Methodology of Narrative Economics: From Anecdote to Theorem
Traditional economics has long prided itself on mathematical rigor and empirical validation, yet it has systematically overlooked one of the most potent drivers of human behavior: stories. Narratives—the vivid, memorable, and socially transmitted accounts of events—shape expectations, alter preferences, and mobilize collective action in ways that conventional models often fail to capture. Robert Shiller (2017) pioneered the recognition of narratives as central to economic phenomena, arguing that the “virality” of stories can drive business cycles, asset bubbles, and consumer trends. However, integrating narratives into economic analysis requires a methodological framework that transforms anecdotal tales into testable evidence. This chapter establishes such a framework, delineating a protocol that elevates narrative from illustrative ornamentation to rigorous data. We distinguish between narrative as mere illustration—used to liven up presentations but not to inform models—and narrative as evidence, where story characteristics become variables in formal analysis. Drawing on Bayesian inference, natural language processing (NLP), and structural equation modeling (SEM), we outline a five-step procedure: (1) select a verifiable everyday story, (2) extract economic variables, (3) formulate testable hypotheses, (4) apply formal models, and (5) validate with real-world data. A technical appendix details coding schemes for narrative quantification, and a case study on the perennial “Christmas toy shortage” demonstrates the protocol in action, linking inventory optimization and elasticity calculations to narrative dynamics. The chapter culminates in the formal derivation of the Narrative Transmission Function $N_t = f(S_t, \theta)$, a mathematical representation of how story traits ($S_t$) and audience susceptibility ($\theta$) generate economic impact ($N_t$). By the end, the reader will possess a concrete toolkit for converting the messiness of human storytelling into the precision of economic theorems.
The Epistemological Shift: From Anecdote to Evidence
At its core, the narrative economics enterprise confronts a fundamental epistemological question: can subjective, qualitative stories serve as legitimate data in a discipline predicated on objectivity and measurement? Traditional economics treats narratives as illustrative anecdotes—colorful asides that humanize models but do not alter their predictions. For instance, a textbook on supply and demand might open with a story about coffee prices after a frost, yet the story merely illustrates a shift in supply; it does not become part of the empirical test. In contrast, narrative economics posits that stories are not just reflections of economic reality but active constituents of it. When a narrative spreads, it changes beliefs, and changed beliefs change behavior, thereby altering the very economic variables we seek to explain. This is not merely philosophy; it demands a operational definition of narrative as evidence.
We define narrative as evidence as a story whose content, structure, and transmission dynamics are systematically measured and included as explanatory variables in economic models. This stands in contrast to narrative as illustration, where a story is invoked post hoc to exemplify a pre-existing theoretical relationship without being formally quantified. The distinction hinges on verifiability and testability. An illustrative anecdote might say, “The 1970s oil crisis showed how supply shocks cause inflation,” but it does not specify which stories about oil were told, how widely they circulated, or how they influenced wage and price-setting behavior. Narrative as evidence, however, would code the prevalence of “oil shortage” stories in media archives, correlate them with inflation expectations surveys, and include that coded variable in a Phillips curve regression.
This shift is epistemologically analogous to the incorporation of expectations into macroeconomics. Once, economists modeled outcomes based solely on current fundamentals; today, rational expectations are central because beliefs about the future affect present actions. Narratives are a vehicle for forming those beliefs. They simplify complex information, evoke emotions, and are more memorable than raw statistics. Shiller (2017) documents how narratives like “the American Dream is dead” or “housing prices always rise” propagated during the 2008 crisis, influencing spending and investment decisions. But documenting is not enough; we need a methodology to move from qualitative description to quantitative inference.
Critics may argue that narratives are too fuzzy, too context-dependent, and too susceptible to selection bias to be data. Our response is twofold. First, all economic data are constructs: GDP is a formula; inflation is a basket of goods. Narratives, once coded, become similarly structured. Second, we mitigate fuzziness through rigorous protocols that ensure replicability. The five-step protocol below is designed to impose the same discipline on narrative data that we expect from conventional datasets. It draws inspiration from fields like psychology (where narrative identity is measured) and computational social science (where text as data is standard), but adapts them to economic questions.
Step 1: Selecting a Verifiable Everyday Story
The first step is to identify a story that is both economically relevant and empirically verifiable. “Everyday” does not mean trivial; it means the story must be grounded in ordinary economic life—a product shortage, a local business closure, a new technology adoption—so that its effects can be observed in market data. “Verifiable” means we can confirm the story’s existence, content, and timeline through independent sources. This step prevents the researcher from cherry picking exotic tales that cannot be systematically studied.
Selection criteria include:
- **Economic salience**: The story must involve economic decisions—buying, selling, investing, producing. A story about a celebrity’s wedding may be viral but irrelevant unless it influences consumer spending patterns.
- **Temporal boundedness**: The narrative should have a discernible start and peak, allowing for before-and-after comparison. For example, the “Christmas toy shortage” recurs annually but spikes around specific product crazes (e.g., Cabbage Patch Kids in 1983, Hatchimals in 2016).
- **Multi-source traceability**: The story must appear in at least two independent channels (e.g., news archives, social media, sales reports) to mitigate single-source bias. A rumor confined to one forum is insufficient.
- **Variation**: For hypothesis testing, we need stories that vary in intensity, sentiment, or demographic reach. Studying multiple instances of toy shortages across years provides this variation.
Selection procedure: 1. Define the narrative domain: e.g., “consumer product shortages during holiday seasons.” 2. Construct a candidate list: Use keyword searches in news databases (e.g., “toy shortage,” “sold out,” “must-have gift”) over a 20-year period. 3. Filter for verifiability: Cross-reference with sales data, inventory reports, or court cases (e.g., price gouging lawsuits). Discard stories lacking corroboration. 4. Assess everyday relevance: Ensure the product is widely consumed (not a niche collectible) and the shortage affects ordinary shoppers. 5. Finalize sample: Select 3–5 cases that maximize variation in narrative characteristics (e.g., media coverage volume, emotional tone).
This step is not about proving the story’s “truth” in an objective sense—stories are social constructs—but about establishing that the story was told and believed by enough people to potentially influence economic behavior. The researcher’s role is akin to a field biologist selecting a observable ecosystem: the story must be accessible and measurable.