The suspension of processing both prevents them from receiving legal permanent residence and, in some cases, causes them to lose their prior legal status. It also prevents them from receiving a more secure status that can prevent their arrest, detention, and removal. This is a deliberate effort to boost ICE arrests by thwarting people’s efforts to stay on the right side of the law.

Subscribe now

Cutting green card approvals in half

The agency responsible—US Citizenship and Immigration Services (USCIS)—has worked with ICE to set these legal immigrants up for arrests by failing to process their applications. Immigrants may apply for green cards—or lawful permanent residence—inside the United States, typically by converting from another temporary legal status or from refugee or asylee status. By obstructing their ability to receive green cards, USCIS degrades the rights of these applicants and may even cause them to lose their underlying status.

Over the course of the year, USCIS has cut grants of legal permanent residence by about half. The cuts have affected all categories except employment-based applicants, but USCIS has reserved the biggest cuts for the humanitarian green card categories: asylees, refugees, Cubans, and other (mostly crime victims). USCIS Director Joseph Edlow has repeatedly emphasized how closely USCIS and ICE work together. “We already have a great relationship with ICE where there is someone at any one of our offices who needs to be arrested,” he said in September last year.

Cutting green cards for family of US citizens

Start with family-sponsored immigrants, since their picture is the most nuanced. After a brief dip following Inauguration Day, the administration appears to have shifted adjudicators from other categories toward the family-sponsored cases, leading to a spike in approvals. But following the confirmation of Joseph Edlow to lead USCIS, the spike ended, and ultimately the number of approvals fell by 54 percent from July 2025 to January 2026. January 2026 approvals are now 22 percent below the January 2025 level.

Why this decline happened is more difficult to pin down than for the other categories. USCIS intentionally shed staff throughout the year, with total employment falling about 10 percent in 2025. Nonetheless, in September 2025, USCIS shifted resources into a new enforcement wing of the agency. In August 2025, USCIS stated it would increase interviews of family-based petitioners (the citizen or legal permanent resident sponsors). In November 2025, USCIS announced it would start using “country-specific” factors to weigh against approving applicants from certain countries. In December 2025, USCIS suspended all green card processing for 19 countries and extended the suspension to 40 countries in January 2026.

This suspension of green card processing allows ICE to arrest immigrants with pending green card applications. We don’t know how often this happens, but the graph below shows how the administration was ramping up enforcement, including against the targeted 40 countries, even as it slashed green card approvals. While only a small fraction of ICE arrests hit relatives of US citizens, the throttling of green card issuances has enabled ICE to arrest their family members. The delays can even cause the immigrant’s underlying status to expire.

ICE has arrested many spouses of US citizens whose status had expired but who are otherwise eligible for green cards. Ukrainian Viktoriia Bulavina—initially allowed entry legally with parole status—was detained in front of her US citizen husband at a USCIS office in December 2025 after her Temporary Protected Status had expired (though she had a pending application to extend that status as well, which USCIS had not acted on). A Swedish student who graduated from a US university, married a US citizen, and had a pending family-based green card application was also detained after her status expired. An active-duty Navy sailor’s wife was detained in the same situation. There are many other such cases.

The law permits people who entered legally to adjust to permanent residence and receive a green card even if their initial temporary status expired, but the law does not prohibit ICE from detaining them—despite their potential eligibility for a green card. KPBS reports, “ICE is transferring people arrested at green card appointments to the Otay Mesa Detention Center—a privately owned facility where it costs taxpayers approximately $200 per day to keep someone detained.”

Ending Cuban green card approvals

As bad as the situation is for family of US citizens, Cuban immigrants have faced an even more concerted targeting. In late February 2025, USCIS suspended all immigration requests, including for permanent residence, Temporary Protected Status, employment authorization documents, and other immigration benefit requests, by applicants who entered under the Cuba, Haiti, Nicaragua, and Venezuela parole processes, while canceling their underlying parole status—again to set them up for ICE arrests. Cubans were the most negatively affected because, under the Cuban Adjustment Act, they all qualified for permanent residence and green cards.

In December 2025, USCIS suspended all work on adjustments of status and other immigration benefit requests for Cubans, along with 18 other nationalities (later expanded to 40 total nationalities).

As a result of these actions, ICE is able to arrest Cuban parolees with green card applications pending. For instance, Jose Miguel Suri Hernández, who entered legally via parole in 2024, was still being detained in March 2026, despite being eligible for the Cuban Adjustment Act with an application pending. The sole reason for his detention is that USCIS is refusing to process his green card application, which has now been suspended for seven months. Although not all Cubans arrested by ICE are eligible to adjust their status, ending Cuban Adjustment Act green card approvals has certainly helped ICE to increase arrests of Cubans by 463 percent.

Ending refugee green card approvals

The most illegal and absurd collaboration between USCIS and ICE has occurred with respect to refugees. In March 2025, USCIS paused refugee and asylee green card applications for a period of about three weeks. Processing resumed at a much lower level. In December 2025, USCIS stopped virtually all approvals on refugee green cards, and in early January 2026, it announced Operation PARRIS, which it claimed would focus on reinterviewing 5,600 refugees in Minnesota who had not yet received lawful permanent residence.

In reality, in December 2025, as USCIS suspended green cards, ICE and USCIS had secretly reinterpreted an old statute to allow ICE to arrest refugees who had not yet received green cards, quietly rescinding a long-standing internal policy that forbade such arrests. Refugees cannot apply for green cards until they have been in the country for a year, and of course, by delaying the processing of these applications, USCIS allowed ICE to arrest the refugees whose legal status is unquestionable. In February 2026, ICE and USCIS director Edlow signed a joint memo permitting the arrest of refugees.

The law requires refugees to return to “custody” for “inspection and examination” after one year if they have not yet received legal permanent residence, while also prohibiting applying for legal permanent residence until one year of residence as a refugee. This means all refugees are technically exposed to arrest after one year. Rather than sending letters to these refugees asking them to appear for interviews voluntarily, USCIS sent ICE agents to their homes to drag them away from their families and send them to ICE detention facilities in Texas—not to interview them and approve or deny their green card applications, but to hold them indefinitely.

ICE shackled and took away Selamawit Mehari, an Eritrean refugee and single mother of three, in front of her children in her home. ICE sent her to Texas. Mehari and other refugees were dumped on the streets in Texas and left to find their way home. The New York Times reported on the arrest of a man from Moldova whom ICE arrested in Minneapolis and moved to Texas on January 16, 2026, despite—according to a judge’s later order releasing him—“ample evidence in support of his refugee status and his pending adjustment application.” One judge wrote about Te Eh Doh La, a Burmese refugee who entered legally and had properly filed an application for adjustment of status to permanent residence: “There is something particularly craven about transferring a nursing refugee mother out of state.”

A district court has temporarily blocked ICE from undertaking further arrests under Operation PARRIS. The court stated clearly: USCIS “could simply notify refugees of their adjustment interview and afford them the chance to attend voluntarily—summoning them only after the interview is scheduled.” ICE arrested approximately 150 refugees before the policy was blocked, but the suspension of green card issuances appears to have continued.

Conclusion

As I’ve explained before, the administration’s mass deportation strategy depends crucially on preventing immigrants from receiving or extending legal status. The mass cancellation of parole status for 1.5 million immigrants, the elimination of Temporary Protected Status for nearly a million more immigrants, the virtual elimination of asylum, and the suspension of various green card pathways have permitted ICE to increase arrests of people who would otherwise have had legal status or been protected against arrest. Removing their legal status is a necessary precondition for removing people from the country.

Subscribe now

America’s tax revenues would suffer severely from banning immigration (as President Trump is attempting to do). The primary reason that immigrants pay more in taxes than the average person is that they are far more likely to be employed than the average person. This means that even though they earn lower hourly wages, they work more total hours, so an immigrant’s per-capita earned income is higher than an American’s.

That’s why in the image below, you can see immigrants’ share of the workforce far above their share of the population. Their share of income is in the middle because they earn below-average wages per worker. However, their share of taxes and income remains disproportionate to their share of the population because there are so many more immigrant workers per capita.

In the table below, you can observe the amount of taxes paid cumulatively over 30 years by type of tax. Immigrants generated more per capita than the US-born for every type of government revenue except federal and state nontax revenues and supplemental medical insurance payments, which are tied to participation in Medicare, which immigrants use at a much lower rate.

Most taxes and other revenues (about 75 percent) generated by immigrants and the US-born are not federal, state, and local income taxes. A very significant portion of taxes is never directly paid by the person on whom the tax incidence ultimately falls. For instance, the employer portion of the payroll tax doesn’t show up on your paycheck, but it still reduces your earnings.

What about illegal immigrants? Illegal immigrants also pay taxes, directly or indirectly. We estimate that, at least before President Trump’s mass deportation campaign, illegal immigrants were complying with income taxes at about 75 percent of the rate of the average person, which was 80 percent of the required amount. This is because illegal immigrants often work with borrowed, fake, or stolen identities under which employers still withhold their taxes about half the time. They also file for refunds at much lower rates. The result is that illegal immigrants have paid about $3 trillion in taxes over the last 30 years.

All of our estimates about the tax effects of immigrants are conservative because—unlike the Congressional Budget Office (CBO)—we do not include any of the indirect effects that immigrants have on the productivity of US workers, which increases their income and tax payments. In its 2024 study, CBO found that about one-third of the revenue effect of recent illegal immigrants, other humanitarian immigrants, and their children came from these indirect effects. The CBO found that President Trump’s early 2025 immigration policy changes have already put the federal government on a trajectory toward adding $500 billion more in deficits over 10 years—primarily by lowering tax receipts.

Of course, taxes are just one side of the fiscal ledger. In our full study, we analyze how immigrants also receive fewer government benefits. The net result after accounting for benefits received is $14.5 trillion in debt reduction from immigrants, $7.9 trillion when we include the entire 2nd generation, and $1.7 trillion from illegal immigrants. To get America’s fiscal house in order, governments must address spending for US-born Americans.

It is important to note that the net effect on the government budget is an indirect consequence of the much larger economic benefits from immigration: the goods and services that they provide to Americans through their work, innovation, and entrepreneurship. If immigrants did increase the deficit, as some subset certainly does, the best solution is to further wall off the welfare state, not the country, so that Americans could still realize these economic benefits without the costs.

You can read the full study and methodology here.

Illegal entries had already fallen before he took office, and while they have fallen further, the declines in legal immigration far surpass the declines in illegal entries. The falloff in illegal immigration continued a prior trend, while the cuts in legal immigration broke the trend of rising legal immigration from 2021 to 2024.

Subscribe now

Illegal immigration had already fallen by over 80 percent before Trump.

The number of Border Patrol arrests at the southwest border had fallen by 82 percent from December 2024 to March 2025, but by December 2024, they had already fallen by 81 percent from their peak in December 2023. Of the decline from the peak, 83 percent occurred under President Biden. Prior to Trump taking office, Border Patrol releases had already dropped 96 percent from their peak. Border Patrol releases have declined further under Trump’s second term, but in absolute terms, releases fell far more before he took office. Of the decline in releases from their peak, 96 percent occurred before Trump took office.

Recorded evasions of Border Patrol—known “gotaways”—also fell about 80 percent under President Trump from December 2024 to September 2025 (the most recent data available). But again, by December 2024, evasions had fallen well before Trump took office, falling by 83 percent from their peak in November 2022. President Trump’s administration continued an earlier trend in declines in illegal immigration.

President Trump’s administration has credited the low levels of illegal immigration to an executive order that requires Border Patrol to ignore US asylum law and immediately return any crosser to their home country. In an amicus brief for a case challenging the order, I detailed some reasons to believe that this claim is likely wrong. Besides the fact that illegal immigration was already falling, I note that arrests by the one population exempt from the order—unaccompanied children—fell just as much, and just as quickly, as other arrests. Better explanations for the changes are the fear of the Trump administration and worsening economic conditions here. 

Asylum seekers entering legally fell 99.9 percent.

Every prior administration allowed some asylum seekers to enter legally at the southwest ports of entry because the law explicitly allows them to apply there. The first Trump administration terminated this practice in 2020, and the second Trump administration did so again in January 2025 by eliminating the CBP One scheduling app and banning asylum. As a result, the number of monthly southwest border legal entries by asylum seekers fell 99.9 percent from almost 40,000 in December 2024 to just 26 in February 2025. Other “inadmissibles”—Customs and Border Protection’s broad classification for undocumented immigrants, asylum seekers, and others trying to enter legally at ports of entry—also fell in 2025.

Update: President Trump shared this chart.

Refugees entering legally from abroad fell by about 90 percent.

The number of refugees—who enter legally and are granted full protection based on their fears of persecution—has also declined by nearly 90 percent. The president initially suspended the refugee program entirely before restarting it to exclusively admit white South African refugees. The drop was from 12,518 in December 2024 to 1,341 in March 2026. The cap for FY 2026 is just 7,500, so the monthly average for the rest of the year will be 500. As I’ve pointed out previously, most refugees blocked by the administration were persecuted Christians.

Immigrant visas for legal permanent residents fell by about half.

The State Department has not published the number of immigrant visa issuances since September 2025. Since then, President Trump has banned all immigrant visas for nationals of 40 countries, including the Palestinian Territories, starting on January 1, 2026. As I’ve previously described, this affected one in five immigrant visa applicants. The State Department also suspended the Diversity Visa Lottery, effective December 23, 2025. Meanwhile, in a three-sentence press release, the State Department suspended immigrant visa issuances for 75 countries (for a total of 92 banned countries) on January 21, 2026. 

As I’ve detailed before, these bans affected half of all legal immigrants coming from abroad, including half of all spouses and minor children of US citizens, based on 2024 immigrant visa processing. However, it’s possible some of this flow could be replaced with immigrants from other countries, but that did not happen when President Trump enacted a narrower ban on certain categories of immigrants from 19 countries in June. Immigrant visa issuances fell for those nationalities, but rather than other countries increasing to compensate, visas for other nationalities also declined.

Theoretically, the State Department could resolve this issue for the new ban, but we should question how quickly. Moreover, only about a third of the visas are employment– and family-based capped categories that would even theoretically be available for reallocation. If these were reallocated, the immigrant visa cut would be 43 percent from December 2024. This projection also assumes a cautious 1‑percent cut per month to account for the pre-ban downward trend and for all other policies intended to reduce issuances, including the new health-related rules that debuted in November 2025.

Visas for fiancé(e)s and spouses of US citizens fell by 65 percent.

Visas for fiancé(e)s and spouses of US citizens were down by at least half by summer 2025, compared to December 2024. It’s not obvious how seasonal these visas are, but the year-over-year decline was 65 percent when compared with the summer of 2024. The 40-country ban—that started in January 2026—also affects them, excluding about 3,000 visas annually, but because the 75-country immigrant visa ban does not, it’s possible that at least some spouses will eventually try to shift from immigrant visas to K‑1 visas. This will not be possible immediately, as restarting the process would be too costly and time-consuming.

International student visas fell by 40 percent last summer. 

