The model and findings of the study
In this section, after analysing the status quo situation, the mean of maximum WTP values of the forest-fringe households for forest conservation programme have been derived. Logit model has been used under single-bounded DC technique since the dependent variable is categorical. The value of the dependent variable in binary form is derived from the response of a specific bid offered to the respondent following a randomized trial.
The logit model used for the study is specified as
Pr (DC_yes/no) = α + \({\beta }_{1}\)DC_bid + \({\beta }_{2}\) hh_size + \({\beta }_{3}\) avg_age + \({\beta }_{4}\)avg_edu + \({\beta }_{5}\) food_suf + \({\beta }_{6}\) loan + \({\beta }_{7}\) eforest_dep + \({\beta }_{8}\) avg_dist + \({\beta }_{9}\) ln_inc +\(\epsilon\)
Dependent variable: Pr (DC_yes/no), where Pr (DC_yes/no) = ln \(\frac{{P}_{i}}{1-{P}_{i}}\) given \({P}_{i}\) = Pr (DC_yes/no = 1). Therefore, \({P}_{i }\) implies the probability of WTP amount greater than an assigned bid and ln\(\frac{{P}_{i}}{1-{P}_{i}}\) is the log odds ratio.
Random error term: ε
The independent variables used in the logit model are described in Table 3.
Table 3
List of explanatory variables with descriptive statistics
Variable | Descriptions | Minimum | Maximum | Mean | Standard Deviation |
hh_size | Number of members in a household | 1 | 11 | 4.920319 | 1.893575 |
avg_age | Ages of household members falling under the working group population have been considered to calculate average age (years) | 17.33 | 62.5 | 36.64386 | 9.126736 |
avg_edu | Level of education of household members in coding [Variants: Illiterate: 0, pre-primary (3–6 years children): 1, primary (classes 1–5): 2, medium (classes 6–8): 3, secondary (classes 8–10): 4, higher secondary and graduate (above class 10): 5, technical: 6] is considered to determine average education level | 0 | 4.17 | 1.63567 | 0.9662872 |
avg_dist | Average distance between the near by town where the forest dwellers sell the NTFPs and the village where they reside (kilometre) | 8 | 55 | 27.31076 | 16.08723 |
ln_inc | Rate of growth of total income of a household | 8.79 | 12.47 | 10.67443 | 0.6195218 |
DC_bid | Bid vector of INR 10, INR 20, INR 30, INR 40, INR 50 and INR 60 | 10 | 60 | 31.35458 | 14.27437 |
food_suf | Food sufficiency; Yes = 1, No = 0 (dummy variable) | 0 | 1 | 0.6294821 | 0.4839084 |
loan | Loan taken; Yes = 1, No = 0 (dummy variable) | 0 | 1 | 0.2788845 | 0.4493466 |
eforest_dep | Extent of forest dependence; High = 1, Low = 0 (dummy variable) | 0 | 1 | 0.7171315 | 0.451293 |
N = 251 |
Source: Done by authors |
[Table 3here]
Description of independent variables helps deliberating the demographics of forest-fringe households upon which the mean WTP value for forest restoration depends. Survey data revealed that almost one third of the dwellers were illiterate. Very few in numbers of 3 to 6 years children were started attending pre-primary schools. Technical education includes involvement of dwellers in different training programmes in search of jobs. Levels of literacy in the study area for 251 households comprising 1181 number of people are depicted in Fig. 2.
[Figure 2here]
The study considers average age of the working group population (14–65 years) of each household. Data of household income were collected segregating the sources of forest and non-forest income. However, total income of each household was considered for analysis so as to capture the true WTP value which was likely to depend upon affordability of the household members collectively. Forestry income is mainly seasonal in nature. Hence the dwellers are also involved in non-forestry works for survival at Simlipal. This becomes one of the primary reasons of varying monthly income of households through-out the year. On the way of analysis, at first annual income was calculated from different sources of earnings by all the working members in a household, and then average monthly income was calculated. Mean of monthly income stands at INR 4371.04, and lies between INR 500 and INR 21666.67. Rate of growth of monthly income has been considered as a determinant of WTP for forest conservation in the econometric model.
People in the study area were found to collect several NTFPs in different seasons with permission from the Forest Department. Conversely, Forest Department shares assured revenue (20–25 per cent) with the forest-fringe dwellers through the Forest Protection Committee at the advent of felling matured trees. This revenue is used for economic and social well-being of the residents. Other than the forestry activities, people in the study area take part in cultivation, work as small scale industrial labour, tourist guide, wage labour in different government schemes. Some of them were also engaged in animal husbandry, small business, etc. Frequencies and percentages of households belonging to different income groups are depicted in Table 4.
