Codebook

Each row in the SPI panel scores an (agency, year) cell against ten L0 risk domains and roughly fifty L1 sub-domains. The L0 layer is the chapter-level taxonomy of the book; the L1 layer is a finer-grained decomposition used in the empirical work.

L0 domains

Full L1 sub-domain definitions and example phrases — coming next pass. They live in data/panels/codebook.json and will be rendered here.