The Trump administration has enacted specific policies targeting international students. In January 2025, President Trump signed an executive order that led to the cancellation of F‑1 status and revocation of between 1,700 and 4,500 student visas between January and April. The administration also arrested and placed in detention several students for their political speech. In May, the administration attempted to suspend the right of Harvard to enroll international students at all, but was blocked by the courts. 

In May and early June, the State Department suspended all student visa issuances for three weeks during the peak season. In June, Trump signed an executive order banning student visas from 19 countries, which affected about 10,188 students annually, and in December, he expanded that ban to 40 countries, affecting about 23,000 students per year. Comparing the peak months when students typically get visas, student visas had already fallen by 40 percent, comparing the summer of 2024 to the summer of 2025. As the figure shows, the country ban does not play a massive role. The visa suspension last year and other policy moves mattered much more last year. What will happen this year is still unclear.

H‑1B visas have likely fallen by about 25 percent.

In the average month, H‑1B visa issuances in 2025 were already lower than in 2024. But in September, President Trump signed an executive order mandating a $100,000 fee to petition for an H‑1B worker outside the United States. There were 65,000 H‑1B approvals for workers outside of the US in FY 2024, and there were  219,659 visas issued abroad, implying that about 30 percent of visa issuances are for workers targeted by the ban. 

The State Department hasn’t been totally transparent about the effects of the fee, but in one court filing, it implies it led to an 87 percent decline in petitions for workers outside the United States. H‑1B visas had already declined somewhat by September, so a reasonable estimate would place current H‑1B visa issuances down by about a quarter. Since the fee only applies to new petitions approved after September 2025, it might take a bit longer than projected for the full effect to be felt. The chart below shows H‑1B and H‑1B dependent (H‑4) visa issuances based on this projection.

Legal entry cuts are now likely 2.5 times higher than illegal entries.

When we put together the full picture of legal and illegal immigration, we see that President Trump has indeed cut legal entries far more than illegal entries in absolute terms. The available information suggests that about 2.5 times as many legal entries are being stopped as illegal entries: 132,000 versus 50,000. Among the cuts to immigration, 72 percent were from legal entries, not illegal entries.

This comparison is misleading in favor of illegal entries because most illegal crossers caught by Border Patrol were not being released into the United States in December 2024. It also assumes that 100% of the cuts to the employment– and family-based categories from banned countries result in increases in family– and employment-based categories from other countries. These are just the major visa categories. If we included every category, the cuts would be more severe. It also does not include all the people who are already in the United States and are banned from receiving legal status or converting from temporary to permanent status.

These cuts to legal immigration are harming US citizens seeking to reunite with their spouses, fiancés, children, and other relatives. They are also undermining US prosperity and increasing the US deficit. But they also illustrate the central misconception about the Trump administration’s immigration agenda.

It is not about stopping “illegal” immigration. It is a broader assault on all types of immigration. As Americans debate the path forward on immigration, that’s a reality everyone should understand.

That conversation has stuck with me because it illustrates a lazy pattern I see in social science and policy commentary on Substack and X. Someone observes a behavioral difference between groups or countries. They can’t immediately identify the mechanism. So, they invoke “culture” as an explanation or, even worse, “the culture.” The word lands with a satisfying thud that sounds like an explanation but isn’t one. It is the terminus of inquiry, not the beginning.

In 1977, economists George Stigler and Gary Becker published a paper that should have inoculated all social scientists and commentators against the culture crutch. In “De Gustibus Non Est Disputandum,” they argued that economists should treat tastes as stable and uniform across people, then search for differences in prices, incomes, and constraints to explain behavioral variation:

The difference between these two viewpoints of tastes is fundamental. On the traditional view, an explanation of economic phenomena that reaches a difference in tastes between people or times is the terminus of the argument: the problem is abandoned at this point to whoever studies and explains tastes (psychologists? anthropologists? phrenologists? sociobiologists?). On our preferred interpretation, one never reaches this impasse: the economist continues to search for differences in prices or incomes to explain any differences or changes in behavior.

Stigler and Becker argue that invoking taste differences is the social science version of quitting. It lets you feel like you’ve explained something when you’ve just inserted a placeholder for a real explanation. The economist’s job is to keep looking for the price or incentive that affects behavior. They admitted the search is often long and frustratingly difficult, but that is the work. Do the hard work of identifying the incentives and tradeoffs. You won’t always find a satisfying answer, but whatever you find will make more sense than substituting the word “culture” for an explanation.

The Definitional Problem

Culture is human behavior that is socially learned and transmitted rather than genetically inherited or individually discovered. In Substack and online debates, culture means whatever the person invoking it needs it to mean. Values. Beliefs. Norms. Attitudes. Customs. Work ethic. Family structure. Trust. Time preference. Cuisine. Music. When someone says “culture explains X,” they’re gesturing at a black box the size of human civilization and calling the gesture a theory.

Keep that definition of culture in mind as I explain how unsatisfying using the word “culture” is as an explanation. You notice a spike in unemployment. Curious what could be causing it, you ask your economist friend why unemployment is rising. He says it’s because of “the economy” and then sits back as if he’s explained something when he has done nothing of the sort. That’s how everybody sounds to me when they say that culture explains a behavior or outcome.

Economists Luigi Guiso, Paola Sapienza, and Luigi Zingales acknowledged the problem directly in a 2006 Journal of Economic Perspectives article. The concept of culture as an independent variable in economics, they wrote, is “so broad and the channels through which it can enter the economic discourse so ubiquitous (and vague) that it is difficult to design testable, refutable hypotheses.” They think econometric advances help. They at least see the problem.

David Oks recently asked why Ghanaians and other sub-Saharan Africans spend fortunes burying their dead? “The standard answer to this is that exorbitant funeral spending is ‘part of the local culture,’ and particularly reflective of a ‘reverence toward elders.’ And indeed that is true. But that only begs the question of why it’s part of the culture.” The cultural explanation redescribes the phenomenon and calls the redescription an answer. He doesn’t fall for that trap.

If you’re going to claim that culture has an effect, you should be able to do four things. First, pinpoint exactly what cultural characteristic you mean. Don’t be vague, be specific by describing the type of behavior. Second, prove that cultural behavior actually exists as a measurable trait. Don’t rely on stereotypes, do the hard work. Third, demonstrate that the cultural behavior differs meaningfully across the groups being compared. Wow, that culture likes food a lot. Which culture doesn’t? Fourth, rule out that the real cultural trait isn’t caused by an exogenous economic force like high real estate prices, rising wages, or different institutions that incentivize behavior. Almost nobody who invokes culture does any of these four things. Culture is endogenous to everything. That’s why you have to do the work to isolate it. That’s also why almost nobody bothers.

The fourth step is the hardest because culture is endogenous to everything. It doesn’t exist outside the institutional, economic, and geographic environment that produces it. The corruption norms in Egypt didn’t fall from the sky. They emerged from decades or centuries of weak rule of law, chaos, and institutional dysfunction. Japanese cooperative norms didn’t spring from the soil of Honshu or grow from their bodies like an appendage. Claiming culture causes an outcome without first ruling out that the outcome’s causes also produced the culture is not an explanation. It is circular reasoning with a dedicated vocabulary.

The cleanest test is the divided-country natural experiment. North Korea and South Korea share a language, ethnicity, history, and culture up to 1945. One is among the richest countries on earth, the other among the poorest. East and West Germany diverged dramatically under different institutions and converged after reunification. Mainland China stagnated under Mao while Taiwan, Singapore, and Hong Kong prospered, all four sharing Chinese culture. In every case, the culture was identical on both sides of the border. The incentives, shaped by the institutions, are what changed. The outcome followed the institution, not the culture. Untangling causality is difficult, sometimes impossible, but that’s no reason to embrace a false explanation like “the culture made them do it.”

At its root, the culture discourse is anti-intellectual. Culture is a faux explanation for social behavior and outcomes that have real explanations. Think harder. Use AI to search the literature if you have to because other researchers have probably already written about the issue you claim is just caused by culture. The cultural explanation is the one you reach for when you’ve decided the search isn’t worth your time. Better to remain quiet if culture is the only explanation you’ve got. Here are some examples.

Fertility Is Not Cultural

Those who want to reverse falling fertility diagnose a cultural disease of secularism, individualism, feminism, and loss of traditional family values. They prescribe cultural medicine that is consistent with their preexisting policy positions, like pronatalist cultural messaging, traditional values campaigns, and more religion. The economic explanation for falling fertility is simpler: Opportunity cost. Women’s wages rose relative to men’s. Entertainment options multiplied. Children became more expensive relative to doomscrolling, flights, restaurants, streaming services, career advancement, sleep, and every other fun activity available to modern couples. The DINK couples on TikTok who occasionally go viral aren’t expressing a cultural stance. They’re reporting where they are in a downward-sloping demand curve.

You don’t need to assume preferences changed. The shadow price of children rose. Immigrants don’t assimilate to a low-fertility culture as my friend thought, they assimilate to high opportunity costs. If it were cultural, pronatalist campaigns would work. They don’t. The policy that works targets prices. The policy that fails targets culture.

Corruption Is Not That Cultural

Raymond Fisman and Edward Miguel’s famous 2007 paper found that UN diplomats from corrupt countries accumulated far more parking violations than those from less corrupt countries in New York City when diplomatic immunity meant no enforcement. Their paper was widely cited as proof that a culture of corruption persists across institutional environments, because what else could explain variation like that? But when New York started towing and seizing plates in 2002, violations collapsed by 98 to 99 percent across every nationality. Rutar found that even the residual correlation was fragile with alternative corruption measures, turning non-significant. That’s overwhelmingly just responsiveness to prices. Turns out prices and incentives explained it. Some small differences remained, but nothing to write home about.

Trust Doesn’t Explain Economic Development

Trust is the respectable face of culture-explains-development in economics. The trust literature claims that societies where more people answer “yes” to the survey question “Can most people be trusted?” grow richer, and that this cultural trait causally drives economic development. Andrew Forrester and I identified five problems with this literature: vague definitions, endogeneity, sample bias, unspecified macro models, and a weak relationship between survey trust and actual trusting behavior. We tested whether trust predicts regional U.S. development across nine regions over 45 years and found that it didn’t. If trust doesn’t survive a within-country test, the cross-country correlation likely reflects the institutions that produce both trust and growth.

Japanese Trains Are Not Cultural

Japanese trains are punctual and heavily used. The standard explanation is culture, but the real explanation is policy and incentives. Private, vertically integrated railway companies own real estate, retail, and hospitals along their lines. Combined with high population density, liberal zoning, privatized parking (23 spaces per hectare in Tokyo versus 263 in Los Angeles), toll-funded motorways, and other legal differences, there is nothing cultural to explain Japan’s use of trains.

Funerals and the Kinship Tax

David Oks identified a fascinating puzzle. A modest funeral in Ghana costs $5,000 in a country where the median income is $1,500. A befitting one runs $15,000 to $20,000. And this is common in sub-Saharan Africa. Oks writes:

The standard answer to this is that exorbitant funeral spending is “part of the local culture,” and particularly reflective of a “reverence toward elders.” And indeed that is true. But that only begs the question of why it’s part of the culture—in fact, why it’s part of so many distinct cultures across very different parts of Africa. Why is heavy funeral spending such a pronounced part of life? And if this is merely a reflection of “reverence toward elders,” why do so many elderly Africans complain that far more attention is given to their funeral than to caring for them while they’re alive?

Oks asks the right questions and identifies some real explanations based on incentives created by kinship-based networks. Kinship networks are mutual aid societies you’re born into and can’t leave. Members claim everything above subsistence. Individual wealth threatens the network because a rich member can defect, so the funeral is ritualistic wealth destruction, a public signal of loyalty. The expense is the point. As Oks puts it, “Social modernity, in the end, is really about not having to do what your family tells you to do—marrying whom you want, taking the job you want, and spending your earnings the way you want.” The funeral isn’t a cultural mystery, it’s the result of bad incentives.

The Graveyard of Famous Cultural Explanations

Max Weber’s Protestant ethic is the most famous cultural explanation in social science. Protestantism supposedly instilled a psychological orientation toward capital accumulation that launched modern capitalism. Economist Davide Cantoni tested this with a natural experiment using the 1555 Peace of Augsburg, which assigned Protestantism or Catholicism to German cities based on their ruler’s religion. Protestant regions didn’t develop faster.

Economists Sascha Becker and Ludger Woessmann identified the actual mechanism that fooled Weber and others into thinking that Protestantism increased economic development. Luther promoted universal literacy so Christians could read the Bible. The economic advantage was human capital, not theology, at least in Prussia. And even the Prussian result is challenged by Kersting and coauthors, who showed that the apparent gap in Prussia disappears once you control for ethnicity. Weber confused a literacy program encouraged by a new religious belief chosen by local rulers that affected some Germans for an exogenous cultural independent variable. Tibor Rutar summarizes it well.

Ronald Inglehart and Christian Welzel argued that self-expression values, progressive cultural attitudes produced by economic development, cause democratization. But 19th-century Britain, France, and Switzerland didn’t have progressive mass values when they democratized. Once you control for endogeneity, self-expression values don’t predict democratic transitions. The culture follows the institution, not the reverse.

Joel Mokyr won the Nobel Prize for work on the prerequisites for sustained technological progress, partly for arguing that the European Enlightenment’s culture of growth explains the Industrial Revolution. If the culture of growth was Europe-wide, why did only England break out first? Institutions and maybe relative price differences. Property rights, patent law, and parliamentary constraints on the crown. The culture was common, but the institutions were particular. But couldn’t cultural characteristics help modify or select the institutions? Yes!

The dominant ideology thesis holds that workers accept capitalism because ruling-class ideology programs them to. Tibor Rutar rightly calls this a conspiracy theory and it’s easy to see why. Workers don’t stay quiet because of cultural hegemony, but because collective action is costly and the free-rider problem is real. Material incentives explain mysterious actions, not cultural programming.

Why Culture Persists as an Explanation

If a country is poor because of its culture, nobody has to examine the bad incentives facing members of that society. Intellectual laziness explains the rest. Finding the price, the constraint, the institutional mechanism that creates an incentive is hard, but invoking culture as if it’s a magical exogenous decider lets you stop searching. Cultural explanations are cheap to produce, requiring only anecdotes rather than data, prices, or evidence. It feels like an answer because it has the grammatical structure of one. “Japanese people ride trains because of their culture” masquerades as an explanation, but it’s just a tautology.

Culture is endogenous to everything. Claiming culture causes an outcome without first ruling out that the outcome’s causes also produced the culture is circular reasoning. Every cultural explanation must first survive a price, incentive, and institutional audit. Few of them do, but those that do are extraordinary findings, which is perhaps another explanation why so many claim it. Nobody would let economists get away with explaining a recession of high unemployment with the explanation, “It’s the economy.” We shouldn’t let others get away with the equally lazy non-explanation of “it’s the culture.”

These May Be Minor Concessions

The most interesting cases are the ones where culture might genuinely be an exogenous explanatory variable. Nathan Nunn and Leonard Wantchekon found that African regions hit hardest by the slave trade still show lower trust centuries later. But the slave trade targeted regions that were already politically fragmented and poor. Poor places were enslaved, so disentangling whether slavery made them poor or poverty made them enslaved is exactly the endogeneity problem this piece is about. The slave trade didn’t just change attitudes. It destroyed political institutions, displaced populations, and created economies designed for extraction. Nunn’s cultural residual may be persistent institutional damage wearing a cultural mask.

