The Causal Question

Does a sudden influx of low-skilled immigrant labour reduce wages or increase unemployment for native workers? This question lies at the heart of the economics of immigration, and answering it from observational data is difficult: immigrants tend to settle in cities with strong labour markets, creating a spurious positive correlation between immigration and native wages that masks any true negative effect.

In 1990, David Card turned to a natural experiment to break the confounding. His study of the Mariel boatlift remains one of the most influential — and contested — empirical papers in labour economics, a case study in both the power and the fragility of quasi-experimental designs.

Identification Strategy

On 20 April 1980, Fidel Castro announced that any Cuban wishing to emigrate could leave from the port of Mariel. Between April and October 1980, approximately 125,000 Cubans arrived in Miami, increasing the city's labour force by roughly 7% virtually overnight (Card(1990)). The immigrants were disproportionately low-skilled: unlike the earlier wave of Cuban exiles who had come largely from the professional class, the Marielitos included a broader cross-section of the Cuban working population.

The key identification insight is that the timing and destination of the boatlift were determined by Cuban political events — specifically, Castro's response to the occupation of the Peruvian embassy by Cubans seeking asylum — rather than by any feature of Miami's labour market. This exogeneity makes the boatlift a natural experiment: Miami effectively received a large, random (with respect to labour market conditions) shock to its low-skilled labour supply.

Card's strategy was a difference-in-differences comparison. He compared wages and unemployment in Miami before and after 1980 with trends in four comparison cities: Atlanta, Houston, Los Angeles, and Tampa–St.\ Petersburg. These cities were selected because their demographic composition and pre-1980 labour market trends resembled Miami's, satisfying the parallel trends assumption that Miami and the comparison cities would have followed similar paths absent the boatlift.

The identifying assumption is: \[\begin{equation}
   \mathbb{E}[Y_{Mt}(0) - Y_{Mt'}(0)] = \mathbb{E}[Y_{Ct}(0) - Y_{Ct'}(0)],
   \label{eq:paralleltrends}
\end{equation}\] where \(Y_{Mt}(0)\) is Miami's potential outcome (wages/unemployment) absent the boatlift in period \(t\), and \(Y_{Ct}\) is the corresponding outcome for comparison cities. If this holds, the difference between Miami's actual post-boatlift outcomes and the counterfactual (proxied by comparison cities) identifies the causal effect.

Data and Setting

Card used two main data sources: the Current Population Survey (CPS) Outgoing Rotation Group files, which provide repeated cross-sections of wages and employment, and the CPS Annual Demographic Supplement (March files), which tracks unemployment and labour force participation. He focused on the period 1979–1985, examining outcomes for workers without a high school diploma — the group most likely to face direct competition from the Marielitos.

Miami's low-skilled labour market in 1979–80 was already distinctive: it had a large Cuban-American community, a service-sector economy, and tourism-dependent employment. The comparison cities were chosen to match these features as closely as possible.

Key Findings

Card's central finding was striking in its simplicity: the Mariel boatlift had virtually no discernible effect on the wages or unemployment rates of low-skilled workers in Miami. The wage series for Miami tracked the comparison cities closely through the early 1980s. Unemployment did rise in Miami during 1981–82, but this spike was also visible in comparison cities — reflecting the national recession rather than any boatlift effect.

Table  summarises the qualitative pattern of Card's findings.

OutcomeMiami trendComparison cities trendWages (high school dropouts)No detectable fallSimilar trajectoryWages (Cuban immigrants)No detectable fall—Unemployment (Black workers)Rose 1981–82Rose similarlyUnemployment (overall)Rose 1981–82Rose similarlySummary of Card (1990) findings: Miami versus comparison cities, 1979–1985

Card argued that this null result was consistent with a model in which immigrants are imperfect substitutes for native workers, or in which the local economy absorbed the shock through expansion of immigrant-intensive industries, spatial sorting, or capital adjustment.

The Borjas Reanalysis and the Aftermath

Card's findings remained largely unchallenged for over two decades. But in 2017, Borjas(2017) published a reanalysis that found large negative wage effects — a 10–25% wage reduction for high school dropouts in Miami relative to comparison cities.

The Borjas–Card divergence stems primarily from two choices:

1. Skill group definition. Card compared all workers without a high school diploma. Borjas restricted the sample to prime-age (25–59) male workers not enrolled in school, dropping women, the very young, and the very old. This dramatically reduced the sample size for Miami, increasing sampling variability.
2. Comparison group. Card used a composite of Atlanta, Houston, Los Angeles, and Tampa–St.\ Petersburg. Borjas dropped Los Angeles (which had its own immigration waves) and used the remaining three, producing different pre-trends.

Peri and Yasenov(2019) and Clemens and Hunt(2019) argued that Borjas's choices dramatically reduced effective sample size, producing noisy estimates driven by a few outlier observations, and that the pre-1980 trends in Borjas's comparison group did not parallel Miami's. Using the synthetic control method with a larger donor pool, Peri and Yasenov found effects consistent with Card's original null.

This debate illustrates a broader lesson: when findings depend on undisclosed or arbitrary choices — comparison group, skill definition, sample restrictions — the credibility advantage of quasi-experimental designs over OLS can be substantially eroded. Pre-registration and transparency about these choices are essential.

Limitations and What We Learn

The DiD estimate identifies the average effect of the boatlift on Miami's low-skilled workers during 1980–85, under the parallel trends assumption. It does not speak to the long-run effects (capital and industry adjustment may take years), to other cities, or to different types of immigration shocks.

The parallel trends assumption is untestable in the post-period and can be evaluated only through pre-trend analysis. Card shows that Miami and comparison city wages tracked each other closely from 1979 to 1980, providing some reassurance, but the comparison group's validity remains contested.

A single-city study may miss general equilibrium effects: if Miami employers responded by opening new immigrant-intensive businesses, or if some native workers relocated, the city-level wage may understate the true wage pressure while other adjustments absorb the shock (Borjas(2003)).

The average null result may conceal distributional effects: even if mean wages are unaffected, the variance might increase, with some workers winning and others losing (Dustmann et al.(2016)).

Despite its limitations and controversies, Card (1990) established an enduring methodological template: exploit geographically concentrated, externally determined shocks to identify causal effects of immigration. The paper demonstrates that even large, sudden immigration shocks need not dramatically reduce native wages, challenging simple competitive labour market models. Whether this result generalises — and why — remains an active research frontier (Dustmann et al.(2016)).

References

1. Borjas, G. J. (2003). The labor demand curve is downward sloping: Reexamining the impact of immigration on the labor market. Quarterly Journal of Economics, 118(4):1335--1374.
2. Borjas, G. J. (2017). The wage impact of the Marielitos: A reappraisal. ILR Review, 70(5):1077--1110.
3. Card, D. (1990). The impact of the Mariel boatlift on the Miami labor market. ILR Review, 43(2):245--257.
4. Clemens, M. A. and Hunt, J. (2019). The labor market effects of refugee waves: Reconciling conflicting results. ILR Review, 72(4):818--857.
5. Dustmann, C., Sch\"onberg, U., and Stuhler, J. (2016). The impact of immigration: Why do studies reach such different results? Journal of Economic Perspectives, 30(4):31--56.
6. Peri, G. and Yasenov, V. (2019). The labor market effects of a refugee wave: Applying the synthetic control method to the Mariel boatlift. Journal of Human Resources, 54(2):267--309.