Does immigration lower wages? — Work, Chapter 6

New faces have been showing up on your shift. They work hard, they took the job at the going rate or a little under, and a thought arrives that you are not proud of but cannot shake: there is a fixed amount of work and a fixed amount of money to pay for it, and every hour that goes to one of them is an hour, and a dollar, that does not come to you. It is close to the first thing anyone reasons out from a chair: add more people willing to do your job, and the price of doing your job should fall, the same way the price of anything falls when more of it turns up for sale. The question is whether the labour market actually works that way, and it is one of the few questions in economics where a century of argument has been forced up against real data from real waves of arrivals.

The short version of what that data says is layered, and each layer has to be kept distinct, or the answer collapses into a slogan for one side. On average, across the whole workforce, the measured effect of immigration on wages is close to zero. That is not the same as no effect, and the averaging hides the part that matters most to the person on the shift: the cost, where it lands, lands on specific people, mainly the immigrants who arrived just before and the native workers whose skills are the closest match to the newcomers’. Meanwhile the gains are real too, and they go to employers, who pay a little less or expand, and to everyone who buys what the enlarged workforce produces, in slightly lower prices. This chapter walks each layer, because the answer is not a single direction; it is a map of who competes with whom.

The lump of labour, and why it leaks

The fear rests on a picture of the economy as a fixed pie of jobs, and that picture is the first thing the evidence complains about. Immigrants are not only sellers of labour; they are also buyers of goods and services, renters of apartments, customers of the same shops that employ them. The arrival that adds a worker also adds a consumer, and the spending creates work. Employers, facing more available labour, do not simply pocket the difference and hold everything else fixed; some expand, some open a location they would not otherwise have opened, some shift toward doing things in a way that uses the labour that is now around. Capital flows toward the workers. None of this makes the newcomer costless to everyone, but it means the pie is not fixed, and the simple subtraction the fear performs, one job in means one job out, leaks at every step. The question underneath is who, inside the enlarged workforce, now finds themselves competing with the arrival and who finds themselves complemented by it.

That is why the aggregate number and the distributional number can both be true and point different ways. A major review by the National Academies of Sciences concluded that, measured over a decade or more, the impact of immigration on the wages of natives overall may be small and close to zero. The same review reports a much wider spread of estimates for the natives most likely to compete with the newcomers: some studies find sizable negative effects on native workers without a high-school education, the closest substitutes for the least-skilled arrivals, others find effects near zero, and the review names prior immigrants alongside them as the most exposed group. Both findings travel together. The figure below shows why the picture has to carry them both.

Figure 6.1 The estimates, and the dispute inside them. The top strip collects the National Academies panel’s summary of study estimates of how much a one-percent, immigration-driven rise in labour supply moves native wages. In the spatial studies, which compare cities, the estimates for men and for natives taken as a whole crowd near zero, and a real tail of large negative estimates runs to the left for high-school dropouts and workers with similar schooling, the closest substitutes for low-skilled arrivals. Two more of the large negatives, from the national skill-cell tradition, are estimates for men as a whole, while other skill-cell estimates for men sit near zero. The episode economists have returned to most often, the 1980 Mariel Boatlift that brought a sudden wave of Cuban workers to Miami, appears three times and pulls in opposite directions: two estimates from one reanalysis land far out in the negative, another lands slightly positive. The lower panel, drawn on its own separate scale because it measures a different thing, shows why they disagree. The large negative comes from a very small sample of Miami dropouts whose composition changed sharply after the boatlift; once the rising share of Black workers in that sample is accounted for, the estimate moves most of the way to zero, and a reanalysis using a larger sample finds no significant decline at all. The two panels use different units and must not be read against each other; each says what its own axis says. Top: National Academies of Sciences, The Economic and Fiscal Consequences of Immigration (2017), Table 5-2, pp. 242–243. Bottom: G. Borjas, NBER Working Paper 21588 (rev. 2016); M. Clemens and J. Hunt, NBER Working Paper 23433 (rev. 2017), Tables 1 and 2; G. Peri and V. Yasenov, NBER Working Paper 21801 (rev. 2017). Retrieved 2026-07-14.

The Mariel fight, which is a fight about counting

The episode in that lower panel is worth setting out in full, because it is where the whole disagreement lives in miniature. In 1980, over a few months, roughly 125,000 Cubans left the port of Mariel for Florida, and a large share stayed in Miami, enlarging the city’s low-skill workforce suddenly and by an amount no policy would ever have designed. It is the kind of event economists call a natural experiment: a big, fast, unplanned supply shock in one city and not its neighbours. The economist David Card studied it and found that the wages and employment of Miami’s existing low-skill workers, including its Black workers, barely moved relative to comparison cities. For years that was the headline result, and it was read as strong evidence that even a large influx need not push wages down.