The Doctrine of First Effective Settlement holds that the institutions created by a country’s earliest settlers disproportionately shape future institutional change and affect long-run economic development. That sounds like it’s cultural, maybe. I don’t dismiss these findings, but endogeneity haunts every one of them. If someone eventually establishes that culture operates independently of the institutional environment it developed inside, that will be among the most important findings in social science. Until then, we don’t know whether we’re measuring culture or institutional scar tissue. Douglass North wrote that culture is just the partial solution to the frequently encountered problems of the past, like how a taboo against pre-marital sex evolved to prevent paternity disputes and then was rapidly undermined by birth control.

Culture can be more than just another endogenous variable. People have values. Norms vary. Trust differs across populations and those differences are real even if they don’t explain different income levels. Sometimes cultural differences are consequential, especially when the stakes are low. My position is a methodological commitment, not a metaphysical claim. Assuming stable tastes and searching for price-and-constraint explanations produces better social science than assuming taste differences and stopping there. The question is how often, how much, and whether the analyst did the work to rule out everything else first.

The culture-as-explanation discourse is largely anti-intellectual. These are faux explanations for social behavior and outcomes that have real explanations. Think harder. Read the literature on a topic yourself or ask AI to search for you. Other researchers have probably already written about the issue you claim is just caused by culture. Before you write the word “culture” in a causal sentence, search for the price. Search for the institution. Investigate the incentives. Search for the constraints. If you exhaust those and culture is still standing, then maybe you’ve found something. But you probably just didn’t look hard enough.

Using publicly available apportionment data and Treasury outlays through February, I estimate the administration has released $114 billion of $191 billion in available OBBBA funds for DHS to spend—including $33 billion to ICE and $56 billion to CBP—with roughly $77 billion still available for apportionment.

While these data can provide limited insight into how the administration plans to use OBBBA funds, it also reveals a larger problem. We don’t know whether OBBBA spending is consistent with statutory instruction or even how much has already gone out the door. Without robust transparency, Congress cannot conduct oversight, course-correct poor policy, or deter the misuse of taxpayer resources. The system of checks and balances that undergirds our constitutional republic suffers as a result.

Congress Knows What Was Appropriated, but Not What’s Been Spent

Congress has provided DHS with roughly $191 billion via the OBBBA. Of that $191 billion, Congress earmarked around $75 billion for ICE and $65 billion for CBP, the federal government’s two primary immigration and border enforcement agencies. That is seven times ICE’s annual budget and four times CBP’s typical annual budget. The next largest line item is a $25 billion appropriation for the Coast Guard. A further $10 billion is provided in Section 90007 as an effectively unrestricted DHS slush fund. Table 1 lists key OBBBA funding categories for DHS.

But knowing what Congress appropriated is very different from knowing how the administration has used these funds. As 21 senators noted in a recent letter to the Congressional Budget Office, there is no comprehensive public accounting of how OBBBA DHS funds are being obligated or spent.

In short, the branch vested with the power of the purse cannot determine how the largest immigration enforcement spending increase ever for CBP and ICE is being used.

What Has Been Apportioned?

By law, OMB must publicly release plans for how the administration intends to use available budgetary resources, called apportionments. These apportionments create a legally binding upper bound on what an agency can spend. Importantly, agencies may spend less than is apportioned, so these plans represent the ceiling of available budgetary resources, not final obligations or outlays.

Based on OMB apportionment data through February 2026, the administration has released $113.9 billion in OBBBA funds for DHS to spend—$12 billion in FY2025 (OBBBA was enacted on July 4, 2025) and $101.9 billion in FY2026. That leaves roughly $77 billion available for apportionment in FY2027 and beyond. The chart below illustrates the cumulative apportionment of OBBBA resources to DHS.

As of February, ICE and CBP were apportioned $33 billion and $56 billion, respectively. Table 2 shows apportionments for each DHS agency.

Within ICE, the single largest category is detention. More than $24.8 billion has been apportioned for ICE detention facilities across FY2025 and FY2026. That’s nearly three times the annual spending for the Federal Bureau of Prisons.

For CBP, border wall construction dominates, with at least $40 billion apportioned. Personnel, facilities, vehicles, and surveillance technology account for most of the remaining $16 billion.

What Has Actually Gone Out the Door?

Apportionments provide only a limited picture of how OBBBA funds will be used. Other data sources that could reveal how OBBBA funds have been deployed are inadequate. Some reporting, for example, has cited USAspend​ing​.gov, a government tool intended to track federal obligations down to the individual award and grant level. Unfortunately, USAspend​ing​.gov relies on agency-submitted data that is often incomplete, delayed, or inaccurate. Comparable data sources have similar problems.

The most reliable alternative is the Monthly Treasury Statement. Per Treasury, total DHS outlays in FY2026 (October 2025 through February 2026) are actually lower than the same period last year: $46.9 billion versus $60.2 billion. But that headline figure is mostly driven by a reduction in FEMA disaster relief outlays. When you look at ICE and CBP, the picture changes (see Table 3).

ICE’s FY2026 outlays through February totaled $10.7 billion. Before the OBBBA took effect, ICE averaged ~$800 million per month between October and February. Post-OBBBA, that average has risen to $1.7 billion per month.

CBP tells a different story. Despite $56 billion in OBBBA apportionments, CBP’s FY2026 outlays through February are $8.8 billion, compared to $8.5 billion in the prior fiscal period, a 4 percent increase.

This discrepancy may be due to the underlying OBBBA provisions funding these agencies. Most of the CBP funding is for border wall construction, which may involve a multi-month to multi-year lag between obligation and outlay. Meanwhile, the OBBBA seems to provide ICE with greater flexibility to spend more quickly, given that it can lease existing private facilities and has a larger budget to hire new personnel. In any case, the CBP funds will eventually appear in Treasury data, just not yet.

There is also a useful natural experiment on the horizon. Regular annual DHS appropriations lapsed in February 2026. New Monthly Treasury Statements for ICE and CBP will soon reflect almost exclusively OBBBA-funded activity, providing the cleanest available signal for measuring OBBBA outlay velocity going forward.

Shades of the Pandemic

During the pandemic, Congress approved a multi-trillion-dollar spending surge across a series of emergency stimulus bills. Waste, fraud, and abuse followed, in part because the sheer scale of spending outstripped the weak oversight capacity of both Congress and the executive branch.

By shifting an unprecedented surge in immigration enforcement spending outside the annual appropriations process, Congress has similarly short-circuited the mechanisms that normally enable accountability. OBBBA’s funding structures mean that neither congressional appropriators nor independent watchdogs can easily track how the money is being used.

Both of these funding episodes—and the associated transparency and oversight problems—are part of a broader trend of institutional erosion in Congress. Over the years, legislators have increasingly ceded spending and policy control to the executive branch. Republicans and Democrats alike should be wary of the gradual decay of the system of checks and balances that secures all Americans’ freedoms.

The Path Forward

As with the pandemic, the cleanest solution is to simply not spend the money in the first place. Ideally, Congress would rescind unobligated OBBBA funds. If Congress still wants to fund these programs, it should do so through regular discretionary appropriations and limit multi-year availability.

Short of rescissions, the Department of Homeland Security, like every federal department and agency, should release clear, detailed spending plans. Congress can compel agencies to do so by strengthening existing reporting requirements and tools, such as USAspend​ing​.gov, or by attaching targeted DHS reporting requirements in appropriations bills.

One recent bipartisan bill, the Expedited Transparency Act, introduced by Representatives Brecheen (R‑OK) and Panetta (D‑CA), would take a step in the right direction by shortening the current 30-day posting window for federal disbursements on USASpend​ing​.gov to three business days.

Regardless of what you think about the administration’s deportation campaign, all public spending decisions deserve transparent accounting. OBBBA should be no different.

We estimate illegal and legal immigrant incarceration rates in the United States by using the 2024 American Community Survey Public Use Microdata Sample from the US Census. We also provide detailed incarceration data on immigrants and native-born Americans by race and ethnicity, region of birth, country of birth, sex, education, age, the number of years present in the United States, and age of arrival.

All immigrants, both legal and illegal, are less likely to be incarcerated than native-born Americans. The 2024 native-born American incarceration rate of 1,195 per 100,000 natives is the highest of the three groups analyzed. Legal immigrants have the lowest incarceration rate, at 303 per 100,000 legal immigrants in 2024. Illegal immigrants have an incarceration rate of 674 per 100,000 illegal immigrants, higher than legal immigrants but also lower than native-born Americans.

Background

According to a recent poll by Gallup, 47 percent of Americans believe that immigrants increase crime in the United States and only 5 percent think that immigrants reduce it.¹ There’s little doubt that many of the respondents who blamed immigrants for higher crimes are specifically thinking of illegal immigrants. This briefing paper is the latest in a series that attempts to answer whether that perception is true by estimating illegal immigrant incarceration rates in the United States by using the 2024 American Community Survey (ACS) Public Use Microdata Sample (PUMS) from the US Census. The data show that all immigrants are less likely to be incarcerated than native-born Americans relative to their shares of the population. By themselves, illegal immigrants are less likely to be incarcerated than native-born Americans and legal immigrants are the least likely of all.

We published the first nationwide estimates of the incarcerated illegal immigrant population in 2017, followed by updates in 2018, 2019, 2020, and 2025.² The 2017 brief analyzed incarceration rates for 2014, the 2018 brief analyzed incarceration rates for 2016, the 2019 paper analyzed incarceration rates for 2017, the 2020 paper estimated incarceration rates using an updated methodology for the entire 2010–2018 period, and the 2025 policy analysis analyzed the 2010–2023 period. This paper updates the estimates using the most recent 2024 inmate data from the ACS.

The illegal immigrant incarceration rate is an important indicator of that population’s criminality, but different measures of criminality used in other studies are valid too.³ They generally find that immigrants do not increase crime rates in small communities, are less likely to cause crime than their native-born peers, and are less likely to be incarcerated, convicted, and arrested than native-born Americans.⁴ Immigrant criminality does vary based on whether they are legal or illegal immigrants, probably because the two groups are distinct demographically, socioeconomically, and on other margins that could make one group more or less crime-prone than the other.⁵ Illegal immigrant incarceration rates, criminal conviction rates, arrest rates, and broader impacts on crime are not well studied because most jurisdictions, law enforcement agencies, state corrections departments, and other organizations in the criminal justice system do not systematically record data on the immigration statuses of those arrested, convicted of crimes, or incarcerated. The three major exceptions are Texas, Georgia, and Oklahoma. Texas has been keeping data on arrests and convictions by immigration status since 2011, Georgia has been publishing data on illegal immigrants incarcerated in that state beginning in 2024, and Oklahoma in recent years.⁶

Cato Institute analysis of the conviction, arrest, and incarceration data in those states found that illegal immigrants had a lower criminal conviction rate and a lower arrest rate in Texas relative to the native-born population, a lower incarceration rate in Georgia relative to the non–illegal immigrant population, and a slightly lower incarceration rate in Oklahoma too. The finding held for all crimes, including the various types of homicide that are the most serious offenses.⁷ Data for the other 47 states are unavailable, so Cato scholars have estimated the illegal immigrant and legal immigrant incarceration rates for the entire country and have found it to be lower than for native-born Americans.⁸

Recent peer-reviewed empirical studies have found no link between violent crime and illegal immigration, a negative relationship between the number of illegal immigrants and most types of nonviolent crime, and lower illegal immigrant criminal conviction and arrest rates in Texas compared to other subpopulations.⁹ Results vary somewhat based on the methods and data, but there is convincing evidence that a larger number of illegal immigrants present in an area increases the rate of identity theft, and there is also some evidence of a small but statistically significant relationship between the size of the illegal immigrant population and drug arrests.¹⁰

Our estimates of a low illegal immigrant incarceration rate are consistent with other research that finds that increasing immigration enforcement and deporting more illegal immigrants does not reduce the crime rate, which we would expect to occur if illegal immigrants were more crime-prone than natives.¹¹ Our research is also consistent with work that finds crime rates either do not increase to a statistically significant extent when states create sanctuary jurisdictions that limit the scope of immigration enforcement, or that the rates for some crimes actually fall, which we would not expect to occur if illegal immigrants were more crime-prone than the rest of the population.¹² However, the difference in immigrant criminality would have to be extremely large or their population would have to be a significant share of the local area to affect the local crime rates to the point where their effect would show up. Thus, comparative arrest, conviction, and incarceration rates are better calibrated to identifying the relative criminality of immigrant subpopulations. For methodological and data limitation reasons mentioned above, these studies are not the final word on the matter.

Methodology

In this policy analysis we use ACS data to estimate the incarceration rate and other demographic characteristics for all immigrants, legal and illegal immigrants separately, and native-born Americans ages 18–54 in 2024. The ACS inmate data are reliable because they are ordinarily collected by, or under the supervision of, correctional institution administrators; however, the quality of the data for the population that includes the incarcerated was not always as reliable. In the 2000 census, the data were for a subpopulation who live in facilities that are owned and managed by others, which includes prisoners incarcerated in correctional facilities, and the response rate to the census was low. Recognizing the problem with data collection, the Census Bureau substantially resolved it in the 2010 census and the ACS, making several tweaks over the years that have continually improved the size and quality of the group quarters sample.¹³

The ACS counts the incarcerated population by their nativity and naturalization status, but local and state governments rarely record whether prisoners are illegal immigrants.¹⁴ As a result, we have to use common statistical methods to identify incarcerated illegal immigrant prisoners by excluding those with characteristics that illegal immigrants are unlikely to have.¹⁵ In other words, we can identify likely illegal immigrants by looking at prisoners with individual characteristics that are highly correlated with being an illegal immigrant.

We identified likely illegal immigrants using a modified residual method developed by economist Christian Gunadi. Our modified method makes larger adjustments for the estimated undercount of the immigrant population and relaxes assumptions about employment and Medicaid access because of legal changes since Gunadi first published his methods.¹⁶ Gunadi’s method imputes legal immigrant status first and then identifies those remaining foreign-born residents as illegal immigrants, which is different from other residual statistical methods that identify illegal immigrants first and then count the remaining foreign-born residents as legal immigrants.¹⁷ Our modified estimation method counts people as legal immigrants if they meet any of the following criteria as recorded in the ACS: is a US citizen; arrived in the US before 1982; served in the armed forces; was born in Cuba and immigrated prior to 2017; received welfare benefits such as Social Security, Supplemental Security Income, Medicaid (with some adjustment based on states extending Medicaid access to illegal immigrants), Medicare, or military insurance; resided in public housing or received rental subsidies or was the spouse of someone who resided in public housing or received rental subsidies; had occupational licenses; and/​or had a spouse who was a legal immigrant or US citizen.¹⁸ The number of legal immigrants estimated using this method includes those present in the United States on temporary nonimmigrant work visas and those who have naturalized and earned American citizenship.

