Why We Ignore the News

Every dataset a research firm uses has a producer, and the producer's incentives are part of the data. Price data is produced by people committing money. Corporate filings are produced under legal liability. Even weather data is produced by instruments that do not care whether anyone reads them. News is the exception. News is produced by an industry that is paid for attention, and it is optimised, headline by headline, for the probability that a person will stop scrolling. That does not make it false. It makes it selected: selected for novelty, for narrative, for emotional charge, and against everything that is typical, gradual, or unresolved, which is to say against most of what actually moves markets over the horizons we trade.

This is why our refusal of the news is not primarily a claim about noise. Many datasets are noisy, and noise can be averaged, filtered, and priced. The problem with news is provenance. A headline arrives pre-interpreted: someone has already decided what the event means, which horizon it matters on, and whether it is good or bad, and those decisions were made by people whose incentives have nothing in common with ours. To trade on the news is to import those decisions into our process without examining them, which is something we would never tolerate from any other input. We would not let a data vendor quietly rewrite our price history. We do not let an editor quietly rewrite our reasoning either.

There is a second, quieter problem. News is a lagging indicator wearing a real-time costume. By the time an event is a story, it has been through reporting, editing, and publication, and the market has been trading on the underlying facts, and on rumours of the facts, for as long as the facts have existed. What arrives as information is usually confirmation, and confirmation feels valuable precisely when it is worthless. It makes a held position feel safer and a missed one feel stupider, and both feelings are trading signals of the worst kind: strong, timely, and uncorrelated with returns.

The strongest counterargument is real, and we want to state it fairly. Some events genuinely change the world a portfolio is priced on: a central bank reversing course, a war, a fraud exposed, a technology that obsoletes an industry. A process that ignores the news will hold positions through some of these events and be hurt by them, and its defenders cannot pretend otherwise. Refusing information has a cost. The question is whether the cost is larger than the alternative, and whether it can be bounded.

Our answer rests on one observation: price is itself a carrier of news, and it is the only carrier whose incentives we trust. When an event matters, it matters because people with money at stake revise what they will pay, and that revision appears in prices, in volumes, and in the relationships between markets, without an intermediary deciding what it means. A rules-based process that reads prices is not uninformed about events. It is informed by events in the only form in which events have ever made anyone money: other people's willingness to transact. What it refuses is the narrative layer, the part that tells us a story about the revision and invites us to agree with it emotionally. The cost of this refusal is real but bounded: we will sometimes be late to a genuine break, by hours or by days, and we accept that cost in advance, the way one accepts an insurance premium. The cost of the alternative is unbounded and, worse, unmeasurable. A process contaminated by narrative cannot tell you afterward which decisions were its own, and a firm that cannot audit its own reasoning cannot improve it.

There is also a matter of what the news does to the people running the process, which firms are reluctant to discuss because it sounds unprofessional. It is unprofessional, and it is true. A steady diet of market journalism trains attention toward the dramatic and the immediate, and away from the statistical and the slow, which is where the effects we study actually live. Researchers begin, without noticing, to study the questions the news has made salient, and salience is a terrible grant committee. A research programme can become, on inspection, an elaborate answer to last year's headlines. Disassociation from the news is therefore also a defence of the research agenda: we study markets historically, in bulk, at the level of decades of data, and we let the questions come from the data's structure rather than from this morning's anxieties.

We apply three tests to every input before it is allowed near a decision, and it is worth stating them because the news fails each one in a different way. The first is provenance: who produced this, and what were they paid to optimise? The second is time: can we say, to the minute, when this information became available to someone who could have acted on it, and can we reconstruct the version that existed then rather than the version that exists now? The third is replay: if we run the same process on the same inputs tomorrow, do we get the same decision? Prices pass all three. A headline fails the first by construction, fails the second because stories are revised, re-headlined and re-ranked after publication, and fails the third because the reading of a story depends on who is reading it and in what mood. An input that fails one test can sometimes be repaired. An input that fails all three is not data. It is weather, and we dress for weather rather than trade on it.

None of this is a claim that news is worthless to everyone. It moves prices in the very short run, and firms built for that horizon use it as raw material, properly, as one signal among many, tested like any other. Our horizon is not that horizon. For a process that holds positions on rules validated over long histories, news fails the tests above: we cannot state what it adds, we cannot measure when it helps, and we cannot reproduce a decision it influenced. An input that cannot be audited cannot be sized, and an input that cannot be sized does not belong in a portfolio, however interesting it is. We read the news the way other citizens do, to know what world we live in. We simply refuse to let it trade.