For each parliamentary vote we estimate what share of named-party voters stayed with the same party across two consecutive elections. This isolates party loyalty from swings in turnout.
How to read it
Stay rate — share of voters who chose the same named party in both elections.
Churn — share of named-party voters who switched parties between cycles.
Top defection — the single largest party-to-party movement (source, destination, share of source-party outflow).
By region — loyalty per multi-mandate region, plus the most stable and most volatile regions.
Persistence is estimated from per-section voter flow (NNLS Goodman regression + RAS scaling). The signal is robust at the regional level but is an aggregate estimate, not an individual-voter measurement.