Modelling Information Incorporation in Markets, with Application to Detecting and Explaining Events

TL;DR

This paper models how information flows into markets, particularly political betting markets, and introduces algorithms to detect and explain significant events by analyzing news vocabulary changes and market data.

Contribution

It presents a new model of information flow in markets and algorithms for event detection and explanation based on news vocabulary analysis.

Findings

Market prices tend to approach correct outcomes over time under certain conditions.

IEM data closely matches the theoretical model of information flow.

The event detection algorithm successfully identifies significant market events and explains their origins.

Abstract

We develop a model of how information flows into a market, and derive algorithms for automatically detecting and explaining relevant events. We analyze data from twenty-two "political stock markets" (i.e., betting markets on political outcomes) on the Iowa Electronic Market (IEM). We prove that, under certain efficiency assumptions, prices in such betting markets will on average approach the correct outcomes over time, and show that IEM data conforms closely to the theory. We present a simple model of a betting market where information is revealed over time, and show a qualitative correspondence between the model and real market data. We also present an algorithm for automatically detecting significant events and generating semantic explanations of their origin. The algorithm operates by discovering significant changes in vocabulary on online news sources (using expected entropy loss) that align with major price spikes in related betting markets.