University of British Columbia
Clustering populations for holistic policymaking : promises and limits in British Columbia’s energy transition
Abstract
dc:descriptionClustering populations across multiple characteristics may enable more holistic policymaking, but policymakers lack guidance for running clustering analysis and interpreting ambiguous results. This thesis develops an artificial intelligence framework and applies it to municipalities in British Columbia to analyse barriers and enablers to energy transitions. It discusses how to evaluate cluster cohesiveness, characterize clusters, and use results to inform policy design. PCA, k-means, and eta-squared are used to cluster and characterize 147 municipalities for targeting residential (heat pump) and transportation (electric vehicle) electrification policies. Ten datasets were cleaned, harmonized, and merged, capturing 14-year trends in residential and transport energy (energy per capita and CO2e per unit energy), 15-year trends in social systems (population, income, housing, education, and labour force participation), geography (remoteness, latitude, and longitude), and institutional capacity for climate action. The residential and transportation scenarios produced two-cluster partitions in which one cluster compounded remoteness, lower education, smaller and stagnant populations, and weaker institutional climate capacity. A new policy could acknowledge the co-occurrence of these barriers within the cluster, and their recurrence across both residential and transportation scenarios. For example, a northern and a southern hub could simultaneously provide training for heat pump and EV charger installation, logistical and financial support for equipment delivery to remote communities, and governance support for local climate planning. Meanwhile, the other, less constrained, cluster could receive lighter instruments such as income-differentiated rebates or regulations. This method can make policy targeting more holistic than existing targeting mechanisms that use only municipal latitude or household income to differentiate policy treatment. Clustering results were sensitive to methodological choices and sometimes ambiguous, and therefore need to be interpreted critically. Clustering quality scores were lower than expected and validation metrics gave conflicting advice on how many cluster partitions to select. Furthermore, the co-occurrence of barriers in a cluster does not tell us why these barriers exist and whether they are the key factors blocking energy transitions. Finally, the analysis excluded the smallest communities for lack of data so that other municipalities could be analysed with more variables. The trade-off between equity and thorough analysis needs to be navigated carefully.
Degree
thesis:*- Name thesis:degree_name
- Master of Science - MSc
- Level thesis:degree_level
- master's
- Discipline thesis:degree_discipline
- Resources, Environment and Sustainability
- Grantor dc:publisher
- University of British Columbia
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Kouzak Campos da Paz, Aloysio
- Contributors dc:contributor
-
- Ellis, Naoko
- Gopaluni, Bhushan
Rights
dc:rights- Statement dc:rights
-
- Attribution-NonCommercial-NoDerivatives 4.0 International
- Language dc:language
- eng
Identifiers
dc:identifier.*- Handle dc:identifier
- http://hdl.handle.net/2429/95426