Quantifying the macroeconomic and spatiotemporal impacts of high-skill immigration and visa restrictions on national productivity, native wages, and regional innovation.

The 3 papers this week all explore what high-skill immigration actually does to the economy that receives it? One looks at expansion, one at crackdown, one at a century of settlement patterns. They disagree about method and quality, but they converge on a finding that is politically inconvenient for everyone: the gains are real, large, and almost entirely invisible to the people receiving them.

Research Roundup

Merrily, Merrily, Merrily, The Gains Are All Downstream

The H-1B visa allows for the immigration of high-value workers into the US, usually STEM workers. From 1999–2003, the cap on these visas was expanded. A new paper used this policy “shock” to reveal that incomes for both native workers and pre-existing immigrants increased, but in messy ways.

The income gains propagate forward through supply chains to downstream industries, but not backward to upstream ones. The authors argue that a labor supply shock would ripple in both directions, but a productivity shock only moves forward: your customers get more for less. [1]

This strongly suggests that the value created by an H-1B hire mostly leaks out of the firm doing the hiring. The company absorbs the sponsorship cost, the legal exposure, and the political heat; the surplus shows up in someone else's margins 3 links down the chain. This textbook positive externality implies firms underinvest in exactly the thing we currently ration by lottery.

Externalities aside, the productivity gains appear to come from better task execution, not patentable invention. The standard case for high-skill immigration is the founder story—the unicorn, the Nobel Prize, the patent count. [2] This data says the actual mechanism is more boring but more generically valuable: complicated projects get finished. Immigration policies weighted toward publications and prizes are screening for the thing that didn't drive broad gains.

The last thing I noticed wasn’t explicitly discussed in the paper. Despite the visas being doled out for STEM labor, the gains are concentrated in non-STEM occupations such as sales, operations, marketing, management. In economic terms, the gains were complements.

So I wonder what happened inside the exposed job categories themselves? If the answer is "flat," then the political economy is exactly backwards: diffuse winners who don't know they won, concentrated losers who know precisely what happened and when.

[1] Of course, the tech superfirms are smart enough to own their customer relationships.

[2] I only respect one of those 3 measures anyhow. It's the one that gets you a free parking space for life at UC Berkeley.

I Slept Through High School

The 2017 "Buy American, Hire American" drove denials of immigrant employment petitions from 7% to 17%, and STEM-specific rejections tripled to 31%. A high school student decided to measure whether this change hurt the US states it affected most using quasi-experimental methods. You know…that teenagers do. [1]

The trouble is in the methods. The paper compares states that lean heavily on H-1B talent against states that don't, and finds the output  of H-1B-heavy states fell 2.8% relative to the others. But those states (e.g. California, New York, New Jersey, Massachusetts, Washington) are nearly the same ones hit hardest by at the same time by a 2017 tax law's cap on state and local deductions. Two things happened to one group at one time, a classic confound.

The author noticed, crediting the tax changes with masking effects on jobs and wages, but a shock big enough to hide one result is big enough to distort the others. The paper's claim of a $218 billion economic hit just isn’t reliable: it scales an uncertain estimate into an overly precise number.

Here’s one way the author might fix this (depending on how ambitious she is):. the order changed how H-1B petitions were reviewed and that varied by office, industry, and job type. So, in theory, pseudorandom variation across firms within a single state would sidestep the tax problem entirely.

All that said, the direction of the result does match cleaner existing studies, and so we can read it as corroboration rather than new evidence. That's a respectable thing for a paper to be, especially a first one.

This paper did get me wondering if the harm reported in all of these papers comes only from losing exceptional individuals or does it emerge from losing variety training, failing, and insight? Those point to opposite policies, and nothing in the literature cleanly separates them yet.

[1] I presume her choice was either kid economist or kid detective, but I would like to put in a plug for Buckaroo Banzi-esque mash-ups. If Jeff Goldblum can be a cowboy brain surgeon in a rock band, why can’t this kids be a standup-economist-whale biologist?

