Lauren Policy

weekly link roundup (May 5, 2026)

Lauren Gilbert — Tue, 05 May 2026 09:01:40 GMT

This week in : writes about Jakarta’s large expansion of transit over the last decade, and trying to tame the traffic of the world’s most congested city.
1 writes about how it would be ideal if writers on agriculture… knew anything about agriculture. An excellent quote from the piece:
They suggest that perhaps cattle producers can transition marginal rangeland used for grazing into other crops. Marginal grazing land is not a close substitute for perennial specialty crop production due to soil constraints, climate risk, and capital requirements. The piece explicitly references a conversation with an Oklahoma cattle producer who allegedly cannot afford to transition to growing stone fruit. There’s a very good reason for that!2
Nick Bloom and coauthors find that the ROI on good government statistics is 25:1 in the US. I would love to see an estimate on this in LMICs.3
has been on a roll this week: first pointing out that development aid will not reduce migration and we shouldn’t pretend it will, and secondly, addressing the ongoing debate on “should dev be more macro vs. micro”. He points out that macro research is often more useful to policymakers - though I will admit to an academic’s preference for clean causal inference.
Daniel Björkegren writes about what AI may mean for developing countries, and why the structure of these economies means it may not have similar effects to those seen in developed economies.
points out that it’s no one’s job to make sure the Prime Minister is good at his job. Shockingly, this often means they aren’t.
argues that Dulles Airport and the mobile lounges actually weren’t futuristic, because they failed to anticipate that the future would be different than present day.
The Blue Lagoon is actually… heated by a nearby geothermal plant? Based.

Abby is one of my favorite people I’ve gotten to know recently.

For those who have never been to Oklahoma: Oklahoma has lousy soil and no water. If you have a sudden desire to grow stone fruits, probably do it in not Oklahoma.

I am maybe a little bit obsessed with the quality of government statistics in LMICs. If you know things about this, please, please pitch me an essay at submissions@indevelopmentmag.com.

weekly link roundup (April 29, 2026)

Lauren Gilbert — Wed, 29 Apr 2026 09:00:26 GMT

This week on , discusses crime in developing countries with Chris Blattman.
There’s a new LLM using only training data from before 1930.
Michael Wiebe’s replication of Moretti 2021 has finally been published. tldr: the results are likely due to a coding error; moving to an agglomeration does not improve inventor productivity.
DHS data quality varies across regions - and northern Nigeria and Niger’s data looks particularly bad.
You know how you can get melatonin over the counter but it’s prescription in the UK? Gavin Leech and team put together a database of such drugs.
Howard French reviews How Africa Works.
has a new blog about growth and development - he opens with a post about just how rare sustained growth has been. In the last twenty years, no new country has met the Commission on Growth and Development’s definition of structural transformation (sustained 7% growth for 25 years).1
Drug development is a very expensive, very risky bet with terrible expected value. explains some of the financial engineering that makes it more profitable.

Vietnam is close, I believe.

Exporters Without Borders: Why You Should Start a Company Instead of Working in Aid

Lauren Gilbert — Thu, 23 Apr 2026 09:35:05 GMT

OK, I won’t link every piece from here, but I’m particularly proud of this week’s.

argues that more young people (from rich countries) who care about development should consider starting a business rather than working in the Aid Industrial Complex.

Daniel speaks from experience - he founded Wasoko, Africa’s largest B2B e-commerce company when he was 19. His company grew to employ 2,000 people and serve 100,000 businesses across the continent.

Read the piece here, and subscribe here.

weekly link roundup (April 21, 2026)

Lauren Gilbert — Tue, 21 Apr 2026 09:02:18 GMT

Have I perhaps mentioned that In Development has launched? Because it has; our first article (Paul Niehaus on cash transfers) is live; there is another article coming Thursday; you should subscribe here.
writes about how the AI people have been right about a lot. As a certified AI Skeptic myself, Dylan is absolutely right, and I should have paid more attention to AI earlier.
and discussed what an African YIMBYism would look like with the founder of the Centre for Affordable Housing Finance in Africa. Recommend!
argues Baumol’s cost disease is part of the reason Americans are so lonely.
GAIN is looking for mentors to help young Africans interested in graduate degrees in economics navigate the application process.
It appears to be generally better for academically precocious youth to skip grades - if they do, they are more successful later in life.1

Feels like this paper would have been more useful to me - or my parents - in 1993, rather than 2013.

