Separating Signal From Noise in Headlines

Most financial headlines are not news. They are restatements, commentary, or a known fact wearing a fresh verb. Sorting them is not a matter of reading more, it is a matter of applying a short test, and the fastest version of that test takes about thirty seconds.

Two headlines, one morning, and only one moved anything

At 09:14 a wire runs: "ECB official warns inflation risks remain elevated." It leads three sites within the hour. EUR/USD moves four pips and is back where it started inside a minute. The official had said a near-identical thing twice in the previous fortnight, and the committee's stance was already in the price.

At 13:31 the US payrolls release lands. The headline figure is close to forecast, so the wire lead is "jobs growth broadly in line with expectations", which is about as unexciting a sentence as financial journalism produces. In paragraph eight of the statistical release, the previous two months are revised down by a combined 112,000. EUR/USD moves 55 pips over the next twenty minutes.

The alarming headline was noise. The dull one carried the information. What separated them was not tone or prominence. It was whether the item changed what the market had already expected, which is the same mechanism described in how forex prices move.

Four questions that sort a headline in thirty seconds

1. Is it new? Has this specific fact been reported before, and if so, did price move then. A restated position is not an event.
2. Is it already priced? If interest rate futures have shown an 85% probability of a hold for a fortnight, a headline saying the bank is expected to hold changes nothing.
3. Who says so? An on-the-record name, a statistical agency or a regulatory filing sits in one category. "People familiar with the matter" sits in another. An anonymous social account sits in a third.
4. What did price do? The market's own answer, available within two minutes, and usually more reliable than your reading of the story.

Recycled news has a tell

Aggregation makes one comment feel like twelve events. A single remark at a conference is written up by a wire, rewritten by four sites, summarised by two newsletters and posted by dozens of accounts, and by the afternoon it reads like a developing situation when it is a single sentence spoken once.

The tell is that the article contains no time-stamped new fact. It quotes a speech from three days ago, it uses "renewed concerns" or "growing fears" in place of an event, and the only genuinely new element in it is an analyst's opinion about the old fact. Once you look for a date on the underlying fact rather than on the article, a large share of daily financial coverage sorts itself.