Sign in

How we calculate immigration court statistics

Every figure on the court, judge, nationality and backlog, representation, bond and postponement pages comes from one public file. This page says what we count, what we leave out, and where the numbers can mislead.

Data through · Source: EOIR case data · How we calculate this

Source

The Executive Office for Immigration Review (EOIR), the Justice Department agency that runs the immigration courts, publishes its case database each month in its FOIA Library. The release has one record for every proceeding, hearing and application for relief, with the court and the judge on each. It is a work of the U.S. government and is in the public domain.

We read the release directly from EOIR and compute every figure ourselves. The current pages use EOIR Case Data 2026-1001, with data through September 30, 2026, and cover 71 courts, 1,072 judges and 75 nationalities.

Updates and checks

EOIR publishes the release as one zip file at a fixed address and replaces it each month, early in the month and on no fixed day. We check that file for a new version every two hours during the first twelve days of the month and daily after that. When a new version is out, we load it, rebuild every page’s figures and publish them the same day, unless a health check fails.

  • The file is complete. Each table must hold exactly the number of rows EOIR’s own Count.txt says it exported, and must not shrink or jump in size against the previous release.
  • The format has not changed. No column may disappear, and the share of empty values in each column must stay close to the previous release.
  • The data moved forward. The data-through date must be later than the previous release’s and end within 45 days before EOIR posted the file.
  • Headline figures are plausible. National pending cases, cases closed, the share of cases with a lawyer, hearings and bond decisions must each stay within set bounds of the previous release. A large swing is far more likely a bad file than news.

A failed check publishes nothing, and the pages keep the last release that passed until we have looked at the file. EOIR keeps only its latest release online, so we archive every one we publish exactly as EOIR posted it. Any figure on these pages can be traced back to the file it came from.

Time periods

Fiscal year
The federal fiscal year, October 1 to September 30. FY 2026 began October 1, 2025.
FY 2022–2026
Asylum figures cover five fiscal years, from October 2021 through September 2026.
Last 12 months
Cases closed and their results, bond decisions and hearings cover October 2025 to September 2026. Pages lead with the asylum grant rate over the same 12 months, the current rate, and show the five-year rate beside it for context. Where fewer than 50 claims were decided in 12 months, the page leads with the five-year rate instead.
Last three years
Appeals cover Board rulings from October 2023 through September 2026.

Pending cases

A case is pending when its proceeding has no completion date. We count removal, deportation, exclusion, asylum-only and withholding-only proceedings, the case types EOIR uses for its own pending count. Credible-fear and reasonable-fear reviews, bond hearings and motions are not counted.

The backlog chart counts the cases that had been received and not yet completed on September 30 of each fiscal year. The age of a pending case is the time since the date EOIR recorded receiving the current proceeding. A proceeding with no receipt date, usually a charging document the court was never sent, is counted from the date on that document.

Cases are counted under the court that holds them now. When EOIR moves a docket from one court to another, the file shows those cases under the new court for their whole history, so a court that opened recently can show a backlog from before it existed.

Cases closed and results

A case is closed when the proceeding is completed at that court for a reason other than a change of venue or a transfer. A case that moves to another court is still open, so we count it at neither end. Results are shown for removal proceedings and grouped this way:

Removal ordered
The judge ordered removal. An order issued when the person did not appear is in absentia.
Terminated or dismissed
The judge terminated or dismissed the proceeding, or the government did not file the charging document and the case was closed for failure to prosecute.
Voluntary departure
The person was allowed to leave without a removal order.
Relief or protection granted
The judge granted relief such as asylum, cancellation of removal or adjustment of status, or ordered removal but granted withholding or protection under the Convention Against Torture.
Administratively closed
The case was taken off the active docket without a decision.
Other
Withdrawals and completions EOIR records under other codes.

Asylum grant rate

The grant rate is asylum claims granted, divided by claims granted or denied. It uses each asylum application EOIR records as granted or denied, placed in the fiscal year the proceeding was completed.

