Narrative contagion is the contemporary formalization of what Keynes called herd behavior in speculative markets. Using natural-language processing to analyze corporate earnings calls, IMF researchers documented a specific pattern: companies tend to adopt the narratives of their peers. When one firm starts talking up the transformative power of AI, others follow suit. This contagion starts within peer-firm groups and spreads to the aggregate level, producing market-wide optimism whose intensity bears no stable relationship to evidence at the firm level. The mechanism is exactly what Keynes identified in Chapter 12 of the General Theory — animal spirits operating at scale, amplified by technology Keynes could not have imagined.
The IMF's research was published in 2024 as part of the Fund's analysis of AI's macroeconomic implications. The methodology — applying NLP to the text of earnings calls — allowed the researchers to quantify what Keynesian theorists had long argued was qualitatively true: that narratives travel, that they shape decisions, and that the aggregate effect of narrative travel is a level of conviction that exceeds any individual firm's evidentiary basis.
The finding has direct implications for the Keynesian beauty contest. Each firm's decision to adopt AI narratives is influenced not primarily by its own assessment of AI's value but by the narratives its competitors are deploying. The firm that fails to match peer narratives appears to be falling behind — a perception that matters to investors regardless of its empirical basis.
Narrative contagion produces asymmetric risk. On the upside, it amplifies adoption and investment beyond what individual firms would rationally undertake. On the downside, when sentiment reverses — when the narrative shifts from enthusiasm to caution — the same contagion mechanism produces synchronized contraction. The Software Death Cross of early 2026 may be an early instance: the trillion-dollar repricing of software companies reflected the narrative shift from 'code is the moat' to 'code is a commodity,' propagating across the industry in weeks.
The Keynesian prescription for narrative contagion is institutional: information infrastructure that distinguishes firm-level evidence from industry-level narrative, regulatory disclosure that surfaces the distinction, and cultural norms that reward skepticism alongside enthusiasm.
The phenomenon is classically Keynesian (General Theory, Chapter 12). Its empirical formalization is Philip Gabriel's and colleagues' 2024 IMF working paper applying NLP to earnings call transcripts.
Empirical formalization of Keynesian herding. NLP methods quantify what Keynes argued qualitatively.
Firm-to-firm, then aggregate. Contagion spreads first within peer groups, then to industry-wide narratives.
Detachment from fundamentals. The aggregate narrative diverges from the aggregate evidence.
Asymmetric risk. Contagion amplifies both enthusiasm and reversal.
Institutional response. Information infrastructure must distinguish evidence from narrative.
Whether narrative contagion is a pathological feature of financial markets or a necessary mechanism for coordinating investment under radical uncertainty.