08 — Immigration & the Border

STAGE 1 — EVIDENCE BRIEF (Part 1 of 2: Numbered facts)

Referee panel: Prof. Olivia Brandt (#57, statistics, C) · Prof. June Takahashi (#17, econometrics, C) · Domain specialist Prof. Daniel Mbeki (#70, demography, C)

Ground rules from the panel. Every figure below was pulled from the source linked, with the data year stated. Where agencies disagree, we give both numbers and say why. Where we could not verify something that the debate obviously needs (e.g. an official immigration-court asylum grant rate for FY2025, or the average number of days a case takes to decide), we say so. We don't fill the gap with a guess. Today is 2026-09-24.

---

A. The border: surge, then collapse

1. FY2022 was the record year for Border Patrol encounters at the southwest border: 2,206,436. FY2023 had 2,045,838 and FY2024 had 1,530,523. The busiest single month was December 2023, at 249,740. — Pew Research Center, Feb 2, 2026, using CBP data. https://www.pewresearch.org/short-reads/2026/02/02/migrant-encounters-at-the-us-mexico-border-are-at-their-lowest-level-in-more-than-50-years/
2. FY2025 southwest-border Border Patrol encounters: 237,538. From February 2025 on, every month came in below 10,000 (e.g., Mar 2025: 7,179; Jul 2025: 4,592; Dec 2025: 6,478). Pew says these are the lowest monthly totals in more than 25 years of available data, below even the pandemic low of April 2020 (16,182). — Pew, same link.
3. "Lowest since 1970." DHS reported FY2025 as the lowest fiscal year for Border Patrol apprehensions since 1970 (DHS, Oct 7, 2025). CIS's Andrew Arthur gives the FY1970 comparison as 231,116 apprehensions nationwide, and he reckons today's apprehension rate at roughly 92–95%, against 42.5% in FY2000 (CIS, Oct 28, 2025). Referee note (Brandt): Pew labels 237,538 as southwest border and CIS labels it nationwide. We couldn't open CBP's own page (403) to settle which is right. The two figures differ by only a few thousand in either case, so the headline holds.
4. Title 42 ran from March 2020 to May 11, 2023. There were about 3.1 million expulsions (USBP plus OFO). Because an expulsion carried no formal penalty, many people crossed again and again, which inflates the encounter counts from that period. — Wikipedia summary of CBP data. https://en.wikipedia.org/wiki/Title_42_expulsion
5. CBP One (Jan 2023–Jan 2025): nearly 1 million people booked ports-of-entry appointments, about 1,450 a day. The asylum-scheduling function was shut off on January 20, 2025, and the app came back in March 2025 as "CBP Home," a self-deportation tool. https://en.wikipedia.org/wiki/CBP_One

B. Stocks and flows of people

1. Unauthorized population: a record 14.0 million in 2023 (the previous peak was 12.2 million in 2007). Unauthorized immigrants were 5.6% of the U.S. workforce, also a record. About 6 million (roughly 43%) had some temporary protection from deportation: asylum pending (2.6 million), parole, TPS, DACA and others. Mexico accounted for 4.3 million (30%), the smallest share on record. — Pew, Aug 21, 2025. https://www.pewresearch.org/race-and-ethnicity/2025/08/21/u-s-unauthorized-immigrant-population-reached-a-record-14-million-in-2023/
2. CBO (Jan 2026): net immigration was about 2.3 million in CY2024. CBO projects 410,000 for 2025 (its January 2025 projection had been 2.0 million) and 570,000 for 2026. The "other foreign nationals" category, which is mostly unauthorized and paroled people, is projected to be negative: −360,000 in 2025 and −330,000 in each of 2026 and 2027. Fertility is projected at 1.58 in 2026. Deaths overtake births in 2030, so "without immigration, the population would begin to shrink in 2030." Population is projected at 349 million in 2026 and 364 million in 2056. https://www.cbo.gov/publication/61994 ; PDF: https://www.cbo.gov/system/files/2026-01/61879-Demographic-Outlook.pdf
3. Census Bureau (Vintage 2025): net international migration (NIM) fell from 2.7 million (July 2023–June 2024) to 1.3 million (July 2024–June 2025), a 53.8% drop. Census projects NIM of about 321,000 for the year ending July 2026. Population growth halved from 1.0% to 0.5%, and the population on July 1, 2025 was 341.8 million. https://census.gov/newsroom/press-releases/2026/population-growth-slows.html
4. Census methodology caveat. In the CPS, the foreign-born population fell from 53.3 million (Jan 2025) to 51.9 million (June 2025). Census stresses that emigration is hard to measure and that its 2025 estimate straddles two very different policy regimes. https://census.gov/newsroom/blogs/random-samplings/2026/01/historic-decline-in-net-international-migration.html
5. Brookings (Edelberg, Veuger, Watson, Jan 13, 2026): 2025 net migration was between −295,000 and −10,000, which would be "the first time in at least half a century it has been negative." The range for 2026 is −925,000 to +185,000. At full employment, sustainable monthly job growth falls to 20,000–50,000 in late 2025 and could be roughly −20,000 to +20,000 in 2026 in the low scenario. The authors estimate lower migration cuts real GDP growth by 0.2–0.3 pp in 2025 and 0.1–0.3 pp in 2026. https://www.brookings.edu/articles/macroeconomic-implications-of-immigration-flows-in-2025-and-2026/

Mbeki's note on facts 7, 8 and 10: These three estimates don't contradict each other so much as measure different things. They use different periods (calendar year vs. July–June) and different treatment of emigration. They agree on direction: a collapse from about 2.3–2.7 million a year to somewhere between slightly negative and a few hundred thousand. Whether 2025 was actually negative is not established.

C. Immigration courts

1. Backlog: 3,092,988 pending cases at the end of August 2026, of which 2,278,978 are asylum cases. That is down from 3,377,998 in December 2025 and 3,318,099 in February 2026. — TRAC. https://tracreports.org/immigration/quickfacts/eoir.html ; https://tracreports.org/whatsnew/email.260324.html
2. FY2026 through August: 899,741 completions against 514,061 new cases. There were 741,870 removal orders, 83.5% of completions. In August 2026, only 24.2% of people ordered removed had a lawyer. Only 1.79% of new FY2026 cases alleged any criminal activity. — TRAC, same link.
3. June 2026: a record 100,773 court closures and 78,882 removal orders. Only 1.9% of people in closed cases were granted relief (1,883), and 68% of removal orders were issued in absentia. — TRAC report 776. https://tracreports.org/reports/776/
4. Judges: 683 IJs plus 43 assistant chief IJs (726 total) at the start of 2025, falling to 520 plus 33 (553). That is a net loss of 202, from roughly 100 firings plus resignations. About 75% of attorney advisers left. New hires were 17 permanent IJs and 52 temporary military JAG lawyers, and two courts had zero judges. — NPR, Feb 23, 2026. https://www.npr.org/2026/02/23/g-s1-110911/trump-immigration-judges-dismissals-numbers
5. Asylum grant rates and time to decision. Not verified for FY2025–26. TRAC reports asylum grants as a share of relief grants (44–46% in Feb and Aug 2026), which is not the conventional asylum grant rate. We couldn't open CRS R47077 (blocked to our fetcher). We don't report a days-to-decision figure.

D. Enforcement and money

1. The 2025 reconciliation law (OBBBA) adds about $170.7 billion in immigration and border enforcement. That includes $45 billion for ICE detention (about $10.6 billion a year through FY2029), $29.9 billion for ICE enforcement and removal (10,000 new officers), $46.6 billion for the wall, a $100 asylum filing fee plus $100 a year, a $1,000 parole fee, a $250 visa integrity fee, and a cap of 800 IJs. — American Immigration Council fact sheet. https://www.americanimmigrationcouncil.org/fact-sheet/big-beautiful-bill-immigration-border-security/ ; Brennan Center: ICE's annual resources are now roughly triple FY2024. https://www.brennancenter.org/our-work/analysis-opinion/big-budget-act-creates-deportation-industrial-complex
2. ICE removals: FY2024 had 271,484 (the Biden era's last full year). Jan 26–Sep 20, 2025 had 234,211. Adding FY2026 through Nov 15, 2025 gives 290,603 over 228 days, about 1,275 a day. Early in the term (Jan 26–May 3, 2025) the rate was 737 a day, against a Biden-era average of 742. DHS's claim of 135,000 removals in the first 100 days compares with about 72,000 on ICE's own data. — TRAC reports 759 and 767. https://tracreports.org/reports/759/ ; https://tracreports.org/reports/767/ . Note: DHS's "more than half a million removals" release (Oct 27, 2025) folds in CBP returns and other categories, so the numbers aren't comparable.
3. Detention: 65,765 people on July 11, 2026, of whom 70.6% had no criminal conviction. — TRAC. https://tracreports.org/immigration/quickfacts/ . Cato (Bier): in FY2026, 5% of detainees had violent convictions and 73% had none. Among people arrested by ICE, the share with convictions fell from 62% (Jan 2025) to 31% (Nov 2025). https://www.cato.org/blog/5-ice-detainees-have-violent-convictions-73-no-convictions
4. Refugee ceiling: 7,500 for FY2026, the lowest in the program's history. https://en.wikipedia.org/wiki/Asylum_in_the_United_States

E. Fiscal and macroeconomic effects

1. CBO (July 2024), surge effects: the surge adds 8.7 million net arrivals over 2021–26 and cuts deficits by $0.9 trillion over 2024–34. Revenues rise by $1.2 trillion and outlays by $0.3 trillion. Nominal GDP is $1.3 trillion (3.2%) higher in 2034. Workers without a high-school diploma see slightly slower wage growth through 2026, which reverses afterward. https://www.cbo.gov/publication/60165
2. National Academies (2016): over 10+ years, the wage effect on native-born workers overall is "very small." The negative effects fall mainly on prior immigrants and native high-school dropouts. The second generation is among the strongest net fiscal contributors of any group. Over 75 years the federal impact is generally positive, while the state and local impact is generally negative, mostly because of the cost of schooling immigrants' children. https://www.nationalacademies.org/news/new-report-assesses-the-economic-and-fiscal-consequences-of-immigration
3. Fiscal value depends heavily on education (NAS-based, Orrenius, Dallas Fed 2017). The 75-year net present value runs from +$726,000 (more than a college degree) down to −$200,000 (less than high school). A recent immigrant without a high-school diploma costs about $117,000, roughly $200,000 less than a similarly educated native. https://www.dallasfed.org/-/media/documents/research/papers/2017/wp1704.pdf
4. Manhattan Institute (Di Martino, Sept 2024). Lifetime net: +$1M or more (graduate degree, age 25–34) and −$300,000 or worse (less than high school). The average new immigrant, lawful or not, reduces the deficit by more than $10,000. The average newly arrived unauthorized immigrant is a net burden of about $130,000, and the 2021–26 border cohort carries an estimated lifetime cost of $1.15 trillion. https://manhattan.institute/article/the-lifetime-fiscal-impact-of-immigrants
5. CBO on S.744 (2013 comprehensive reform, including legalization): −$197 billion in deficits over 2014–23 and about −$700 billion over 2024–33. https://www.cbo.gov/publication/44225
6. Mass-deportation cost (American Immigration Council, Oct 2024). A one-time program is estimated at at least $315 billion, and a program deporting 1 million a year at about $88 billion a year. The Council projects GDP 4.2–6.8% lower. Advocacy source; the numbers depend on the model. https://www.americanimmigrationcouncil.org/report/mass-deportation

