Endogenous Detection: Rate-Free Equilibrium Capacity and Sharp Information-Incentive Frontiers
Reliable statistical detection can depend on evidence that independent agents must privately pay to produce. This paper studies preannounced level-$\alpha$ tests with private capability and cost types that are independent across agents, hidden one-time production, and an action-invariant null. Coordinate experiments may be heterogeneous, and informative laws need not be dominated by their null laws. The cost-quantile Hellinger capacity $V_n^{\mathrm{het}}$ measures the greatest aggregate squared Hellinger information attainable under a unit budget of expected production costs normalized by agents' prior-weighted benefits. The main theorem establishes a rate-free equilibrium capacity separation. Every level-$\alpha$ test admits an independent behavioral Bayesian Nash equilibrium with power at most $\alpha+\sqrt{V_n^{\mathrm{het}}}$. If $V_n^{\mathrm{het}}\to\infty$, however slowly, a sequence of preannounced level-$\alpha$ tests achieves power tending to one at every independent equilibrium. The constructive proof combines bounded discrimination scores, Gaussian smoothing, uniform pivotality bounds, and a cost-quantile argument that forces production by low-cost types. A uniform Hellinger information spectrum yields sharp strict information-incentive frontiers for power-law private costs, with explicit results for logarithmic Poisson signals, exact singular certificates, and mixed regular and singular information. Bounded capacity can nevertheless coexist with power one at every equilibrium. The constructed tests use publicly specified experiment and cost laws.
Authors
- K. C. Lalropuia (ORCID: https://orcid.org/0009-0008-4861-602X)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-21
- DOI
- https://doi.org/10.5281/zenodo.24417116
- Primary Topic
- Game Theory and Applications
- Type
- preprint