Despite growing recognition that narratives shape economic outcomes, we lack a formal, empirically tractable definition of what constitutes a “story” or a “narrative.” This project develops a conceptual and computational framework to address that gap. I define a story as a temporally ordered sequence of economically meaningful events and a narrative as a set of stories sharing the same meaning. An axiomatic framework formalizes story similarity and disciplines narrative clustering. Empirically, I implement a three-stage pipeline: large language models extract structured event sequences from unstructured text; stories are compared using a multi-stage similarity measure that combines semantic and structural features; and network-based clustering aggregates similar stories into endogenously emerging narratives. I construct narrative-prominence indices which display systematic co-movement with financial indicators of interest (e.g., prices, returns, volatility, etc.), consistent with markets dynamically reweighting competing narratives over time. Interpreted through a revealed-preference lens, these patterns suggest markets act “as if” particular narratives are believed or attended to at different points in time. Ongoing work develops models of dynamic narrative credibility and evaluates whether narrative-based measures provide incremental forecasting power relative to existing text-based approaches. By providing both a formal definition and an empirically implementable identification strategy, the framework offers a new approach to studying belief formation and aggregate, market-level belief elicitation in financial markets.