Decades later the economist George Borjas reopened the case and reached the opposite conclusion. Narrowing the sample to the workers he judged most exposed, non-Hispanic men without a high-school diploma, and using a particular labour survey, he found a sharp fall in their wages after the boatlift, which his paper puts at ten to thirty percent across its estimates. That finding is why the question is still live. Two teams of economists then reanalysed the reanalysis. Giovanni Peri and Vasil Yasenov showed that Borjas’s result depended on the narrow sample and the particular survey extract; widen either and the decline fades. Michael Clemens and Jennifer Hunt found the mechanism. Borjas’s sample of Miami dropouts was tiny, seventeen to twenty-seven workers in some years, and its racial composition lurched after 1980: the share of Black workers in that little sample rose from about 36 percent in 1979 to about 91 percent by 1985, partly because of the arrival of Black Haitians and improved survey coverage of low-wage Black men already there. Because Black workers earned less on average, a sample tilting sharply Black would show falling average wages even if no individual’s pay changed at all. Adjust for that shift, keeping the comparison cities the headline numbers use, and Borjas’s large negative shrinks to something small and statistically insignificant, close to what Card and to what Peri and Yasenov had found; with the alternative comparison cities Borjas also reports, the adjusted estimates stay negative and significant, at roughly half their original size. Clemens and Hunt point, in support of their diagnosis, to a matching pattern: Borjas found effects about three times larger in one survey than another, and the jump in the Black share of the sample was likewise about three times larger in that same survey. Borjas rejected the diagnosis in a published reply, reporting that in his data the low-skill Miami wage still fell significantly, relative to other labour markets, shortly after 1980, recovering by 1990.

Set out this way, the layers separate, and it is worth being exact about which is which. On the aggregate question, whether a typical native’s wage falls when immigration rises, most estimates sit near zero, and the National Academies review reads them that way. On the Mariel question, whether that particular shock cut the wages of Miami’s least-skilled, the literature is split in print and stays split. What no side disputes is that the sample under the largest negative estimates is very small and that its racial composition moved sharply; what the published exchange disputes is whether adjusting for that shift removes the decline or leaves it standing. That is where the evidence stands, layer by layer.

Who competes, and who gains

Behind the estimates is a mechanism, and it is the one the aggregate average keeps hiding. A newcomer competes with the workers whose skills are the closest substitute for theirs, and complements the ones whose skills fit alongside; the competition is never with everyone at once. A new arrival with limited English and strong hands presses most directly on the last arrival with limited English and strong hands, and on the native worker doing the same manual task; that is where the downward wage pressure concentrates, and it is why the previous wave of immigrants, not the long-settled native, is often the group that loses most. At the same time the same arrival raises the value of workers who do the things the newcomer cannot yet do: coordinate, supervise, deal with customers, work in the language. Those workers become more productive with more hands to organise, and their pay tends to rise. The schematic below lays out that division, because the phrase “immigration lowers wages” collapses a structure into a slogan.

Figure 6.2 Where the effects settle. An arriving worker presses hardest on the workers most like them, the earlier arrivals and the natives doing the same readily substitutable task, and that is where the downward wage pressure concentrates. The same arrival raises the productivity, and the pay, of workers whose skills fit alongside rather than compete, those who supervise, deal with customers, or work in the language the newcomer has not yet mastered. Employers gain from the lower labour cost or the room to expand, and customers gain from slightly lower prices. The near-zero average wage effect is the net of the first two boxes, the concentrated pressure and the spread-out wage gains; the third box holds the gains that arrive as profits and prices rather than as wages. The argument about immigration and wages is very often an argument about which of these boxes the speaker is standing in. Schematic. The incidence of a labour-supply inflow across substitutes and complements, as described in the National Academies review (2017) and the task-based literature it draws on.

The gains are as real as the costs, and worth stating concretely. When employers pay less for labour or expand because labour is available, some of that saving stays as profit and some passes to customers as lower prices for restaurant meals, construction, cleaning, childcare, and the rest of what a low-wage workforce produces. Those customers are, in large part, other workers. So the same event that presses on the wages of the closest substitutes lowers the cost of living for a much broader group, which is a real gain even though it is diffuse and unnamed in a way the concentrated loss is not. The event genuinely helps most people a little, and genuinely hurts some people more than a little, and does not, on the average wage, do much at all at the inflows that have actually occurred.

That last qualifier matters as much here as the tested-levels qualifier did for the minimum wage. The near-zero aggregate is measured over the range of immigration flows countries have actually experienced, and identified mostly from natural experiments and cross-city comparisons rather than from anything randomized. It is a statement about the flows on the record, not a law that any inflow of any size and speed must be harmless. The evidence is direct about what has happened and cannot speak to a hypothetical surge far outside that range.

Immigrants take jobs and push wages down.

