Your portfolio can tell when you're tired

2026-10-21 · behavioral-finance, investing · about the paper: Hee Seo Han et al., 2025

I do my worst financial thinking at 9:47pm, phone above my face in the dark, checking a position I already checked this morning. It always felt harmless, a little entertainment with my thumb. A recent working paper suggests that habit has a price tag, and the price tag is in basis points.

The paper asked a question that sounds like a joke until you see the results: what happens to your trading when the sun stays up later? (Han, Hirshleifer, Sheng and Sun, NBER Working Paper 33477)

TL;DR: An hour of later sunset cost retail traders about 29 basis points a month on their trades. At time-zone borders the gap was 40 bps, in northern summers it hit 64. The mechanism looks like sleep, and the specific failure is missing news about companies you own.

How you even study sunsets

The authors took one of the most famous datasets in finance, the Barber-Odean collection of retail brokerage accounts (10,000+ accounts, 1991 to 1996), and matched every household to the sunset time in their ZIP code. Then they got creative about isolating the effect, because "people in sunny places trade differently" is not a finding, it's a confound festival.

They ran four separate tests:

• The same neighbourhood, month to month. Does performance drop when local sunset gets later? (ZIP and year-month fixed effects, so you're comparing a household to itself.)
• Time-zone borders. Two similar households a few kilometres apart, one hour apart on the clock, roughly the same actual daylight. If the later-clock-sunset side trades worse, the clock is the suspect.
• Latitude. Northern cities get much later summer sunsets than southern ones, then the pattern reverses in winter. Sleep disruption predicts one shape, seasonal depression predicts another.
• Daylight saving transitions. Compare performance around the spring clock change in places that observe DST against places that don't.

Returns were measured as abnormal returns on the traders' portfolios, meaning returns adjusted for what the market was doing anyway.

The scoreboard

One basis point is a hundredth of a percent. On $10,000 of actively traded money, 29 bps a month is about $29 a month of performance the tired traders left behind, and the worst test found closer to $64.

The latitude result is my favourite, because it's the one that kills the boring explanation. Seasonal affective disorder should bite hardest in the dark northern winter. The damage instead tracks later evening light, peaking in northern summers, which is exactly the shape a sleep story predicts and a mood story does not. The result holds even after controlling for the standard SAD measure.

The part that got me: it's an attention problem

Here's the detail I can't shake. Traders with later sunsets were less likely to trade a held stock during its earnings announcement window, and reacted more weakly to earnings news when it landed. They also tilted harder against earnings surprises, which is what you'd expect from someone who missed the announcement and is now guessing at why the price moved.

That's not "tired people pick worse strategies." That's tired people missing the memo and then rationalizing whatever their screen shows them. The impaired asset isn't your portfolio, it's your attention, and the portfolio just inherits the damage.

The Canadian angle

This is not a tropical problem. Around the summer solstice, sunset in southern Ontario slides past 9pm, and in cottage country it's later still. Add the spring daylight saving jump, which the paper finds costly in its own right, and Canadian summers serve up months of exactly the evening-light exposure the study flags. The sun goes down, the phone comes out, and the market opens in fifteen hours whether you sleep or not.

The honest caveats

The brokerage data runs 1991 to 1996, before commission-free trading and push notifications, so today's effect could be bigger or just different. The numbers are monthly abnormal returns on trades, not a permanent annual drag on a buy-and-hold portfolio, so a person who trades rarely inherits little. Nobody wore a sleep tracker, the mechanism is inferred from the patterns, and the daylight saving leg of the study is the shakiest by the authors' own admission. It's a working paper, not a finished, peer-reviewed one.

What to actually do

• Put money decisions on a morning schedule. The market's entire trading day happens while you're awake and functional anyway, so there is no good reason the decision-making has to happen at 10pm.
• Treat late-night portfolio tinkering the way you'd treat late-night texting an ex. Nothing you send at that hour improves your life, it just feels urgent.
• Automate the recurring stuff so fatigue never gets a vote: contributions on payday, rebalancing on a calendar date, decisions made once instead of nightly.
• The boring portfolios sleep through all of this. That's a feature.

How strong is the evidence?

Scale: Strong = replicated across multiple peer-reviewed studies · Moderate = one solid peer-reviewed study or consistent findings in limited settings · Suggestive = working paper, small samples, or not yet replicated · Thin = non-peer-reviewed source or conflicting evidence.

Next steps

• Even the pros sell worse than a coin flip — another way ordinary attention quietly taxes a portfolio.
• The skim view: digest + full PDF — every claim above, one tap from the source.
• The paper itself — Han, Hirshleifer, Sheng and Sun, NBER Working Paper 33477.