APAC Tertiary-Institution Quality Index

A cluster-level index of the quality trajectory of top Asia-Pacific university clusters, framed around local economic value-capture rather than global prestige.

Live site: https://curioputterings.github.io/apac-tertiary-quality-index/

• 32 clusters across SE Asia, Korea, Japan, India, Bangladesh, Australia, New Zealand, Central Asia & Mongolia, Hong Kong
• 50 anchor institutions
• Four pillars: research frontier · innovation transfer (university→domestic-firm) · human capital · talent retention
• Data: OpenAlex (research, industry co-authorship, retention proxy) + World Bank (STEM share, enrolment)

Framing hypothesis

Local institutional quality is a necessary condition for sustained economic competitiveness — so quality is measured by locally-captured value, and shown as a level × trajectory map rather than a single league table.

Two lenses

• Absolute capture — how much university value the economy actually retains and feeds to local firms (the competitiveness-relevant quantity).
• Model-adjusted — university quality net of economic structure (value-capture metrics scored against each cluster's own economic-model baseline), which fairly credits hub and resource economies.

The gap between the two is the insight.

Repository

• index.html — the Pages site (two maps, findings, method links)
• apac-index-fullroster-map.md — full results, rankings, two-lens framing, sensitivity analysis, caveats
• apac-institution-quality-index-spec.md — methodology specification (pillars, formulas, normalisation, weights)
• apac-index-provisional-map.md — earlier 5-cluster pilot (superseded)
• pilot_data/ — reproducible pipeline: resolve.py → stage2.py … stage8_dualmodel.py, plus aggregated .tsv outputs

Raw OpenAlex API samples are intentionally not committed (they contain third-party researchers' email addresses); only aggregated metrics are published. Re-run the scripts to regenerate them.

Caveats

A research instrument, not a ranking authority. Some Pillar-A figures (Japan, Hong Kong) and hub-retention proxies rest on imputation or bibliometric approximation; small-output clusters have noisy samples; the index does not establish causation. See the full-roster doc for the complete list.

Support

If this is useful, you can buy me a coffee.

Contact: curioputterings@proton.me