Reform Window Tracker

When do governance windows open — and do the gains last?

Every significant democratic improvement episode since 1945, detected algorithmically from the V-Dem dataset: 261 windows across 138 countries. The headline findings reframe how reform happens. Improvement is episodic, not gradual — two-thirds of all democratic progress since 1945 occurred in just five per cent of country-years. Windows close fast: the median lasts two years. They open unpredictably: three-quarters open from stagnation or outright decline. And the gains mostly stick: ten years on, the median episode has kept everything it won.

Episode detection: V-Dem v16 electoral democracy index (v2x_polyarchy), 1945–2025. A window = clustered year-on-year jumps of ≥0.05 with a cumulative gain of ≥0.10 (10 points on a 0–100 scale). Retention measured ten years after the window closes. This detects democratic openings; capability-led improvements without electoral opening (e.g. Uzbekistan after 2016) follow a different path and are deliberately not captured here.  ·  Governance Compass  ·  Risk Horizon
67%
of all improvement since 1945 happened in 5% of country-years
2 yrs
median window duration · 89% close within 3 years
73%
of windows open from stagnation or decline — not momentum
103%
median share of gains retained 10 years later
12%
of windows fully reverse within a decade
14 yrs
median wait between windows in the same country
Eighty years of windows

Each dot is one episode, placed by the year it opened (horizontal) and the size of its democratic gain (vertical, on V-Dem’s 0–1 polyarchy scale). Colour shows what happened in the ten years after the window closed. Click a legend item to isolate a category. The 1990s tower over everything — 68 windows in one decade — but windows have kept opening at a near-normal rate through the “democratic recession”.

Recent windows — opened since 2017

These are the openings still inside or just beyond the historical danger zone. The base rates above are their odds: most will keep what they won; roughly one in eight will lose all of it.

All 261 episodes
CountryWindowDuration FromToGain Pre-trendRetained @10yOutcome
Reading this honestly. The tracker detects windows from outcomes, so an opening becomes visible here only once it has begun to register in V-Dem’s expert coding — typically with a one-to-two-year lag. It describes base rates; it does not predict. The polyarchy index measures electoral democracy, so administrative or capability-led reform without political opening is invisible to this method by design — the Compass and Risk Horizon capture that dimension. Retention at ten years compares the level a decade after the window closed with the pre-window level; values above 100% mean the country kept its gains and added more.