Donald Trump has focused on the immigration issue more and restricted unauthorized immigration more severely than any other President in recent history.
A hypothetical: if Trump and his people understood that the median voter specifically hates illegal immigration but likes skilled immigration, would they have done the actions you flag as unpopular? Eg the recent green card announcement
For what it’s worth, sometimes it seems like Trump does basically agree with the median voter and wants zero illegal immigration and plenty of legal skilled immigration. But seems to spend a lot on who he talked to most recently
It seems to me that populists in general, and Trump in particular, have a good intuitive understanding of what is popular with many people. However, their understanding is from perfect. This may partly explain why their policymaking does not appear, at least to me, precisely to the point. Instead, it appears broad brushed, combining a lot of popular things with unpopular policies. Of course, there may be different explanations for this too.
On a positive note, this suggests that there is much room for improvement here by, being more to the point, for people who want to enter politics.
People aren’t allowed to say they are against “skilled” immigration, but I don’t see a lot of people bemoaning the massive overhaul of the H1b program.
The best test case here is Canada. It went all in on recruiting “skilled” Indian immigrants. It’s been a disaster for Canada and they abandoned it.
From what I can tell the backlash in Canada was more about fast food workers, students, and a sense that immigration was out of control. Not skilled workers per se
People definitely don’t want out of control immigration, but I don’t see the evidence that people don’t like controlled immigration of high skilled people
Students have degrees don't they (hence "skilled"). The pro-H1B crowd wanted to "staple a green card to every degree". And Canada's immigration policy was very similar to the H1B. Its salary minimum was exactly the same as the American H1B. I would regard Canada's experiment as "what if we took the existing H1B system and 10x its per capita rate."
What this really comes down to is what constitutes "skilled". If skilled means top 1% global stem talent, then obviously we want more skilled immigration. I would pay people in that tier a million dollars to immigrate.
If "skilled" means mediocre white collar workers from India, then I'm not signed up.
I think Canada proves that mediocre tier white collar workers are not the bottleneck on economic growth. If anything when you add significant cultural differences they just gum up the works.
So yes, the new H1B rules are good because they increased the skill level of the applicants. But remember that it took a civil war for that to happen. Elon Musk and the rest of the "tech right" said they are going to "rip the faces off" anyone that tried to change the H1B program. They wanted to replicate Canada's poor experience. They lost the debate and we are all better off for it.
HB-1s aren’t all skilled workers and the skilled ones aren’t better than Americans. Now with AI, we don’t need so many. Real geniuses and top tier workers can get an O visa or something sillier. Hb-1 needs to die and be replaced by a visa that benefits Americans.l, not just companies and Indians who have cornered tbr market and not because they’re so smart.
Thanks, Laurenz! I agree with the general thrust that "[Trump's] immigration policy is comparatively popular," which is something important for smart people to finally recognize.
That said, I don't think the relative increase/decrease Gallup item is a good way to diagnose the representation gap. It mixes up too many things at once: what people actually want from immigration, what politicians in power and opposition are doing, and thermostatic reactions to current events. A better existing measure would be voters' approval/trust for both parties on immigration, which gets closer to whether anyone actually feels represented.
I'd also push back on the broader read of Trump's "immigration policy." He (or Stephen Miller?) did a bunch of popular things (asylum restrictions, border security) alongside a bunch of unpopular ones (legal green card cuts, travel bans, high-skilled immigration cuts). Bundling them all together as "Trump's immigration policy" obscures more than it elucidates.
In particular, thanks for suggesting another existing measure that addresses some of the shortcomings that you criticise (of course, all these measures are imperfect, but it's better to have an imperfect measure than none at all...).
A brief analysis suggests that during Trump-2, Americans think Republicans handle immigration better than Democrats (even though the Republican advantage has decreased strongly recently, it's still there), which is consistent with my main finding here. But this masks that both parties are unpopular on immigration.
The reason I did not use this measure is that it compares the actual policy of Republicans, who are in power, to some hypothetical policy (proposals or whatever people imagine) of Democrats. This asymmetry can bias their comparison. In addition, I think this measure also mixes many of the things you rightly say are mixed up in the item I use. That said, my post would have benefited from using more items, including relative approval.
