also the only way to obtain meaningful responses. Carson (1991) has further argued that WTP is more appropriate than WTA if a quantity or price change of a public good affects the same group of agents from both sides of the transaction. For example, with the Mantadia National Park, WTP would be appropriate if losing access to the park area resulted in an increase in non-use benefits to the villagers, and they were willing to pay for these benefits emerging from conservation. Pre-survey information did not suggest that the villagers were willing to pay for these benefits. The numeraire used In the survey to obtain WTA bids is rice. Rice is the main crop in this region and its value Is well understood. Furthermore, some amount of rice is also sold, and transactions of rice are thus known and understood by the local people. The unit of measure used was a "vata", which is a locally-used unit for rice transactions, equalling 30 kilograms of rice. The elicitation method used in the survey was the dichotomous choice method. This method, also known as the "take it or leave it" method (Mitchell and Carson 1989), was first developed by Bishop and Heberlein (1979, 1980). In this method, respondents are presented with a bid amount which they have to agree to take or not to take. This method, relative to other methods, is alleged to be easier for respondents to answer. It also best imitates regular market transactions. The simplicity of this elicitation method was the primary reason for choosing dichotomous choice over other methods. Prior to posing the contingent valuation question, the respondents were asked a series of questions prompting them to begin thinking about the benefits drawn from the park. These questions probed perceptions on different aspects related to the forests like flooding, soil erosion, ancestral traditions, wildlife as destroyers of crops, availability of primary forests in the future, etc. Respondents were also asked If they knew about the park and about their perceptions on the use of buffer zones as alternatives to the forests in the park. The authorities in Madagascar have plans for buffer zones around Mantadia within which development activities will take place. Therefore, in presenting the CV question, buffer zones were proposed as a compensatory mechanism for the loss of lands within the park. Thus, the contingent valuation exercise enables one to estimate the welfare change as a result of the park, given the possibility of planned buffer zones within which economic activities will take place. The contingent valuation question used was: Suppose you are asked to use only the buffer zone set aside for collecting forest products and for growing crops and are asked not to use the rest of the forests any more. Suppose in order to make up for asking you not to use the forests in the park, you are given vata of rice every year from now on. Would this make you as content as before when you could use the forest in the national park? Respondents were randomly assigned to seven groups, corresponding to the seven different amounts of vata (1 to 7 vata) used as the offered bid levels. -21. Theoretical and Econometric Specification for the Dichotomous Choice Analysis The following section puts forth the theoretical assumptions and understanding of how villagers value forests which underlie dichotomous choice analysis. This section follows theoretical developments in McConnell (1990) and Cameroon (1988). Let Yj be the true WTA bid of household i. Yj is the bid which will make the household as well off with the buffer zones and no access to park lands, as it would have been with access to park lands and no buffer zones. Yj is a compensated demand function for within park lands, given that households are denied access to these lands. Yj is defined by the following relationship: YA.) = e(BZ,T',p".p4,p1w.U";s,) - e(0,T,p,,p,prw.U;s) + (4- * Y(.) - '(BZ,T',p,p',p,w-U,;s,) - Ye + a, where T' and T refer to lands accessible to the households with and without the park (T'< T); BZ to potential buffer zones; p.,, p, to vectors of prices of marketed goods, agricultural goods, and forested products faced by the household; w to the wage rate; U. to the initial utility level; and s, to a vector of socio-economic characteristics of household i. The e(.) function is the household expenditure function which is known to households, but known with a margin of error, epsiloN, to investigators. Yj' is current income. Y, is therefore the difference in the minimum expenditure required to make the village household as well off without the lands in the park as it was with the lands. It is a measure of the Hicksian Compensating Surplus (Mitchell and Carson 1989), in that utility or household real income is held constant. Yj is a continuous function of a number of different variables which appear in the expenditure function. It can be expressed as: Y, = X'a + U, where U, - (,b) and X, is a vector of explanatory variables including household size, hectares of land used for cultivation, the number of years the household has lived in the vicinity of forests, a dummy variable which reflects preference for forests over buffer zones, the education level of the head of the household, household rice yield, which acts as a