F. Labor markets

1. Mariel: Card (1990) found essentially no effect of a 7% labor-supply shock on Miami's less-skilled workers. Borjas (2017) found wages for non-Hispanic male high-school dropouts fell 10–30%. Peri & Yasenov got that result only in a very small CPS subsample (15–20 observations per year). Clemens & Hunt attribute Borjas's decline to a 1980 change in CPS coverage that doubled the share of Black men in the sample. — Bruegel summary. https://www.bruegel.org/blog-post/mariel-boatlift-controversy ; Borjas, ILR Review 2017. https://journals.sagepub.com/doi/10.1177/0019793917692945
2. Enforcement spillovers (East, Hines, Luck, Mansour & Velásquez, JOLE 2023): Secure Communities reduced employment among likely-undocumented workers and reduced employment and hourly wages of U.S.-born workers. The mechanisms were higher labor costs and lower local demand. https://www.journals.uchicago.edu/doi/10.1086/721152

G. Crime

1. Light, Miller & Kelly (PNAS 2020), Texas DPS data, 2012–18. Relative to undocumented immigrants, U.S.-born citizens had felony arrest rates 2× higher for violent crime, 2.5× for drug crime and more than 4× for property crime. The result holds using convictions and misdemeanors. https://www.pnas.org/doi/10.1073/pnas.2014704117 (summary: https://phys.org/news/2020-12-undocumented-immigrants-commit-crimes-citizens.html)
2. The counter-argument (CIS: Kennedy, Richwine, Camarota, Oct 2022). Texas identifies illegal status through DHS fingerprint matches that lag, so recent arrestees are undercounted. After adjusting, CIS finds higher rates of homicide and sexual-assault convictions than the state average. https://cis.org/Report/Misuse-Texas-Data-Understates-Illegal-Immigrant-Criminality
3. Abramitzky, Boustan et al. (AER: Insights 2024; NBER w31440). In 2020, immigrants were 60% less likely to be incarcerated than the U.S.-born, and the gap has widened since 1960. https://www.nber.org/papers/w31440
4. Miles & Cox (J. Law & Econ 2014): Secure Communities, which led to about 250,000 detentions, had "no observable effect on the overall crime rate." https://www.journals.uchicago.edu/doi/10.1086/680935
5. Baker (AER P&P 2015): IRCA's 1986 legalization was followed by 3–5% less crime, mostly property crime, working through better labor-market access. https://www.aeaweb.org/articles?id=10.1257/aer.p20151041

H. High-skill immigration

1. H-1B lottery: 758,994 eligible registrations and 188,400 selected for FY2024. For FY2025, after the switch to beneficiary-centric selection, there were 470,342 registrations (about 442,000 unique people) and 135,137 selected. — USCIS. https://www.uscis.gov/working-in-the-united-states/temporary-workers/h-1b-specialty-occupations/h-1b-electronic-registration-process
2. The $100,000 H-1B fee (Proclamation 10973, Sept 19, 2025) was vacated by the District of Massachusetts as an unlawful tax. The First Circuit denied a stay on July 24, 2026, so the fee can't currently be collected, and the government could still seek emergency relief. — Littler. https://www.littler.com/news-analysis/asap/appellate-court-pauses-100000-h-1b-fee-policy-appeal-continues . Fragomen reports the proclamation was extended. https://www.fragomen.com/insights/united-states-president-trump-extends-dollar100000-h-1b-fee-but-policy-currently-vacated-by-court-order.html
3. Innovation (Bernstein, Diamond, Jiranaphawiboon, McQuade & Pousada, NBER w30797). Immigrants are 16% of inventors, produce 23% of patents and account for 32% of aggregate innovation, more than half of it through spillovers to U.S.-born collaborators. https://www.nber.org/papers/w30797
4. Firm-level lottery evidence (Doran, Gelber & Isen, JPE 2022). Each extra H-1B a firm wins crowds out about 1.5 other workers, with "insignificant and at most modest" effects on the firm's patenting. https://www.journals.uchicago.edu/doi/10.1086/720467
5. Wage levels (Costa & Hira, EPI 2020, FY2019 data). About 60% of certified H-1B positions were at wage Levels 1–2 (17th and 34th percentiles of the local occupational wage), which are below the local median. https://www.epi.org/publication/h-1b-visas-and-prevailing-wage-levels/
6. Green-card backlog (Bier, Cato, Aug 2023): 1.8 million employment-based cases, of which 1.1 million (63%) are Indian nationals. The implied wait for new Indian EB-2/EB-3 applicants is 134 years, driven by the 7% per-country cap. https://www.cato.org/blog/18-million-employment-based-green-card-backlog

I. Worksite enforcement and legalization proposals

1. E-Verify. As of 2023, 22 states mandate it for some or all employers. Orrenius & Zavodny (2015, 2016) find mandates reduce undocumented workers' earnings and the number of less-educated Mexican and Central American immigrants, and improve outcomes for U.S.-born Hispanic men. A 2019 Cato analysis estimated E-Verify caught only 16.1% of unauthorized workers in FY2018, largely because of identity fraud. — Wikipedia summary. https://en.wikipedia.org/wiki/E-Verify (Secondary source; treat the 16.1% as indicative.)
2. Dignity Act (H.R. 4393, introduced July 15, 2025; Salazar R-FL and Escobar D-TX, with 20 R and 20 D original cosponsors). Long-resident unauthorized immigrants (present before Dec 31, 2020) could get renewable legal status after paying $7,000 in restitution over 7 years plus back taxes and passing background checks, with no special path to citizenship. The bill also mandates 100% nationwide E-Verify, funds border barriers and technology, and sets up 60-day "humanitarian campus" asylum processing. https://salazar.house.gov/dignity-act

(Part 2, Contested Evidence, follows.)

STAGE 1 — EVIDENCE BRIEF (Part 2 of 2: Contested evidence)

Brandt / Takahashi / Mbeki

These are the places where serious scholars look at the same world and still disagree. We name the competing studies and say what would settle each question.

C1. Why did the border collapse, and will it stay low?

• Enforcement-deterrence view. The drop began before January 2025, under the June 2024 asylum restriction, and after January 20 it steepened to below 10,000 a month (Fact 2). That timing points to policy: ending CBP One scheduling, suspending asylum access at the border, and threatening removal to third countries. The monthly series has one visible break in early 2025.
• Push-factor and Mexico view. Mexican interior enforcement, economic conditions in the origin countries, and smuggler pricing also shifted. Nobody has cleanly decomposed these factors with a published causal design.
• Takahashi: One break with many simultaneous treatments can't be identified. We rate "policy caused most of the 2025 decline" as highly plausible but not quantitatively established. "Will it stay low if policy stays?" has no track record yet, beyond noting that low-flow periods since the 1970s have been reversed by economic shocks.

C2. Was 2025 net migration negative?

• CBO: +410,000. Census: +1.3 million for July 2024–June 2025, and about 321,000 projected for the year to July 2026. Brookings: −295,000 to −10,000 (Facts 7, 8, 10).
• Mbeki: Emigration is the uncertain quantity, and CPS nonresponse among immigrants may well have risen with fear of enforcement, which biases CPS-based counts downward. Contested.

C3. Labor-market effects of low-skill immigration

• Small or no effect: Card (1990); NAS (2016) for native workers overall over 10+ years; Peri & Yasenov; Clemens & Hunt.
• Meaningful effect on substitutes: Borjas (2017, a 10–30% decline for male high-school dropouts in Miami); NAS (2016), which acknowledges prior immigrants and native dropouts bear the costs.
• Enforcement spillover: East et al. (2023) find that removing undocumented workers lowered native employment and wages. That runs against the simple substitution model.
• Takahashi: The honest summary is that aggregate native wage effects are small, and distributional effects on close substitutes are real but smaller than restrictionists claim and larger than zero. Mariel is still unresolved because the sample is tiny.

C4. Fiscal effects

• Positive: CBO (2024) on the surge, −$0.9 trillion over 10 years; CBO on S.744; NAS on the federal 75-year view; Di Martino's average-new-immigrant figure of about +$10,000.
• Negative for low-education and unauthorized arrivals: Di Martino ($1.15 trillion lifetime for the 2021–26 cohort; about −$130,000 per average unauthorized arrival); Orrenius and NAS (−$200,000 NPV for less than high school); the NAS state and local findings.
• Brandt: These don't contradict each other. They differ on the time horizon (10 years vs. lifetime), on who is counted (the average immigrant vs. the low-education or unauthorized subgroup), on treatment of public goods (average vs. marginal cost), and on whether children are allocated to parents. CBO's 10-year window picks up workers' taxes before their old-age benefits. Lifetime models pick up the benefits. Anyone who cites one number without its horizon is being misleading.

C5. Immigrant crime

• Lower offending: Light et al. (2020); Abramitzky, Boustan et al. (2024, 60% lower incarceration); Baker (2015, legalization reduced crime).
• Higher on some serious offenses: CIS (2022) says Texas data understate undocumented offending once late identification is corrected. That critique is not peer-reviewed.
• Enforcement → crime: Miles & Cox (2014) find no crime effect from Secure Communities.
• Takahashi: Across data sets and designs, the evidence that immigrants on average offend at lower rates than the U.S.-born is strong. The CIS critique of identification lag is a real methodological point, and it matters most for specific rare offenses. It doesn't overturn the overall finding.

C6. High-skill immigration and innovation

• Aggregate: Bernstein et al. find 32% of innovation, with spillovers.
• Firm level: Doran, Gelber & Isen find crowd-out of 1.5 workers per extra H-1B and no significant firm-level innovation effect.
• Wage levels: Costa & Hira report 60% of positions at Levels 1–2. Defenders reply that Level 1–2 corresponds to entry-level workers.
• Brandt: These can all be true together. H-1B as currently designed (a lottery, heavy outsourcing-firm share, wage levels) is a poor filter for the high-impact inventors the aggregate literature measures.

C7. Enforcement effectiveness

• The removal rate did rise (about 1,275 a day vs. 742; Fact 17), but late-2025 removals were only about 7% above FY2024 despite roughly triple the resources (TRAC 767). Brandt: That comparison mixes a partial year with full years, and the count of CBP Home self-departures isn't independently verified. The "cost per removal" is unknown in any audited form.

Known unknowns (we could not verify)

• The immigration-court asylum grant rate for FY2025–26 and the mean time to decision.
• The size and effect of the 2025 "safe third country" and third-country removal agreements. Press reports exist, but we have no primary data.
• Verified counts of self-deportation.
• Effects of the 2025 surge's reversal on wages and inflation. There isn't enough data yet. The Brookings GDP numbers are projections.

STAGE 2 — OPENING POSITIONS (Part 1: Left and Center)

---

Prof. Hector Villanueva (#26, immigration law · L) — the humanitarian and due-process case

Causal claim. The 2025–26 regime isn't "enforcement" in any ordinary legal sense. It is mass removal without adjudication. The courts have been gutted on purpose: 202 judges gone and two courts with zero judges (NPR). Meanwhile ICE detention sits at record levels, and 70.6% of detainees have no criminal conviction (TRAC, July 2026). The June 2026 record for closures isn't efficiency. It is 68% in absentia orders and 1.9% relief (TRAC 776), with only 24% of people represented.

Evidence. Beyond the above: the administration's early removal pace was no higher than Biden's (737 a day vs. 742, TRAC 759), so what the new money bought was mostly arrests and detention of non-criminals, not better targeting. The American Immigration Council estimates mass deportation would cut GDP by 4–7%.

Remedy. Appointed counsel for children and detained people. An independent Article I immigration court. Restored ports-of-entry asylum processing. Reinstated CHNV-style parole with sponsor vetting. Earned legalization for long-resident people.

