Patent extension: report-relevant sensitivities
Participation and grant-event totals
| Grant-event window comparison | 2015–2019 | 2020–2024 | Change |
|---|---|---|---|
| All Arkansas-inventor grants | 2,045 | 2,201 | +7.6% |
| Excluding grants naming one largest source assignee ID | 1,634 | 1,543 | -5.6% |
The aggregate rises while the series excluding one source assignee ID falls. The purpose is to test sensitivity to concentrated participation, not to explain why either series changes or to measure invention creation. Source IDs are not corporate families; joint records are removed once.
Technical influence and counting choices
| Mature grant cohort | Window | Arkansas documents | Defined score coverage | Whole-counted relative mean | Arkansas-inventor-weighted mean |
|---|---|---|---|---|---|
| 2015–2022 | 3 years | 3,423 | 99.7% | 1.06 | 0.98 |
| 2015–2020 | 5 years | 2,631 | 99.7% | 1.05 | 0.95 |
A value of 1 is the reference mean for the same grant year and first inventional at-issue CPC subclass. The reference is all US utility grants, including foreign inventors; it differs from the domestic-inventor benchmark used for regional activity. The high-citation tail matters: the top approximately 1% of documents account for about a fifth of the sum of relative scores. These descriptive averages support a cautious near-reference interpretation across counting choices, not a statistically established advantage.
Unknown fields and zero expected means remain undefined. Patent citations are references in later granted documents, not observed sales, licensing, adoption or causal programme returns. Three- and five-year rows use different eligible cohorts and cannot establish a time trend.