In Development is live!

Lauren Gilbert — Thu, 16 Apr 2026 11:46:23 GMT

So it’s time to unveil the thing I’ve been working on for the last few months. is a new magazine dedicated to exploring how progress happens — or doesn’t happen — in the developing world.

We’ve just posted our first story. It is by Paul Niehaus, on cash transfers, evidence, and what empowerment really means.

It also contains the line “In 2005 my co-founders and I finagled invitations to a kickoff event for the International Year of Microcredit at the United Nations. The mixed drinks were, as I recall, stronger than the evidence.”

weekly link roundup (April 13, 2026)

Lauren Gilbert — Mon, 13 Apr 2026 09:01:39 GMT

Now would be a good time to subscribe to , because our first article launches Thursday!
Works in Progress’ summer program for ambitious 18-to-22 year-olds, Invisible College, is back this summer. I’ve had the privilege to hang out with the Invisible College participants the last couple years, and they’ve raised my ambitions. Highly recommend.
Speaking of residencies and programs, the African School of Economics Zanzibar and the Charter Cities Institute also have some exciting opportunities. Applications for their professional diploma program in urban development (2026-2027 cohort) are open and there will be a two-week fellowship for builders and technologists in Zanzibar this summer as part of Zanzalu.
Would you like to do a dumb thing to find out information about the world? Inspired by the Citrini Research employee who took a speedboat out into the Strait of Hormuz,1 there’s a fellowship for that.
The Berkeley Existential Risk Initiative is looking for a new Deputy Director. I may be biased here, but Liz (the current director) is a gem and a joy to work with.
writes on kinship cultures and how they make economic development hard. I do think that 2 has a good counterpoint, though; East Asia also has a strong kinship culture and has seen extremely rapid economic growth.
writes on what a development economics of growth would look like.
writes for about how life has become low-variance for the American upper-middle class (and what that means for having a child).

Probably do not do this.

Whose work is also coming to In Development!

If You're Reading Substacks About How You're Never Going To Get Married, You're Probably Going To Get Married

Lauren Gilbert — Tue, 07 Apr 2026 09:02:27 GMT

There has been a recent spree of articles about the decline of marriage, particularly focused on young, educated women who want to marry men.1 Matt Yglesias discusses some here, while Aria Schrecker’s series about dating has also gained quite a bit of attention.

Her second post is called “You’ll probably die alone”, and includes the following statement:

“If you were born in 1960, over three quarters of your peers would be married by 30. Only about 15% of people should expect to be terminally unwed.

This is the world of Sex and the City and Friends and it’s totally lost to us.”

This is false. While timing of marriage has shifted later, there is no lost world of marriage for the American upper-middle class.

Educated American women are as likely to be married as their counterparts 40 years ago.

(from Scarcity of College Men and the Decline in Marriage Among Non-College Americans)

This bears some emphasis. Marriage rates for college-educated women have been flat since the cohort born in the 1940s. There is no lost world of marriage.

Note that this graph does include recent cohorts; this includes those who turned 45 in 2025.

It does not include most Millennials or Gen Z, because it is not yet possible to determine if they will be married by 45. Given the stability of this number over time, though, I’d be surprised if Millennial women suddenly decided marriage was Not For Them.

Indeed, marriage has become increasingly class-linked in the United States. Consider the following graph of the percentage of women that are married by income:

(from Marry for Money or Time? Explaining New Marriage Trends in the U.S.)

The likelihood of a woman in the top income quartile - an income of around $85k for a 30 year old in the US2 - being married3 has increased over time. If you’re an elite woman, your prospects for a successful marriage may be better than they were 25 years ago.

You’ll probably only spend a couple fewer years married than your forbears - and they’ll probably be happier ones

There is also a broader point to be made about all of these essays, even if you are not a highly educated elite. In general, most people’s goal for marriage is not marriage per se, it is a successful marriage - that is, one ideally not ending in divorce.4

Therefore, I think it is reasonable to compare the amount of your life spent married rather than one’s relative success at getting to the altar. There are several demographic changes that affect the expected number of years married besides marriage rates. At the same time that marriage rates have fallen, divorce rates have also fallen; life expectancy has increased.5

Let us compare two women: one born 1958, one born 1998. The first was born during the Baby Boom, the prototypical high-marriage-rate environment. The second is a member of Gen Z, a generation supposedly uninterested in marriage.

