The Mariel Boatlift

What happens when a city adds 125,000 immigrants almost overnight?

Some housekeeping first:

This is the first post in a living literature review on migration.  Basically, I’ll review what the academic literature has to say about a particular topic in migration.  Topics may be as narrow as “the Mariel Boatlift” or as broad as “what is the effect of migration on economic growth”.  You can see an example of a more complete living literature review at New Things Under The Sun.

I plan to put up a new post on migration about once a month (though it may be more frequent in the next couple of months, as public interest is high).  All migration related posts will live in the migration section of my Substack; you can subscribe to just that if so desired.  Since this is a living literature review, I'll also go back and update these posts as new research comes out; those updates will live on a future website - subscribe to keep in the loop.

This post - and the others in the living literature review - are supported by a grant from Open Philanthropy.¹  There is a fringe benefit to readers of this Substack being grant supported - since it is being supported by external funding, this newsletter will be entirely free.

All opinions and analysis are my own, and not Open Philanthropy’s.  And so, without further ado:

The Mariel Boatlift

On April 20, 1980, Fidel Castro announced anyone who wanted to leave Cuba via the port of Mariel was welcome to do so.²  There were no restrictions on exit, other than requiring emigrants to provide their own transportation off the island.  Within hours, Cuban-Americans organized missions to go pick up refugees; within days, a flotilla of hastily rented vessels was ferrying Cubans across to Miami.

(Image by the US Coast Guard)

By the end of the boatlift in October, about 125,000 Cubans had arrived in the US.  Almost all landed in Miami;³ 60% would still be living there a decade later.

This was a large influx; about 1% of Cubans left Cuba for Miami.  The influx increased the size of the labor force in Miami by 7%; 16% of all Cubans living in America in 1990 arrived during the Mariel Boatlift.  8% of all Cuban-Americans (that were in the country in 1990) arrived in the single month of May 1980.

So what happens to a city when you add 125,000 immigrants almost overnight?  The size and suddenness of the influx has made Mariel a particularly popular way for economists to examine the effects of immigration on destination labor markets.

The Mariel Boatlift was unusual among immigration shocks because the receiving country exercised essentially no selection of those who entered.  The US had no control over the composition of the influx.⁴

Countries rarely accept an unexpected influx of 100,000 new migrants.  Usually, there are changes in the receiving country that make it more likely to want to receive migrants and there is some policy in place that shapes where they go and what they do.  In the case of the Mariel Boatlift, no one was particularly expecting to drastically increase the number of Cuban-Americans and the US was caught largely off-guard.

Mariel is a good natural experiment because Cubans came to Miami not by Miami’s (or the US’) choice; they came to Miami because it was close to Mariel.  It was as-if random that Miami gained 100,000 new immigrants and Detroit did not.

What were the labor market effects of a sudden influx of workers in fields with low barriers to entry?

Most of the economics literature on Mariel has focused on the labor market outcomes from a large immigrant influx.

Over the course of five months, the labor force expanded almost 10%; really, since the majority of Marielitos arrived in May, the labor force in Miami expanded about 5% in a single month.  Most Marielitos tried to get jobs in fields with low barriers to entry and relatively low wages, such as construction.⁵

Mariel is thus a hard test of how migrants who enter industries with a low barrier to entry can affect the wages of natives in the same industries.  If wages of natives did not decline from this large shock, it seems unlikely that there are many - indeed, any - circumstances in which adding additional migrant workers in low-wage positions would drive down native wages in an advanced economy.

The first - and still most famous study - of this question is Card 1990.  Card finds no evidence that the influx of Marielitos drove down wages or increased unemployment in low-wage jobs.  Card uses a difference-in-differences approach.  In this design, you look at how a control group and a treatment group evolve over time, and see how much more (or less) the treatment group changed relative to the control group.⁶  In this paper, he compares how labor market outcomes changed in Miami between 1974 and 1984 to how labor market outcomes changed in the relatively similar cities of Atlanta, Los Angeles, Houston, and Tampa-St. Petersburg. The key assumption is that without the boatlift, wages in Miami would have evolved in parallel with wages in these other cities.

However, empirical economics has come a long way since 1990 (with some of those innovations coming from David Card himself).  Card doesn’t give a strong reason to choose those comparison cities, and it’s possible Miami differs from them in important ways.  There have been two major re-analyses of the labor market impacts of the Mariel Boatlift with decidedly different results: Borjas 2017 and Peri and Yasenov 2019.

