Background growth, count noise and the first accepted year of a patent take-off rule

Abstract Patent take-off rules assign retrospective years by combining growth, level, persistence and scale conditions. One fixed five-condition rule is examined here. For a permanent log-growth step δ on a monotone background of rate g, its level condition at offsets d = −1, 0, 1 reduces at large counts to 4g + (d+2)δ ≥ log 1.5. Background growth can therefore advance acceptance relative to the imposed change although the growth contrast removes constant growth. On European Patent Office reporting-year counts for 35 fields and 175 geographies in 2011–2025 it accepts 201 events, where the level ratio has the smallest relative slack in 132 and the count floor in 52. Removing the ratio advances 56 first dates and adds 191 series, removing the floor advances 23 and adds 592, so both affect timing and membership in different proportions. Wider growth windows change the earliest-labelled geography in 24 of 33 fields with events under both definitions; applicant and inventor attribution produce different calendars across five priority-year populations and a second field taxonomy. Fitted scenarios generate sizeable calendars: with parameters from 2011–2015 alone, 1,002 series carrying 183 observed events yield a scenario mean of 256.5, so calendar size is not evidence of discrete change, though no individual event is classified. Count perturbations reproduce an event's year in 45 per cent of replicates on the main panel and 41 on a second. Source version, window completion, pre-period growth and criterion margins make such labels interpretable; accuracy against independently dated onsets remains untested.