Geomagnetic Storms and the Stock Market: What the Research Found

Geomagnetic storms and the stock market are linked in one published study: a 2003 Federal Reserve Bank of Atlanta working paper found that unusually high geomagnetic activity in the previous week was associated with lower stock returns today. The authors call the effect statistically and economically significant for the world index and most of the international indices in their sample. That is a real result from a serious institution, and it is worth understanding properly. It is also an average over a past sample, not a signal for next week.

What the study says

The paper is "Playing the field: Geomagnetic storms and international stock markets" by Anya Krivelyova and Cesare Robotti, Federal Reserve Bank of Atlanta Working Paper 2003-5. It documents the impact of geomagnetic storms on world and country-specific stock returns. In its own words, for the world index and most of the international indices in the sample, unusually high geomagnetic activity in the previous week is associated with lower returns today.

The proposed mechanism is psychological, not magnetic. The authors draw on research on "misattribution of mood": geomagnetic storms have been linked in other studies to poorer judgment and behaviour, and that could carry over into investors' decisions. In plain terms, the idea is that a storm affects how people feel, and people who feel off make slightly different choices.

One honest note on our own reading. We could read the abstract and the publisher's citation record, but the full text did not load in this run. So we are not quoting the sample years, the country list, the exact threshold for "unusually high" or the size of the effect. If those details matter to you, read the paper itself from the Atlanta Fed or FRASER before relying on anything here.

What a geomagnetic storm is, and how it is measured

A geomagnetic storm is a disturbance of Earth's magnetic field, usually driven by solar wind and eruptions from the Sun. Ground magnetometers around the world feed the planetary Kp index, which runs from 0 to 9 in three-hourly steps. NOAA's Space Weather Prediction Center turns Kp into a storm scale:

Source: NOAA Space Weather Prediction Center, fetched 13 October 2026. Kp below 5 is not counted as a storm. The practical point is that minor storms are common, and the strongest ones are rare. Any rule built on "a storm last week" depends heavily on where you set the line.

A live example from this week

The Kp feed is public and free, so you can look at it yourself. The feed we fetched on 13 October 2026 covers 61 three-hourly readings from 5 October to 12 October. Four of them reached Kp 5 or higher:

• 5 October, 03:00 UTC: Kp 5.33
• 10 October, 15:00 UTC: Kp 7.00, which is G3 (Strong) on the NOAA table
• 10 October, 18:00 UTC: Kp 6.67
• 10 October, 21:00 UTC: Kp 5.33

So there was a strong storm on Saturday 10 October. These are preliminary values, and NOAA revises them later.

A storm on a weekend is a neat case for the study's idea, because the first trading day afterwards is Monday 12 October. Using Yahoo Finance daily closes for the S&P 500 (ticker ^GSPC), the close was 7,811.54 on Friday 9 October and 7,805.09 on Monday 12 October. That is a change of about minus 0.08%, calculated as 7,805.09 divided by 7,811.54, minus one. One caution: the Monday row carried a much lower volume than the other days, which suggests the data was captured before the session ended, so treat that close as provisional.

This is not a test of the paper and we do not present it as one. It is one storm and one day. The paper's claim is about an average across many storms and many markets, and a single small move in either direction tells you nothing about it. We include it only to show how quickly you can line up a Kp date against a price file.

What the paper does and does not give you

What it gives you:

• A published, peer-visible reason to think a physical variable could touch investor behaviour.
• A clear, checkable definition of the input: Kp and related indices are public and measured, not opinion.
• A testable idea. You can line up storm dates against any index you like and see whether the gap is bigger than chance.

What it does not give you:

• A trading signal. An average effect across a long sample can be real and still be far too small, or too inconsistent from year to year, to trade once you pay brokerage and spreads.
• Proof of the mechanism. "Misattribution of mood" is a proposed explanation. The statistical finding and the story are separate things.
• Anything about next week. Storms are driven by the Sun's activity, and the count of strong storms varies a great deal across the 11-year cycle.

How to test it yourself

You do not need to take anyone's word for it, ours included. A fair check looks like this:

1. Download Kp history from NOAA and pick a clear definition of a storm day, for example Kp of 5 or more, before you look at prices.
2. Get daily closes for an index over a long period, ideally including more than one solar cycle.
3. Compare the average return over the following week after storm days with the average over all other weeks.
4. Ask the key question: is that gap larger than what you would get by picking the same number of random dates? Run the random version many times.

Step four is the one that most informal write-ups skip. Our backtesting primer covers how to build that comparison against chance, and the Bartels test post shows one way to judge whether a repeating pattern is more than luck.

This is the same approach we take to every timing idea. The Crohamhurst app computes planetary positions from the JPL ephemeris and gives you dates in seconds, and it lets you test an idea against chance before money goes near it. Space weather is outside what the app computes today, so for this one you use NOAA's data and your own spreadsheet or script.

The takeaway

There is published research behind the idea that space weather touches markets, and it comes from a Federal Reserve Bank, not a forum. Take that seriously. Also take seriously that it is a working paper's average over a past sample, that we have only read its abstract, and that one quiet Monday after a G3 storm proves nothing in either direction.

This is research and education, not financial advice.

Try it yourself

Create a free account and get 50 compute units a day to run the modules on your own dates. The ephemeris guide shows how to list planetary aspect dates, and the cycle scanner guide shows how to look for repeating cycles in any price series and check them against chance.