Project topics and materials
Get Complete Project Material File(s) Now! »
Additional Survey Data. As noted above, the survey included questions
related to a number of socio-economic variables, including the sex of the respondent, the age and education (both in years) of the household head, the size of the household, the household’s non-food expenditure, the household’s ownership of livestock (measured in tropical livestock units, where 1TLU=250kg), a measure of forest access, based on a GIS data, distances to the nearest town, land holdings, measures of wealth (whether a household has corrugated metal on their house or not) and experimentally determined household rates of time preference.
Descriptive statistics of this data are presented in Table 2.2.
We postulate that the demand for community forestry depends on covariates vindicated by economic theory. These include income and wealth, the price of the good, other prices and other taste shifters. From this list, covariates were sorted into three broad categories: (1) wealth and income – ownership of a house with corrugated roofing, land holdings and non-food expenditure; (2) the price of the good – livestock ownership, rate of time preference and education; (3) other prices – access to alternative forests, household size, and the distance to town.
Whereas proxies for wealth and income are relatively clear, variables used as price proxies merit further explanation. With regard to the price of the good, community forestry involves both temporal and intertemporal trade-offs. Starting with livestock, we expect that livestock has two opposing effects; wealth effect and prices effect. The wealth effect arises from the importance of livestock as major asset holding with the implication that the demand for community forestry increases with increased livestock wealth (more capacity to pay). In contrary, the establishment of community forestry on grazing land implies a potential income loss from livestock production, as grazing land is a major input of production. The foregone income is what we describe as the prices to be paid for community forestry establishment. We thus, expect that higher prices (higher holding livestock implying higher income to be given up) lowers demand for the goods under consideration. The net effect of livestock holding wills thus, either is negative or positive depends on the strength of either effect.
Moreover, given that community forestry establishment and management requires labour, income from alternative employment may have to be sacrificed, the value of which depends on the level of education. Therefore, such opportunity cost should be construed as part of the cost of establishing the community forest, in addition to the direct contribution suggested by the proposed bids. We, therefore, hypothesize that the level of education is expected to reduce the demand for community forestry.
Likewise, community forestry involves an intertemporal trade-off, in the sense that the benefits given up today to establish the programme must be weighed against benefits that accrue at later dates. We capture these intertemporal trade-offs through the household head’s rate of time preference, assuming that this rate is inversely related to the demand for community forestry6. Moreover, we presume that time preferences are dependent on household wealth7 measures (education, landholding and ownership of corrugated house etc.) and household head characteristics, such as age and sex.
With regard to proxies for other prices, recall that both access to alternative forests, typically open access natural forests, and the opportunity to buy from markets, as measured by the distance to town, are potential community forestry substitutes8. We, therefore, argue that better access to alternative forests and shorter distances to town will lower the prices of forest products obtained from these alternative sources. Subsequently, these measures are expected to be associated with reduced demand for community forestry. Moreover, the size of the household is likely to reduce WTP, partly because larger households have less discretionary income per capita and partly because a larger household increases the supply of labour available for collecting forest products from open access forests.
Welfare and Estimation Efficiency.
Although the existence of preference anomalies is interesting, on its own, the primary purpose of CVM is the elicitation of preferences. Preference anomalies should be controlled in the analysis, such that appropriate welfare estimates can be obtained. Upon calculation of the welfare effects, we found that the shift-effect model yielded the lowest median willingness to pay, ETB20.14, whereas the anchor-effect model and the shift-anchor model yielded slightly higher WTP estimates, ETB22.80 and ETB30.41, respectively10. However, WTP values for either the ascending or descending bid sequence subsamples were generally lower than for the full sample. As elucidated earlier, the likelihood ratio test results for model selection support the choice of the shifteffect and shift-anchor random effects probit models, although the shift-anchor effects model yields an inconsistent, with respect to theory, negative anchor effect.
As such, we report the willingness to pay estimate for these models as our measure of the community forestry welfare impact, which ranged between ETB20.14 and ETB22.80. However, our preferred estimate is the lower estimate of ETB20.14, as the higher estimate includes the inconsistent negative anchoring effect. In addition to examining the welfare effect, efficiency is also relevant, given the fact that the double-bounded model has been shown to be more efficient. The welfare estimate, median WTP, is computed from the model parameters, and, hence its distribution depends on the distribution of the parameters. The Delta method is used to derive the standard errors of the welfare estimates (Greene 1997). On the basis of efficiency, as measured by the relative standard errors, all of the random-effects probit models (naïve, shift, anchoring, and shift and anchoring models) outperform the single-bounded models. Amongst the randomeffects probit models, the shift-effect model yielded the lowest standard error estimate.
Chapter 1 General Introduction
Chapter 2 Contingent Valuation of Community Forestry in Ethiopia: ShouldWe Care About Preference Anomalies in Double-Bounded CVM?
Chapter 3 Does One Size Fit All? Heterogeneity in the Valuation of Community Forestry Programs
Chapter 4 The Welfare Effect of Common Property Forestry Rights: Evidence from Ethiopian Villages
Chapter 5 General Conclusion
Reference
