Of all the statutes examined in this series, the Immigration and Nationality Act of 1965, Public Law 89-236, is the one whose effects are most widely asserted and least precisely understood. Ask what the act did to American demographics and you will hear confident answers that point in opposite directions. One version holds that the law’s designers engineered the transformation deliberately, steering admissions away from Europe and toward Latin America and Asia with full knowledge of the result. The other holds that the act was a minor technical adjustment to a system that was already changing, and that its demographic footprint is overstated. The census record, the statutory text, and the peer-reviewed research support neither story as told. What they support is a third account, more mechanical and more interesting: a small number of specific provisions, chief among them the exemption of immediate relatives from every numerical limit and a family preference structure that compounds as each newly naturalized citizen becomes a petitioner, produced a transformation far larger than the law’s nominal ceilings suggested. The demographic facts that resulted are not in dispute. The economic and fiscal effects that followed are genuinely contested, and the research on both sides of those disputes deserves to be named rather than summarized away.

This article is built for a single reader test. When a reader finishes, they should be able to state precisely what changed in the size and the origin composition of American migration after 1965, explain the statutory mechanism that produced each change, separate the demographic facts that are not in dispute from the economic and fiscal effects that genuinely are, and name the research on both sides of the contested questions. That test dictates the article’s architecture. The first layer walks the census series from 1970 to 2018, decade by decade, and the origin composition series from 1960 to 2010, with the sources and years attached to every figure. The second layer explains the mechanism: the unlimited immediate relative category, the compounding family preference structure, and the small initial base from Asia and Latin America that explains why growth from those regions started slowly and then accelerated. The third layer presents the causal argument of Douglas Massey and his coauthors about how the imposition of Western Hemisphere numerical limits, combined with the ending of the agricultural labor program, converted a long-standing circular and largely lawful labor flow from Mexico into an unauthorized one, and how later border enforcement reduced return migration more than entry, increasing the settled unauthorized population. The fourth layer works through the contested economics: the National Academies of Sciences consensus report on the economic and fiscal consequences of migration, and the unresolved methodological dispute between George Borjas and David Card over the same natural experiment, the Mariel boatlift. A reader who can name all four layers, with sources, has the article’s substance.

The statutory identity of the 1965 act matters because the outcomes are measured from it. Public Law 89-236, 79 Stat. 911, signed October 3, 1965, took effect in two stages: the transition provisions, including the pooling of unused national quotas on a first-come first-served basis, took effect December 1, 1965, and the full abolition of the national origins quota system together with the new hemispheric caps took effect July 1, 1968. The principal provisions thus operated between 1965 and 1968, and the effects are measured from the act and its three major successors: the 1976 amendments, Public Law 94-571, the 1978 amendments, Public Law 95-412, and the Immigration Act of 1990, Public Law 101-649. A fuller treatment of the 1965 act’s provisions and its full statutory profile belongs to the statute guide in this series, and the legislative history of its passage to the passage article; this article measures outcomes. The discipline of the article is that every demographic fact carries a census citation with years, every economic and fiscal effect is anchored to the 2017 consensus report plus named dissenting researchers, and no claim appears that cannot be sourced to a named study or a government data series with a period attached.

One more boundary belongs at the start, because this article operates under the strictest neutrality rules in the series. Demographic facts get census citations with years. Economic and fiscal effects get the consensus report plus named dissenting researchers. No claim about crime, culture, or assimilation appears without a named peer-reviewed source and its stated limitations, which means in practice that no such claims appear in this article at all. The article never characterizes any national origin group, never estimates any current population figure, and never references a present-day policy dispute. Claims about the act’s effects are measured against what the statute actually did, with the disputed questions labeled disputed, which is the series thesis thread carried through this evidence article.

The reader who wants the short version can hold this much and the article will defend it in detail. Between the 1970 census and the 2018 American Community Survey, the foreign-born share of the population rose from 4.7 percent, an all-time low, to 13.7 percent, approaching the level last seen in the early twentieth century. Between the 1960 decennial census and the 2010 American Community Survey, the origin composition of the foreign-born population shifted from a large European majority, 75 percent from Europe in 1960, to a Latin American and Asian majority, 53 percent from Latin America and 28 percent from Asia in 2010. The engine was statutory, not demographic drift: immediate relatives of citizens entered without regard to any numerical limit, family preference categories compounded as each new citizen gained the power to petition, and the small initial base from Asia and Latin America meant that growth from those regions started slowly and then accelerated. On the economics, the 2017 National Academies consensus reported very small overall wage effects over periods of ten years or more, with possible negative effects concentrated on workers without a high school credential and on prior migrants, positive long-run effects at the federal budget level and negative effects at the state and local level driven by schooling costs. The disagreement between George Borjas and David Card over the Mariel boatlift remains an unresolved methodological dispute, and this article presents it as one.

The foreign-born share: what the census records from 1970 to 2018

The single most important demographic fact about the post-1965 period is also the simplest to state with precision, because the census series runs without a break from 1970 forward. At the 1970 decennial census, the foreign-born population of the United States stood at 9.6 million people, or 4.7 percent of the total population. That figure was the all-time low in the recorded series, lower than at any point since the census began asking about nativity, and well below the 14.8 percent recorded at the 1890 census, the historic peak. By the 2018 American Community Survey, the foreign-born population stood at 44.8 million people, or 13.7 percent of the total population. Between those two points the foreign-born share roughly tripled, the foreign-born count more than quadrupled, and the country moved from the smallest foreign-born share in its recorded history back toward the level last seen in the early twentieth century. Every step of that walk can be dated and sourced, and walking it step by step matters, because the pace and the timing are the evidence that bears on which statutory provisions did the work.

The 1970 figure deserves attention first, because it fixes the starting point from which the 1965 act’s effects are measured. The 9.6 million foreign-born residents counted in 1970, 4.7 percent of the population, reflected the long demographic tail of the national origins quota system that the 1965 act replaced. Under the 1924 system, admissions had been held to low annual levels for more than four decades, and the foreign-born population had aged and shrunk as a share of a growing native population. The 1970 census was taken after the 1965 act had been signed but before its principal provisions had fully taken effect, the full abolition of the quota system and the hemispheric caps arriving July 1, 1968, and before any compounding mechanism had time to operate. The 1970 figure is therefore best read as the baseline of the old regime rather than as evidence about the new one. The source for the pre-2000 figures in this series is the decennial census long-form data, tabulated in the Census Bureau’s historical series, and the source shift to the American Community Survey for the post-2000 figures is noted where the series crosses it.

By the 1980 decennial census, the foreign-born population had risen to 14.1 million, or 6.2 percent of the population. The increase from 1970 to 1980, roughly 4.5 million people in a decade, was the first census-to-census gain that the 1965 act’s machinery can be seen operating inside, though the decade also carried the effects of refugee admissions and the end of the transition period. The rise from 4.7 percent to 6.2 percent looks modest against what followed, but it was the first upward movement in the foreign-born share in more than half a century, and it established that the direction of the series had changed. The 1980 figure is drawn from the decennial census long form, consistent with the 1970 figure, so the 1970 to 1980 comparison is an apples-to-apples comparison within a single data series.

By the 1990 decennial census, the foreign-born population had risen to 19.8 million, or 7.9 percent of the population. The 1980s were the decade in which the family preference compounding described in the mechanism layer of this article began to show in the aggregate numbers, as the relatives of those admitted in the 1970s naturalized and petitioned for their own relatives. The decade also contained the 1986 Immigration Reform and Control Act, which legalized approximately 2.7 million people and thus moved a large population from unauthorized to lawful permanent status, an event that belongs to the unauthorized migration layer of this article and that complicates any simple reading of the 1990 figure as a pure measure of new arrivals. The legalization figure is documented in the fact base of the series as approximately 2.7 million approvals for permanent residence, combining about 1.6 million under the general legalization provisions of INA section 245A with about 1.1 million under the Special Agricultural Worker program. The 1990 census figure of 19.8 million, 7.9 percent, is the last figure in the series drawn from the decennial long form before the source shift.

The 2000 decennial census recorded 31.1 million foreign-born residents, or 11.1 percent of the population. The 1990s were the decade of the largest absolute gain in the series, an increase of more than 11 million foreign-born residents in ten years, and the decade in which the foreign-born share crossed into double digits for the first time since the early twentieth century. The 2000 figure was the last decennial long-form figure; from this point forward the series shifts to the American Community Survey, and the article notes that shift wherever decennial and ACS figures sit near each other. The 2009 American Community Survey recorded 38.5 million foreign-born residents, or 12.5 percent of the population. The 2010 American Community Survey recorded 40.0 million foreign-born residents, or 12.9 percent of the population. The Pew Research Center’s tabulation of the 2013 data placed the foreign-born population above 41 million, or 13.1 percent of the population, and the 2018 American Community Survey recorded 44.8 million, or 13.7 percent. The sources for the post-2000 figures are the American Community Survey one-year estimates for the stated years and the Pew Research Center tabulations of the 1960 through 2018 censuses and surveys via IPUMS, as published in the 2015 report on the modern migration wave and the 2020 trend data compilation covering 1960 to 2018.

Two features of this walk deserve emphasis because they are the evidence most often mishandled. The first is the source shift. Figures for 1970 through 2000 come from the decennial census long-form data, while figures from 2009 forward come from the American Community Survey. The two series are not identical instruments, and a careful reader should not treat a comparison across the shift as precise to the decimal point. The article has paired the series the way the Census Bureau’s own long-term trend table pairs them, 1850 through 2000 from the decennial censuses and the 2010 point from the ACS, and the Pew tabulations carry the series forward through 2018. The direction and the scale of the change are robust to the source shift; the exact percentage points are not worth disputing because the instruments differ.

The second feature is the approach to the 1890 record. At 13.7 percent in 2018, the foreign-born share approached but had not reached the 14.8 percent recorded at the 1890 census, the all-time high. The phrase approaching the early-twentieth-century level, used in the brief for this article, is the correct characterization: the country had not returned to the 1890 peak by 2018, but it had come back into the neighborhood of the levels that prevailed when the foreign-born share stood at its historic highs. This matters because both overreach stories misuse the comparison. The intentional-design story treats the return toward early-century levels as though the 1965 act’s authors aimed at a specific number. The minor-amendment story treats the comparison as proof that nothing unusual happened. The record supports neither gloss. What the record shows is a measured fact: a rise from the all-time low of 4.7 percent in 1970 to 13.7 percent in 2018, with the pace of increase accelerating in the 1990s and continuing through the 2010s, driven by the statutory mechanism this article explains in its second layer.

The 1990s: the decade of the largest gain

Every decade in the census walk contributed to the post-1965 transformation, but the 1990s contributed the most in absolute terms and therefore deserves a closer examination. Between the 1990 and 2000 decennial censuses, the foreign-born population rose from 19.8 million to 31.1 million, an increase of more than 11 million people in ten years, and the foreign-born share rose from 7.9 percent to 11.1 percent, crossing into double digits for the first time since the early twentieth century. No other decade in the recorded series added as many foreign-born residents. Understanding why the 1990s were the peak decade requires pulling together several threads this article has treated separately: the maturation of the family preference compounding, the IRCA legalization, and the 1990 act’s expansion.

The compounding mechanism explains the underlying acceleration. By the 1990s, the family preference system had been operating for more than two decades, which meant that multiple cycles of admission, naturalization, and petitioning had completed. The relatives admitted in the 1970s had naturalized in the 1980s and were petitioning for their own relatives in the 1990s; the relatives admitted in the 1980s were beginning the same cycle. Each completed cycle enlarged the petitioner base, so the admissions the system generated grew even without any change in the statute. The 1990s were thus the decade in which the compounding reached a scale large enough to dominate the aggregate figures, the point at which the multiplier operating on an enlarged base produced the largest absolute gains in the series. The small-base explanation for the slow start in the late 1960s and the acceleration thereafter finds its confirmation in the 1990s: the mechanism predicted that growth would accelerate as the base grew, and the census recorded exactly that.

The IRCA legalization complicates the 1990s figures in a way the article has noted but that deserves emphasis here. The approximately 2.7 million people approved for permanent residence under the 1986 act, about 1.6 million under the general legalization and about 1.1 million under the Special Agricultural Worker program, were counted in the 1990 census as foreign-born residents, and their presence in the 1990 baseline means that the 1990 to 2000 gain cannot be read as a pure measure of new arrivals in the 1990s. Some portion of the decade’s growth reflects the legalization’s movement of people from unauthorized to lawful status in the late 1980s rather than new migration in the 1990s. The legalization also had a compounding consequence of its own: the newly legalized residents became eligible, after the required years of permanent residence, to naturalize and petition for relatives, which meant that the 1986 legalization fed the family preference compounding of the 1990s and 2000s. The statute’s effects thus operated across decades, with each decade’s legislation shaping the next decade’s numbers.

