US immigration courts have no caseload-linked workforce plan — GAO asked for one in 2017 and reported on 2023-10-18 that none had been developed, and whether immigration judges may be removed without cause was still awaiting appellate argument as of 2026-07-27

What is happening?

The United States has no published rule that converts the number of pending immigration cases into a number of immigration judgeships. The Government Accountability Office recommended in 2017 that the Executive Office for Immigration Review build a strategic workforce plan tied to caseload. Its follow-up report GAO-24-107046, issued 2023-10-18, recorded that the review office had not developed a strategic workforce plan and had not set workforce planning goals, and that 8 of the 17 recommendations across the two reports remained open at that point.

The caseload the plan would have to size against is large. One opened source counts 3,195,137 cases pending before the immigration courts at the end of June 2026, of which 2,310,698 carry an asylum application that has been filed and is awaiting decision.

The Department of Justice puts a different movement on the record. It announced on 2025-09-04 that the immigration courts had completed more than 722,000 cases in the first 11 months of FY2025 and that the pending total had fallen by more than 447,000 since 2025-01-20, from above 4.18 million to below 3.75 million, which it described as the steepest decline in the history of the review office.

The second gap is about the seat rather than the number of seats. Immigration judges are employees of the Department of Justice. Two of them were removed in February 2025 and challenged the removals before the Merit Systems Protection Board, which held in docket MSPB No. CF-0752-26-0069-I-1 that immigration judges are inferior officers under Article II of the Constitution and may be removed by the Attorney General without cause. Section 7513 of Title 5 of the United States Code provides that a covered employee may be removed only for cause and only after 30 days of written notice. The two judges appealed on 2026-03-24, the United States Court of Appeals for the Federal Circuit granted rehearing en banc, and as of 2026-07-27 the case, docketed as No. 26-1575, was awaiting oral argument.

Between those two gaps the bench moved sharply in both directions. One opened source counts the corps of immigration judges together with assistant chief immigration judges falling from 726 to 553 between early 2025 and February 2026, a decline of about 24 percent. On 2026-05-21 the department announced that 77 immigration judges and 5 temporary immigration judges had been sworn in, described the class as the largest in the history of the review office, and put the size of the corps at approaching 700.

Whose problem is this?

The body that would have to build the staffing plan is the same body that appoints and removes the judges the plan would count, and no outside body sets either number.

Where does this problem end?

The boundary matters because the number of judges and the fairness of any single decision are argued in the same breath, and only the first of those two is measured here.

What is the state now, and what should it be?

Now

Needs a new measurementthe numeric target: no source opened here names a target number of immigration judgeships, a target caseload per judge, a target pending total, or a date by which any of them should be reached. The follow-up report of 2023-10-18 names the end state as a strategic workforce plan with workforce planning goals and stops there, because the goals are what the plan was supposed to produce. No number is supplied here from any other quarter to stand in for one the government has not set.

How big is it?

3,195,137 cases were pending before the immigration courts at the end of June 2026. Of those, 2,310,698 carried an asylum application that had been filed and was awaiting decision. Both figures come from the same opened page and share the same closing date.

The unit is cases and not people. A family matter is split into separate cases, so one person can hold more than one, and the count therefore bounds the number of distinct people from above rather than equalling it. No source opened here publishes a person count or a ratio that would convert one into the other.

This total is not the only total in circulation. The department reported on 2025-09-04 that the pending total had fallen by more than 447,000 since 2025-01-20, from above 4.18 million to below 3.75 million. Totals in circulation run between about 3.2 million and 3.9 million depending on which source and which closing date is taken, and no source opened here states the counting basis behind the difference.

The pending count measures how many matters are waiting on a bench whose size nobody derives from them. It does not measure how long any one of them waits, and no opened source converts this total into a waiting time for this closing date.

Under what conditions does it arise?

1. Nothing in law or in a published plan converts caseload into judgeships. The request to build such a plan came from the audit office rather than from a statute, and an audit recommendation carries no deadline. Eight of the 17 recommendations across the two reports were still open when the follow-up was issued on 2023-10-18.

2. The number can move fast in both directions. Between early 2025 and February 2026 the corps fell from 726 to 553 on one counting basis, with 12 courts losing more than half of their judges and two left with no full-time judge. New hiring in the same window fell to 46 in calendar 2024 and 14 through the third quarter of FY2025, against a peak of 133 in 2023, and then on 2026-05-21 the department swore in 77 immigration judges and 5 temporary immigration judges at once.