A problem with this analysis is you ignore the pre-1995 data. If you include that data you will find the net pressure to reduce immigration in 1965 (when the 1924 immigration restrictions were still in place and rates commensurably low) was higher than it has been since 2010. The public has *always* been in favor reduced immigration, even when it had been reduced. One might subtract this baseline level from the values in Figure. Here you would find that even with the adjustment, people were upset about immigration and wanted to reduced, leading to the Secure Fence Act under the Bush administration, but by the end of the Bush administration and afterward this has not been the case *except* for the end of the Bush administration when it became publicly well know just how many people were pouring into the country across the Southern Border (there was a Sixty Minutes segment on this in 2024). This explains why Tom Tancredo's immigration restrictionist campaign in 2008 went nowhere and why there was pushback from other Republicans on Trumps centering on this issue in 2015. They did not see that there was a constituency for this. And there wasn't. Trump's success in 2016 was based on his skill as a personal marketer and brand creator.
I doubt any other outsider running on the same issue mix could have snagged the nomination as easily as Trump did.
I agree that the analysis would profit from a longer time horizon. I ignore pre-1995 here on purpose because in this period coverage of the item is use is more sparse and for some presidents pre 1995 there is no coverage at all.
In general, the fact that people were still anti-immigration even after immigration had been reduced somewhat may be explained by the reduction simply being too small or the underlying immigration preferences having changed. Possibly, people in the 1960s were fine with much lower levels of immigration than people nowadays, but it is difficult to test and I am unsure.
In general, I agree this data shows, as described in the text, that part of the success of Trump-1 is probably his marketing. But, generally speaking, how concerned people are also lines up really well with the actual objective inflows of undocumented immigrants, so there may also be a large rational component (and Trump-2 did deliver).
I guess my main message for politicians would be to take this rational part serious and do something about it. Thinking that politics is purely marketing is extremely dangerous because it provides a justification to just ignore all problems and purely focus on talking them down. Even though public opinion does diverge from the objective truth for some time on some issues there is usually a rational core to each issue policymakers can address.
This border crossing data (i.e. the data your link points to, NOT the data I use) is very unreliable because most border crossings are not observed and the number observed is endogenous to policy decisions. I think regarding illegal immigration I use the best (different) data for all periods it exists.
I dont use border crossings. The measures I use are clean and for these measures I use all data that exists. This does lead to different durations for the time series but it is, to my best knowledge, the best we can do with the existing data. As scientists we are unfortunately limited by data availability.
I see how this can look cherry-picked at first but using all clean data that exists is not cherry picking. Thanks for raising that point I hope that makes it clearer.
I found the data set you used. It is better in that in has more detail. I plotted the encounters data set from the CBD against the entering data in that set to see how good the correlation was. The r was 0.988, which is pretty good.
One has to weigh the detailed accuracy of the data one uses against the loss of information one loses by not considering a sufficiently long period of time.
By using only the 2013-2025 period one is considering only a single cycle: Obama-Trump I (low), Biden (high) and Trump II (low) By using lower-quality data (which still captures the trends that you care about) back to the 1990’s (for which you have reasonably dense Gallup data) you get the Clinton-Bush peak, capturing two cycles of immigration dynamics, increasing the power of your analysis.
My point was even when the percent foreign born was at very low levels (4.7%) the polls showed support for lower immigration was 27 points higher than support for higher immigration. I like your increase-decrease metric for immigration feeling. I tried regressing these against immigration rate per 1000 population data that I already had. I got a moderate correlation (r = 0.37) with the 2022 and 2023 points being way out line. If I take these out, I get a better correlation (r = 0.57). Now the 1965 value looks like an outlier, so I took that out and got a pretty good correlation (r = 0.66) with the remaining 26 data points. It suggests that immigration feeling would be neutral were immigration rate at 0.25%.
When I tried to do a regression using percent foreign born it was a nonsense correlation (negative slope) that is the higher percent foreign born the lesser the objection to immigration.
A hypothetical: if Trump and his people understood that the median voter specifically hates illegal immigration but likes skilled immigration, would they have done the actions you flag as unpopular? Eg the recent green card announcement
For what it’s worth, sometimes it seems like Trump does basically agree with the median voter and wants zero illegal immigration and plenty of legal skilled immigration. But seems to spend a lot on who he talked to most recently
I agree.