The 1990 act’s expansion took effect during the decade, with the 700,000 worldwide level applying to fiscal years 1992 through 1994 and the 675,000 level from fiscal year 1995. The larger employment-based track and the diversity lottery added new admissions on top of the family-channeled flow, contributing to the decade’s totals. The composition figures for 2000 reflect the combined operation of all three dynamics: Latin America at 51.7 percent, Asia at 26.4 percent, and Europe at 15.8 percent of the foreign-born population, with Latin America plus Asia at 78.2 percent, up from 28.3 percent in 1970. The 1990s were the decade in which the post-1965 transformation became unmistakable in the aggregate numbers, not because any single provision caused it, but because the compounding mechanism, the legalization’s aftereffects, and the 1990 expansion all pushed in the same direction at once.

The 2000s and 2010s: continuation and compositional drift

The transformation did not stop with the 1990s; the two decades that followed continued the rise in the foreign-born share while shifting the composition within the new majorities. The 2000 decennial census recorded 31.1 million foreign-born residents, 11.1 percent of the population. The 2010 American Community Survey recorded 40.0 million, 12.9 percent. The 2018 American Community Survey recorded 44.8 million, 13.7 percent. The pace of increase in percentage-point terms slowed compared with the 1990s, which is what a maturing compounding process predicts: the petitioner base was by then so large that each cycle added large absolute numbers but smaller proportional gains, and the foreign-born share was approaching the upper range of the historical series. The 2013 tabulation at 13.1 percent and the 2009 ACS at 12.5 percent fill in the path between the decade points, all from the American Community Survey series that the article has used consistently for the post-2000 period.

The compositional drift within the period is as instructive as the continued growth. In the Pew tabulation, the Mexican share of the foreign-born population stood at 29 percent in 2010 and declined to 25 percent in 2018, while the Asian share rose from 25 percent to 28 percent and the share from elsewhere in Latin America held at 24 to 25 percent. Europe and Canada together stood at 14 percent in 2010 and 13 percent in 2018. The Mexican decline and the Asian rise reflect the changing composition of new admissions in the 2010s, as the family preference compounding from Asian countries, operating on bases built in the 1980s and 1990s, generated growing admissions, while net migration from Mexico slowed. The Census Bureau’s country-level comparison reinforces the point: Mexico remained the leading country of birth by a wide margin at 11.7 million in 2010, but the next nine countries were all in Asia or Latin America, with China at 2.2 million, India and the Philippines at 1.8 million each, and Vietnam, El Salvador, Cuba, and Korea each above one million. The drift shows the mechanism still operating two generations after the act: the compounding that began with small Asian and Latin American bases in the 1960s was, by the 2010s, producing the compositional changes visible in the survey data, with no new statute required.

The origin shift: from a European majority to a Latin American and Asian majority

The change in the size of the foreign-born population is only half of the demographic story the 1965 act produced. The other half is the change in where the foreign-born population came from, and it is the more dramatic of the two. At the 1960 decennial census, taken five years before the act was signed, the foreign-born population of the United States was overwhelmingly European: 75 percent of foreign-born residents had been born in Europe, 10 percent in Northern America, principally Canada, 9 percent in Latin America, and 5 percent in Asia, with the remainder elsewhere. By the 2010 American Community Survey, the composition had inverted: 12 percent of foreign-born residents had been born in Europe, 2 percent in Northern America, 53 percent in Latin America, and 28 percent in Asia. The European share fell from three quarters to roughly one eighth. The combined Latin American and Asian share rose from about one seventh to more than four fifths. These are the figures from the Census Bureau’s chart pairing the 1960 decennial census with the 2010 American Community Survey, and they are the undisputed core of the origin story. State them with the sources and the years and do not editorialize: 1960, Europe 75 percent, Northern America 10 percent, Latin America 9 percent, Asia 5 percent; 2010, Europe 12 percent, Northern America 2 percent, Latin America 53 percent, Asia 28 percent.

The Pew Research Center’s tabulation of the same period, grouping Europe with Canada, tells the same story in slightly different numbers: in 1960, 84 percent of the foreign-born population came from Europe or Canada, with 6 percent from Mexico, 3 to 4 percent from elsewhere in Latin America, and 4 percent from South and East Asia. In 2010, by the Pew tabulation, 29 percent came from Mexico, 24 percent from elsewhere in Latin America, 25 percent from Asia, and 14 percent from Europe or Canada. The two tabulations differ in their regional groupings and therefore in their exact percentages, but they agree completely on the direction and the scale of the change. The 1970 decennial census, tabulated by Pew, shows the composition partway through the transition: 68 percent from Europe or Canada, 8 percent from Mexico, 11 percent from elsewhere in Latin America, and 7 percent from Asia. The 1970 figure is a useful intermediate point because it captures the composition after the act’s signing but before the compounding mechanism had time to reshape it, and it shows how much of the shift was still to come.

The 2000 decennial census provides the midpoint of the transformation in the Census Bureau’s own regional classification. In 2000, 51.7 percent of the foreign-born population had been born in Latin America, 26.4 percent in Asia, and 15.8 percent in Europe. Latin America plus Asia accounted for 78.2 percent of the foreign-born population in 2000, up from 28.3 percent in 1970 on the comparable classification. The 2018 American Community Survey, in the Pew tabulation, shows the composition continuing to evolve: 25 percent from Mexico, 25 percent from elsewhere in Latin America, 28 percent from Asia, and 13 percent from Europe or Canada. The Mexican share had declined from its 2010 level of 29 percent to 25 percent in 2018, while the Asian share had risen from 25 percent to 28 percent, a shift that reflects the continuing operation of the preference system and the changing composition of new admissions in the 2010s. The Census Bureau’s 2009 place-of-birth report summarized the arc in a single sentence that has been widely quoted: in 1960, 75 percent of the foreign born were from countries in Europe, and by 2009, over 80 percent of the foreign born were from countries in Latin America and Asia. The 2010 ACS figures of 53 percent Latin America and 28 percent Asia, summing to 81 percent, confirm the report’s characterization.

The country-level data make the composition shift concrete in a way that regional percentages cannot. The Census Bureau’s comparison of the top countries of birth in 1960 and 2010 reads like a changing of the guard. In 1960, the ten leading countries of birth for the foreign-born population were Italy with 1.3 million, Germany with 1.0 million, Canada with 1.0 million, the United Kingdom with 0.8 million, Poland with 0.7 million, the Soviet Union with 0.7 million, Mexico with 0.6 million, Ireland with 0.3 million, Austria with 0.3 million, and Hungary with 0.2 million. Seven of the ten were European countries, and the list was dominated by the nationalities that the national origins quota system had favored. In 2010, the ten leading countries of birth were Mexico with 11.7 million, China with 2.2 million, India with 1.8 million, the Philippines with 1.8 million, Vietnam with 1.2 million, El Salvador with 1.2 million, Cuba with 1.1 million, Korea with 1.1 million, the Dominican Republic with 0.9 million, and Guatemala with 0.8 million. No European country appears in the 2010 top ten at all. Mexico alone, at 11.7 million, accounted for more foreign-born residents in 2010 than the entire 1960 top ten combined. The comparison of the two lists, both drawn from the Census Bureau’s place-of-birth tabulations, is the most vivid single illustration of what the origin shift meant in human terms, and it requires no editorial comment to land.

A persistent question about the origin shift is whether it was caused by the 1965 act or whether it merely coincided with broader forces, such as the economic development of Asia, the demographic transitions of Latin America, and the decline of European emigration as European living standards rose. The honest answer is that both operated, and the article’s mechanism layer explains how the statute translated global conditions into American admissions. The quota system the 1965 act replaced had allocated the overwhelming majority of visas to Europe by design; the national origins formula allocated quota slots by national origin in a way that heavily favored Europe. When the 1965 act abolished that formula and replaced it with hemispheric ceilings and a uniform preference system, it removed the structural barrier that had kept Asian and Latin American admissions low, and the family preference compounding then amplified the admissions from the regions whose migrants had established the initial footholds. The global forces supplied the demand for migration; the statute determined which demand could be satisfied through lawful channels and how each admission generated further admissions. Neither the global-forces story alone nor the statute-alone story is complete. The mechanism layer of this article works through the statutory half of that interaction in detail.

How did the 1965 Act structure its numerical limits?

The 1965 Act set a 170,000 annual ceiling for the Eastern Hemisphere, a 20,000 per-country limit within that hemisphere, and a 120,000 annual ceiling for the Western Hemisphere, while exempting immediate relatives of citizens from all numerical limits. The Eastern provisions took full effect July 1, 1968.

The architecture of the limits matters more than the numbers, because the numbers were not what drove the demographic outcome. Section 201(a) of the amended Immigration and Nationality Act, codified at 8 U.S.C. 1151, set the Eastern Hemisphere ceiling at 170,000 immigrants per fiscal year, with a quarterly sub-cap of 45,000, operative from July 1, 1968. Section 202(a), codified at 8 U.S.C. 1152, set a limit of 20,000 entries per year for natives of any single foreign state within the Eastern Hemisphere. Section 21(e) of the act capped Western Hemisphere special immigrants at 120,000 per fiscal year, effective for the fiscal year beginning July 1, 1968 and every fiscal year thereafter. Section 201(b) defined immediate relatives, the children, spouses, and parents of a United States citizen, with the petitioning parent required to be at least 21 years of age, as admitted without regard to the numerical limitations of the act. That exemption is the provision that matters most for everything that follows in this article, because it meant that the fastest-growing category of admissions faced no ceiling at all.

The preference system that allocated the capped visas is codified at section 203, 8 U.S.C. 1153, and its structure is worth stating precisely because the compounding mechanism runs through it. The Eastern Hemisphere preferences reserved the capped visas for, in order, unmarried sons and daughters of citizens, spouses and unmarried sons and daughters of permanent residents, professionals and persons of exceptional ability, married sons and daughters of citizens, siblings of citizens, and workers in occupations facing labor shortages. The family preferences, the first, second, fourth, and fifth, were the ones that compounded: each admission in a family category created a potential new petitioner once the admitted relative naturalized, and each new petitioner could sponsor further relatives, who could in turn naturalize and petition. The employment preferences did not compound in the same way, because employment-based admission did not create new petitioning power in the family categories. The statute thus contained two different growth dynamics inside one preference system, a linear one for employment and a compounding one for family, and the demographic record reflects the compounding one.

A nuance about the per-country limit belongs here because it becomes central in the unauthorized migration layer of this article. The 20,000 per-country limit in the 1965 act applied to the Eastern Hemisphere only. The Western Hemisphere received no per-country limit under the 1965 act; it received only the 120,000 hemispheric ceiling. The per-country limit was not extended to the Western Hemisphere until the 1976 amendments, Public Law 94-571, which applied the 20,000 limit hemisphere-wide, with initial exceptions of 35,000 each for Canada and Mexico. That 1976 extension is the hinge of the Massey argument presented in this article’s third layer, because it imposed for the first time a binding per-country cap on the Western Hemisphere country whose lawful admissions most exceeded it, Mexico. The 1965 act’s own Western Hemisphere ceiling of 120,000, though, was itself a new restriction: before 1965, Western Hemisphere migration had faced no numerical limit at all, so the 120,000 ceiling was the first cap ever imposed on the hemisphere and the first binding constraint on the circular labor flow from Mexico that the third layer of this article examines.

The mechanism: the immediate relative exemption and the compounding family preference structure

The demographic transformation after 1965 followed from two specific provisions, and the article states the finding once, verbatim, as the brief requires: “The mechanism-not-intent finding: the demographic transformation after 1965 followed from two specific provisions, the unlimited immediate relative category and the compounding family preference structure, and explaining it by reference to anyone’s intentions, in either direction, is a weaker account than explaining it by reference to the text.” The rest of this section explains the two provisions and shows how they operated on the census record.

The first provision is the immediate relative exemption. Section 201(b) of the amended act admitted the immediate relatives of United States citizens without regard to numerical limitation, and immediate relatives were defined as the children, spouses, and parents of citizens, with a citizen child required to be at least 21 to petition for a parent. The exemption had two properties that made it the dominant engine of post-1965 admissions growth. First, it was uncapped, so it could grow without any act of Congress and without competing against other categories for a fixed pool of visas. Second, it was self-expanding, because every immediate relative admitted was a future citizen who could petition for their own immediate relatives. A citizen who sponsored a spouse created a new citizen, the naturalized spouse, who could sponsor a parent, who after naturalization could sponsor a sibling under the family preferences, whose admission created further petitioning capacity. The chain had no statutory terminus. The compounding was not metaphorical; it was the literal operation of the petitioning rules, in which each link in the chain was a person exercising a right the statute granted.

