Who are Europe's immigration worriers?

A contemporary sociological profile of citizens with anti-immigrant attitudes

Look, I’m open to being wrong. So correct me if I am missing something. And if it helps you feel better, my initial intuitions (before looking at the research or the data) back in 2015, 2016 were wholly materialist and not really politically correct.

That is, when the whole “migrant crisis” thing happened in Europe around a decade ago, my immediate reaction was: people are typically rational actors, so if they seem to be sperging out about immigration, let’s try and understand what about their objective situation is making them do that. I thought it likely that though people dress their complaints in vulgar and sometimes racist rhetoric, which my tribe (the left) rightly denounces, there must be substantial underlying truths to what they’re saying. Yes, they are expressing their fears in an uncouth manner, but rational actors that they are, they’re roughly just pursuing their material interests and channeling their veridical experience of objective reality.

So, I thought, it’s very likely that the most rabid anti-immigrant natives (who help elect authoritarians) are, almost without exception, those who:

1. actually have the most to fear from the migrants economically, owing to their (i.e., the natives’) objective economic location, which pits them against each other;

2. live in now-crowded places where they’re constantly bumping up against the migrants, some of whom really are bad people (as goes for any group), which is what the natives are correctly experiencing in their personal daily lives (even if they end up overemphasizing and overgeneralizing);

3. live in abandoned, left-behind, rust-belt regions that are falling apart due to globalization/neoliberalism and are at the same time filled with migrants, who then make for a predictable (if not really justified) scapegoat, given how human nature works. People correctly see things falling apart around them, they correctly see an influx of a new group, and they blame it on them or just generally lash out at them because they, instead of neoliberalism, are an easy target (no propaganda from the ruling politicians necessary).

That was just based off of my gut feeling and general materialist inclinations I had/have. Then, some years later, I read books like Caplan’s The Myth of the Rational Voter, Somin’s Democracy and Political Ignorance, and Brennan’s Against Democracy. Those showed me how, precisely because the (sophisticated) rational choice model is approximately true, one shouldn’t expect voters to accurately perceive complicated and abstract social reality. And the data bear out significant voter ignorance even about basic political facts.

Relatedly, one shouldn’t think of voters in straightforward, intuitive egocentric or pocketbook ways – sociotropic voting is more likely. If the former were true, voters would make decisions based on their personal circumstances, which they’re more likely to be right about (since they know their immediate lives and finances relatively well). If the latter is true, voters make decisions on what they feel is happening to (and is good for) the broader society, regardless of their own objective individual position. Unfortunately, they’re far more likely to be mistaken about this broader picture, because they don’t get to experience “the society” directly, unlike their own personal, immediate lives. So, they might quite likely make biased and misinformed sociotropic decisions based on remote and probably unmoored vibes they get from “the uncle,” or media or fake news. After all, the average voter is not reading up on statistics of immigrant crime or studies about the likely effects immigration has on the labor market.

More importantly, and theory aside, I simply looked at the relevant research and data (specifically for Europe, which is often contrasted to the US). What I found out was that my materialist hypotheses under (1), (2), and (3) above are mostly false or at least not corroborated.

Here are my results from several multivariate models using the latest European Social Survey (ESS) data for both individuals and regions across many European societies.¹ Let me take you through them.

A quick note on how I measure “anti-immigrant” sentiment before we begin. I follow the existing literature and create two indexes, named “Impact” and “Admission” in the coefficient plot above. They’re both made up of several standard immigration questions from the ESS. The first one measures what people think immigration does to society; the latter measures people’s opinions on how open or closed to immigrants the society should be.

And though I employ many controls (alongside country-fixed effects), this is all observational. Can’t read off causality here, this is just descriptive.

Regions with more immigrants

If immigrants are objectively bad for social life in Europe and if natives are correctly perceiving that badness, you’d expect natives in regions with higher immigrant share to have more anti-immigrant attitudes, right? But we don’t see that. In fact, natives in regions with more migrants are less anti-immigration in their attitudes.

You might respond: “Sure, but that’s just because immigrants tend to sort themselves into wealthier regions where life is better off for everyone.” Well, of course, but I obviously controlled for regional GDP per capita. Once wealth/development is controlled for, natives who have to live in places with more immigrants still exhibit lower anti-immigrant sentiment.

Weird that precisely those people who don’t have to deal with as many (ostensibly problematic) immigrants, and whose regions haven’t yet been touched by immigrant presence, have stronger anti-immigrant attitudes. They don’t even know what immigrants are like! It’s less weird, however, if you abandon the “citizen as objective perceiver” hypothesis.

Income deciles

It’s the same with household income. There’s just no effect. It isn’t true to say that poorer Europeans, who have more to fear (both culturally and economically) from low-skilled immigrants, are any more anti-immigrant in their attitudes than their wealthier counterparts. Whether I introduce income as a continuous variable that goes from the bottom to the top deciles by household income or whether I pit, say, 2nd decile against 8th, or 5th decile against the top, or whether I do a spline check, or whether… There’s just nothing there. Objective economic location, which should be relevant if people form their immigration fears objectively, reveals nothing.

Think about it. Poor people (in contrast to wealthy people) have no choice but to use public transportation – just like the immigrants. Or at least, both these groups are much more likely to use it and thus be in contact with each other. Wealthier people are “shielded” from encountering immigrants on their way to work and just generally, given that they live in gated communities or upscale districts. Poor people also don’t tend to employ (immigrant) maids, as their wealthier counterparts do. So there should be a clear gradient on the objective-perception hypothesis.