Your Great-Grandparents Set Today's Immigration Policy

Where immigrants settle today is largely predicted by where people from the same country settled a century ago. Newcomers go where they already know someone. Those networks are shockingly durable, an accident of history rather than a response to current conditions.

That quirk of American immigration history allows a new paper to separate the effect of immigration from everything else happening in a county. The analysis reveals that immigration raises local innovation and wages within 5 years, and that economic boost more than offsetting the short-run drag of more workers competing for the same jobs. Immigration since 1965 has raised American innovation and wages by roughly 5%.

But while gains out pace cost in the long run, those costs land immediately, in one place, on identifiable people. Gains arrive years later and spread across a much wider area. Even that “5 years” finds—fast, by the standards of economic policy—is slower than an election. A democracy will underbuild immigration for the same reason it underbuilds bridges: whoever authorizes it doesn't collect.

So time is bad for good policy, but space by contrast is an easier argument. Innovation and wages rise in the county that receives the migrants. Struggling regions don’t need protection from newcomers; they're the places with the most to gain…5 years later.

There’s another policy challenge. The same historical social networks that make the study work also limit what it implies. People go where they already know someone, and so the counties that would benefit most from immigration are the ones least able to attract anyone. Nobody's great grandmother settled there 100 years ago.

For policy to actually redraw the map, it needs to build community.

Takeaway

The gains from high-skill immigration are downstream, delayed, diffuse, and mostly show up in occupations that have no idea they're connected to the visa system. The costs are immediate, concentrated, and attach to a face and a job title. That asymmetry isn't an artifact of bad data; it's what the good data says. A country could know all of this with certainty and still vote against reasonable immigration policy every time, because the winners can't identify themselves and the losers can.

Perhaps the argument cannot be entirely empirical. Here's my idealized definition of an American: anyone who believes anyone else can be an American. It's recursive, which is weird [2], and I mean it as an aspiration rather than a filter. For me, it captures something real about the cultural looseness, rather than strict individualism, that is this country at its best. Let anyone in who is ready to work with a complete stranger to build something better. It's a definition that can't be used to build a wall…by definition.

Media Mentions

Sep 8, at the super{summit} 2026: Come hear me talk about my new research of hybrid intelligence and innovating innovation (w/ Possibility Sciences).

The entire amazing event runs September 8–10 at Convene in downtown San Francisco. Get tickets here: https://rewirecon.com/

Come for me; stay for the pie! [1]

[1] Or all the event awesomeness that isn’t me :)

SciFi, Fantasy, & Me

From the author of The Midnight Library comes the seemingly unrelated The Midnight Train. It was very cosy, a warm fantasy blanket somewhat in the vein of Life After Life, but with a decidedly uncomplicated attitude about what makes a good life. A literal dream about second chances.

Stage & Screen

• September 8, Online: How might AI change the world of investing?
• September 8, San Francisco: I'm going to talk all about Hybrid Intelligence at
• September 15, SF: Innovation Day with INSEAD!
• September 16, DC: AI and education–beyond dreams and dread.
• September 19, Phoenix: I'm giving the keynote for the Association of Science & Technology Centers annual conference.
• September 21, Stanford: We're still working on the details, but hopefully I'll be talking about my research on machine learning and neurodiversity for Stanford's Neurodiversity Project.
• September 24, UC Berkeley: It's my annual Berkeley Change-makers Lecture!
• September 24, NYC: Culture Shifting Deal Making Summit
• September 29, Cincinnati: Still baking...
• September 30, Irvine: Hybrid Intelligence for innovation!
• October 6, SF: UCSD Alumni Association
• October 6, SF: Giving a talk at the Draper Richards Kaplan Foundation
• October 6, Park City: It's Robot-Proof in the Rockies with setups.
• October 21-23, Warsaw: So much good stuff is in the works for my first visit to Poland
• October 27, Cologne: Maybe, maybe a visit to Germany!
• October, Toronto: The Future of Work...in the Future
• November 19, NYC: Secrets in the dark!

Vivienne L'Ecuyer Ming