How it's counted

The decisions

The group is every written federal district court decision Passim holds in a suit asking a court to make an immigration agency act: mandamus petitions and claims that an agency unlawfully withheld or unreasonably delayed action under the APA. Rules over each decision's text place it in the group, and model readings check them. Habeas petitions and suits that only challenge a denial are outside the group.

Each decision is read for the agency the suit is about, the forms or filings at issue, what the suit asked the court to make the agency do, the motion decided, the disposition, the prevailing party, how long the application had been pending, the court's findings on the TRAC factors, and its rulings on threshold questions (section 1252(a)(2)(B), consular nonreviewability, a duty to re-adjudicate a 221(g) refusal, mootness). Every value carries the passage it was read from and how it was read: by rule, from fixed text patterns, or by a model whose answer was checked against that passage. Most decisions were read by two models, with a third where they disagreed; a decision not yet read that way shows its rule values, and a value no reading supports shows as not recorded.

Disposition is the ruling on the principal motion: granted, granted in part, denied, dismissed, dismissed in part (a motion to dismiss granted in part, so some claims go on), dismissed as moot, or remanded to the agency. A motion to dismiss granted in full counts as dismissed. Two kinds of order are listed apart and counted as neither side's win: a magistrate judge's recommendation, which the district judge decides whether to adopt, and an order transferring the case to another court, which decides where the suit is heard. Because a granted motion can be the government's, read the disposition with the prevailing party: the applicant, the government, each side in part, or neither (a dismissal as moot, a consent order, a recommendation or a transfer). A delay claim that outlives the government's motion can read denied, or dismissed in part when another theory fell, so the list's delay claim filter selects the delay claim survived the government's motion whichever of these a decision shows.

Months pending runs from the date the application was filed with the agency to the date the suit was filed (at filing) and to the date of the decision (at decision), as the decision states those dates; where it gives no filing date, the court's own figure for the delay is used for the decision-date measure. Where the decision holds that the agency acted (a dismissal as moot, or a claim held moot), only the months at filing are shown: the months to the decision ran on after the agency acted and are not a delay, and the counts of months at decision leave those decisions out. The readings do not date the agency's action, so a decision that misses an earlier denial can still overstate the months at filing. In a section 1447(b) naturalization suit the months run from the application; the statute's 120 days run from the interview, which the readings do not date.

Cases

One suit can produce several written decisions: a motion to dismiss denied, then summary judgment for the government months later. Two decisions are in the same case when the docket record links them to the same docket, or when they carry the same docket number in the same court and their captions name the same plaintiff (so the suits consolidated under one lead docket stay apart). Every count of outcomes (disposition, prevailing party, what the suit asked for, agency, form, months pending at filing) counts each case once, by its latest decision in what is being counted: the newest decision with a recorded disposition, so an early win reversed by a later ruling counts as the later ruling. A year's page counts a case by its latest decision that year. A count needs at least 5 cases behind it.

Every decision stays listed: the decision list shows each one (or, with "Each case once", each case by its latest decision), a judge's page lists each case's other orders under its latest decision, each decision's page lists the other orders in its case, and the spreadsheets hold every decision. Counts of decisions themselves (how many weigh the TRAC factors, decisions by year, months pending when the court decided, threshold rulings) count each decision.

The dockets

Most of these suits never produce a written decision: the agency acts and the parties dismiss. The docket record holds the mandamus, APA and section 1447(b) naturalization suits in the federal courts' Integrated Database since 2015, with newer filings and endings from CourtListener. Each suit that has ended is put in one class:

• Ended once the agency acted

  Voluntary or stipulated dismissal, or moot because the agency acted: how suits end once the agency acts (the docket rarely says why).

• Dismissed on jurisdiction or the merits

  Dismissed for want of jurisdiction, on the trac factors, or judgment for the government.

• Dismissed, ground not recorded

  Court dismissal with no recorded ground (idb 'other' or want of prosecution).

• Judgment for the plaintiff

  Consent, default or other judgment for the plaintiff.

• Remanded to the agency

  Remanded to the agency (1447(b) or idb remand to a u.s. agency).

• Transferred

  Transferred to another district.

• Still pending

  No termination recorded as of meta.freshness.

• Ended, outcome not recorded

  Terminated; no disposition the record explains.

"Ended once the agency acted" counts voluntary and stipulated dismissals and dismissals as moot. It is the usual sign that the agency decided, not proof of it: the docket rarely says why a plaintiff dismissed, and it does not say whether the agency approved or denied. "Dismissed, ground not recorded" is a court's dismissal for which neither the federal courts' database nor the docket gives a reason; it can be a dismissal as moot after the agency acted, a dismissal for want of prosecution, or a loss. The public docket list leaves out party names; the docket number opens the docket on CourtListener.

The Passim connector's docket tool reports the finer outcome codes these classes are made of. Its agency_acted_moot share counts only the suits the record says ended as moot because the agency acted, which is rare; its adjudicated_or_mooted share adds the voluntary and stipulated dismissals and is "Ended once the agency acted" here. Its dismissed_other is "Dismissed, ground not recorded" here.