Table 4
Distribution of households among different income groups
Income groups (Annual ) in INR | Frequency | Valid per cent | Cumulative per cent |
< 25000 | 39 | 15.54 | 15.54 |
25001–50000 | 122 | 48.61 | 64.14 |
50001–75000 | 35 | 13.94 | 78.09 |
75001–100000 | 30 | 11.95 | 90.04 |
> 100000 | 25 | 9.96 | 100.00 |
Source: Calculated by authors based on primary data |
[Table 4here]
In the study region, almost all the villagers were found collecting fire wood through out the year for household needs. Collection of other NTFPs remain for self consumption or/and sale. Our study categorises households with high forest dependence who use forest products for both - self consumption and sale for sustenance. The extent of forest dependence is captured in binary form through the categorical variables ‘1’ and ‘0’. Profile of the study area indicating the extent of forest dependence is depicted in Table 5.
Table 5
Extent of forest dependence
Blocks | Total number of surveyed households (after omitting outliers in the data) | Number of households highly forest dependent with their percentages |
Bisoi | 43 | 32 (74.41) |
Bangriposi | 40 | 35 (87.5) |
Jashipur | 35 | 5 (14.28) |
Shamakhunta | 35 | 30 (85.71) |
Khunta | 54 | 34 (62.96) |
Thakurmunda | 44 | 44 (100) |
Total | 251 | 180 (71.71) |
Source: Calculated by authors based on primary data |
[Table 5here]
In the study area about 27 per cent households were found having loans at the interest rates higher than the market rate charged by the local money lenders. Mostly the women population being the members of Self Help Groups are entitled for taking loans from formal financial institutions. About 37 per cent households were found facing food insufficiency as they were not used to of meals twice a day through out the year. Among the households possessing agricultural land (75 per cent), survey data explored that mostly the farmers were marginal ones with having land less than one acre (47 per cent). Households who held domestic animals were considerable in number (87 per cent). Average distance between the near by towns and the villages where the forest dwellers reside came around 27 kilometres, which in most of the cases posed a serious problem to the household members in conducting market activities. Further, the data exposed that almost 95 per cent of people in the study region belong to tribal population. Hinduism was the dominant religion (82 per cent). Male-female ratio was 51 : 49. Estimation result of the logit model to determine the key factors affecting WTP is depicted in Table 6.
Table 6
Estimation result of Logit Model
Variable | Coefficient | Standard Error | VIF |
DC_bid | -0.0198259* | 0.0116008 | 6.19 |
hh_size | -0.0634285 | 0.0919813 | 10.54 |
avg_age | 0.031046* | 0.0173994 | 16.09 |
avg_edu | 0.2311699 | 0.1706739 | 4.80 |
food_suf | 0.2464419 | 0.3257945 | 2.92 |
loan | 0.7300865** | 0.3137494 | 1.51 |
eforest_dep | 0.2154173 | 0.3024925 | 1.86 |
avg_dist | -0.0186696** | 0.0098851 | 4.14 |
ln_inc | 0.3609471 | 0.2739205 | 36.53 |
Constant | -5.316048 | 2.832014 | - |
N = 251 Wald chi2(9) = 22.64 Prob > chi2 = 0.0071 Pseudo R2 = 0.0762 Log pseudolikelihood = -141.40857 Mean VIF = 9.40 |
* Significant at 10 per cent level * *Significant at 5 per cent level |
Source: Estimated by authors |
[Table 6here]
In Table 6, the Wald chi-square value is 22.64 and the associated p-value is 0.00, which show that the model as a whole has a good fit. The reported pseudo-R2 value of 0.07 confirms the fact. The fitted model shows that the independent variables like household size, bid vector and average distance have negative coefficients, while the coefficients of the other variables are positive. This negative coefficient of a variable implies that with a unit change in the respective variable, the odds ratio (or the probability) for payment decreases. The results are as per expectation since WTP increases with the rate of increase in total monthly income, food sufficiency, loan, forest dependence, average age and literacy level of the households.
