Nine Tickers, Zero Timestamps: Auditing the Trump Energy Disclosures Like a Protocol

The Tape Without a Clock

On April 7, an account tied to U.S. President Donald Trump sold ExxonMobil shares valued between $500,000 and $1,000,000. Roughly two and a half hours later, a ceasefire with Iran was announced. The stock opened more than 6% lower the next session.

Now reverse the tape. On March 2, the first trading day after the joint U.S.–Israel strike on Iran, the same account bought shares in eight oil and gas companies, including ExxonMobil in the $100,000–$250,000 band. On March 23, after a planned strike on Iranian energy facilities was delayed before the market opened, Brent crude fell nearly 11% on the day, and the account filed sixteen buy transactions in oil and gas names totalling roughly $163,000 to $570,000.

CNBC estimates that nine major oil and gas holdings — ExxonMobil, Chevron, ConocoPhillips, Occidental Petroleum, plus several refining and pipeline companies — appreciated somewhere between $1.5 million and $4.4 million across the window from February 27 to August 31. As of June 29, the account reported at least 23 transactions involving related stock sales.

I don't trade oil. I read tapes. And this tape is missing its timestamps.

Nine tickers. Twenty-three transactions. No share counts. No execution prices. No sale batches. No block height. The most consequential structure in this entire story is not the profit line — it is the field schema of the disclosure document that produced it.

That is the anomaly worth opening a terminal for.

What the Disclosure Layer Actually Publishes

Start with the mechanics, because the mechanics determine what can and cannot be concluded.

Executive branch financial disclosure in the United States runs on a two-part protocol. The annual filing — OGE Form 278e — reports holdings and income as of a snapshot date, in broad ranges. The periodic transaction report covers individual trades, and it is the more interesting object. A covered filer must report a transaction within 30 days of receiving notification of it, and in no case later than 45 days after the trade itself. Amounts are bucketed into wide bands. Share quantities are not reported. Execution prices are not reported. Execution timestamps are not reported.

So a purchase executed on February 27 can enter the public record in mid-April. Two transactions executed minutes apart can appear as separate line items with no ordering information. A sale of 4,000 shares and a sale of 40,000 shares can appear as the same line, because the band is the unit of resolution.

This is an oracle. It is not a bad oracle by accident — it is a low-frequency, interval-valued, event-censored oracle by design, because its stated objective is conflict-of-interest disclosure for the public, not forensic attribution of intent. Those are different objective functions. Concealing exact position sizes protects the filer and their family from a specific class of harm. The trade-off is real. But we should be precise about what we are reading: every oracle is a trust assumption wearing a Merkle proof, and this one wears its range buckets the way a light client wears a checkpoint.

The White House position, as reported, is that the portfolio is managed entirely by independent managers. CNBC explicitly states that it found no evidence that Trump directed the trades, had prior knowledge of the relevant decisions, or that personal interests shaped policy. I take that statement at face value. It is also, structurally, the only statement the instrument is capable of producing. The disclosure format cannot generate directed-trade evidence in either direction, because directed-trade evidence requires co-location of three things — actor identity, execution timestamp, and decision timing — and the format publishes exactly one of them.

That is the first information gain of this piece, and it is worth stating plainly: the current disclosure regime is structurally incapable of distinguishing an informed trader from a lucky systematic sleeve, and every headline you will read about this story inherits that limitation whether it admits it or not.

Four Anchors, One Schema Problem

Strip the narrative and you are left with four data points with unusually clean geometry.

| Date | Event | Reported account activity | |---|---|---| | Mar 2 | First session after joint U.S.–Israel strike on Iran | Buys in eight oil and gas names; ExxonMobil $100k–$250k | | Mar 23 | Strike on Iranian energy facilities delayed pre-open; Brent −11% | 16 buy transactions, ≈$163k–$570k | | Apr 7 | ExxonMobil sold $500k–$1M | Ceasefire announced ~2.5h later; −6% next open | | Feb 27–Aug 31 | Holding window | Nine-name basket est. +$1.5M–$4.4M |

Four events, one sector, one latent variable. That is not a stock-picking profile. That is a factor expression.

Here is the schema problem in one line: I can tell you the date of the transaction and the date of the announcement. I cannot tell you the time of either, relative to the other, at any resolution finer than a day — and on March 2 and April 7, the entire question lives in a window measured in hours.

