Informational inertia in a decentralized prediction market: evidence from the May 2024 CPI leak

Official URL: http://dx.doi.org/10.1016/j.frl.2026.110378

On May 15, 2024, the U.S. Bureau of Labor Statistics inadvertently uploaded the CPI report to its public server at 08:00 ET, thirty minutes before the scheduled release. We use this verified timing shock as a single-episode descriptive case study of how quickly a decentralized retail prediction market incorporates a high-signal macroeconomic surprise. Polymarket’s Fed rate-cut contracts exhibit a near-zero top-of-book response throughout the 35-minute interval between the inadvertent upload and the first detectable price adjustment at 08:35 ET, while CME 30-Day Federal Funds futures begin adjusting within seconds of the upload and complete the bulk of their move before the official release. The 35-minute interval has two parts. The pre-release thirty minutes are consistent with rational non-adjustment inside a 1–2 percentage-point no-arbitrage band given the modest realized shock. The remaining five minutes, by contrast, fall after the official release: by 08:30 ET the information is universally public and CME has completed the bulk of its move, so the bid–ask logic that rationalizes the pre-release window cannot account for the continued absence of repricing. Change-point detection, randomization inference, synthetic control, and quote-update hazard analysis, reported as bounding exercises rather than sharp hypothesis tests, are jointly consistent with limited updating rather than noisy immediate repricing. The episode highlights a sizable monitoring-infrastructure difference between a thin decentralized retail venue and the corresponding institutional benchmark, and provides a pre-growth baseline for future measurement of real-time efficiency on Polymarket.