Does immigration help the economy?
The incoming Trump administration promised the “most spectacular migration crackdown” in US history. The Peterson Institute ran the numbers: GDP 7.4 percent below baseline by 2028, prices 9.1 percent higher. The dispute isn’t whether immigration matters to the economy — it’s who gets the gains, who eats the losses, and whether the politics will let the economics speak.
以辩论图谱查看Spectacular crackdown
“Trump will unleash the vast arsenal of federal powers to implement the most spectacular migration crackdown.”
— Stephen Miller, to the New York Times, November 2023
The clearest articulation of deportation-as-economic-policy from inside the incoming administration. Ten months later, the Peterson Institute handed back a number for what spectacular costs.
The PIIE study modeled two scenarios. The low case — 1.3 million removals over four years — put GDP 1.2 percent below baseline by 2028. The high case — 8.3 million removals, roughly the entire undocumented workforce — put GDP 7.4 percent below baseline and pushed the price level 9.1 percent higher. Construction, agriculture, hospitality, food processing, and elder care take the hit first. The mechanism isn’t mysterious: subtract several million workers from sectors that can’t substitute capital for them in eighteen months, and output falls while wage costs and goods prices rise.
This is the standard production-function logic from Economics Ch.5 (Consumer and Producer Theory): labor supply enters as a factor input, and short-run elasticities are bounded by how fast a sector can reorganize. Ch.13 (Growth Theory) handles the longer-run population and human-capital channel.
PIIE’s estimate uses a standard CGE model with a Cobb-Douglas-nested production structure $Y = A \cdot K^{\alpha} L^{1-\alpha}$, augmented for sectoral heterogeneity. The 7.4 percent contraction reflects a direct labor-input reduction of roughly 5 percent of the workforce, a sectoral-reallocation cost, and a supply-side price-level adjustment that the Fed’s reaction function partly offsets but doesn’t cancel.
If you pulled one in every twenty workers out of the 2025 economy — concentrated in the sectors that build houses, harvest food, and staff kitchens — output falls and prices rise. That’s the whole mechanism. Nothing about “will they replace native workers” comes in yet, because the question here is subtraction, not addition.
There is a version of the deportation argument that takes the GDP number on the chin. It says: the loss is real, but it falls on businesses that built their model on cheap labor the political community never authorized, and the price increase is the unwinding of a subsidy native workers were paying through wage suppression. PIIE measures the dollar cost; it doesn’t measure the distributional benefit to the bottom decile of natives competing against this workforce. That argument deserves a hearing — and it’s exactly the distributional question the next stage takes up. But the narrower point survives: mass deportation is not a costless reset. Any honest case for it has to argue that distributional benefits exceed the aggregate cost rather than pretend the cost isn’t there.
Where this leaves us
The political framing treats restriction as costless. The mainstream production-function model says you pay for it in lost output and higher prices, in measurable units, on a four-year horizon. Reasonable people can disagree about whether 7.4 percent is the right number; no one inside mainstream econ disputes the sign. Subtracting labor from the economy makes the economy smaller.
Deportation has a known cost. The implicit other side — that adding immigrants symmetrically helps — runs into a famous twenty-year fight inside labor economics. The question isn’t whether the aggregate rises. It’s whether the gains land everywhere, or come out of the paychecks of the most-substitutable native workers.
The Mariel fight
“[The Mariel boatlift] profoundly affected the labor market in Miami, depressing wages for low-skill men with less than a high school education by 10 to 30 percent... The earlier consensus that immigration had little impact on native wages rested on a methodological error.”
— George Borjas, “The Wage Impact of the Marielitos: A Reappraisal,” NBER WP 21588, 2015 (published ILR Review, 2017)
Borjas isn’t a fringe figure. Harvard Kennedy School, three decades of labor economics, and a direct attack on the famous Card 1990 result that for a generation had been the closest thing the field had to settled. The argument reopens.
The 1980 Mariel boatlift dropped 125,000 Cuban migrants into Miami over five months — a 7 percent shock to local employment, concentrated in low-skill workers. Card’s 1990 analysis compared Miami against four control cities and found essentially no effect on native wages, including for workers without a high-school diploma. The result became the canonical evidence that immigration doesn’t hurt native wages, and justified a research program (Peri, Ottaviano, Foged-Peri) built on the complementarity premise.
Borjas’s 2017 reanalysis re-cut the data on prime-age native men without a high-school diploma and found a 10–30 percent wage decline relative to controls in the years immediately after Mariel. Peri and Yasenov, replying in 2017, argue the sub-sample is too small for the gap to be statistically reliable, that pre-Mariel trends already pointed downward, and that excluding Cuban-Americans from the “native” pool creates compositional bias.
