In 2012–2026 the announcement-day beta premium in 35 US portfolios was +5.3 bp a day (95% CI −12 to +24); this neither confirms nor rejects the 1978–2011 value of +11.6 bp

Question: Savor and Wilson (2014) found that on scheduled PPI, employment and FOMC announcement days (a-days), high-beta stocks earned much more than low-beta stocks, but not on other days (n-days). Their sample ended in 2011. Did the pattern continue, and could a paper investor use it? Approach: we froze a protocol before we downloaded return data. We used daily Kenneth French portfolios (CRSP build 202608), with 25 size/book-to-market and 10 industry portfolios as primary test assets. We took dates from BLS archives, ALFRED vintage dates and Federal Reserve FOMC pages. Betas used only the previous 12 months of daily returns. Daily Fama-MacBeth regressions gave the security market line (SML) slope. We tested five holdout hypotheses (2012-01 to 2026-08, 3,686 days, 454 a-days) with Holm correction, Newey-West errors and a stationary block bootstrap. Results: the method reproduced the original result in 1978-2011 (a-day minus n-day slope +11.6 bp per day, 95% CI +2.6 to +20.9). In the holdout, the difference was +5.3 bp (CI -12.4 to +23.6), and the a-day slope was +6.8 bp (t 0.74). No hypothesis passed (all Holm p = 1). The market excess return was not higher on a-days (-0.6 bp, CI -12.1 to +10.6; in-sample +8.9 bp). Holding the market only on a-days earned +0.35% a year at 2 bp costs, against a 13.9% excess return for buy-and-hold. The SML slope portfolio held only on a-days lost 9.3% a year at 5 bp costs. Post-hoc: FOMC days had the largest holdout slope (+23.8 bp, t 1.29). The minimum detectable difference was 23 bp, two times the in-sample effect. At holdout noise, about 59 years of new data are necessary to detect the in-sample effect. Thus the holdout neither confirms nor rejects the original size. Limitations: low power; French portfolios instead of stock beta deciles; current (not point-in-time) data versions; costs are scenarios. Code, protocol, dates, daily slopes, strategy returns and figures are included. Raw CRSP-derived returns are not included, but they are retained as a hashed data item.

announcement days · security market line · beta · FOMC · macroeconomic announcements · post-publication decay · holdout test · Fama-MacBeth · pre-registration · calendar strategy · statistical power · trading costs