News as a scenario signal — an academic paper
Research co-authored with an academic supervisor: a measure of a news event's market impact, tested against a sample of 73 events.
- A news event becomes a measurable quantity rather than an occasion for interpretation
- Conclusions rest on an event sample instead of hand-picked examples
- Practical value stated separately from theoretical novelty
The problem
News impact on markets is usually discussed after the fact: the event happened, the move got explained. That conversation cannot be checked. What's needed is a measure computable before the explanation and comparable across different events.
How it works
The work is built around a single impact measure, tested on a sample of 73 events across three years in the technology sector against the broad market. Novelty, the justification of the measure itself, and the interpretation of results are separated into their own sections, so the contribution is visible without reading between the lines.
The hard part
The longest part was not the arithmetic but the language. Text written with language models is recognised instantly: 'firstly, secondly' scaffolding, triads, the obligatory 'let us emphasise', passive voice in the abstract. A dedicated editorial pass removed those traces without touching the substantive sections the supervisor had already read and approved.
Evidence
- Every supervisor revision incorporated and mapped to its section
- A sample of 73 events across 2022–2024, sector compared against the broad market
- The text went through a separate stylistic pass removing traces of machine generation