Skip to content
Project Curia

Watching the government

Four ways to watch the Administration

Tracking Presidential Pardons.

When a president grants clemency, we read the record and say who was pardoned, what was forgiven, and how the grant moved: whether it went through the Justice Department’s own process, how fast, and what the pardoned had given to the pardoning side. Every claim here is tied to federal files, measured the same way for every administration since 1989, where itemized donor records begin.

First, the word

Clemency is the umbrella

The Constitution gives the president one power that, outside cases of impeachment, no one can appeal, veto or review: mercy for federal offenses. Clemency is the umbrella word for every use of that power, and almost all of it takes one of two forms.

Pardon

Forgiveness, after the fact

A pardon forgives a federal conviction and restores the civil rights it took, like sitting on a jury or owning a firearm.

It does not erase the conviction, does not overturn the verdict, and, in the Justice Department’s own words, does not signify innocence. The department’s rules ask petitioners to wait five years after release, or after conviction where no prison term was imposed, before even applying.

Commutation

A shorter sentence, nothing more

A commutation cuts short a sentence still being served: prison time, or a fine. The person walks out sooner, or owes less.

The conviction stands and no rights come back. Someone released by commutation is still, on paper, a convicted person; that is why most commutations go to people serving long sentences.

A president can also forgive an entire class of people in one signed proclamation (the Vietnam draft cases, simple marijuana possession, January 6), remit a fine, or delay a punishment with a reprieve. The power reaches federal offenses only: no president can touch a state conviction, which is why a person can be pardoned and imprisoned at the same time.

The record so far

Every grant on the record.

7,289
named clemency grants since 1989
54,137
named petitions refused since 1989
106
recipients with a corroborated donation to the pardoning side
$2.8 million
given to pardoning sides before those pardons
Why the page starts at 1989

The page begins at Bush 41 because 1989 is where itemized federal donor records and the published denial lists both begin; the 2,341 earlier named grants stay in the reserve, counted in About rather than drawn here.

Scope

A mass day is one on which an administration commuted more than 50 sentences at once: one presidential act covering a list, mostly Obama's clemency initiative and Biden's final-year batches. Setting those 5,710 grants aside compares individually signed clemency like for like; pardons are never set aside. The record band above and the full record below always count every grant.

The money

Recipients with a corroborated donation to the pardoning side

Checked recipients whose donation record to the president’s own committees, party, inaugural fund or aligned super PAC survives timing, offense and location corroboration: the same search, run against each administration’s own committees. None of it is a verdict on any person, and the search surface itself is uneven across eras, a limit the method tuck below spells out.

Measure
  • Bush 41
    1.3%1 of 77 checked
  • Clinton
    0.9%4 of 460 checked
  • Bush 43
    2.5%5 of 200 checked
  • Obama
    2.0%7 of 346 checked
  • Trump I
    17.5%29 of 166 checked
  • Biden
    2.8%5 of 177 checked
  • Trump II
    7.9%10 of 127 checked

Floors, not totals: gifts under $200 are never itemized, and checked means the recipient could be searched against itemized records at all. Dollar figures are conservatively summed: a transfer routed through a joint fundraising committee counts once, one person is one unit, and money after a pardon is computed separately and never added here. The axis runs from 0 to a printed maximum, never to 100.

The control group, the proclamation cohort, and the method

The rates in this table are NOT the corroborated rates above: they are bare, uncorroborated name-match rates, and they run far higher because both populations carry people who merely share a donor’s name. That noise is roughly symmetric, which is the point: running the same match on the people granted clemency and on the people who petitioned and were denied, the DIFFERENCE between the two rates is an institutional signal net of the noise. It validates the pattern; the corroborated tier above is a separate, stricter per-person gate. The interval beside each difference is its honest width, with a spanned zero said out loud.