A limitation of the ACS data is that they include prisoners in correctional facilities and other types of facilities. Although most inmates in the public-use microdata version of the ACS are in correctional facilities, the data also include those in mental health and elderly care institutions and in institutions for people with disabilities.¹⁹ These inclusions add ambiguity to our findings about the illegal immigrant population but not to our findings about the immigrant population as a whole, because the ACS releases macrodemographic snapshots of inmates in correctional facilities, which allows us to check our work.²⁰

The above-mentioned ambiguity in illegal immigrant incarceration rates prompted us to narrow our analysis to those who are ages 18–54. This range excludes most inmates in mental health and retirement facilities. Few prisoners are under age 18, many in mental health facilities are juveniles, and many of those over age 54 are in elderly care institutions. Additionally, few illegal immigrants are elderly, whereas those in elderly care institutions are typically over age 54.²¹ As a result, narrowing the age range does not exclude many individuals from our analysis. We identified 1,742,385 prisoners in the 18–54 age range in adult correctional facilities in 2024, compared to approximately 1,591,028 identified by the ACS in the same year and age ranges. Winnowing the age range reduces our estimated number of incarcerated individuals in the 18–54 age range to about 8.7 percent below that of the ACS 2024 snapshot.²² Natives in our results are those born as American citizens, and the group includes both those born in the United States and those born abroad to American parents.

Controlling for the size of the population is essential to compare relative incarceration rates between the native-born, illegal immigrant, and legal immigrant subpopulations. Thus, we report the incarceration rate as the number of incarcerations per 100,000 members of that particular subpopulation, just as most government agencies do.²³

Incarcerations

An estimated 1,605,032 native-born Americans, 79,825 illegal immigrants, and 57,528 legal immigrants between the ages of 18–54 were incarcerated in 2024. The incarceration rate for native-born Americans was 1,195 per 100,000; 674 per 100,000 for illegal immigrants; and 303 per 100,000 for legal immigrants in 2024 (Figure 1). Illegal immigrants are about 44 percent less likely to be incarcerated than native-born Americans. Legal immigrants are 75 percent less likely to be incarcerated than natives. If native-born Americans were incarcerated at the same rate as illegal immigrants, about 701,000 fewer native-born Americans would be incarcerated. If native-born Americans were incarcerated at the same rate as legal immigrants then there would be 1.2 million fewer incarcerations.

The ACS data include illegal immigrants incarcerated for immigration offenses and those in Immigration and Customs Enforcement’s (ICE) detention facilities.²⁴ Those individuals are not detained for violent or property crimes but only for immigration violations. If we were to remove the 37,684 people in ICE detention facilities at the end of fiscal year 2024, that would lower the illegal immigrant incarceration rate to 356 per 100,000—17 percent above the incarceration rate for legal immigrants.²⁵

Robustness Checks for Counting the Illegal Immigrant Population

The results of the modified residual method we used to estimate the number of illegal immigrants is sensitive to the specific ACS variables chosen. Thus, we decided to use a robustness check to alter some of the variables to see whether the results significantly changed. First, we loosened the identifications for illegal immigrants, counting some of those who lived in households with users of means-tested welfare benefits as illegal immigrants. Illegal immigrants do not have access to those benefits, but US citizens and some lawful permanent residents in their households do. This adjustment increased the illegal immigrant population and their incarceration rate to 759 per 100,000, reduced the legal immigrant incarceration rate to 221 per 100,000, and did not affect the native incarceration rate.

Our second robustness check excluded all immigrants who entered the United States after 2009. Immigrants on lawful permanent residency can apply for citizenship after five years, guaranteeing that most of the lawful permanent residents who are able to naturalize have done so, which decreases the pool of potential illegal immigrants in our sample. This robustness check shrinks the size of the nonincarcerated illegal immigrant subpopulation relative to those incarcerated and, thus, slightly raises the rate of illegal immigrant incarceration to about 957 per 100,000 and the legal immigrant rate to 390 per 100,000. These variable changes did not alter our results enough to undermine confidence in the findings.

Illegal Immigrant Incarceration Rates Over Time, 2010–2024

Figure 2 shows how incarceration rates for native-born Americans, illegal immigrants, and legal immigrants have changed during the 2010–2024 period. In every year, the illegal immigrant incarceration rate is between 31 percent and 56 percent below that of native-born Americans. In every year, the legal immigrant incarceration rate is between 65 percent and 75 percent below that of native-born Americans. Furthermore, the incarceration rate has declined for every group. From 2010 to 2024, the native-born incarceration rate fell by 25 percent, the legal immigrant incarceration rate fell by 41 percent, and the illegal immigrant incarceration rate fell by 30 percent.²⁶ However, recently there was an increase in the illegal immigrant incarceration rate of 25 percent from a low of 538 per 100,000 in 2022 to 674 in 2024.

Demographic and Social Characteristics

Incarceration rates vary widely by race and ethnicity in the United States, even within each immigrant category (Table 1). Legal and illegal immigrants have a lower incarceration rate than native-born Americans of the same race or ethnicity. For instance, the incarceration rate for black native-born Americans is 9.5 times that of black legal immigrants, and the incarceration rate for white native-born Americans is 5.1 times higher than it is for white legal immigrants. The incarceration rate for all illegal immigrants is 6.5 percent lower than the incarceration rate for white native-born Americans.

Immigrants from some parts of the world are more likely to be incarcerated than others (Table 2).²⁷ Legal immigrants from Oceania have the highest incarceration rates, followed by legal immigrants from Latin America, but both are substantially below the incarceration rate for native-born Americans. For illegal immigrants, those from Latin America have the highest incarceration rates, followed by those from Africa. Again, all groups of illegal immigrants have lower incarceration rates than native-born Americans.

More than 72 percent of all immigrants in the United States come from the top 20 countries of origin for the foreign-born population.²⁸ Illegal immigrants from Jamaica and legal immigrants from Guatemala have the highest incarceration rates; these are likely exacerbated by their presence in immigration detention facilities for immigration offenses for Guatemalans, and by the small sample sizes for Jamaicans (Table 3). The distribution of prisoners by their immigration status and region of origin shows that 7 percent of all those incarcerated are immigrants from Latin America; 1 percent each from Europe, East Asia, and Africa; and 91 percent are native-born Americans regardless of their location of birth (Table 4).

On January 6, 2023, the Biden administration created a parole program to allow American residents to sponsor up to 30,000 Cubans, Venezuelans, Nicaraguans, and Haitians (CVNH) to come to the United States legally through a parole sponsorship program.²⁹ As of November 2024, 531,670 Cubans, Venezuelans, Nicaraguans, and Haitians arrived lawfully in the United States and were granted parole. Although the incarceration data in this paper only go through 2024, the incarceration rate for immigrants from the CVNH countries could indicate how criminally inclined the parolees from there will end up being. The incarceration rate for all immigrants from the CVNH countries in the 18–54 age range was 387 per 100,000 in 2024, well below the incarceration rate for all immigrants of 446 per 100,000. Venezuelans had the lowest incarceration rate, at 277 per 100,000, while Nicaraguans had the highest, at 634 per 100,000. Haitians and Cubans had incarceration rates of 440 and 406, respectively. Incarceration rates for CVNH countries, whether separately, combined, or divided by immigration status, had incarceration rates substantially below those of native-born Americans and all immigrants.

Whereas only 11.6 percent of all prisoners are women, 88.4 percent are men (Table 5). Legal and illegal immigrant women are a smaller share of prisoners in their respective subpopulations at 9.9 percent and 3.4 percent, respectively. Native-born American women are significantly more likely to be incarcerated than immigrant women.

Prisoners in every subpopulation are less educated than their total subpopulation (Table 6). About 66.6 percent of all native-born adults have some college education, which can include a community college course or above, whereas 18.7 percent of native-born prisoners have the same level of education.³⁰ A total of 22.7 percent of legal immigrant prisoners and 12.5 percent of illegal immigrant prisoners have some college education or above; these percentages are lower than the percentages of their subpopulations with the same level of education, which are 55.8 percent and 46.4 percent, respectively. Both highly educated native-born Americans and highly educated immigrants tend to avoid incarceration.

Native-born Americans, illegal immigrants, and legal immigrants all have higher incarceration rates when they are younger (Figure 3). Native-born American and illegal immigrant incarceration rates peak in the 35–39 age range at 1,596 and 833 per 100,000, respectively. Legal immigrant incarceration rates peak younger, in the 18–24 age range. The incarceration rates for legal and illegal immigrants generally increase with the amount of time they have spent in the United States, with minor variations in the different age ranges (Table 7).

Illegal and legal immigrants who immigrate at a younger age are more likely to be incarcerated, which is related to the amount of time those immigrants have spent in the United States—but it is still distinct (Table 8). Illegal immigrants who arrive between ages 0 and 17 are almost 180 percent more likely to be incarcerated than those who arrive between 18 and 24, suggesting that illegal immigrants who were old enough to choose to come here illegally are more law-abiding than those who were brought here as minors. The pattern is less pronounced, albeit still notable, for legal immigrants. Those who immigrated between the ages of 0 and 17 were 150 percent more likely to be incarcerated than legal immigrants who came at older ages. This again suggests that those who were old enough to choose to come to the United States legally are more law-abiding regardless of their legal status.

At least two non-mutually exclusive theories can explain why those who immigrated in their youth have higher incarceration rates. First, spending part of one’s childhood in the United States could assimilate some immigrants to a relatively high-crime culture. A second theory is that those who decide to come here by choice as adults have systematically different characteristics from the people who remain in their countries, whereas those who are too young to make the decision to immigrate do not. For instance, victims of crime in Latin America are more likely to consider immigrating to the United States.³¹

Policy Implications

A substantial percentage of the American public believes that immigration increases crime and that illegal immigrants disproportionately contribute to the problem.³² However, the evidence presented here shows that the addition of a less crime-prone immigrant population to the United States mechanically reduces the overall incarceration rate in the country. The facts uncovered in this paper should point the government toward immigration policies that would reduce crime further.

For instance, federal officials should stop their efforts to convince so-called sanctuary cities to fully abandon those policies because such cooperation will likely have no effect on violent and property crime rates nationwide. Illegal immigrants have a lower incarceration rate than native-born Americans, so scarce law enforcement resources should not be spent on identifying and deporting a subpopulation with such a low crime rate. If the purpose of law enforcement is to deter crime and to punish criminals, their resources would be inefficiently allocated if targeted at illegal immigrants. However, the federal government should convince sanctuary jurisdictions and others to turn over any noncitizen guilty of committing a violent or property-related offense for removal from the United States. Immigration and Customs Enforcement should be a pickup and delivery service for convicted illegal immigrant criminals, nothing more.

Second, the federal government already has effective programs to identify illegal immigrant criminals who have been arrested, convicted, or incarcerated. The federal government should continue those policies and make the removal of illegal immigrant criminals a priority, but it should not widen its reach to include illegal immigrants who have not committed criminal offenses or have not otherwise put Americans at risk.³³

Third, the government should collect better data on illegal and legal immigrant criminality. Incarceration rates are just one measurement of criminality that are used to fully understand relative crime rates in the United States. Unfortunately, the paucity of data means that we must estimate the number of illegal immigrants who are incarcerated, which adds some uncertainty to our final numbers. Every state should collect—and make available—data on the immigration statuses of those convicted and arrested for crimes, just like Texas does, as well as those who are incarcerated, as Georgia and Oklahoma do.³⁴ To be clear, this proposal would only require documenting the immigration status of people who are arrested for crimes, convicted of crimes, or incarcerated for crimes. There is no excuse for the lack of data on this important public policy issue.

Conclusion

Legal and illegal immigrants were less likely to be incarcerated than native-born Americans in 2024, and even going back to 2010. Those who were incarcerated do not represent the total number of immigrants who can be deported under current law or the complete number of convicted immigrant criminals who are in the United States, but merely those who are incarcerated. The younger that immigrants are upon their arrival in the United States, and the longer that they are here, the more likely they are to be incarcerated as adults. This analysis provides numbers and demographic characteristics to better inform the public policy debate over immigration and crime. The government should expeditiously remove violent and property criminals who are noncitizens, whether they are legal immigrants or illegal immigrants, but a general mass deportation policy indiscriminately targeted at all illegal immigrants will not reduce crime rates, nor will reductions in legal immigration. Lastly, we recommend that governments at all levels in the United States focus on collecting better data so that we can more precisely understand how illegal immigrants and legal immigrants contribute to crime in the United States.

But there is another issue with polls of Muslims, even in the United States: they significantly understate the extent of immigrant assimilation because so many Muslim immigrants leave Islam. According to the Pew Research Center’s 2023–2024 Religious Landscape Study, nearly one in four people raised Muslim leave Islam. This biases polls against showing how the opinions of these immigrants and their descendants change over time. Indeed, leaving Islam is a form of assimilation into the broader society.

As an illustration of the issue, here is one question from Pew’s survey on whether homosexuality should be accepted. It shows that Muslims who were raised Muslim disapproved of homosexuality by 21 points more than they approved of it. Net approval among all Muslims, including converts, was negative 14 points. But among all individuals raised Muslim, as many respondents approved as disapproved. That’s a 21 percentage point difference from Muslims who were raised Muslim.

The Pew poll allows us to compare responses over immigrant generations. As we would expect based on the opinions of Muslims abroad, first-generation Muslims are the most hostile toward homosexuality, with a net disapproval of 27 percentage points. But including ex-Muslim immigrants immediately drops that share by 14 percentage points. Among the 2nd generation who were raised Muslim, a majority accepted homosexuality.

The sample size for Muslims in the poll is relatively small, so the specific results should be interpreted with appropriate caution, but the pattern broadly makes sense. Many Muslim immigrants come to the United States. The ones who leave Islam tend to be the most liberal and assimilated. This polling actually understates the extent of Muslim assimilation because it excludes the second-generation children of Muslim immigrants who were not raised Muslim themselves. We also should not infer that everyone who does not think homosexuality is acceptable believes in authoritarian measures to stop it.

Americans who were raised Muslim diverge sharply from people raised in Muslim countries. Looking at the Pew Research Center’s Spring 2019 Global Attitudes Survey and the Arab Barometer Survey from 2019 reveals a stark contrast between people raised in the United States and most other majority-Muslim countries. The highest level in any majority-Muslim country included in those surveys was 26 percent acceptance, compared with 48 percent in the United States.

This post cannot address every aspect of Muslim assimilation. We have previously discussed issues related to terrorism, crime, and other aspects of assimilation. But it is important to understand that Muslims do assimilate. In prior posts on this topic, I have shown how this pattern repeats on a variety of other issue areas—including support for violence against civilians, which Muslims are less supportive of than the general population.

It is worth emphasizing that most Muslim immigrants have come to the United States in just the last two decades, so the period for assimilation has been short, and an extraordinary convergence is already underway. We should be hopeful that these trends continue and that the Muslim liberalism advocated by my colleague Mustafa Akyol and other American Muslims radiates from the United States around the world.

During testimony before the US Senate, I called it the largest fraud in the history of the US immigration system. In fact, it is likely the first $1 billion fraud—a theft of processing fees for services never rendered. Even Republican Senator John Kennedy of Louisiana told me at the hearing that he was shocked it was happening. It is happening, and it is shocking.