Level of education and age augment the realization of sustainable use of natural resources. In the study, households with higher values of average education level and average age signify the presence of more educated and aged members in the families compared to the others. The matured thinking and awareness for resource preservation help enhancing the probability of payment of the forest-fringe households. People belonging to low income group and having dependence on forest require a steady flow of natural resources to maintain their livelihoods. Though, they eager to pay for the benefits received from forests, their payments remain low compared to the high income group for having less affordability. Therefore, increase in the rate of growth of monthly income and forest dependence of the forest-fringe households boost the probability of payment. Burden of loan often forces the people towards forest conservation and helps escalating WTP because of the fact that inability of repayment increases the dependence on natural resource base for present and future sustenance. Agriculture in the study area is found not to be profitable and mostly for self consumption. The dwellers collect different seasonal fruits and vegetables from the forest for daily consumption. Therefore, to get the provisioning services sustainably in order to avoid food insufficiency; probability of payment for conservation increases. Conversely, increase in household size, average distance between the near by town where the dwellers sell the forest products and the village where they reside, and bid amount lessen the probability of payments. The reasons lie in the facts that larger household size shrinks affordability; increase in average distance is involved with greater time cost and effort of the forest dwellers to get engaged in market activities; increase in bid amount reduces marginal utility of resource use. The study shows bid and average age are significant at 10 per cent level; loan and average distance are significant at 5 per cent level.
Problem of heteroskedasticity (which calls for lower precision of the coefficient estimates and increases the likelihood that the coefficient estimates are further from the correct population value), is an inherent problem in logit model, which is addressed using robust estimators in the study. It is recognized that some sort of multicollinearity is common in any cross section data including the one used in the present study. Therefore, the test of multicollinearity is performed to identify the correlation between the independent variables and the strength of that correlation. Variance inflation factors (VIF) for the model were estimated and the mean VIF is 9.40, which is within the tolerable scalar value of 10. This result exposes that the model does not suffer from any severe multicollinearity problem. The collinearity diagnostics is also presented in Table 6.
A correctly specified model generates a significant (_hat) and an insignificant (_hatsq). Testing for model misspecification is depicted in Table 7.
Table 7
Test for model misspecification
Variable (DC_yes/no) | Coefficient | Standard Error |
_hat | 0.8825663* | 0.4783854 |
_hatsq | -0.0750765 | 0.2736611 |
_cons | -0.0136891 | 0.2277569 |
N = 251 LR chi2(2) = 23.40 Prob > chi2 = 0.0000 Log likelihood = -141.3704 Pseudo R2 = 0.0764 |
* Significant at 10 per cent level |
Source: Estimated by authors |
[Table 7here]
The high value of Pearson chi2 (241) = 248.59 with Prob > chi2 = 0.3548, from the Hosmer-Lemeshow test, also reveals that the model is a good fit. Further, the study reveals that area under the Receiver Operating Characteristic (ROC) curve for the estimated model is 0.6854, as a measure of good predictive power (where, a model with no predictive power possesses an area of 0.5 and a perfect model has 1).
After confirming the validity of the estimated logit model, marginal effects have been estimated. Mean predicted probability of accepting a bid for the entire sample is 0.27. The marginal effects show that for a unit change in the bid amount, household size and average distance; the expected probabilities of saying ‘YES’ to the bid assigned decrease by 0.003, 0.012 and 0.003 units respectively. However, for a unit change in the average age, average education, food sufficiency, loan, forest dependency, rate of change of total income; the expected probabilities of saying ‘YES’ to the bid amount increase by 0.006, 0.046, 0.048, 0.156, 0.043 and 0.072 units respectively. The estimated marginal effects are depicted in Table 8.
Table 8
Marginal effects of the estimated logit model
Variable | dy/dx | Standard Error |
bid | -0.0039935* | 0.00232 |
hh_size | -0.0127762 | 0.01854 |
avg_age | 0.0062535* | 0.0035 |
avg_edu | 0.0465636 | 0.03419 |
food_suf^ | 0.0489128 | 0.06347 |
loan^ | 0.1560465** | 0.0698 |
eforest_dep^ | 0.0437883 | 0.0617 |
avg_dist | -0.0037606** | 0.00197 |
ln_inc | 0.0727041 | 0.05564 |
(^) dy/dx is for discrete change of dummy variable from 0 to 1 * Significant at 10 per cent level * *Significant at 5 per cent level |
N = 251 Marginal effects after logit y = Pr(yes_no) (predict) = 0.27960484 |
Source: Estimated by authors |
[Table 8here]
To end with the DC technique used, the mean WTP value for the sample was calculated and it turned around INR 16.38 per month for the proposed forest conservation programme. The estimation result of mean WTP is shown in Table 9.
