Two Economists, One Boatlift, and the Fight Over Immigration and Wages

Economics has a reputation for being unable to run experiments. Then, in the spring of 1980, Fidel Castro ran one for them, without asking anybody's permission, using a fleet of fishing boats and a harbor east of Havana.

The experiment was crude, it was unplanned and it was real. A medium-sized American city's labor force grew by several percent in a matter of months. Two serious economists have been fighting over what happened next for more than thirty years, and the fight is the best classroom we have for the question every voter asks: when more workers arrive, what happens to the pay of the workers already here?

What happened at Mariel

The facts first, at the level where historians agree. In early April 1980, a group of Cubans drove a bus into the grounds of the Peruvian embassy in Havana seeking asylum, and within days thousands more crowded in. Castro responded by announcing that anyone who wished to leave could do so from the port of Mariel. Cuban Americans in South Florida set out in boats to collect relatives. The Carter administration, which had just signed the Refugee Act of 1980 in March, first spoke warmly of welcoming the arrivals and then spent months trying to control a flow it could not stop. By the time the boatlift ended in the fall of 1980, roughly 125,000 Cubans had reached Florida.

A large share of them stayed in Miami. The commonly cited result, and the one both sides of the later fight accept as a starting point, is that the Miami labor force grew by roughly seven percent in a very short time. Many of the arrivals had little schooling and little English. A good number were young men. A smaller number had been released from Cuban prisons and mental institutions, a fact Castro used as a weapon and one that made the American reception more difficult. The vast majority were ordinary people looking for work.

For economists the beauty of the event is its shape. It was sudden. It was large relative to a single local labor market. It was driven by a political decision in Havana rather than by anything happening in the Miami economy, so the usual problem, that immigrants go where jobs already are, does not apply. If you wanted to design a clean test of whether a labor-supply shock cuts wages, you would be hard pressed to improve on it.

The simple model, and why it matters

Let me state the theory the experiment tests, because the stakes are not about Cuba. They are about a model that every first-year student learns.

In the basic model, the price of labor, the wage, is set where the supply of workers meets the demand for workers. Add workers to a particular kind of job without changing anything else, and the supply curve shifts out. The wage for that kind of job falls, and the people who compete most directly with the newcomers, who have similar skills, similar language and the same sort of résumé, feel it first. The workers who do different jobs, or who own capital, or who employ labor, come out ahead. Economists call the winners complements and the losers substitutes.

Here is the premise in one sentence: a new worker who does what you do is competition, and competition for a fixed number of jobs lowers the price. If you reject the premise entirely, you are saying a surge of new workers has no effect on the price of labor, which would make labor the only good in the economy for which supply does not matter. A fair-minded economist does not say that. The real argument is over how fast the labor market adjusts, whether employers expand to use the extra labor, and which workers are close substitutes for which.

Card's finding

In 1990, David Card, an economist then at Princeton, published a paper on the Miami labor market after Mariel in the Industrial and Labor Relations Review. His method was what economists call a natural experiment, or a difference-in-differences design. He took Miami as the "treated" city and compared its wage and unemployment trends before and after 1980 against a handful of other American cities that looked like Miami before the boatlift. His comparison group included Atlanta, Houston, Los Angeles and Tampa-St. Petersburg.

What he found surprised many people. Over the period he looked at, the wages of the workers he studied in Miami did not fall relative to the comparison cities, and the unemployment rate among those workers did not rise relative to them. That included workers without much schooling and, in particular, Black workers, who were the group many people thought would be hit hardest. His conclusion was that the Miami labor market had absorbed the new arrivals with little visible effect on the wages and employment of the people already there.

Card's paper became a landmark. It was a central reference for the argument that immigration does not hurt native-born workers' wages, and his name is attached to the broader body of work on natural experiments that was recognized with a Nobel Prize in economics in 2021. I mention the prize because it is fair to a man whose work you are about to hear questioned: this was a serious paper by a serious scholar, and it was a real contribution to how economists test claims.