The Policies Defrauding Legal Immigrants

This massive fraud is the result of three policies promulgated by President Trump, Secretary of State Marco Rubio, and the head of US Citizenship and Immigration Services (USCIS), Joseph Edlow.

1. First, in December, President Trump signed a proclamation that bars legal entry and almost all visas for citizens of 40 countries (including the Palestinian territories) who were then outside of the United States. An earlier June proclamation had already partially suspended entry for 19 countries, but it had exempted immediate relatives of US citizens—the largest category of legal immigration. The upshot of these bans is that any citizens from these countries will be denied if they apply for an immigrant visa and most types of temporary visas. Nonetheless, consular officers are being instructed not to “counsel applicants or advise them in advance of the interview that they are subject to the [Proclamation].”

This Trump entry proclamation, which uses the “travel ban” authority under section 212(f) of the Immigration and Nationality Act (INA), was based on the purported inability to screen those immigrants to the United States, and if his first term is any indication, this policy will likely never be rescinded. By itself, this policy blocks one in five legal immigrants from entering the United States. Among others, it covers major origin countries of Cuba, Venezuela, Nigeria, Iran, and Haiti. A disproportionate number of countries are in Africa.

2. Second, in November, Edlow signed a memorandum that extended the 19-country ban to applications for benefits for immigrants already in the United States if the beneficiary is from one of the targeted countries. In January, he extended this processing freeze to the 40-country adjudication freeze, and he ordered a re-review of any approved benefits for individuals from these countries under the Biden administration. The benefits freeze—which applies to everything from employment authorization documents to permanent residency applications—covers even people who have lived in the United States for decades. It has no end date, and although it cites no statutory authority, it is unlikely to be rescinded. One lawsuit describes the devastating effects that this policy can have on immigrants in the country.

3. Third, the State Department, through public statements and informal guidance, has frozen all immigrant visa applications for 75 countries. The purported justification for these restrictions is that immigrants from these countries use welfare at higher rates than immigrants from other countries. The announcement was just three sentences long and included no evidence to support its conclusion. It was later amended to exempt adopted children of US citizens, but not their biological children, stepchildren, spouses, parents, or other relatives.

The 75-country ban covers major sending countries of Jamaica, Bangladesh, Ethiopia, Colombia, Nepal, Egypt, Ghana, and Guatemala. This policy also has no end date, and the most detail provided to the public is from State Department social media posts, which say, “The freeze will remain active until the US can ensure that new immigrants will not extract wealth from the American people.”

The legal justification is that immigrants must prove that they are not “likely to become a public charge.” But this is an individualized assessment, not a group-based one. There is no authority in statute to impose a blanket ban on all immigrants based on their nationality’s welfare use rates. Nearly all immigrant visa applicants must show the ability to support themselves and are ineligible for welfare for the first five years after they enter. All family-sponsored immigrants must have a financial sponsor who agrees to support them and demonstrates an income above the poverty line. The State Department has cited nationalities like Somalis who primarily entered as refugees, who are not subject to the public charge rule or subject to the usual 5‑year waiting period for welfare benefits.

The two policies—the 40-country presidential proclamation entry restriction and this indefinite immigrant visa “pause”—mean that citizens of 92 countries (including Palestine) are completely banned from receiving immigrant visas and immigrating permanently to the United States. These nationalities account for half of all legal immigrants to the United States entering from abroad. Over 320,000 immigrant visa applicants—based on 2024 flows—are now blocked. Once accounting for the immigrants in the United States, the number of potential legal permanent residents blocked rises to 561,000.

There is a fourth policy that has separately shut down the diversity visa lottery. The State Department writes, “DV applicants may submit visa applications and attend interviews, and the Department will continue to schedule applicants for appointments, but no DVs will be issued.” Once again, the State Department is letting people pay fees and go to interviews, while simultaneously promising not to process them. However, over 90 percent of diversity visa lottery winners were already banned based on the other three policies.

Fees to Immigrate

How much money has the federal government accepted to not perform the services required by law?

To immigrate to the United States or to obtain authorization to work or travel internationally, noncitizens must usually pay a fee to have their applications processed. USCIS’s immigration fee revenues were nearly $7 billion, and the Consular Affairs budget was about $6 billion.

The fees stack up. For instance, to sponsor a spouse, a US citizen must pay a $675 fee to USCIS to petition for their spouse to obtain lawful permanent residence. Then, the immigrant must pay $1,440 to adjust status from temporary to permanent residence. That application takes so long that people usually pay $560 for the spouse to receive an employment authorization document, so the total fees can add up to $2,675.

But the exact fees are extraordinarily difficult to track. Every application has a different fee. The fee amounts often depend on the applicant’s age, and which applications are required depends on whether the immigrant is the primary beneficiary or a dependent, such as a spouse or child. The government does publish detailed statistics sufficient to establish the number of beneficiaries affected, let alone the number of fees. Many statistics on applications lack even a breakdown by country. We also have no up-to-date statistics. We are forced to rely on 2023 and 2024 statistics because the current administration has simply stopped publishing most statistics.

That said, I made assumptions, which I believe will likely understate reality. That’s because it only includes people who would have otherwise been approved for a benefit in the next year, so it basically ignores the entire backlog of cases that the government may not get to this year, as well as everyone who would have otherwise been denied under normal grounds who are equally entitled to an answer. Regardless, it provides some insight into the enormous scale by which the government has bilked immigrants of their money.

Overall, we can estimate that the 2 million applications affected by the ban, which would’ve otherwise been approved, came with over $1 billion in fees. The fees came primarily from applications for work permits and filings to adjust status to permanent residence or immigrant visas—the two ways in which people receive green cards to stay permanently in the United States.

Cubans represent the largest group of affected applicants with nearly a million affected applications at a combined cost of $543 million. The second most common was Venezuelans, with 239,000 applications at a cost of $138 million. Nigerians, Afghans, Haitians, and Iranians are the other top nationalities. Interestingly, about 90 percent of the applications and fees come from countries targeted by the initial 40-country ban, not the later State Department immigrant visa restriction. That’s largely because the State Department charges much lower fees than USCIS does, and USCIS’s ban applies to so many different types of applications.

Congress should immediately require the administration to start processing applications and fairly adjudicate those applications without regard to a person’s birthplace. If someone cannot establish their eligibility, they can be denied under the law, but there is no reason to steal people’s fees and fail to provide the service the law entitles them to.

Jalloh’s dastardly attack is one of five deadly foreign-born terrorist attacks during the second Trump administration that have killed a total of eight people. Six of them were victims of foreign-born terrorists motivated by Islamism, two by a right-wing terrorist who killed his parents to steal money to pursue his plot to assassinate President Trump to start a white supremacist ethnostate. In comparison, a single foreign-born terrorist killed one person during the Biden administration, and he was motivated by Chinese nationalism.

Figure 1 shows the number of victims of foreign-born terrorism. It’s easy for the number of victims to vary so much over just a few years when there are so few. The number of victims of foreign-born terrorism in 2025 was the highest since 2017, when Trump was also in office, but just 0.03 percent of homicides last year were committed by foreign-born terrorists in attacks. The annual chance of being murdered in a foreign-born terrorist attack during the second Trump administration is about one in 85 million per year, about 16 times higher than during the Biden administration so far, which was about one in 1.3 billion per year. Regardless, it’s a tiny risk even though fewer people were murdered in foreign-born terror attacks during Biden’s presidency than any other president since Gerald Ford and likely even earlier.

Over 85 percent of all people killed by terrorists in attacks on US soil were killed by foreign-born terrorists since 1975. The 9/11 attacks account for 83 percent of those victims because they were the deadliest terrorist attacks in world history by a factor of nine.[i] Things have changed since 9/11, which occurred smack dab in the middle of my dataset.

More people have been killed in terrorist attacks committed by native-born Americans since 9/11 than by foreign-born terrorists, and it’s not close (Table 1). Native-born terrorists murdered 6.3 people in attacks for every one person murdered by a foreign-born terrorist. Of the 335 people murdered in terrorist attacks on US soil since September 12, 2001, 289 were murdered by native-born Americans and 46 by foreign-born attackers. The share of victims in foreign-born attacks is 13.7 percent, almost exactly their average share of the US population since 2001.

What does all this mean? Jalloh didn’t survive his attack, but terrorists who do should be prosecuted and punished to the fullest extent of the law. Keeping foreign-born terrorists out of the United States is a legitimate function of government that nobody disputes. But the data are clear that foreign-born terrorism is a tiny risk, even during the periods when it is most elevated, and each attack is committed by a handful of ideologically motivated killers who are wildly unrepresentative of the tens of millions of foreign-born people living peacefully in this country. Punishing innocent immigrants and people abroad who want to come to this country, launching more foreign wars, or stripping Americans of civil liberties in response to attacks this rare would be an expensive overreaction that would cost far more than any benefits.

[i] If you count the 10/7 attack in Israel as a terrorist attack, then 9/11 was two-and-a-half times deadlier than the next deadliest attack, but some don’t count Hamas’ attack because Hamas was the government of Gaza and therefore their attack is not terrorism by definition. Brutal, regardless of how you classify it, but plausibly not terrorism.

Here is my five-minute opening statement, as prepared:

Chairman Graham, Ranking Member Merkley, and distinguished members of the Committee, thank you for the opportunity to testify.

Subscribe now

For half a century, the Cato Institute’s research has shown that people—whatever their ancestry, background, or birthplace—can thrive in a free society. Immigrants come to work for us and with us, to make our lives better.

They work at higher rates and are nearly twice as likely to start businesses. Immigrants have reduced government deficits by at least $14.5 trillion over the last 30 years.

And yet Congress has not updated the legal immigration system during that entire time, so just 3 percent of all legal immigrant applicants received approval in 2024.

Rather than fix this broken system, the new administration banned half—half—of all the legal immigrants who had made it to the end of that process, including the spouses and children of US citizens.

The fallout from this flawed system is left to our cities to manage. In many areas, one in five residents lives with someone here illegally. Mass deportation is not an option for these places. It is an attack on them.

It is taking Americans—Americans—away from their spouses, parents, pastors, parishioners, employees, employers, customers, and friends. It’s not us Americans versus them immigrants. Cities know immigrants are part of us.

They also know immigrants are here to build, not destroy. A mountain, literally a mountain, of research proves that immigrants—and reasonable limits on participation with ICE—do not increase crime rates.

Immigrants—even illegal immigrants—are half as likely to be incarcerated for crimes. According to DHS, only about 3 percent of illegal immigrants have any criminal convictions.

Immigrants reduce crime rates, which means you’re less likely to be a crime victim. Mass deportation—not immigration—is the threat to cities.

Let me be clear: when a noncitizen does victimize someone, they should be convicted, imprisoned, and deported.

But it is DHS that rejects that formula. Just 28 percent of the people detained by ICE right now were convicted of any crime.

This is no surprise. DHS has abandoned its public safety mission. It stripped millions of vetted immigrants of legal status just so it could deport them.

Instead of protecting us, DHS violates the Constitution. Its masked agents invade our homes without warrants.

They detain without evidence and deport without due process. They arrest legal immigrants for writing opinion articles.

They intimidate, beat, and even kill Americans who protest. They lie to the courts, and then they ignore them.

With all this unconstitutional mayhem, Congress should not itself violate the Constitution by threatening local law enforcement—not for interfering with ICE—but rather, for following their own laws that they swore an oath to uphold.

Under our system of federalism and the Constitution, Congress can’t force cities to do its bidding—whether on immigration, guns, or the environment.

If Congress wants cities to help DHS, it’ll need those cities to trust DHS. But to get trust, you need to be trustworthy. And this DHS is not.

Consider this: before this administration, noncitizens were more likely to report serious crimes to the police than US-born citizens.

Now DHS is arresting immigrant crime victims. For example, in Iowa, a shooting victim, who helped bring to justice violent robbers, was arrested by ICE before he had even healed.

But since he had a pending charge for an unpaid traffic ticket, ICE considers him a criminal. This has caused filings for U visas—for crime victims working with police—to drop by 63 percent. This is not making Americans safer.

Congress must require DHS to focus on real threats, and it must empower Americans to sue federal agents for violating the Constitution. We need accountability and constitutionality to restore trust.

Thank you.

You can read my full written testimony here.

Here’s my exchange with Sen. Jeff Merkley (D-OR):

Here’s my exchange with Sen. Chris Van Hollen (D-MD):

Here’s my exchange with Sen. Ron Johnson (R-WI):

Here’s my exchange with Sen. John Kennedy (R-LA). Apparently, it is not widely known that President Trump is going to the Supreme Court for the right to deport US-born Americans by terminating birthright citizenship for children of anyone whose parents did not have permanent residence at the time of their birth.

Here are the social media posts Sen. Kennedy references:

Here’s the video from the final post, showing that it was a judge targeted with threats for ruling that the attempt to deport US-born Americans is unconstitutional:

When asked how long the process should take, 66% of Republicans, 65% of Independents, and 72% of Democrats said under a year. In other words, the immigration system that even the modal Republican voter wants is far less restrictive than the one the United States already has. They just have no idea how difficult and restrictive the existing immigration system already is. These results motivated us to study the issue more carefully.

As researchers who study immigration policy and public opinion, we hear some version of “I support legal immigration” all the time. It is one of the most common things Americans say when the topic comes up. Unfortunately, the people who say it usually cannot explain what legal immigration actually involves: the nearly two hundred visa categories, the decades-long backlogs, the $1,140 application fee for a green card (before thousands more in legal fees), or the 75-country travel ban that now bars a quarter of the world’s population from most visa categories. That is not the fault of prospective immigrants. The system is genuinely hard to understand, even for graduate-level trained immigration researchers like us.

Take immediate migration for spouses of US citizens, which is about as close to a slam dunk in terms of public support as immigration policy gets. Almost nobody opposes letting citizens bring their husband or wife to the country. Despite its popularity, the process takes 12 to 18 months, costs thousands in fees, and involves mountains of paperwork. Learning about these delays angers even the harshest critics of current immigration levels, including supporters of President Trump (who has famously married several immigrants himself). If the system is this difficult for spouses, imagine what it looks like for everyone else.

The gap between what people think the system looks like and what it actually looks like has real consequences for the debate.

Almost nobody knows how the system works

We tested this phenomenon more systematically in our paper, recently published in the Journal of Experimental Political Science. In our survey of 1,000 Americans, we asked respondents how long it would take various types of people to legally immigrate. We also asked whether certain family members are even eligible. The average respondent got 25% of the questions right, barely above the 20% you’d expect from random guessing. For example, only about 8% correctly identified that aunts and uncles of U.S. citizens are not eligible to immigrate at all.

The ignorance was remarkably uniform. Young and old, rich and poor, college-educated and not, Democrat and Republican: none of these groups performed meaningfully better than any other. Republicans were, if anything, slightly less likely to know the rules, though the differences were small. The point is not to single anyone out. Almost everyone is in the dark.

This should not be surprising. Alex Nowrasteh and David Bier at the Cato Institute have spent years documenting just how byzantine the system is. Their analysis found that fewer than 1% of people who want to move permanently to the United States can do so legally. They also created the Green Card Game, an online simulation that lets you try to work through the actual rules. Most players discover that there is no “line” to get into, or that the line they’d be placed in lasts longer than a human lifetime. Our research was directly inspired by their work, and we wanted to test whether learning about these realities actually changes people’s minds.

