Stock Reaction to Earnings: The First Hour Is Not Yours

The stock reaction to earnings is decided in the first seconds after the release, and not by people who read the filing. It is decided by programs that scan the text for keywords and send orders before you have opened the file. The first hour does not belong to you. Your work starts the next morning and it is a different job: checking which assumption in your thesis this quarter confirmed, and which one it broke.

What the stock reaction to earnings prices in#

The usual explanation is correct as far as it goes. Markets trade expectations, not results. Before the release the price already contains what analysts expect, so on the day the company is measured with somebody else's ruler.

Say the market expects 10% revenue growth and the company delivers 12%, then cuts guidance for the next quarter from 10% to 6%. Quarter won, year lost. The buyer is paying for twelve months of promise, and the quarter that just ended is already in the price.

The consensus itself is softer than it looks. It is an average of forecasts from a dozen or so people, some updated last week and some updated last spring. The market trades the deviation from that average, which means it trades the deviation from other people's assumptions about your company.

This is where the empty headline "the company beat expectations" comes from. Beating by a cent is routine, because management teams spend years learning to set a bar they can clear, and the information is whatever was not in the consensus at all.

The same goes for a sentence that looks harmless in the filing. A delayed launch, a change in the manufacturing process, one fewer name in the top ten customers: any of them can outweigh record revenue, because they concern quarters that are not in the table yet. Sentences like that usually turn up in the Q&A rather than in the release.

Who is actually trading in the first minute#

Not you, and not the fund that reads filings. In its literature review from March 2014 the SEC repeated its own earlier finding: estimates of high frequency trading at the time exceeded 50% of total volume in US-listed equities. So the other side of roughly every second trade was a program, not a person with a view on margins. That measurement is old, and I know of no newer one with the same standing. The hardware has not slowed down since.

Those programs do not wait for an analyst. The same review summarises research in which the most aggressive of them reacted to the flow of positive and negative newswire items about a specific company, which means they reacted to the text itself before anyone had interpreted it. In an earnings release one word off a short list is enough: miss, guidance, recall, investigation. You are still waiting for the PDF to load.

So "the market discounts the future" means something other than we assume: the price had already moved before the headline reached your retina, and at minute zero half of the material does not exist.

The rest of the advantage is physical. Listing the characteristics of this kind of trading, the regulator names co-location services and individual data feeds in the same document: a server in the same room as the exchange's matching engine, and prices delivered by a shorter route than everyone else gets. What that buys is metres of cable.

So "the market discounts the future" does not mean what we assume it means. The price finished changing before the image of the headline reached your retina.

How much of a head start has to be handed back before somebody calls the race even? IEX built a deliberate delay into its order handling, and the SEC approved the exchange in 2016 with that delay in place: 350 microseconds.

The threshold for what counts as negligible sits somewhere else. In its press release on the decision the SEC said that staff guidance accompanying the Commission's interpretation treats delays of less than one millisecond as de minimis. Three hundred and fifty microseconds is 0.00035 of a second, roughly a third of that threshold. It was enough for an exchange to consider the field level, and it tells you the units in which a reaction to a quarterly report is now measured.

The race finishes before you hear the starting gun.

The quarter the numbers lost to a sentence#

On 28 August 2024 Nvidia reported a quarter most companies never get to have. Its own release shows revenue of $30.0 billion, up 122% year over year, and guidance of $32.5 billion for the following quarter. The stock fell nearly 7% in after-hours trading.

What traded that evening was the manufacturing of the next generation of chips. Asked about the change to the Blackwell mask, Jensen Huang answered on the earnings call: "The change to the mask is complete. There were no functional changes necessary." The chief executive was reassuring the room, and the fact that he had to reassure anyone was itself the news.

The proportions are worth remembering. Revenue up 122% year over year is a change you do not see for years at that size of business, and it lost to two sentences about a lithography mask.

So the record quarter had already been priced, and the only fresh content in the whole report was a risk nobody had modelled. It works this way every time: a result everyone expected moves a price far less than a sentence nobody expected.

For a shareholder the same quarter carried entirely different material. One question mattered: whether the schedule of the product the whole thesis rests on had just slipped. Answering it takes a few days and cannot be extracted from a chart. Answering it takes days rather than minutes.

