Economist here who studies exactly this type of counterfactual analysis. You are completely right: the effect of Gemini can only be estimated if we factor in wh

Economist here who studies exactly this type of counterfactual analysis. You are completely right: the effect of Gemini can only be estimated if we factor in what the Alphabet stock price would have been in the same time but in a world without Gemini. This is actually very standard in financial economics. This type of effect can be calculated with econometric techniques that compare before/after for “treated” vs “untreated” units, but in instances such as these, where only one or a few units were affected, like Alphabet stock amongst hundreds of other companies, one could use techniques such as “synthetic controls”. The intuition is to use other companies’ data to estimate before Gemini how Alphabet stock prices move over time, and then use that relationship to estimate a post-Gemini version of no-Gemini Google. The difference between the actual stock price and that counterfactual is the effect of interest; whether it is a significant effect or just random noise can be established by a number of auxiliary statistical tests. For more info, see [0]. [0] Abadie, A. (2021). Using synthetic controls: Feasibility, data requirements, and methodological aspects. Journal of Economic Literature, 59(2), 391-425.