Table 9
Measure | Coefficient | Standard Error | [95% Conf. Interval] |
WTP | 16.38199 | 28.1886 | -38.86664 71.63062 |
Source: Estimated by authors |
[Table 9here]
As mentioned, the study considers both closed-ended and open-ended bidding formats to get the mean WTP values. After getting response from the specific randomized bid offered to a particular respondent in terms of ‘YES’ or ‘NO’ answer under DC technique, we have gone through increasing or decreasing the bid amounts within the specified range (Rs 10 to Rs 60, determined from FGDs), respectively, until the first ‘NO’ or ‘YES’ answer comes to capture the actual WTP value in quantitative term under the bidding game framework. Repeating the same process with every respondent, the ultimate bid amounts accepted by the respondents were marked as a payment for forest conservation. Subsequently, mean WTP from those responses was calculated using arithmetic mean, and the value stands at INR 21.63 per month.
In any CVM study, it is likely to have two different values of mean WTP from the two different bidding formats. The presence of starting point or anchoring biases in the DC elicitation method makes this difference. Therefore, to avoid these types of biases, bidding game was used in the subsequent phase. Similar result has also been reflected in the study of Kohlin (2001), where while determining the actual WTP, the author found the difference in WTP values between asking for the consent for a discrete question and open-ended question followed by the discrete question. Since there is always a risk that the open-ended bid could be anchored to the price stated in the discrete question, our study utilises bidding game instead of asking the WTP value from the respondent directly.
Related Discussions for development planning
Preferences of the respondents to pay for nature conservation are very heterogeneous which depend on personal experiences and perceptions (e.g., whether respondents feel anxious for forest degradation), political views (e.g., the acceptance of strict legal protection of natural resources) and opinions on forest policy (e.g., preferences regarding privatization of public land) (Getzner et al. 2018). Zaiton et al. (2019) conducting a research work in Kuala Perlis, Malaysia, revealed that besides the bid level, socio-demographic factors like education level and marital status played important roles in influencing respondents’ WTP towards conservation of nature. Chukwuone et al. (2008) explored that the variables like wealth in heritance, occupation type, number of years of schooling and number of females in a household had positive and significant relationship with WTP. Conversely, gender (male-headed households), number of males in a household, distance from home to forests and starting point bias in offering bids were negatively related to WTP. In the study of Hema and Devi (2015), income from mangrove dependent activities exerted significant positive influence on WTP by the fishermen and farmers staying at Ernakulam and Kannur districts of Kerala, India. The factors influencing WTP of the respondents were estimated using multiple regression model with WTP as a dependent variable. It was observed that age, educational level and awareness index had significant positive influence on WTP of the direct dependent population. About one fourth of the respondents expressed their willingness to contribute for conservation both - in cash and kind or in a combination.
Alike the above stated research works on environmental valuation, our study explores that the proposed bid amounts have significant impact on determining the mean WTP of forest-fringe households for forest conservation. However, the acceptance of the proposed bid amounts is subjected to matured judgment and prevailing socio-economic conditions of the surveyed population. Therefore, government policies should be directed towards creating more livelihood opportunities which might trim down lending habit, forest dependence of people and boost up WTP for resource conservation. Importance should be given on agriculture and allied activities to make them profitable. Promotion of ecotourism and channelisation of indigenous knowledge of the tribal population to the greater world regarding the use of medicinal plants might help improving the socio-economic conditions of the residents. Further, exploration and accessibility of market for agriculture and forest produce might help regional development. In the study area, households have been found dependent on firewood for fuel use. But when stored carbon in plants is released and transferred back to the atmosphere or soil, the plants get die (Kayler et al. 2017). Even the dry leaves or twigs have certain role to play to retain ecological integrity which should be a serious concern to the policy makers. Hence knowledge based forestry with blended social and cultural factors may help efficient conservation.
The most favourable affair in the study is that almost all the forest-fringe households (99 per cent) have shown their concerns regarding sustainable use of forest resources, and therefore, their interest on payment for proposed conservation programme. The dwellers were willing to make the payments either from their savings or reducing their daily consumption expenditure. Therefore, government’s initiative for forest conservation might work well in the study region in collaboration with local people if their economic well-being is properly taken care of by enhancing livelihood opportunities (Agarwalla and Saha 2022). Also, the existing participatory forest management system should be made strong enough so that incentive-sharing mechanism can work well through the Forest Protection Committees. Alike this study, Bahuguna et al. (1994) also cited the need of collective management in rehabilitation of village ecosystems towards people's welfare in Harda forest division of Madhya Pradesh, India.