Why the after-hours price misleads you#

US companies usually report after the close, so the first reaction you see happens in the after-hours session. Same instrument, completely different conditions.

There is a fraction of the daytime crowd, the order book is thin, and the gap between bid and offer is the widest it will be all day. A single large order can push the price several percent, because there is almost nobody on the other side of it. The number you see half an hour after the close mostly describes whoever sent that order.

Which is why the open of the next session can look nothing like the number from the evening. Not because the market changed its mind, but because the market has only now walked into the room.

The only reaction worth looking at is the close of the next session.

Why the price moves between earnings#

If the fundamentals change once a quarter, the price should sit still for the other three months. It does not. That is not the fundamentals' fault either.

One of the oldest algorithmic strategies watches pairs of companies whose prices have historically moved together. When one rises and the other does not, the program assumes the old correlation will return and sells the laggard. So your company can fall because a competitor had a good day.

How do we know this is not a lottery? Because every one of those moves has a reason, it simply is not a reason about the business. These programs are written by a great many people, each working from a private hypothesis, so on any given day a single stock hosts independent wagers about things that have nothing to do with what it sells. Price moves outside results season have a list of causes of their own, most of which appear in no release.

What the first hour does not tell you#

Here is the honest counter-argument. If the information in a quarter is that valuable, surely some of it can be earned on the day.

The research has said the opposite for a long time. Bernard and Thomas on post-earnings-announcement drift, published in 1989, showed that prices keep drifting in the direction of the surprise for weeks after the announcement. Which means the information in one quarter is not consumed in an hour. It is consumed at a pace a human can keep up with.

That is not an argument for trading the print. The drift was measured across hundreds of companies and portfolios held for weeks, not on one company in the first hour, when the spread is the widest it gets all quarter.

I will put my own view on the record: the first hour after a release is the most expensive hour of the quarter for a private investor, because you pay the full cost of entry for information you could read calmly the next morning.

There is also an arithmetic nobody does. A wager on a print resolves overnight, so to break even you have to call the direction more often than half the time and cover your costs on top. Analysis of a business pays out over years of ownership, and it gets confused with that wager only because the two involve the same piece of paper.

The risk in that claim deserves naming. If your company is small, thinly traded and covered by nobody, the market genuinely is slow, and a filing can take days to be read properly. I also do not know how strongly the same mechanism applies to smaller companies on the Warsaw exchange, where the algorithmic share is lower and public measurement is close to non-existent. That changes the pace, not the principle, because even a slow market will price a filing faster than you can understand it.

What to do in earnings week#

So why open the report on the day at all? To have read it yourself before anyone tells you what was in it. The useful reading comes a day later, when a quarter answers three questions: whether the assumption your thesis stands on got confirmed, whether management met a deadline it set itself, and whether a risk appeared that was not in this business a year ago.

The whole job fits into four steps, none of which requires being fast.

1. Before the release, write one sentence: what you expect from this quarter, and which number would confirm or break your assumption.
2. On the day, read the release and note the difference between your sentence and what the company showed. Nothing beyond that.
3. The next morning, read the transcript and write down the sentences that appear for the first time. A full day later, because only by then is the set complete: the numbers, the transcript and the first analyst notes.
4. After a week, go back to your thesis and delete or rewrite the assumptions this quarter undermined.

There is no decision in it and no adrenaline, only a crossed-out sentence and a date. A year later those deletions are what tell you whether you understand the company or merely like it. The version with four quarters lined up side by side fits into four columns: topic, before, now, difference.

Before the release Taufolio sends a Pre-earnings brief, a short preparation: what to expect and where to look. The Investment thesis, generated from a Full report, breaks your reason for owning the company into assumptions with a measurable condition, so after each quarter it is visible which of them fell apart.

A calendar, not a stopwatch#

At Taufolio this rhythm sits inside Monitoring. After the release, the Earnings brief collects the numbers and what management said. Nothing in Monitoring costs credits, and on the Free plan it covers one company. What that looks like from the inside is on the product page.

My own rule is one sentence long: on the day of a release I make no decision, I only put the coming Friday in the calendar.