The connector's code	Counted here as	What the record shows
agency_acted_moot	Ended once the agency acted	The record says the case ended as moot because the agency acted
voluntary_dismissal	Ended once the agency acted	The plaintiff's notice of dismissal
stipulated_dismissal	Ended once the agency acted	A dismissal both sides signed
dismissed_jurisdiction	Dismissed on jurisdiction or the merits	Dismissed for want of jurisdiction
dismissed_trac_merits	Dismissed on jurisdiction or the merits	Dismissed on the TRAC factors
decided_for_government	Dismissed on jurisdiction or the merits	Judgment for the government
dismissed_other	Dismissed, ground not recorded	Dismissed with no ground recorded (the federal courts' "dismissed: other" or want of prosecution)
granted_relief	Judgment for the plaintiff	Relief granted, or a consent or default judgment
remanded_1447b	Remanded to the agency	Remanded to USCIS under section 1447(b)
transferred	Transferred	Transferred to another district
pending	Still pending	No ending recorded
unknown	Ended, outcome not recorded	Ended with no disposition the record explains

The docket record seldom names the form or filing a suit is about. A suit's forms are the forms its docket names and the forms its written decisions name, so most suits with a form are suits that reached a written decision, and their endings lean toward the contested ones. These pages never count suits by form for that reason: Check a delay counts a court's suits against the form's agency, and a court's page counts every suit. The connector's docket tool does take a form; read its count of suits with a written decision before quoting what a form filter returns.

Where a suit ended by a docket entry, "Docket text" shows that entry, and any remand order or consent judgment, as the docket record holds it, with the suit's own party names replaced. From its words alone, fixed rules read the agency's deadline, what the plaintiff may do if the agency does not act, what follows if the plaintiff does not return in time, and whether the court kept the case; each is shown beside the words it was read from, and a term the entry does not state is left out.

The counts

Every number on these pages counts past decisions and dockets. None is a prediction of how a case will be decided, and none is a rate at which a judge or court rules for one side: decisions differ in their facts, their posture and what the parties asked for, and the suits that reach a written decision are the contested minority.

A distribution (counts by outcome, the middle and quartiles of a delay or of days to the end) appears only when at least 5 decisions or 5 dockets are behind it. Smaller groups are listed, not summarized. Every count links to the decisions or dockets it counts.

A judge's decisions are the decisions the judge signed; a judge's dockets are the suits assigned to the judge in the docket record or decided by the judge in a linked decision. The docket record names the assigned judge for only part of the suits, so a judge's docket counts are incomplete.

Supreme Court and circuit authority

District decisions bind no one. Each decision page lists the Supreme Court and court of appeals decisions on the questions the decision raises: the TRAC test for unreasonable delay, the reach of section 1252(a)(2)(B), consular nonreviewability and 221(g), and section 1447(b). The Supreme Court's decisions and the circuit's published decisions bind the decision's court; a court of appeals disposition marked nonprecedential was not published as precedent and is persuasive only. Each entry was verified in the corpus when it was added, and its ImmCite treatment status is read again every night.

Which total to quote

The line at the top of every page counts the whole group and names the group's newest decision; a court's, a year's or a judge's page gives its own count of decisions and cases, and its own newest decision, under "How the decisions came out". A count of written decisions counts every order; a count of cases counts each suit once, by its latest decision, so a judge can have more decisions than cases.

These pages are rebuilt from one snapshot, so every page counts the same decisions and suits. The Passim connector counts the live collection, which changes during the day as decisions are added, corrected or retired, so its totals can differ from these by that day's changes; the group is the same (every decision Passim holds in a mandamus or agency-delay suit), and decisions not yet read for these pages are in the connector's count and not here. The connector's outcome counts count decisions, each order separately; these pages count cases. Suit counts also differ by what they cover: a court's page counts every suit in its docket record, filed in any year, against any immigration agency; Check a delay counts the court's suits of the last three years whose lead defendant is the form's agency (USCIS, DHS or the Department of Justice for a USCIS form).

Coverage and freshness

These pages are checked for new decisions, readings and docket updates four times a day and rebuilt when any have come in. This rebuild finished November 3, 2026 at 1:24 a.m. CT; the decision readings were last updated November 2, 2026 at 5:03 p.m. CT and the docket record November 2, 2026 at 4:51 a.m. CT.

Passim does not hold every public decision. Every week, and in occasional fuller runs, Passim searches CourtListener and GovInfo for district decisions since January 1, 2025 that mention mandamus or unreasonable delay with USCIS, the State Department, a consulate or the National Visa Center, and checks which it holds. The latest check, on November 3, 2026, found 1,606 and Passim held 1,371 (85%). Some of what that search finds are not mandamus decisions, so the share of real mandamus decisions held is somewhat higher than that, and missing decisions are added as they are found.

If you know of a decision that is missing or read wrong, tell us.

Spreadsheets, data files and the search API

Every list on these pages can be downloaded as it is filtered, and ready-made spreadsheets hold every decision by court, year, agency, form and judge, and every court's suits. Each row links to the decision's page here and to the full decision. The data page lists the JSON files the pages are computed from, their fields, and the search API's parameters and answer.