Welfare Implications of Tropical Forest Conservation: The Case of Ruteng Park by

In 1993, the Indonesian government established the Ruteng Nature Recreation Park in western Flores. Subsequently, the government banned all timber extraction in and around the park's sub-tropical forest to promote biodiversity and watershed protection. This study quantitatively examines the role that tropical forest conservation has on the development of the local economy, and tests hypotheses regarding the local use of and dependence on the tropical forest. Microlevel data is taken from a rich socioeconomic survey administered to local Manggarai households residing within the park's buffer zone. This information is used to quantify the size of forest incomes, and to estimate a household profit function to characterize forest use, test for wealth differences, and identify policy levers for Ruteng Park. Given the large number of households without forest profits or an observable reservation price, a sample selection model ("Heckit analysis") is used on the forest profits censored data. Results indicate that there is a significant reliance on the forest, as forest products make up approximately 30% of total income. Furthermore, poorer households disproportional depend on forest access than do their wealthier counterparts. The policy options explored include limiting forest entry, via a tax or subsidy system. 1 Duke University 2 Research Triangle Institute, 3040 Cornwallis Road, PO Box 12194, RTP, NC 27709-2194 INTRODUCTION There is much debate over the influence tropical forest conservation has on local economies. Conservation programs are preferred to generate benefits (particularly globally) that greatly outweigh the costs, however, their success depends critically on how these programs affect the wellbeing of the local people. If a program provides a wealth of benefits to the world as a whole, but fails to better or retain, the welfare of those reliant on the resource, it will be unsustainable. Without adequate compensation, local populations have little incentive to facilitate forest conservation and change their behavior, which is often steeped in tradition. Previous research has analyzed watershed protection benefits, ecotourism, and the potential of Ruteng Nature Recreation Park located in Flores, Indonesia (Pattanayak, 1997). An economic welfare analysis of logging has not been conducted, though data about the logging community in Ruteng was collected by Kramer et al. (1997). The average logger admitted to spending about 14 days a month in the forest cutting and processing timber, which totals 84 trees per year per logger. Thus, even with the harvesting ban there appears to be a large amount of income derived from illegal harvesting, which would disappear if the ban was effectively enforced. This study quantitatively examines the role that tropical forest conservation (Ruteng Park) has on the development of the local economy (Manggarai), and tests hypotheses regarding the local use of and dependence on the tropical forest. The objectives of this project are three-fold. First, it addresses how an effective timber ban affects the well being of the timber harvesters. It is assumed that well being can be measured as a function of profit (total revenue minus total costs), so welfare losses can be measured as the difference between the current and an assumed no-harvest profit level. Secondly, the distributional effects are examined. This includes how particular socioeconomic factors influence the level of loss, and which economic class stands to lose the most. The third objective is to determine if any policy adjustments might mitigate the loss, perhaps through a transfer mechanism or an alternative policy. BACKGROUND The Ruteng Nature Recreation Park (Taman Wisata Alam Ruteng or TWAR) was created in 1993, on the western side of Flores, Indonesia. The park is 32,246 hectares, with a 56,000 hectare buffer zone surrounding the park. Consequently, there are about 17,000 indigenous Manggarai inhabitants residing within the buffer zone (Kramer, et al. 1997). The area is mainly a sub-tropical forest, that is rich with biodiversity. Ecotourism is a planned park activity, as it is believed its presence will stimulate the local economy. Therefore, it is important that endemic species of cave bats, Komodo rats, monkeys, wild boar, civets, cobra, and vipers have ample habitat. The terrain is extremely steep, as the park is located on the tops of several volcanic ridges that is the source of many rivers and streams. The forest plays a central role in watershed protection. The majority of Manggarai people are farmers who grow a variety of crops, including coffee, rice, bananas, cassava, and vanilla. Since coffee is only harvested three months a year and is the main revenue generator, other sources of income are extremely important (Ministry of Forestry, 1995). Timber harvesting provides additional income for a significant number of the population. All timber harvested comes from government owned forests, which includes the park reserve. The timber is processed into boards and beams, and is sold as construction lumber to local government agencies, churches, and private builders (Ministry of Forestry, 1995). In 1989, the government banned timber extraction within several government owned forests and included Ruteng Park in 1993. The ban was aimed to protect the park’s 12,800 hectares of accessible forested lands (Ministry of Forestry, 1995). The ban, however, has not been very effective. The government has failed to adequately enforce the logging ban and illegal harvesting continues. Assuming an effective ban is feasible and is achieved, the benefits and cost of such a policy are numerous. The ecotourism industry will benefit by accruing amenities, and other communities (some far downstream from the forest) will benefit through carbon sequestration, biodiversity preservation, and watershed protection. The increased watershed protection will also specifically benefit the local farming community (Pattanayak and Kramer, 1997). The loggers will gain little that will translate into real benefit, and stand to take the most substantial loss. The majority of loggers surveyed did express concern for the forest system, but 26% felt that the timber regulations were not warranted. Therefore, the majority of the costs will be borne by the logging community who derive their income from timber harvesting and processing. For example, if harvesting of construction wood is effectively banned, it is estimated that 9.2 million dollars (US) will be lost over a 30 year period (Kramer, et al. 1997). THEORETICAL FRAMEWORK The Profit Function In order to determine a representative logging household's level of welfare, or utility derived from the production, sale, and consumption of park's resources, a profit function is developed. Specifically, the profit function is used to determine the level of household forest dependency