  • One decision per proceeding. If a proceeding holds more than one asylum application with a decision, it counts once, as a grant if any was granted.
  • Merits decisions only. Applications withdrawn, abandoned, not adjudicated or closed administratively are left out of the rate.
  • Asylum only. A denial of asylum counts as a denial even when the judge granted withholding of removal or protection under the Convention Against Torture in the same case.
  • No double counting across courts. When a case changes venue, EOIR’s file repeats the decision on the earlier proceeding. We count it only where the case was closed.
  • Two periods. The headline rate covers claims decided in the last 12 months; the five-year rate covers FY 2022–2026. Grant rates fell sharply in FY 2025 and FY 2026, so the two can differ widely.
  • Small numbers. We print a rate for 50 or more decisions in the period, and for 20 or more in a single year. Below that the page shows the counts.

Representation

A case counts as represented when a lawyer or accredited representative had filed a notice of appearance (Form E-28) for the person in immigration court by the date of the decision, or by September 30, 2026 for a pending case. Help from a lawyer who never filed an E-28 does not show in the data.

  • Pending cases. The share of cases pending on September 30, 2026 with a notice of appearance on file.
  • Asylum with and without a lawyer. The asylum grant rate, split by representation at the decision, for claims decided on the merits in FY 2022–2026.
  • In absentia. Of removal cases closed from October 2025 to September 2026, with and without a lawyer, how many ended in a removal order issued because the person did not appear.

The gap between represented and unrepresented cases is not all the lawyer’s doing. People with stronger claims find counsel more easily, and detained people are represented less often.

Bond decisions

Bond requests an immigration judge decided from October 2025 to September 2026, counted at the court and under the judge that heard them. A bond is set when the decision records a bond amount. Amounts, the median and the middle half (25th to 75th percentile) are over the decisions in which a bond was set.

EOIR does not publish what each of its bond decision codes means, so we report the share of bond requests decided in which the judge set a bond, and do not split the rest into denials and other results. Bond figures print from 25 decisions.

Hearings and postponements

Master and individual calendar hearings set for dates from October 2025 to September 2026, by the reason EOIR recorded for how each one ended. EOIR charges every reset to the respondent, to DHS or to the court itself. Hearings with no reason recorded yet, data-entry corrections and cases that never reached the court are left out.

Case decided
The case was decided at or before the hearing.
Moved on to a merits hearing
A master calendar hearing moved the case on to its individual hearing.
Moved earlier
The court advanced the hearing to an earlier date.
Reset for the respondent or for DHS
Reset for a reason EOIR charges to that party, such as time to find a lawyer or to prepare. The respondent’s resets include those their lawyer asked for.
Reset by the court
Reset for the court’s own reasons: a judge reassigned or on leave, docket management, a priority case, or no time left to finish.

Breakdowns print from 200 hearings. The live counts on the postponements page come from Lexfill’s Court Watch, not from EOIR’s release, and appear for a court only once enough cases are watched there.

Motions to pretermit

A motion to pretermit asks the judge to deny an asylum application without a merits hearing. We count records in EOIR’s court motions table with motion type PM filed by DHS, from June 2021, when the type was added. Filed is the date the court received the motion. The share granted is motions granted, divided by motions granted or denied in the last 12 months; withdrawn motions and other outcome codes are left out. A share prints from 50 rulings. See the study.

One nationality at one court

A nationality has a page at a court once judges there decided 300 or more of its asylum claims in FY 2022–2026. Judges are listed with 10 or more of those decisions.

Appeals

Rulings by the Board of Immigration Appeals from October 2023 through September 2026 on appeals of an immigration judge’s decision in the case, charged to the judge and court that decided it. A case is sent back when the Board remanded it or sustained the appeal. Withdrawals, closures and returns for paperwork are left out. Appeal figures print from 30 rulings.

Hearing calendar

The calendar figures count hearings dated after September 30, 2026 that have not been adjourned, for cases still pending, at the court where the case now sits. Master calendar hearings are short preliminary hearings; individual hearings are where the judge hears the merits. “Half are set before” is the median hearing date. A case can have more than one hearing on the calendar.