The second provision is the family preference structure inside the capped system, codified at section 203. The preferences for unmarried sons and daughters of citizens, spouses and children of permanent residents, married sons and daughters of citizens, and siblings of citizens allocated the capped visas in a way that reinforced the compounding. Each family preference admission, like each immediate relative admission, created a future petitioner. The employment preferences and the refugee provisions did not compound in this way, because they did not create family petitioning rights. The result was that the family side of the system grew as a function of its own past admissions, while the employment side grew only as Congress adjusted it. Over the decades, the family side came to dominate total admissions, not because Congress repeatedly voted to expand it, but because the statute’s own logic expanded it automatically. The detailed statutory anatomy of these categories is treated in this series’ guide to the family preference visa system, and this article takes that anatomy as the engine whose output the census recorded.

The third element of the mechanism is not a provision but a condition: the small initial base from Asia and Latin America. In 1960, only 5 percent of the foreign-born population had been born in Asia and 9 percent in Latin America, which meant that the pool of potential citizen petitioners from those regions was small when the 1965 act took effect. The compounding mechanism operates as a multiplier on the existing base, and a multiplier applied to a small base produces small absolute numbers at first. This is why growth from Asia and Latin America started slowly in the late 1960s and early 1970s and then accelerated: the base had to be built before the compounding could show in the aggregate figures. The 1970 composition figures reflect this early stage, with Asia at 7 percent and Latin America, including Mexico, at 19 percent of the foreign-born population in the Pew tabulation, up from 1960 but still far from the 2010 levels. By 2000, with several compounding cycles complete, Latin America and Asia together accounted for 78.2 percent of the foreign-born population. The acceleration was not a change in the mechanism; it was the mechanism operating on a larger base, which is exactly what a compounding process predicts.

The mechanism also explains a feature of the record that the intent-based stories cannot explain: the act’s effects were largest in the categories the statute left uncapped and smallest in the categories it capped most tightly. The Eastern Hemisphere ceiling of 170,000 and the per-country limit of 20,000 did bind, and they bound most visibly in the employment preferences and in the oversubscribed family preferences of high-demand countries. But the immediate relative category, facing no ceiling, grew past every capped category, and the family preferences compounded within their caps as backlogs grew. A reader who understands this asymmetry understands why the nominal ceilings of the 1965 act are a poor guide to its demographic output. The ceilings described the part of the system that Congress controlled annually; the exemption described the part that grew by its own logic.

One further implication of the mechanism deserves to be stated plainly, because it bears on the contested economics layer. The compounding family structure meant that the skill and education composition of admissions was determined less by the employment preferences, which selected for labor market criteria, than by the family categories, which selected for kinship. The economic research on wage and fiscal effects, presented in this article’s fourth layer, must be read with that composition in mind: the population whose effects the economists measured was disproportionately admitted through kinship rather than through employment selection, and the National Academies consensus findings about small overall wage effects and mixed fiscal effects are findings about that population, not about a hypothetical skill-selected one. The mechanism thus connects the demographic layers of this article to the economic layer, through the composition of the flow the statute produced.

How the preference system allocated visas in practice

The mechanism layer of this article described the compounding logic of the family categories in outline. This section walks the preference system provision by provision, because the demographic record becomes legible only when the reader sees exactly how the statute rationed the capped visas and which categories escaped rationing entirely. The exposition stays inside the verified statutory facts: the ceilings, the per-country limits, the preference order, and the exemption.

The Eastern Hemisphere system that took full effect July 1, 1968, worked as follows. Section 201(a), codified at 8 U.S.C. 1151, capped Eastern Hemisphere admissions at 170,000 per fiscal year, with a quarterly sub-ceiling of 45,000 that prevented the year’s visas from being exhausted in a single quarter. Section 202(a), codified at 8 U.S.C. 1152, limited natives of any single foreign state to 20,000 entries per year within the hemisphere. Section 203, codified at 8 U.S.C. 1153, then allocated the capped visas among seven preferences in order: first, unmarried sons and daughters of citizens; second, spouses and unmarried sons and daughters of permanent residents; third, professionals and persons of exceptional ability in the sciences or arts; fourth, married sons and daughters of citizens; fifth, siblings of citizens; sixth, workers in occupations facing labor shortages as certified by the Labor Department; and seventh, refugees. The first, second, fourth, and fifth preferences were the family categories that compounded, because each admission in those categories created a future petitioner. The third and sixth were the employment categories, which selected on labor market criteria and did not create family petitioning rights. The seventh preference reserved a capped channel for refugees within the Eastern Hemisphere system.

The Western Hemisphere system that took effect the same day was simpler and, for the Massey argument, more consequential. Section 21(e) of the act capped Western Hemisphere special immigrants at 120,000 per fiscal year, but the act imposed no preference system and no per-country limit on the hemisphere. Admissions were allocated first-come, first-served up to the hemispheric ceiling. This was the first numerical limit the Western Hemisphere had ever faced, and its practical effect was to ration a flow that had previously been unlimited. The absence of a per-country limit meant that Mexican admissions, which were by far the largest in the hemisphere, could run up to the hemispheric ceiling without a country-specific constraint, a feature the 1976 amendments would remove.

The immediate relative exemption sat outside both systems. Section 201(b) admitted the children, spouses, and parents of United States citizens without regard to numerical limitation, with the requirement that a citizen child be at least 21 years old to petition for a parent. Because the exemption applied hemisphere-wide and preference-wide, it created a third channel alongside the two capped systems, and it was the only channel with no ceiling. The demographic significance of the three-channel structure is the asymmetry this article has emphasized: two capped channels that Congress could adjust, and one uncapped channel that grew by its own logic. Over the decades, the uncapped channel outgrew the capped ones, not through any further legislation but through the arithmetic of naturalization and petitioning.

The per-country limit deserves a closer look because it was the provision that translated the hemispheric ceilings into country-level outcomes. The 20,000 annual limit on natives of any single foreign state meant that high-demand countries quickly exhausted their allocations, creating backlogs in the oversubscribed family preferences while low-demand countries left visas unused. The limit thus had two faces: it restrained the largest sending countries, and it protected access for smaller ones. In the Eastern Hemisphere, the limit bound most visibly on the countries whose family networks were largest, which after the compounding cycles of the 1970s and 1980s increasingly meant Asian countries. In the Western Hemisphere after 1976, it bound on Mexico alone, because Mexico was the only Western Hemisphere country whose admissions substantially exceeded 20,000 annually. The per-country limit was therefore not a neutral administrative detail; it was the provision that determined which national flows the statute restrained and which it left alone, and its extension to the Western Hemisphere in 1976 was the single most consequential change in the geographic incidence of the system’s restraints.

The quarterly cap of 45,000 on Eastern Hemisphere admissions is a smaller provision worth noting for what it reveals about congressional intent in the statute’s design. By preventing the year’s 170,000 visas from being issued in a single quarter, the quarterly cap smoothed admissions across the fiscal year and prevented administrative surges. It did not affect the annual totals or the demographic outcomes, but it shows the statute’s drafters thinking about the flow as a managed stream rather than a single annual event. The provision is a reminder that the 1965 act was a complete administrative system, with ceilings, sub-ceilings, per-country limits, preference orders, and exemptions fitted together, and that the demographic outcome emerged from the interaction of all of these parts rather than from any single number.

When the 1978 amendments replaced the two hemispheric ceilings with a single worldwide ceiling of 290,000, the preference mechanics were carried forward into the unified system: the same seven-preference structure, the same 20,000 per-country limit now applied worldwide, and the same immediate relative exemption outside the ceiling. When the 1990 act raised the worldwide level to 700,000 for fiscal years 1992 through 1994 and 675,000 thereafter, it reorganized the preferences into the three-track structure of 480,000 family-sponsored, 140,000 employment-based, and 55,000 diversity admissions, with five employment-based preference categories replacing the old third and sixth preferences. Through both reorganizations, the provisions that drove the demographic outcome, the uncapped immediate relative category and the compounding family preferences, survived intact. The ceilings changed; the engine did not.

The Naturalization Lag: The Compounding Trigger

The compounding engine described in layer two has a trigger, and the trigger is naturalization. The distinction between what a lawful permanent resident can petition for and what a citizen can petition for is one of the most demographically consequential distinctions in the statute, and it is frequently overlooked because it operates as a time delay rather than as a number.

A permanent resident’s petitioning power is narrow. Under the preference system, a permanent resident can petition for a spouse and for unmarried sons and daughters, the second preference, and for no one else. A citizen’s petitioning power is broad: spouses, children of any marital status, parents, and siblings, with the spouses, minor children, and parents admitted outside the numerical limits entirely through the immediate relative exemption. The full branching power of the family system, the power that turns each admission into the seed of future admissions across multiple relationship categories, therefore activates only at naturalization. Between admission as a permanent resident and naturalization as a citizen lies a period of years during which the immigrant accumulates eligibility but cannot yet exercise the full petitioning power.

The demographic consequence is a generational lag built into the statute’s operation. The cohorts admitted under the new preference system in the late 1960s and early 1970s became citizens across the late 1970s and 1980s, and their petitions, filed in those years, produced the admission cohorts of the 1980s and 1990s. Those cohorts naturalized in turn across the 1990s and 2000s, generating the petitions behind the continued growth recorded in the 2010 and 2018 readings. Each wave of the census curve is therefore the echo of an earlier wave, separated by the naturalization lag, and the curve keeps rising long after the rules stabilize because the earlier cohorts are still working through the pipeline. The stability of the rules across the 2000s and 2010s did not freeze the demography. It let the lagged compounding run its course.

The 1986 legalization illustrates the lag at scale. The approximately 2.7 million people approved for permanent residence under the Immigration Reform and Control Act, about 1.6 million under section 245A and about 1.1 million as Special Agricultural Workers, entered the permanent resident stock in the late 1980s. Their petitioning power at that stage was limited to the second preference. As they naturalized across the 1990s, they acquired the full citizen petitioning power, including the unlimited immediate relative channel, and became a large new source of family petitions. The legalization was therefore not only a one-time addition to the legal stock. It was a delayed addition to the petitioner base, with its full demographic effect arriving years after the grants themselves. Any analysis that dates the legalization’s consequences to 1986 alone misses the second wave that the naturalization lag produced.

The age requirement on citizen petitioners adds a further lag dimension. A citizen child must be at least twenty-one years of age to petition for parents, which means the parent-petition channel opens a generation after the child’s own admission or birth. The sibling and married-children preferences similarly operate on family-formation timelines rather than administrative ones. The statute’s family logic is therefore paced by the life cycle: admission, naturalization, marriage, children reaching adulthood, each step adding years before the next round of petitions. The compounding is real, but it is slow compounding, and the slowness is why the demographic transformation took five decades rather than one.

This is the final piece of the mechanism account, and it completes the explanation of the census curve’s shape. The small initial base explains the slow start. The branching logic explains the acceleration. The naturalization lag explains the persistence: why the growth continued through decades when no new statute was enacted, and why the foreign born share was still rising in the 2018 ACS, half a century after the caps took effect. The provisions installed in 1965 set all three dynamics in motion, and they operated on the timetable the statute’s own eligibility rules dictated.

Why did the 1976 amendment matter so much for the Western Hemisphere?

The 1976 amendments extended the 20,000 per-country limit to the Western Hemisphere for the first time, which imposed a binding cap on Mexico, the one Western Hemisphere country whose lawful admissions substantially exceeded it. This made 1976 the hinge year for the unauthorized migration argument.

Public Law 94-571, the Immigration and Nationality Act Amendments of 1976, enacted October 20, 1976, extended the Eastern Hemisphere preference system and the 20,000 per-country annual limit to the Western Hemisphere, keeping the hemispheric ceilings of 170,000 and 120,000 in place. The initial implementation carried exceptions of 35,000 each for Canada and Mexico, but the structural change was the per-country cap itself: for the first time, a migrant from Mexico faced the same 20,000 annual country limit as a migrant from Italy or the Philippines. The 1976 change mainly affected Mexico, because Mexico was the only Western Hemisphere country whose admissions substantially exceeded 20,000 visas annually. For the rest of the hemisphere, the new per-country limit was not binding, and the change was administrative rather than demographic. For Mexico, it was a sharp restriction imposed on a flow that had been running for decades, and the restriction arrived in combination with the end of the agricultural labor program to produce the consequence this article’s third layer examines.

The timing of the 1976 amendment within the longer sequence matters for understanding why the Massey argument treats it as the hinge. The 1965 act had imposed the first Western Hemisphere ceiling, 120,000 from the fiscal year beginning July 1, 1968, but it had left the hemisphere without per-country limits, so Mexican admissions could still run up to the hemispheric ceiling without a country-specific constraint. The 1976 amendment closed that gap by imposing the per-country limit, and in doing so it converted the Western Hemisphere system from a first-come, first-served hemispheric pool into a country-capped system like the Eastern Hemisphere’s. The 1978 amendments, Public Law 95-412, then replaced the two hemispheric ceilings with a single worldwide ceiling of 290,000, retaining the immediate relative exemption, and created the Select Commission on Immigration and Refugee Policy. The 1990 act, Public Law 101-649, raised the worldwide level to 700,000 for fiscal years 1992 through 1994 and 675,000 from fiscal year 1995, organized around family-sponsored, employment-based, and diversity tracks, and created the H-1B category and temporary protected status. Each of these successors reshaped the lawful admissions system, and their demographic effects are treated in the H3 that follows this section. But the 1976 per-country extension is the one the unauthorized migration literature identifies as the turning point for the Mexican flow, because it was the first provision to cap Mexican lawful admissions at a level far below the established demand.

