Research

Better data needed to make good immigration policy

Few of the national surveys on immigrant populations ask respondents about their legal status and insurance coverage. Credit: © iStock Photo PoparticAll Rights Reserved.

UNIVERSITY PARK, Pa. -- As debates rage about the legal status of immigrants, researchers still lack enough data -- and enough of the right data -- to help policy makers make better, more informed decisions, according to a team of sociologists and statisticians.

Few of the national surveys on immigrant populations ask respondents about their legal status and insurance coverage, leaving gaps in the current understanding of, for example, the number of undocumented immigrants in the country, said Jennifer Van Hook, professor of sociology and demography and director of the Population Research Institute. To compensate, researchers use -- imputation methods -- a process to substitute data for the missing information to estimate these numbers.

However, Van Hook said that many of these methods may deliver inaccurate results, which in turn could affect planning and implementation of policies designed to manage immigration.

"Requiring researchers to estimate this key information is like 'trying to spin straw into gold,'" said Van Hook. "Policy makers won't get the complete picture and could, for example, end up facing a range of unintended consequences if they enact legislation based on inaccurate data."

Inaccurate estimates may hit local groups hardest, according to Van Hook.

"Determining what resources are needed at the local level -- for example, what types of resources a community would need for language classes -- is very difficult to assess without knowing the size of these immigrant communities and where they are living," said Van Hook, who worked with James Bachmeier, assistant professor of sociology, Temple University; Donna Coffman, research associate professor of health and human development, Penn State and Ofer Harel, associate professor of statistics, University of Connecticut.

The researchers measured the accuracy of estimates by comparing the most common statistical models used to estimate immigration status and insurance coverage against true population data from the Survey of Income and Program Participation -- SIPP -- one of the few surveys that does inquire about legal status.

Although SIPP does provide information on legal status and insurance coverage, it is considered too small for some types of analyses, particularly studies at the state level, said the researchers, who report their findings online in Demography.

According to the researchers, none of the models produced accurate results when compared to the SIPP data.

Organizations avoid questions about legal status on surveys because of the issue's sensitivity.

"Most surveys do not ask questions, such as 'are you a citizen,' or 'do you have a green card?' " said Van Hook. "It's just too sensitive an issue because many families are facing the threat that they could be split up and even the threat causes them stress."

Facing this type of pressure, many immigrants would likely avoid taking the survey or provide inaccurate information skewing results, according to Van Hook.

Although it may do little to diminish the fear of some immigrants, the researchers suggest that acknowledging the sensitivity of questions about legal status could be one way to gather more accurate information.

Another option is to require cooperation between researchers and government agencies. As it is, researchers cannot easily link and cross-reference all the available databases to create better estimates.

"The most inexpensive and timely way to accomplish this would be to permit administrative record linkages to be used to logically impute legal status," the researchers said.

Still another option is to employ a method that was developed and validated by Van Hook and her team called the cross-survey multiple imputation method. While limited in its usefulness, it would allow researchers to use the SIPP to infer immigrants' status in a much larger data set.

The National Institutes of Health supported this work.

Last Updated July 28, 2017

Contacts