Borjas 2017 argues that one should consider the wage effects on native high school dropouts, specifically.  Most of the Marielitos did not complete high school; thus, they would compete in the labor market against other high school dropouts.  Marielitos probably would have no impact on the market for labor for college graduates; it makes sense to focus specifically on the effects on the workers that were most similar to the Marielitos.  He finds that wages for these workers specifically dropped by 10-30%.

Peri and Yasenov 2019 uses a synthetic control method to compare post-Mariel Miami to what Miami might have looked like without Mariel.  In a synthetic control design, you construct a fake city that is similar to your treatment city.  It’s built as a combination of other cities - for instance, your city might be most similar to Houston, but also a little bit similar to Tampa; your synthetic control will be a weighted average of data from these other (not-treated cities).  You can then compare fake-Miami (built out of other cities that didn’t receive the treatment) to real Miami (which did).

Peri and Yasenov find no evidence of a drop in wages or increase in unemployment among high school dropouts in Miami.  This is a true null; they can exclude an increase in unemployment and a decrease in wages for high school dropouts.

So what gives?  Did high school dropouts suffer or not?  Here we turn to Clemens and Hunt 2019.  They point out that Borjas and Peri and Yasenov use different samples.⁷  In particular, Borjas looks at a small portion of the native labor market, excluding “women, Hispanics, workers under age 25, workers over age 59, and workers who have finished high school or its equivalent” in a small survey set (the March Current Population Survey).  That is quite a bit of the population to exclude; Clemens notes he excludes 91% of workers in low-wage jobs.

Furthermore, Borjas’ choice of the Current Population Survey data - and then excluding large portions of it - does seem to matter.  At the same time as the Mariel boatlift, Miami experienced an influx of Haitian refugees.  These refugees had little formal education - only 5% had completed high school - and struggled to find work, much more so than Marielitos.⁸  These refugees make up a substantial portion of Borjas’ sample, because he has excluded almost everyone else.

And if you compare a pre-1980 group of mostly African-Americans, and a post-1980 group of African-Americans and Haitian refugees, the latter group will have lower wages since so many of the refugees struggled to find work.  It’s not because Cubans took their jobs, though.⁹

There are other issues with the Borjas sample - for instance, the Census was making an effort to survey all African-Americans, and therefore, there was increased coverage of low-income African-Americans just around the time of the Mariel Boatlift.  This meant that mechanically, wages for African-Americans decreased because more low-wage African-Americans were added to the sample.  They had already existed before 1980; they just weren’t included in the data.

All of this means that, in my judgement, Peri and Yasenov 2019 contains the best estimates we have on the effect of a sudden influx of workers into fields with low barriers to entry.  And their conclusion was: there was no effect on wages for native workers.

One could imagine that there was no wage effect on natives just because Marielitos didn’t find jobs.  I don’t think this was the reason, though.

It is true that Marielitos struggled to find work once in the US.  In surveys conducted three to four years after the Boatlift, 39% of Cubans were not currently working.¹⁰  But those same surveys seem to indicate that Marielitos were in and out of low-wage jobs, rather than consistently unemployed, as average wages for Marielitos in 1983 were only 20% lower than the average for Cuban-Americans in Florida.  Marielitos were also ineligible for government assistance; the Marielitos who didn’t have family in Miami¹¹ had to work or starve.

So… you added a hundred thousand people to a city nearly overnight.  They all needed jobs; the supply of labor has increased dramatically.  If one considers pure supply and demand, you would expect wages to fall; your supply has suddenly increased but the demand for them is unchanged.  Why didn’t wages fall?

Bodvarsson, Van den Berg and Lewer 2008 has (at least some of) our answer.  Workers are not simply suppliers of labor; they also increase demand for labor.  They buy food; they go to stores; they interact with the local economy.  Adding new people to a city doesn’t just increase supply of labor; it increases demand for the goods and services they provide.  Immigrants might take some jobs, but they also create jobs; these effects appeared to cancel out in Miami.

It is possible there are other explanations as well. Card 1990 and Monras 2020 posit the effects of the Mariel Boatlift were lessened because other immigrants and natives moved out of Miami (or at least the rate of inflow slowed). However, Peri and Yasenov 2019 sees no offsetting outmigration of the natives most similar to the Marielitos. There is no Clemens-esque paper reconciling these two results, so I think it remains unclear how - and if - internal migration played a part in how the labor market in Miami reacted to the Mariel Boatlift.

