People Leave Footprints: Millions More Unauthorized Immigrants Cannot Be ‘Hidden’ in Data Estimates
Amid the current roiling political debate in the United States around immigration, and particularly illegal immigration, there is little doubt that an academic article out today in PLOS One contending the unauthorized immigrant population is millions larger than has long been estimated will attract widespread attention.
It is deeply unfortunate, therefore, that this thought experiment from a team of academics who specialize in management studies is based on seriously flawed assumptions leading them to the conclusion that there were at least 16.2 million, and as many as 29.5 million, unauthorized immigrants in the United States in 2016.
This accounting exercise departs dramatically from the estimates generated independently by several organizations in and out of government, using variations of a method whose accuracy has been proven successful in a real-world setting. These estimates, tested against other datasets to ensure their accuracy, range from a low of about 10.8 million to a high of 12.1 million, the latter the most recent estimate from the Department of Homeland Security (DHS).
Even researchers in immigration restrictionist groups have concurred there cannot be millions upon millions of extra unauthorized immigrants hidden in the United States, because, in short, people leave footprints that are seen in statistical records—namely in birth, death, school enrollment, housing, and other records.
We believe these new numbers represent at most an interesting academic exercise, but are ultimately greatly off-base and thus counterproductive to the public’s very real need to understand the true scope of illegal immigration and how best to address it.
Where This Thought Experiment Goes Wrong
While we welcome fresh thinking and creative new methods to estimate a population that is by definition difficult to count, the theory articulated in the article has serious flaws, as we explain here and in greater detail in a formal response, also published today in PLOS One. The journal’s editors invited us to write this response, after we served as peer reviewers for the article.
In brief, we believe that the method:
Fails to sufficiently account for circular migration patterns prevalent in the 1990s. Because the government did not estimate the rate of successful illegal border crossings in the 1990s, the authors apply 2005-10 DHS estimates of detection rates to the 1990s. These rates estimate how many people successfully snuck across the border for each person apprehended.
However, crossing patterns were very different in the 1990s than in the mid- to late 2000s. In the 1990s, many migrants crossed multiple times in the same year, or they came for just a year or two before permanently leaving. Back then, illegal crossers faced few consequences, so little deterred them from coming, leaving, and returning again. As border enforcement increased strongly over the 2000s, resulting in higher smuggling costs and growing consequences for illegal entry (and in particular illegal re-entry), people who crossed illegally tended to remain in the United States. Therefore, in the 1990s, far more individuals were apprehended repeatedly than was the case in the 2000s. Applying 2005-10 apprehension rates to the 1990s leads the researchers to overestimate how many people crossed the border illegally in the 1990s.
Separately, they also overestimate how many of those who came actually stayed, by applying departure rates from studies of immigrants overall—not just unauthorized immigrants—to border crossers. The result of both of these flawed assumptions: the authors vastly overcount how many unauthorized immigrants came and stayed during the 1990s. They conclude the unauthorized population numbered at least 13.3 million in 2000, while DHS put the total at 8.5 million. By overestimating the number who arrived in the 1990s, the researchers’ estimates into later years thus build on a shaky foundation.
Is misaligned with Census data. Demographers have long agreed that the decennial Census undercounts unauthorized immigrants. But the 13.3 million number is highly implausible. It would imply that the 2000 Census missed almost 5 million more unauthorized immigrants than demographers thought, for an undercount rate of 42 percent. That is well in excess of even the highest assessment of the Census undercount. When demographers compared the U.S. Census data to U.S. birth and death records and data from the Mexican Census on changes in the Mexican population, they concluded that the 2000 Census could have undercounted unauthorized Mexican immigrants by at most 26 percent—a far cry from 42 percent. Other assessments were far more modest, including one survey of unauthorized immigrants in Los Angeles, which found that just 10 percent said they had not taken the 2000 Census.
Allows mistakes to snowball over time. Beyond flawed assumptions, the authors also use a flawed process. While they begin with the widely accepted estimate of the size of the unauthorized population in 1990—3.5 million—their method for extrapolating its future size quickly falls apart. From the 1990 number, they estimate the size of the unauthorized population for the following years by adding in their estimates of each year’s illegal border crossers and visa overstayers and subtracting those who die, leave the country, or transition to legal status in that year. But in this accounting method, any error in estimating border crossers or those who leave the country—as explained above—gets compounded over time, leading their mistakes to snowball. By the time they build their estimate of the unauthorized population in 2000, their number is far higher than could be validated by any other means. And building from that highly questionable estimate for later years only leads to even higher, and more flawed, numbers for 2016.
What the Traditional Method Gets Right
The demographers who use the traditional method for estimating the unauthorized population—known as the residual method—start fresh each year, looking at Census data and government data on visas granted to legal immigrants. And then these demographers, who work independently at DHS, the Pew Research Center, and the Center for Migration Studies of New York, to mention the most notable users of this method that MPI also employs, double check their estimates against other sources: Data from Mexico, birth and death records, school enrollment records, and other datasets. In this way, traditional estimates have guardrails: They stay aligned with the best Census and administrative data available on immigrant populations in the United States.
The residual method was put to the real-world test successfully in the 1980s, with estimates generated with this methodology largely similar to the actual number of unauthorized immigrants who came forward to get legalized under a broad legalization offered in the Immigration Reform and Control Act of 1986.
It is also worth noting that while several organizations use the residual method, each has its own proprietary methodology for developing datasets on the unauthorized, and that while these have some variations, their results fall within the same relatively narrow range. That is a far cry from the 13 million swing that the Yale researchers generate in their exercise, depending on which of their assumptions they use.
Over several years, MPI carefully developed its methodology with demographers at The Pennsylvania State University’s Population Research Institute and Temple University. Together, we have been transparent about the assumptions undergirding our methodology, sharing them in leading demographic journals, through presentations at academic conferences, and inviting critiques from others in the field.
It is imperative for the public and decisionmakers alike to know how many unauthorized immigrants are in the country, so that we can determine the effectiveness of our immigration and border-control policies and make any necessary adjustments. Inaccurate, inflated estimates only serve to inflame and confuse the debate, and could lead to poorly designed and wasteful enforcement and policy overreactions.