proxy for household income, a series of dummy variables for each of the interviewers in the surveys, and a series of variables which reflect forest use for non-agricultural purposes. The error term U, is distributed logistically with mean 0 and standard deviation b. This model is estimated with the LIMDEP econometric package. Dat used The dichotomous choice analysis is based on information obtained in the household surveys. Along with the responses to the WTA bid question, data on a number of other socio-economic variables have been used. Table 7 presents the data used, defines the variables, and specifies the names used for each variable in the logit model presented in the next section. -22- Table 7: Definition of Variables fousthold level variables used in Dichotomnour Choice Analysis Variable labels es/No responses to the bid (the dependent variable) WTA ffered bid (ranging from I to 7) BID umber of household members HHMEM umber of hectares of land available for cultivation LAND umber of years the household has lived near forests FORYR A dummy representing preference for buffer zones over forests (Buffer Zones = 1, PREFDUM Forests - 0) Years of education of head of household EDUCN ice yield per household RTOT dummy for the first interviewer in group 1 INTIDUM dummy for the second interviewer in group I INT2DUM A dummy for the first interviewer in group 2 INT3DUM A dummy for the second interviewer in group 2 INT4DUM A dummy for the first interviewer In group 3 INTSDUM A dummy for the second interviewer in group 3 INT6DUM A dummy for the first interviewer in group 4 INTDUM A dummy for the second interviewer in group 4 INT8DUM dummy for the first interviewer in group 5 INT9DUM requency of construction wood collection per household CNFREQ requency of fuel wood collection per household FWDFREQ requency of Herana* collection per household OGRFREQ requency of Raphia" collection per household ROFFREQ requency of Harefo* collection per household HARFREQ Herana and Harefo are two different types of grass collected by households, and used for making household articles. Raphia is a palm used for making fibre. Results The results of the multinomial logit model are presented in Table 8 which provides the maximum likelihood estimates of the parameters of the fitted logit function relating the dichotomous response variable to the explanatory variables. As indicated by signs, significant coefficients, and goodness of fit measires, the logit model performs well in explaining variations in responses to the CV questions. The sign of the estimated regression coefficients indicates the direction of influence of independent variables on the CV responses, but the estimated - 23. coefficients cannot be interpreted as marginal influences on the probability of a positive response to the CV questions. Table 8: Maximum Likelihood Estimates of the Logit Regression Variable Coefficient Std. error t-ratio Constant -1.5822 1.522 -1.039 HHMEM -0.11920 0.1102 -1.082 LAND 0.67927 0.2702 2.514 FORYR 0.038293 0.01969 1.945 PREFDUM 2.2867 0.6409 3.568 IDUCN -0.064296 0.07398 -0.869 RTOT -0.0018125 0.0006585 -2.752 INTIDUM -0.56075 1.268 -0.442 INT2DUM -2.1709 0.9613 -2.258 INT3DUM -4.0150 1.131 -3.549 INT4DUM -11.069 3.804 -2.909 INTSDUM -6.7610 1.730 -3.908 INT6DUM -7.2817 1.948 -3.738 INT7DUM -0.92649 1.003 -0.924 INT8DUM -0.14597 1.151 -0.127 INT9DUM -6.6743 1.600 -4.173 CNFREQ -0.25724 0.1236 -2.082 FWDFREQ 0.004275 0.002516 1.699 OGRFREQ -0.76782 0.5010 -1.533 ROFFREQ 1.9439 0.9399 2.068 HARFREQ 0.070036 0.04609 1.520 BID 0.26043 0.1282 2.031 Maximum Log Likelihood -68.27945 McFadden's* R2 .47 * McFadden's R 1 - (Log Likelihood/Log Likelihood (slopes =0)) The offered bid level is significant and has the expected sign, i.e. with an increase in the offered bid amount, the probability that the respondent would say "yes" to the bid increased. With regard to the other independent variables, the parameter estimates which are significant are the number of hectares households use for cultivating crops, the number of years households have lived amidst forests, the dummy on household preference of buffer zones over forests, household rice yield, frequency of construction wood and raphia harvests and, some of the interviewer dummies. The coefficient on hectares suggests that an increase in the number of hectares households' control, would result in an increase in the probability of the bid being accepted. This apparently surprising result has a good explanation. The larger the acreage of land a . 