Social Volume Spike Detection

WHY ATTENTION IS WORTH MEASURING AT ALL. Investors cannot evaluate every asset, so attention is a scarce resource that determines which assets get considered. Barber and Odean (2008) found that individual investors are net buyers of ATTENTION-GRABBING stocks - those in the news, with high abnormal volume or extreme prior returns - because attention solves their search problem on the buy side but not the sell side, where they can only sell what they already own. That asymmetry is a genuine mechanism, and it predicts buying pressure specifically from retail participants following attention.

Da, Engelberg and Gao (2011) made attention directly measurable using search volume, showing it is distinct from news coverage and captures retail attention in particular.

WHAT THIS MEASURES. Mentions of each tracked asset across recent NEWS coverage, against the average across assets, tiered as trending at roughly twice average, active, or quiet.

TWO THINGS TO BE PRECISE ABOUT.

• THIS IS MEDIA COVERAGE, NOT SOCIAL ATTENTION. Journalists publishing about an asset is not the same as people searching for or discussing it. The Barber-Odean mechanism runs through what retail investors NOTICE, and media volume is an upstream proxy for that.
• A RATIO TO A CROSS-ASSET AVERAGE IS A FRAGILE THRESHOLD. Assets have wildly different baseline coverage: Bitcoin is written about constantly, a small token almost never. Comparing each asset to the average across ALL assets means a well-covered asset can sit above the threshold permanently while a rarely covered one crosses it on two articles. The stronger comparison benchmarks each asset against ITS OWN history - abnormal coverage standardised by that asset's own variability - though that is not automatically sufficient either: sparse, trending, zero-inflated counts need a minimum-history rule and a model suited to count data. It is the same discipline the narrative-shift analysis needs, for the same reason.

A SPIKE IS NOT DIRECTIONAL. Coverage rises around good news and bad news alike, and around volatility of either sign. Attention says something is happening; it does not say what.

BE CAREFUL WITH 'BEFORE A MOVE'. Coverage frequently FOLLOWS a price move rather than preceding it - journalists write about what just happened - so a mention spike is often a lagging record. Establishing a lead requires a time-ordered test against subsequent returns, which this does not perform.