Science has been on a mini-surge of publishing refugee related articles. The first that came across my desk was on the relationship between asylum applications and climate change. While I appreciated the spirit of it, I’m not the only one who thinks it’s seriously flawed, and was surprised it was accepted in a journal as highly ranked as Science. The other article that recently came out was on the use of algorithms and demographic data for selecting placements for refugees. This article was more compelling since it’s methodology is better, and I think it speaks to a policy problem in a direct way. It also provides people like me a relief against which to discuss ongoing political gaps in refugee placement and integration that need to be dealt with before a placement algorithm can really work.
So, how does this algorithm work? The gist of it starts with a lot of demographic data on refugees to the U.S. and Switzerland, and a lot of data on placement and job acquisition, and then matches it up. The algorithm would look at the attributes of a new refugee, and match that person with geographic regions where people with similar profiles had found work. Without going into the nitty-gritty math, if the algorithm had been used to place refugees from 2011-2016 the results indicate the job placement rate could have been up to 40% higher in the U.S. and 10% higher in Switzerland. This difference is huge for economic and social reasons.
Finding work is a key component of settling into a new country and culture. Self-sufficiency is a major benefit, and work is a key place where people develop social networks. There are also significant costs that come with not being able to work (either since you couldn’t find it, or were not allowed to work): when refugees aren’t allowed to work, even for a few months, it can make integration harder and is expensive for host countries. How expensive? Moritz Marbach and a team from the Stanford/ETH Zürich Immigration Policy Lab estimate that in Germany a seven month employment delay for refugees cost the German government 40 million Euros per year in social benefits, plus lost income tax revenue. The negative effects on economic and social integration that came with a seven month employment ban had a long horizon of the future too; refugees who had to wait out bans felt the effects as much as ten years down the road.
Unfortunately, an excellent algorithm and good economic data is only as useful as the political environment around refugee settlement allows it to be. In a perfect world, politicians would look at the German example, recognize how inefficient employment bans are, and then start using the Stanford team’s algorithm to place people more efficiently…if only politicians were so accepting of science. Indeed, the biggest hurdle facing humane, efficient refugee placement is often politicians who are happy to propagate falsehoods about refugees and migration, leading to a voting public that has a radically warped view of migration and refugee policy. Lenka Dražanová wrote an insightful post for the LSE EUROPP blog explaining the political and historical factors that have combined to make the Czech Republic hostile to refugees, even though migration and asylum seeking is, in reality, a very small issue there. The best algorithm in the world won’t help if people and politicians actively oppose the fundamental notion of granting asylum.
The gap between the Czech antipathy toward migration and refugees and the scientific precision available to help integrate newly arrived asylum seekers is a troubling microcosm of the populism-driven gap between the average voter and scientific expertise. When a political scenario where decency prevails emerges, and the body politic decides to deal with refugees on empirically and humanely grounded terms, it’s good to know that the science and statistics will be there to help things along.