The Bracero program and the circular flow it organized

The Massey causal chain begins before the 1965 act, with the labor migration system the act’s restrictions displaced. Understanding that prior system is necessary because the argument is not that the 1965 act created Mexican labor migration; it is that the act, together with the end of the agricultural labor program, changed the legal character and the settlement pattern of a flow that already existed. The flow predated the statute by decades, and its circularity was its defining feature.

The Bracero program was the bilateral agricultural labor program under which Mexican workers entered the United States on short-term contracts to perform seasonal farm labor. In its 22 years of operation, the program brought more than four million Mexican agricultural workers to the United States, making it the largest lawful channel for Mexican labor migration in American history to that point. The program’s authority derived from Public Law 78, and Congress allowed that authority to expire, ending the program on December 31, 1964. The termination was a legislative choice made on labor market and labor standards grounds, debated in its own terms and for its own reasons; this article does not relitigate that debate. What matters for the demographic story is the structural consequence: the closure of the largest lawful channel for seasonal Mexican labor at the end of 1964, months before the 1965 act was signed and more than three years before its Western Hemisphere ceiling took effect.

The circularity of the pre-1965 flow deserves a precise description, because the Massey argument turns on what circularity meant in practice. A circular labor flow is one in which workers cross for employment, work for a season or a period, and return to their place of origin, repeating the cycle as labor demand recurs. The Bracero contracts institutionalized this pattern: admission was tied to agricultural employment, the stay was temporary by design, and return was the expected conclusion of each contract. Beyond the program itself, a wider pattern of seasonal movement operated in the same circular fashion, with workers entering, working, and returning in rhythm with agricultural seasons and with the wage gap between the two countries supplying the persistent incentive. The flow was, in the Massey literature’s characterization of the period, largely lawful: the dominant institutional channel was a legal guest worker program, and the movement it organized was temporary and recurrent rather than permanent and one-directional.

Two structural facts about the pre-1965 legal environment complete the picture. First, Western Hemisphere migration faced no numerical limit at all before the 1965 act. A Mexican worker seeking lawful permanent admission was not subject to any quota, which meant that the legal immigration system, for all its restrictiveness toward the Eastern Hemisphere, placed no numerical barrier on the hemisphere that supplied the circular labor flow. Second, the demand for that labor was structural rather than cyclical: American agriculture in the Southwest and elsewhere had organized its production around seasonal Mexican labor over decades, and the wage gap between the two countries ensured a steady supply of workers willing to cross for the season. The combination of open lawful channels, institutionalized circularity, and persistent demand produced a stable equilibrium: large numbers crossed, large numbers returned, and the settled unauthorized population remained small because the system gave workers little reason to stay permanently and every reason to repeat the cycle.

The significance of this equilibrium for the Massey argument is that it establishes the baseline against which the post-1965 changes are measured. The argument does not claim that Mexican labor migration began with the 1965 act or with the Bracero termination; it claims that a stable circular system was dismantled in two steps, the program’s end in December 1964 and the imposition of numerical caps beginning in 1968 and tightening in 1976, and that the dismantling changed the flow’s legal character from largely lawful to largely unauthorized and, later, changed its settlement pattern from circular to permanent. A reader who grasps the pre-1965 equilibrium understands what was lost when the channels narrowed: not the migration itself, which persisted because the underlying demand persisted, but the circularity and the lawfulness that had kept the settled unauthorized population small. The next section works through the four-element chain by which that loss occurred.

The unauthorized migration consequence: the Massey causal argument

The single most under-covered element of the 1965 story is the causal argument developed by Douglas Massey and his coauthors about how the statute, in combination with the end of the Bracero program and later border enforcement, transformed a circular labor flow into a settled unauthorized population. The argument is serious, well-cited, and built on decades of survey and administrative data, and this article presents it at the length its importance warrants. The presentation follows the brief’s discipline: the mechanism only, never the original authors’ characterizing language, and no characterization of any national origin group.

The argument begins with the character of Mexican labor migration before the mid-1960s. For decades, a large flow of Mexican workers entered the United States under short-term arrangements, worked, and returned to Mexico, in a circular pattern that was largely lawful. The Bracero program, the bilateral agricultural labor program, was the institutional form of that circularity: in its 22 years of operation it brought more than four million Mexican agricultural workers to the United States under short-term contracts, and Congress ended the program on December 31, 1964, when its Public Law 78 authority expired. The Bracero program was not the whole of the circular flow, but it was the largest lawful channel for it, and its termination removed the principal legal mechanism through which seasonal Mexican labor entered the country. The timing is the first element of the causal chain: the lawful circular channel closed at the end of 1964, just as the 1965 act was being written.

The second element is the 1965 act’s Western Hemisphere ceiling. Before 1965, Western Hemisphere migration had faced no numerical limit at all. The 1965 act imposed the first cap, 120,000 per fiscal year from the fiscal year beginning July 1, 1968. The demand for Mexican labor migration, driven by the wage gap between the two countries and by employer demand in American agriculture and other sectors, did not fall when the cap was imposed; the lawful channel for satisfying it simply narrowed. The third element is the 1976 amendment’s extension of the 20,000 per-country limit to the Western Hemisphere, which imposed for the first time a binding country cap on Mexico, the one Western Hemisphere country whose admissions substantially exceeded it. The Massey literature identifies the 1976 amendment as the hinge of the argument: it was the provision that converted the Western Hemisphere system from a hemispheric pool, within which Mexican admissions could still be substantial, into a country-capped system that restricted Mexican lawful admissions to a level far below the established flow. The combination of the Bracero termination, the 1968 hemispheric ceiling, and the 1976 per-country cap is, in the Massey account, what converted a long-standing circular and largely lawful labor flow from Mexico into an unauthorized one: the demand persisted, the lawful channels narrowed, and the flow continued outside the lawful channels.

The fourth element of the argument concerns what happened after the flow became unauthorized. Beginning in the mid-1980s and accelerating through the 1990s, the United States expanded border enforcement substantially. The Massey, Durand, and Malone volume, “Beyond Smoke and Mirrors: Mexican Immigration in an Era of Economic Integration,” published by the Russell Sage Foundation in 2002, documents how the 1986 to 1996 enforcement expansion shifted Mexican migration from seasonal, circular movement toward permanent settlement. The precise statement of the enforcement-backfired mechanism appears in the 2016 article by Massey, Durand, and Karen A. Pren, “Why Border Enforcement Backfired,” published in the American Journal of Sociology, volume 121, number 5, in March 2016, pages 1557 to 1600. The mechanism is this: enforcement raised the cost and the risk of crossing the border, and higher crossing costs deterred return migration more than they deterred initial entry. A migrant who had crossed faced a choice between returning to Mexico and risking a costly and dangerous re-entry later, or remaining in the United States; as the cost of crossing rose, remaining became the rational choice, and the circular flow of workers became a settled population of families. The earlier comparative volume, “Worlds in Motion: Understanding International Migration at the End of the Millennium,” by Massey, Joaquin Arango, Graeme Hugo, Ali Kouaouci, Adela Pellegrino, and J. Edward Taylor, published by Clarendon Press in 1998, supplies the broader theoretical framework within which the Mexican case sits.

The argument’s empirical content is worth stating carefully, because it is an argument about rates and directions rather than about levels. The claim is not that enforcement had no effect on entry; the claim is that it reduced return migration more than it reduced entry, and that the net effect was to increase the settled unauthorized population. This is a claim about the difference between two elasticities, the responsiveness of entry to enforcement and the responsiveness of return to enforcement, and the Massey literature’s contribution is to have measured both and found the return elasticity larger. The finding is counterintuitive, which is why the article states it plainly: a policy designed to reduce the unauthorized population increased the settled portion of it, because it changed the behavior of people already inside the country more than it changed the behavior of people deciding whether to come. The 2016 article’s analysis draws on the Mexican Migration Project’s survey data, a long-running binational survey that tracks the migration histories of individuals and households, which is what allows the authors to observe return behavior directly rather than inferring it from border apprehension counts.

The 1986 Immigration Reform and Control Act belongs in this layer as the legislative response to the unauthorized population the earlier restrictions had helped create. Public Law 99-603, the Simpson-Mazzoli Act, signed by President Ronald Reagan on November 6, 1986, combined employer sanctions for knowingly hiring unauthorized workers with two legalization programs: the general legalization under INA section 245A, for persons with continuous residence since January 1, 1982, under which about 1.6 million people were ultimately granted lawful permanent residence, and the Special Agricultural Worker program, for persons with 90 days of seasonal agricultural work, under which about 1.1 million were granted permanent residence. The documented total is approximately 2.7 million approvals for permanent residence, the INS and Department of Labor derived figure that this series uses. The legalization moved a large population from unauthorized to lawful permanent status, which is why the 1990 census figure of 19.8 million foreign-born residents cannot be read as a pure measure of new arrivals: it includes the formerly unauthorized population that IRCA had legalized. The fuller legislative history of the 1986 act is treated in this series’ article on the Immigration Reform and Control Act, and this article uses the 2.7 million figure only as the documented scale of the legalization that complicates the 1990 census reading.

The Massey argument, taken as a whole, is the third layer of this article’s evidence table, and it carries the rating the brief assigns: contested, in the sense that it is a causal argument advanced in the scholarly literature and debated there, not a census fact like the foreign-born share. The article presents it as a serious, well-cited causal argument because that is what the publication record shows: a 1998 comparative volume from a major academic press, a 2002 monograph from the Russell Sage Foundation documenting the 1986 to 1996 policy history, and a 2016 article in one of sociology’s leading journals stating the enforcement mechanism with survey data. A reader who finishes this section should be able to state the four-element chain, Bracero termination at the end of 1964, the 1968 Western Hemisphere ceiling, the 1976 per-country extension as the hinge, and the enforcement asymmetry between entry and return, and should understand that the chain is an argument about the statute’s unintended demographic consequences, distinct from the intended operation of the family preference mechanism in the second layer.

What did the 1978 and 1990 amendments change about admissions?

The 1978 amendments replaced the two hemispheric ceilings with a single worldwide ceiling of 290,000 while keeping the immediate relative exemption, and the 1990 act raised the worldwide level to 675,000, organized admissions into family, employment, and diversity tracks, and created the H-1B category.

Public Law 95-412, enacted October 5, 1978, established a single worldwide annual immigration ceiling of 290,000, replacing the separate Eastern Hemisphere ceiling of 170,000 and Western Hemisphere ceiling of 120,000. The arithmetic of the consolidation is exact: 170,000 plus 120,000 equals 290,000, so the 1978 act did not change the total nominal level; it changed the structure, eliminating the hemispheric distinction that had organized admissions since 1968. The immediate relative exemption was retained, so the uncapped category continued to grow outside the ceiling. The 1978 act also created the Select Commission on Immigration and Refugee Policy, the body whose deliberations shaped the next decade’s legislation. Demographically, the 1978 consolidation mattered less for the totals than for the flexibility it gave the system: with a single worldwide ceiling, the allocation of visas across regions was governed by the preference system and the per-country limits rather than by the hemispheric division, which allowed the compounding family mechanism to operate across the whole system without hemispheric boundaries.

Public Law 101-649, the Immigration Act of 1990, enacted November 29, 1990 and signed by President George H. W. Bush, was the largest statutory expansion of lawful admissions since 1965. It raised the worldwide level to 700,000 for fiscal years 1992 through 1994 and 675,000 from fiscal year 1995, and it organized admissions around three tracks: family-sponsored admissions at 480,000, employment-based admissions at 140,000, and diversity admissions at 55,000. The act created five employment-based preference categories, the diversity visa lottery, H-1B visas for highly skilled temporary workers, and temporary protected status. It eliminated the exclusion ground for sexual deviancy and created naturalization testing accommodations. The demographic significance of the 1990 act was twofold. First, it substantially increased the employment-based track, which meant that a larger share of admissions after the early 1990s was selected on labor market criteria rather than kinship, a compositional shift that bears on the economic findings in this article’s fourth layer. Second, the diversity lottery created a new channel for admissions from countries with low recent migration to the United States, which modestly diversified the origin composition beyond the family preference compounding. The 1990 act did not alter the immediate relative exemption or the compounding logic of the family preferences; it added new capped channels alongside the uncapped one, and the uncapped one continued to grow.

