Statistical analysis for the development of national average weighting factors—visualization of the variability between each individual’s environmental thoughts

Abstract

Purpose

Weighting is one of the steps involved in life cycle impact assessment (LCIA). This enables us to integrate various environmental impacts and facilitates the interpretation of environmental information. Many different weighting methodologies have already been proposed, and the results of many case studies with a single index have been published. However, a number of problems still remain. Weighting factors should be based on the preferences of society as a whole so that the life cycle assessment (LCA) practitioner can successfully apply them to every product and service. However, most existing studies do not really measure national averages but only the average of the responses obtained from the people actually sampled. Measuring the degree of uncertainty in LCIA factors is, therefore, one of the most important issues in current LCIA research, and some advanced LCIA methods have tried to deal with the problem of uncertainty. However, few weighting methods take into account the variability between each individual’s environmental thoughts. LIME2, the updated version of life cycle impact assessment method based on endpoint modeling (LIME), has been developed as part of the second LCA national project of Japan. One of the aims of LIME2 is to develop new weighting factors which fulfill the following requirements: (1) to accurately represent the environmental attitudes of the Japanese public, (2) to measure the variability between each individual’s environmental thoughts and reflect them in the choice of suitable weighting factors.

Methods

This study adopted the technique of conjoint analysis, which is currently the most advanced methodology available in the field of environmental economics. Using a random sampling process, 1,000 individual responses were collected. Every response was based on an interview survey designed to minimize bias. We used a random parameter logit model to estimate the preferences of society. Statistical values based on this model can be considered to reflect the variability between each individual’s environmental thoughts. The calculated results can then be used to develop integration factors in LIME2, enabling us to express LCIA results as a single index, such as external cost.

Results and discussion

The calculated values were significant statistically at the 1% level (all p values for the safeguard subject coefficients were less than 0.0001), with the exception of “social assets.” Based on the calculated results, two types of weighting factor, an economic valuation and a dimensionless index, were obtained. A relative comparison of importance among these four categories indicates that “biodiversity” receives the highest level of recognition, followed by “human health” and “primary production,” while the weight of “social assets” rate lower than the other safeguard subjects, in comparison. Using the calculated results produced by the RPL model, the probability density of the variables for individual preferences could then be derived and displayed. The coefficients of variance for the estimated weighting factors were relatively small (in the range from 0.1 to 0.3).

Conclusions

Accurate weighting factors representing the environmental attitudes of the Japanese public are needed in order to conduct general-purpose LCA for Japanese products. Random, unbiased sampling throughout Japan and an interview survey carried out on 1,000 respondents enabled us to address and solve the problems found with past weighting methodologies. We confirmed that the results of comparisons carried out among safeguard subjects were statistically significant, and showed that the contents of the questionnaires were well understood by the respondents. This study succeeded in visualizing the variability between each individual’s environmental thoughts in order to improve the transparency of the weighting factors—expressing the difference in individual preferences within a certain range. This data can be used to develop integration factors with statistical values which can then be applied to uncertainty analysis in future LCA case studies.