In Narrative and Numbers (2017) Damodaran lays valuation out as an assembly line: story → inputs → numbers → feedback that revises the story. Its value is not precision but that it forces you to state the magnitude of your story. "It will change how people move" cannot be argued with; "in ten years it takes 10% of the urban mobility market at a 40% operating margin" is a claim you can contest line by line.
The reverse constraint matters as much: the numbers must be explicable by the story. A model with "8% perpetual growth, margins rising every year, capex flat" is three mutually contradictory stories stuffed into one spreadsheet. Damodaran's test is three questions — is it possible, is it plausible, is it probable — and most expensive stories are stuck between the last two.
In June 2014 Damodaran argued against Uber's $17bn financing price with a valuation near $5.9bn, built on "a global urban car-service market of roughly $100bn, of which Uber takes 10%." In July, investor Bill Gurley replied in "How to Miss By a Mile": the taxi market is not a constant, more supply and lower prices create their own demand, and the boundary of the market is itself part of the story. Each was half right: on May 10, 2019 Uber listed at $45 a share (about $82bn) and closed the day at $41.57, down roughly 7.6% — the growth packed into a price is delivered more slowly than the story promises.
The failure mode is specific: when the story itself moves the boundary of the market, anchoring on today's market size is wrong — Gurley's rebuttal stands. The other misuse is treating numbers as a source of precision: a DCF exists to expose implicit assumptions (Day 6), not to produce an exact target price.
Shiller borrows the SIR model from epidemiology: a narrative has an infection rate (how easily it is retold) and a recovery rate (how fast it is forgotten), and their ratio decides whether it becomes an epidemic. Three consequences. Transmissibility is orthogonal to truth — the version that travels has characters, a turn, a moral. A narrative is self-fulfilling for a while: believers buy, prices rise, and the rise becomes fresh evidence for the story (reflexivity, Day 33, operating at the level of information). Narratives mutate and recur — "this new technology makes the old valuation methods obsolete" reappears across railroads, radio, the internet, and AI: different wording, identical structure.
The practical implication: treat prevalence as an observable independent of fundamentals — a rise in it is not a reason to buy, and a high level means the pool of incremental buyers is shrinking.
The late-1990s "new economy" narrative is the textbook sample. The Nasdaq Composite closed at a record 5048.62 on March 10, 2000, and Shiller's CAPE hit roughly 44 in December 1999, the highest in the series on record; the index then fell to 1114.11 on October 9, 2002, a decline of about 78%. Cisco became the most valuable company in the world in March 2000 (about $555bn) and by October 2002 was down roughly 86% from its high. And the story — "the internet will reshape commerce" — was true. What was wrong was never the story. It was the price.
The most dangerous misuse is as a timing tool. When Shiller testified about "irrational exuberance" to the Federal Reserve in December 1996, the S&P 500 stood near 744; the market rose for more than three further years before it peaked. Identifying a bubble narrative and knowing when it breaks are different problems — the first is doable, the second is not. A second limit: not every popular narrative is a bubble, and equating "everyone is talking about it" with "sell" will make you miss genuine trends systematically. The boundary is to land the conclusion on position size and price, not on "in or out."
Kahneman supplies the mechanism — WYSIATI, "what you see is all there is": the mind assembles a coherent story from whatever material is at hand, then allocates confidence according to the story's smoothness rather than the strength of the evidence. Hence the most hidden error in investing: the less information there is, the cleaner the story, and the more certain people become.
Three harms follow. Hindsight bias: after a crisis everything looks like it should have been foreseen, so you overrate your foresight next time. Survivor narrative: a methodology is reverse-engineered from successes while the identically-run failures never enter your sample (Day 27). Explanatory trading: you read "fell on news X," treat it as new information, and adjust — trading on causality a reporter attached before deadline.
On October 19, 1987 the S&P 500 fell 20.47% in a single session and the Dow 22.61%, with no news event remotely proportionate to the move. Post-hoc accounts offered program trading, portfolio insurance, stretched valuations; none of them issued a signal beforehand. The systematic evidence comes from Cutler, Poterba and Summers, "What Moves Stock Prices?" (1989): macroeconomic news explains only a small share of the variance of market returns, and after going through the largest fifty single-day post-war moves one by one they wrote that "many of the largest market movements in recent years have occurred on days when there were no major news events."
This does not license the nihilism of "no explanation is valid." Real causality exists: the effect of a rate change on long-duration assets can be derived in advance and verified afterwards (Day 39). The test is one question: would this explanation have produced the same prediction before the event? If yes, it is a model; if it only holds afterwards, it is a narrative. The danger is not telling stories; it is treating coherence as evidence.
In that same letter Buffett supplies the tool via Aesop: "a bird in the hand is worth two in the bush" is only useful once three numbers are added — how many birds are actually in the bush, how soon you get them, and what the risk-free rate is. That is the minimum grammar for translating a narrative into cash flow: quantity, time, discount rate.
The second sheet is the one more often skipped. What makes a narrative dangerous is not that it is wrong but that it has no boundary — a fall means "the market doesn't get it," a rise means "the story is playing out," and both directions are self-consistent. The remedy is to write observable falsifiers in advance: not "sell if it drops 30%" (that is a price, not a fact) but "four consecutive quarters of zero user growth." Pair it with a pre-mortem (Day 27): if this position is down 70% in three years, what is the most likely cause? Turn that answer into something you monitor.
Amazon is the cleanest sample of "story right, price wrong": the share price (unadjusted for later splits) fell from about $106.69 in December 1999 to about $5.51 in September 2001, roughly 94% — while the narrative that online retail would change commerce was never falsified in those three years and has since been thoroughly confirmed. Same story; buying it in 1999 and buying it in 2001 differ by an order of magnitude. The difference lay not in the judgment but in the price.
The discipline carries a real cost: demanding that everything be written as falsifiable numbers will systematically make you miss great businesses that genuinely could not be computed at the time. Buffett is the counter-example himself — at the 2017 annual meeting he said flatly that he had been "too dumb to realize what was going to happen," having missed Amazon for years. The lesson is not to abandon the discipline but to accept that the opportunity cost outside your circle of competence is the fair price of it (Day 1). The other failure mode: falsifiers set too finely trigger on normal volatility and turn long-term ownership into frequent trading.