AdministrationGranted, name matchDenied, name matchDifference95% interval
Bush 41R11.7%8.0%+3.6-3.7 to 11.0 (includes zero)
ClintonD16.3%10.0%+6.32.7 to 9.8
Bush 43R33.5%17.3%+16.29.6 to 22.9
ObamaD36.4%26.0%+10.47.8 to 13.0
Trump IR45.6%25.6%+20.111.0 to 29.1
BidenD34.4%32.5%+1.9-0.2 to 4.1 (includes zero)
Trump IIR39.4%29.9%+9.50.9 to 18.1

The 91 people named in the January 6 commutations and the 2020 election pardons are political professionals, donors by occupation, and are never pooled into the list above; measured apart, 23 of 91 carry a corroborated record, on $111,461 given before the grant.

The registry is the president’s own side: campaign committees, the party, and, where they exist, joint fundraising committees, the inaugural committee and one labeled aligned super PAC. The same query runs for every administration, but the searchable surface is not equal across eras and the gap runs one way: aligned super PACs exist only from 2010 forward, Bush 41 and Clinton have no findable inaugural or joint fundraising entity, and state party and nonprofit routes are not built for anyone. An older administration’s rate is therefore a deeper floor than a recent one’s. A surname must match whole; a match is promoted only with a gift before the pardon, a plausible offense profile and a location signal, either the district of conviction or a distinctive surname whose gifts cluster in one place. The Trump-Vance 2025 inaugural's $247.7 million from 951 disclosed donations was recovered from the committee's own filings after the FEC's tables indexed five rows of it.

Computed 2026-09-08 from federal clemency and contribution records.

The process

Regular order, and the distance from it

The standard route to a pardon runs through the Justice Department: an application, a review, and a five year wait after release. No president is required to use it. Each dot is the share of a president’s clemency that took the route, on one true 0 to 100 scale.

Measure

The share of each president's classifiable, dated pardons that pass both tests at once: filed with the department, and past the five year wait. This is what a by the book pardon looks like in the record.

Bush 41
96.9%63 of 65 classifiable
Clinton
78.7%295 of 375 classifiable
Bush 43
94.4%170 of 180 classifiable
Obama
98.5%200 of 203 classifiable
Trump I
41.2%54 of 131 classifiable
Biden
92.4%61 of 66 classifiable
Trump II
14.6%14 of 96 classifiable
How a pardon is supposed to work, and what these measures read

The standard route runs through the Justice Department. A person applies to the Office of the Pardon Attorney; the application is docketed, which is the department’s word for filed; the department’s own rule asks pardon seekers to wait five years after release, or after conviction where no prison term was imposed, before applying at all; the file is reviewed and a recommendation goes to the White House. None of that is in the Constitution. A president can sign a pardon for anyone, any time, with no application on file, and skipping the route is entirely legal. It is also measurable, and that is what this section does.

Filed through the department asks: did this person’s pardon start as a docketed application with the Pardon Attorney? Docketed means filed, never that the department recommended the grant; whether it recommended any individual case is not public, and no number here claims it. The docket match is by name and recovers 94 to 99 percent of the department’s own published applied-through counts, so every docket share here is a floor.

Pardons inside five years asks: how many pardons came faster than the waiting rule contemplates? This page measures from sentencing. For recipients who served prison time the rule’s clock starts later, at release, so their flag is a floor; for the smaller set sentenced to probation or a fine alone, the clock runs from conviction instead, and their rows say what the printed sentence was.

Textbook grants are the ones that pass both tests at once: filed with the department and past the five year mark. This is what a by the book pardon looks like in the record, and most pardons that can be classified, under most presidents, look exactly like it; that is what makes the exceptions measurable.

Two different things drive a low share: mass initiative cohorts were categorical by design, while individually signed grants with no docketed petition reached the president another way. Coverage varies by era because the department printed sentencing dates unevenly, which is why every row prints its own denominator.

The pace

Every grant, by day in office

Cumulative clemency over each presidency on one shared term clock, the same way the executive orders page draws the pace of the pen.

0200400InaugurationYear 1Year 2Year 3Year 4Year 5Year 6Year 7Year 8Bush 41 77Clinton 461Bush 43 200Obama 346Trump I 174Biden 182Trump II 139

Each line is one administration’s individually signed grants, cumulative from its own inauguration; an eight year presidency spans the whole width, a term still in office stops where the record does, and the vertical cliffs are single signing days. Mass commutation days are set aside under the scope above.

