On 30 August, we published a measure showing that what determines the market reaction to AI company results is not the size of the beat versus estimates, but the basket's share of its daily stock movement. We concluded that research with a stated test: seven names reporting the following week, with their shares published in advance, and explicit falsification criteria. Six reported between 1 and 2 September — the last was Broadcom, which beat estimates on both revenue and earnings and then fell in the next session, as it did in June. This article is the test result: the line got four right and two wrong, and the errors reveal what the original measure could not see.
A free account gives you 4 full research papers every month. Or subscribe to EL7 Pro for full access to more than 300 papers.
$700 per year · or $99 per month