24 - household controls or has the labor to utilize for agriculture, the smaller the need for the more fertile lands in the park, and the higher the probability of agreeing to not use the land within the park. The coefficient on rice yield has a negative sign. This suggests that production decisions are not independent of consumption decisions. Households with more mouths to feed are likely to produce more of the staple, rice. As long as rice yields are directly correlated to rice demand, the negative coefficient on rice yields can be taken to reflect the consumption needs of the household, i.e. the probability of saying yes to the offered bid decreases, the larger the rice requirements of the household. The coefficient on the dummy PREFDUM is positive as would be expected. Respondents who found buffer zones acceptable were more likely to agree to the offered bid. Several of the dummy variables for interviewers have significant coefficients. These variables likely reflect village level differences not captured in other variables, since the interviewers worked in different geographical areas. Interestingly, the coefficients on the frequency of harvest of construction wood and Raphia have opposite signs. The more frequently construction wood is harvested, the less likely the respondent is willing to accept the bid, while the opposite is true for the harvest of Raphia. This suggests that the harvesting of construction wood is threatened by the park, while that of Raphia is not. One way of deriving a goodness of fit measure for a dichotomous choice model is to estimate McFadden's R1, also known as the Likelihood Ratio Index (Green 1990), which reflects how well the independent variables can explain the variation in the dependent variable. This is similar to a regular R2 calculated in ordinary least square estimations except that the maximum likelihood estimator is not chosen such that the R2 is maximized. The McFadden R derived in this case is .47, which is relatively high for cross sectional data. Another goodness of fit measure often used for logit and probit models is the proportion of correctly predicted responses. As shown in Table 8, this model predicts 165 of the 191, or 86 percent of the observations correctly. - 25- Table 8: Frequencies of actual & predicted outcomes for logit model Frequencies of actual and predicted outcomes Predicted Actual 0 1 TOTAL 0 99 12 111 1 14 66 80 Total 113 78 191 Actual outcome - predicted outcome for 165 observations out of a total of 191 (I.e. 86%) The Mean WMingness to Accept A number of different approaches have been suggested for estimating the appropriate WTA or WTP value (Cameron 1988; Hanemann 1984; Duffield and Patterson 1991). Alternative methods to estimate the mean bid, and the appropriateness of using the mean as a sumnmary measure of WTP or WTA relative to the median are also extensively discussed in the literature on the CVM (Johansson et al. 1989; Hanemann 1989). Cameroon (1988) suggests that the appropriate measure of the mean WTA is the predicted value of the willingness to accept function, estimated at mean values of the covariates. This is the mean obtained from the conditional distribution of WTA given the vector of X variables. More specifically, the predicted mean value of the WTA bid is given by: E(1) - E(XA + U) a X + E(U) *2 Re-parameterizing the logit parameter estimates allows us to obtain the parameters of the valuation function. The predicted mean willingness to accept is then obtained by using these estimates in conjunction with the mean values of the covariates. The mean willingness to accept is estimated to be 8.03 vata of rice. A vata of rice, as mentioned earlier, is equal to 30 kilograms of husked rice and each kilogram of rice is worth fmg 500. Translating this to dollar terms, the mean WTA is estimated to be $108.34 per household per year. Aggregating over the relevant population in the area and assuming a twenty year project life at 10 percent discount rate, the aggregate net present value of the welfare loss is $673,078.1s '*The real exchange rate used is US$1 fmg 1,111.8. - 26 - CONCLUSIONS AND RECOMMENDATIONS The Mantadia National Park has been established with the intention of preserving Madagascar's unique biological heritage. While the benefits of conserving the fauna and flora and the biological diversity within the park are large, some very significant opportunity costs must be considered, as a necessary condition, to avoid open access problems that will threaten the existence of the park in the long run. Empirical Results Approximately 3,400 people in three sets of villages will be negatively affected by the park. The first set of villages is a group of villages close to the north-eastern boundaries of the park, the second set is villages along the eastern limits of the park, and the third is a set of villages along the south-western boundaries of the park. Most of the villagers in these villages are forest farmers, i.e. they depend on swidden agriculture for a major part of their subsistence. Most of these villagers have in the past used lands which currently lie within the park for their agricultural activities. Thus, as a result of the establishment of the park, they have lost access to a portion of the territory they traditionally used for subsistence. The cash flow analysis shows that the aggregate net present value of the opportunity costs of establishing the Mantadia National Park to the local people is $566,038. Aggregating separately over the populations in the eastern and southern villages, $270,600 is obtained as the NPV of costs borne by the people in regions east and north-east of the park. The NPV of costs borne by people south of the park is $295,438. The NPV of costs to the average household, across all the different sets of villages, is $775. This amounts to an annualized opportunity cost borne by the average household of $91 as a result of the Mantadia National Park. The cash flow analysis also shows that the incidence of the