Poor people should be more anti-immigrant, because they don’t gain much from immigrants and might even be economically or otherwise hurt by them. Wealthy people should be much more pro-immigrant, because they don’t compete with them in the labor market, don’t encounter them as much on the street, and if anything, use them to make their life more comfortable at home.

But we see none of that reflected in the distribution of anti-immigrant sentiment. If voter perceptions don’t reflect objective reality (at least with regard to migrants), that makes more sense.

The unemployed, the left behind regions, and the low skilled workers

Are the unemployed any more or less fearful of immigrants than other people? No, there’s nothing here either. What about regional-level unemployment and thus sociotropic concerns about “the left behind” regions? Nada.

(In contrast to regional-level unemployment, regional-level GDP is statistically significant in some models and has a coefficient size of around 0.14. That’s worth mentioning and lends some modest and mixed support to the sociotropic left-behind thesis. Note though that the result isn’t robust in the coefficient plot above; it’s also non-significant under an additional, larger sample specification I ran.)

What about occupations? No, not really. Skilled manual workers are no more anti-immigrant than workers with elementary occupations. Pitting skilled manual workers vs. service/clerical jobs is likewise non-significant (as is the contrast between service/clerical and elementary occupations). Given that some of these are much more exposed to labor market competition with low-skilled immigrants, the objective-perceiver hypothesis is again in a bit of a pickle.

It’s mostly subjective impressions and ideology

But there must be some things that explain difference in immigrant attitudes between people in Europe? Sure. There’s a correlation with perceived, subjective income strain. Those claiming to have difficulties with their income are somewhat more anti-immigrant. Funny, how objective indicators show virtually nothing, but then when you think you’re doing badly, all of a sudden you are a bit more anti-immigrant in your attitudes. I’m not saying people are necessarily wrong in their personal finance perceptions; I actually don’t think that’s likely. But even if there’s something real here, the correlation is vanishingly small: around 0.07.

What else is there?

Well, ahem, being educated makes you less anti-immigrant in attitude (controlling for a host of confounders, obviously). But people don’t like hearing about that one, so let’s quickly pass it by. Being a right-winger makes you much, much more anti-immigrant in attitude (a correlation of almost 0.3, after controlling for loads of confounders). The same goes for people who don’t trust others. They’re much more likely to be anti-immigrant in sentiment. People from rural areas are more anti-immigrant (funny, given that immigrants cluster in the cities).

Overall, I’m not impressed. Now, before you get mad, I’m sure anti-immigrant attitudes in Europe are not wholly baseless. And it’s most definitely true, as a raw fact, that in many European countries, immigrants are overrepresented in prison populations. We have to be honest about that. (It’s a completely different question whether natives actually know a statistical fact like that and use it to form their anti-immigrant attitudes; given how widespread voter ignorance is, I wouldn’t bet on it.) I know about cultural persistence and Jones’ The Culture Transplant (see here, here, and here).

I’m not saying there is nothing to anti-immigrant attitudes or that we shouldn’t be thinking hard about immigration policy. But I just can’t buy the whole “natives are wholly or mostly objective perceivers of immigrants and social facts around immigrants, hence they’re being correctly anti-immigrant in attitude, and, oh my god, why are you snobby bourgeois elites not listening to them!” narrative. At the very least, it seems to be much, much more complicated than that.

What about other research?

Is it just me who finds these correlates? Nope, the broader recent European literature shows roughly the same. Once you move beyond bivariate storytelling and actually throw a decent multivariate model at the question, the big, stable predictors of anti-immigrant sentiment are not usually objective economic position. Just as in my case, they’re things like education, ideology, and generalized social trust. Education keeps (statistically, not necessarily causally) pushing people in a more pro-immigration direction; right-coded ideology goes the other way; low-trust people are reliably more anti-immigrant.

The literature also finds that natives living in regions with a larger foreign-born population are generally less anti-immigrant, not more. There is one important caveat that I have to mention here though. The literature distinguishes between stocks and flows, so between regions with a high existing immigrant share and regions experiencing a short-term increase. The former often goes with less hostility (like in my case), but the latter can go with more. People can be unhappy with sudden changes, whatever the change may be. For what it’s worth, I myself have looked at that in a previous post with a synthetic diff-in-diffs event study and found nothing (at least as far as support for the anti-immigration hard right in elections goes).

Lastly, just like my analysis, the recent literature doesn’t give much comfort to the idea that poorer people, the unemployed, or those in more exposed labor-market positions are consistently more anti-immigrant (once other things are held constant). You do sometimes see something for subjective economic strain, dissatisfaction, or felt insecurity. But that is precisely my point. The “objective” variables are typically weak, while the subjective and ideological ones turn up as significant (though some of them, like income strain, are quite small in terms of size).

1. The models use a common-support sample size of around 14,000 respondents, more than 100 regions, and 12 countries. If I drop regional-level unemployment as a variable, I can almost double the sample size to 21 countries and 24k respondents. The results remain the same. If I drop all regional-level variables, the sample size goes up to 28 countries and more than 30k respondents. Again, the results remain the same. Everything is estimated with native-citizen samples only.