Did the Mariel Boatlift increase crime in Miami?

Of course, jobs aren’t the only reason that people worry about immigration.  There is a strong stereotype that allowing immigrants into “our” neighborhoods can increase crime, and Marielitos in particular were perceived to be undesirables.

There was widespread press at the time that Castro used the Boatlift to rid himself of “hardened criminals, mental patients, and other deviants” (Portes and Stepick 1985).  The Washington Post estimated ~22,000 of the Marielitos had felony records, about 17% of the influx.  The criminal Marielito even shows up in Scarface.

These concerns were overblown, but not entirely baseless.  Miami had just accepted 70,000 young(ish) men that were loosely attached to the labor market.  This is the demographic group most likely to be involved in criminal activity, and engage in criminal activity they did.  And some of them did have criminal records, though nowhere near 17%.¹²

Billy and Packard 2022 use a synthetic control method to estimate crime rates in Miami had the Mariel Boatlift not happened, and compare against Miami’s real crime rate to determine how the Boatlift affected crime.  They find the Boatlift caused a “nearly 25-32% expansion in property crime”. Robberies increased by 70% and murders by 41%.

Billy and Packard hypothesize that unemployed young men, some of whom had a criminal background prior to arrival in Miami, turned to property crime as a method of making money.  They also note that much of the increase in violent crime appears to have been driven by violence by Marielitos on other Marielitos; other Miami residents may not have experienced an increase in violent crime.

Still: these are very large increases, particularly when one considers baseline crime rates in the 1980s were considerably higher than they are today.  For once, then, it seems that fearmongering about immigrants causing an increase in crime was correct.  This 1985 Sun Sentinel article seems likely to be correct; the legacy of Mariel was not any labor market impacts, but increased crime.

Does that mean immigrants cause crime?  Not necessarily.  It is important to report the facts as we know them, even if I don’t particularly want them to be true, and Billy and Packard’s study does indicate that the Mariel Boatlift did increase crime rates in Miami.

But: it is worth noting (as Billy and Packard do) that the Mariel Boatlift was an atypical immigration episode, where a large number of young men (some with felony records) all arrived in one location in a short time period.  A future post will look at the broader literature on immigration and crime under more common circumstances.

What were the electoral consequences of the Mariel Boatlift?

In 1980, Jimmy Carter was running for re-election against Ronald Reagan.  In November, he lost very badly.  51.9% of Floridians voted for Carter in 1976; only 38.5% voted for him four years later.  Did the chaos around the Boatlift contribute to Carter’s loss in Florida?

According to Thompson 2022, probably not.  Thompson also uses (primarily) a synthetic control design, this time to look at how electoral outcomes changed because of the Boatlift.  He finds that while Reagan did gain vote share in Miami, it doesn’t appear to be because of the Boatlift.  Reagan picked up vote share among all Cuban-Americans, not just those in Miami - likely because he was so vocally anti-Communist.

Other Republicans did not gain as much vote share as Reagan did in Miami in the 1980 election.  Voters did not seem to blame the county or city’s Democratic mayors; Democrats remained in power in Miami.  This is a bit of a surprising result, given the previous section on crime!  Why didn’t Miami voters punish Democrats for the increase in crime?  I don’t really know.

What don’t we know about the Mariel Boatlift?

There are still quite a lot of things we don’t know about the impacts of Mariel Boatlift. A few stand out:

• 1% of Cubans left the island; what did that do to the Cuban labor market?¹³ Did wages in Cuba change?

• What were the long-term outcomes of Marielitos? As far as I’m aware, there are no papers on outcomes for Marielitos after 1990. What happened to Marielitos and their children >10 years after arrival?

So, what can we learn from the Mariel Boatlift?

I have three major takeaways from the Mariel Boatlift:

1. This was a weird policy choice that has not that much applicability to other circumstances and types of immigration.

Mariel was a very large, very sudden influx of people with little vetting, some with prior criminal records.  This is not how immigrant inflows normally go!  Most immigrants have to go through an extensive process to get a visa and there is considerable selection on the part of the receiving country on who gets to enter the country.¹⁴

The Mariel Boatlift is a good natural experiment because the usual vetting and selection process didn’t happen.  But that also means that it’s very different from other immigrant inflows.¹⁵

We can’t really learn anything about most immigration policies from Mariel.

