Useful study. Interesting literature review. That may not sound like high praise, but it is.
You raise a hypothesis in the introduction, however, that neither your study nor the other ones you cited seems to address: namely that contact with immigrants or direct negative effects from immigrants might reduce support for immigration.
Anecdotes are dangerous, of course, but I lived through the rise in anti-immigrant sentiment in Brooklyn. (Not upscale Brooklyn, to be clear.) My next door neighbor became vociferously anti-immigrant when hotel rates spiked after the refugee influx, forcing him to put up his in-laws in his house when they came to visit. My wife's support for immigration dropped when she began to see little kids selling candy on the subway for the first time, well, ever. The owner of the neighborhood hardware store complained that contractors were hiring utterly unqualified Venezuelans; conversely, my morning coffee conversations at the corner bodega often featured mini-rants about how much cash was being showered upon them making them unwilling to work.
FWIW, everyone involved here is of Puerto Rican or Dominican descent, and the neighborhood swung hard to Trump in 2024. (Not that he got a majority, of course, just a 20-point swing.) In a sense, then, the negative reaction controlled for race and partially-controlled for culture, although I'm still just giving one anecdote.
The question, then, is how much results from surveys reflect how people react when they are directly confronted with effects they find unpleasant or distasteful. The strong effect you find on annual cost per taxpayer, even when presented abstractly and hypothetically, makes me worry that the survey effect won't actually travel to real public opinion.
If my worries give you an idea for future surveys or experiments, then I will consider this comment a success!
P.S. Your findings on linked fate among Latinos don't fit my priors, but I'll admit that I don't have great controls since my personal experience is mostly in New York and Florida. Assuming the result is correct, however, I worry about unobserved variable bias. For example, public education might be more likely to be biased towards immigrants or celebrate ethnic identities. Your finding on Latino-linkage v. immigrants in general leans against that particular hypothesis, but you might want to consider other more direct ways of addressing unobserved variables.
Hi Noel, thanks for your thoughtful comment! Your first point is alluding to the intergroup relations theories about contact theory and group threat – namely, does exposure to an out-group reduce animosities due to increased contact or does it antagonize conflict over limited resources. There’s a lot of work out there on this. One piece that comes to mind is about brief exposure to migrants in Greece increasing support for the far right (Dinas et al. 2019). The first suggested explanation I posit relates to this. The research I’ve done that I describe here does not measure exposure to migrants directly though. However, I have gone down this road for NYC in particular in trying to collect the exact hotel/shelter locations of migrants during the influx. I’ve submitted a few FOIL requests to the city for this information and the response is regularly deferred. And, even if I got the precise data, it is an imperfect proxy for contact as there is no guarantee that residents on the block actually saw or interacted with the migrants.
Your anecdote about the rise of anti-immigrant sentiment is well taken and is consistent with what I’d expect. The portability from a survey to the real world is something I acknowledge. Ideally, one would like to have pre and post attitudinal data that convincingly measures people’s exposure to migrants and tests whether the shock affects their attitudes. This requires panel data that happens to cover the right places and times, which is rare and expensive.
To your last point, the Latino immigrant linked fate paper includes models with and without education. It is a potential downstream variable though, as state policy can plausibly affect immigrants’ educational outcomes (not just educational level achieved, but also the content of the education), so the preferred specification I use in the main portion of the paper does not include it as a control. However, in the appendix I include it as a robustness check. The findings of that paper are indeed in tension with some related research on linked fate. I agree more direct tests are needed. I am working on a project that directly tests (to the extent that a survey can) how state policies can affect immigrant linked fate and political behavior.
If I may, there are two useful ways to deal with unobserved variable bias. The first one is to just run the unconstrained regression and then add in your controls one by one -- if your coefficient of interest doesn't change very much, then you're golden. I think Ted Miguel has a paper that gives you the rough rule of thumb for when a change becomes concerning; I can dig up the cite if it would be helpful. (Although I suspect that dumping the above paragraph into a talking computer will be enough for you to find it.)
The second more complicated way comes from Emily Oster. In her method, you choose artificial values for R2 and see if the coefficient flips sign or shrinks to some economically significant number. (That is, you're pretending that your controls explain more of the variation than they actually do -- which is another way of saying that you're assuming the unobserved variables are correlated with the observed ones. After all, if they're not then you don't usually have to worry about them!) If your results hold by the time you've pushed your fake R2 above 0.35, then you're also probably golden. There's a Stata package for it that I can dig up for you, although see the above paragraph about the utility of talking computers.
About contact theory, I can't remember, did you cite that famous Ryan Enos experiment on Boston commuter rail? I'd imagine the immigration studies union requires it, the way economic historians had to cite Douglas North at some point. You'd get disinvited to cocktail parties if you didn't. More seriously, in addition to his book, there's also Abrajano and Hajnal on the backlash to immigration.