costs will vary among the different areas of population concentration. For example, the average village household in the eastern and north-eastern region will lose $21 every year if it is constrained from acquiring forest products from the park. On the other hand, the average household in the south-western region will lose an annual amount of $95. The forest products harvested range from timber for construction, to different types of grasses used for making a variety of mats, clothing etc. to wildlife which are consumed. The cash flow analysis also indicates that the average household in regions east of the park will bear a loss of $30 of net agricultural income every year because of loss of access to the park, but in villages south-west of the park this loss amounts to $10. In this study, the contingent valuation methodology provides results which are consistent with the cash flow analysis. The contingent valuation analysis indicates that the aggregate welfare loss to households from the loss of access to lands within the park is $673,078. This implies that the average household perceives a loss of $108 each year as a result of the park. -27 - Application of Methodology The cash flow approach is a relatively simple, but data intensive, form of analysis with practical applications for conceptualizing development programs and in involving people in the management of programs. It is also a powerful tool in understanding the inter-relationship among microeconomic factors relating to use and management of parks. Non-market valuation techniques, which have been extensively used in developed countries to directly measure the welfare changes occurring from a given change in a public good, have been underutilized in developing countries (Whittington 1990). In this study, contingent valuation was used to estimate the welfare change perceived by local residents as a result of loss of access to lands currently within the Mantadia National Park. The analysis indicates that non-market valuation techniques, rigorously applied, can be effectively used in the developing country context. The econometric analysis undertaken indicates a systemettic association between various socio-economic variables of interest and the expressed WTA compensation. The results show that responses to the CV question were decidedly non-random. Also, the opportunity cost (or market based) approach and the CV method provided remarkably comparable estimates of costs borne by villages. All this is encouraging evidence to support the use of CV in such a context, but further research is required to improve its widespread applicability. Lessons Learned Several lessons can be drawn from this study. This research has involved a survey of village households, collection of data on various quantities and prices, and rigorous quantitative analysis. Research of this kind is time-intensive. It is apparent that there is a strong need for a significant amount of pre-survey work to draft a useful survey instrument. There is a need for focus groups and a formal pre-test to sharpen the wording of the questions so that the desired information can be collected. For example, it was found that units of measure for forest products varied between villages only a few kilometers apart. There is a need for involving local sociologists and cultural anthropologists who can ensure that the questions are posed appropriately for the local cultural context, and to advise researchers on the appropriate protocol for approaching local village leaders to ensure their cooperation. In this instance, it was found advantageous to provide an incentive for survey participation by arranging to have a health team accompany the interviewers Of course, careful translation into local languages is also a necessary step, as is thorough training of interviewers. The study team worked with an experienced rural survey group, but found that extensive training was still necessary. This was in part due to the fact that they were unfamiliar with the contingent valuation method. Finally, it is necessary in a study of this type to make the best possible use of secondary data. This reduces the amount of information necessary to collect in the survey and provides information difficult or impossible to ask about directly, e.g. rates of deforestation. Despite the considerable effort required to collect the data gathered for this village study, this information is critically important for implementing conservation projects, and can be -28- collected when baseline Information about residents within or around conservation areas is gathered. Economic valuation is therefore, analysis which can (a) be undertaken when plans for conservation areas are being formulated, and (b) be useful in understanding what the focus of such plans should be. Broader Implications If conservation is to be successful, it is important that local residents participate in conservation, I.e. that they gain as a result of conservation (Wells, Brandon and Hannah 1992). This is a much needed pre-requisite for the success of conservation efforts. Often cost and compensation issues are relevant at the local and regional levels within a country. But there are also some national and global implications if large areas of tropical