Where a national number hides the most
A national figure averages a whole country into one number, and in doing so it erases the places where risk actually concentrates. Two patterns stand out in the subnational data: countries whose highest-risk regions sit far above their national average, and the corridors themselves, clusters of high-risk regions the country score averages away.
| Country | National tier | Highest-risk region (risk) | Range across regions |
|---|
The Philippines is the clearest case in the data: nationally middling, but its highest-risk regions (the Cordillera highlands, Masbate, Western Samar, and Sultan Kudarat) reach a local risk near 0.71. In Thailand, the northeastern Isan belt dominates the highest-risk regions globally (local risk near 0.74–0.78). These regions are read directly from the regional surface, not chosen as illustrations.
| Highest-risk regions in the data | Country | Local risk |
|---|
Where the national number misleads most
Some countries look unremarkable on the national map yet hide a region among the most at-risk places on earth. This is measured as divergence: the gap between where a country ranks nationally and where its highest-risk region ranks globally. The three cases below have the widest gaps in the data.
The bar runs from the lowest to the highest regional risk worldwide. The grey mark is the country’s national percentile; the red mark is its highest-risk region’s. Every value is read straight from the regional surface. Source: divergence_headline.csv.
What each layer is, and is not
Country composite: the published national structural-risk score (184 of 195 scored). Subnational surface: admin-1 risk derived from IPUMS-International census microdata and GDIS disaster locations — 1,732 regions across 92 countries on the map, of which ~1,450 regions in 79 countries pass reliability filtering and carry a score; countries with no surface carry no fill and confirm “no data” on hover (not low-risk, just not measured), and the surface weights (0.70 precarity + 0.30 shock) are a documented choice. Labor migration pressure: remittance flows as a share of GDP, both directions: blue marks origin economies reliant on labor sent abroad (inflows received), and red marks destination economies whose residents send the most home (outflows paid), the demand side, where migrant labor concentrates (the Gulf, Luxembourg, Switzerland). Corridors are proxied from two observed series — outbound remittances sent as a share of GDP (the origin side) and inbound migrant population share (the destination side) — not from tracked movements of people; not part of the composite score. Spatial clusters (LISA): statistically significant High-High / Low-Low / High-Low / Low-High clusters across admin-1 regions (FDR-corrected, global Moran’s I = 0.81). A separate country-level analysis (not mapped) shows the cross-border clustering largely tracks governance; see the methodology.
Coverage ceilings are real. No layer is global. An unfilled or absent country always means “no reliable data here,” never “zero risk”; hover to confirm. The composite is the published product; the other layers are second-pass analytical overlays, labelled as such.
See the full rankings · Open a country profile · How the spatial layers are built · Limitations & sources