CoCoCo: Evaluating the ability of LLMs to quantify consequences