Published on Development Impact

Can information reduce anti-immigration biases?

This page in:
Also available in: Français

Let’s start with a little quiz. Grab a piece of paper and pencil.   What’s the share of legal immigrants in the US population? (or you can choose the Germany, UK, Italy, Sweden or France).  A legal immigrant is defined as someone living legally in the country and born abroad. 
OK.  If the suspense is killing you on the answer, you can skip to somewhere below.   But if not, this question, and why so many people get it wrong, how it relates to information, and if information can offset it, are at the heart of a fascinating new paper by Alberto Alesina, Armando Miano, and Stefanie Stantcheva.  
Alesina and co. run a large scale set of surveys covering 22006 respondents in the six countries listed above during January to March of this year.   The surveys were run online and designed to be nationally representative.   In addition to the usual socio-economic characteristics, the survey focused on two main things.    First, it asked a set of questions about immigration.    These covered the question above, but also a range of factual questions about immigrants in the country where the survey was taking place: where do they come from, what religion are they, their employment levels, their education level, the share living below the poverty line, and the average (government) transfer that they get relative to someone born in that country.    Then they ask a set of attitudinal/perception questions about whether immigrants are poor because of lack of effort or circumstances beyond their control, a question where an immigrant (Mohammed) is compared to a native (John) in terms of paying taxes and receiving benefits, and then a set of questions on immigration policy.  Taken together these questions form what Alesina and co. call the immigration block.  
The second set of questions are about redistribution.  Here Alesina and co. work hard to separate out views on the the total size of government from how to raise funds (taxes) and how to spend them.   They also set up a mechanism to incentivize meaningful answers: respondents are told they have been entered into a lottery to win $1000, and before they can find out if they win, they have to decide how much of it (zero is an option) to donate to charity, where the charity is about low-income folks but not immigrants in particular.   Taken together, these questions form the redistribution block.  
Before getting to the results, one interesting methodological note.  Before they ask the immigration questions, Alesina and co. ask the respondents whether they have paid careful attention to the preceding questions and whether the responses should be used in the analysis.  Apparently, this is a tool from a paper by Meade and Craig (2012) and the answer doesn’t matter (most people say yes) but it gets people to pay more attention to the questions that follow. 
OK, so how do people respond?   To start with: the fraction of the population that are immigrants are as follows 10 percent (US), 13.4 (UK), 12.2 (France), 10 (Italy), 14.8 (Germany), and 17.6 (Sweden).   And people get this massively wrong – for example in the US the respondents think that the share is 36.1 percent (so don’t feel bad if you were off).   The Swedes do the best, but they’re still coming at 10 percentage points too high. 
There is some heterogeneity in how wrong people are, but as Alesina and co. note, it is important to note that no single group (taken across countries) actually gets closer than 15 percentage points to the actual estimate.   So who does worse?   People who have low education and also work in sectors with more immigrants, those without a college education, those who have an immigrant parent, the young and women.    Interestingly (at least for me with an immigrant parent), the folks with an immigrant parent have the largest over estimate of the share of population that are immigrants.   Also interesting is the fact that the political orientation of respondents does not correlate with how much they overestimate the share of immigrants.  
People are also pretty bad at guessing where the immigrants in their countries come from.  They are much more likely to say the Middle East, North African or sub-Saharan Africa than is actually the case (although there is a bit of cross-country variation in which of these three regions they overestimate for).   Folks (except the French) are also significantly likely to overestimate the fraction of immigrants who are Muslim.    The US does this the most.   On the flip side, folks significantly underestimate the fraction that are Christians.   Again, the misperceptions are higher among those without a college degree, women, and the less educated working in a high immigration sector.   In contrast to the fraction of the population question, here the older respondents and those who identify with right-wing political parties get it more wrong.  
Respondents are also more likely to believe that immigrants are poorer than they are (except for Sweden) and they also overestimate immigrant unemployment (even in Sweden).   These latter numbers are large – in the US it is overestimated by 25 percentage points, and by 35 percentage points in Italy.  Respondents also overestimate the share of low educated immigrants (not finishing high school) in all countries except Germany.    The heterogeneity in views here is similar to the region of origin of migrants.   In terms of saying whether migrants are poor because of a lack of effort, in three countries (the US, UK and Sweden), respondents pretty much have the same views of migrants and natives.   However, in Italy and France, respondents see this lack of effort as significantly more likely to cause poverty for migrants than natives.  