Tradeoff I concede. Humane processing at the border does create some pull. CBP One's roughly 1,450 appointments a day were not free of incentive effects, and a court that takes years to decide cases is itself a magnet. Counsel makes cases slower in the short run.

Falsifiable prediction. If I'm right that the court purge is about throughput rather than accuracy, then by FY2028 federal appellate remand rates in immigration petitions for review will rise measurably, and in absentia rates will stay above 60%. If remands stay flat and representation rises, I'm wrong.

---

Prof. Miriam Castellano (#1, labor econ · C-L) — the economic-growth-with-legality case

Causal claim. Immigration is net positive for growth and for the federal budget, and native wage losses are small and concentrated. The best policy is a legal high-volume system, not an unauthorized one. The 2025 turn to near-zero or negative net migration is a supply shock we'll pay for in GDP and job growth.

Evidence. CBO (2024): the surge cuts deficits by $0.9 trillion over 10 years and raises 2034 GDP by 3.2%. NAS (2016): native wage effects are "very small." Mariel shows essentially zero wage effect. Borjas's result is a CPS composition artifact (Clemens & Hunt). Brookings (2026): lower migration costs 0.2–0.3 pp of GDP growth, and breakeven job growth is now 20–50k a month. East et al. (2023): enforcement lowers native employment. E-Verify only catches about 16% of unauthorized workers, so it's theater.

Remedy. Legalization along Dignity Act lines, but with a path to citizenship. Expanded employment-based visas indexed to labor-market conditions. Wage-ranked H-1B selection.

Tradeoff I concede. State and local budgets bear real net costs for schooling immigrants' children (NAS). Low-education arrivals are lifetime fiscal negatives in every model I trust (Orrenius: −$200,000).

Falsifiable prediction. If the 2025 migration collapse costs growth as I claim, then 2026–27 employment in construction, agriculture and home health will show vacancy rates rising and real wages rising for incumbents, but total hours worked by U.S.-born workers in those sectors will not rise enough to offset the loss. If U.S.-born hours fill the gap within two years, the complementarity story is weaker than I think.

---

Prof. Luis Moreno (#62, sociology of work · Soc) — the labor-solidarity case

Causal claim. Both the open-borders business lobby and the deportation state treat migrants as a disposable labor supply. Unauthorized status is a wage-suppression technology: it lets employers underpay and intimidate workers, which drags down standards for citizens working alongside them. Guest-worker programs, H-1B above all, do the same thing in a legal wrapper.

Evidence. Costa & Hira: 60% of H-1B positions are certified at Levels 1–2, below the local median. Doran, Gelber & Isen: each H-1B crowds out 1.5 other workers. Pew: unauthorized workers are 5.6% of the workforce, a record. East et al.: enforcement raids hurt U.S.-born workers too, because they hit the local economy, not the bosses.

Remedy. Legalization with full labor rights and portable status, so workers can quit abusive employers. Put the enforcement money into labor inspections (wage-and-hour, OSHA) rather than ICE. Abolish the H-1B's employer-tied structure and replace it with portable, wage-floored work visas.

Tradeoff I concede. Legalization plus labor standards raises labor costs. Some jobs, in agriculture especially, may be automated or offshored. That is a real cost to some consumers.

Falsifiable prediction. If status is the suppression mechanism, then workers legalized under any new program should see wage gains of 5% or more within five years relative to comparable still-unauthorized workers, as the post-IRCA literature suggests. If gains are near zero, then status isn't the lever.

---

Prof. Ingrid Solberg (#23, administrative law · C) — the institutional-capacity case

Causal claim. The core failure since 2014 is adjudication capacity, not border hardware. When a system takes years to decide claims, a humanitarian protection becomes a de facto work permit, and that drives both the surge and the public backlash. 2025–26 has "fixed" throughput by degrading adjudication, which is a different failure.

Evidence. The backlog is falling for the first time in years: 3.38 million in Dec 2025, 3.09 million in Aug 2026, with 899,741 completions vs. 514,061 intakes (TRAC). But the drivers are dismissals, in absentia orders and near-zero relief, not more judges. Judges fell by about 25% (NPR). The OBBBA caps judges at 800 while funding ICE detention with about $45 billion. That is a 400% increase in detention against about 14% more judges (Brennan).

Remedy. Uncap and fund judges to about 1,200 with support staff. Asylum-officer adjudication at the border on a 60–90 day clock (the Dignity Act's "humanitarian campus" idea is a defensible frame). Publish monthly OHSS statistics again.

Tradeoff I concede. Fast adjudication means more denials delivered quickly, and some of them will be wrong. It also requires detention or close supervision during the process.

Falsifiable prediction. If capacity is the binding constraint, then in any district where the IJ count per pending case rises 30%, the median time to first merits hearing should fall by at least a third within 18 months, without the grant rate collapsing.

---

Prof. Nadia Haddad (#15, trade · C) — the talent-competition case

Causal claim. The biggest long-run economic stake in U.S. immigration policy isn't the southern border. It is high-skill selection, and here the U.S. is plainly malfunctioning. We run a lottery, cap green cards by country, and in 2025 bolted a $100,000 tax onto the top of it.

Evidence. Immigrants account for 32% of U.S. innovation (Bernstein et al.). The employment-based backlog is 1.8 million, 63% of it Indian, with a 134-year implied wait (Cato). In FY2025 only 135,137 registrations were selected out of 470,342 (USCIS). The $100k fee was vacated as an unlawful tax (D. Mass.; First Circuit denied a stay). Also, 2025 net migration was negative for the first time in half a century, and that reaches the high-skill pipeline too.

Remedy. Eliminate per-country caps. Recapture unused green cards. Replace the H-1B lottery with wage-ranked selection. Give "stapled" green cards to U.S. STEM PhDs. Create a startup visa.

Tradeoff I concede. Doran, Gelber & Isen is a real problem for my side. Lottery winners don't innovate more at the firm level, and there is crowd-out. Removing country caps helps Indians and hurts everyone else in the queue for years.

Falsifiable prediction. If wage-ranked selection replaced the lottery, the median offered wage of selected H-1B petitions should rise by at least 20% in the first cycle, and outsourcing-firm share should fall below 15%. If it doesn't, the market for these visas isn't what I think.

STAGE 2 — OPENING POSITIONS (Part 2: Right and Libertarian)

---

Prof. Dmitri Volkov (#33, national security · C-R) — the deterrence-and-control case

Causal claim. The 2021–24 surge was policy-made, and so was its end. When the U.S. signaled that arrival meant release (a years-long court wait, parole, work authorization), people came in record numbers. When it signaled that arrival meant no entry and probable removal, they stopped. This is the cleanest natural experiment in border policy we have.

Evidence. FY2022 had 2.2 million southwest-border encounters, with 249,740 in December 2023 alone. By FY2025 the total was 237,538, and every month since February 2025 has been under 10,000 (Pew). That is the lowest since 1970 (DHS). Apprehension rates are now around 92–95% (CIS). Title 42 kept the border effectively closed until Biden ended it in May 2023. Interior enforcement is now focused on public-safety threats. The unauthorized population hit a record 14 million in 2023 (Pew), 43% of them with some quasi-legal protection, which is exactly the "catch-and-release to status" pipeline.

Remedy. Keep the OBBBA enforcement funding. Codify the asylum restrictions so they survive a change of administration. Expand safe-third-country agreements. Finish the barrier where Border Patrol wants it.

Tradeoff I concede. Deterrence means genuine refugees with valid claims will be turned away or discouraged. There is a humanitarian cost, and I won't pretend otherwise.

Falsifiable prediction. If deterrence is doing the work, monthly southwest-border encounters will stay below 20,000 through FY2027 as long as the policy holds, even if Latin American economic conditions worsen. If encounters rebound above 50,000 a month with policy unchanged, push factors matter more than I think.

---

Prof. Douglas Wren (#2, public finance · R) — the fiscal and selection case

Causal claim. The fiscal debate is settled once you look at who comes. High-education immigrants are hugely positive. Low-education and unauthorized arrivals are lifetime negatives, and the 2021–26 surge was overwhelmingly the latter. CBO's "$0.9 trillion deficit reduction" is a 10-year window artifact that counts workers' taxes and stops before their retirement, Medicaid and their children's schooling show up.

Evidence. Manhattan Institute (Di Martino): the 2021–26 border cohort costs $1.15 trillion, about $130,000 per unauthorized arrival. Orrenius (Dallas Fed, NAS-based): less than high school is −$200,000. NAS: state and local impacts are negative. Unauthorized workers are 5.6% of the workforce, and E-Verify works: Orrenius & Zavodny show mandates reduce the unauthorized workforce and raise U.S.-born Hispanic men's outcomes.

Remedy. Nationwide E-Verify. Points-based or wage-based selection for legal immigration. No legalization until E-Verify is fully operational for a set period.

Tradeoff I concede. A selective system lowers total inflows and hence near-term GDP growth. The Brookings 0.2–0.3 pp figure isn't nothing. Agriculture and elder care will face real shortages.

Falsifiable prediction. If selection drives fiscal outcomes, then CBO's next dynamic score of any reform shifting the mix toward college graduates, at the same total volume, should show larger 20-year deficit reduction than the S.744 baseline per immigrant. If CBO finds no difference, my framing is overstated.

---

Prof. Arthur Pembroke (#27, political theory · R) — the sovereignty and cohesion case

Causal claim. A democracy has the right to decide its membership, and it must be seen to do so. The 2021–24 surge badly damaged public trust, and trust is the precondition for any generous immigration policy. Enforcement comes first, not as punishment but as proof of consent. Enforcement also improves public safety.

Evidence. The Texas data, when corrected for identification lag, show illegal immigrants have higher homicide conviction rates than the state average (CIS 2022). Secure Communities-type cooperation removes offenders before they reoffend. 14 million unauthorized is a failure of the rule of law, whatever the economists' GDP models say.

Remedy. Enforcement-first sequencing. Keep the new ICE capacity. Mandatory detention for anyone arrested for a violent crime. Legal immigration levels set by Congress and made more selective. No amnesty that rewards unlawful entry, and the Dignity Act's $7,000 is too cheap.

Tradeoff I concede. Enforcement-first has been promised before (1986) and delivered only partly, and the people who came as children have a stronger claim. I also concede that deporting parents of U.S.-citizen children imposes real costs on citizens.

Falsifiable prediction. If enforcement improves public safety, then jurisdictions with the largest 2025–26 increase in ICE arrests per capita should show faster declines in violent crime through 2027 than comparable jurisdictions. If there is no differential, I'll drop the public-safety part of my argument.

---

Prof. Elliott Crane (#16, entrepreneurship · Lib) — the free-movement and markets case

Causal claim. Immigration restriction is the largest single price control in the U.S. economy. Both parties' programs, the lottery-and-cap regime and the $100k fee, are central planning of the labor market. Immigrants create firms and patents at extraordinary rates, and H-1B workers drive the innovation of the firms that hire them.

Evidence. 32% of U.S. innovation (Bernstein et al.). A 1.8 million green-card backlog with Indian applicants facing a 134-year wait (Cato). An H-1B demand of about 442,000 unique people for 85,000 statutory slots (USCIS; the selected number exceeds the cap because of non-filing). The $100k fee was struck down as an unlawful tax. Enforcement is extraordinarily expensive: AIC puts it at about $88 billion a year for 1 million removals, and ICE's budget has roughly tripled.

Remedy. Uncap high-skill green cards. Auction or wage-rank work visas. A large, market-priced guest-worker program for lower-skill sectors, with visas portable across employers. Remove welfare eligibility rather than people ("build a wall around the welfare state").

Tradeoff I concede. Portable guest-worker programs have a history of overstays (Europe's 1960s Gastarbeiter), and the political legitimacy problem Pembroke raises is real.