Telling people the truth moves opinion

We are generally skeptical that information alone changes political attitudes. On most issues, people already know what they want to believe, and new facts get filtered through existing priors. But immigration policy might be different, precisely because the knowledge gap is so enormous. People are not filtering out inconvenient truths. They simply have no idea what the current rules are.

To test this, we randomly assigned respondents to read one of two short paragraphs: one describing how burdensome the system is (costs, wait times, complexity) and another describing how restrictive it is (numerical caps, country quotas, multi-decade backlogs). A control group read a neutral paragraph about migration definitions.

Both treatments worked. Respondents who learned about the system were 13 percentage points more likely to say they preferred increasing legal immigration or making the process easier. That is a large effect for a survey experiment, roughly a 35% increase over the baseline. And the shift was not limited to Democrats. It showed up across the political spectrum, including among Republicans and self-identified conservatives.

Why did it work? We think it is because the information was relevant to what people already care about. Telling people “immigrants are good” is unlikely to move many minds because it runs into existing beliefs and identities. Telling people “the system you think you support is actually far more restrictive than you realize” is a different kind of message. It speaks to voters’ own preferences rather than trying to override them.

What this means

We should be clear about what this does and does not show. A one-shot survey experiment cannot tell us whether these effects last, or whether they survive the counter-messaging that voters encounter in real political campaigns. We also studied legal immigration specifically, not the broader debate about border security or unauthorized migration, which triggers different reactions. We are now working on a larger follow-up study across multiple countries to test whether the effects hold up over time, whether they extend to attitudes about unauthorized immigration and naturalization, and whether some groups (such as moderates and conservatives) are more persuadable than others.

But the finding already points to something practically useful. Most Americans, across the political spectrum, want a legal immigration system that is faster and simpler than the one they have. They do not know that the current system already fails to meet their own standards. That gap is an opportunity, both for changing minds and for changing policy.

For those who want to make immigration more politically sustainable, the lesson is that persuasion works best when it is about things people care about but don’t yet know. Generic appeals to the benefits of diversity or immigrant contributions are easy to dismiss. Concrete facts about wait times, costs, and eligibility rules are harder to argue with because they speak to the voter’s own sense of fairness.

The best outcome, of course, would be to actually make legal immigration less difficult rather than simply informing people that it is difficult. But the latter may be important for the former. When voters learn that the system is more restrictive than what they want, the political space for reform expands. Spread the word.

Alexander Kustov is an associate professor at the University of Notre Dame’s Keough School of Global Affairs and the author of In Our Interest: How Democracies Can Make Immigration Popular (Columbia University Press, 2025). Michelangelo Landgrave is a faculty member in the Department of Political Science at the University of Colorado. Their paper “Immigration is Difficult?!” was published in the Journal of Experimental Political Science in 2025.

The Manhattan Institute (MI) published a criticism of Cato’s immigration report on the effects of immigrants on government budgets. In Part 1, I explained how MI misrepresented the plausibility of our conclusion and, in so doing, revealed flaws in its own methods that understate the net effect of immigrants on government budgets. Now I will turn to MI’s methodological criticisms, which are bad-faith, irrelevant, and false.

Subscribe now

MI concedes the point: immigrants reduce deficits.

MI’s main objection to our study actually has nothing to do with our methodology. Indeed, it has no serious methodological criticisms at all (see below). Instead, it focuses on the presentation of our results. Specifically:

the [Cato] researchers assign the cost of every service received by native-born children of immigrants to the children rather than to the immigrant parents. This removes large expenditures—especially the cost of education—from the fiscal cost of immigration, even though those children would not cost the state anything if their parents had not immigrated.

In other words, MI wants us to—not change our model—but to instead treat US-born children as if they are immigrants. But they are not immigrants, and we wanted to determine whether immigrants are the driver of America’s deficits and debt, or whether Congress needs to reduce spending on US-born Americans—including US-born children of immigrants—to address them.

The fact that some (ha!) US-born Americans are descendants of immigrants does not undermine our thesis at all. Arbitrarily lumping some US-born Americans in with immigrants—without first clearly differentiating the two—would obscure who is driving the fiscal situation in the United States.

Moreover, it is impossible to make accurate comparisons between types of immigrants if all child costs are credited to adults. For instance, is an adult immigrant traveling alone less costly than a child immigrant coming with parents? If the child’s costs are awarded entirely to the parents, we couldn’t use our model to answer this question. Similarly, conflating US-born and immigrant costs confuses policy implications of guest worker programs, immigrant welfare reforms, and deportations. How much do deportations of immigrant adults reduce spending on their US-born children? At a minimum, we can say that it is not 100 percent, as MI claims.

But of course, the US-born children of immigrants exist. They impose costs, and they wouldn’t exist without immigration. They should be analyzed. And you’ll be shocked (if you only read MI’s criticism) to learn that we did estimate the effect of immigrants and their children, finding that they collectively reduced deficits by $7.9 trillion over 30 years. Not only that, but we find that in the long term, the second generation—who are mostly below the age of 20—are already the most fiscally positive at any given age.

We did the very estimates MI demanded. In fact, we overstate these costs by nearly $4 trillion by including the children of immigrants who came in the early 20th century. These early-20th-century immigrants are clearly not the immigrants in our study and do not represent even the indirect costs of post-World War II immigration.

What’s particularly surprising about this criticism is that MI’s 2025 paper essentially uses the same method as the Cato Institute. Like Cato, child benefits—everything from food stamps and Medicaid to education and parks—are credited to children, either US-born or immigrants. Like Cato, it first estimates the effects of immigrants alone and then the effects of immigrants with their descendants, indistinguishable from our report at least in this aspect.

The only difference, as far as I can tell, is that MI treats child tax credits as a cost of adults. This is inconsistent with their treatment of all other types of child benefits, including cash benefits. From our standpoint, a benefit should be credited to the child if the spending would not have occurred without the child’s presence in the United States. Any other method leads to erroneous and misleading results. For instance, if a parent is deported, the value of the child tax credit could increase, since household income would fall. MI does not specify the criteria it uses to decide these questions.

But this issue is tiny compared to the theoretical stance that MI’s critique lays out. Readers get the impression that Di Martino and MI hold a consistent position requiring treating the descendants of immigrants as “immigrants.” But they don’t.

In fact, it is MI, not Cato, that published a 2024 paper by Di Martino that explicitly “excludes the impact of second-generation Americans.” In other words, MI is accusing Cato of doing the very thing that it did.

MI’s methodological objections don’t undermine Cato’s conclusions.

Besides the newly discovered concern about the second generation, MI levels three methodological criticisms relevant to our main analysis that are almost too trivial to address.

1. The first is:

Cato’s [estimates] do not account for the cost of providing public goods—like spending on national defense or roads—and the proportional rise of discretionary spending with population growth. When the U.S. population grows, so does spending on highways and roads, policing, firefighters, and a variety of subsidies.

False. We account for the cost of roads, transportation, policing, fire, and other “public goods” that increase with population growth, and we discuss this issue repeatedly throughout the paper. In fact, our assumption on this point, that costs grow immediately and proportionally in these areas with any increase in the immigrant population, is far more conservative than MI’s own assumption. MI assumes that the growth is more gradual, so MI is criticizing Cato over an issue where Cato adopted a more conservative position than MI did.

We also have a long explanation of why immigration does not increase defense spending. But it doesn’t matter: MI yet again misrepresents its own model on this point, which “exclud[es] defense and other pure public-goods spending completely.”

2. The second criticism regards nontax revenues, which account for about 6 percent of government revenue and 14 percent of the net positive effect from immigrants. MI is right that the National Academies did not incorporate these government revenues into its model, but it provided no explanation for this exclusion and admitted that it made the deficit seem worse than it was. Nonetheless, MI comments:

In the study’s estimates, such revenues account for more than one third of the net payments attributed to low-skilled immigrants. Yet there is little reason to assume that low-income immigrants are generating thousands of dollars per person for government enterprises.

First, MI also employs a subtle rhetorical shift from “low-skilled” to “low-income,” when these categories aren’t the same. You can have low educational attainment and be high income or vice versa. More importantly, MI is confused about what these revenues are. They have little to do with government enterprises, though immigrants are consumers who contribute to their profits. They include a wide range of revenue sources, including leases or rental income from government property, but the biggest amount comes from fees and fines, largely related to vehicles, licenses, and the like, which are notoriously regressive.

While MI claims to find our numbers implausible, the fact is that these revenues exist. For fees and revenues from businesses, we attributed them to shareholders’ dividend income. For state and local fines and fees unrelated to businesses, such as vehicle fees and revenue from violations, we attributed them based on sales tax revenue. This was a conservative assumption since flat fees are more regressive than sales taxes. MI never offers any argument for why Cato should not have modeled them at all, nor does it give any reason that a different modeling choice would have significantly changed our results.

3. The third criticism is technical and also irrelevant to our main conclusion. MI states:

On the revenue side, the study assigns 30 percent of corporate tax revenues to immigrant workers—more than prior analyses—while simultaneously assuming income, payroll, and sales taxes are exclusively borne by workers. But if corporate taxes are partially borne by immigrant workers, then other taxes should only be partially borne by native employers as well. The methodology therefore maximizes the revenues attributed to immigrants while minimizing the attributed costs.

There is indeed a minor tension in our analysis here, but not in the way MI presents at all. In a fiscal analysis like ours, we aren’t trying to identify who specifically sent the money to the government. We are trying to track which individuals are ultimately responsible for the growth in revenue. We assume (as does the National Academies) that when an immigrant comes, the fact that sales taxes are paid by a business isn’t relevant. The taxes wouldn’t be paid without the immigrant consumers. The same principle applies to payroll taxes. No immigrant workers on payroll, no more payroll taxes. MI has the same approach.

MI accuses us of being inconsistent in our analysis of the corporate income tax. Instead of attributing all the profits generated by the corporation from employing the immigrant, the National Academies assigns 20 percent to the worker and 80 percent to the shareholders. For our part, we decided to follow the more recent research on “tax incidence,” which finds that more than 70 percent of corporate tax increases are borne by workers’ wages.

MI is right: that is inconsistent. Economist Michael Clemens has written a paper arguing that the National Academies got this wrong and should have attributed 100 percent of corporate income taxes to workers, because whether the tax comes out of profits or wages is irrelevant, since the profits wouldn’t exist at all without the immigrant workers. In other words, MI is right, but the consistent position would have increased the amount attributed to immigrant workers. MI has a different method that indirectly accounts for future growth in capital, which, as far as I can tell without actual access to the model, is Clemens’ approach. In other words, yet again, MI is criticizing Cato for methods that were more conservative than its own.

Regardless, since many immigrants are business owners, this issue is irrelevant to our aggregate conclusion for all immigrants. While it matters more in the analysis of the effects of different types of immigrants, sticking with the original 20–80 distribution would reduce the net effect for all immigrants by less than 1 percent (Table A3).

MI ignores all the ways in which Cato inflates the costs of immigrants.

MI claims that Cato is biased:

Its conclusion rests on accounting assumptions that inflate immigrant tax payments, underestimate costs, and attribute fiscal effects to immigrants in ways that systematically favor immigrants.

Despite this bold claim, MI identifies no ways in which we did this—literally none. The issue about corporate tax revenue is revealing because it demonstrates the lengths to which MI went to ignore all the ways in which our methods lead to a more conservative result than is warranted. Here are a few other examples:

1. Again, we assume that shareholders pay 30 percent of the cost of the corporate income tax, rather than 100 percent as suggested by economist Michael Clemens and Di Martino’s model.

2. Again, we assumed that all congestible public goods spending—everything from parks and roads to fire protection and most law enforcement—increases immediately and proportionally with any increase in the population from immigration. This is obviously severely overstated, so much so that MI’s 2025 study doesn’t adopt it. MI should know that we didn’t use every means to favor immigrants. Why not criticize us for that? Instead, MI claims we didn’t even account for road costs at all, which is obviously false. We mention accounting for transportation spending seven times in the report.

3. We likewise assumed that public education spending increases immediately and proportionally with any increase in the population from immigration. Again, this is overstated and not adopted by, for instance, the CBO.

4. We followed the National Academies in using the highest available estimate of the cost of bilingual education, based on an ancient (1991) estimate from one state (Florida) that vastly exceeds any recent estimates.

5. We attributed all non-tax government revenues from businesses entirely to the business owners, not to the workers. This is unfair. If the business had no workers, it couldn’t operate. At least a portion of these revenues should be attributed to immigrant workers.

6. We followed the National Academies’ assumption that immigrants cause a proportional share of Border Patrol spending, such that current immigrants are deemed to cause spending intended to prevent future immigrants from coming. Is that fair?

7. Again, when we calculated the cost of children of immigrants, we included children of immigrants who were not themselves immigrants in our study—that is, they were the children of immigrants who arrived in the early 20th century. When people dispute our findings about recent immigrants, they want to know the effects of recent immigrants’ children. If we had excluded these elderly retirees, the fiscal benefit from immigrants with their children would have risen by nearly $4 trillion.

8. We ignored all the ways in which immigrants increase the productivity of US-born workers, which increases their incomes and tax payments. In 2024, the CBO estimated that newly arrived illegal immigrants, asylum seekers, and their families would reduce the federal deficit by between $700 and $900 billion over 10 years, and of this amount, 45 to 55 percent would come from increasing the productivity of US workers. Mark Colas and Dominik Sachs have shown that even using the most conservative estimate of how much low-skilled immigrants affect the wages of US workers would increase their fiscal effect beyond what we find. MI should know all this because we wrote about these issues in our paper, but also because Di Martino wrote about the CBO report.

Conclusion

MI states:

The result of Cato’s unrealistic assumptions is a conclusion that every additional resident—regardless of income or education—is a large fiscal asset in a country running persistent trillion-dollar deficits.

We never make such an absurd statement. There are certainly immigrants who exacerbate the country’s fiscal challenges. Unemployed immigrants, children, and most retirees are fiscal drags, but our analysis shows that these immigrants are far from the norm. The average immigrant—even the average low-skilled immigrants—improve the fiscal situation of the United States, but that’s far from the misleading claim that MI makes.

After MI published its first fiscal effects paper in 2024, we carefully analyzed it and published a response. We found numerous problems that rendered the results invalid. MI published another paper last year that “fixes several assumptions from previous research on the fiscal impact of immigrants,” which we suggested, but as I indicated, has many other problems. MI would not share its model to fully replicate that version. Nonetheless, MI cites its earlier, 2024 paper as refuting our work—even though it changed its methods in response to our criticism.

Rather than crowing about these changes, I offered to work with MI’s Di Martino to identify any actual problems with Cato’s model. He never did. If he ever does, it is unlikely to affect the robustness of our conclusion: immigrants reduced the deficits by trillions.

Instead, MI misleads its readers about our methods, its methods, and the basic facts.

Subscribe now

MI’s own study proves trillions in upside to immigration.