2. But most likely, immigrants do not drive down the wages of native workers.

As I note above, the Mariel Boatlift is an extremely hard test of how an influx of immigrants affects the wages of native workers.  If adding 5% to your city’s labor force in a month doesn’t change wages for even those most similar to the new workers, probably nothing will.

3. Being careful about methodology matters.

As Borjas 2017 and Peri and Yasenov 2019 show, different researchers can look at the same event and draw opposite conclusions based on the methodology they use to study it.  Since these conclusions can influence policy decisions, it’s important to be careful and rigorous about how you study a question.

If one took Borjas 2017 at face value, accepting a large number of low-wage workers would really hurt the least educated Americans.  That might lead one to pursue policies that limited the number of low-wage workers allowed to enter the US.

But my read of the evidence is that Borjas’ result isn’t true; it’s an artifact of unrelated changes in the dataset he uses.  One could then end up making policy based entirely on a change in sample composition, rather than a true effect.

Empirical social science is more than academic; it informs real debates that impact real people’s lives.  It is therefore incumbent upon the empirical social scientist to make sure their results are as robust and as accurate as possible, and communicate transparently about how they got the results they did.  Methodology matters.

Many thanks to Denise Melchin, Jeff Fong, Karthik Tadepalli, Matt Clancy, Luzia Bruckamp, Richard Nerland, Oscar Sykes, and Jesse Smith for their comments and edits on this post.

1.  For more information on this program, see this blog post.

2. Cuba, unlike many developed countries, requires potential emigrants to get permission to leave the country.  This process is difficult even today.

3. Approximately 7500 people left Cuba in April 1980 via flights to Costa Rica and then on to the US; this population did not land in Miami.

4. Indeed, Card 1990 notes that the exact number of Marielitos was unknown, as was the composition of the Mariel influx. The information we do have suggests that Marielitos were younger than the average Cuban-American (three years younger on average), mostly male (55%), had lower levels of education than the average Cuban-American (~40% less likely to have a high school diploma) and were 40% less likely to speak English well (Portes and Stepick 1985, Card 1990)

5. This is partially because language barriers made higher-wage jobs less accessible to them.

6. An explanation of DIDs can be found here, but basically, you’re estimating the treatment effect as the difference between P2 and Q in this graph.

7. For information about labor market outcomes in Miami, Peri and Yasenov turn to the May Outgoing Rotating Group for data rather than the March Current Population Survey, as it has a larger sample size and measures point in time wages rather than recalled yearly wages.  Borjas primarily uses the smaller March Current Population Survey, which had as few as 55 individuals included in his sample in some years (table 1, Peri and Yasenov 2019).  Both of these features increase the size of the error in his estimates, because it is a small sample and people’s memories aren’t that great.

8. Most of the facts in this piece about the labor market for Haitians and Cubans are from Portes and Stepick 1985.  Portes and Stepnick conducted representative surveys of both groups three years after the Boatlift.

9. Perhaps unsurprisingly, Borjas still disagrees with this conclusion, as he does with most studies that show null to positive effects from immigration.

10. The aforementioned Portes and Stepick 1985 surveys

11. Many did have relatives in Miami; indeed, some Miami Cubans went to Mariel specifically to get their relatives and bring them back.  However, this was not the case for all Marielitos, and most did need to get a job to support themselves.

12. Billy and Packard 2022 estimates that about 5000 Marielitos had previously committed crimes the US would consider felonies; about 2000 of those were eventually deported, so about 3000 Marielitos with previous criminal records were able to stay in the US. This is about 2.5% of the total.

13. Richard Nerland, who proofread this post, said “we have like seven papers on possible downward pressure on wages in Miami and none on potential upward pressure in Cuba”.

14. Especially unusually, Mariel immigrants were even sometimes negatively selected - people Castro wanted to get rid of - and some did have criminal backgrounds.

This is unusual for immigrant flows, even when one considers undocumented people; the average undocumented person is considerably less likely to commit a crime than the average native-born person.

15. Perhaps the only comparable recent event was Germany’s admission of a large number of Syrians in 2015 (most of whom were young men), but the policy consequences there seem to have been different.  In Germany, refugees were not particularly likely to commit crimes against Germans but the vote share for right-wing parties did increase - exactly the opposite of what happened in Miami.