The Friday Deep Dive — The Payday Effect

The test that changed everything

We took our entire analysis and asked one simple question: does the signal survive if we remove all Fridays from the dataset?

If the clearing-lag effect is about payroll settlement timing, it should work on any day. Money settling on a Tuesday should produce the same run-up as money settling on a Friday. But if something specific about Friday itself is driving the signal, removing Fridays should kill it.

We re-ran the full semi-monthly lag sweep twice: once with all trading days, and once with every Friday excluded. Same regression, same controls, same 65 years of data — just minus one day of the week.

The results

Semi-monthly run-up coefficient: with vs. without Fridays

Grouped bar chart comparing β_run at each lag with all days (blue) vs. Fridays removed (gray). Stars indicate statistical significance. Every significant result vanishes without Fridays.

Every single significant result vanishes when Fridays are removed. The signal is 100% Friday-dependent. With Fridays: lag+7 (p=0.029*), lag+8 (p=0.019*), lag+17 (p=0.013*), lag+18 (p<0.001***). Without Fridays: not one of these lags reaches even p<0.10.

Why Fridays?

The complete Friday dependence opens four hypotheses, none mutually exclusive:

Friday settlement clustering

Employers batch payroll on Fridays. ACH transfers initiated Friday arrive Monday through Wednesday. The clearing pipeline concentrates settlement activity on Fridays because that is when the bulk of payroll processing kicks off. The lag+7 window from a semi-monthly payday frequently lands on or near a Friday, capturing the accumulated settlement flow.

Options expiration interaction

Monthly options expire on the third Friday. Gamma hedging by market makers creates predictable directional flow on expiration Fridays. Our earlier test of FOMC, OpEx, and CPI dates showed options expiration didn't fully explain the signal, but the Friday concentration suggests a partial interaction — some of the elevated Friday returns may reflect hedging activity layered on top of settlement flow.

Weekend effect inversion

The classic "Monday effect" — negative returns on Mondays — may interact with Friday buying pressure. If settlement flow pushes prices up on Fridays, it partially offsets the typical Monday decline. The payday signal on Fridays could be the mechanism behind the well-documented weekend-effect asymmetry.

Biweekly-semi overlap

All biweekly paydays ARE Fridays. Biweekly flow shows significant volume spikes (see below) but no detectable return signal on its own. Since biweekly paydays hit every other Friday (~50% of Fridays), the semi-monthly signal on Fridays could capture combined flow from multiple pay schedules — semi-monthly and biweekly workers settling on the same Fridays.

The biweekly volume story

Biweekly is the most common pay schedule in America, covering roughly 36% of workers. Every biweekly payday is a Friday. So what happens when we look at biweekly paydays?

Biweekly volume z-score by clearing lag

Orange bars are statistically significant (p < 0.05). Volume spikes are massive at lag+3, +7, and +8 — but there is zero return signal at any lag.

Biweekly paydays show highly significant volume spikes at lag+3, +7, and +8 (all p < 0.001) but zero return signal. The money IS flowing — we can see it in volume — but it's too diffuse (every other Friday = 50% of Fridays) to create detectable price pressure.

Think of it like traffic. Biweekly flow is like adding 10% more cars to a highway that's already busy — you can measure the extra cars (volume) but they don't change the speed (returns) because the flow is spread across too many days.

Decomposing the signal

The semi-monthly signal at lag+7-8 is actually a composite. To understand what's driving it, we separated the two anchors (the 15th and end-of-month) and compared them against the combined semi-monthly schedule and the military schedule (1st and 15th).

Run-up coefficient by payday type and clearing lag

Each line represents a different payday anchor. Filled dots mark statistically significant lags (p < 0.05). The semi-monthly composite (red) only becomes significant at lag+7-8 when both anchors are combined.

Neither the 15th alone NOR the end-of-month alone shows significance at lag+7-8. The signal only appears when they're COMBINED (semi-monthly). The military schedule (1st + 15th) independently confirms lag+7 at p = 0.001 — the strongest single-lag signal of any schedule.

How to read this table:

Anchor = the payday date used to start counting.
Signal location = which lag has the strongest statistical signal.
β = how much extra daily return (in %) during the run-up window.
Confidence = how sure we are this isn't random luck.
What it means = plain-English interpretation.

Anchor	Signal location	Extra daily return	Confidence	What it means
End of month	lag+0	+0.056%	High ** p = 0.004	Classic turn-of-month effect — immediate settlement from the fastest processors.
15th	lag+16-18	+0.057%	High ** p = 0.004	Points to the NEXT month's turn-of-month — same calendar dates, different anchor.
Semi (combined)	lag+7-8	+0.055%	Solid * p = 0.019	Composite — only appears when both anchors are combined, doubling measurement opportunities.
Military (1st + 15th)	lag+7	+0.073%	Very High *** p = 0.001	Independent confirmation — different payday dates, same lag, strongest single-lag signal.

What this means — the revised theory

Your 401(k) money enters a pipeline. Some arrives fast (day 1-2 after payday), some slow (day 5-7). But the statistical signal only shows up on Fridays. This likely means Friday is when the bulk of settlement activity concentrates — either because employers batch payroll on Fridays, or because Friday trading captures the accumulated flow from the entire week.

The core finding — a detectable calendar pattern tied to payroll timing — survives. But the mechanism is more specific than "money arrives at lag+7-8." It's a Friday-concentrated clearing effect that interacts with settlement infrastructure. The cost to investors is real; the cause is structural, not adversarial.

What changes: the "manipulation" framing weakens further. If the effect is driven by Friday settlement mechanics, it's infrastructure, not intent. But the per-person cost estimates remain valid — regardless of mechanism, the price difference between optimal and typical settlement days is real and measurable.

Next: did this pattern always exist, or did it emerge with 401(k)s? →