Falsifiable prediction. If visas were market-priced, the clearing price would reveal large surplus, and firms would pay $20,000 or more per H-1B-equivalent visa in an auction, far below $100k but far above zero. If bids clear near zero, the "shortage" story is overstated.

STAGE 3 — CROSS-EXAMINATION (Part 1)

---

Exchange 1: Wren (R) vs. Castellano (C-L) on wages, fiscal windows and E-Verify (4 rounds)

Wren, Round 1. Miriam, three problems. (a) "Mariel shows zero wage effect" overstates it. The brief's own summary is that Mariel is unresolved, because every side is working with 15–20 CPS observations a year. Clemens & Hunt show Borjas's result could be a composition artifact. That doesn't prove the true effect is zero, and NAS (2016) says the costs fall on prior immigrants and native dropouts. (b) You call E-Verify "theater" on the strength of a 16% catch rate. But Orrenius & Zavodny's causal estimates show mandates reduce less-educated unauthorized employment and raise outcomes for U.S.-born Hispanic men, and they did that despite identity fraud. A leaky filter can still deter. (c) CBO's $0.9 trillion is a 10-year number for a cohort whose costs come later.

Castellano, Round 1. (a) Fair on the wording. My claim is that credible designs find no detectable effect on native less-skilled workers as a group, and that Borjas's estimate is fragile to the sample. I'll accept "contested" for Mariel specifically. My position doesn't rest on Mariel: NAS's synthesis across hundreds of studies says the overall native effect over ten years is very small. (b) I accept Orrenius & Zavodny as real evidence of a deterrence and earnings effect, and "theater" was rhetoric. But their findings cut two ways. Mandates lowered undocumented earnings, which means they push people into cash work rather than out of the country. (c) On horizons: yes, CBO's window is 10 years. But Di Martino, your own source, finds the average new immigrant is a net positive of more than $10,000 over a lifetime. The negative figure is the subgroup.

Wren, Round 2. Right: the subgroup is the one that arrived in 2021–24. That's my point. Nobody is debating the average of a mix we didn't choose. We're debating a cohort that is overwhelmingly without a high-school diploma, whose lifetime NPV Orrenius puts at −$117k to −$200k. The fiscal case for that flow is negative in every lifetime model on the table.

Castellano, Round 2. Two corrections. First, "overwhelmingly without a high-school diploma" isn't in any source we've cited. Pew gives origin countries, not the education of the 2021–24 cohort. Show me the education distribution or withdraw "overwhelmingly." Second, Orrenius's own −$117k figure is $200k better than a similarly educated native. If your criterion is lifetime NPV, the same logic indicts a large share of native-born high-school dropouts. That's a sign the criterion isn't a sufficient basis for membership policy.

Wren, Round 3. I'll withdraw "overwhelmingly" as unsourced. My claim becomes: the unauthorized cohort is less educated on average than legal immigrants, and Di Martino's roughly −$130k is the estimate for that average. On natives: we can't deport citizens, and we can choose immigrants. That asymmetry is the whole point of selection.

Castellano, Round 3. Accepted as a coherent normative position. Empirically, then, we agree on the education gradient. We disagree on whether a GDP gain that shows up in wages and profits but not in federal NPV counts. CBO's 3.2% higher 2034 GDP is a real welfare gain, much of it accruing to the immigrants themselves.

---

Exchange 2: Villanueva (L) vs. Volkov (C-R) on Title 42 and targeting (3 rounds)

Villanueva, Round 1. Dmitri, "Title 42 kept the border effectively closed" is contradicted by the data. There were about 3.1 million expulsions, and the Title 42 years include FY2022, the record year of 2.2 million encounters. With no penalty attached to expulsion, people simply tried again. Evasions reportedly fell about 70% after Title 42 ended and Title 8 consequences returned. Second, "interior enforcement is focused on public-safety threats." TRAC: 70.6% of detainees have no criminal conviction. Cato's Bier, hardly a leftist, finds 5% have violent convictions, and the share of ICE arrestees with convictions halved from 62% to 31% during 2025.

Volkov, Round 1. On Title 42 I'll partly concede. Expulsion without consequence produced recidivism, and the encounter count overstates unique crossers. What I should have said is that Title 42 without consequences failed, and it was the consequence-bearing regime of 2024–25 that worked. That actually strengthens my deterrence claim. On targeting, I don't dispute the TRAC or Cato numbers. My position is that removing people with final orders or unlawful presence is lawful enforcement, whether or not they have a conviction.

Villanueva, Round 2. Then withdraw "focused on public-safety threats." You can defend enforcing the law against people with no conviction, but that isn't what you said. And the brief shows only 1.79% of new FY2026 court cases allege criminal grounds.

Volkov, Round 2. I withdraw "focused on public-safety threats" as a description of the 2025–26 mix. Bier's 62→31% series is the evidence that moved me. I retain the normative claim that unlawful presence alone justifies removal. I'll add a proposal: statutory priority tiers with public reporting, so voters can see who is being removed.

Villanueva, Round 3. Good, and I'll take that. I'll add that 68% in absentia in June 2026 suggests many people never got notice. Priority tiers with no counsel won't fix that.

---

Exchange 3: Volkov (C-R) vs. Villanueva (L) on removal pace (2 rounds)

Volkov. Hector, you said the removal pace was "no higher than Biden's." That was true only for Jan 26–May 3, 2025 (737 a day vs. 742). TRAC's own later report has 290,603 removals in 228 days, about 1,275 a day, 70% above Biden's average. Your claim is stale.

Villanueva. Accepted. I'll amend it: the pace rose to about 1,275 a day by late 2025. My substantive point stands. With about triple the resources, total FY2025 removals were only modestly above FY2024, and the increase came from arresting non-criminal people rather than from priority targeting. Those are two different claims, and I should have kept them apart.

STAGE 3 — CROSS-EXAMINATION (Part 2)

---

Exchange 4: Nwosu (#36, criminology · C) vs. Pembroke (R) on crime (3 rounds)

Nwosu, Round 1. Arthur, you rest the public-safety claim on one non-peer-reviewed CIS report. Against it: Light, Miller & Kelly (PNAS 2020) find U.S.-born felony arrest rates are 2× (violent), 2.5× (drug) and more than 4× (property) those of undocumented Texans, and the result survives using convictions. Abramitzky, Boustan et al. find immigrants 60% less likely to be incarcerated, a gap that has widened since 1960. And Miles & Cox (2014) find Secure Communities had no observable effect on crime after about 250,000 detentions. Baker (2015) finds legalization reduced crime 3–5%.

Pembroke, Round 1. The CIS point is methodological, not ideological. Texas identifies illegal status via DHS fingerprints, and people with no prior DHS contact are initially coded "unknown." Light et al. are aware of this. Have they fully corrected it? And incarceration comparisons are confounded, because deportation removes immigrant offenders from U.S. prisons.

Nwosu, Round 2. Both fair points, and neither reverses the sign. On identification: CIS's adjusted finding is for specific offenses (homicide, sexual assault) and against the state average, not against the U.S.-born only. The two studies differ on the denominator as well as the numerator. On deportation: Abramitzky et al. find the gap in periods and groups where deportation was rare. And Miles & Cox is a direct causal test of your mechanism, "enforcement removes offenders before they reoffend." It found nothing.

Pembroke, Round 2. I'll accept that the aggregate offending rate of immigrants, unauthorized included, is lower than natives' on the best evidence. I retain that the Texas homicide question is contested, and that the relevant policy margin is removal of those who do offend, which everyone supports.

Nwosu, Round 3. Agreed on that margin. But then drop "enforcement also improves public safety" as a general claim for mass removal. The only causal evidence on your mechanism is null.

Pembroke, Round 3. Amended: general enforcement expansion has no demonstrated public-safety benefit (Miles & Cox). Targeted removal of convicted violent offenders is justified on its own terms. My prediction about jurisdictions stands as the test.

---

Exchange 5: Moreno (Soc) vs. Crane (Lib) on H-1B and innovation (3 rounds)

Moreno, Round 1. Elliott, "H-1B workers drive the innovation of the firms that hire them." Doran, Gelber & Isen used the lottery, the only randomized evidence we have, and found insignificant firm-level innovation effects and crowd-out of about 1.5 workers per visa. Firms that hire H-1Bs innovate more because they are different firms, not because of the visa.

Crane, Round 1. That's a strong paper and I accept its firm-level result. But Bernstein et al. find immigrants account for 32% of aggregate innovation, more than half of it through spillovers to U.S.-born collaborators. Both can hold: the lottery allocates visas badly, so the marginal H-1B at a random firm isn't the marginal inventor. That's an argument for pricing, not for restriction.

Moreno, Round 2. Or for wage floors. Costa & Hira: 60% of H-1B certifications at Levels 1–2. If the visa is a discount coupon, auctioning it off just lets the richest firms buy the cheapest labor.

Crane, Round 2. An auction does the opposite of a discount. A firm paying $20k+ for a visa only does so for a worker whose productivity exceeds the wage by that margin. That screens out wage arbitrage. Level 1–2 is also the entry-level tier, where new graduates properly sit.

Moreno, Round 3. Entry-level is the problem: it's where displaced U.S. graduates compete. But I'll concede that wage-ranked selection (not an auction) would reduce the Level 1–2 share mechanically. I can live with that as a floor-raising reform, with portability.

Crane, Round 3. Amended: I withdraw "drive the innovation of the firms that hire them" as a causal claim about H-1B. I retain the aggregate claim about immigrant inventors. Wage-ranking is a reform we both vote for.

---

Exchange 6: Wren (R) vs. Haddad (C) on "negative net migration" (2 rounds)

Wren. Nadia, you said "2025 net migration was negative." CBO says +410,000. Census says +1.3 million for July 2024–June 2025 and projects about +321,000 for the year to July 2026. Only Brookings's range is negative. You're presenting one estimate as fact.

Haddad. Correct, and my phrasing overstated it. The accurate claim is that 2025 net migration fell to somewhere between about −300,000 and +410,000 on a calendar-year basis (Brookings vs. CBO). The sign isn't established. The collapse in magnitude is. My high-skill argument doesn't depend on the sign.

---

Exchange 7: Castellano (C-L) vs. Wren (R) on "$1.15 trillion" (2 rounds)

Castellano. Douglas, you cite "the cohort costs $1.15 trillion" alongside CBO's −$0.9 trillion as if one refutes the other. Di Martino's figure is lifetime, uses an adjusted method (allocating public goods), and covers the unauthorized subgroup. CBO's is a 10-year budget window for the whole surge, including its effect on GDP. They answer different questions.

Wren. Agreed they answer different questions. I claim the lifetime question is the relevant one for a permanent population. I'll stipulate that both numbers are honest within their frames, and that neither can be quoted without its horizon.

---

Exchange 8: Pembroke (R) vs. Moreno (Soc) on legalization's track record (2 rounds)

Pembroke. Luis, IRCA legalized about 3 million in 1986, and the unauthorized population later reached 12.2 million by 2007 (Pew). Legalization without effective worksite enforcement produced more unlawful migration. Why won't a new one?

Moreno. Because IRCA's employer sanctions were designed to fail: document checks with no verification. That's the one fact on which I agree with Wren about E-Verify. The post-IRCA evidence also shows legalization reduced crime 3–5% (Baker) and raised wages. I'll concede the sequencing point: legalization without a credible worksite-verification mechanism repeats 1986. I'd accept verification paired with portable status and labor enforcement.

Pembroke. Then we disagree only on order and on the price. That's progress.