What was forgiven

The same crime, administration by administration

Each cell is one offense family in one administration’s own grants; the tint tracks the shown measure on one scale for the whole grid.

Measure
Bush 41
77 grants
Clinton
461 grants
Bush 43
200 grants
Obama
346 grants
Trump I
174 grants
Biden
182 grants
Trump II
139 grants
White collar
27%
36%
28%
19%
48%
16%
33%
Drug
9%
22%
20%
62%
16%
54%
15%
Tax
8%
8%
7%
2%
9%
2%
7%
Firearms
6%
6%
7%
8%
8%
13%
4%
Military
5%
1%
1%
1%
0%
1%
0%
Public corruption
3%
3%
0%
0%
8%
4%
8%
Violent
3%
3%
3%
1%
7%
21%
1%
Immigration
1%
1%
1%
0%
0%
1%
0%

Families are read from the offense text the Justice Department printed and they overlap: one case can be both a drug case and a firearms case, and every family it touches is counted. Grants whose pages publish no offense text match no family and are absent from every row, which is a fact about how the record was kept, never a count of zero.

The record, one grant at a time

Every named grant on the federal record

7,289 clemency grants as the Justice Department published them, quoted from its own pages and linked back to them. Each row carries what the record shows about its own grant: the offense, the sentence, the process, and any corroborated donation.

Loading the clemency record.

About this data

Sources and honest limits

This page reads three federal sources directly: the Justice Department’s Office of the Pardon Attorney, which publishes every grant, every named denial, and the docket of every petition filed; the Federal Election Commission, whose itemized contribution records supply the donor side; and the Senate’s Lobbying Disclosure Act filings, which report who paid a registered lobbyist to seek a pardon. Nothing here is drawn from a secondary account of any of these records.

The limits that shape every number here

The page starts at 1989 because the donor record does. Practical itemized contribution coverage begins in 1989: the Reagan era was searched and returns no corroborated matches, the three administrations before it cannot be searched at all, and their DOJ pages publish no offense text, no sentence text and no denial lists to compare. The 2,341 named grants from the Nixon, Ford, Carter, Reagan administrations stay in the reserve rather than drawn as rows of absence.

Contributions under $200 are not itemized. A gift below that floor leaves no record to find, so a person with no match may simply have given quietly, or given nothing. The two cannot be told apart, and absence is never read as innocence.

Only the corroborated tier is ever shown beside a name. A bare name match is routinely a different person with the same name; the medium and low confidence matches are a review queue that is never printed against anyone.

Docketed is not recommended. The petition-track figures read the Pardon Attorney’s own public docket: docketed means a petition was FILED through the department’s process. Whether the department recommended any individual grant is not public, and no number here claims it.

The five-year figures are floors, on uneven coverage. The waiting rule’s clock runs from release, or from conviction where no prison term was imposed; this page measures from sentencing, so a flagged pardon of someone who served time was inside the window by any reading, while a probation-only row carries its printed sentence for the reader to weigh. Sentencing dates parse unevenly by era and every rate prints its own denominator.

A case group is a set of candidates, not a verified docket. Grants made the same day, in the same district, for the same offense are grouped as one likely case; confirming co-defendants requires the court record itself, which this page does not yet read.

Trump I’s control comparison carries the widest interval on the page. The Justice Department published only 180 named denials for that term, all from a single day, against a full term of named grants; a shorter denial list widens every estimate built against it.

How the record itself was read

The offense families and the state of conviction are derived from the free text the Justice Department printed, with the same parsers the reserve itself uses. Offense families overlap and a grant can match several. The state is the venue where a case was tried, not where the person lived: 6,949 of the 7,289 grants resolve to one of 51 jurisdictions, the states and the District of Columbia, and the remaining 340 name a district this record cannot place, most of them military venues.

Absence of a match is never innocence, and a match is never guilt. A person with no recorded gift may have given under the $200 itemization floor, given to a different pardoning side entirely, or simply not given at all; none of those can be told apart from this record alone.

Clemency record computed 2026-09-07, severity profile 2026-09-07, institutional profile 2026-09-07, from federal clemency and contribution records. Internal, dark, pending review.