RESEARCH & FINDINGS

We tried to explain the premium and could not

It tracks dividend yield closely enough to look solved, then two stocks break it and an experiment we ran on the day came back empty. What follows is what we measured, including the parts that do not fit.

Intraday research1 bp = 0.01%Explanation unresolved

Usable samples: 1,025
Observed time: 13.5 hours
Sampling cadence: 3 minutes
Research status: Unresolved

01 / THE TIME SERIES

First, is it even stable?

Everything below this is one measurement per stock at one moment, which cannot tell a structural premium from a number that happened to be there when we looked. So we sampled every three minutes and asked a different question: does each sample predict the next one?

Premium over the sampling period

Basis points · session gaps removed

Hover or tap a point. Focus the chart and use arrow keys to inspect.

Select a stock to highlight · select again to reset

The shape is the argument. SPY holds a band roughly five basis points wide for hours at a time; Tesla crosses its own zero line again and again. Same sampler, same three-minute cadence, same moments.

Series summary - premium in basis points
Stock	Samples	Mean ± SD	Persistence (lag 1)
SPY	260	54.9 ± 5.9 bp	0.929
AAPL	261	27.4 ± 11.1 bp	0.846
NVDA	247	12.3 ± 10.5 bp	0.439
TSLAControl	257	-3.5 ± 8.6 bp	0.034

Persistence is the correlation with the next sample. TSLA control: the one stock with no premium, and the only one with no structure.

Tesla is what makes the rest readable. Every stock carrying a premium predicts itself three minutes later at 0.85 or better. Tesla, which carries none, sits at 0.05 — indistinguishable from noise. The sampler finds structure where a premium exists and none where it does not, which is what rules out our own measurement as the thing producing it.

1,025 usable samples over 13.5 hours in 5 sessions. Three were discarded by a stated 1% spread rule — one a genuine after-hours route failure at 36%, two merely wide at 1.2%; the 99th percentile of every other sample is 0.80%. Premium correlates with oracle staleness at −0.34 to +0.02, so a stale price is not it. This is intraday evidence only — we have no overnight or weekend coverage.

02 / CROSS-STOCK EVIDENCE

It is not pool size, and it is not liquidity

If demand drove the premium, the biggest wrappers would carry the most of it. Dividend yield plotted against each stock’s premium:

Dividend yield vs premium

6-stock snapshot

Dashed line: ordinary least-squares fit across all six stocks, with an intercept. Circled NVDA falls below the dividend trend. Circled TSLA is the zero-yield control that challenges the size/demand explanation; it does not break the dividend trend. Correlation is not a causal explanation.

Reported snapshot - original stock order
Stock	Yield	Premium	AUM
SPY	0.98%	57.1 bp	$72.9m
QQQ	0.40%	27.3 bp	$60.8m
AAPL	0.32%	26.6 bp	$51.9m
GOOGL	0.23%	19.3 bp	$56.5m
NVDA	0.46%	9.2 bp	$70.6m
TSLA	0.00%	0.0 bp	$83.6m

NVDA / EXCEPTION

pays more than AAPL, carries a third of its premium

TSLA / EXCEPTION

largest wrapper here, premium of exactly zero

premium vs yield: 0.887
excluding NVDA: 0.986
premium vs liquidity: 0.780
premium vs AUM: −0.235

Tesla settles it. It is the largest wrapper in the set, it trades actively, and its premium is zero to the decimal. No size or demand variable predicts that. Paying no dividend does.

03 / THE NATURAL EXPERIMENT

The experiment, and the null result

SPY went ex-dividend during the build. A $1.90 dividend against a $759 share predicts a step of roughly 25 bp at the opening bell — widening, since the share drops while the token keeps the claim. We watched it happen.

Predicted opening step~25 bp upward

Observed opening move5.4 bp downward

SPY spread3 bp

The step that did not appear

Change from the pre-open premium (bp)

Dashed gold: the hypothesized +25 bp step, held constant for comparison. Solid blue: the four recorded observations, joined as a visual guide. Observations are evenly spaced; the opening marker is schematic, not an exact event timestamp. The first post-open observation is -5.4 bp from the 70.5 bp baseline.

View the four recorded values

Recorded time	Observation	Premium
03:50	before the open	70.5 bp
11:59	after the open	65.1 bp
12:11	Later observation	60.5 bp
15:07	Later observation	53.1 bp

Across the opening bell the premium moved 5.4 bp, and downward. The spread on SPY is 3 bp, so a 25 bp step in either direction would have been impossible to miss. There was no step.

04 / AN OPEN QUESTION

Where that leaves it

The premium follows dividend yield and sits at zero for the one stock paying none. It also ignores an ex-dividend date, implies seven to ten months of accrual on a wrapper that launched fourteen months ago, and inverts on Nvidia. We did not force a story onto that.

What holds without qualification: the premium is measurable, it reproduces, and no quote anywhere shows it. That is what the guard checks against — not the reason behind it.

Back Price another trade