Presidential self-pardon ban

The proposal

The idea is a single sentence added to Article II: the President may not grant a reprieve or pardon to himself or herself. It is the narrowest of the pardon-limiting proposals, and the only one whose legal answer may already be no. A 1974 Office of Legal Counsel memorandum concluded that “under the fundamental rule that no one may be a judge in his own case, the President cannot pardon himself.”¹ That opinion binds no court, none has been attempted, and no court has ruled; the Constitution Annotated treats the question as unresolved.² The variant in circulation is a bundled version, introduced in every Congress since 2017, that also bars pardons of relatives, administration members, and paid campaign staff.³ We rejected this category on 2026-08-06 on the polling alone, and reopened it on 2026-08-18 as a candidate clause of the Pardon Reform amendment.

The evidence

Two independent national surveys, both by the same pollster, have asked the same question about prohibiting a president from pardoning themself, twenty-one months apart. In May 2026, 72% supported it — 85% of Democrats, 71% of independents, 61% of Republicans.⁴

The August 2024 reading of the same question was 64% overall — 81% of Democrats, 67% of independents, 43% of Republicans.⁵ Republican support rose eighteen points between the waves and Democratic support four, most of it out of uncertainty rather than opposition: the Republican “not sure” share fell from 27% to 16%.⁴ The two fieldwork dates fall on either side of a change in which party holds the presidency, and the 2024 wave ran three months before a presidential election while federal charges were pending against one party’s nominee.⁵

The rest of that instrument places the clause inside the pardon package, on the same respondents: publishing pardon details including reasons drew 79% (79% of Republicans, 83% of Democrats), a family-member ban 64% (57% and 74%), and giving Congress power to reverse a pardon 48% (30% and 73%).⁶ The self-pardon ban is the second-strongest provision in that battery.

Older evidence measured the norm rather than the rule, and has been consistently strong; those framings and their numbers are set out below. Polls about a self-pardon by a named president measure a person rather than a rule and are excluded from the floor (no party breakdown published).⁷

Where it stands

The bipartisan floor is 61% of Republicans on the 2026 reading — Strong — against 43% on the identical wording in 2024 — Weak. The result is time- and context-dependent, and that is what holds the category open rather than closed.

Support may move with who holds the office. The two readings sit in opposite alignments, and the corpus records the same mechanism on an adjacent question: asked in early 2025 whether it is acceptable for a president to issue pardons to family members, 72% of Republicans and 42% of Democrats called it unacceptable, each party’s number falling on the provision its own president had just used.⁸ Regime coding would ordinarily predict Republican support for a limit on the presidency to fall, not rise, under a president of that party — a reason to test the 2026 reading again, not a reason to treat one reading as settled.

The gap between condemning the act and adopting the rule has closed. In 2024, 63% of Republicans called a self-pardon inappropriate while 43% backed the prohibition; in 2026, 64% called it inappropriate and 61% backed the prohibition.⁹¹⁰ The norm held steady while the remedy caught up to it.

High topline over a low floor — the test the category failed in 2024 — is cleared on the current reading. The 2026 topline sits eleven points above its floor, with a 24-point party gap; in 2024 it sat 21 points above its floor, with a 38-point gap.⁴⁵ Depth is untested in both years: no survey has put a cost to respondents — that the rule would bind a president of their own party, or that it would settle by text a question no court has answered.

The category passes the three non-polling tests, as it did at rejection: a president who may pardon himself is, as to federal crime, outside the law he executes; the text binds every future president of either party identically, and none has ever attempted a self-pardon, so no partisan abuse record attaches to it; and the Supreme Court has placed the pardon power among the President’s conclusive and preclusive powers, so settling the question definitively requires an amendment; whether a statute could declare a self-pardon ineffective is itself contested.¹¹¹² Part of the principle already lives in the Pardon Reform amendment, whose self-benefit provision voids a pardon granted to shield the President from legal liability; this page concerns the remaining case, in which the President is the named recipient.

Framings we tested

The framing that would be drafted — “prohibiting a president from pardoning themself” — now clears the bar, at 61% of Republicans and 85% of Democrats (YouGov, 2026),⁴ having read 43% of Republicans in 2024.⁵ The norm framing clears in every wave: “inappropriate for a president to pardon themself” reached 64% of Republicans and 88% of Democrats in 2026 and 63% and 88% in 2024,⁹¹⁰ and “unacceptable for presidents to pardon themselves if charged with a crime” reached 85% overall in 2018, with about 75% of Republicans and more than 90% of Democrats (AP-NORC, 2018; party split from the AP write-up).¹³ An advocacy-commissioned survey of battleground House districts found 78% disagreeing that “presidents should pardon themselves” — 59% of Republicans, 96% of Democrats, 76% of independents (YouGov for Protect Democracy, 2024), below the independent 2026 reading of the rule itself.¹⁴ The framing that still fails is the generic one: asked whether the Constitution should be amended to limit the president’s ability to grant pardons, 56% agreed — 34% of Republicans, 58% of independents, 76% of Democrats (Economist/YouGov, 2025), an instrument that contains no self-pardon item.¹⁵ The specific categorical rule outperforms the general one by 27 points among Republicans. Asked in 2026 whether the president has the power to self-pardon, 28% said yes, 36% no, and 36% were not sure (no party breakdown used here; the item is a knowledge question, not a preference).¹⁶

What would move it

In: one independent national poll asking the drafted sentence in forward-looking, both-parties wording — “This would apply to every future president, of either party, including presidents you vote for” — clearing 60% in both parties, and holding at 55% in both after a follow-up naming the costs on each side. It should be fielded in the same instrument as the article’s other provisions, so the item is measured against publication, naming, family, and sale within respondents.

Out: a Republican result below 45% on that wording, which would establish that the 2026 reading measured an alignment rather than a rule and return the category to rejection on the test it failed in 2024. A result of 45–59% among Republicans would leave it a candidate, at a floor comparable to the family ban the article already carries at 57% of Republicans.⁶

← All candidates