Underneath the statistical dispute is the elasticity of substitution between native and immigrant labor. Ch.5 (Consumer and Producer Theory) sets up the framework. Borjas reads the elasticity as substitution-dominated for the bottom decile; Peri reads it as complementarity-dominated even there. The deeper reason a single boatlift can’t settle the question is identification: Mariel is the textbook natural experiment, but comparison group, time window, wage measure, and subgroup definition each pull the estimate around enough to flip the sign. Ch.10 (Econometrics Foundations) is where that machinery lives.
In a two-skill production function $Y = A \, F(N, M)$, the wage effect on natives from inflow $dM$ is $\partial w_N / \partial M = A \, F_{NM}$. The sign depends on the cross-partial: $F_{NM} > 0$ (complements) gives a wage increase; $F_{NM} < 0$ (substitutes) gives a decrease. Borjas’s CES estimate puts the dropout-vs-dropout substitution elasticity high enough that substitution dominates; Peri’s keeps complementarity intact by re-categorizing imperfect substitutes inside the “dropout” cell.
If a Miami high-school dropout doing construction and a Marielito doing construction are basically interchangeable, more Marielitos means lower wages for the dropout. If they end up doing different jobs — the Marielito framing, the dropout running equipment — both wages can rise. The fight is whether the “interchangeable” case describes the actual margin.
David Card’s own restatement, three decades on:
“The Cuban arrivals weren’t perfect substitutes for American workers — they had different skills, more likely to work in industries that weren’t traditionally dominated by Americans.”
— David Card, UBS Nobel Perspectives interview, 2021
The restrictionist case at full strength doesn’t need the headline 10–30 percent figure. It needs only the claim that the prior consensus papered over a real wage effect on the most-substitutable subgroup. When labor economists report “immigration has no effect on native wages,” they are reporting an average across a population where most workers aren’t competing directly with immigrant labor. The relevant object is the conditional effect on workers who are, and that effect is plausibly negative, persistent, and large enough to matter.
A US-born man with no high-school diploma in 2024 earns less in real terms than his 1979 counterpart. Trade competition (see Is free trade always good?), declining unionization, technological change, and the disappearance of the manufacturing wage premium all contributed — the long-run native-wage-stagnation framing carried by Is inequality a problem economics can solve?. Immigration is plausibly one more. Borjas’s contribution is to insist that mainstream economics not treat the question as settled when the identification strategy required to make it look settled is the one that would also mask the effect a restrictionist would predict.
The honest Peri-Card response is not that the dropout wage effect is zero. It is small (closer to a few percent than 30), it dissipates over a decade as labor markets adjust, complementarity within the dropout cell partly offsets it, and the average native worker outside that cell gains. Those four claims survive Borjas’s reanalysis. The 10–30 percent figure does not survive the sample-size objections. What survives is the more modest finding that distributional incidence on the bottom decile is real and policy-relevant.
“The earlier consensus that immigration had little impact on native wages rested on a methodological error.”
— George Borjas, 2017
Borjas is right about the methodology — that’s why he still loses the policy question
The mainstream consensus did average the wage effect away on the subgroup where it should be most visible. That’s a real win for restrictionist econ. It is also not what the post-2017 immigration drawdown actually delivered for the workers the policy was meant to protect.
Where this leaves us
The aggregate gains from current US immigration are positive and robust. The distributional incidence is the contested layer. Estimates range from “small negative effect on native dropouts, decaying over a decade” to “substantial negative effect, persistent.” The truth sits closer to the small-and-decaying end — Peri’s sample-size and trend-control objections are serious — but the prior “essentially zero” consensus was overconfident. A real distributional effect exists; it is policy-relevant; it does not flip the sign of the aggregate.
Either way, the Borjas fight is inside a model where US immigration is between a small drag and a small boost on native wages. The other flank of the debate doesn’t care about Miami. It thinks the entire labor-economics literature is staring at a rounding error.
Trillion-dollar bills
“The gains from eliminating migration barriers dwarf — by an order of a magnitude or two — the gains from eliminating other types of barriers... For labor mobility barriers, the estimated gains are often in the range of 50 to 150 percent of world GDP. When it comes to policies that restrict emigration, there appear to be trillion-dollar bills on the sidewalk.”