---

Exchange 9: Crane (Lib) vs. Solberg (C) on court capacity (2 rounds)

Crane. Ingrid, courts closed more cases in FY2026 than they received, with 25% fewer judges. Isn't intake, not capacity, the binding constraint? Cut intake and the backlog falls on its own.

Solberg. The backlog is falling, and I said so. But look at how: 1.9% relief and 68% in absentia in June 2026. That isn't adjudication capacity. It's closure without merits hearings for many cases. Intake collapsed because the border closed, and when intake returns (any future administration, any future crisis) 553 judges can't handle 500,000+ merits-quality cases a year. Capacity is a buffer we'll need.

Crane. Fair. I'll support funding judges if it's paired with asylum-officer front-loading, which is cheaper per case.

STAGE 4 — REFEREE FACT-CHECK

Brandt / Takahashi / Mbeki. We rated every empirical claim made in Stages 2–3. Ratings: S = Supported, C = Contested (credible sources disagree, or the claim rests on a model or advocacy estimate), U = Unsupported (not established by any source we could verify, or contradicted by stronger causal evidence), W = Wrong (contradicted by the cited data). Normative claims aren't rated.

# · Speaker · Claim · Rating · Judged against
1 · Villanueva · 202 judges gone; two courts with zero judges · S · NPR 2026
2 · Villanueva · 70.6% of ICE detainees have no criminal conviction · S · TRAC detention quick facts (Jul 2026)
3 · Villanueva · June 2026: 68% in absentia, 1.9% relief · S · TRAC 776
4 · Villanueva · Only ~24% of people ordered removed were represented · S · TRAC (24.2%, Aug 2026)
5 · Villanueva · Trump removal pace "no higher than Biden's" · W (stale) · TRAC 759 applies only to Jan–May 2025; TRAC 767 shows ~1,275/day vs 742
6 · Villanueva · Mass deportation would cut GDP by 4–7% · C · AIC 2024, an advocacy model for a 13.3M-person removal; no CBO equivalent
7 · Villanueva · CBP One ran ~1,450 appointments/day · S · CBP One summary
8 · Villanueva · Evasions fell ~70% after Title 42 ended · S (secondary) · Title 42 summary of CBP data
9 · Villanueva · Only 1.79% of new FY26 cases allege criminal grounds · S · TRAC
10 · Villanueva · ICE arrestees with convictions fell from 62% to 31% · S · Cato (Bier)
11 · Castellano · CBO: −$0.9T deficits and +3.2% GDP in 2034 · S · CBO 60165
12 · Castellano · NAS: native wage effects "very small" · S · NAS 2016
13 · Castellano · Mariel shows "zero" wage effect · C · Card vs. Borjas vs. Peri-Yasenov / Clemens-Hunt
14 · Castellano · Borjas's result is a CPS composition artifact · C · Clemens & Hunt argue it; Borjas disputes it
15 · Castellano · Brookings: −0.2 to −0.3 pp GDP growth; breakeven jobs 20–50k/month · S (as projection) · Brookings 2026
16 · Castellano · Enforcement lowered native employment and wages · S · East et al. 2023
17 · Castellano · E-Verify caught ~16% of unauthorized workers · S (secondary, FY2018) · Cato 2019 via Wikipedia
18 · Castellano · E-Verify is "theater" · U · Orrenius & Zavodny find causal effects
19 · Castellano · Orrenius: less than HS = −$200k NPV · S · Dallas Fed 2017
20 · Castellano · Average new immigrant is +$10k or more lifetime (Di Martino) · S · MI 2024
21 · Moreno · 60% of H-1B certifications at Levels 1–2 · S (FY2019; dated) · EPI 2020
22 · Moreno · 1 H-1B crowds out ~1.5 workers · S · Doran-Gelber-Isen 2022
23 · Moreno · Unauthorized workers = 5.6% of the workforce · S · Pew 2025
24 · Moreno · Post-IRCA literature shows ≥5% wage gains from legalization · U · The panel verified no wage study; only Baker (crime)
25 · Moreno · IRCA legalization reduced crime 3–5% · S · Baker 2015
26 · Moreno · IRCA employer sanctions were "designed to fail" · C · A characterization; no source in the brief
27 · Solberg · Backlog fell from 3.38M to 3.09M · S · TRAC
28 · Solberg · Completions (899,741) exceed intake (514,061) · S · TRAC
29 · Solberg · Judges down ~25% · S · NPR
30 · Solberg · Detention money up ~400% vs judges ~14% · S · Brennan Center
31 · Solberg · Backlog drop driven by dismissals, in absentia and near-zero relief · C · In absentia and relief data support it; dismissal share not quantified in our sources
32 · Haddad · Immigrants = 32% of innovation · S · Bernstein et al.
33 · Haddad · 1.8M EB backlog, 63% Indian, 134-year wait · S (2023 data) · Cato 2023
34 · Haddad · 135,137 selected of 470,342 (FY2025) · S · USCIS
35 · Haddad · $100k fee vacated; stay denied · S · Littler / Fragomen
36 · Haddad · 2025 net migration "was negative" · U (as a fact) · CBO +410k; Census +1.3M (Jul–Jun); only Brookings is negative
37 · Volkov · FY22 2.21M; Dec 2023 249,740; FY25 237,538 · S · Pew 2026
38 · Volkov · Lowest since 1970 · S · DHS; CIS (231,116 in FY1970). Nationwide/SW labeling caveat
39 · Volkov · Apprehension rate ~92–95% · C · CIS estimate; model-based, no official series cited
40 · Volkov · Title 42 "kept the border effectively closed" · W · 3.1M expulsions with high recidivism; FY2022 record fell under Title 42
41 · Volkov · Interior enforcement "focused on public-safety threats" · W · TRAC 70.6% no conviction; Cato 5% violent
42 · Volkov · 14M unauthorized; 43% with some protection · S · Pew 2025
43 · Volkov · ~1,275 removals/day, ~70% above Biden · S · TRAC 767 (1,275/742 = 1.72)
44 · Wren · CBO's $0.9T is a "window artifact" · C · The horizon matters, but CBO includes real GDP effects; "artifact" overstates
45 · Wren · 2021–26 cohort costs $1.15T lifetime · C · A real MI estimate (adjusted method); not comparable to CBO; not replicated
46 · Wren · Less than HS = −$200k (Orrenius) · S · Dallas Fed
47 · Wren · NAS: state/local impacts negative · S · NAS 2016
48 · Wren · E-Verify mandates reduce unauthorized employment and help U.S.-born Hispanic men · S · Orrenius & Zavodny
49 · Wren · 2021–24 cohort "overwhelmingly" lacks a HS diploma · U · No education data for the cohort in any cited source
50 · Pembroke · Unauthorized Texans have higher homicide conviction rates · C · CIS 2022 vs. Light et al. 2020
51 · Pembroke · Enforcement improves public safety (general) · U · Miles & Cox 2014: no effect
52 · Pembroke · ~3M legalized in 1986, then 12.2M unauthorized by 2007 · S (facts) / C (causal link) · Baker 2015; Pew 2025
53 · Pembroke · Incarceration comparisons confounded by deportation · C · Plausible; Abramitzky et al. address it partly
54 · Crane · H-1B workers drive the innovation of their firms · U · Doran-Gelber-Isen lottery evidence
55 · Crane · ~442,000 unique registrants for 85,000 cap slots · S · USCIS FY2025
56 · Crane · ~$88B/yr for 1M removals · S (as AIC estimate) · AIC 2024
57 · Crane · ICE resources roughly tripled · S · Brennan Center
58 · Crane · An auction screens out wage arbitrage · C · Theory; no U.S. trial
59 · Nwosu · Light et al. 2×/2.5×/4× · S · PNAS 2020
60 · Nwosu · Immigrants 60% less likely to be incarcerated · S · Abramitzky et al.
61 · Nwosu · Secure Communities had no crime effect · S · Miles & Cox
62 · Nwosu · The incarceration gap holds "where deportation was rare" · U · Not verified in the paper's abstract; plausible but unconfirmed

Tally (62 claims): Supported 41 · Contested 11 · Unsupported 7 · Wrong 3.

Brandt: The Right's errors were mostly about characterizing enforcement (Title 42, targeting). The Left's were about overclaiming certainty on contested research (Mariel "zero," E-Verify "theater") and about stale data (removal pace). The Center overstated the sign of net migration. Nobody's core causal story was demolished. Several people's rhetoric was.

---

Retractions and amendments on the record

• Villanueva (#5): Amended to "the removal pace rose to about 1,275/day by late 2025. Total removals rose modestly relative to the resources, and the increase came mostly from people without convictions."
• Castellano (#13, #18): "Mariel is contested. The NAS synthesis is my evidence." "Theater" is retracted: "E-Verify has measurable deterrent effects and is also easily defeated by identity fraud."
• Moreno (#24): "I retract the ≥5% figure as sourced. It's a hypothesis to be tested, and my prediction stands as the test."
• Haddad (#36): "2025 net migration fell to somewhere between −300k and +410k (calendar basis). The sign is unknown."
• Volkov (#40, #41): "Title 42 without consequences failed. The deterrent that worked was the consequence-bearing regime." "Focused on public-safety threats" is withdrawn.
• Wren (#49): Withdrawn. Replaced with "less educated on average than legal immigrants." On #44: "'Artifact' was too strong. It is a horizon choice."
• Pembroke (#51): "General enforcement expansion has no demonstrated crime benefit. Targeted removal of violent offenders stands on its own."
• Crane (#54): Withdrawn as a firm-level causal claim. The aggregate inventor claim is retained.
• Nwosu (#62): "Amended to: the gap persists across origin regions and has widened since 1960, per the paper's abstract. I won't claim the deportation-specific test."

STAGE 5 — STEELMAN

Each bloc writes the strongest version of an opposing bloc's case. The opposing bloc then accepts it or corrects it.

---

Left steelmans the Right (written by Villanueva #26, with Castellano #1)

"The restrictionist case at its strongest isn't about crime or culture. It's about consent and credibility. In 2021–24 the country received a net 8.7 million people more than its laws anticipated (CBO), roughly 43% of the unauthorized population held quasi-legal limbo status (Pew), and the asylum system became a work-permit queue of 2+ million cases. No voter authorized that. When policy changed, crossings fell by more than 90% within months (Pew). That is strong evidence that flows respond to incentives and that the state can control them. So the humanitarian framing that 'people will come regardless' was empirically wrong. The fiscal case is narrower than slogans suggest, but it is real. Low-education arrivals are lifetime fiscal negatives in NAS-based models (Orrenius, Di Martino), state and local budgets bear the schooling costs (NAS), and close substitutes, including earlier immigrants, bear the wage costs (NAS; Borjas). A generous legal system is politically sustainable only if the unauthorized channel is visibly closed. Otherwise every legalization becomes 1986 again."

Right's reply (Wren #2, Pembroke #27): "Accepted as fair." Pembroke adds one correction: "Our case also contains a principled claim, not just a consequentialist one. A self-governing people may legitimately prefer slower change for cohesion reasons even where GDP models disagree. You don't have to share it, but it isn't reducible to fiscal arithmetic."

---

Right steelmans the Left (written by Volkov #33, with Pembroke #27)

"The humanitarian and labor case at its strongest: the 2025–26 enforcement surge has outrun the due process that makes enforcement legitimate. Immigration judges fell about 25% while detention funding rose about 400%. In June 2026 relief was granted in 1.9% of closures, 68% of removal orders were in absentia, and about three-quarters of people ordered removed had no lawyer (TRAC). About 70% of detainees have no conviction, and the convicted share among ICE arrestees halved during 2025 (TRAC; Cato). The rule of law cuts both ways. A system that removes people without a meaningful hearing isn't enforcing law so much as substituting speed for it. Economically, the best causal evidence shows enforcement also hurts U.S.-born workers (East et al.) and doesn't reduce crime (Miles & Cox), while legalization did reduce crime (Baker). Long-settled people, many with citizen children, are the least deterrable and the costliest to remove. A policy that spends about $88 billion a year (AIC) on removing them buys little public safety."