rainforests are involved. Protection of large areas would generate benefits to people within a country and to those in other countries, but would entail opportunity costs. The financing of these costs falls on the country in question, as well as the world community at large. To this end, further research is needed to study the cost and compensation issues involving industrialized countries and preservation of large areas of tropical rainforests. The key components of such a research program would be to determine the total costs associated with preservation, the appropriate level of compensation, the mechanism for financing and channeling funds, the institutional and legal framework for managing the funds, and the means of distributing funds to affected parties based on equity and intergenerational considerations. Costs and Compensation Considerations This study demonstrated that there is large opportunity cost associated with the establishment of a national park. The affected party is the people who live in forest areas or at the fringes of these areas who have enjoyed access to this resource. The local residents may gain from the establishment of the park by maintaining the benefits from environmental sources, such as watershed protection and regulation of microregional climate, but they will also bear the burden of the conservation efforts. The average annual opportunity costs borne by the average household ranges in the study area are $90 to $110. This cost requires that decision makers establish compensatory mechanisms in development projects as safety nets for the people adversely affected from setting aside forest areas which were previously used by these people. Any compensatory mechanism would also provide incentives to local people from exploiting reserve areas and causing environmental degradation. Per household foregone opportunity costs and willingness-to-accept estimates tend to be low since they reflect income earning rates of subsistence farmers. This raises an issue that has been problematic in benefit-cost analysis for many years, i.e. the costs (and benefits) accruing to the poor carry little weight. Economists have sought to compensate for this effect by arguing that the marginal value of additional income is higher for poorer groups and therefore their costs and benefits should be weighted more than those for poor groups. Another approach is to impose external constraints on maximum (minimum) costs (benefits) that accrue to the poor as a result of a project. In this case study, the latter approach is more practical. More specifically, - 29- when compensation is considered, a parallel option might be to ensure that some minimum level of access to protected areas is assured for local villagers. This would have other advantages of protecting cultural or religious heritage. In project design, planners need to consider different options for compensating local people. The project could provide direct payments to the affected households. Other forms of compensation could include (a) establishment of a buffer zone providing access to people; transfer of land outside the national park to local communities to manage as common resources with clearly defined rules governing access and management; (b) subsidization of fertilizer, seeds, farm implements, and technical support to intensify agriculture; (c) transfer of a portion of revenues generated from the park through user fees to the villages to support social programs such as health, water supply, and education; (d) clarification of property rights by surveying, demarcating and issuing land titles to local people and villages; and (e) involvement of local people in the management of the park. REFERENCES Andriamampianina, 1. 1985. "Traditional land use and nature conservation in Madagascar." In J.McNeely and D.Pitt, ed., Culture and Conservation: The Human Dimension in Environmental Planning. London, U.K.: Croom Helm. Associates for International Development (AID). 1992. Regional Growth and Specialization in Madagascar (Draft Report), Vol 1. Cambridge, MA: AID. Bishop, Richard C., and Thomas A. Heberlein. 1979. "Measuring Values of Extra Market Goods: Are Indirect Measures Biased?" American Journal of Agricultural Economics 61(5):926-930. 1980. Simulated markets, hypothetical markets and travel cost analysis: Alternative methods of estimating outdoor recreation demand. Staff Paper Series No. 187. University of Wisconsin: Department of Agricultural Economics. Cameroon, Trudy Ann. 1988. "A New Paradigm for Valuing Non-Market Goods Using Referendum Data: Maximum Likelihood Estimation by Censored Logistic Regression." Journal of Environmental Economics and Management 15:335-379. Carson, Richard T. 1991. "Constructed Markets." In John B. Braden and Charles D. Kolstad. ed., Measuring the Demand for Environmental Quality. North Holland: Elsevier Science Publishers. Desvousges, W.H. and K.V. Smith. 1988. "Focus Groups and Risk Communication: The 'Science' of Listening to Data." Risk Analysis 8(4):479-484. Direction G6nirale de la Banque des Donn6es de l'Etat. 1992. "Madagascar: Consumer Price Indices." Washington D.C. Duffield J.W. and D.A.Patterson. 1991. "Inference and Optimal Design .for a Welfare Measure in Dichotomous Choice Contingent Valuation." Land Economics 67(2):225-239. Dufresne, Alain and Antoine Cloutier. 