The Execution-Certainty Wedge: Why Contention-Linked Instruments Redistribute at Near-Zero Cost, and Per-Entity Caps Do Not

In several important markets the party doing the allocating does not rank offers by value. It ranks them by effective value — what is offered, multiplied by the probability the transaction actually completes. Where one class of participant holds a systematic advantage in that second factor, the winner is not reliably the highest-value participant, and the disadvantage falls on those who cannot supply certainty: the buyer who needs a mortgage, the applicant who must compare financial-aid offers before committing, the patient whose insurer loses one claim in six. This paper names that gap the execution-certainty wedge, documents it in three markets, and establishes what follows for policy. The measured case is residential real estate. Reher and Valkanov (2024) estimate a mortgage-cash premium of 11.7% across 426,256 repeat-sales transactions, 8.6% in offer-level brokerage data and 10.5% in a homeowner survey, against a rationally justified compensation for transaction failure of 3.3% for a representative seller, or 6.9% adjusting for heterogeneity in local denial rates. The premium exceeds its actuarial justification by roughly four to eight percentage points, and that excess, not the whole premium, is the distortion available for correction. An agent-based model of a shared-inventory market compares four allocation regimes against a policy-invariant benchmark, and the central result is a ranking of instruments. At the calibrated wedge, a smooth duty triggered by a local contention signal cuts owner-occupier starvation from 41.9% to 36.6% at an aggregate welfare cost of zero; a work-conserving fair share cuts it to 24.7% for 1.8% of aggregate welfare; a static per-entity acquisition cap — the instrument most often proposed — cuts it to 8.7% but costs 11.3% of aggregate welfare, 17.3% once supply responds, and decays to a statistical no-op once the same capital is held across twenty-five shell entities. Constraints that bind on observed contention survive every robustness test applied; constraints that bind on entity identity survive none. Two results are methodological and transfer beyond this setting. Measuring allocative efficiency in a market where the policy determines who is waiting requires a benchmark fixed independently of the policy; a contemporaneous benchmark computed over the bidders each regime leaves in the queue moves by up to 14% across regimes and reverses the sign of the conclusion. And a latency statistic computed over completed transactions is uninterpretable without the accompanying rate at which participants leave unserved — a rule that applies identically to housing search times and to clinical waiting lists. Four boundary conditions state when the underlying scheduling correspondence transfers to institutions — assignable identity, flow rather than stock, an insulated scheduler, and unified rather than partitioned enforcement — and each is matched to a regime in force whose operating record displays the predicted behaviour. Because identity is assignable in admissions and in health but manufacturable in property markets, the conditions generate a testable prediction: the per-entity cap that fails in housing should succeed in the other two. Replication code, simulation output, figures, and a pre-analysis plan for the paper's proposed empirical test are included in this deposit.