— Michael Clemens, “Economics and Emigration: Trillion-Dollar Bills on the Sidewalk?” Journal of Economic Perspectives, 2011
The most-cited summary of the gains-from-migration literature, paired with Bryan Caplan running it to its open-borders conclusion. The magnitude is staggering on purpose. The Mariel fight was a few percentage points of wages for high-school dropouts in Miami. Zoom out, this frame says, and the entire developed-world labor-market literature has been computing the wrong derivative.
Clemens’s argument runs through the place premium. The same worker, with the same skills, doing the same job, earns four to fifteen times more in the United States than in their country of origin. The premium isn’t about productivity differences in the human; it’s the institutional environment around the work — capital stock, infrastructure, contract enforcement, technology. Move the worker, and the productivity moves with them. Removing labor-mobility restrictions would roughly double global output, depending on absorption parameters.
This is the gains-from-trade logic of Is free trade always good? applied to labor. Ch.13 (Growth Theory) handles the population and human-capital channel; Ch.20 (Development Economics) covers Clemens’s claim that labor mobility is the largest unrealized development intervention in the world, dwarfing aid and trade-barrier reduction by an order of magnitude. The lineage is older than Clemens — History of Economic Thought Ch.3 (Classical Political Economy) covers Smith’s argument that free labor mobility within a country is the basis of gains from specialization. The open-borders case applies the same logic at a larger scale.
The historical record is consistent. Economic History Ch.10 (Imperialism and Colonial Economies) covers the great labor reorganizations of 1850–1914 — indentured Indian and Chinese workers across the British and French empires, mass European migration to the Americas, internal African labor mobilizations — flows an order of magnitude larger as a share of population than anything post-1970. The post-1970 reopening of cross-border labor mobility is the smaller wave, covered in Ch.16 (Stagflation and the Neoliberal Turn); the migration that built modern American agriculture, construction, and food services sits inside that wave.
Bryan Caplan’s Open Borders runs Clemens’s magnitude to the conclusion the policy community has refused to draw. Every other intervention the profession spends time on — tax reform, monetary policy, industrial policy — operates where the achievable welfare gain is a few percent of GDP, optimistically. Migration liberalization operates where the gain is 50 to 150 percent of world GDP. The discipline writes the papers, puts the numbers in tenure files, then walks past them to debate whether the optimal corporate tax rate is 21 or 25 percent.
This is a claim about what gets acted on. The political-feasibility constraint that blocks engagement with the migration result is, in Caplan’s frame, itself the largest market failure in the world — a political community sitting on a welfare improvement of unprecedented magnitude and choosing not to take it. Caplan grants that absorption capacity is bounded, fiscal pressure is real, and legitimacy of large transfers requires shared institutions. What he refuses to grant is that those constraints justify the current level of restriction — one to two orders of magnitude below the welfare-optimum — rather than some larger flow.
The honest mainstream-policy response is that the welfare gains are computed in a model where receiving-country institutions are exogenous — the place premium exists because US institutions are what they are. Flows large enough to capture a fraction of the trillion-dollar number could alter the institutions that produce the premium, and then the computation changes. The open-borders position has to argue either that the institutional-survival threshold is much higher than restrictionist intuitions suggest, or that the institutional changes from larger flows are themselves positive. Caplan makes the first case; the second is contested.
Where this leaves us
The Clemens magnitude is real. The discipline genuinely computes, on its own preferred model, that current migration restrictions are the largest welfare-reducing policy distortion humans currently impose on themselves. The open-borders limit isn’t the right place to land — the institutional-survival argument is serious — but the policy-mainstream consensus is sitting much closer to the restrictionist limit than the welfare-maximizing point. The marginal product of moving the dial toward more migration, from current US levels, is high. The political economy doesn’t act on that fact.
Aggregate gains are real and large. Distributional incidence is non-trivial but doesn’t flip the sign. That should be enough to settle the policy direction. It isn’t. The reason has nothing to do with the economics and everything to do with what the political economy does with the economic facts — sometimes much worse than ignoring them.
Springfield, Ohio
“If I have to create stories so that the American media actually pays attention to the suffering of the American people, then that’s what I’m going to do.”
— Senator JD Vance, CNN interview with Dana Bash, September 15, 2024
PolitiFact’s Lie of the Year for 2024. The Haitian immigrants in Springfield were legally recruited by local manufacturers to fill a labor shortage — the exact mechanism the mainstream-econ case relies on. The hoax inverts it.