3. The seat has no settled tenure rule while it is being emptied and filled. The Merit Systems Protection Board held in MSPB No. CF-0752-26-0069-I-1 that the Attorney General may remove immigration judges without cause, section 7513 of Title 5 provides removal only for cause and after 30 days of written notice, and the question of which governs sits with the Federal Circuit en banc, which had not heard argument as of 2026-07-27.

4. There is no agreed denominator. Opened sources count the bench on different bases and give different pending totals for different closing dates, so a ratio of cases to judges can be built several ways and none of them is official.

5. Most people in front of the court arrive without a lawyer. Representation in immigration court is not appointed, and one opened source reports the representation rate falling from 65 percent at the close of FY2019 to 30 percent as of 2023-12 while the pending total rose from 1,023,767 to 3,287,058, with represented cases growing about 1.5 times over that span and unrepresented cases about 6.3 times.

What has been tried?

The corps was reduced and then enlarged inside the same period, and neither movement followed from a published rule about how many judgeships there should be.

What was found?

Why is it still unsolved?

Institutional gap — the number of judgeships is not derived from anything published, and the rule that would protect a judgeship is not settled.

A recommendation is not a requirement. The audit office asked in 2017 for a workforce plan tied to caseload, and the follow-up report of 2023-10-18 recorded that no such plan had been developed and that eight of the seventeen recommendations across the two reports were still open. Nothing in that sequence carries a deadline, a penalty or a court order, so a plan that is never written looks exactly like a plan that is still being written.

The second half of the gap is that the seat has no settled rule. The Merit Systems Protection Board held in docket MSPB No. CF-0752-26-0069-I-1, in the decision the two judges appealed on 2026-03-24, that immigration judges are inferior officers under Article II and may be removed by the Attorney General without cause. Section 7513 of Title 5 provides that a covered employee may be removed only for cause and only after 30 days of written notice, and the question of which of those governs sits with the Federal Circuit en banc, which had not heard argument as of 2026-07-27.

The third part is that there is no agreed denominator. Opened sources count the bench on different bases and give different pending totals depending on the publisher and the closing date, so a ratio of cases to judges can be assembled in several ways and none of them is the official one. A plan that has never been written also carries no counting rule, which means the input such a plan would need does not exist in a single agreed form either.

The fourth part is that the structural remedy has been on the table since 2022 and has not been taken. A bill to establish the immigration courts under Article I was reported out of a House committee in May 2022, lapsed with that Congress, and was reintroduced on 2026-03-05 as H.R.7836, where it remained at committee referral. Moving a court system out of a department is the kind of change that has to be voted, and it has not been.

What observation would mean it is solved?

Candidates — (a) the department publishes a workforce plan that states how the number of immigration judgeships is derived from caseload, with a target and a date (b) the removal question is settled, by a final appellate decision or by legislation (c) the pending caseload and the size of the bench hold a stable published relation across several years.

(a) alone is thinner than it looks. A plan can be published and leave the number exactly where it was, because nothing obliges the department to staff to its own plan and no statute attaches a consequence to missing it. The 2017 recommendation is the measure of how long a document can stay unwritten with nothing following from that.

(b) settles the rule and not the number. A decision either way tells everyone what protects the seat, and it does not create a formula for how many seats there should be. The same question can also be reopened by legislation under a different instrument, in which case the answer is settled twice and the staffing rule still does not exist.

(c) alone counts a ratio nobody agrees on. The pending total differs by source and closing date, the bench is counted on at least three different bases across the opened sources, and one of those bases may or may not include temporary detailees. A ratio built on two contested numbers can move without either underlying quantity moving at all.

The three have to be read together, and the second of them runs on a clock that neither the department nor the reader controls.

What is it connected to?

Fills with researchthe appellate backlog at the Board of Immigration Appeals, detention capacity and its effect on case scheduling, the supply of immigration attorneys and the absence of appointed counsel, federal personnel law after recent decisions on removal protection for officers, and adjudicator staffing formulas in other federal benefit and appeal systems. Relation type and evidence grade were not confirmed in this round.