One thing about migrant contact is that you don't have to have direct contact with migrants to be negatively affected by them. For example, you might live on the block with them, but see them only as poor suffering compatriots, or even feel good about your community has opened its hearts. Meanwhile, some contractor over in the unfashionable eastern part of Bushwick is seriously pissed off that his mother-in-law is now sleeping in his kids' bedroom. (That is to say, Irvin, my neighbor.) But I guess that's why studying this stuff is hard!
Thank you for taking the time to explain but yes I’m familiar. The paper uses and reports the results across different varieties of models and measurement choices.
The essay and the research papers the essay is based on do not include Enos as the focus of the work I describe is on how policy bundles affect policy support and perceptions of fairness. The hope is that better designed policies that align with the public’s perceptions of fairness could reduce backlash.
You’re right about migrant contact and why those studies can be tricky.
It would be very helpful to understand how educational migration can play a more prominent role in integration.
Similarly, it would be helpful to understand if a generous student lending program can fund services to balance costs. Then that allows a higher bar over time. We understand you don't speak Dutch/German/Japanese when you arrive but we will lend you money for classes and we expect you to have language acquisition by year 5. Similarly, we understand you don't have a degree now but we lend you money for certain programs and after 10 years we expect you are over a high bar.
I genuinely don't know how punitive left wing people view lending or how generous right wing people view it. Or similarly how punitive a left wing person views refusing a visa renewal if someone can't learn the language in 5 years.
Sorry that I am asking such a specific question! But I genuinely really appreciate the scholarship and want to learn more.
Hi Richard, thanks for your question. The rules that earned a "premium" in my studies were rules about access that citizens already face. Proof of employment, which is closer to an “earned” condition that you're describing, had no detectable effect once cost was part of the experiment. What you describe speaks to the investment dimension of fairness (rewarding contribution) and I'd expect it would not gain as much support as universal, civic conditions do, at least based on my two studies and the US context. I'll also admit I expected the "earned condition" to play a larger role in the results than it did.
On your partisan comment, my guess is that many on the right would find a loan easier to accept than a grant because repayment signals reciprocity. Republicans penalized direct cash payments in my first study. Many on the left would likely see debt plus a residency deadline as punitive, especially for refugees or older arrivals. That’s just my speculation though. It’d be a good experiment.
Last year a warehouse team lead at our Dublin depot spent 14 months in sponsorship limbo; after a new compliance checklist landed, the same process took five. The loudest queue-jumper complainers turned out to be the first to admit the wait felt unfair. Your +11-point background-check bump mirrors what I saw at the depot. Would the effect survive once rules actually get implemented?
Useful study. Interesting literature review. That may not sound like high praise, but it is.
You raise a hypothesis in the introduction, however, that neither your study nor the other ones you cited seems to address: namely that contact with immigrants or direct negative effects from immigrants might reduce support for immigration.
Anecdotes are dangerous, of course, but I lived through the rise in anti-immigrant sentiment in Brooklyn. (Not upscale Brooklyn, to be clear.) My next door neighbor became vociferously anti-immigrant when hotel rates spiked after the refugee influx, forcing him to put up his in-laws in his house when they came to visit. My wife's support for immigration dropped when she began to see little kids selling candy on the subway for the first time, well, ever. The owner of the neighborhood hardware store complained that contractors were hiring utterly unqualified Venezuelans; conversely, my morning coffee conversations at the corner bodega often featured mini-rants about how much cash was being showered upon them making them unwilling to work.
FWIW, everyone involved here is of Puerto Rican or Dominican descent, and the neighborhood swung hard to Trump in 2024. (Not that he got a majority, of course, just a 20-point swing.) In a sense, then, the negative reaction controlled for race and partially-controlled for culture, although I'm still just giving one anecdote.
The question, then, is how much results from surveys reflect how people react when they are directly confronted with effects they find unpleasant or distasteful. The strong effect you find on annual cost per taxpayer, even when presented abstractly and hypothetically, makes me worry that the survey effect won't actually travel to real public opinion.
If my worries give you an idea for future surveys or experiments, then I will consider this comment a success!
P.S. Your findings on linked fate among Latinos don't fit my priors, but I'll admit that I don't have great controls since my personal experience is mostly in New York and Florida. Assuming the result is correct, however, I worry about unobserved variable bias. For example, public education might be more likely to be biased towards immigrants or celebrate ethnic identities. Your finding on Latino-linkage v. immigrants in general leans against that particular hypothesis, but you might want to consider other more direct ways of addressing unobserved variables.
Hi Noel, thanks for your thoughtful comment! Your first point is alluding to the intergroup relations theories about contact theory and group threat – namely, does exposure to an out-group reduce animosities due to increased contact or does it antagonize conflict over limited resources. There’s a lot of work out there on this. One piece that comes to mind is about brief exposure to migrants in Greece increasing support for the far right (Dinas et al. 2019). The first suggested explanation I posit relates to this. The research I’ve done that I describe here does not measure exposure to migrants directly though. However, I have gone down this road for NYC in particular in trying to collect the exact hotel/shelter locations of migrants during the influx. I’ve submitted a few FOIL requests to the city for this information and the response is regularly deferred. And, even if I got the precise data, it is an imperfect proxy for contact as there is no guarantee that residents on the block actually saw or interacted with the migrants.
