Measured reactions, by event type
Six analyst event classes, measured across 4,520 distinct names — 14,380 name-by-class cross-sections in all, each read against the price at its own news timestamp. Every measured ticker →
The event class does not tell you which way the name moved.
50% of names have a positive median 30-minute reaction to analyst upgrades. For analyst downgrades — the opposite call — it is 50%. Not a rounding artefact of this sentence: the two are 50.1% and 49.8%. Across all six classes the figure sits between 45% and 50%. Knowing an event was an upgrade tells you close to nothing about the reaction; the reaction has to be measured per event, per name, which is what the pages below do.
The classes are not ranked by that share, and the spread between them is not published as a finding: these are medians of small per-name samples, and the six values sit close enough together that an ordering would imply a precision the floor does not carry. No class median reaction is published at all — the median across names is zero in every class. See any class page for that in full.
The six measured event classes.
Rating reiterations
An analyst restates an existing rating. The highest-volume, lowest-information analyst event — the baseline. 3,515 names measured; 45% have a positive median 30-minute reaction.
Analyst upgrades
An analyst raises their rating on a stock. The market's response is the label — not the upgrade itself. 2,679 names measured; 50% have a positive median 30-minute reaction.
Analyst downgrades
An analyst cuts their rating. Whether the stock actually drops — and for how long — is an empirical question. 2,583 names measured; 50% have a positive median 30-minute reaction.
Coverage initiations
An analyst starts covering a name. New coverage tends to be constructive — but the reaction varies by name. 2,268 names measured; 48% have a positive median 30-minute reaction.
Price-target raises
An analyst lifts their price target while keeping the rating. Often the news is already priced in. 1,797 names measured; 50% have a positive median 30-minute reaction.
Price-target cuts
An analyst lowers their price target. The reaction is rarely a clean one-directional drop. 1,538 names measured; 45% have a positive median 30-minute reaction.
Last 3 years, pulled 2026-08-31. Reaction = percent move vs. the price at the news timestamp. Every number here is a cross-section over NAMES, one vote each: a name with 2,000 measured events counts the same as one with 40, and none of these are pooled event medians. A name appears in as many classes as it was measured in, which is why the six counts sum to 14,380 against 4,520 distinct names.
The same cross-section by sector → · Every measured ticker →