Two and a half hours. That is the interval. It is almost exactly the kind of gap that a low-latency actor exploits and exactly the kind of gap that a long-horizon systematic sleeve ignores. The disclosure format rounds both possibilities to the same integer.

Any competent anomaly detector will tell you the same thing: you cannot score a signal you cannot timestamp. You can only score the outcome, and scoring outcomes is how you generate false positives at industrial scale.

The Mempool We Can't See

Let me put this in the language I actually think in, because the structural analogy is tighter than it looks.

On Ethereum, a searcher and an insider are both latency arbitrageurs. The searcher watches the mempool and front-runs a pending transaction; the edge is verifiable because the pending transaction is public before it lands. The insider watches a decision process and positions before the decision is public; the edge is not verifiable because the decision process has no mempool. There is no pending-transaction pool for a National Security Council meeting.

That asymmetry is the whole story. The March 2 entry is structurally a front-run of a known pending event: the strike had already happened, so the trade is post-event, not pre-event, and the relevant question becomes whether the follow-on escalation path was priced. The March 23 entry is structurally a dip-buy on a policy-generated dislocation — the same actor generated the news that moved the tape 11%, then bought the tape. The April 7 exit is structurally the inverse: close the long-volatility expression before the volatility resolves.

Symmetric around a volatility event. Entry, add, exit. If I saw that pattern in a wallet, my first hypothesis would be a sophisticated directional trader with a thesis and my second would be a sandwich bot with visibility. I could distinguish them on-chain in about ten minutes, because I would have block.timestamp, gas price, and the full state transition history.

Off-chain, I have a date and a range bucket.

The consequence is not that the pattern is meaningless. The consequence is that the pattern is underdetermined. And underdetermined patterns are exactly what confirmation bias feeds on. Show this table to someone who already believes the story and they see a smoking gun. Show it to someone who already disbelieves it and they see a wealthy man with a diversified energy sleeve and a volatile year.

Both readers are doing the same thing: extracting a narrative from a schema that cannot support one.

Hormuz Is the Settlement Layer

Now the part where I have standing, because this is an infrastructure question before it is a stock question.

The Strait of Hormuz carries something on the order of a fifth of global petroleum liquids consumption — crude plus refined products — along with a substantial share of seaborne LNG. Think of it as the settlement layer for physical energy. Everything downstream — refinery runs, crack spreads, tanker routes, war-risk insurance — settles against whether that channel stays open.

A military strike on Iranian energy infrastructure is a congestion event on that layer. The first thing that reprices is not spot crude. It is the fee market: war-risk insurance premiums for vessels transiting the strait. Premiums move before barrels do. That is the priority-fee analogue, and it is the leading indicator that actually carries information.

Brent's nearly 11% single-day move on March 23 — on news of a delay, not an escalation — tells you how much tail probability the market had loaded into the wings. A delay removes a slice of near-term strike probability, and the front month unwinds accordingly. That is not a stock market event. That is a probability mass shift on a binary outcome with a physical delivery mechanism attached.

And this is where the disclosed instrument stops being the interesting instrument. The nine-name energy basket — integrated majors, refiners, pipelines — is a low-beta proxy for that probability mass. Integrated majors are hedged across upstream and downstream; refiners are exposed to crack spreads; pipelines are largely contracted and volume-insensitive. Buying the basket is buying roughly 0.3 to 0.6 beta to the underlying war-probability factor, not the factor itself.

Gas wars are just ego masquerading as utility. The bidding war for the marginal barrel of Hormuz-linked crude is not where the utility lives — the utility lives in insurance, inventory, and optionality. The equity basket is the retail-facing expression of that utility, and it is the one expression the disclosure regime can actually see.

Which raises the question that the entire news cycle skipped.

Where the Edge Would Really Live

If an actor had genuine informational edge on the timing of strikes and ceasefires, the efficient expression would not be common stock. It would be the volatility surface.

A long straddle on Brent, or a call spread on front-month crude, or war-risk reinsurance exposure, or simply a well-timed position in the equity options chain of an integrated major — all of these express the same view with an order of magnitude more convexity per dollar of capital and a fraction of the disclosure footprint. Equity common stock in a $400 billion company is a blunt instrument for a timing trade. It is the instrument you use when you want the exposure to be legible, not when you want it to be large.