It seems to me that populists in general, and Trump in particular, have a good intuitive understanding of what is popular with many people. However, their understanding is from perfect. This may partly explain why their policymaking does not appear, at least to me, precisely to the point. Instead, it appears broad brushed, combining a lot of popular things with unpopular policies. Of course, there may be different explanations for this too.
On a positive note, this suggests that there is much room for improvement here by, being more to the point, for people who want to enter politics.
Makes sense!
Very thought provoking - I've written up my reactions to this piece in a bit more detail here. Would be very curious your feedback/thoughts: https://lukeeure.substack.com/p/the-bouncer-and-the-recruiter
People aren’t allowed to say they are against “skilled” immigration, but I don’t see a lot of people bemoaning the massive overhaul of the H1b program.
The best test case here is Canada. It went all in on recruiting “skilled” Indian immigrants. It’s been a disaster for Canada and they abandoned it.
Agreed there has been relatively little backlash to the H-1B program - that’s because it’s a pretty good policy! On net it increases the skill level of immigrants coming to US companies https://lukeeure.substack.com/p/trumps-h-1b-lottery-policy-kind-of?r=bx8en&utm_medium=ios
From what I can tell the backlash in Canada was more about fast food workers, students, and a sense that immigration was out of control. Not skilled workers per se
People definitely don’t want out of control immigration, but I don’t see the evidence that people don’t like controlled immigration of high skilled people
Students have degrees don't they (hence "skilled"). The pro-H1B crowd wanted to "staple a green card to every degree". And Canada's immigration policy was very similar to the H1B. Its salary minimum was exactly the same as the American H1B. I would regard Canada's experiment as "what if we took the existing H1B system and 10x its per capita rate."
What this really comes down to is what constitutes "skilled". If skilled means top 1% global stem talent, then obviously we want more skilled immigration. I would pay people in that tier a million dollars to immigrate.
If "skilled" means mediocre white collar workers from India, then I'm not signed up.
I think Canada proves that mediocre tier white collar workers are not the bottleneck on economic growth. If anything when you add significant cultural differences they just gum up the works.
So yes, the new H1B rules are good because they increased the skill level of the applicants. But remember that it took a civil war for that to happen. Elon Musk and the rest of the "tech right" said they are going to "rip the faces off" anyone that tried to change the H1B program. They wanted to replicate Canada's poor experience. They lost the debate and we are all better off for it.
Skilled is doing a a lot lifting and is purposefully http://ambiguous.it
HB-1s aren’t all skilled workers and the skilled ones aren’t better than Americans. Now with AI, we don’t need so many. Real geniuses and top tier workers can get an O visa or something sillier. Hb-1 needs to die and be replaced by a visa that benefits Americans.l, not just companies and Indians who have cornered tbr market and not because they’re so smart.
Thanks, Laurenz! I agree with the general thrust that "[Trump's] immigration policy is comparatively popular," which is something important for smart people to finally recognize.
That said, I don't think the relative increase/decrease Gallup item is a good way to diagnose the representation gap. It mixes up too many things at once: what people actually want from immigration, what politicians in power and opposition are doing, and thermostatic reactions to current events. A better existing measure would be voters' approval/trust for both parties on immigration, which gets closer to whether anyone actually feels represented.
I'd also push back on the broader read of Trump's "immigration policy." He (or Stephen Miller?) did a bunch of popular things (asylum restrictions, border security) alongside a bunch of unpopular ones (legal green card cuts, travel bans, high-skilled immigration cuts). Bundling them all together as "Trump's immigration policy" obscures more than it elucidates.
Thanks for your feedback!
In particular, thanks for suggesting another existing measure that addresses some of the shortcomings that you criticise (of course, all these measures are imperfect, but it's better to have an imperfect measure than none at all...).
A brief analysis suggests that during Trump-2, Americans think Republicans handle immigration better than Democrats (even though the Republican advantage has decreased strongly recently, it's still there), which is consistent with my main finding here. But this masks that both parties are unpopular on immigration.
The reason I did not use this measure is that it compares the actual policy of Republicans, who are in power, to some hypothetical policy (proposals or whatever people imagine) of Democrats. This asymmetry can bias their comparison. In addition, I think this measure also mixes many of the things you rightly say are mixed up in the item I use. That said, my post would have benefited from using more items, including relative approval.
Otherwise, I agree with you.