and the associated welfare. Forest dependency is positively related to quantity of timber extracted, thus to timber profits. Profit level can be used to proxy welfare, as profits to the household are a function of benefits net of costs. Profits were modeled as a function of several exogenous variables and determined as (Sadoulet and de Janvry, 1995): π = p’q(p,w,z,f(b)) – w’x(p,w,z,f(b)) where q is the vector of output quantities, x is the vector of input quantities, z is the vector of fixed factor quantities, f(b) is accessible forest size given the existence of a ban (b), p and w are the output and input prices, and where q(p,w,z,f(b)) and x(p,w,z,f(b)) are the output and input supply functions (respectively). If an effective ban (strictly enforced, unlike the current situation in Ruteng) exists, then b = 1, so that f = 0, thus q = 0 or that there will be no output supply of forest products, so that π(p,w,z,f(1)) = 0, and π(p,w,z,f(0)) π(p,w,z,f(1)) = loss. Profits are maximized when the first-order condition of the profit function is set to zero and solved for the respective variables. Household Production Theory In this study, household production theory is applied to describe the relationship between household behavior and forest dependency. The use of household production theory is necessitated as the households examined exhibited simultaneous production and consumption decision making behavior (Pattanayak and Kramer, 1998). That is, households are both producers and consumers of products that do not always have completely functioning markets. Then it follows that production and consumption activities are not separate, and the relevant price is some household specific price and not the market price. A perfect market requires that "all prices are exogenous to the household and all products and factors are tradable with no transaction costs"(Sadoulet and de Janvry, 1995). These choices are to be made independently of one another, and their only "linkage" is through the level of income generated by the production process (Sadoulet and de Janvry, 1995). Given the imperfect market structure, demand-side socioeconomic factors are included into the profit model (Pattanayak and Kramer, 1997). Education, household labor availability, non-agricultural activities, agricultural productivity, and household wealth (among others) appear to significantly influence forest dependency when markets are incomplete (Gunatilake, 1998). Therefore, profit becomes a function of prices, wages, environmental factors, and household characteristics. Additional socioeconomic characteristics are used to test differences in dependency between the two wealth groups (high and low). EMPIRICAL MODEL Sample Selection Model A censored regression model is used to estimate the statistical relationship between forest profits (forest dependency) and prices, wages, household's head years of logging and education, village forest size, average age of the household, and a wealth dummy variable. A censored regression model is used when information about the dependent variable is missing, yet there is information available for the independent variables. In the Ruteng case, forest profits are given for 49 out of the 97 observations, thus there appears to be some sample or self-selection bias. Forest profits are known for the actively harvesting/collecting households, but the "reservation" price, or the minimum profit that nonharvesting household require to harvest is not (Pindyck and Rubinfeld, 1998). A censored regression model is preferred over an ordinary leastsquare estimated model since ordinary least-square estimation will yield heteroscedastic (E(μi) ≠ 0) estimators which are biased and inconsistent. With the Ruteng case, heteroscedasticity (non-constant error variance) exist due to the self-selected problem noted above. The censored regression model can be calculated using a two-stage estimation technique, first suggested by Heckman (1979), which has been shown to produce efficient and consistent estimators. Heckman argued that specification error could be effectively eliminated if the values of the omitted variables are estimated, and then used as model regressors. First, Heckman suggested the estimation of the model using probit analysis to calculate the probability of a response or in the Ruteng case the probability that a household actively harvests timber or collects fuel wood. Probit analysis calculates estimators via a maximum-likelihood function, which has been shown to create efficient and consistent estimators of β and σ. The probit maximum-likelihood function will yield estimates of β /σ, thus predicted values of φi, Φi, and Zi can be calculated where φi and Φi are, respectively, the probability density of a standard normal variable and the corresponding cumulative distribution function (Heckman, 1979). Probit analysis determines the probability of response. With φi, Φi, and Zi determined, the inverse of the Mill's ratio, or the probability of harvesting, λi, is calculated as below. λi "is a monotone decreasing function of the probability that an observation is selected into the sample, Φ(-Z) = (1-Φ(Zi)). In particular, limΦ(-Z)→1 λi = 0, limΦ(-Z)→0 λi = ∞, and ∂λi/∂Φ(-Z)<0 (Heckman, 1979)." In the second stage of the estimation technique, λi is added as an additional model regressor to accommodate self-selection bias. Since 'λi predicted' approaches 'λi actual' as sample size increases, the mean of μi is normalized to zero and now ordinary least-square estimation will yield consistent estimates of the model parameters (Pindyck and Rubinfeld, 1998). Two-Stage Estimation using the Cobb-Douglas Form An empirical counterpart of the theoretical model is presented below. The model form was based on a Cobb-Douglas profit function, but is estimated using the two-stage estimation technique. The probit regression analysis is defined as Prob(yi>0) = β + βTi In PTi + βF In PF + βL In PL + βC In PC + βZ Z + error where Z is the vector of capital inputs/socioeconomic factors such that e z Z = e ZH + β ze ZE + β zy ZY + β zf ZF + β zw ZW + β zs ZS + β zd ZD +β zg ZG + β zi ZI) Where PTi is the price of harvested timber products, PL is the daily wage rate of labor, and Z as the vector of the socioeconomic variables. These variables include average age of household (H), forest size (F), a wealth dummy variable (W), head of household's education level (E), number of years logging (Y), size of household (S), age of the head of household (D), and whether the household head believed that loggers are often fined or arrested for logging within the park (I). The second stage estimates the Cobb-Douglas profit function using ordinary least-squares regression, but now includes λi, Ordinary least-squares estimation assumes model linearity. The functional form of the Cobb-Douglas model is only linear in parameters. The log linear specification is described as In Π = β + βTi In PTi + βF In PF + βL In PL + βZ Z + error