Your anecdote about the rise of anti-immigrant sentiment is well taken and is consistent with what I’d expect. The portability from a survey to the real world is something I acknowledge. Ideally, one would like to have pre and post attitudinal data that convincingly measures people’s exposure to migrants and tests whether the shock affects their attitudes. This requires panel data that happens to cover the right places and times, which is rare and expensive.
To your last point, the Latino immigrant linked fate paper includes models with and without education. It is a potential downstream variable though, as state policy can plausibly affect immigrants’ educational outcomes (not just educational level achieved, but also the content of the education), so the preferred specification I use in the main portion of the paper does not include it as a control. However, in the appendix I include it as a robustness check. The findings of that paper are indeed in tension with some related research on linked fate. I agree more direct tests are needed. I am working on a project that directly tests (to the extent that a survey can) how state policies can affect immigrant linked fate and political behavior.
If I may, there are two useful ways to deal with unobserved variable bias. The first one is to just run the unconstrained regression and then add in your controls one by one -- if your coefficient of interest doesn't change very much, then you're golden. I think Ted Miguel has a paper that gives you the rough rule of thumb for when a change becomes concerning; I can dig up the cite if it would be helpful. (Although I suspect that dumping the above paragraph into a talking computer will be enough for you to find it.)
The second more complicated way comes from Emily Oster. In her method, you choose artificial values for R2 and see if the coefficient flips sign or shrinks to some economically significant number. (That is, you're pretending that your controls explain more of the variation than they actually do -- which is another way of saying that you're assuming the unobserved variables are correlated with the observed ones. After all, if they're not then you don't usually have to worry about them!) If your results hold by the time you've pushed your fake R2 above 0.35, then you're also probably golden. There's a Stata package for it that I can dig up for you, although see the above paragraph about the utility of talking computers.
About contact theory, I can't remember, did you cite that famous Ryan Enos experiment on Boston commuter rail? I'd imagine the immigration studies union requires it, the way economic historians had to cite Douglas North at some point. You'd get disinvited to cocktail parties if you didn't. More seriously, in addition to his book, there's also Abrajano and Hajnal on the backlash to immigration.
One thing about migrant contact is that you don't have to have direct contact with migrants to be negatively affected by them. For example, you might live on the block with them, but see them only as poor suffering compatriots, or even feel good about your community has opened its hearts. Meanwhile, some contractor over in the unfashionable eastern part of Bushwick is seriously pissed off that his mother-in-law is now sleeping in his kids' bedroom. (That is to say, Irvin, my neighbor.) But I guess that's why studying this stuff is hard!
Thank you for taking the time to explain but yes I’m familiar. The paper uses and reports the results across different varieties of models and measurement choices.
The essay and the research papers the essay is based on do not include Enos as the focus of the work I describe is on how policy bundles affect policy support and perceptions of fairness. The hope is that better designed policies that align with the public’s perceptions of fairness could reduce backlash.
You’re right about migrant contact and why those studies can be tricky.
Great article!
It would be very helpful to understand how educational migration can play a more prominent role in integration.
Similarly, it would be helpful to understand if a generous student lending program can fund services to balance costs. Then that allows a higher bar over time. We understand you don't speak Dutch/German/Japanese when you arrive but we will lend you money for classes and we expect you to have language acquisition by year 5. Similarly, we understand you don't have a degree now but we lend you money for certain programs and after 10 years we expect you are over a high bar.
I genuinely don't know how punitive left wing people view lending or how generous right wing people view it. Or similarly how punitive a left wing person views refusing a visa renewal if someone can't learn the language in 5 years.
Sorry that I am asking such a specific question! But I genuinely really appreciate the scholarship and want to learn more.
Hi Richard, thanks for your question. The rules that earned a "premium" in my studies were rules about access that citizens already face. Proof of employment, which is closer to an “earned” condition that you're describing, had no detectable effect once cost was part of the experiment. What you describe speaks to the investment dimension of fairness (rewarding contribution) and I'd expect it would not gain as much support as universal, civic conditions do, at least based on my two studies and the US context. I'll also admit I expected the "earned condition" to play a larger role in the results than it did.
On your partisan comment, my guess is that many on the right would find a loan easier to accept than a grant because repayment signals reciprocity. Republicans penalized direct cash payments in my first study. Many on the left would likely see debt plus a residency deadline as punitive, especially for refugees or older arrivals. That’s just my speculation though. It’d be a good experiment.
Thanks again for your engagement!
Last year a warehouse team lead at our Dublin depot spent 14 months in sponsorship limbo; after a new compliance checklist landed, the same process took five. The loudest queue-jumper complainers turned out to be the first to admit the wait felt unfair. Your +11-point background-check bump mirrors what I saw at the depot. Would the effect survive once rules actually get implemented?