Involuntary Turnover Rate

Number of jobs inside a public funder's jurisdiction that were RETAINED across the reporting period — a FLOW, not a stock, and the second half of the created/retained pair SSBCI 2.0 and most US state incentive programmes require. CONVENTION (pinned deliberately — retention has genuinely contested definitions, and two boards reporting incomparable numbers into one portfolio total is the failure this pin exists to prevent): an in-jurisdiction position that was FILLED AT THE START of the reporting period AND is STILL FILLED BY THE SAME EMPLOYEE, in-jurisdiction, AT THE END. Three consequences follow, and they are the convention: (1) BACKFILLS DO NOT COUNT — same seat, new person is not a retained job (that person is `hr.in_jurisdiction_jobs_created` if their first day fell in the period); (2) it is ENDPOINT CONTINUITY, not continuous employment on every day of the period, because endpoints are what payroll records can actually evidence; (3) retained and `hr.in_jurisdiction_jobs_created` are DISJOINT — a job is one or the other in a period, never both, so the two are safe to add. The invariant this yields, which validates the pair against the stock series: retained employees are a subset of the in-jurisdiction workforce at BOTH endpoints, so `retained <= min(hr.in_jurisdiction_headcount[t-1], hr.in_jurisdiction_headcount[t])`. Same FTE-equivalent, employees-only, primary-work-location test as `hr.in_jurisdiction_headcount`; contractors excluded. REPORTED BY THE COMPANY (`production: "primary"`) — retention has no stock analogue and cannot be computed from headcount snapshots at all: it is a statement about continuity of specific roles across a period, not a count at an instant. — HR KPI, I'mBoard-authored (editorial tier).