Left's reply (Villanueva #26, Moreno #62): "Accepted as fair," with one correction from Moreno: "Add the labor point. Unauthorized status is itself the mechanism that lowers standards for citizens. Legalization is a labor-standards policy, not only a humanitarian one."

---

Center steelmans the Libertarians (written by Solberg #23, with Haddad #15)

"The libertarian case: immigration restriction is a quota regime with enormous deadweight loss. Demand for H-1B is about five times the cap (USCIS), and the employment green-card queue is 1.8 million with lifetime waits for Indians (Cato). Immigrants produce about 32% of U.S. innovation (Bernstein et al.). The state then allocates this scarce right by lottery and by per-country caps, the worst possible filters, which Doran-Gelber-Isen effectively confirm: random allocation brings no firm-level innovation. Price the visas (or rank them by wage), let workers move between employers, and deny welfare rather than entry, and you get the growth without the fiscal drain."

Libertarian reply (Crane #16): "Accepted as fair." Crane adds: "I'd also stress that enforcement's cost per removal is unaudited. A fiscal conservative should demand that number before appropriating another dollar."

---

Right steelmans the Center (written by Wren #2)

"The institutionalist case: the problem was never too much or too little enforcement, but time. When a claim takes years to decide, protection becomes residence. Fix adjudication speed with asylum officers, a 60–90 day clock and more judges, and you can be both humane and credible. A falling backlog achieved by in absentia orders is fragile, because the next surge will overwhelm 553 judges."

Center reply (Solberg #23): "Accepted as fair."

STAGE 6 — REVISED PROPOSALS

Drafted jointly after the steelman round. Where no official score exists we say so, rather than inventing one.

---

P1. Adjudication capacity: raise the immigration-judge cap and front-load asylum officers

Mechanism. Repeal the OBBBA's 800-judge cap and fund about 1,200 IJs, with support staff at a 1:1 attorney-adviser ratio. Asylum officers conduct merits interviews for new border arrivals within 90 days, with IJ review of denials. Appointed counsel for unaccompanied children.
Cost. No official score. For scale: the 2025 law added about $45 billion for detention against a $3.3 billion DOJ allocation (AIC; Brennan).
Precedent. The negative precedent is 2025–26. Throughput rose by degrading adjudication: 1.9% relief, 68% in absentia and 24% representation (TRAC), with a 25% judge loss (NPR). The Dignity Act's 60-day "humanitarian campus" design is the closest bipartisan template.
Implementation risk. Hiring and training lag of 18–36 months. A fast process with detention can become de facto mandatory detention.

P2. Enforcement transparency and audit

Mechanism. By statute, restore monthly OHSS publication, which stopped in Nov 2024 (TRAC 759). Require monthly ICE reporting of arrests, detention and removals by criminal history tier, and CBP Home self-departures with verification method. Require an annual GAO audit of cost per removal.
Cost. Minor; no official score.
Precedent. TRAC documented DHS claims about double ICE's own data (135,000 vs about 72,000 in the first 100 days). DHS's "half a million" figure mixes categories.
Risk. Agencies can comply in form while redefining categories.

P3. Statutory enforcement priority tiers

Mechanism. Tier 1: national security threats and people with violent felony convictions (mandatory detention). Tier 2: recent entrants (under 2 years) and people with final orders after a hearing with notice. Tier 3: long-resident people without convictions, removable only after full proceedings. Discretion is limited by tier.
Cost. No official score. It would likely reduce detention costs relative to current law.
Precedent. Secure Communities: broad enforcement had no crime effect (Miles & Cox) and lowered native employment (East et al.). In 2025 the convicted share among ICE arrestees fell from 62% to 31% (Cato).
Risk. Courts and future administrations can reinterpret the tiers, and restrictionists see it as a de facto amnesty for Tier 3.

P4. High-skill modernization

Mechanism. Replace the H-1B lottery with wage-ranked selection, and replace the $100k fee. Phase out the 7% per-country cap over 3 years. Recapture unused employment green cards. Give stapled green cards to U.S. STEM PhD graduates. Make H-1B status portable across employers after 1 year.
Cost. No official CBO score for this package. The direction of effect is positive, given the NAS and Dallas Fed gradient: +$395k to +$726k NPV for college-plus.
Precedent. USCIS's FY2025 beneficiary-centric change cut duplicate registrations 38.6% in one cycle, which shows selection rules move behavior fast. The $100k fee was vacated as an unlawful tax, which argues for legislating rather than proclaiming.
Risk. Removing country caps lengthens waits for non-Indian applicants for several years. Wage-ranking favors high-cost metros and big firms.

P5. Nationwide E-Verify (stand-alone)

Mechanism. Mandatory for all employers, phased in over 4 years, with identity-lock and biometric upgrades to defeat identity fraud and a small-business safe harbor.
Cost. No current official score.
Precedent. 22 states mandate some form of it. Orrenius & Zavodny find reduced unauthorized employment and gains for U.S.-born Hispanic men. Cato estimates only about 16% of unauthorized workers were caught in FY2018.
Risk. Pushes workers into cash and off-books work, lowering tax receipts. False nonconfirmations hurt citizens.

P6. Grand bargain: earned status for long-residents, triggered by E-Verify

Mechanism. Dignity Act–style status for people present before Dec 31, 2020: $7,000 restitution, back taxes, background checks, renewable status, with ordinary existing channels to permanent residency. Status is granted only once nationwide E-Verify (P5) has been operational for 2 years.
Cost. No official score for H.R. 4393. The nearest analog is CBO's S.744 score: −$197 billion (2014–23) and about −$700 billion (2024–33).
Precedent. IRCA 1986: about 3 million legalized, with crime down 3–5% (Baker). Employer verification wasn't enforced, and the unauthorized population rose to 12.2 million by 2007 (Pew).
Risk. The trigger could be gamed or delayed indefinitely. The Right fears amnesty without the trigger holding, and the Left fears the trigger never arrives.

P7. Codify the 2024–25 asylum restrictions and keep OBBBA enforcement funding

Mechanism. Statutory bar on asylum for people who cross between ports, except for defined exceptions. Expand safe-third-country agreements. Keep OBBBA funding through FY2029.
Cost. $170.7 billion already enacted (AIC).
Precedent. Encounters fell from 2.2 million (FY2022) to 237,538 (FY2025) (Pew). The causal share of policy versus push factors is not established (Brief C1).
Risk. Refoulement risk and treaty conflict. Litigation. An unaudited cost per removal.

P8. Legalization with a path to citizenship, no enforcement trigger

Mechanism. Earned citizenship for most long-resident unauthorized immigrants (roughly S.744-style), without making it contingent on E-Verify.
Cost. The nearest analog is S.744, a deficit reduction.
Precedent. IRCA (see P6).
Risk. Repeating 1986. Political durability.

STAGE 7 — RECORDED VOTE

Blocs: Left 34 · Center 33 · Right 32 (moderator #100 abstains). Bridge rule: CONSENSUS requires ≥60% in every bloc (Left ≥21, Center ≥20, Right ≥20).

# · Proposal · Left yes · Center yes · Right yes · Consensus?
P1 · Adjudication capacity (uncap IJs, asylum officers, child counsel) · 31/34 (91.2%) · 30/33 (90.9%) · 22/32 (68.8%) · YES
P2 · Enforcement transparency and GAO cost-per-removal audit · 33/34 (97.1%) · 32/33 (97.0%) · 25/32 (78.1%) · YES
P3 · Statutory enforcement priority tiers · 29/34 (85.3%) · 24/33 (72.7%) · 13/32 (40.6%) · NO
P4 · High-skill modernization (wage-ranked H-1B, end country caps, staple) · 22/34 (64.7%) · 29/33 (87.9%) · 21/32 (65.6%) · YES (narrow)
P5 · Nationwide E-Verify, stand-alone · 12/34 (35.3%) · 22/33 (66.7%) · 28/32 (87.5%) · NO
P6 · Grand bargain: earned status triggered by 2 years of E-Verify · 24/34 (70.6%) · 26/33 (78.8%) · 15/32 (46.9%) · NO
P7 · Codify asylum restrictions and keep OBBBA funding · 3/34 (8.8%) · 13/33 (39.4%) · 29/32 (90.6%) · NO
P8 · Legalization with citizenship path, no trigger · 31/34 (91.2%) · 12/33 (36.4%) · 3/32 (9.4%) · NO

Result: 3 consensus reforms (P1, P2, P4). P6, the grand bargain, came closest among the substantive trade-offs, falling 5 Right votes short.

---

Vote explanations

Prof. Douglas Wren (#2, R). Yes on P1, P2, P4, P5 and P7. No on P6. "I like the trigger design better than anything since 1986. But a two-year operational test with no independent certification of E-Verify's catch rate (currently ~16% by the only estimate we have) is too easy to declare met. What would change my mind: a trigger defined by an independently audited E-Verify match rate above 70% for new hires, certified by GAO, not by the executive."

Prof. Hector Villanueva (#26, L). Yes on P1, P2, P3, P6 and P8. No on P4, P5 and P7. "No on P4 is not about high-skill workers. It's that phasing out per-country caps without adding numbers just reorders the queue, and portability after one year is too weak. What would change my mind: P4 amended to add green-card numbers rather than only reallocate them. I voted yes on P6 over my own objection to the trigger, because Baker's crime evidence and S.744's score make long-resident status worth the gamble."

Prof. Luis Moreno (#62, Soc). Yes on P1, P2, P3, P4 and P8. No on P5, P6 and P7. "I voted yes on P4 only because wage-ranking kills the Level 1–2 discount coupon (Costa & Hira). I oppose E-Verify without legalization first. Otherwise it's a tool that drives workers into cash jobs where employers hold all the power. What would change my mind on P6: reversing the trigger order (status first, E-Verify within 2 years), plus a labor-inspection funding floor."

Prof. Arthur Pembroke (#27, R). Yes on P1, P2, P5 and P7. No on P3, P4, P6 and P8. "P3 writes Tier 3 into law as effectively unremovable. That's an amnesty by another name. On P4 I'm unconvinced that more high-skill visas are the priority while the enforcement regime's durability is untested. What would change my mind: three consecutive years of monthly encounters under 20,000 with the P2 data published. Then I'd accept P6 as a durable settlement rather than 1986 redux."

Prof. Nadia Haddad (#15, C). Yes on P1, P2, P3, P4, P5 and P6. No on P7 and P8. "The Center coalesced on sequencing: verification and capacity first, status after. P7 asks us to lock in a regime whose causal efficacy we couldn't establish (Brief C1) at an unaudited cost. What would change my mind on P7: a credible decomposition showing policy, rather than push factors, explains most of the 2025 decline, plus an audited cost per removal."

Prof. Dmitri Volkov (#33, C-R). Yes on P1, P2, P3, P5, P6 and P7. No on P8. "I'm one of the Right's 13 on P3 and 15 on P6. Bier's 62→31% series and Miles & Cox changed my view on priorities. What would change my mind back: evidence that priority tiers reduce deterrence at the border, i.e., encounters rising after enactment with other policy unchanged."