The 1990 Act’s demographic footprint: new channels alongside the old engine

The H3 above stated what the Immigration Act of 1990 changed in the statute’s numbers and structure. This section examines what those changes meant demographically, because the 1990 act was the first legislation since 1965 to alter the composition of the flow in a deliberate way, and its effects show up in the origin and skill composition figures of the 2000s and 2010s. The verified facts are the starting point: Public Law 101-649, enacted November 29, 1990, raised the worldwide level to 700,000 for fiscal years 1992 through 1994 and 675,000 from fiscal year 1995, organized around 480,000 family-sponsored, 140,000 employment-based, and 55,000 diversity admissions, created five employment-based preference categories, the diversity visa lottery, H-1B visas for highly skilled temporary workers, and temporary protected status, eliminated the sexual deviant exclusion ground, and created naturalization testing accommodations.

The most demographically significant change was the expansion of the employment-based track to 140,000. Under the 1965 system, the employment preferences had been the third and sixth preferences within the 170,000 Eastern Hemisphere ceiling, a modest channel compared with the family categories. The 1990 act’s 140,000 employment-based allocation, organized into five preference categories covering priority workers, advanced-degree professionals, skilled workers, special immigrants, and employment-creation investors, created for the first time a large-scale skill-selected channel operating alongside the kinship-selected channels. The compositional consequence was that a growing share of admissions after the early 1990s was selected on labor market criteria, education, and employer sponsorship rather than on family relationship. This matters for reading the economic findings in this article’s fourth layer: the National Academies consensus findings about wage and fiscal effects describe the post-1965 migrant population as a whole, which was dominated by the family channels, and a more skill-selected flow would be expected to produce different economic effects. The 1990 act thus began a slow compositional shift within the flow even as the family engine continued to dominate the totals.

The diversity visa lottery was the 1990 act’s most novel demographic instrument. By allocating 55,000 visas annually to natives of countries with low recent migration to the United States, the lottery created an admissions channel explicitly designed to diversify the origin composition beyond what the family preference compounding produced. The lottery’s demographic effect was modest in absolute numbers compared with the family channels, but it was structurally significant: it was the first provision since 1965 to allocate visas on the basis of origin-country underrepresentation rather than kinship or employment, and it opened a lawful path for migrants from regions, particularly Africa and parts of Europe, that the family compounding had largely bypassed. The lottery did not reverse the Latin American and Asian majorities documented in the 2010 census figures, but it widened the origin distribution at the margins and introduced a new principle into the allocation system.

The H-1B category and temporary protected status belong to a different demographic category: temporary rather than permanent admission. The H-1B visa for highly skilled temporary workers created a large-scale channel for employer-sponsored skilled migration that did not lead directly to permanent residence, though many H-1B holders later adjusted status through the employment-based preferences. Temporary protected status created a humanitarian channel for nationals of designated countries experiencing armed conflict or natural disaster. Neither channel appears directly in the foreign-born stock figures that anchor this article’s first two layers in the same way that permanent admissions do, but both contributed to the flows that fed those stocks, and the H-1B channel in particular became a major pathway for skilled migration from Asia in the 1990s and 2000s, reinforcing the Asian share gains visible in the 2010 and 2018 composition figures.

The provisions the 1990 act did not change are as important as the ones it did. The immediate relative exemption survived intact, and the family preference compounding continued to operate exactly as before, which meant that the demographic engine of the post-1965 period kept running alongside the new channels. The 1990 act added to the system; it did not replace the engine. A reader comparing the 2000 composition figures, Latin America 51.7 percent, Asia 26.4 percent, Europe 15.8 percent, with the 2010 figures, Latin America 53 percent, Asia 28 percent, Europe 12 percent, can see both dynamics at work: the continued dominance of the family-channeled flows from Latin America and Asia, and the incremental diversification and skill-selection effects of the 1990 act’s new channels. The statute that produced the post-1965 transformation was thus not a single act but a layered system, with the 1965 provisions supplying the engine and the 1990 provisions supplying the modifications.

The contested economics: wages, growth, and public budgets

The demographic layers of this article report facts that are not in dispute. The economic and fiscal effects of the post-1965 migration are genuinely contested, and this article treats them with the discipline the brief requires: the consensus report as the anchor, the named dissenting researchers beside it, and the disputed questions labeled disputed. The anchor is “The Economic and Fiscal Consequences of Immigration,” the consensus report of the National Academies of Sciences, Engineering, and Medicine, edited by Francine D. Blau and Christopher Mackie, published by the National Academies Press in 2017, with an advance copy released in September 2016. The report runs 508 pages, synthesizes the economics literature through the mid-2010s, and is the closest thing the field has to a settled statement of what is known. Its findings are the starting point, not the last word, and this section works through them layer by layer before presenting the dissent.

On wages, the report’s central finding is stated in the panel’s own summary language: when measured over a period of ten years or more, the impact of migration on the wages of native-born workers overall is very small. The qualification about the measurement period matters, because short-run effects can differ from long-run effects as labor markets adjust, and the report’s finding is explicitly a long-run finding. Where negative wage effects are detected, the report finds that two groups are most likely to be affected: prior migrants, who are the closest substitutes for new migrants in the labor market, and native-born workers without a high school credential, who compete most directly with new arrivals in low-skill labor markets. The logic of substitutability is the key to the finding: new arrivals compete most directly with workers who do similar work, and those workers are disproportionately earlier arrivals and less-educated natives, not the native workforce as a whole. The report’s wage finding is thus a differentiated finding, not a blanket one: very small effects overall, with the detectable negative effects concentrated in specific segments of the labor market.

On economic growth, the report finds that migration has an overall positive impact on long-run economic growth, increasing both GDP and the GDP growth rate. The mechanism is straightforward: migrants add to the labor force and to the population, and a larger labor force produces more output. The report also notes the role of innovation and entrepreneurship, though the growth finding does not depend on any claim about exceptional migrant productivity; it follows from the arithmetic of a larger workforce. On the fiscal effects, the report draws the distinction that structures the entire fiscal debate: the first generation of migrants is more costly to governments than the native-born population, but the second generation, the children of migrants, are among the strongest fiscal and economic contributors in the United States. The long-term fiscal effects are generally positive at the federal level but remain negative at the state and local level, driven especially by the costs of educating the children of migrants. The federal versus state and local split is the finding most often misreported, because advocates of each side quote one half of it. The report’s actual finding is the split itself: the federal budget benefits in the long run, while state and local budgets bear net costs, principally because public schooling is financed at the state and local level and the children of migrants attend public schools.

The report’s method deserves a paragraph, because the authority of a consensus report rests on how it handles disagreement. The panel reviewed the published literature, commissioned new analyses, and reported where the literature converged and where it did not. On the overall wage effect, the literature converged on very small long-run effects. On the distribution of those effects across skill groups, the literature was less settled, which is why the report phrases the negative findings for dropouts and prior migrants as likelihoods rather than certainties. On the fiscal effects, the report built generational accounting models that track the taxes paid and the benefits received by migrants and their descendants over lifetimes, and the federal positive versus state and local negative finding is the output of those models under the panel’s stated assumptions. A reader who understands that the report is a synthesis with stated assumptions, not a measurement with a single answer, is equipped to read both the report and its critics fairly.

The critics and the dissenting researchers belong in this section by name, because the brief requires the consensus report plus named dissent rather than the consensus alone. The most prominent dissenting voice on the wage question is George Borjas of Harvard, whose work on the Mariel boatlift is treated in the H3 that follows, and whose broader research program has consistently found larger negative wage effects for low-skill native workers than the consensus reports. The most prominent voice on the other side is David Card of Berkeley, whose 1990 Mariel study found no discernible wage or employment effects and whose subsequent work has defended the small-effects finding. The Borjas versus Card dispute is the subject of its own section because it is the sharpest and best-documented disagreement in the literature, but it is not the only one: the fiscal findings have drawn criticism from researchers who argue that the generational accounting models understate long-run costs or overstate the fiscal contributions of the second generation, and the growth findings have drawn the standard objection that GDP growth driven by population growth does not imply growth in GDP per capita. This article does not adjudicate these disputes. It names them, states what each side’s best evidence shows, and leaves the reader with the map rather than the verdict, which is what the series thesis thread requires: assessing the statute against what it actually did, with the disputed questions labeled disputed.

The second generation and the long-run fiscal picture

The National Academies consensus report’s most consequential fiscal finding is the generational one, and it warrants a section of its own because it is the finding that determines whether the post-1965 migration is judged a fiscal gain or a fiscal cost. The report’s formulation is precise: the first generation of migrants is more costly to governments than the native-born population, but the second generation, the children of migrants, are among the strongest fiscal and economic contributors in the United States. Everything in the fiscal debate flows from how those two halves are weighted, over what time horizon, and at which level of government.

The first-generation cost finding rests on the composition of the post-1965 flow that this article’s mechanism layer described. Because the family preference system selected on kinship rather than on earnings capacity, the first generation included large numbers of migrants with modest formal education and correspondingly modest earnings, which meant lower tax payments, combined with eligibility for public services that did not depend on earnings. The report’s generational accounting models, which track taxes paid and benefits received over lifetimes, capture this as a net first-generation cost relative to the native-born population. The finding is not a claim about individual deservingness or about any group’s characteristics; it is an arithmetic result of the interaction between the skill composition the statute produced and the structure of American taxes and benefits. A more skill-selected flow, of the kind the 1990 act’s expanded employment track began to produce, would change the first-generation arithmetic, which is why the composition of the flow and the fiscal findings cannot be read separately.

The second-generation finding runs in the opposite direction and is the stronger of the two. The children of migrants, raised and educated in the United States, attain education and earnings levels that converge toward and in many cases match the native-born population, while their parents’ lower earnings mean the second generation starts from a lower fiscal baseline. The report’s finding that the second generation ranks among the strongest fiscal and economic contributors reflects this convergence: lifetime tax payments rise sharply between the first and second generations, while the costs associated with the first generation’s lower earnings do not repeat. The mechanism behind the finding is educational attainment, and the policy implication the report draws is correspondingly focused on schooling: the costs of educating the children of migrants are the principal driver of the state and local fiscal burden, and they are also the investment that produces the second generation’s fiscal contributions.

The federal versus state and local split is where the generational finding meets the structure of American public finance. The long-term fiscal effects are generally positive at the federal level because the federal government collects the income and payroll taxes on the second generation’s higher earnings while bearing a smaller share of education costs. The effects remain negative at the state and local level because states and localities finance public schooling, the single largest cost associated with the second generation’s upbringing, while collecting a smaller share of the resulting tax revenue. The split is thus not a disagreement among researchers but a structural feature of fiscal federalism interacting with the generational pattern: the level of government that pays for the investment is not the level that collects most of the return. Any honest summary of the fiscal effects of post-1965 migration must state both halves together, and any summary that quotes only the federal surplus or only the state and local deficit is misreporting the report.

The disputes around these findings are genuine and belong in the article by name. Critics of the generational accounting approach argue about the discount rates applied to future tax payments, about the treatment of public goods whose costs do not rise proportionally with population, about whether the models adequately capture the fiscal effects of the unauthorized population, and about the sensitivity of the long-run results to assumptions about future earnings convergence. Defenders respond that the panel’s assumptions were stated transparently, that the qualitative findings are robust across a range of assumptions, and that the federal positive versus state and local negative pattern appears under every reasonable specification. This article does not adjudicate the dispute, consistent with the series thesis thread. It reports the consensus findings, names the lines of criticism, and leaves the reader with the structure of the argument: first generation costly, second generation strongly contributory, federal positive, state and local negative, with the magnitude of each element depending on assumptions the report states and the critics contest.

The National Academies panel: how to read a consensus report

The 2017 consensus report anchors this article’s entire economics layer, and a reader who understands what kind of document it is will read its findings more accurately than one who treats it as a single study. “The Economic and Fiscal Consequences of Immigration” is the product of the Panel on the Economic and Fiscal Consequences of Immigration, convened under the Committee on National Statistics in the Division of Behavioral and Social Sciences and Education of the National Academies of Sciences, Engineering, and Medicine, edited by Francine D. Blau and Christopher Mackie, and published by the National Academies Press in 2017 as a 508-page volume, with an advance copy released September 22, 2016. The institutional form matters: a National Academies consensus report is not one research team’s paper but a panel’s synthesis of the field, produced through deliberation among appointed experts and reviewed before publication. Its authority comes from the breadth of literature it surveys and the explicitness with which it states where that literature converges and where it does not.

The report’s wage finding illustrates how to read convergence. The panel’s summary states that when measured over ten years or more, the impact of migration on the wages of native-born workers overall is very small, and that where negative effects are detected, prior migrants and native-born high-school dropouts are the most likely to be affected. The phrasing is careful in two ways a reader should notice. First, the ten-year measurement period is part of the finding, not a footnote to it: the report is making a long-run claim and declining to make the same claim about short-run adjustments. Second, the distributional qualification is stated as a likelihood rather than a certainty, which signals that the literature on which workers bear the negative effects is less settled than the literature on the overall effect. A consensus report communicates uncertainty through phrasing, and the reader who attends to the phrasing gets the report’s actual position rather than a simplified version of it.