Borjas's reanalysis

George Borjas, a Harvard economist and himself a Cuban-born immigrant who arrived in the United States as a boy, has spent a career studying the economics of immigration, and he was never persuaded that the Mariel result closed the question. In 2017 he published "The Wage Impact of the Marielitos: A Reappraisal" in the same journal where Card had published. His argument was about whom Card had counted.

Borjas's first point was that the group at risk from the Marielitos was narrow. The boatlift's arrivals were disproportionately people without a high school diploma. So the right group to look at is native workers without a high school diploma, not all Miami workers, because averaging them with everyone else dilutes any effect on the people who are truly exposed. Borjas also narrowed to non-Hispanic men aged 25 to 59, to avoid mixing in earlier Cuban arrivals and women, whose labor-force participation was changing for unrelated reasons in those years. And he used a stricter comparison group than Card did, selecting comparison cities on whether their wage trends looked like Miami's before 1980.

With those choices, Borjas found a large wage drop. His estimates showed wages of the group he studied falling sharply in Miami relative to the comparison cities, by amounts on the order of 10 to 30 percent depending on the specification. If you take that estimate at face value, the Marielitos were devastating to the workers who competed with them directly. He made the point in terms that fit the simple model: the workers who looked most like the new arrivals were the ones whose pay fell.

The method fight

Now we reach the part where a fair reader has to slow down, because both sides have made real points and the dispute is mostly about samples.

The first issue is sample size. The surveys that economists use to measure wages, the Current Population Survey among them, draw a small sample of people in any given city. When you narrow the group to non-Hispanic men aged 25 to 59 without a high school diploma, in one city, in one year, you may be looking at very few people. Small numbers are noisy, and noise can look like an effect. Borjas's critics argued that his sharper estimate came from slicing the data so thin that it became unstable.

The second issue is composition. Critics, including Michael Clemens and Jennifer Hunt, pointed out that the way the Current Population Survey sampled Miami changed around the time of the boatlift, in a way that raised the share of Black respondents in the low-skilled sample. Since Black workers had lower average wages in the data, the sample mix by itself could pull measured wages down, without any change in what any given worker earned. Their claim was that a good part of the apparent wage drop was an artifact of who was in the sample, not a change in pay. Borjas answered that the sample shift was real but that his result held up when he adjusted for it, and the argument went another round.

The third issue is the comparison cities. Card chose a group of cities by hand. Borjas chose a stricter group by trend-matching. Other economists, among them Giovanni Peri and Vasil Yasenov, used a statistical technique called the synthetic control method, which builds a weighted comparison out of many cities, and reported that they found no wage effect. Borjas objected that the technique, applied to this question, can be steered by the choice of which cities to include.

The fourth issue is the timing and the scope. Even if low-skilled wages dropped for a while, how long did it last? Did other Miami workers move to other cities, so that the measured effect was spread out? Did employers shift toward industries that use unskilled labor, like apparel and construction, dampening the wage effect? Each of those is a plausible channel, and each of them can make the local wage effect look smaller than the true effect of the supply shock. A city is an open economy. Native workers can leave, and capital can move in. That is a reason to think the local estimate understates the competitive effect, and also a reason to think any large effect would be temporary.

What each side gets right

Let me be fair to both of them, because both deserve it.

Card is right that the Mariel episode is not evidence of catastrophe. The Miami economy did not collapse. Unemployment among existing workers did not explode. The claim that a surge of low-skilled immigrants will automatically crush wages in a city is not supported by the best-known test case.

Borjas is right about the logic of the question and about the discipline of the method. The workers most exposed to a labor-supply shock are the ones most like the new arrivals, so averaging them with everyone else is the wrong way to find an effect. If a policy changes the competitive position of one group, you look at that group. His principle, that you aim the study at the people at risk, is sound, whatever you think of his estimate.

And on the sampling problem, the honest answer is that nobody can resolve it from the data available. The surveys are what they are. They do not contain enough observations of this specific group in this specific city to settle a dispute of this kind.

What a fair reader can conclude

I would draw five conclusions, and I will mark which are facts and which are judgments.

Fact one: The Mariel boatlift added roughly 125,000 people to South Florida in 1980, and it raised Miami's labor force by roughly seven percent. Both economists work from that.