The disclosure format is blind to precisely the instrument class where the edge would be largest. Periodic transaction reports capture common equity, funds, and certain other holdings at the band level. They do not capture the options surface. They do not capture OTC structures. They do not capture the timing distribution within a band.

This is the core technical finding: the disclosure regime is high-resolution on the instrument where the edge is smallest and effectively zero-resolution on the instrument where the edge is largest. Any inference drawn from the visible trades is therefore an inference drawn from a biased sample — the sample selected for visibility, not for significance.

I have spent audit cycles on reward-distribution functions that looked clean in the ABI and were catastrophic in state transition. The lesson transfers. When the observable surface of a system is well-instrumented and the consequential surface is not, you do not conclude that nothing happened. You conclude that your monitoring is pointed at the wrong layer and go find the layer that matters.

For this story, the layer that matters is the one nobody publishes.

Why Prediction Markets Didn't Catch It

You would think this is a job for prediction markets. It is not, and the reasons are instructive.

Geopolitical markets — the Iran-strike type — exist on platforms like Polymarket and Kalshi. In principle they are exactly the right instrument: continuous, timestamped, sized, and public. An insider with edge should be forced to pay up to express it, and that pressure should be visible as price impact with a block height attached.

In practice, four things break it.

First, liquidity. Tail-event markets are shallow. Order books that clear a few tens of thousands of dollars cannot absorb a position large enough to matter to a nine-figure portfolio, and shallow books mean the informed trader's price impact is either negligible or self-defeating.

Second, resolution ambiguity. 'Strike on Iran' has to be defined precisely enough to settle. Whether a specific operation counts, whether a cyber action counts, whether a proxy action counts — these definitions are contestable, and contestable settlement turns a market into a poll. A poll does not have a term structure.

Third, jurisdictional friction. Access restrictions have historically limited who can participate, which limits how much informed flow reaches the book.

Fourth, and most important, there is no term structure across the decision horizon. The relevant question on March 2 was not 'will there be a strike' — the strike had happened. It was 'what is the conditional probability of escalation given the strike.' That is a multi-period, path-dependent object. Short-dated binary markets price a snapshot. Policy decisions are a sequence.

So the market that could have produced forensic evidence did not exist at the required depth. And the market that did exist — equities, futures, the options surface — produced a tape with a clock but no names attached, while the disclosure regime produced names attached to a tape with no clock.

We have a tape without identity and an identity without timestamps, and the two datasets are never co-resident in the same structure. That is the precise, mechanical reason this story cannot be resolved by journalism, by subpoena, or by public argument. It is not a cover-up. It is an architectural gap between two systems that were never designed to be joined.

The Basket Has One Factor

Now model the alternative explanation honestly, because a falsifiable analysis has to include the boring hypothesis.

Suppose the portfolio is managed by independent managers running a systematic or discretionary energy sleeve. What would that sleeve look like?

It would be factor-driven. Nine large-cap energy names — integrateds, refiners, pipelines — share a dominant loading on a single latent variable: the probability of a supply disruption out of the Gulf. Cross-sectional dispersion within that basket is driven by second-order factors — crack spreads, regulatory exposure, leverage. The first-order variance is one factor.

Which means buying eight names on the same day is not eight decisions. It is one decision expressed eight ways. The 'broad basket' framing that makes the activity look diversified is misleading in the opposite direction from what the headlines imply: it makes a single factor bet look like a portfolio.

Second, a systematic sleeve rebalances on a calendar or a volatility target, not on geopolitical anchors. Vol-targeted energy exposure, for instance, would have mechanically increased the position after the volatility spike that followed an escalation and decreased it as vol decayed. Some of the reported March activity is consistent with that. Not all of it. The April 7 exit, in particular, is a pre-resolution de-risking move, which is a discretionary signature more than a systematic one — though a vol-target overlay can produce the same footprint if realized vol is compressing.

Third, base rates. On March 2, thousands of independent accounts bought energy equities. On April 7, thousands sold. Selecting the one account that appears in a disclosure filing and then noting that its activity coincided with policy events is exactly the selection procedure that manufactures false positives. If you mine a large enough trade database for accounts whose activity correlates with a given news calendar, you will find them. You will find dozens.