The fiscal findings illustrate the report’s use of modeling with stated assumptions. The generational accounting that produces the first-generation-costly, second-generation-contributory, federal-positive, state-and-local-negative pattern is a model, not a measurement: it projects taxes and benefits over lifetimes under assumptions about earnings trajectories, discount rates, and the allocation of public goods costs. The report states those assumptions, which is what makes the findings criticizable and what makes the criticism legitimate rather than dismissive. The disputes this article names, over discount rates, public goods treatment, and earnings convergence assumptions, are disputes about the model’s inputs, and they are the normal scientific accompaniment to a modeling exercise of this kind. The panel’s contribution is not to end those disputes but to show what the best available literature implies under transparent assumptions, which is the most a consensus process can honestly deliver.

The report’s handling of the Borjas versus Card dispute is the clearest example of the consensus method at work. The panel reviewed both the 1990 Card study and the 2017 Borjas reappraisal, along with the surrounding literature, and reported very small overall wage effects with possible negative effects concentrated on the least-skilled, a formulation that accommodates both papers without choosing between them. A reader looking for the report to declare a winner will be disappointed; a reader looking for the report to map the disagreement will find exactly that. This is why the article treats the consensus report as the anchor and the named dissenters as essential companions: the anchor tells the reader where the field converges, and the dissenters tell the reader where it does not, and both are needed for the honest summary the series thesis thread requires.

Why do Borjas and Card disagree about the Mariel boatlift?

David Card’s 1990 study of the Mariel boatlift found no discernible effect on Miami wages or employment, while George Borjas’s 2017 reappraisal found a sharp wage decline for low-skill men; they used the same event but different worker subgroups and methods, and the dispute remains unresolved.

The Mariel boatlift is the most studied natural experiment in the economics of migration, and the reason both sides of the wage dispute claim it is that it offers something rare: a sudden, large, and geographically concentrated increase in labor supply that was not caused by local labor market conditions. In April 1980, Fidel Castro declared that Cubans wishing to leave could do so through the port of Mariel, and the arrivals beginning April 23, 1980 increased the city’s workforce by about 7 percent in a matter of months. The suddenness of the inflow meant that researchers could compare Miami’s labor market before and after the boatlift against the labor markets of comparable cities that did not receive the inflow, a research design that approximates an experiment. David Card’s 1990 article, “The Impact of the Mariel Boatlift on the Miami Labor Market,” published in the Industrial and Labor Relations Review, volume 43, number 2, in January 1990, pages 245 to 257, executed that design and found no discernible effect on the wages and employment of Miami workers, including low-skill workers. The finding was striking because the textbook prediction was that a 7 percent increase in labor supply should depress wages, and Card’s result suggested that labor markets absorbed the inflow through channels the textbook model missed: out-migration of competing workers, adjustments in industry mix, and complementarities between new arrivals and existing workers.

George Borjas’s 2017 reappraisal, “The Wage Impact of the Marielitos: A Reappraisal,” published in the same journal, the ILR Review, volume 70, number 5, in October 2017, pages 1077 to 1110, with a working paper version circulating as NBER Working Paper 21588 in 2015, reanalyzed the same event and reached a different conclusion. Borjas focused on a narrower subgroup, low-skill men defined as high-school dropouts, and within that subgroup on non-Hispanic men, on the reasoning that the Mariel arrivals competed most directly with the least-skilled workers in the Miami labor market and that the Hispanic sample in the survey data was contaminated by the arrival of the Mariel migrants themselves, who were counted as Hispanic residents of Miami in the post-boatlift surveys. With that subgroup definition, Borjas found a sharp decline in wages that took about a decade to recover. The two papers thus disagree not about the event but about whom to measure: Card measured the Miami labor market broadly and found no effect, while Borjas measured the narrowest skill subgroup and found a large effect.

The methodological dispute between the two positions is genuine and unresolved, and this article presents it as such rather than settling it. Card’s defenders argue that Borjas’s subgroup is too small for reliable measurement in the survey data, that the wage decline Borjas finds reflects sampling variation and the changing composition of the subgroup rather than a true wage effect, and that subsequent replications with larger samples recover Card’s original finding. Borjas’s defenders argue that Card’s broad measurement diluted the effect by averaging over workers who did not compete with the arrivals, that the subgroup definition follows directly from the economic logic of substitutability, and that the decade-long recovery pattern is exactly what a supply shock should produce. Both papers appeared in the same journal 27 years apart, which gives the dispute a clean bibliographic form: ILR Review 43(2) in 1990 for Card, ILR Review 70(5) in 2017 for Borjas. The National Academies consensus report reviewed both and reported very small overall wage effects with possible negative effects concentrated on the least-skilled, a formulation that accommodates both findings without choosing between them. A reader who finishes this section should be able to state each paper’s result, explain why the subgroup definition is the hinge of the disagreement, and understand why the dispute is methodological rather than ideological: it is a disagreement about how to measure, not about whether migration is good or bad.

Reading the Mariel literature: the research design beneath the dispute

The Borjas versus Card dispute is argued in the language of subgroups and samples, but underneath that language is a disagreement about research design, and a reader who understands the design understands why the dispute has proven so durable. The Mariel boatlift offered economists a natural experiment: a sudden, geographically concentrated labor supply shock that was not caused by conditions in the receiving labor market. The Cuban government’s April 1980 decision to permit departure through Mariel was a political act, not a response to Miami’s wages, which meant that the timing and the destination of the inflow were independent of the outcome the researchers wanted to measure. That independence is what makes the episode valuable: in ordinary circumstances, migrants choose destinations with strong labor markets, which makes it impossible to tell whether wages are high because migrants arrived or migrants arrived because wages were high. The boatlift broke that simultaneity by sending a large inflow to one city for reasons unrelated to that city’s economy.

Card’s 1990 design exploited the experiment by comparing Miami’s labor market outcomes before and after the boatlift with the outcomes in comparable cities that did not receive the inflow. The logic is that of a controlled comparison: if Miami’s wages behaved differently from the comparison cities’ wages after April 1980, the difference could be attributed to the boatlift, since the political decision that caused the inflow had nothing to do with Miami’s economy. Card’s finding of no discernible effect on wages and employment, including for low-skill workers, was therefore a finding that the treated city’s labor market did not diverge from its controls. The result’s influence came from the cleanliness of the design as much as from the finding itself: it was the closest thing migration economics had to an experiment, and the experiment seemed to show that labor markets absorb supply shocks without the wage declines the textbook model predicts.

Borjas’s 2017 reappraisal accepted the experimental setup but disputed the measurement. His argument was that the relevant labor market for the shock was not Miami’s workforce as a whole but the narrow segment of workers who competed directly with the arrivals: men without a high school credential, and within that group, non-Hispanic men, because the survey data counted the Mariel arrivals themselves as Hispanic Miami residents in the post-boatlift years, contaminating the Hispanic sample. On that subgroup, Borjas found a sharp wage decline that took about a decade to recover. The reappraisal thus kept Card’s experiment and changed the outcome variable, and the dispute became a question of which outcome variable the experiment was designed to measure. Card’s defenders respond that the subgroup is too small in the survey data for reliable inference, that the measured decline reflects sampling variation and compositional change rather than a true wage effect, and that the working paper’s 2015 circulation as NBER Working Paper 21588 drew replications that recovered Card’s finding with larger samples. Borjas’s defenders respond that averaging over non-competing workers dilutes the effect the theory predicts, that the subgroup follows from the substitutability logic the National Academies report itself endorses, and that the decade-long recovery matches the dynamics of a supply shock working through a labor market.

The reason the dispute remains unresolved is that both positions have a valid methodological point and neither can fully defeat the other on the available data. The survey samples for a narrowly defined subgroup in a single city are thin enough that reasonable researchers can disagree about what the numbers show, and the theoretical argument for the narrow subgroup is strong enough that dismissing it requires more than pointing to the thinness of the data. The National Academies panel’s formulation, very small overall effects with possible negative effects concentrated on the least-skilled, is best read as the field’s acknowledgment of exactly this impasse: the overall finding is secure, the distributional finding is not, and the Mariel literature is the reason. A reader who finishes this section should understand that the dispute is not a proxy war between political positions but a genuine scientific disagreement about measurement, and that treating it as unresolved is not evasion but accuracy.

Inside the Earnings Dispute: What the Specifications Decide

Layer four presented the Borjas-Card dispute as unresolved. This section goes inside the disagreement to show what “unresolved” means operationally, because a reader who understands why the specifications decide the answer will never again mistake a single estimate for a settled fact.

Both papers study the same event, the 1980 Mariel boatlift’s sudden addition of roughly seven percent to the Miami labor force, and both draw on the same underlying survey data, the Current Population Survey. The divergence begins with the choice of whom to study. Card compared broad groups: the Miami labor market as a whole, and within it low-skilled workers defined broadly, against the same groups in a set of comparison cities that did not receive the shock. Borjas narrowed the lens to the workers most directly substitutable for the Mariel arrivals: men without a high school credential, further restricted to non-Hispanic men to hold the ethnic composition of the sample constant across the years. The two choices embody a genuine tradeoff rather than a mistake on either side. Broader groups yield larger samples and more stable estimates, but they dilute any concentrated effect by averaging the most affected workers together with workers the shock never touched. Narrower groups isolate the affected workers, but the survey samples for a narrowly defined subgroup in a single city in a single year are small, and small samples produce noisy estimates that move with the specification.

The comparison cities are the second specification choice that matters. A natural experiment needs a control group, a set of labor markets that resembled Miami but did not receive the boatlift, so that the difference between Miami and the controls can be attributed to the shock. But no two cities are identical, and the choice of which cities count as comparable affects the estimated difference. Cities with different industrial structures, different prior immigration histories, or different business cycles will diverge from Miami for reasons unrelated to the boatlift, and the researcher’s choice of controls determines how much of the divergence the estimate absorbs. The subsequent literature has replayed the event with different control groups and produced different answers, which is evidence about the method’s sensitivity rather than about any researcher’s good faith.

The third choice is the treatment of time. Card’s analysis focused on the years immediately around the boatlift, asking whether the shock depressed wages and employment in its aftermath. Borjas traced the subgroup’s earnings over a longer horizon and found a decline that took about a decade to fade. The horizon matters because labor markets adjust: workers move, industries adapt, and the initial shock dissipates through channels the short-run analysis misses. A finding of no short-run average effect and a finding of a decade-long subgroup effect can both be true of the same event, because they measure different things over different periods. The dispute looks like a contradiction only if the two papers are read as answering the same question, and the careful reading shows they were not.

The state of the question, then, is this: the profession knows that the overall effect of immigration on native-born earnings is very small over long horizons, because the 2017 consensus report synthesized the literature and said so; it does not know the precise magnitude or duration of the concentrated effect on the least-skilled workers, because the hardest case study in the literature supports different answers from defensible specifications. “Unresolved” does not mean that nothing is known. It means that the known findings have a confidence interval the profession has not yet narrowed, and that narrowing it requires better data or better methods rather than louder advocacy. The reader who can state the tradeoff between sample size and subgroup precision, the role of the comparison cities, and the difference the time horizon makes has understood the dispute at the level the evidence supports, which is the level this article aims to produce.

What resolution would look like is worth stating, because it clarifies what the dispute is not. It is not a disagreement about values or about immigration policy. It is a disagreement about statistical identification under data constraints, and it would narrow with larger samples for narrowly defined subgroups, with longer and more consistent time series, and with specifications chosen before the results are known rather than after. Until those conditions are met, the honest presentation is the one this article gives: the consensus finding on the average, the named disagreement on the subgroup, and the methodological reasons the disagreement persists. A reader who encounters a confident claim about the Mariel wage effects in either direction should ask which specification produced it, what the comparison cities were, and how the subgroup was defined, because those three choices contain the entire disagreement.

What the demographic record cannot measure

An evidence article owes its readers an account of its evidence’s limits as well as its evidence’s contents, and the census record that anchors this article’s first two layers has limits worth stating plainly. None of these limits undermines the undisputed facts the article reports; they define the boundary between what the figures show and what they do not show, which is exactly the boundary the article’s discipline requires.

The first limit is the distinction between stocks and flows. The census and the American Community Survey measure the resident foreign-born population at a point in time, which is a stock: the accumulated result of past admissions, departures, deaths, and status changes. They do not directly measure the annual flow of new admissions, which is recorded in administrative data from the immigration system. The foreign-born share rising from 4.7 percent in 1970 to 13.7 percent in 2018 tells us about the accumulated stock, not about how many people were admitted in any given year, and the origin composition figures describe the resident population’s birthplaces, not the composition of each year’s admissions. The mechanism layer of this article connects the statute to the stock through the compounding logic, but a reader should not mistake the stock figures for admission figures. The 11.7 million Mexican-born residents counted in 2010, for example, are the accumulated result of decades of admissions, legalizations, and settlement, not the number admitted in any single period.