1989. "Plans the Gestion. Pare National Andasibe. Parc National Ankarafantsika." Projet GPF. Volet Conservation et Protection. Unpublished document for the IUCN. Antananarivo, Madagascar. Gray, Clive, J. Dirick Stryker and Jeffrey Metzel. 1991. "The Current Macro Economic and Progress of Economic Reform in Madagascar." Prepared for the U.S. Agency for International Development. Cambridge, Massachussetts: Harvard Institute for International Development and Associates for International Resources and Development. Green. Glen Martin. 1993. "Remotely Sensed Determination of Deforestation History and Topographic Controls: Mantady National Park and Surroundings." Report prepared for the Center for Resource and Environmental Policy Research, Duke University. Durham. NC. -31- Green, William H. 1990. Econometric Analysis. New York: MacMillan Publishing Company. Gronau, Reuben. 1977. "Leisure Home Production and Work: The Theory of the Allocation of Time Re-visited." Journal of Political Economy 85:1099-1124. Hanemann, W.M. 1984. "Welfare Evaluations in Contingent Valuation Experiments with Discrete Responses." American Journal of Agricultural Economics 66:322-341. Hanemann, W.M. 1991. "Willingness to Pay and Willingness to Accept: How Much Can They Differ". American Economic Review 81: 635-647. Harcourt, C. and Thornback J. 1990. Lemurs of Madagascar and the Comoros. The IUCN Red Book. Gland, Switzerland and Cambridge, UK: IUCN. Kramer, Randall A., Robert Healy, and Robert Mendelsohn. 1992. "Forest Valuation." In Managing the World's Forests. Ed. N. Sharma. Dubuque, Iowa: Kendall/Hunt Publishing Company. McConnell, K.E. 1990. "Models for Referendum Data: The Structure of Discrete Choice Models for Contingent Valuation." Journal of Environmental Economics and Management 18:19-34. Minist6re de la Production de l'Agriculture et Syst6me des Statistiques Agricoles (MPASSA). 1987. "Etude M6thodologique de D6termination des Cofts de Production (le cas de paddy)." Projet Recensement National de l'Agriculture et Syst6me Permanent des Statistiques Agricoles. Madagascar. Minist6re du D6veloppement Rural. 1989. Annuaire Statistiques Agricoles. Antananarivo, Madagascar: Service de la Statistique Agricole. Mitchell, R.C. and R.T. Carson. 1989. Using Surveys to Value Public Goods: The Contingent Valuation Method. Washington, D.C.: Resources for the Future. Munasinghe, M. 1993. Environmental Economics and Sustainable Development. Washington, D.C.: World Bank. Olson, S. 1984. "The Role of the Ancestors: Forests in the History of Madagascar." Journal of Forest History (October) 174-186. Oxby, Clare. 1985. "Forest Farmers: The Transformation of Land Use and Society in Eastern Madagascar." Unasylva 37(148):42-51. Projet Gestion et Protection des For6ts (GPF). 1991. Etude Socio-Economique des Zones Piriphdriques des Aires Proregies d'Andasibe. Madagascar. Rasolofarinaro. 1988. "Quelques Facies de la Couverture Foresti6re en Relation avec L'Environnement dans la R6gion d'Andasibe." Th6se de Doctorat de 3* Cycle. Antananarivo, Madagascar. - 32 - Sharma, Narendra P. 1992. Managing the World's Forests. Ed. N. Sharma. Dubuque, Iowa: Kendall/Hunt Publishing Company. Shyamsundar, Priya. 1993. "Economic Implications of Tropical Rain Forest Protection for Local Residents: The Case of the Mantadia National Park in Madagascar." Ph.D. Dissertation. Shyamsundar, Priya, Randall A. Kramer, and Narendra Sharma. 1993. "Estimating the Costs to Local Inhabitants of Establishing the Mantadia National Park in Madagascar". Working Paper 93-1, Center for Resource and Environmental Policy Research, Duke University. Singh Inderit, Lyn Squire, and John Strauss. 1986. Agricultural Household Models. Extensions, Applications and Policy. Washington D.C.: Johns Hopkins University Press for the World Bank. Strauss, John. 1986. "The Theory and Comparative Statistics of Agricultural Household Models: A General Approach." In Inderjit Singh, Lyn Squire, and John Strauss, ed., Agricultural Household Models. Extensions, Applications and Policy. Washington D.C.: Johns Hopkins University Press for the World Bank. U.S. Department of Commerce and Bureau of Economic Analysis. "Producer Price Indices July 1986-June 1992." Survey of Current Business, pp. 66-72. Washington D.C. Walker, Ernest P., F. Warnick, S.E. Hamlet, K.I. Lange, M.A. Davis, H.E. Uible, P.F. Wright, 1964. Mammals of the World, Vol II. Baltimore: Johns Hopkins Press. Wells, Michael, Katrina Brandon, Lee Hannah. 1992. People and Parks: Linking Protected Area Management with Local Communities. Washington D.C.: World Bank, World Wildlife Fund and USAID. Whittington, Dale, John Briscoe, Xiaming Mu, and William Barron. 1990. "Estimating the Willingness to Pay for Water Resources in Developing Countries: A Case Study of the Use of Contingent Valuation Surveys in Southern Haiti." Economic Development and Cultural Change 38: 293-311. World Bank. 1988. "Forests Management And Protection Project - Staff Appraisal Report. Democratic Republic of Madagascar." Report No. 6994-MAG. Washington D.C.. World Bank. 1989. "Madagascar. Adjustment in the Industrial Sector and an Agenda for Future Reforms." Report No. 7784-MAG. Washington, D.C.