MI begins by asserting:

There is a rich literature that estimates the fiscal impact of immigrants in the United States and none—even the most pro-immigrant analyses—reach conclusions of a magnitude similar to Cato’s.… immigration can, under some circumstances, produce fiscal benefits, but those benefits are measured in tens of thousands of dollars—not the trillions implied by the Cato estimates. [my emphasis]

MI is using the fact that our work is the first-ever analysis of the fiscal effects of immigrants to cover the last 30 years against us. Naturally, no one can have found exactly what we found. We thought our novelty was good, but MI is clear: immigration cannot produce fiscal benefits in the trillions—or is it clear?

To prove the point, MI states that the National Academies found “the average lifetime fiscal impact of immigrants to be roughly –$23,000 in 2012 dollars.” This is wrong. This scenario estimated the net future fiscal value for immigrants present in the country in 2013 over the next 75 years, so it excludes all the taxes that those immigrants already paid in. It was also the worst-case scenario. In projections where Congress does anything to address the fiscal crisis we are barreling toward, the results were positive. Looking at immigrants from the time that they arrive onward—the true lifetime effect—the net benefit ranged from a low of positive $92,000 to a high of $186,000 (Table 8–14). In 2026 dollars, that would be between $6 and $12 trillion in deficit reduction for another 50 million immigrants.

The Congressional Budget Office (CBO) estimated that just the illegal immigrants and asylum seekers who entered under President Biden (and their US-born children) would reduce the federal debt by between $700 and $900 billion over 10 years. The deficit savings rise to $167 billion annually in the tenth year. The CBO also found that immigration changes under President Trump will increase the debt by $500 billion over the next 10 years. And these debt figures are from just a small subset of the 50 million immigrants in the United States and only over 10 years. When extended, the deficit reduction (or increase) would easily reach the trillions over 20 to 30 years. MI should remember this report because it also criticized it (inaccurately).

In 2013, the CBO found that increasing population growth from immigrants and their children by 10 million would reduce deficits by about $900 billion over 20 years ($1.3 trillion in 2026 dollars). The same year, the conservative American Action Forum (AAF) found that immigration reform that increased population growth would reduce the federal deficit by over $2.7 trillion over 10 years. In 2015, AAF published a paper on mass deportation, reiterating and expanding on that conclusion, that MI has even cited. The Bipartisan Policy Center has a more recent analysis with findings of a similar scale.

But we don’t need to consult all these other studies. Di Martino’s own 2025 report for MI states:

The Cato Institute’s plan [to double immigration levels] would reduce the national debt by about $1 trillion in 10 years and $11 trillion in 30 years…. A legal immigration moratorium would increase the national debt by $567 billion over 10 years and by over $6.6 trillion over 30 years, all while causing an 8% economic contraction, increasing the debt-to-GDP ratio by approximately 22 percentage points. … the immigration proposal of the [anti-immigrant] Center for Immigration Studies would shrink the size of the U.S. economy by over $3.3 trillion while adding nearly $5.5 trillion to the debt over 30 years.

Here is the graph of MI’s findings directly from its paper. Note the y‑axis is in billions, so $10,000 represents $10 trillion.

When MI states that everyone who has ever studied this issue has found immigrants cannot produce trillions in debt reduction, it ignores even its own work. Of course, all these analyses are not apples-to-apples comparisons with Cato’s work, but MI cites its work to suggest that Cato’s work is too “implausible” to be believed. Instead, it shows the opposite. And remember all these studies are projections of future budgets with far more spending and lower taxes assumed than the actual budgets of the 1990s and 2000s—the period covered by Cato’s analysis—so as a starting point, there is nothing implausible about immigrants reducing deficits by trillions.

MI’s analysis proves the fiscal upside to immigration.

MI then turns to our finding that the average immigrant had a positive net fiscal effect of $10,349 in 2013. Citing data from the Census’s Current Population Survey (CPS) that the average income for immigrants (including children, the unemployed, and retirees) was $25,000, it comments:

A group earning $25,000 annually cannot plausibly reduce the national deficit by $10,000 per person. That would imply tax payments approaching 40 percent of income, even after accounting for government transfers.

It admits that the CPS shows net positive federal tax payments of $1,900—which at least confirms our analysis is directionally correct: immigrants reduce the deficit. But MI’s analysis is fatally flawed.

1. First, it compares our 2013 finding, expressed in 2024 inflation-adjusted dollars, with nominal income and taxes in 2013.

2. Second, MI only includes federal Social Security and income taxes. This excludes about 30 percent of federal revenue—corporate, excise, Medicare supplemental payments, tariffs, and non-tax fees and revenues—and most federal, state, and local revenues. MI inaccurately claims that these revenues are only 10 percent of federal revenue and that non-federal revenues don’t matter, but in fact, state and local revenues accounted for 44 percent of immigrants’ net fiscal effect in 2013.

3. Third, as the National Academies study cited by MI discusses, the Current Population Survey famously underreports earnings and taxes. MI says tax payments of $10,000 are impossible with incomes of $25,000, but the actual amount of per capita taxes in 2013 was over $15,000 with an average income of just $22,700—lower than the $25,800 for the immigrants—using CPS’s narrow definition of earnings and underreporting. In reality, per capita personal income was $44,520 and per capita GDP was $53,400 in 2013 dollars. Tax payments have been consistently about 27 percent of GDP.

MI should agree with this because its paper also estimates that earnings are less than half of a person’s effect on GDP, so its comparison between CPS-measured income and CPS-measured taxes is misleading.

The chart below shows the actual per capita values for taxes and spending in 2013, compared to actual personal income and GDP per capita. It shows that the Manhattan Institute does not account for most tax revenues in its example. These aren’t estimates or survey results. These are the real numbers from the Bureau of Economic Analysis’s National Income and Product Accounts. Check for yourself. Given that immigrants’ earnings are higher than the average, it would be difficult for immigrants not to be more fiscally favorable than the average person.

Everyone who does this work—the National Academies, Congressional Budget Office, the Cato Institute, among others—takes the total amounts of income, taxes, and spending from administrative records, the true totals, and then uses the survey only for immigrant versus US-born distribution. The fact that MI does not take this step invalidates its prior studies on this issue by excluding a significant amount of revenue.

A better way to intuitively understand our conclusions is to start with the aggregate numbers. In 2013, average per-capita GDP was $53,400 and per-capita government revenue was $15,643. Average per-capita benefits were $13,082, setting aside fixed costs like payments on past debt and the US military, so the average marginal person paid about $2,558 more in taxes than they received in benefits. “Benefits” in our report include everything from welfare and policing to parks and roads.

Were immigrants better than average? Unambiguously yes. As already noted, immigrants had higher per capita incomes, which translated into higher per capita revenues ($17,307 versus $15,396). These taxes widen the average net effect to $4,224. When we account for how many fewer benefits immigrants receive, the effect nearly doubles to $8,000—or $10,000 in 2024 dollars.

Our study identifies the two main sources for these savings: fewer retirement and education costs. Immigrants are half as likely to be in schools, saving on education spending, and they draw much less from Social Security (because many are ineligible due to insufficient work history or legal status) and public pensions (because they are half as likely to work for the government). MI does nothing to refute these findings.

MI provides a false theory for Cato’s result.

Instead of refuting our findings, MI tries to explain our results by stating:

By limiting their analysis to the years 1994 to 2023—the period of fastest growth in the foreign-born population in U.S. history—Cato captures the period when most immigrants were at an age that generated revenue, while largely omitting the future entitlement spending owed to a greying population.

First, we limited our analysis to 1994 to 2023 because the data series starts in 1994, not because we cherry-picked a favorable date. Second, we tested MI’s theory that its young workers entering that maintained the positive net effect, and it’s not true. Immigrants were not significantly underrepresented in the retiree population during this time. Immigrants averaged about 12 percent of the population from 1994 to 2023, while they were 11 percent of the over 65 population. They used 34 percent fewer retiree benefits because they were less likely to qualify for public pensions and because they were subject to unique eligibility limits for Social Security and Medicare.

Third, we tracked a specific cohort of immigrants who entered from 1990 to 1993 throughout the period to see if it is new workers entering that drive our results. In fact, the average immigrant in the 1990 to 1993 cohort was more fiscally positive than the average immigrant ($11,064 versus $9,572 per immigrant overall), despite having lower educational attainment than the other cohorts. If MI had read our paper, it would know these facts. They are prominently highlighted.

In Part 2, I will discuss MI’s irrelevant, inaccurate, and bad-faith methodological criticisms and why Cato’s model is extremely conservative.

Sign up for this newsletter to get that delivered to your inbox. If you can’t wait, it is already on Cato’s website.

There are many good counterarguments to that position such as the fact that we include immigrant children, native-born children are not immigrants, and the logic of Camarota’s argument means we should count all welfare consumed by native-born Americans as actually coming from immigrants because none of us would be here without immigrants. We’ve made those arguments before, so this time we decided to rerun our analysis under some assumptions Camarota sounds like he would approve of. This time, we grouped immigrants together with their U.S.-born dependent children and compared their per-capita welfare use to that of natives (excluding children with two immigrant parents). The results are clear: even under this revised framework, immigrants and their children use significantly less welfare per capita than native-born Americans do.

Camarota’s preferred metric is the share of households that use at least one welfare program. But household use rates are an imprecise way to measure fiscal impact. A household with one member receiving a modest benefit counts the same as a household with 10 people receiving far more in total assistance. Use rates alone do not measure fiscal burden, while dollar amounts do. There’s no good methodological reason to compare welfare use in households of different sizes when we can compare individuals. Just to hammer this point home, individuals all contain the same number of people (one) while households contain different numbers.

Our analysis focuses on per-capita welfare expenditures. Our metric captures the actual fiscal cost per person and allows for meaningful comparisons across populations with different household sizes and demographic structures. We also adjust for survey undercounting in both use rates and benefit amounts. Administrative data would be ideal, but in its absence, careful adjustments to the SIPP survey data are the best available method.

Camarota does raise a fair point: U.S.-born children of immigrants would not exist in the United States had their parents not immigrated. If the goal is to estimate the total fiscal effects attributable to immigration, then including those children alongside their parents is a reasonable exercise. Of course, that argument also means we should include the grandchildren of immigrants in the analysis because they wouldn’t be here without immigrants either. Great-grandchildren too. On second thought, Camarota doesn’t have such a good argument.

After reclassifying immigrants’ dependent U.S.-born children as part of the immigrant population, we found that immigrants and their children still use less welfare per person than native-born American adults and their dependent children. On a per-capita basis, immigrants and their U.S.-born children consumed $2,734 less in welfare benefits annually than natives on average. They used less in Social Security, Medicare, SNAP, SSI, and TANF. They did use more Medicaid, the Earned Income Tax Credit (EITC), and WIC. But in total, they used 25 percent less welfare per capita (Figure 1).

Native-born adults and their dependent children used an average of $10,878 of welfare per capita, in comparison to $8,144 for immigrants and their dependent children. Of that, native-born adults and their children used $7,370 in old-age entitlements and $3,508 in means-tested welfare. Immigrants and their children used $3,993 in old-age entitlements and $4,150 in means-tested welfare. While immigrants and their children used 18 percent more means-tested welfare under this new measure, the U.S.-born and their children used 85 percent more old-age entitlements, driving the nearly $3,000 extra use among the U.S.-born and their children.

That reveals another reason why Cato’s research on immigrant welfare use is better than Camarota’s work at the Center for Immigration Studies. He published the rate of welfare use by households, but he doesn’t publish the amount of welfare consumed by households, nor does he reveal the duration of their use. That leads to whacky results. For instance, Camarota counts a hypothetical immigrant who uses $200 of food stamps for one month of the year and no other welfare for the rest of the year the same as a native-born American who consumes $1,500 in benefits from Medicaid, Social Security, TANF, and SNAP every month. Camarota’s rates of welfare use by households hide extreme variation, while Cato’s analysis of the amount of money consumed will give you a more accurate picture of what’s going on.

Think of it this way: Federal welfare and entitlement reform that lowers the welfare consumed by the hypothetical native-born American above from $1,500 to $200 a month would be a monumental achievement. Scholars would write books, academic papers, and a million blog posts and news stories about it. American taxpayers would save money and the scale of welfare in this country would shrink to a level so low that European stereotypes about American state stinginess would start to match reality. But you’d never know it happened if you relied on Camarota’s method of measurement.

Camarota concludes that because an estimated 51 percent of legal immigrant households use at least one welfare program, the United States should select immigrants based on their likelihood of welfare use and more aggressively enforce immigration laws. But this conclusion depends entirely on a metric that does not measure fiscal burden. When measured properly on a per-capita basis, immigrants and their children consume less welfare than natives, whether or not dependent children of immigrants are included.

Policymakers concerned about welfare expenditures should use this simpler policy rather than trying to predict which immigrants might someday use public benefits or to lock out the enormous economic gains from immigration: reform welfare itself. Build a wall around the welfare state, not around the country. Better yet, combine that with shrinking the welfare state for everyone. Nativists have long been skeptical of these methods because, we suspect, they want to use the existence of the welfare state as an argument for reducing immigration. We propose a simpler and more popular way to resolve that complaint: Use immigration as an argument to reduce welfare. Even though immigrants use less, we can make sure they use less still. A system that limits immigrant access to public benefits removes the incentive problem altogether and avoids turning the government into a central planner tasked with forecasting individual immigrants’ future economic trajectories.

The assassination of conservative activist Charlie Kirk was one of those events. Tyler Robinson was motivated by his progressive beliefs to murder Kirk on September 10, 2025. The video of the assassination is shocking, and Kirk was a well-known fixture on college campuses and in conservative circles. His face was especially recognizable to the young and terminally online.

Shortly afterward, the airwaves and social media filled with stories about how common these types of killings are and how left-wingers are often killing right-wingers. The Trump administration responded by shifting the domestic counter-terrorism state toward targeting left-wingers. Normally sober people had inane comments like, “This time it just feels different.” The facts disagree, as I argued shortly after Kirk’s murder. Today, we just released an improved version of that research titled “Politically Motivated Killers: 51 Years of Terrorist Murders on US Soil, 1975–2025.”

Politically motivated killings are rare. Of the over 1 million murders committed on US soil since 1975, a mere 3,577 were committed by killers with political motivations. Over 88 percent of those murdered were committed in the two largest attacks that are statistical outliers: 9/11 and the Oklahoma City bombing, committed by Islamists and right-wingers, respectively. The remaining 430 deaths committed by 186 killers who were motivated by political ideology represent the extent of the threat of the last 51 years.

The annual chance of being murdered in a politically motivated attack is about 1 in 33.1 million since 1975. In more recent years, it’s about one in 19 million. That is a tiny risk roughly equal to the chance of being killed in a lightning strike in the UK. Every murder is a tragedy, and some have more serious political consequences than others, but that doesn’t justify exaggerating the danger. There isn’t even a statistically significant increase in the number of victims over time, as the deadliest triennial period was 2014-2016, and the safest was 2005-2007.