Fact two: The two best-known analyses of the wage effect reached opposite conclusions, and the gap between them comes from choices about whom to include and which cities to compare, not from different raw facts.

Judgment one: The result for low-skilled workers is genuinely uncertain, and anyone who tells you the matter is settled in either direction is overselling. If the estimates range from "zero" to "large," then the honest position is wide error bars.

Judgment two: The simple model is not refuted. A model that predicts a wage effect on close substitutes is not defeated by a study that fails to find one in a noisy dataset, any more than the model is proven by a study that does. It is a prior, and it deserves a respectable weight.

Judgment three: The effect on low-skilled workers deserves more attention than the effect on the average worker. Averages hide the people at the bottom of the labor market, and a policy that is good for the average and bad for the least advantaged is a policy with a moral cost, whatever the headline says. This is also where the National Academies, in the report covered in What the National Academies Found on Immigration's Fiscal Costs, placed the most likely wage losers: earlier immigrants and native-born workers without a high school diploma.

Why the answer matters to a policy

Now the stakes. If Card is right, a country can admit large numbers of low-skilled workers and the wage effect is small. If Borjas is right, it cannot, and the cost falls on the workers who already have the least. Which one you believe changes whether you think immigration rules should consider the skill mix.

I would argue that the policy conclusion does not depend as heavily on settling the dispute as people think. Suppose the effect is small. Then a policy that favors higher-skilled arrivals costs us little in the labor market, because a small effect means the lower-skilled flow was not that important, and gains something in the fiscal ledger. Suppose the effect is large. Then a policy that favors higher-skilled arrivals protects the most vulnerable workers. In one world the skill-based system is a cheap improvement, and in the other it is an important one. That is what a decision under uncertainty looks like, and the answer that works either way is the sensible one.

There is also a consideration the wage studies leave out. The Constitution lets Congress set the terms of naturalization, and the Founders assumed that the national government would weigh the national interest in doing so. Article I, Section 8, Clause 4 gives Congress the power "To establish an uniform Rule of Naturalization." A rule has to be written by somebody who is accountable to voters, and the people who bear the cost of any wage effect are voters. Concern for the worker at the bottom of the ladder is not a prejudice against anyone. It is the ordinary duty of the legislature that makes the rule, and it connects to the argument in Jobs Americans Won't Do Is an Argument About Wages, Not About Americans, which asks the question from the employer's side.

There is a related story about how an economy adapts when the cheap labor goes away, told in When the Bracero Program Ended, California Built a Tomato Harvester. And the enforcement side of the labor market, which determines whether employers compete on wages or on who will break the law, is covered in The Employer Who Hires Off the Books Is Breaking the Law, Not Helping.

A word about the people

One thing I want to say before the numbers swallow the point. The Marielitos were human beings, many of them fleeing a regime that denied them the right to leave, and the Miami community that took them in did so with a generosity that deserves respect. The question of wages is a question about policy design, not about the worth of anybody who arrived in 1980. You can believe that the Mariel migrants made good lives, and many did, and still ask whether a country that admits a large number of people at once, without regard to the match between their skills and the labor market, is being fair to the low-wage workers already standing in that labor market. Both things are true, and a conservative approach to immigration holds both.

The case in brief

1. The Mariel boatlift of 1980 brought roughly 125,000 Cubans to Florida and raised Miami's labor force by roughly seven percent, a rare natural experiment.
2. David Card's 1990 study compared Miami with other cities and found no clear wage or unemployment effect on low-skilled workers.
3. George Borjas's 2017 reanalysis narrowed the sample to non-Hispanic men aged 25 to 59 without a high school diploma and found a large wage drop, and critics answered that his slicing and the survey's changing sample drive the result.
4. A fair reader concludes the question is unsettled, the simple supply-and-demand model deserves real weight, and the people at the bottom of the labor market deserve more attention than the average.
5. Because skill-based selection is a cheap improvement if the effect is small and an important one if it is large, it is the sound policy either way.

What you can do: read the abstracts of both papers on the website of the Industrial and Labor Relations Review and then write to your member of Congress through the representatives page asking how immigration rules account for the workers at the bottom of the wage scale.