Oversimplified Moderate confidence

The claim states as a general law something the evidence supports only in a specific corner, and denies the offsetting effects that come with it. It has a true kernel, which is why the ruling is not the reverse of the claim: the workers whose skills most closely substitute for new arrivals, chiefly earlier immigrants and native workers without a high-school education, do face real downward pressure on their wages, and that cost is concentrated enough to be felt sharply by the people it lands on. But three things break the general claim. First, on the average native wage the measured effect over a decade sits close to zero in most studies, because immigrants add demand and spending as well as labour and because employers and capital adjust; the workforce is not a fixed pie. Second, the largest negative estimates rest on narrower ground than the claim: the reanalysis of the Mariel Boatlift discussed above turns on a very small sample whose composition shifted, and whether adjusting for that shift removes the decline is itself a published dispute rather than a settled point. Third, the claim counts only the loss and ignores the gains that fall to employers and to the customers, many of them workers, who pay lower prices. So the claim is not simply false, and it is not the opposite of the truth; it takes a real, concentrated cost borne by the closest substitutes and inflates it into a wage cut for everyone, while leaving out the near-zero average and the gains on the other side of the ledger.

Sources

The aggregate near-zero and the distributional exception: National Academies of Sciences, Engineering, and Medicine, The Economic and Fiscal Consequences of Immigration (2017), Chapter 5 and Table 5-2 — the review concludes the impact on the wages of natives overall, measured over a decade or more, may be small and close to zero, while estimates for the natives most exposed span a wide range, from near zero to sizable and negative for high-school dropouts, with prior immigrants named alongside them. In the review’s own table, spatial estimates for men and all natives cluster near zero; the large negatives are spatial estimates for dropouts and skill-cell estimates for men.
The Mariel dispute, named on both sides: G. Borjas, “The Wage Impact of the Marielitos: A Reappraisal,” NBER Working Paper 21588 (rev. 2016), finding a sharp wage fall for non-Hispanic male dropouts (about −0.20 to −0.37 log points), against D. Card’s original 1990 finding of little effect. G. Peri and V. Yasenov (NBER WP 21801, rev. 2017) show the result depends on the narrow sample and survey extract; M. Clemens and J. Hunt (NBER WP 23433, rev. 2017) show the Black share of Borjas’s tiny Miami sample rose from about 0.36 to 0.91 between 1979 and 1985, and that adjusting for it, with the comparison cities the headline numbers use, collapses the estimate to a small, insignificant figure. G. Borjas’s published reply (“The Wage Impact of the Marielitos: The Role of Race,” ILR Review, 2019; NBER WP 23504) disputes the correction and reports race-adjusted declines that stay significant in his specifications. The dispute is staged in the chapter, with the composition shift reported and the published exchange over its meaning left standing, not adjudicated away.
The incidence and the gains: a labour-supply inflow presses on the closest substitutes (earlier arrivals and similar-task natives) and complements workers who supervise, serve, or work in the language, whose pay tends to rise; employers gain from lower costs or expansion and customers from lower prices. The average effect is the sum of a concentrated loss and a diffuse gain.
Confidence is moderate under the rubric, which scores the weaker of evidence directness and construct match. Construct match is good on the claim’s wage clause: the plotted estimates measure the native wage response to an immigration-driven labour-supply increase. The claim’s jobs clause rests on the demand and adjustment evidence the chapter describes rather than on the plotted estimates, which measure wages, not employment. The binding, weaker leg is evidence directness. The findings come from natural experiments and cross-city comparisons rather than randomized trials, the largest opposing estimate turns on a contested sample and a published dispute over the correction, and every estimate is bounded to the immigration flows that have actually occurred. That range limit and the quasi-experimental design are what cap the confidence.

Why this one rewards precision

Immigration and wages is a question where the average and the incidence pull in different directions, and where a person can be told a true fact that points them exactly wrong. Tell a low-skilled worker that immigration has no effect on wages and you have hidden the concentrated cost that may be landing on them in particular. Tell them that immigrants push wages down and you have hidden that the estimates for their own group are contested, running from near zero to sizable, that most estimates for the average worker sit near zero, and that they gain as a customer from what they may lose as a competitor. The precise answer, that the measured effect on the average wage sits near zero, that a real and concentrated cost falls on the closest substitutes, and that gains flow to employers and consumers, is harder to chant than either slogan.

The thread running under this chapter and the last is leverage, the same thread the volume opened on when it argued that pay tracks how replaceable a worker is. The pay gap turned on which jobs and hours each worker could bargain into; the immigration fear is, at bottom, a fear about the supply of replaceable labour and what a larger supply does to a worker’s hand. The next flashpoint follows the same thread to the opposite end of the pay scale, where a single worker is treated as nearly irreplaceable and the leverage runs entirely one way. What happened to executive pay, and whether a chief executive who takes home three hundred times the median is worth three hundred times as much, is where the volume turns next.