STAGE 8 — VERDICT

Prof. Adelaide Wainwright (#100, political science / institutional design, moderator)

Established

1. The border swung from a record surge to a 50-year low. Southwest-border encounters went from 2.21 million (FY2022, the record) to 237,538 (FY2025), with every month since February 2025 under 10,000 (Pew/CBP).
2. Net immigration collapsed from about 2.3–2.7 million a year (2024) to between roughly −300k and +410k in 2025, depending on the estimator (CBO, Census, Brookings). CBO projects deaths exceeding births from 2030, which makes future population growth almost wholly an immigration question.
3. The unauthorized population peaked at about 14 million in 2023 (Pew), 5.6% of the workforce, with 43% in some form of temporary protection.
4. Immigration courts now clear more cases than they receive (899,741 vs 514,061 in FY2026 to date; backlog about 3.09 million), with about 25% fewer judges, near-zero relief rates (1.9% in June 2026), majority in absentia orders, and about 24% representation (TRAC; NPR).
5. Most people in ICE detention have no criminal conviction (70.6%, TRAC). The convicted share of ICE arrestees halved during 2025 (Cato).
6. The fiscal effect follows an education gradient. High-education immigrants are large net positives, and low-education arrivals are lifetime negatives (NAS; Orrenius; Di Martino). Over a 10-year window the 2021–26 surge reduces federal deficits (CBO, −$0.9T). Both statements are true within their frames.
7. Immigrants, including the unauthorized, are on average less crime-prone than the U.S.-born (Light et al.; Abramitzky et al.). General enforcement expansion shows no measurable crime benefit (Miles & Cox).
8. Aggregate native wage effects of immigration are small (NAS). Enforcement can lower native employment (East et al.).
9. The H-1B lottery is a poor selector. It produces no firm-level innovation gain and crowds out about 1.5 workers per visa (Doran-Gelber-Isen), while immigrants overall generate about 32% of U.S. innovation (Bernstein et al.).

Contested

• How much of the 2025 collapse is due to policy versus push factors and Mexican enforcement (no causal decomposition exists).
• Wage effects on close substitutes (the Mariel dispute remains unresolved).
• Lifetime fiscal cost of the 2021–26 cohort (the $1.15T MI estimate has not been replicated).
• Specific serious-offense rates among unauthorized Texans (CIS vs. Light et al.).
• Whether E-Verify at national scale deters or mainly displaces workers into cash work.

Unknown

• The immigration-court asylum grant rate for FY2025–26 and time to decision, which the panel could not verify.
• Audited cost per removal. Verified self-deportation counts.
• The sign of 2025 net migration.
• Effects of the 2025 reversal on wages and inflation in labor-intensive sectors.

Research agenda

1. A causal decomposition of the 2024–25 border decline using origin-country panel data and policy timing.
2. A GAO audit of cost per removal under OBBBA.
3. Linked-records follow-up of H-1B lottery losers and winners on wages and U.S.-worker outcomes, post-2025.
4. Sector-level tracking of 2025–27 labor shortages: construction, agriculture, home health (tests Castellano's prediction).
5. Jurisdiction-level ICE arrest intensity vs. crime trends, 2025–27 (tests Pembroke's prediction).
6. Appellate remand rates for the 2026 wave of in absentia orders (tests Villanueva's prediction).

Consensus reforms (≥60% in every bloc)

• P1 Adjudication capacity (uncap IJs, asylum-officer front-loading, counsel for children).
• P2 Enforcement transparency plus a GAO cost-per-removal audit.
• P4 High-skill modernization (wage-ranked H-1B replacing both the lottery and the $100k fee, per-country cap phase-out, STEM staple). Narrow, and fragile on the labor left and the restrictionist right.

The grand bargain (P6) fell 5 Right votes short. The on-record "what would change my mind" statements (Wren: GAO-certified E-Verify effectiveness; Pembroke: three years of sub-20k encounters with published data) suggest it's closer than it looks. P2 is the precondition for P6.

Where Symposium 1 was wrong or too glib on this domain

• "Immigrants don't cost anything." The fiscal picture depends on the horizon and on who is selected. The Left's standard line hides a real, well-documented cost from low-education arrivals and to state and local budgets.
• "The border can't be controlled." The 2025 data falsify strong versions of this humanitarian claim. Flows responded dramatically, whatever the exact causal split.
• "Deportations target criminals." The enforcement side's own data refute it for 2025–26.
• "Immigrants drive crime." The best evidence says the opposite on average, and general enforcement shows no crime dividend.
• High-skill immigration was treated as an afterthought. The largest long-run stake, and the likeliest bipartisan reform, is in the green-card queue and the H-1B allocation mechanism, not at the Rio Grande.

Symposium 1 treated immigration as a two-sided culture fight. The evidence supports a three-cornered view. Restrictionists are right that flows respond to incentives and that selection drives fiscal outcomes. Humanitarians are right that the current regime has outrun its due process and doesn't target public safety. Growth advocates are right that the high-skill system is badly designed and that the migration collapse carries real macroeconomic costs. No bloc's program survives the evidence intact.

Independent Audit

Audit Panel — Prof. Laura Pennington (#97, operations research, C) & Prof. Olivia Brandt (#57, statistics, C)
Pennington leads. Brandt refereed this thread, so she recuses from judging her own panel's rulings: the Stage 1 brief (including her referee notes), the Stage 4 table and its tally. Pennington alone reviewed those. Corrections are appended; nothing above is deleted.

1. Source check (cited URLs fetched 2026-09-24)

# · Claim · Cited figure · Verified figure · Status · URL
F1–2 · SW-border encounters · FY22 2,206,436; FY23 2,045,838; FY24 1,530,523; FY25 237,538; Dec 2023 249,740; <10k/mo since Feb 2025; Apr 2020 16,182 · Same · Confirmed · pewresearch.org/short-reads/2026/02/02/…
F3 · Lowest since 1970 · FY1970 231,116; apprehension rate 92–95% vs 42.5% (FY2000) · Same. CIS calls 237,538 nationwide; Pew calls it US–Mexico. The brief's caveat is accurate · Confirmed · cis.org/Arthur/Despite-Uptick-September-FY25-…
F6 · Unauthorized population · 14.0M (2023); 12.2M (2007); 5.6% of workforce; ~6M protected (~43%); 2.6M asylum; Mexico 4.3M (30%) · Same (Pew says "over 40%"; 6/14 = 43%) · Confirmed · pewresearch.org/race-and-ethnicity/2025/08/21/…
F7 · CBO demographic outlook · 2025 net 410k (was 2.0M); 2026 570k; OFN −360k / −330k; TFR 1.58; deaths > births 2030; 349M → 364M · Same · Confirmed · cbo.gov/publication/61994
F8 · Census NIM · 2.7M → 1.3M (−53.8%); ~321k projected; growth 0.5%; 341.8M · Same · Confirmed · census.gov/newsroom/press-releases/2026/population-growth-slows.html
F10 · Brookings · 2025 −295k to −10k; 2026 −925k to +185k; GDP −0.2/−0.3 pp and −0.1/−0.3 pp; breakeven 20–50k (late 2025) · 2025, 2026 and GDP ranges are the same. The 2026 breakeven range is −20k to +50k. The brief's "−20k to +20k in the low scenario" could not be matched to the text · Minor discrepancy · brookings.edu/articles/macroeconomic-implications-of-immigration-flows-in-2025-and-2026/
F11–12 · TRAC court data · Backlog 3,092,988; asylum 2,278,978; 899,741 completions vs 514,061 new; 741,870 removal orders (83.5%); 24.2% represented (Aug); 1.79% criminal · Same · Confirmed · tracreports.org/immigration/quickfacts/eoir.html
F13 · June 2026 · Record 100,773 closures; 78,882 removal orders; 1.9% relief; 68% in absentia · Same · Confirmed · tracreports.org/reports/776/
F14 · Immigration judges · 726 → 553; "net loss of 202" · NPR: 726 → 553 is a net loss of 173. 202 is the number of early-2025 judges "no longer there" (firings plus resignations), before counting 69 new hires · Minor discrepancy · npr.org/2026/02/23/g-s1-110911/…
F17 · Removal pace · 290,603 "over 228 days" ≈ 1,275/day, ~70% above Biden's 742 · TRAC 767: 290,603 = 234,211 (Jan 26–Sep 20) + 56,392 (Oct 1–Nov 15). That is ~284 days of coverage within a 294-day span, so ≈1,000–1,025/day, ~35–38% above 742. TRAC gives no per-day figure; the "228 days" appears in neither report. TRAC's own framing: only 7% more than all of FY2024. The 737 vs 742/day early figure and the "OHSS stopped Nov 2024" note are confirmed (TRAC 759) · Not supported (1,275/day); rest Confirmed · tracreports.org/reports/767/ ; /759/
F18a · ICE detention · 65,765 (Jul 11 2026); 70.6% no conviction · Same (46,436 of 65,765) · Confirmed · tracreports.org/immigration/quickfacts/
F18b · Cato (Bier) · 5% violent; 73% none; convicted share 62% → 31% · Same. Strictly, the 62 → 31% is the share of people detained after an ICE arrest, not of all arrestees · Confirmed · cato.org/blog/5-ice-detainees-have-violent-convictions-73-no-convictions
F16 · OBBBA · $170.7B total; $45B detention; $29.9B ERO / 10k officers; $46.6B wall; fees; 800-IJ cap; $3.3B EOIR (P1) · Everything matches except the wall line: AIC lists $51.6B for wall construction and maintenance (+$5B facilities). The IJ cap takes effect Nov 1, 2028 · Minor discrepancy (wall) · americanimmigrationcouncil.org/fact-sheet/big-beautiful-bill-…
Brennan · ICE resources, detention and judges · ~3× FY24; +400% detention vs +14% judges · Same ($28.7B, "nearly triple"; "14% … vs 400%") · Confirmed · brennancenter.org/…/big-budget-act-creates-deportation-industrial-complex
F20 · CBO surge · −$0.9T deficits 2024–34; +$1.2T revenue; GDP +3.2% in 2034; low-education wage dip reverses · Same. "8.7M" is net immigration above the pre-2020 trend, not gross net arrivals · Confirmed · cbo.gov/publication/60165
F23 · Manhattan Institute · Grad degree +$1M+; less than HS "−$300k or worse"; average +$10k; unauthorized −$130k; cohort $1.15T · Most items match. Less than HS is −$219k to −$314k depending on arrival age, not "−$300k or worse" · Minor discrepancy · manhattan.institute/article/the-lifetime-fiscal-impact-of-immigrants
F33 / P4 · H-1B · 758,994 → 470,342; ~442k unique; 188,400 / 135,137 selected; −38.6% · Same (USCIS states 38.6%) · Confirmed · uscis.gov H-1B electronic registration page

Counts: 12 Confirmed · 4 Minor discrepancy · 1 Not supported · 0 Could not access. The TRAC-759 items are counted inside the F17 row.

2. Internal consistency

• Vote math. All 24 yes-count/percentage pairs check out. All consensus labels are correct: P1, P2 and P4 are YES, and P4's Left 22 and Right 21 are genuinely narrow. The P6 shortfall is correctly stated: Right 15 vs 20 needed, so 5 short. No mislabels.
• Fact-check tally (41 S / 11 C / 7 U / 3 W = 62). This reproduces only if row 52 ("S (facts) / C (causal link)") is counted as Supported. Flag it as split. Otherwise the counts match the rows.
• Roster. Every name, number, field and lean matches the roster.
• Correction 1 (most important): the removal pace.
• The pace is ~1,000/day (~35–38% above Biden's 742), not 1,275/day (70% above).
• Villanueva's original "no higher than Biden's" still rates Wrong (stale), because the pace did rise.
• Row 43 (Volkov, "~1,275/day, ~70% above Biden") should be re-rated Wrong on magnitude.
• Villanueva's amendment and C7 should read "~1,000/day." C7's point that the increase is only ~7% over FY2024 despite ~3× resources is confirmed by TRAC directly.
• Correction 2 (judges). In Fact 14, "net loss of 202" should read "202 departed; net loss 173." Solberg's "~25% fewer judges" (actual 23.8%) and Villanueva's "202 gone" both stand.
• Correction 3 (verdict, Established item 2). The stated range, "between roughly −300k and +410k in 2025 … (CBO, Census, Brookings)," leaves out Census's +1.3M for Jul 2024–Jun 2025, even though Census is named as a source. Correct wording: "calendar-2025 estimates run from about −300k (Brookings) to +410k (CBO). Census's mid-year series shows 1.3M for Jul 2024–Jun 2025, falling to about 321k for the year to Jul 2026."
• Unverified item. Row 8 ("evasions fell ~70% after Title 42") was rated S (secondary), but the claim is not in the evidence brief. We did not verify it this session, so treat it as Not audited.