The second limit concerns the unauthorized population. The census aims to count all residents regardless of legal status, but unauthorized residents are harder to reach and harder to count than lawful residents, and the degree of undercount is itself a subject of demographic estimation rather than direct measurement. This article follows its neutrality rule here strictly: it never estimates any current population figure for the unauthorized population, and it presents the Massey argument about the growth of the settled unauthorized population as a causal argument about rates and directions, not as a headcount. The IRCA legalization figure of approximately 2.7 million is an administrative count of approvals, not an estimate of a resident population, and it is the only unauthorized-adjacent number this article reports. A reader who wants unauthorized population estimates must look to the demographic estimation literature, which this article does not summarize because its brief confines the unauthorized discussion to the Massey mechanism.

The third limit is the source shift this article has already flagged: the pre-2000 figures come from the decennial census long-form data and the post-2000 figures from the American Community Survey, which are different instruments with different sampling designs. The article pairs the series the way the Census Bureau’s own long-term trend table pairs them, and the Pew tabulations carry the series forward on a consistent basis, but comparisons across the shift should not be treated as precise to the decimal point. The direction and the scale of the post-1965 transformation are robust to the instrument change; the exact percentage point in any given year is not the level at which the evidence operates.

The fourth limit is compositional: the origin figures describe birthplaces, not the legal channels through which people were admitted. The 53 percent Latin American and 28 percent Asian shares in the 2010 ACS describe where the foreign-born population was born, not whether they entered through family preferences, employment preferences, the diversity lottery, refugee channels, or unauthorized entry followed by legalization. The article’s mechanism layer argues that the family channels dominated the flow, and the IRCA section notes the legalization’s contribution to the 1990 stock, but the composition percentages themselves do not decompose by admission channel. A reader should hold the composition facts and the mechanism argument as two distinct kinds of evidence that support each other, not as a single measurement.

Stating these limits is not a concession to either overreach story. The intentional-design story is not rescued by measurement limits, because the mechanism-not-intent finding does not depend on precise headcounts. The minor-amendment story is not rescued either, because the scale of the stock transformation is robust to every limit stated here. The limits define the article’s precision, not its conclusions, and a reader who understands them can distinguish the article’s undisputed facts from the places where the evidence runs out, which is the same discipline the article applies to the contested economics.

Attributing Causation: Stocks, Flows, and Counterfactuals

With the four layers stated, the article can address the methodological question that underlies all of them: how does a researcher attribute demographic change to a statute when the statute operated across decades alongside economic cycles, sending-country conditions, later statutes, and enforcement regimes? The answer is a disciplined sequence of moves, and this section makes the sequence explicit because it is the method the article has been modeling throughout.

The first move is to distinguish stocks from flows. The census measures the stock of the foreign born population at a point in time. The statute governs the flow of admissions across time. The stock at any census date is the accumulated result of decades of flows minus mortality, emigration, and legalization episodes, plus the unauthorized population the census captures. No single year’s admissions explain any census reading, and no census reading directly measures any statute’s output. The article’s mechanism section connected provisions to flows and flows to stocks, but the connection is inferential, and the inference has to be built provision by provision rather than asserted from the correlation of the signing date with the census trend.

The second move is to specify the counterfactual. To say the 1965 Act caused a demographic change is to say the change would not have occurred, or would have occurred differently, under the prior law. The prior law was the national origins quota system, which allocated the overwhelming share of numbers to northern and western Europe by a formula derived from the 1920 census. Under that system, the compositional shift toward Latin America and Asia could not have occurred through the legal admission channels, because the formula reserved the numbers for Europe. The counterfactual is therefore unusually clean on the compositional question: whatever else changed in the world, the quota formula would have continued to produce a predominantly European inflow. On the question of the total numbers, the counterfactual is less clean, because economic demand, the 1990 expansion, and the unauthorized flows all contributed to the growth of the stock, and the article does not claim the 1965 Act alone produced the 44.8 million figure recorded in 2018.

The third move is to prefer the mechanism account over the intent account, which is the article’s namable claim restated as method. Intentions are inferred from statements, and statements are cheap, contradictory, and made for political purposes. Mechanisms are read from the text, and the text operated the same way regardless of who said what about it in 1965. The unlimited immediate relative category compounded whether or not anyone predicted the compounding. The per-country limit distributed the numbers whether or not anyone modeled the distribution. The Western Hemisphere cap narrowed the lawful channel whether or not anyone foresaw the unauthorized consequence. A causal account built from the provisions is therefore more robust than a causal account built from the floor speeches, because the provisions are the part of the history that actually did the work.

The fourth move is to hold the competing causes in view without letting them dissolve the attribution. The economic boom of the 1990s increased the demand for immigrant labor. Conditions in sending countries pushed emigration. The 1990 Act raised the ceilings. The 1986 legalization added 2.7 million people to the legal stock in a single episode. None of these negates the mechanism account. They complete it. The 1965 Act built the admission structure within which the later causes operated: the economic demand was expressed through the preference categories the act created, the 1990 expansion enlarged the ceilings the act had established, and the legalized population naturalized and petitioned through the family preferences the act had installed. Attributing causation to the statute does not require the statute to be the only cause. It requires the statute to be a necessary part of the causal structure, and on the evidence of the provisions and the census series, it was.

How do researchers separate the 1965 Act’s effects from everything else that happened?

They work from the statutory text outward: the provisions define the maximum lawful flow, the admission records show what the provisions produced, and the census stocks show the accumulated result. Competing causes, later statutes, economic cycles, and sending-country conditions, are measured against that baseline rather than assumed.

The separation is never perfect, and the honest researcher states the imperfection. The cleanest separation is on the compositional question, where the counterfactual is sharp: under the national origins formula, the inflow would have remained predominantly European, so the shift to Latin American and Asian majorities is attributable to the abolition of the formula and the preference system that replaced it. The separation is messier on the totals, where the 1990 expansion, the 1986 legalization, economic demand, and unauthorized flows all contributed alongside the 1965 mechanism. The article’s discipline is to claim only what the separation supports: the statute caused the structure of the transformation, its direction and its compositional logic, while the magnitude reflects the statute plus everything that operated within the structure it built. That is a weaker claim than either overreach wants, and it is the claim the evidence supports.

Two overreaches about the 1965 Act

The complication this article was commissioned to address comes in two directions, and both deserve a direct answer because both misuse the evidence the article has assembled. The first overreach treats every subsequent demographic change as the intended design of the act, reading the 2010 composition figures back into the motives of the 1965 Congress as though the legislators aimed at the outcome and achieved it. The second overreach treats the act as a minor technical amendment, a cleanup of an embarrassing quota formula whose demographic footprint is overstated by those who want a dramatic story. The statutory mechanism supports neither.

Against the intentional-design story, the evidence of this article offers the mechanism-not-intent finding stated in the second layer: the transformation followed from the unlimited immediate relative category and the compounding family preference structure, and explaining it by reference to anyone’s intentions, in either direction, is a weaker account than explaining it by reference to the text. The compounding mechanism is intention-independent: it operates regardless of what any legislator wanted, because it is the arithmetic of petitioning rights, not the psychology of the 1965 Congress. The small-base acceleration is likewise intention-independent: growth from Asia and Latin America started slowly and accelerated because the initial petitioner base was small, not because anyone planned a timetable. And the Massey unauthorized migration chain is a story of unintended consequences, in which restrictions imposed for control produced settlement through the enforcement asymmetry. None of these mechanisms requires a designer, and attributing the outcome to design adds nothing to the explanation while subtracting the precision the text provides. The myths and misreadings that surround the 1965 act are collected in this series’ article on immigration law myths, and the intentional-design version is the most persistent of them.

Against the minor-amendment story, the census record is the answer. A statute that presided over the foreign-born share rising from 4.7 percent at the 1970 census to 13.7 percent in the 2018 American Community Survey, and over the origin composition shifting from 75 percent European in 1960 to 81 percent Latin American and Asian in 2010, is not a minor technical amendment regardless of how modest its nominal ceilings looked. The minor-amendment story typically points to the 170,000 Eastern Hemisphere ceiling and the 120,000 Western Hemisphere ceiling as evidence that the act was restrained, and then treats the demographic transformation as the product of forces outside the statute. But the ceilings were never the whole statute, and the uncapped immediate relative category plus the compounding family preferences were the provisions that did the demographic work. To judge the act by its ceilings alone is to misread the text, and the misreading is precisely what the mechanism layer of this article corrects. The act was not minor; it was mismeasured by those who looked only at the numbers Congress capped and missed the numbers Congress left uncapped.

The series thesis thread runs through both answers: assess the statute against what it actually did, with the disputed questions labeled disputed. What the 1965 act actually did, on the undisputed demographic record, was to preside over a tripling of the foreign-born share and an inversion of the origin composition, through the specific provisions this article has named. What is disputed, on the economic record, is how those demographic facts translated into wage, growth, and fiscal effects, and this article has named the consensus and the dissent without adjudicating between them. A reader who holds both halves, the undisputed mechanism and the disputed economics, has the article’s full substance and is equipped to recognize both overreaches on sight.

The four-layer evidence table

Layer	Outcome measured	Period	Principal source	Rating
1	Foreign-born share of the total population	1970 to 2018	U.S. Census Bureau decennial censuses (1970 to 2000) and American Community Survey (2009 to 2018); Pew Research Center tabulations 1960 to 2018	Not disputed
2	Origin composition of the foreign-born population	1960 to 2010	U.S. Census Bureau 1960 decennial census and 2010 American Community Survey; Pew Research Center tabulations	Not disputed
3	Conversion of a circular Mexican labor flow into a settled unauthorized population via Western Hemisphere caps and enforcement asymmetry	1964 to 2016	Massey, Durand, and Pren, American Journal of Sociology 121(5) (2016); Massey, Durand, and Malone, Russell Sage Foundation (2002)	Contested
4	Wage, growth, and fiscal effects of post-1965 migration, including the Borjas versus Card dispute over the Mariel natural experiment	1980 to 2017	National Academies of Sciences, Engineering, and Medicine, “The Economic and Fiscal Consequences of Immigration” (2017); Card, ILR Review 43(2) (1990); Borjas, ILR Review 70(5) (2017)	Contested

Closing: a study guide for the evidence

A reader who has worked through the four layers owns the article’s argument, and the closing task is to make that ownership durable. The discipline to carry forward is the one this article practiced throughout: demographic facts get census citations with years, economic and fiscal effects get the consensus report plus named dissenting researchers, and no claim appears without a named study or a government data series with a period. The habit to build is the separation of the undisputed from the contested: the foreign-born share rising from 4.7 percent in 1970 to 13.7 percent in 2018 is not disputed, the origin composition shifting from 75 percent European in 1960 to 81 percent Latin American and Asian in 2010 is not disputed, the statutory mechanism of the uncapped immediate relative category and the compounding family preferences is a matter of text, while the wage effects, the fiscal balance, and the Massey causal chain are argued in the literature and should be cited as argued rather than settled.

For readers who want to test their command of the material, the VaultBook legislation study notebook provides a structured format for recording the four layers, the sources and years for each demographic fact, the statutory citations for the mechanism, and the named researchers on each side of the contested economics. Working through the table above row by row, writing out the source and the period for each outcome from memory and then checking against the article, is the fastest way to convert a reading into knowledge. The article’s highest value is not any single figure but the method: measure the statute against what it actually did, keep the mechanism separate from anyone’s intentions, and label the disputed questions disputed.

Frequently Asked Questions

Q: How did the Immigration and Nationality Act change American demographics?

The act presided over two large demographic changes that the census records without dispute. The foreign-born share of the population rose from 4.7 percent at the 1970 census, an all-time low, to 13.7 percent in the 2018 American Community Survey, approaching the level last seen in the early twentieth century. The origin composition shifted from 75 percent European in the 1960 decennial census to 53 percent Latin American and 28 percent Asian in the 2010 American Community Survey. The statutory engine was the exemption of immediate relatives of citizens from all numerical limits, combined with family preference categories that compounded as each new citizen gained petitioning power. The economic and fiscal effects of these changes are genuinely contested in the research literature.

Q: Did the Immigration and Nationality Act increase immigration numbers?

The census record shows that the foreign-born population grew from 9.6 million, 4.7 percent of the population, in 1970 to 44.8 million, 13.7 percent, in 2018, with the largest decade gain in the 1990s. The increase did not come from the act’s nominal ceilings, which were modest: 170,000 for the Eastern Hemisphere and 120,000 for the Western Hemisphere from 1968. It came from the uncapped immediate relative category and the compounding family preference structure, which grew by their own logic without further congressional action. Other factors also contributed, including refugee admissions and the 1986 legalization of approximately 2.7 million people. The act’s measurable demographic footprint is therefore large, but it ran through the provisions Congress left uncapped rather than the ceilings it set.