Judges

  • Attribution. A decision belongs to the judge recorded on the completed proceeding.
  • Names. EOIR identifies judges by code. Codes that carry the same name are combined. Placeholder codes, such as visiting-judge slots, are left out.
  • Who has a page. A judge has a page with 100 or more asylum decisions in FY 2022–2026, 200 or more cases closed in the last 12 months, or a live docket of 100 pending cases or 50 hearings on the calendar. The last two rules cover judges who are new to the bench or hear few asylum claims; their pages lead with caseload, and they are left out of the grant rate rankings.
  • New to the bench. A judge whose first hearing in EOIR’s data falls in the last 12 months is marked as new. Judge codes are sometimes reused, so a code first used by an earlier judge can hide a new one.
  • Pending cases assigned. The cases with no completion date that carry the judge’s code. Courts reassign cases, so this moves from one release to the next.
  • Court. A judge is listed under a court after closing 50 or more cases there in the last 12 months. The court named on a judge’s page is the one where the judge closed the most. Judges often hear cases for several courts, in person and by video.
  • Still sitting. The file does not say whether a judge is still on the bench. A page says so when a judge has no decisions in the last six months of data and no hearings on the calendar.

Nationalities

  • Whose nationality. The nationality EOIR records on the case. Codes EOIR uses when no country is recorded, and codes for countries that no longer exist, have no page.
  • Who has a page. A nationality has a page with 200 or more asylum decisions in FY 2022–2026.
  • By court. A court is listed under a nationality with 25 or more of its asylum claims decided there, and a rate is printed at 50.
  • Reading the rate. A grant rate by nationality mixes very different claims. It moves with who is in court, whether they are detained or have a lawyer, and which judges and courts hear them.

States, rankings and reports

  • States. A state page adds up the courts in that state. States with one court have no separate page. Median ages are not shown for a state because a median cannot be added up.
  • Change over a year. Pending cases on September 30, 2026 against the same date a year earlier, both counted from the current release. A percentage is left out for a court that had fewer than 1,000 cases then.
  • Rankings. The rankings order the same figures and add none. The judge ranking leaves out judges with no decisions in six months and no hearings on the calendar.
  • Monthly figures. Pending cases on the last day of each month, and cases closed during it, counted from the current release. Earlier months can differ slightly from what was published at the time.
  • Monthly reports. A monthly report is written once, from that month’s release, and is not revised afterwards.

Other published figures

EOIR’s own statistics can differ from ours. EOIR counts from its live database with its own definitions, and the public release can omit records it withholds. We report what the release contains.

TRAC, the Syracuse University research project, publishes judge reports from the same file. TRAC counts a case as denied only when no relief of any kind was granted, and covers a different span of years, so its denial rate is not one minus our grant rate. Counts of asylum grants for the same judge and years should match closely.

Some sites count every case that held an asylum application and did not end in a grant as a denial, including cases closed for abandonment, withdrawal, pretermission or a removal order with no merits decision. That denominator is several times ours and gives a much lower rate. Ours counts only applications a judge granted or denied.

Limits

  • A rate is not a prediction. Nationality, detention, representation and the kind of claim differ from one docket to the next and drive much of the difference between judges and between courts.
  • The file changes. EOIR staff correct and update records, so a figure for a past month can move slightly from one release to the next.
  • Recent months are thin. Decisions from the final weeks before a release can be entered late.
  • Many records carry no decision. A large share of asylum applications in the file end without a grant or a denial. They are outside the rate by design.
  • Figures can be wrong. EOIR’s file carries data-entry errors, and our calculations can contain mistakes. Check a figure against the court record before you rely on it, and tell us if one looks wrong.
  • Judges are matched by code. EOIR files each decision under a judge code, and we take the judge’s name from EOIR’s table of codes. Codes that carry the same name are combined. When EOIR reuses a code for another judge, or two judges share a name, one page can mix two judges’ records.

Release history

Every EOIR release these pages have been built from, newest first.

Data throughEOIR postedPublished hereProceedings
September 30, 2026October 1, 2026November 16, 202616,899,699

Reuse and corrections

Our figures, charts and downloadable files are free to reuse under the Creative Commons Attribution 4.0 license (CC BY 4.0). Credit “Lexfill (lexfill.ai), from EOIR immigration court case data” and link to the page you used. The Lexfill name and logo are not part of the license. Aggregate files are on the press and downloads page.

If a figure looks wrong, tell us which page and which number, and we will check it against the release.

Lexfill is not a law firm and is not affiliated with EOIR or the U.S. Department of Justice. Nothing here is legal advice.