Actually knowing an immigrant is associated with significantly lower misperceptions across all measures, even when controlling for all individual characteristics.   It’s also associated with different views – these folks are less likely to say migrants are poor because of a lack of effort.  In the US, where the data will allow for this, Alesina and co. also look at whether or not the density of migrants in your commuter zone matters.   They find that folks in high migrant commuter zones have a higher degree of misperception about the share of migrants in the population.    So it seems it’s the personal connection which helps one get better information.  
In terms of attitudes towards immigration (e.g. is immigration a problem, when should immigrants get citizenship, etc) Alesina and co. find that the US is the country that is most supportive of immigrants, with France, German and Italy coming in as the least supportive.   There is some heterogeneity by respondent characteristics here as well: the left-wing folks have the most favorable attitudes among any group and the right-wing folks the least. 
Moving on to attitudes on redistribution, Alesina and co. find that support for immigration and redistribution are highly correlated (including conditioning on individual characteristics).   And this will be important in the neat experiments they run with their survey sample.    These are two: 1) they randomize the order of questions, showing some folks the redistribution question block first, others the immigration block first, 2) giving respondents a (randomized) information treatment.   This information treatment takes three forms: a) information on the actual share of immigrants in their country (with benchmarking within the OECD –who knew Switzerland had the highest share at just over 29 percent?), b) data on the origin of immigrants, and c) a vignette describing how hard an immigrant works (based on a composite of actual people).  
So, what do they find? 
First off, the order of questions matters.    The folks who get the immigration questions first are less excited about redistribution.   They are also less likely to think inequality is a big problem, and they donate less to charity (under the lottery mechanism described above).   Alesina and co. show us that the magnitude of these effects are meaningful.   They also show us that these effects are particularly pronounced for those affiliated with right-wing parties, non-college educated, and the low educated folks working in high immigrant sectors.  
The information treatments have the effects they are intended to (even if they don’t completely fix misconceptions).   Folks who get information on the share of immigrants end up 5 percentage points closer to the actual number.    The origin of immigrants information results in a big decline (42% relative to the control) in the reported share of migrants from the Middle East and North Africa, and a smaller, but meaningful decline in the over-reporting of the share of immigrants who are Muslim.  And the story of the hard-working immigrant results in a 5 percentage point drop (14% relative to the control group) in the fraction of respondents who say immigrants are poor because they don’t work hard.  
Does this information stick?   In the US, Alesina and co. do a second, follow-up survey a couple of weeks after the first one (and the information treatments).   They find that the hard work story and the origins of immigrants effects are persistent, but that the share of migrants isn’t.   They chalk this last (non)effect up to the fact that remembering an exact number is harder (10 percent in the US, in case you forgot).
Do these information treatments have any other effects beyond the directly related questions?   Information on the share of immigrants and the story of hard work significantly increase support for immigration.   And the story of hard work also moved the needle on redistribution – folks who got this information scored significantly higher (i.e. more in favor) on a redistribution index.   But, alas, there are two bits of grim news.   First, these effects are somewhat weaker for those who have less favorable views on immigration (e.g. the low educated folks in higher immigrant sectors of the economy).  Second, on average, the positive effects from the information treatments aren’t strong enough to offset the negative effects of shifting the order of questions. So, information could help a bit, but it isn’t going to overcome the (negative) priming effects of reminding people about immigration before you ask about redistribution.  
OK, this has been a fairly long post.    Mostly that’s because this is a super interesting paper – there are a significant number of results and a nice literature review and theoretical discussion that aren’t included here.   And so, I urge you to take a look at the paper.   And then go out and meet an immigrant, if you don’t know one already


Markus Goldstein

Lead Economist, Africa Gender Innovation Lab and Chief Economists Office

Join the Conversation

The content of this field is kept private and will not be shown publicly
Remaining characters: 1000