4.2. A Conditional Logit Model with Two-Way Interactions
The result of model 2 could have given further insight into how income level affected the option choice. The signs of the agricultural profit that were to be interacted with for all of the income levels was positive. The variables, namely, low income level*agricultural profit and middle income level*agricultural profit, were highly statistically significant at the 1% significance level. The variable, high income level*agricultural profit, was statistically significant at the 5% significance level. It implied that the low and middle income households were sensitive to the agricultural profits. The variable, ASC, was positive and statistically significant, which meant that the respondents preferred the current status to the hypothetical scenarios. The biodiversity attribute for each income group was not statistically significant. This meant that the biodiversity in the interaction with the income levels was not a significant determinant of the option choice. This was in contrast to the results, which showed that the economic value, regarding the biodiversity attribute, could offer a reliable information in order to estimate the welfare losses by the reduction of the biodiversity levels [30
]. However, the interaction terms, low/middle/high income level*water quality, were highly statistically significant. It implied that the water quality in all of the income levels was an important factor for the choice option of the respondents. This was in line with the previous studies, which showed that the customer placed a high value on maintaining a clean water supply [15
]. The concern for the water quality of the watershed in the study regions, as well as the lack of significant differences in the MWTA between the upstream and downstream residents, could have been explained by the importance of the drinking water use.
Furthermore, the coefficients for the water quality, with the interactions of the upstream farm and downstream households, and with income levels, were highly statistically significant, with a positive sign. It implied that the respondents were prone to being significantly concerned about the water quality in the economic characteristics and districts that affected the water quality of the Soyang watershed. This result could be explained by the fact that, even though the roles of the respondents were different along the Soyang watershed, they were especially concerned about the water quality level in the Soyang watershed. This was in line with the result, which showed the importance of the socio-economic determinants in the heterogeneous choice of the respondents [33
We further considered the estimation results by the MWTA, in order to identify the preference for water quality, which was differentiated by each income range. The implicit costs were calculated using the coefficient of the agricultural profits. In the case of the Soyang watershed, the MWTA values for the water quality implied a change from one water quality level to another, which meant that the increase in one unit of improved water quality was as a result of the reduction of fertilizers and pesticides, which led to a reduction in agricultural production. With respect to the low, middle, and high income levels, the calculated annual MWTAs of the respondents ranged from KRW 2,874,638 to KRW 7,415,775, for the reduction of the water pollution from advanced agricultural farming practices. The result implied that the MWTA for the water quality was significantly different between the low, middle, and high income levels. They suggested that the improvement in water quality was considerably important for each income level and district in South Korea.
4.3. A Conditional Logit Model with Three-Way Interactions
The results with the three-way interactions showed that the coefficients of the upstream/downstream*low/middle income level*agricultural profit were statistically significant at the 1% significance level. The parameter of the upstream*high income level*agricultural profit was statistically significant at the 10% significance level. The coefficient of the downstream*high income level*agricultural profit, was, however, not statistically significant. It implied that the upstream and downstream respondents in the low and middle income levels tended to prefer the increase of water quality of the Soyang watershed. This meant that the improvement of water quality was not a significant determinant of the downstream high income respondents, regarding the choice. Moreover, the parameters that were related to the biodiversity level with the three-way interactions were not a significant determinant on the choice of the upstream and downstream respondents, for each income group. This might have been explained by the different local communities that had less perception about biodiversity conservation.
Moreover, we estimated the implicit costs, or MWTAs, for each of the water quality attributes in the different income levels by the interaction with the different stakeholders. When the two respondent groups were compared, it could be seen that the marginal values of the attributes were different at different income levels. The annual MWTAs of the upstream respondents, for the water quality range, were from KRW 3,484,673 (low), KRW 3,746,120 (middle), to KRW 9,616,920 (high), while those for the downstream respondents, for the water quality, varied from KRW 1,773,511 (low), KRW 2,532,524 (middle), to KRW 5,420,074 (high). Regarding the result from the upstream respondents, the difference between the low and middle income levels was relatively small. The upstream respondents with the high income level had the highest costs for improving the water quality. With respect to the downstream respondents, however, the implicit costs of the upstream respondents were shown to be about three times higher between the low and high income levels, while the implicit cost of the low income level was different to that of the middle income level, by about 1.4 times. Overall, the difference of the implicit costs between the low and middle income levels was relatively small in the upstream and downstream respondents. With respect to the high income level, the result of the MWTA showed a big difference. Their disparity was larger in the interaction terms that were associated with the districts (upstream/downstream households) and income levels. It implied that the estimated marginal values were different in each income level and each district. The MWTA disparity between the upstream and downstream householders in each income group was based on the fact that the downstream citizens had a higher annual income than upstream farmers’ annual net income.
