Narrative Cycle Assessment

WHY NARRATIVES ARE WORTH TAKING SERIOUSLY. Capital in crypto rotates between thematic groups in a way that is visible and persistent enough to be a genuine feature of the market rather than commentary. Shiller's work on narrative economics makes the general case: stories spread through populations like contagions and can drive economic behaviour, so a narrative is not merely a description of a move but sometimes part of its mechanism.

WHAT THIS COMPUTES. Category tags per coin, grouped into themes, with average 30-day price change per theme.

AND HERE IS THE CIRCULARITY TO NOTICE. A narrative's 'strength' is measured by the PRICE PERFORMANCE of its members. So 'the AI narrative is strong' decodes to 'coins tagged AI have gone up'. That is a true and useful statement about what has happened; it is not independent evidence that a narrative is gaining force, because the measure and the thing measured are the same series. Anything derived this way describes the past by construction, and calling it strength invites reading it as a cause.

WHAT WOULD BE INDEPENDENT EVIDENCE: mentions and search volume for the theme, capital flows into its assets, developer activity, or new launches within it - measures that move separately from price and could therefore lead or diverge from it.

TWO MEASUREMENT PROBLEMS ON TOP.

• CATEGORY TAGS ARE NOT A TAXONOMY. Crypto has no GICS equivalent, tags are assigned by data vendors with differing rules, and one asset commonly carries several. Group membership decides the answer, so the result is partly an artefact of tagging.
• A SMALL GROUP IS ONE ASSET IN A COSTUME. Theme averages over a handful of coins, one of which dominates by size or volatility, are that coin's return with a label attached. Equity sector averages cover dozens of comparable firms; crypto themes often do not.

AND THE COMMON FACTOR AGAIN. Crypto assets co-move strongly, so most of any theme's 30-day return is the market's return. Genuine rotation is what remains after that is removed, and comparing raw averages does not remove it - the same objection as in the sector-rotation analyses.