A data-driven framework and decision-support tool for regional diversification and workforce readiness in Canadian cities

Canadian regional economies have diverged over the past half century. Existing tools observe either what a region produces or who works there, rarely both, and rarely over long horizons. This thesis makes four contributions using two independent Canadian datasets. The first constructs a fifty-year trajectory typology of Canadian census metropolitan areas and agglomerations. Each region-year is placed on a grid crossing economic complexity with portfolio breadth, and the sequences are clustered into long-run paths. Mean complexity declines steadily across them while mean portfolio breadth does not, so the two are distinct dimensions of industrial structure that a single complexity score collapses. The second works at the job-cell level, the intersection of a region, an industry and an occupational group. Entry into a previously absent job-cell is modelled with a linear probability model carrying fixed effects for region, transition period and industry. Industrial relatedness density predicts entry across a nested specification ladder and under alternative entry definitions. Canadian regions take up new work next to what they already hold, at a resolution finer than the industry. The third operates at the occupation level on a separate cross-section, adapting a requirement-sufficiency framework to British Columbia metropolitan areas. Readiness is the employment mass lying within a proximity threshold of a target occupation's requirements, and depends heavily on where that threshold sits. The metropolitan advantage widens as the bar rises: the binding constraint in smaller regions is not a general shortage of skilled workers but a shortage of workers positioned in the attribute neighbourhood the target requires. The fourth is a prototype decision-support tool, the Regional Industry Explorer, developed as design science research in which the design is itself the contribution. It is designed to carry the outputs of the first three to practitioners through selection alone, requiring no statistical software or re-running of the pipeline, and it holds the two datasets in separate views so that no quantity it displays is computed across them. Informal practitioner feedback shaped its development; it has not yet been formally evaluated.