Q: What is the foreign born share of the US population since the Hart-Celler Act?

At the 1970 decennial census, taken after the act’s signing but before its principal provisions had compounded, the foreign-born share stood at 4.7 percent, or 9.6 million people, the all-time low in the recorded series. It rose to 6.2 percent in 1980, 7.9 percent in 1990, and 11.1 percent in 2000, all from decennial census long-form data. The series then shifts to the American Community Survey: 12.9 percent in 2010, 13.1 percent in 2013, and 13.7 percent, or 44.8 million people, in 2018. The 2009 ACS recorded 12.5 percent. Readers should note the source shift between the decennial long form and the ACS when comparing figures across the year 2000 boundary, though the direction and scale of the rise are robust to it.

Q: Where did immigrants come from before and after the Hart-Celler Act?

In the 1960 decennial census, 75 percent of foreign-born residents had been born in Europe, 10 percent in Northern America, 9 percent in Latin America, and 5 percent in Asia. In the 2010 American Community Survey, 12 percent had been born in Europe, 2 percent in Northern America, 53 percent in Latin America, and 28 percent in Asia. The country-level comparison is equally stark: the 1960 leading countries of birth were Italy, Germany, Canada, the United Kingdom, and Poland, while the 2010 leaders were Mexico at 11.7 million, China, India, the Philippines, and Vietnam, with no European country in the top ten. The 1970 census shows the composition partway through the transition, and the 2000 census shows Latin America and Asia together at 78.2 percent of the foreign-born population.

Q: Did the Hart-Celler Act’s sponsors predict what happened?

This article does not adjudicate what the act’s sponsors predicted or intended, because its method is deliberately mechanism-based rather than intent-based. The article’s namable finding states that the demographic transformation followed from two specific provisions, the unlimited immediate relative category and the compounding family preference structure, and that explaining the outcome by reference to anyone’s intentions, in either direction, is a weaker account than explaining it by reference to the text. The compounding mechanism operates as petitioning arithmetic regardless of what any legislator expected, and the small initial base from Asia and Latin America explains the slow start and later acceleration without any reference to forecasts. Readers seeking the legislative history should consult this series’ passage history article.

Q: What does research say about immigration’s effect on wages after the Hart-Celler Act?

The anchor is the 2017 National Academies consensus report, edited by Francine D. Blau and Christopher Mackie, which found that when measured over ten years or more, the impact of migration on the wages of native-born workers overall is very small. Where negative effects are detected, they are concentrated on two groups: prior migrants, the closest substitutes for new arrivals, and native-born workers without a high school credential. The most prominent dissenting researcher is George Borjas, whose 2017 reappraisal of the Mariel boatlift found a sharp wage decline for low-skill men, while David Card’s 1990 study of the same event found no discernible effect. The report’s formulation accommodates both findings, and the dispute remains an unresolved methodological disagreement about subgroup measurement rather than a settled question.

Q: Is immigration under the Hart-Celler Act a net fiscal gain or cost?

The 2017 National Academies consensus report gives a split answer that is frequently misreported by quoting only one half. The first generation of migrants is more costly to governments than the native-born population, but the second generation, the children of migrants, are among the strongest fiscal and economic contributors in the United States. The long-term fiscal effects are generally positive at the federal level but remain negative at the state and local level, driven especially by the costs of educating the children of migrants, since public schooling is financed at the state and local level. Researchers continue to dispute the generational accounting models behind these findings, with critics arguing about the treatment of long-run costs and second-generation contributions. The honest summary is the split itself: federal positive, state and local negative.

Q: Did the Hart-Celler Act cause unauthorized migration from Latin America?

The serious causal argument on this question comes from Douglas Massey and his coauthors, and it identifies a chain rather than a single cause. Congress ended the Bracero agricultural labor program on December 31, 1964, removing the principal lawful channel for circular Mexican labor migration. The 1965 act then imposed the first Western Hemisphere ceiling of 120,000 from 1968, where no cap had existed before, and the 1976 amendments extended the 20,000 per-country limit to the Western Hemisphere, imposing a binding cap on Mexico. With demand persisting and lawful channels narrowed, the flow continued outside lawful channels. Later border enforcement then raised crossing costs, which deterred return migration more than entry, increasing the settled unauthorized population. This is presented as a well-cited scholarly argument, stated in Massey, Durand, and Pren’s 2016 American Journal of Sociology article, not as a settled census fact.

Q: When did the foreign-born share reach its historic peak and its historic low?

The historic peak in the recorded series came at the 1890 census, when the foreign-born population reached 14.8 percent of the total population. The historic low came at the 1970 census, when the foreign-born share stood at 4.7 percent, or 9.6 million people. The low reflected more than four decades of the national origins quota system, which had held admissions to low annual levels while the native population grew. The 2018 American Community Survey recorded 13.7 percent, approaching but not reaching the 1890 peak. The article’s framing of the post-1965 period as a movement from the all-time low back toward early-twentieth-century levels rests on these two bookend figures, both drawn from the Census Bureau’s long-term historical series on the foreign-born population.

Q: What is the immediate relative exemption and why does it matter demographically?

Section 201(b) of the amended act provides that the immediate relatives of United States citizens, their spouses, children, and parents, with a citizen child petitioner required to be at least twenty-one, are admitted without regard to the numerical limitations in the act. Admissions in this category do not count against any ceiling or per-country limit. Demographically it matters because it is uncapped: while the preference categories queue behind annual limits, the immediate relative channel has no annual fill point, and it grows automatically as naturalization creates new citizen petitioners. Across the decades it became one of the largest admission channels in the system. Any account of post-1965 demographic growth that omits the exemption misses the provision that removed the ceiling from family reunification.

Q: Why did migration from Asia and Latin America accelerate after a slow start?

The compounding family mechanism operates as a multiplier on the existing base of potential petitioners, and in 1960 that base was small for Asia and Latin America: only 5 percent of the foreign-born population had been born in Asia and 9 percent in Latin America. A multiplier applied to a small base produces small absolute numbers at first, which is why growth from those regions started slowly in the late 1960s and early 1970s. As admissions accumulated, naturalization created more petitioners, who sponsored more relatives, enlarging the base on which the next cycle operated. By 2000, Latin America and Asia together accounted for 78.2 percent of the foreign-born population. The acceleration was not a change in the mechanism but the mechanism operating on a larger base, exactly what a compounding process predicts.

Q: What did the 20,000 per-country limit do to the composition of immigration?

The per-country limit caps annual admissions to natives of any single foreign state at 20,000. In the 1965 text it applied to the Eastern Hemisphere; the 1976 amendments extended it to the Western Hemisphere. The limit does not reduce total admissions. It redistributes them, preventing any one country from capturing the preference categories and spreading the available numbers across many nationalities. The compositional effect is toward diversity of origins within the flow. The limit binds hardest where demand most exceeds the allocation, which in the Eastern Hemisphere meant the large Asian countries. Where the limit applied late or not at all, single-country concentrations could grow larger, which is part of why the Mexican-origin share of the foreign born population reached the levels recorded in the 2010 ACS.

Q: Why was Mexico the only country the 1976 per-country limit bound?

Public Law 94-571, enacted October 20, 1976, extended the Eastern Hemisphere preference system and the 20,000 per-country annual limit to the Western Hemisphere for the first time, while keeping the hemispheric ceilings of 170,000 and 120,000 in place. Before 1976, the Western Hemisphere had a 120,000 ceiling but no per-country limits, so admissions were allocated first-come, first-served across the hemisphere. The extension imposed a binding country cap on Mexico, the only Western Hemisphere country whose admissions substantially exceeded 20,000 annually, with initial exceptions of 35,000 each for Canada and Mexico. The Massey literature identifies this 1976 extension as the hinge of the unauthorized migration argument, because it was the first provision to cap Mexican lawful admissions far below the established flow.

Q: How did the 1978 worldwide ceiling differ from the hemispheric ceilings?

Public Law 95-412, enacted October 5, 1978, replaced the separate 170,000 Eastern Hemisphere and 120,000 Western Hemisphere ceilings with a single worldwide ceiling of 290,000. Arithmetically the new ceiling was the sum of the two it replaced, so it was not an increase. What changed was allocation: under the hemispheric system, unused numbers in one hemisphere could not be used in the other, while the worldwide ceiling let the full 290,000 flow wherever the preferences and per-country limits directed. The practical effect was to make the system more responsive to demand as expressed through the preference categories, which reinforced the compounding family dynamic. The 1978 amendments also created the Select Commission on Immigration and Refugee Policy, whose review led toward the 1986 and 1990 legislation.

Q: What demographic effects did the 1990 Act’s diversity category produce?

The Immigration Act of 1990 created a 55,000 diversity visa category for countries with low admission levels in the preceding years, allocated by lottery rather than by family relationship. Unlike the family preferences, which compound through petitioner networks, the diversity category works against the compounding logic by directing numbers to nationalities with small existing bases. Its demographic effect is therefore diversifying rather than compounding: it added African and Eastern European nationalities to the flow in larger numbers than the family preferences alone would have produced, and it is visible at the margins of the origin composition in the 2000s and 2010s. At 55,000 it was small relative to the 480,000 family-sponsored track, so its effect on the overall census figures was modest, but its effect on the diversity of origins within the flow was deliberate and measurable.

Q: How many people gained permanent residence through the 1986 legalization?

Approximately 2.7 million people were approved for permanent residence under the Immigration Reform and Control Act of 1986, Public Law 99-603, signed November 6, 1986. The total combines about 1.6 million grants under the general legalization provisions of INA section 245A, for persons with continuous residence since January 1, 1982, with about 1.1 million grants under the Special Agricultural Worker program, for persons with 90 days of seasonal agricultural work. The 2.7 million figure is the INS and Department of Labor derived total used in this series. The legalization complicates readings of the 1990 census, since the 19.8 million foreign-born residents counted that year included the formerly unauthorized population IRCA had moved to lawful permanent status.

Q: What is the Massey argument about border enforcement and return migration?

The argument, stated precisely in Massey, Durand, and Pren’s 2016 American Journal of Sociology article “Why Border Enforcement Backfired,” is that expanded border enforcement raised the cost and risk of crossing, and that higher crossing costs deterred return migration more than they deterred initial entry. A migrant who had crossed faced a choice between returning to Mexico and risking a costly re-entry later or remaining in the United States, and as crossing grew costlier, remaining became the rational choice. The net effect was to convert a circular flow of workers into a settled population of families, increasing the settled unauthorized population even though the policy was designed to reduce it. The 2002 Russell Sage Foundation volume by Massey, Durand, and Malone documents the 1986 to 1996 enforcement expansion behind the shift.

Q: What did the National Academies report find about wages over ten years or more?

The 2017 consensus report “The Economic and Fiscal Consequences of Immigration,” edited by Francine D. Blau and Christopher Mackie, found that when measured over a period of ten years or more, the impact of migration on the wages of native-born workers overall is very small. The ten-year qualification matters because labor markets adjust over time, and short-run effects can differ from long-run ones. Where negative wage effects are detected, the report finds them concentrated on prior migrants, the closest substitutes for new arrivals, and on native-born workers without a high school credential. The finding is therefore differentiated rather than blanket: very small effects for the native workforce overall, with detectable negative effects limited to specific labor market segments.

Q: What was the Mariel boatlift and why do economists study it?

In April 1980, Fidel Castro declared that Cubans wishing to leave could do so through the port of Mariel, and the arrivals increased the city’s workforce by about 7 percent in a matter of months. Economists study it because the suddenness, size, and geographic concentration of the inflow created a natural experiment: Miami’s labor market could be compared before and after the boatlift against comparable cities that received no such inflow, approximating an experimental design for measuring how a labor supply shock affects wages and employment. David Card’s 1990 study found no discernible effect, while George Borjas’s 2017 reappraisal found a sharp wage decline for low-skill men, making the boatlift the most contested single episode in migration economics.

Q: What measurement choices produced the Borjas and Card results on Mariel?

They studied the same 1980 boatlift but measured different workers. Card’s 1990 ILR Review study examined the Miami labor market broadly, including low-skill workers as a group, and found no discernible wage or employment effects. Borjas’s 2017 ILR Review reappraisal narrowed the focus to low-skill men defined as high-school dropouts, and within that group to non-Hispanic men, reasoning that the arrivals competed most directly with the least-skilled workers and that the Hispanic survey sample was contaminated by the arrivals themselves. On that subgroup, Borjas found a sharp wage decline taking about a decade to recover. Card’s defenders argue the subgroup is too small for reliable measurement; Borjas’s defenders argue broad measurement diluted the effect. The dispute is methodological and remains unresolved.