Prior estimates of undocumented immigrants in the US were way off

At a time when immigration and border security have become contentious issues in American politics, the number of undocumented immigrants living in the U.S. gets a lot of attention. Officially, most estimates put the figure at roughly 11.3 million.

As an immigrant myself, this data point is of personal interest. I’ve long wondered: How do they really know how many undocumented immigrants are here?

{mosads}The current estimate is derived from population surveys and legal immigration records. But it’s not as though all the people living in the U.S. without permission from the American government are filling out surveys and asking to be identified.

Even though the vast majority of undocumented immigrants are law-abiding citizens, they’re part of a “hidden population” — and they don’t necessarily want to be counted. It’s like trying to tally the number of people who evade taxes or who use drugs. Where do you even begin?

But just because it’s hard to count them, should we not bother to try? If you want to solve a problem, you need to know the scope of it.

Debates about the amount of resources to devote to undocumented immigrants and the relative benefits and disadvantages of various policies — including deportation, amnesty and border control — depend on having a correct estimate of just how many of them are living here. The number sets the scale and accuracy is critical. 

Recently, my colleagues, Edward Kaplan and Jonathan Feinstein, both from Yale School of Management, and I sought a better approach to counting this population. 

Using demographic and mathematical modeling, and armed with the best available data, including some that have only recently become available, we arrived at a higher number. Much higher. 

According to our analysis, published this week in the journal PLOS ONE, there are roughly 22.1 million undocumented immigrants in the U.S. Even using extremely conservative model parameters, we estimate a population of 16.7 million undocumented immigrants.

The discrepancy in the numbers reflects the challenge of relying on survey-based methodologies to identify a hidden population.

For one, the approach entails reaching a representative sample of all those born outside of the U.S. Second, it requires achieving accurate responses from survey respondents to questions about where they were born and whether they are American citizens.

It’s likely that undocumented immigrants are more difficult to locate and survey than other foreign-born residents, and if contacted, they may be inclined to misreport their country of origin, citizenship and number of household residents, fearing the legal consequences of revealing their status.

That’s why a different approach is required. My colleagues and I used a new method based on operational data — such as border apprehensions, deportations and visa overstays — and demographic data — including mortality rates and emigration rates.

We combine these data using a mathematical model that follows a simple logic: The population today is equal to the initial population plus everyone who came in minus everyone who went out.

While the logic is simple, bringing all the different sources of data together was a complex task. The inflows and outflows components of the model are each comprised of numerous subcomponents.

Each subcomponent must be aggregated from different sources, evaluated for its specific level of certainty, and then incorporated into the mathematical model in a consistent way. 

Some of the data we use is new, so our approach is timely. For instance, the U.S. government only started systematically collecting data on the number of people who overstayed their visa in 2015. 

Our study also includes estimates of unlawful border crossings. We don’t necessarily know the number of people who cross the border successfully — they don’t advertise it. But because the Department of Homeland Security fingerprints every person who gets apprehended, we do know how many get caught trying.

From the apprehension data, it’s possible to reverse engineer an estimate of how many people must have tried to cross the border.

Our research informs the debate around immigration in several ways. Take the question of whether immigrants bring crime to America, for instance. A common argument in favor of a tougher immigration policy is that people who have entered the country illegally elevate levels of violent criminal activity.

Whatever the extent of criminality that is assessed, it’s clear that crime statistics be thought of in relation to a substantially larger population of undocumented immigrants. This lessens the risk in per capita terms. 

Social services are another consideration. Our findings suggest that the agencies and nonprofits tasked with determining the appropriate service levels for undocumented immigrants might need to revise their estimates. What’s acceptable for a population of 11 million is unlikely to be sufficient for a population of 22 million. 

Are our numbers foolproof? No. But while our models are based on a number of assumptions and uncertainties, we are confident that the number of undocumented immigrants has always been higher than what was previously understood. The question now is: What are we going to do about it? 

Mohammad Fazel-Zarandi is a senior lecturer in the Operations Research and Statistics group at the Sloan School of Management at MIT.