Estimating the Household’s Preference and spatial dependence for a Solid Waste Management System in Nepal; A Choice Experiment Approach
Description
Solid waste management system in the urban cities is always considered as a prerequisite for development. However, the household waste management system in most of the developing countries' urban cities significantly lacks proper strategy. According to the World Health Organization (UN Habitat, 2017), close to 54.0% of the global population live in urban cities, and urban growth is approximately 2.0% per year. The immense pressure on cities from this emerging population creates an unfavorable condition to society and the environment especially from an upward shift of household waste production. Illegal dumping beside the river and roads, shortage of municipality resources of collection, and a lack of awareness have a great negative impact on the overall ecosystem mostly incurred by water and air pollution. Most of the municipalities are currently providing a waste management service at lower levels than the accepted standard. However, the municipalities are barely aware of the preference and opinions of the households while formulating the policies about waste management services. By including the preference of households, a better waste management service can be developed to decrease environmental degradation as well as the overall quality of life. This study investigates the preference for a better waste management service and their willingness to pay for Siddharthanagar Municipality, Nepal by employing a Choice Experiment (CE) method. Additionally, we analyze the spatial dependence of marginal willingness to pay (MWTP) throughout the study area by employing hot spot analysis. Finally, we run a separate Spatial Autoregressive Model (SAR) to understand the neighborhood effect on MWTP for each attribute of the waste management system. This study uses primary survey data at the household level that was conducted in Siddharthanagar municipality (13 wards‡) on 593 households in June 2019. The CE section comprises 593 household’s responses from only the adults (>= 18 years in age). Each respondent was shown three choice sets and three alternatives in each choice set. In each choice set, two are the proposed alternatives (Solid Waste Management Program A and B) and one is the current management program (Status-quo). Each alternative has five attributes and they have a varying level. The attributes are recycling, composting, dumping, percent of waste collection, and price. The respondent chose 1 alternative from each of the choice set. This procedure makes a total of 1779 of observations in total. The choice sets were designed using the R (DoE. Base function), choosing from an orthogonal array from the full design. The orthogonal design provided 72 choice sets which are blocked into 12 versions of the questionnaire with 3 choice sets in each questionnaire (3 choice sets * 2 alternatives * 12 versions = 72 choice sets). We use the Random Utility Framework (RUM) to analyze the preference for a waste management system. As individual-level data can exhibit multiple levels of heterogeneity in making choice, we employ several models to account for heterogeneity in finding the MWTP for management attributes. A conditional Logit (CL) model is estimated as the baseline model. However, due to the two limitations of CL, Independent and Identically distributed error (iid) and Independence of Irrelevant Alternatives, we proceed to estimate a Random Parameter Logit (RPL) model which account for preference heterogeneity. Additionally, to account for the ‘scale’ heterogeneity, a Generalized Multinomial Logit Model (GMNL) is estimated. We also incorporate the stated level of certainty of individuals in making choices into the GMNL model to explore the heterogeneity even better. Finally, a Latent Class Model (LCM) is estimated to explore the heterogeneity over different unobserved classes in the data. To find whether MWTP of each attribute does follow any spatial pattern, a hot spot (cold spot) analysis is used to find the spatial cluster of high (low) MWTP values over the geographic surface of the study area. To determine whether a specific spot is hot or cold, a Getis-ord Gi* statistic is calculated whose outcome follows a Z distribution. As Getis-ord Gi* gives a vector of points of Z values, we interpolate those points over the surface by kriging Interpolation method. Finally, we run a separate Spatial Autoregressive Model (SAR) to understand the neighborhood effect on MWTP for each attribute of the waste management system. The MWTP of each attribute is extracted from the GMNL model with the level of certainty based on the lowest Akaike Information Criteria (AIC). We find significant positive MWTP for each attribute and the highest WTP is for the dumping attribute which is 157.86 Nepali Rupees. Individuals have the lowest level of WTP for recycling attribute which is 12.82 Rupees. The survey shows that almost 60% of household waste is organic waste. This fact explains the lower WTP for recycling. Because households prioritize the management of organic waste more that the recyclable waste, they have higher WTP for composting attribute which is 20.57 Rupees. The GMNL model shows the sign of both scale and preference heterogeneity in choices. The LCM shows significant heterogeneity. The hot spot analysis shows a significant difference spatially in MWTP values. Out of five attributes, ward number 11 (there are 13 wards in total) exhibits the hot spots in three attributes. The reason behind this fact is the limitation of access of this ward to the services provided by the municipality. The municipality currently does not provide any service to this ward which prompts households to pay higher for the proposed waste management system. Overall, this study provides ample information about the household’s perspective to the municipality. The hot spot analysis has a significant implication in terms of policy perspective towards the differential provision of service or payment scheme.