A problem with this analysis is you ignore the pre-1995 data. If you include that data you will find the net pressure to reduce immigration in 1965 (when the 1924 immigration restrictions were still in place and rates commensurably low) was higher than it has been since 2010. The public has *always* been in favor reduced immigration, even when it had been reduced. One might subtract this baseline level from the values in Figure. Here you would find that even with the adjustment, people were upset about immigration and wanted to reduced, leading to the Secure Fence Act under the Bush administration, but by the end of the Bush administration and afterward this has not been the case *except* for the end of the Bush administration when it became publicly well know just how many people were pouring into the country across the Southern Border (there was a Sixty Minutes segment on this in 2024). This explains why Tom Tancredo's immigration restrictionist campaign in 2008 went nowhere and why there was pushback from other Republicans on Trumps centering on this issue in 2015. They did not see that there was a constituency for this. And there wasn't. Trump's success in 2016 was based on his skill as a personal marketer and brand creator.
I doubt any other outsider running on the same issue mix could have snagged the nomination as easily as Trump did.
I agree that the analysis would profit from a longer time horizon. I ignore pre-1995 here on purpose because in this period coverage of the item is use is more sparse and for some presidents pre 1995 there is no coverage at all.
In general, the fact that people were still anti-immigration even after immigration had been reduced somewhat may be explained by the reduction simply being too small or the underlying immigration preferences having changed. Possibly, people in the 1960s were fine with much lower levels of immigration than people nowadays, but it is difficult to test and I am unsure.
In general, I agree this data shows, as described in the text, that part of the success of Trump-1 is probably his marketing. But, generally speaking, how concerned people are also lines up really well with the actual objective inflows of undocumented immigrants, so there may also be a large rational component (and Trump-2 did deliver).
I guess my main message for politicians would be to take this rational part serious and do something about it. Thinking that politics is purely marketing is extremely dangerous because it provides a justification to just ignore all problems and purely focus on talking them down. Even though public opinion does diverge from the objective truth for some time on some issues there is usually a rational core to each issue policymakers can address.
You analysis of flows also suffers from too short of a series.
https://substackcdn.com/image/fetch/$s_!wMd2!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F325709e0-9c1b-40df-818e-5fd41d966bd7_771x320.png
This border crossing data (i.e. the data your link points to, NOT the data I use) is very unreliable because most border crossings are not observed and the number observed is endogenous to policy decisions. I think regarding illegal immigration I use the best (different) data for all periods it exists.
Then why would you use it at all? My point is if you use it, then use all of it, because otherwise to looks cherry picked.
I dont use border crossings. The measures I use are clean and for these measures I use all data that exists. This does lead to different durations for the time series but it is, to my best knowledge, the best we can do with the existing data. As scientists we are unfortunately limited by data availability.
I see how this can look cherry-picked at first but using all clean data that exists is not cherry picking. Thanks for raising that point I hope that makes it clearer.
I found the data set you used. It is better in that in has more detail. I plotted the encounters data set from the CBD against the entering data in that set to see how good the correlation was. The r was 0.988, which is pretty good.
One has to weigh the detailed accuracy of the data one uses against the loss of information one loses by not considering a sufficiently long period of time.
By using only the 2013-2025 period one is considering only a single cycle: Obama-Trump I (low), Biden (high) and Trump II (low) By using lower-quality data (which still captures the trends that you care about) back to the 1990’s (for which you have reasonably dense Gallup data) you get the Clinton-Bush peak, capturing two cycles of immigration dynamics, increasing the power of your analysis.
My point was even when the percent foreign born was at very low levels (4.7%) the polls showed support for lower immigration was 27 points higher than support for higher immigration. I like your increase-decrease metric for immigration feeling. I tried regressing these against immigration rate per 1000 population data that I already had. I got a moderate correlation (r = 0.37) with the 2022 and 2023 points being way out line. If I take these out, I get a better correlation (r = 0.57). Now the 1965 value looks like an outlier, so I took that out and got a pretty good correlation (r = 0.66) with the remaining 26 data points. It suggests that immigration feeling would be neutral were immigration rate at 0.25%.
When I tried to do a regression using percent foreign born it was a nonsense correlation (negative slope) that is the higher percent foreign born the lesser the objection to immigration.
I see, I misunderstood you. Thanks for clarifying and sharing this!