Springfield’s underlying facts are unusually clean. The town’s population had been shrinking for decades. Local employers — a forklift manufacturer, an auto-parts plant, a logistics warehouse — had openings they couldn’t fill at affordable wages. Roughly 12,000–15,000 Haitian migrants, most on temporary protected status, settled in starting around 2020. Population stabilized. Factories ran. The school district expanded. Housing tightened, the ER got busier, the city council asked for more federal support. That is the textbook signature of a labor-supply increase: production rises, native wage growth in affected sectors compresses, transition costs land on local public services, and legitimacy depends on whether those costs are funded.
Labor markets aren’t free disposal. The adjustment-cost models in Ch.5 and the comparative-institutional work covered in Ch.20 (Development Economics) agree: the gains from a labor-supply shock are bounded by the receiving community’s capacity to manage transition — housing supply elasticity, school capacity, ER funding, language services, employer screening. The literature on Springfield-like episodes (Foged-Peri on Danish municipalities, Mayda et al. on European receiving regions) is consistent: managed flows under stable institutions produce gains; unmanaged flows under stressed institutions produce backlash, even when the underlying labor-market math is identical.
Borjas gave the restrictionist-economics case its sharpest form. The labor-protectionist case deserves the same hearing — and its sharpest version is older than Trump and came from inside the left:
“Open borders? No, that’s a Koch brothers proposal. That’s a right-wing proposal, which says essentially there is no United States... It would make everybody in America poorer. You’re doing away with the concept of a nation state, and I don’t think there’s any country in the world that believes in that.”
— Senator Bernie Sanders, interview with Ezra Klein, Vox, July 28, 2015
Sanders is arguing that the economic case for open borders is a labor-arbitrage case — capital benefits from wage suppression, US workers pay for it, and the “progressive” label depends on identifying with global-welfare abstraction rather than the concrete interests of US workers. That Charles Koch funded major open-borders advocacy is, in this frame, not coincidence but giveaway. The lineage runs through the labor tradition History of Economic Thought Ch.4 (Marx) covers — the structural argument that capital and labor have asymmetric interests in cross-border factor mobility, and that frames presenting liberalization as universally good have already decided whose interests count.
The frame generalizes too far. It assumes the welfare-gain triangle accrues entirely to capital and to the migrant; the literature finds substantial native-worker gains via complementarity, sectoral expansion, and demand effects. It also assumes current restriction levels deliver what proponents claim; the post-2017 drawdown produced labor shortages, sector-concentrated wage spikes, and price increases that hit the same low-income native workers the restriction was supposed to protect. Sanders 2015 is a coherent left-restrictionist position; it is not currently delivering the best material outcomes for the workers it defends.
Springfield 2024 demonstrates the other failure mode. When a sitting senator avows the manufacture of false stories about a community whose labor stabilized a deindustrialized town, the economic facts get hoaxed past, and the policy that follows isn’t calibrated to Sanders’s labor-arbitrage worry. It is calibrated to the political-mobilization value of a constructed grievance. Those are not the same policy.
The verdict
Here is the position, stated plainly. Moderate immigration flows under stable institutions clearly help the US economy on net. Distributional incidence on the bottom decile of native workers is real and policy-relevant; it does not flip the sign, it is not large enough to justify mass removal, and it deserves direct attention via channels that actually address it — wage floors (the monopsony-and-minimum-wage apparatus of Do minimum wages cause unemployment?), sectoral enforcement, training and transition support — rather than blunt restriction that hits the same workers through the price channel.
The Clemens magnitude is the upper bound; current US policy sits closer to the restrictionist lower bound; the welfare-maximizing point is meaningfully above where US policy is now, and the discipline knows it. The Borjas-Card fight is an intra-mainstream dispute over magnitudes inside a frame that agrees on the sign. Sanders 2015 is the most-honest version of the labor-protectionist case, and it loses on its own terms once the post-2017 record is read off.
The right answer for the US in 2026: pull back from the deportation trajectory PIIE costed in Stage 1; expand the legal pathways that recruited Springfield’s workforce; rebuild absorption infrastructure — housing, schools, ERs — that determines whether gains land cleanly or translate into backlash; and treat the Borjas distributional finding as a reason to invest in the bottom-decile labor market through every channel other than the one being tried.
The pattern here — real aggregate gains, distributional incidence the political economy refuses to address through the channels that would actually help, and a measurement gap between the economy and what the political-mobilization layer claims — isn’t unique to immigration. The same shape shows up in Will AI replace jobs?, where the policy conversation can’t reach the distributional incidence on cognitive workers, and Why is housing so expensive?, where demand-side scapegoating has stood in for supply-side capacity for forty years. Three walkthroughs, one pathology: the political economy keeps treating economic facts as optional inputs to a conclusion the politics has already chosen.