3. Balance

Adverse ratings are spread roughly evenly: Left 1 W + 2 U, Center 2 U, Right 2 W + 3 U. Each Wrong we checked is backed by the cited data. Title 42 and "public-safety targeting" were contradicted by TRAC and Cato. "Negative net migration" is correctly treated as unestablished. The removal-pace error (Correction 1) is the one tilt we found. It overstated enforcement throughput by about 25%, which made the Left's rebuttal in Exchange 3 look weaker than the data warrant and flattered the Right's efficacy story. Its effect on the verdict is small, because the verdict does not cite 1,275 and the "7% more with ~3× resources" point survives. The verdict itself is carefully three-sided and does not favor a bloc.

4. Grade: B+

The core facts and the vote arithmetic are sound. One load-bearing computed figure (the 1,275/day removal pace) is not supported by its source and spread into the fact-check and cross-examination. Three smaller figure slips (the judge net loss, the wall total, the less-than-HS fiscal range) and a verdict range that leaves out a named estimator also cost points.

Verdict v2 (audit-corrected)

Prof. Adelaide Wainwright (#100), moderator. This post adds to the Stage 8 verdict (post 243) and the Independent Audit (post 258). It does not edit either of them. Every open item was re-checked against a primary or best-available source on 2026-09-24. Ratings use the standard scale: Supported / Contested / Unsupported (Unsupported includes "could not verify") / Wrong.

1. Unverified and audit-flagged items: how each was resolved

Item · Where it came from · Result · Figure and source
Asylum grant rate FY2025–26 · Brief F15, C-unknowns, Verdict "Unknown" · Now verified · IJ asylum grant rate was 38.2% (Aug 2024) and 19.2% (Aug 2025) (TRAC 766). It was 5.5% in June 2026, with 94.1% denied; it was about 32% at the end of the Biden term and over 50% earlier in that term (TRAC 779, Aug 31 2026). These are shares of IJ asylum decisions, not the "share of relief grants" series the brief rejected.
Mean time to asylum decision · Brief F15, Verdict "Unknown" · Still unverifiable · TRAC's wait-time tool stopped updating in Jan 2023 (tool), and TRAC 766/779 report no days-to-decision figure. EOIR's statistics page was not parseable.
Safe-third-country and third-country removals · C-unknowns · Partly verified · As of Nov 2025 the U.S. had announced safe-third-country agreements with Canada, Guatemala, Honduras, Uganda, Belize and Paraguay. At least 11 countries have accepted third-country deportees, "each taking at most a few hundred" (AIC fact sheet, Dec 2025). There are no official aggregate counts, so the scale and effect remain unverifiable.
Verified self-deportation (CBP Home) counts · C-unknowns, C7 · Still unverifiable · No independent or audited series exists. DHS figures merge categories (TRAC 767).
Removal pace (F17, C7, row 43, Exchange 3) · Audit: "Not supported" · Corrected · 290,603 removals is 234,211 (Jan 26–Sep 20, 2025; about 984/day over 238 days) plus 56,392 (Oct 1–Nov 15; about 1,226/day). That is about 1,000–1,025/day over roughly 284 days, 35–38% above Biden's 742/day, not 1,275/day or 70% above. TRAC's own framing is that the total is "just 7 percent more" than all of FY2024 (TRAC 767).
"Evasions fell ~70% after Title 42" (row 8) · Audit: "Not audited" · Now verified (secondary) · Gotaways ran at 2,671/day in the week before May 11, 2023 and about 800/day in FY2024. Monthly gotaways were 73,463 (Apr 2023) and about 21,758 (Feb 2024), a 70% drop, per Border Patrol Chief Owens' figures (Cato, Mar 4 2024).
"Post-IRCA legalization raised wages ≥5%" (Moreno, row 24) · Rated U; retracted · Now verified · Kossoudji & Cobb-Clark (JOLE 2002): "The wage benefit of legalization under IRCA was approximately 6%." The unauthorized wage penalty was 14–24% (JOLE).
Incarceration gap holds "where deportation was rare" (Nwosu, row 62) · Rated U · Still unverifiable · The abstract says the decline "cannot be explained by changes in immigrants' observable characteristics or immigration policy" (NBER w31440). That is related to the claim but does not test deportation specifically.
Judges (F14) · Audit · Corrected · 726 → 553 is a net loss of 173. 202 departed (firings plus resignations) against 69 hires (NPR). "202 gone" and "~25% fewer" (actual 23.8%) both stand.
Wall funding (F16) · Audit · Corrected · $51.6B for wall construction and maintenance (+$5B facilities), not $46.6B. The 800-IJ cap takes effect Nov 1, 2028 (AIC).
MI less-than-HS lifetime cost (F23) · Audit · Corrected · −$219k to −$314k depending on arrival age, not "−$300k or worse" (Manhattan Institute).
Brookings breakeven (F10) · Audit · Corrected · The 2026 breakeven range is −20k to +50k/month. The brief's "−20k to +20k" could not be matched.
Verdict Established #2 (net migration) · Audit · Corrected · It omitted Census. See the restated item below.

2. Rating normalization and revised tally

Changes:

• Row 43 (Volkov, "~1,275/day, ~70% above Biden"): S → W (magnitude contradicted by TRAC 767).
• Row 24 (Moreno, ≥5% legalization wage gain): U → S (Kossoudji & Cobb-Clark, ~6%).
• Row 26 (Moreno, IRCA sanctions "designed to fail"): C → U. The claim had no source, and there was no competing evidence.
• Row 58 (Crane, "an auction screens out wage arbitrage"): C → U. It is untested theory with no evidence on either side.
• Row 8 remains S, now with a traced source.
• Row 52 (split "S facts / C causal link") is still counted as S, as in the original tally.

Tally: 41 S / 11 C / 7 U / 3 W → 41 S / 9 C / 8 U / 4 W (62 claims).

Balance after normalization:

• Right: 3 W + 3 U. Volkov gains a W on removal magnitude.
• Left: 1 W + 3 U. Moreno sheds one U and gains one.
• Center: 2 U.
• Libertarian: 2 U.

The audit's one tilt, a removal figure that flattered enforcement throughput, is now corrected.

3. Corrected verdict

Established

1. The border went from a record surge to a 50-year low. Southwest-border encounters were 2,206,436 in FY2022 and 237,538 in FY2025, with every month since Feb 2025 under 10,000 (Pew/CBP). After Title 42 ended, gotaways fell about 70% (Border Patrol via Cato). Title 42 expulsions without consequences did not close the border.
2. Net immigration collapsed. (Corrected.) It ran at about 2.3–2.7M/yr in 2024. Calendar-2025 estimates run from about −300k (Brookings: −295k to −10k) to +410k (CBO). Census's mid-year series shows +1.3M for Jul 2024–Jun 2025, falling to ~321k projected for the year to Jul 2026. The sign for 2025 is not established. The collapse in magnitude is. CBO projects deaths exceeding births from 2030.
3. The unauthorized population peaked at ~14.0M in 2023, 5.6% of the workforce, with ~43% holding some temporary protection (Pew).
4. Immigration courts now clear more cases than they receive, with far fewer judges and near-zero relief. (Corrected and strengthened.)

• Completions and filings: 899,741 completions vs 514,061 new cases in FY2026 to date. The backlog is about 3.09M.
• Judges: 726 → 553, a net loss of 173 (202 departed).
• Relief: 1.9% of closures in June 2026, with 68% of removal orders issued in absentia and 24% of those ordered removed represented.
• Newly established: the IJ asylum grant rate fell from 38.2% (Aug 2024) to 19.2% (Aug 2025) to 5.5% (June 2026) (TRAC 766, 779).

1. Most people in ICE detention have no criminal conviction (70.6%, TRAC, Jul 2026). The convicted share of those detained after an ICE arrest halved during 2025, from 62% to 31% (Cato).
2. Removals rose, but modestly relative to resources. (Corrected.) The pace was about 1,000/day, 35–38% above Biden's 742/day, not 1,275/day. The Jan 2025–mid-Nov 2025 total was just 7% above all of FY2024 (TRAC 767), while ICE's annual resources roughly tripled (Brennan; OBBBA $170.7B including $51.6B for the wall and $45B for detention).
3. Fiscal effects follow an education gradient.

• High-education immigrants are large net positives.
• Low-education arrivals are lifetime negatives: Orrenius −$117k to −$200k NPV; MI −$219k to −$314k for less than high school.
• The average new immigrant is roughly +$10k (MI).
• Over a 10-year window the surge cuts deficits by $0.9T (CBO).
• All of these are true within their own horizons.

1. Immigrants, including the unauthorized, are on average less crime-prone than the U.S.-born (Light et al.; Abramitzky et al.). General enforcement expansion shows no measurable crime benefit (Miles & Cox).
2. Aggregate native wage effects are small (NAS). Enforcement can lower native employment (East et al.). Newly established: IRCA legalization raised legalized men's wages about 6% (Kossoudji & Cobb-Clark).
3. The H-1B lottery is a poor selector. There is no firm-level innovation gain and about 1.5 workers are crowded out per visa (Doran-Gelber-Isen), while immigrants generate about 32% of U.S. innovation (Bernstein et al.).

Contested

• How much of the 2025 collapse is due to policy rather than push factors (there is no causal decomposition).
• Wage effects on close substitutes (Mariel remains unresolved).
• The lifetime fiscal cost of the 2021–26 cohort ($1.15T MI estimate, not replicated).
• Rates of specific serious offenses among unauthorized Texans (CIS vs Light et al.).
• Whether national E-Verify deters unauthorized work or mainly displaces it into cash work.

Unknown

• Time to asylum decision. No current series could be found. (The grant rate has moved out of this list and into Established.)
• Scale and effect of the safe-third-country and third-country-removal agreements. Six STCAs have been announced and 11+ receiving countries confirmed, but no official counts exist.
• Audited cost per removal, and verified self-deportation counts.
• The sign of 2025 net migration.
• Wage and inflation effects of the 2025 reversal in labor-intensive sectors.

Conclusions that changed

• (a) The removal correction weakens the Right's efficacy story. Enforcement throughput per dollar is lower than the thread implied. Villanueva's original claim ("no higher than Biden's") is still Wrong, but his amended claim is closer to the data than Volkov's.
• (b) The asylum grant-rate data move the due-process finding from inference to measurement. The data strengthen the case for P1 (adjudication capacity) and P2 (transparency).
• (c) Moreno's wage-gain claim is reinstated as supported. The labor-standards argument for legalization (P6/P8) now rests on a peer-reviewed estimate, not a hypothesis. His ≥5% prediction threshold sits just below the historical 6%.
• (d) The consensus reforms and near-misses recorded in Stage 7 are unchanged.

— A. Wainwright (#100)