What these sources do not say

• How many immigration judges there are. The opened sources count on different bases. One gives 553 for February 2026, counting immigration judges together with assistant chief immigration judges. One gives 685 for the third quarter of 2025 and 735 at the close of FY2025 without stating what either count includes. The department put the corps at approaching 700 on 2026-05-21 without stating whether temporary military detailees sit inside that figure. None of them states its counting rule, so the totals cannot be compared or added.
• What the per-judge figure counts. No source opened here states whether the FY2025 figure of 5,286 counts cases pending before each judge or cases each judge completed, and the same silence covers the FY2015 figure of 566.
• Why the pending totals differ. Totals in circulation run between about 3.2 million and 3.9 million depending on the source and the closing date. No source opened here states whether the counts take the case or the application as the unit, or how consolidated matters are handled.
• What the record completion total is made of. The department announcement gives a completion count and does not separate decisions reached after a full evidentiary hearing from in-absentia orders and administrative closures. One opened source reports that mass calendar sessions increased and does not set that against any department indicator.
• Why the workforce plan was never developed. No source opened here carries an on-the-record explanation from the department or the review office. The follow-up report records the absence and its date and stops there.
• Whether a ruling would reach removals already made. No source opened here says whether a decision in Federal Circuit No. 26-1575 would apply to removals carried out before it or only to later ones. The case had not been argued as of 2026-07-27, so an answer may not exist yet rather than being merely unreported.
• How the fall in the corps is fully accounted for. One opened source reports more than 130 removals and 43 acceptances of a departure programme. No source opened here reconciles those against the total change in the corps over the same window, or separates retirements, resignations and unfilled vacancies.
• Whether the change of job name altered anything. One opened source reports the posts being advertised from November 2025 under the name deportation judge. No source opened here says whether the duties, the qualifications or the appointment authority differ from those of the position under its earlier name.
• What the representation rate is now. The 30 percent figure is as of 2023-12, and no source opened here gives a later value, so the direction it has moved since is not established here.

See the evidence

No primary document from the removal litigation was read here. The order granting rehearing en banc is listed above with its date and could not be extracted from the published PDF, so the status of the appeal rests on two secondary sources that agree on the docket number, on the grant of rehearing and on the absence of argument as of 2026-07-27. The audit report page and the two department announcements were opened directly and carry the workforce-plan finding, the completion total and the sworn-in class, which are the three claims this document leans on hardest. Where the opened sources disagree the disagreement is left visible rather than resolved: the size of the bench appears as 553, as 685, as 735 and as approaching 700 on four different counting bases, and the pending total appears between about 3.2 million and 3.9 million depending on the publisher and the closing date. The department position that the backlog fell sharply and that the most recent class was the largest in the history of the review office is recorded in the same blocks as the fall in the corps, and not only here. Three rows above are federal primary documents in PDF form that the extraction step could not read, and they are kept in the table rather than dropped, because a source that was sought and not read is a different state from a source that does not exist. This is a Path A output, so observation_refs is empty and provenance_mode is press-derived.

People affected

Estimated range 3,195,137–3,195,137 As of 2026-06-30

Sensitivity The interval has zero width and is not a confidence interval. It is a single published figure for a single closing date, and it would move in two directions that could not be computed from the sources opened here. Downward, because the unit is cases rather than people, and no opened source publishes a person count or a ratio that would convert one into the other. In both directions, because pending totals in circulation run between about 3.2 million and 3.9 million depending on which source and which closing date is taken, and no opened source states whether the counts take the case or the application as the unit, or how consolidated matters are handled. The Department of Justice announced on 2025-09-04 that the pending total had fallen by more than 447,000 since 2025-01-20, from above 4.18 million to below 3.75 million, which is a different publisher and a different closing date from the figure used here and is not a restatement of it. The subset of 2,310,698 cases carrying a filed asylum application is not a lower bound on this count. It answers a different question, namely how many pending cases turn on a claim whose refusal ordinarily leads to a removal order, and the people in the remaining cases stand inside the same gap. This count measures exposure to the absence of a staffing rule, not harm. It does not say how long any case waits, and no opened source converts the pending total into a waiting time for this closing date.

Regional breakdown No source opened here gives a court-by-court count of pending cases. Court-level figures appear in the opened sources only for judge losses, and using those to split the national pending total would be proportional allocation on an unrelated denominator. Filings are concentrated by where matters are docketed rather than distributed with judge count or with population, so such a split would be wrong in both directions at once.