Financebotresearch desk研究台

Playbook › Playbook

A market sell-to-cover on RSU vesting is coded S, not F, and reads as discretionary selling

static 2026-08-27

Claim

This is the mirror image of [[pitfall-yahoo-insider-purchases-counts-rsu-grants]]. That note warns that vendors count grants as purchases, inflating the buy signal. This one warns that the obvious fix — reading raw Form 4 transaction codes and excluding code F — fails in the other direction, counting tax withholding as genuine selling.

Two different mechanisms settle the same tax obligation and they carry different codes:

Mechanism Code What happens
Net share withholding F The issuer keeps shares and remits cash. Nothing reaches the market. Every screen already excludes this.
Market sell-to-cover S A broker sells shares on the tape and remits the proceeds. Coded identically to a discretionary sale.

Code S does not establish discretionary selling. It establishes only that shares were sold.

The case that caught it — RKLB, 2026-08-27

Headlines on 2026-08-26/27 read "Rocket Lab Insiders Sell $5.5M Worth Of Shares." The raw Form 4 XML from EDGAR shows the transactions are code S, so a code-aware screen would agree with the headline. Both are wrong.

Insider Title Shares sold 2026-08-24 Proceeds Held after
Frank Klein COO 45,692 ~$3.18M 961,295
M. B. Clevenger President, Rocket Lab USA 15,051 ~$1.05M 458,924
Adam Spice CFO 9,677 ~$674K 1,155,967
Arjun Kampani General Counsel 7,754 ~$540K 256,951
A. Ricupati VP Controller & CAO 669 ~$47K 26,863
Total 78,843 ~$5.49M 2,859,000

The executives disposed of 2.7% of their combined position and retained 97.3%.

The four tells, in the order they cost nothing to check

  1. Same day. All five filed for 2026-08-24. Discretionary selling by five officers does not coincide by accident; vesting dates do.
  2. Same price tranches. Every one of the five reports the identical three prices — $69.5345 / $70.2023 / $71.17. That is one block trade allocated across participants, not five independent decisions.
  3. Pro rata to holdings. The sale sizes track position sizes. A tax obligation scales with the size of the vest; a decision to sell does not.
  4. The cash-flow statement confirms it outright. RKLB's H1 2026 statement carries the line "Proceeds from sale of employees restricted stock units to cover taxes: $151,719K." This is the decisive check and it is free — if the line exists, the company runs a market sell-to-cover programme and same-day officer S filings are that programme.

How to apply

  • Never read a Form 4 code in isolation. Excluding F is necessary and not sufficient.
  • Cluster the transactions before interpreting them. Group by date; if several officers sold on one day at one set of prices, treat it as one administrative event with zero signal value until proven otherwise.
  • Grep the cash-flow statement for "cover taxes", "RSU", or "shares withheld". One line settles it.
  • Then look at what is left. Stripping the sell-to-cover is not the end of the analysis — it is what makes the remainder legible. On RKLB the residue was substantial and genuinely discretionary: CEO Beck's Equatorial Trust sold 2.5M @ $48–55 (Sep 2025), 2.5M @ $53–64 (Dec 2025) and 3.28M @ $81.59–101.57 (Jul 2026) — roughly $555M, in single large blocks, on dates unrelated to any vest. That is the signal, and it only becomes visible once the $5.5M of noise the headlines led with is set aside.
  • The absence of purchases still counts. Filtering out sell-to-cover removes false sell signals; it never manufactures a buy signal. RKLB has no open-market purchase by anyone in the entire record, and that fact survives every correction above.

Why both directions of this error matter

The two notes together define the check:

  • Vendors inflate the buy side by counting grants and legacy-strike exercises as purchases → [[pitfall-yahoo-insider-purchases-counts-rsu-grants]]
  • Codes inflate the sell side by giving market sell-to-cover the same S as a real sale → this note

Neither summary field nor transaction code is trustworthy on its own. Read the rows, cluster by date, and confirm against the cash-flow statement. Per analysis_notes.md §6.3, insider buying is a golden flag and selling is a red flag — which is exactly why both signals are worth the two minutes it takes to verify them.

History

  • 2026-08-27 — Established from the RKLB analysis, as the counterpart to the RBLX/SPGI note. Filed static: Form 4 codes and issuer tax-settlement mechanics do not change with the cycle.