This result provided meaningful insights for the policy makers, with the importance of income-specific and district-specific differences that were associated with environment protection through agriculture in further researches. Moreover, in South Korea, less studies focused on the use of CE and research that was related with biodiversity stated preferences was needed to enhance the perception of public. Further research could have investigated the non-market benefits of biodiversity conservation and elicited the WTP/A for the biodiversity attributes, which used the CE method, considering the income and district effects among the different stakeholders.
As both the upstream and downstream residents utilized the clean water, many studies highlighted the importance of the appropriate water use charge that the downstream citizen and metropolitan areas had to pay for the water protection areas and to compensate the upstream residents who were regulating the economic activities, such as housing and farming [37
]. The water use charge that was obtained from the downstream citizens could be applicable to support for upstream low income farmers suffering from economic restriction and change in farming techniques. However, despite the continuous change in the water policy of South Korea and the attempts to solve the reclusive water quality conflicts, over the water rights between the upstream and downstream regions, which were as a result of the non-point source pollution in the highland areas, the Gangwon Province had less support and excluded the resident support programs that had received support from the water use charge. This was in line with the result, which showed that it was necessary to improve the institutions and laws that were associated with the residents’ support program created by the Han River Management Fund [37
Methodologically, it was a limitation that our study considered only CLM and CLM with interactions. In order to compare the results with the advanced models, we recommended the advanced models, such as error component models, which allowed for the control of the heteroscedasticity over the choice sets, should be utilized. This suggested, therefore, the use of different model specifications that accounted for the heterogeneous preferences, with careful construction of the choice sets and effective data collection. More importantly, the incorporated preferences’ differences, between the districts at different income levels, should have been investigated in the decision-making for the sustainable water management.
Once the specific costs and WTP in the CE model had been provided to the policy decision makers, the specific monetary values for the environment could be the reference points for the sustainable water management planning and designing. Thus, we suggested that the inclusion of a cost attribute, such as direct payments or agricultural program participation, should be done with care in the choice experiment, in order to elicit monetary values for the environmental improvement. Further studies should focus on how the choice experiment could be used to provide both welfare estimates, which corresponded to the policy changes involving one or more attributes with offered direct payments, and community ranking of multiple policy options.
Regarding the sustainable water and land management, a recent study provided a win-win strategy and was challenged under the nexus of the water–energy–food perceptive, as demonstrated by the synergies and trade-offs for implementing the efficient governance and institutions [39
]. With the emphasis on soil conservation measures, some studies focused on quantifying the costs and benefits for sustainable land use management practices [42
]. Previous studies addressed land use and land cover changes in the spatial and temporal changes in the watershed as well as sustainable land use management [44
With respect to the effectiveness and efficiency of the sustainable use of resources and services, an integrated water management interconnected land use was needed. Regarding the environmental issues with excessive abundant nutrients, many studies highlighted the water pollution treatment for water quality improvement and for efficient diffuse source management, under different land-use practices [45
]. Moreover, further studies will be needed to implement urban water management into future urban water infrastructure, which is associated with sustainable land use changes and planning.
Many studies have considered cost-effective solutions for combating the eutrophication of coastal ecosystems in sustainable land management [48
]. In addition, with an attempt to provide various policy programs for cost-effective conservation practices, a focus on the water quality trading markets, trading costs in the river basin were necessary for improving the water quality. Further studies could elicit the preferences of multiple stakeholders for different positive nitrogen management practices, which would show the trade-off between the land use change and economic consequences.