I have run that experiment in a smaller domain. During the 2021 minting congestion, I looked at batched mint transactions against a naive baseline and found apparent 'optimal' minters everywhere — until I controlled for the fact that whichever contract landed in a favorable block got a favorable gas outcome. The pattern was real. The inference was not. The block producer assigned the outcome.

Same structure here. The policy calendar assigned the outcome. The account happened to be positioned. Whether the positioning preceded the assignment is the whole question, and it is the one question the data cannot answer.

The Null That Cannot Be Falsified

Here is the contrarian angle, and it is not the angle either side of the political aisle will find comfortable.

The interesting failure here is not that a public official may have traded on non-public information. The interesting failure is that the disclosure apparatus cannot falsify that hypothesis, and the public has been trained to treat its silence as exoneration.

'No evidence of directed trades' is a true statement about the instrument. It is not a statement about the subject. Running a full node that syncs once every forty-five days and then reporting that you observed no reorg does not mean reorgs do not happen. It means your observation window cannot see them. This is not cynicism — it is the mechanical consequence of resolution limits, and every engineer who has worked with censored data knows it in their hands.

Push harder, though, because the counter-argument is legitimate and I do not want to strawman it. Granularity has costs. Publishing exact position sizes and timestamps for a sitting official and their dependents creates a specific, measurable harm: a map for targeting, for kidnapping, for coercion, for private-sector retaliation. That harm is not hypothetical either. So the regime is trading one harm against another, and it has chosen the configuration that protects individuals and sacrifices forensic capability. That is a defensible design choice.

What is not defensible is then using the output of that design choice as if it were forensic evidence. You cannot optimize for privacy and then advertise the residue as integrity. Pick one objective function and be honest about the trade-off.

The structural fix is not more granular disclosure of the same shape. It is a different shape. The strongest version of accountability here is prospective and verifiable: publish the perimeter, not the positions. A sitting official with an energy-heavy portfolio could commit, publicly and in advance, to a defined restricted sector set for their tenure, implemented as a manager mandate with an auditable constraint. Then any trade inside the perimeter is either permitted and disclosed, or it is a policy violation detectable against a written rule rather than against a profit line.

That converts an unfalsifiable inference problem into a falsifiable compliance problem. Compliance is a solvable engineering problem. Inference about intent from range-bucketed timestamps is not, and never will be, no matter how many hours of cable news are pointed at it.

Contrast that with mechanisms that do work. Retroactive public-goods funding through mechanisms like Optimism's RetroPGF is effective precisely because it rewards outcomes that anyone can verify against a public record, rather than rewarding the judgment of a committee that meets in private. The mechanic is: verifiable output, public input, no discretion at the point of settlement. Every other grant committee I have watched allocate capital ran on soft signals — relationships, narrative, proximity to the decision-maker. It is the same failure mode as this disclosure regime, just at a smaller scale and with less money on the table.

What Changes When the Feed Gets Denser

So where does this go?

The direction of travel is compression. Multiple jurisdictions have moved toward machine-readable financial disclosure with finer bands and shorter reporting windows. The periodic-reporting window is already tighter than most casual readers assume — 30 to 45 days is a compression regime compared to the annual-only filing that preceded it. The next step is not philosophical. It is a schema change: transaction-level records with execution timestamps, published as structured data rather than as PDFs that someone has to hand-transcribe into a spreadsheet before they can be plotted against a news calendar.

When that happens, the analysis moves from inference to enumeration. You stop arguing about whether the geometry is suspicious and start querying whether the timestamps fall inside or outside the announcement window. The noise floor drops by an order of magnitude. You get what on-chain analysts have always had and off-chain analysts never did: a legible ordering of state transitions.

Until then, the correct posture is not skepticism or credulity. It is methodological discipline. State the null hypothesis. State the resolution of your instrument. Refuse to draw conclusions finer than your measurement granularity. February 27 through August 31 is a six-month window with four interesting anchors; the honest reading is that the anchors are consistent with several hypotheses and the data discriminates among none of them.

Which brings me back to the schema. Nine tickers, one factor, four dates, zero timestamps. The record breathes in ranges and forty-five-day windows; that is the entire problem. Code does not lie, but it often forgets to breathe — and this particular record has been holding its breath since February.

If the tape had block height, we would not be having this argument. We would be running a query. The distance between those two worlds is not a question of willpower or subpoena power. It is a question of schema design, and schema is something we can actually fix.