Estimating the Household’s Preference and spatial dependence for a Solid Waste Management System in Nepal; A Choice Experiment Approach
Solid waste management system in the urban cities is always considered as a prerequisite for development. However, the household waste management system in most of the developing countries' urban cities significantly lacks proper strategy. According to the World Health Organization (UN Habitat, 2017), close to 54.0% of the global population live in urban cities, and urban growth is approximately 2.0% per year. The immense pressure on cities from this emerging population creates an unfavorable condition to society and the environment especially from an upward shift of household waste production. Illegal dumping beside the river and roads, shortage of municipality resources of collection, and a lack of awareness have a great negative impact on the overall ecosystem mostly incurred by water and air pollution. Most of the municipalities are currently providing a waste management service at lower levels than the accepted standard. However, the municipalities are barely aware of the preference and opinions of the households while formulating the policies about waste management services. By including the preference of households, a better waste management service can be developed to decrease environmental degradation as well as the overall quality of life. This study investigates the preference for a better waste management service and their willingness to pay for Siddharthanagar Municipality, Nepal by employing a Choice Experiment (CE) method. Additionally, we analyze the spatial dependence of marginal willingness to pay (MWTP) throughout the study area by employing hot spot analysis. Finally, we run a separate Spatial Autoregressive Model (SAR) to understand the neighborhood effect on MWTP for each attribute of the waste management system. This study uses primary survey data at the household level that was conducted in Siddharthanagar municipality (13 wards‡) on 593 households in June 2019. The CE section comprises 593 household’s responses from only the adults (>= 18 years in age). Each respondent was shown three choice sets and three alternatives in each choice set. In each choice set, two are the proposed alternatives (Solid Waste Management Program A and B) and one is the current management program (Status-quo). Each alternative has five attributes and they have a varying level. The attributes are recycling, composting, dumping, percent of waste collection, and price. The respondent chose 1 alternative from each of the choice set. This procedure makes a total of 1779 of observations in total. The choice sets were designed using the R (DoE. Base function), choosing from an orthogonal array from the full design. The orthogonal design provided 72 choice sets which are blocked into 12 versions of the questionnaire with 3 choice sets in each questionnaire (3 choice sets * 2 alternatives * 12 versions = 72 choice sets). We use the Random Utility Framework (RUM) to analyze the preference for a waste management system. As individual-level data can exhibit multiple levels of heterogeneity in making choice, we employ several models to account for heterogeneity in finding the MWTP for management attributes. A conditional Logit (CL) model is estimated as the baseline model. However, due to the two limitations of CL, Independent and Identically distributed error (iid) and Independence of Irrelevant Alternatives, we proceed to estimate a Random Parameter Logit (RPL) model which account for preference heterogeneity. Additionally, to account for the ‘scale’ heterogeneity, a Generalized Multinomial Logit Model (GMNL) is estimated. We also incorporate the stated level of certainty of individuals in making choices into the GMNL model to explore the heterogeneity even better. Finally, a Latent Class Model (LCM) is estimated to explore the heterogeneity over different unobserved classes in the data. To find whether MWTP of each attribute does follow any spatial pattern, a hot spot (cold spot) analysis is used to find the spatial cluster of high (low) MWTP values over the geographic surface of the study area. To determine whether a specific spot is hot or cold, a Getis-ord Gi* statistic is calculated whose outcome follows a Z distribution. As Getis-ord Gi* gives a vector of points of Z values, we interpolate those points over the surface by kriging Interpolation method. Finally, we run a separate Spatial Autoregressive Model (SAR) to understand the neighborhood effect on MWTP for each attribute of the waste management system. The MWTP of each attribute is extracted from the GMNL model with the level of certainty based on the lowest Akaike Information Criteria (AIC). We find significant positive MWTP for each attribute and the highest WTP is for the dumping attribute which is 157.86 Nepali Rupees. Individuals have the lowest level of WTP for recycling attribute which is 12.82 Rupees. The survey shows that almost 60% of household waste is organic waste. This fact explains the lower WTP for recycling. Because households prioritize the management of organic waste more that the recyclable waste, they have higher WTP for composting attribute which is 20.57 Rupees. The GMNL model shows the sign of both scale and preference heterogeneity in choices. The LCM shows significant heterogeneity. The hot spot analysis shows a significant difference spatially in MWTP values. Out of five attributes, ward number 11 (there are 13 wards in total) exhibits the hot spots in three attributes. The reason behind this fact is the limitation of access of this ward to the services provided by the municipality. The municipality currently does not provide any service to this ward which prompts households to pay higher for the proposed waste management system. Overall, this study provides ample information about the household’s perspective to the municipality. The hot spot analysis has a significant implication in terms of policy perspective towards the differential provision of service or payment scheme.