People are most interested in the ideological breakdown of the killings. With all the outliers excluded, right-wing politically motivated killers are responsible for 195 deaths, or 45 percent of all killings, followed by Islamists at 32 percent with 139 deaths, and left-wingers bringing up a distant third place with 68 murder victims at just under 16 percent. With the outliers removed, that’s about 2.9 people murdered by right-terrorists for each person murdered by left-wingers in attacks. Some critics strenuously objected to removing outliers. Keeping them in, the ratio of right-wing deaths to left-wing deaths rises to 5.3, meaning there are 5.3 people murdered in right-wing terrorist attacks for every person murdered in a left-wing terrorist attack. Islamism also becomes the deadliest ideology by a long shot with outliers included.

My original blog post on this topic, which I posted after Kirk’s murder, identified 3,597 people who were politically motivated killers. Many critics sent me objections and some identified killers whom I should not have included. Their hard work convinced me to reduce the number of victims by 0.56 percent to 3,577.

I used an intercoder reliability test to make sure I wasn’t biased in assigning ideological motivations to the killers. I gave three other researchers my ideological definitions for Islamism, right-wing, left-wing, foreign-nationalism, separatism, and unknown/other, and the list of killers. They then independently researched and coded the ideologies and agreed with my assignment of ideological motivations 95.7 percent of the time.

The raw percentage chance of agreement isn’t good enough because it doesn’t control for chance alignment. Thus, we use Cohen’s kappa as a chance-corrected measure. It estimates agreement according to

where Po is observed agreement and Pₑ is expected agreement under random coding. The estimated result here is κ ≈ 0.939, which indicates near-perfect agreement and is consistent with Krippendorff’s alpha, another measure that is robust to missing data and uneven category distributions. Scott’s pi, another method that relies on pooled category proportions rather than coder-specific margins, returned an almost identical estimate (π ≈ 0.939). The convergence of these different statistics is evidence that coder agreement is not an artifact of the underlying distribution of ideological categories.

The remaining coding disagreements were mostly limited to boundary cases where the coders classified foreign nationalism as right or, in a few instances, as unknown/other. No category was large enough to inflate agreement, and the coders were in perfect agreement for rare categories such as separatism (n = 3). Even if every disputed case were recoded against my judgment, the ideological distribution of the attackers would not meaningfully change although there would be a few more victims of right-wing attacks.

My paper also includes explanations for my most controversial inclusions and exclusions, such as listing Rahmanullah Lakanwal, who murdered a national guard member in November 2025, as an Islamist even though there is little evidence that he was ideologically motivated and he hasn’t been charged with a terrorist offense. Contrary to what you’ve heard, very few people support committing violence for political ideologies. Bad polling data is ubiquitous on this topic.

The best criticism of my work here is that some victims are more politically significant than others and their murder can cause serious problems, whereas a greater number of victims in politically motivated mass shootings has less of an effect. The 1960s assassinations of President John F. Kennedy, Rev. Martin Luther King Jr., and Sen. Robert F. Kennedy Jr. were just three murders in a turbulent decade, but they shaped political trends, fomented widespread riots, and prompted other forms of political violence that affected millions. Thus, statistically weighting the murders of prominent people such as Charlie Kirk more heavily than the victims of a mass shooting is intuitively appealing when judging the intensity of politically motivated violence. If there were a consistent and transparent means to weight those lives more heavily when judging the political effects of violence, then it would be worth employing that method, but such a means is absent here and I don’t know of a good way to do so that isn’t just arbitrary.

Below is a list of all the politically motivated killers.

Here are our methods before the results. You can skip this paragraph if you just want to see the results. Like our brief, we use the Survey of Income and Program Participation (SIPP) to evaluate welfare use by program for immigrants and native-born Americans by poverty and at different income quartiles. Our brief looks at individual welfare use, which is almost always the best variable for any analysis because individuals always contain the same number of people: one. The other choice would be a household-level analysis, which would include many native-born Americans and different numbers of people. Here, like our brief, we use the individual as the unit of analysis. SIPP is an unfortunately small survey, so we are limited to reporting only welfare programs where the sample sizes are large enough to be statistically significant.

We first use the poverty variable from the SIPP. This variable indicates whether an individual belongs to a household in poverty by providing the household income divided by the poverty threshold for that household. Immigrants, both in and out of poverty, report less welfare consumption than native-born Americans across almost all programs, with a few exceptions. Immigrants who are not in poverty report slightly higher consumption of Medicaid and WIC, and immigrants in general report higher rates of EITC recipiency. However, immigrants generally use less welfare than native-born Americans with the same level of poverty (Figures 1 and 2).

This pattern is also consistent across the income distribution. SIPP allowed us to break out welfare use by income quartile. We find that immigrants report lower welfare participation than native-born for nearly all programs in every income quartile. This holds not only among people in poverty, but also among those not in poverty. The only exceptions are that immigrants in the fourth (highest) income quartile report more Medicaid and SNAP use than natives (Table 1). There’s no good reason for that, as nobody in the highest quartile of income should be receiving those benefits, even if they are old-age entitlement programs, to say nothing of food stamps.

Another exception to this pattern is the rate of earned income tax credit (EITC) recipiency. Immigrants across all poverty and income subgroups consistently receive EITC at a higher rate than native-born Americans. Unlike means-tested and old-age benefit programs, EITC is framed as a tax credit and does not entail the same kinds of eligibility constraints as the traditional programs. That likely explains its higher use among immigrants (Table 2). Regardless, there is no compelling public policy reason for this, even if you otherwise like redistribution.

Interestingly, the only groups that reported statistically significant Temporary Assistance for Needy Families (TANF) use were native-born in both the lowest income quartile and out of poverty. No grouping of immigrants reported significant TANF use. These results, taken together, show the boundaries of reported TANF use: Immigrants tend not to get TANF, and anyone above the bottom 25 percent of incomes tends not to get it at all.

Note that the Figures above are not directly comparable to those in our brief. This analysis here uses only the weighted means for the demographic subgroups in question, but it does not adjust for undercounting using demographic or administrative data. Thus, the Figures here should not be used to calculate aggregate usage amounts across groups and should only be compared to each other (i.e., immigrants in poverty compared with native-born Americans in poverty).

The data consistently show that immigrants report less welfare use than native-born Americans across income levels and poverty statuses. Whatever one’s views on immigration policy, the claim that immigrants systematically consume more welfare than native-born Americans isn’t supported in the data. Rather than vilifying immigrants or closing the border, the more fiscally responsible policy position is to build a wall around the welfare state and thus limit welfare to US citizens only. The optimal policy is to remove the welfare state entirely.

We’re lucky to have smart and dedicated readers who closely examine our research. Thank you to the reader for emailing us this question. To him and our other readers, please keep the questions coming.

Subscribe now

In this paper, we wanted to accomplish two main things:

1) provide the first-ever assessment of the total net fiscal effect of all immigrants from 1994 to 2023, rather than a one-year snapshot or forward-looking projection like many other studies. We wanted a sufficiently long period to assess claims like those by White House Deputy Chief of Staff Stephen Miller, asserting that immigrants have already sucked us dry.

2) provide the clearest explanation for the mechanisms driving the fiscal effects of immigration on government budgets.

Immigrants have reduced the deficit every year

Every year since 1994—when data collection began—immigrants have paid more in taxes than they received in benefits from the federal, state, and local governments. The fiscal benefits have continued to rise, reaching their highest level ever in 2023.

The fiscal surplus from all immigrants from 1994 to 2023 was $14.5 trillion, compared with a deficit of $48 trillion without immigrants. That means that immigrants cut deficits by nearly a third in real terms over the last three decades.

Why the average person is fiscally positive

How can immigrants be so fiscally beneficial when the country overall is running such extreme deficits? The answer is that a big part of the US budget is pure public goods—primarily the military and interest payments on past debt accrued before the immigrants came—which don’t scale with population growth. These are essentially fixed costs or sunk obligations that the United States will have to cover whether immigrants come or not.

The figure below shows how, in most years, tax revenue exceeds the costs of providing benefits—that is, everything that does require scaling with population growth. Thus, immigrants will be fiscally positive so long as they are at least average in their revenue creation and benefits received. In fact, immigrants are significantly better than average in both aspects of the fiscal equation.

Immigrants pay more taxes, receive fewer benefits

Immigrants pay more in taxes than the average person. This is counterintuitive because they have lower wages on an hourly basis, but because they work at so much higher rates (the blue line), they end up having higher per capita incomes (the gray line) and paying more in taxes than their share of the population predicts (the dotted line). Thus, immigrants have been better at generating revenue for the government than the average person.

Are their tax revenues overwhelmed by the costs they impose? Here’s everything the federal, state, and local governments spent money on over the last 30 years in per capita dollar amounts. Immigrants did not create significantly higher costs for any items and saved the government enormously in two areas: old age benefits and education costs.

Immigrants cost less as retirees: First, the savings on old age benefits is not because immigrants are significantly less likely to be retirees. Instead, it is because they are far less likely to be receiving a government pension since they were less likely to have government jobs, so they were less likely to receive expensive government pensions. The main reason, though, is that they were simply barred from applying for Social Security and Medicare because they either arrived too late in life to earn the necessary qualifying work history or they are here illegally or in a temporary status and ineligible for that reason.

Immigrants cost schools less: Immigrants arrive in the United States at the average age of about 25, meaning that the United States gets workers without having to pay to educate them. Even though they are more costly when they are in school—due to bilingual education needs—they are much less costly overall since they are so much less likely to be in school. The result is that immigrants cost the US education system about half as much as the US-born population.

Immigrants aren’t big welfare users. The savings on education aren’t lost in the welfare state. Immigrants are much more likely to be in poverty but use roughly an average amount of what we call “needs-based” assistance. That includes traditional welfare, food assistance, Medicaid, refundable tax credits, and unemployment insurance. The entire reason for this disconnect is that many immigrants are here illegally and so are ineligible to apply for welfare in most states. This conclusion, that immigrants use welfare at the same rate as the US-born population, matches the Trump administration's conclusion in 2018.

Here is the full picture of spending and taxes for immigrants from 1994 to 2023. Immigrants—legal and illegal—paid more in taxes every single year than they received in benefits, broadly defined, and the gap has grown over time.

Immigrants don’t cause deficits.

Here’s another way to look at our main conclusion. Immigrants accounted for 14 percent of tax revenue and 7 percent of government spending from 1994 to 2023. Even if the government had not spent a dollar on immigrants, while somehow still getting all their tax revenue, the US government at all levels would still have run a $20 trillion deficit. Immigrants are not to blame for government deficits. Indeed, they reduced the deficit by about $14.5 trillion.

We use the highest quality data available for this report, using the best methods for this type of analysis. Although there are undoubtedly methodological finer points that can be debated, these broad conclusions are inescapable:

1. The average additional person is fiscally positive because pure public goods are such a big portion of the budget.

2. Immigrants generate more tax revenue. Immigrants’ rates of employment are well-documented. The correlation between income and taxes is well-established.

3. Immigrants use fewer benefits. The effects of status-based limits on welfare and entitlements are clearly apparent in numerous data sources. The savings on education are indisputable because immigrants are less likely to be enrolled in school.

Since these effects are not driven by the absence of immigrant retirees, we shouldn’t expect our conclusion to reverse after tracking a specific cohort of immigrants over time. Indeed, when we do follow the cohort that entered from 1990 to 1993, we find that after three decades, the cohort was still paying far more in taxes than they received in benefits and that the fiscal gains had grown over time. In total, this cohort reduced the deficit by $1.7 trillion.

Our paper also concludes:

• Without the contributions of immigrants, public debt at all levels would already be above 200 percent of US GDP—nearly twice the 2023 level and a threshold some analysts believe would trigger a debt crisis.

• Even low-skilled immigrants—those without bachelor’s degrees—reduced the debt by $2.8 trillion.

• Immigrants in all categories of educational attainment, including high school dropouts, lowered the ratio of deficit to gross domestic product (GDP) during the 30-year period.

• Illegal immigrants likely reduced the deficit by at least $1.7 trillion.

• Even including the second generation, who are mostly still children who will become taxpayers soon, the fiscal effect of immigration was positive every year, reducing the debt by $7.9 trillion.

The tables at the end of this post provides the effects on the deficit-to-GDP by educational attainment and citizenship status for both 2022-2023 and 1994-2023.

Concluding thoughts

Overall, the main conclusion of our paper is that there is nothing systematically wrong with US immigration policy with respect to the fiscal effects of immigrants. There is nothing unsustainable about the US immigration system. We could have scaled immigration as it existed without burdening government budgets. For years, nativists in Congress and the administration have wrongly claimed that immigrants are behind the growth in debt and that the US immigration system allowed foreigners to take advantage of Americans’ generosity. Our data completely repudiates this view. Immigrants are subsidizing the US government.

The best way to balance the budget is to reduce spending—particularly on wealthy retirees—but rather than hinder our efforts to control deficits, immigrants are helping.

You can read the entire study here: Immigrants’ Recent Effects on Government Budgets: 1994–2023

Subscribe now

In my earlier analysis, I used data from the US Sentencing Commission to show that noncitizens were 8 percent less likely to be convicted of welfare fraud in 2024 and that the number of noncitizens who were convicted of welfare fraud had fallen by 57 percent since 2015. In this analysis, I expand on those results to examine the dollar amounts stolen from the federal government over the period 2013 to 2024.

The US Sentencing Commission data start in 2013. In inflation-adjusted 2024 dollars, the average citizen was convicted of stealing $1.68 per year from government benefits programs from 2013 to 2024, while the average noncitizen was convicted of defrauding $1.16 per year, 31 percent less than the average citizen.

Altogether, the US Sentencing Commission reports on $5.5 billion in benefits fraud losses attributed to convicts during this period—or $6.5 billion in inflation-adjusted 2024 dollars. Of this amount, citizens accounted for 95 percent of the losses in real terms. Noncitizens do commit benefits fraud, but they are not a significant part of the problem.

The US Sentencing Commission does not record a person’s birthplace or whether a citizen has naturalized. But we have reasons to believe that the noncitizen results shed light on the naturalized citizen rate since all naturalized citizens start as noncitizens. Indeed, naturalized citizens have a much lower rate of criminality measured by their propensity to commit crimes, their rate of conviction, and their rate of incarceration because serious noncitizen criminals are barred from naturalizing.

These results are counterintuitive since public benefits fraud is one area where we should naturally expect noncitizens to be overrepresented. Noncitizens are subject to unique restrictions on eligibility for benefits that do not apply to citizens, so we should naturally expect the additional limits would lead to a higher rate of fraud by noncitizens.

As I said in my prior blog post on this topic, benefits fraud is not a compelling justification for broad-based restrictions on immigration. The primary economic benefit of immigration to Americans is the goods and services that they produce. But on the whole, immigrants also reduce budget deficits and lower the debt-to-GDP ratio, and their positive fiscal impact will only grow as the US debt burden rises.

Benefits fraud is a serious issue, and whoever commits it should be held accountable, including through deportation in the case of a noncitizen. Unfortunately, this current Congress chose to spend hundreds of billions on a chaotic, indiscriminate mass deportation program in July that will increase the deficit rather than invest in better systems and investigations to control benefits fraud. Congress should redirect that money toward welfare fraud enforcement.

See also: Noncitizens Were Underrepresented in Welfare Fraud Convictions in 2024