There is obviously, and properly, a rough-and-tumble debate going on about immigration reform. I am not a close student of the topic, but the arguments I see casually right now strike me as heavy on moral outrage and “studies” masquerading as knowledge. Here’s something I think we could do that might help somewhat.
Under the Diversity Immigrant Visa (DIV) program, America offers green cards each year to people from countries that have not been a large source of immigrants over the preceding five years. The program has operated since 1995. There are a total of 50,000 DIV green cards available each year. Generally, on the order of 10 million people around the world apply each year. A lottery is held each year to pick the lucky winners. Subject to proving that they have a high-school education, work experience or the equivalent, and passing an interview, they get a green card.
My hearing the word “lottery” is like a dog smelling bacon.
We are already randomizing applicants via the lottery. What if we were to randomly assign them into different groups, some of which would be subject to exactly the current process, but others subject to different selection processes? A different process could mean something more like Australian or Canadian skills ranking, in which people get more credit for higher levels of educational attainment and demonstration of tangible skills that are in short supply in America. Another could be the administration of something like the battery of aptitude tests that the U.S. military gives to applicants. Another could be a more intensive interviewing process. With N = 50,000, we could have several different methods tested each year.
We could make a requirement of entering the lottery that applicants agree that their IRS, arrest records, medical records, and other administrative data will be available for confidential analysis, and that they will complete follow-up interviews. We could then measure things like employment, arrest history, taxes paid versus benefits consumed, family formation, educational attainment, and much else for each of the groups. Because they would have been randomly assigned, we could know that we had gold-standard evidence about the impact of the various selection methods for immigrants on these outcomes.
These would not be a complete picture of immigrant success, never mind a complete measurement of their impact on the total society. For example, it would be hard to measure things such as softer contributions to the community, or what the very long-term multigenerational differences between immigrants selected under one method versus another would be. Further, it would not necessarily be the case that what worked for immigrants from this list of countries in 2003 would be what would work for immigrants from other countries in 2014. Finally, there would not be complete unanimity about how to weight various outcomes. But it sure seems to me that it would be really useful to know whether, for example, skills-based immigration selection ends up bringing in immigrants with much lower crime rates, higher incomes, higher pay-in rates for social security, higher rates of participation in health insurance and lower utilization use of public welfare than does our current system. Or the reverse. The most likely outcome, of course, would be that some unanticipated combination of selection methods would turn out to work best.
This is not going to inform the debate about the law in 2013, but had we embedded this kind of testing in the original law, we would have ten-year measurements of the impacts of dozens of alternative methods for selecting immigrants. We could have observed the differences in results in the early tests, and by 2005 or so been trying to hybridize and improve various methods. I’m not that old, and I remember 1995. If we started now, within ten years, we could start to know this kind of thing.
In short, we could have useful engineering knowledge that would help us start to treat at least some parts of immigration policy as a practical method for improving our political economy.