On average, the woman born in 1958 is more likely to marry (90% would eventually marry), and when she does marry, she marries earlier (age 22). But Gen Z marriages are much more stable - divorce rates for first marriages have declined from ~45% to ~30%.6

I calculate out the expected number of years each will spend married in the appendix, and find that the Baby Boomer would (on average) be married 33 years of her life. The Gen Z woman? About 30 years of her life.

Three fewer years of marriage is a decline, to be sure, but “a disaster, the likes of which our civilisation has never seen” feels more than a bit overdramatic. (It also seems entirely possible that relationship-years are similar, since engagements tend to be longer; few people now marry after being together six or eight months.7)

The story for college-educated women is even more positive. More college-educated women marry8 and fewer divorce. Given that college-educated women are just as likely to be married at 45 as they were for previous cohorts, and are less likely to divorce, it is entirely possible that they will spend more of their lives married.

Even including the non-college-educated, though, a 28-year-old woman today will spend just as much time married as one born in 1958. She is also considerably less likely to go through the messy, painful, expensive process of getting a divorce.

Spinsterhood As A Great Western Tradition

Someone, I’m sure, is going to point out that 70% of Gen Z college women marrying is not 100%. This is true. But 70% is… not that abnormal for a Western country. The US and the UK have never had universal marriage, especially for educated women.

In the US, the percentage of white women never married in the age range 45-54 has varied from around 4% (the Baby Boom) to around 10% (those born around 1870). Now, 17% of women have never married by 45, with rates of spinsterhood (unmarried by ~45) for college-educated women around 13%. This is an increase, certainly, but a catastrophic collapse of marriage it is not.

Historical spinsterhood rates in the UK were higher; in Tudor England, the likelihood of remaining unmarried for your entire life was about 20%. By the Victorian era, spinsterhood rates exceeded 30% in some parts of the UK. Now, about 36% of women in the UK have never married by 45. Again, this is an increase, but there is no vanished world of marriage.

Historically, marriage rates for the educated were especially low9 - only 56% of the first ten classes at Vassar married. Only 72% of the Harvard men of the classes 1867-76 married. If about three-quarters of your highly educated, elite social circle is married by 46, you are… historically pretty normal.

Lack Of A Marriage Is Not A Lack of A Relationship

You may note that all of my marriage rate data is from the US. The US is a more marrying place than the UK, likely because it is also more religious.

In the UK, cohabitation replaces some percentage of relationships that would likely be marriages in the US. I think roughly 70% of Gen Z will marry in the US; in the UK, it is likely to be closer to 50%.10 In societies like this, though, the relationship decline is considerably smaller than the marriage decline.

Currently, about 25% of non-married adults in the UK are in such non-marital but cohabiting relationships.11 Many of these will eventually become marriages, but not all will; long-term cohabitation has become more common as an alternative to marriage rather than a prelude to it.

So: those 30% of never-marrieds in the US (and perhaps 40-50% never-married in the UK)? They won’t necessarily die alone. And if what you want is the right partner to make achieving your other goals easier, cohabitation clearly counts. Cohabiting couples have many of the benefits of marriage.12

So while it is true that the UK cohort born in 1970 will see 20% fewer of its members marry than the 1940 cohort, it does not mean that those 20% will be single for the rest of their lives.13

Indeed, the total fraction of people in committed couples has been relatively stable over the last decade and a half. In the 2011 census, 57.8% of the population in England and Wales was living in a couple. The 2021 census found almost exactly the same percentage of the population living as a couple - even as marriage rates continued to decline.

Marriage, then, is not the whole story. Particularly in the UK, cohabitation without marriage is common enough that “never married” does not mean “forever alone”.

Conclusion

The data shows no collapse in relationship formation among the demographics most likely to worry about a collapse in relationship formation. Indeed, a woman today will spend about as much time married as a woman born during the Baby Boom, and she’s quite a bit less likely to get divorced.

Is it true that “a lot of beautiful and interesting women… are going to end up single”? Sure. But for the college-educated, that’s no more likely now than it was in the past - even if you don’t count the uptick of non-marital long-term relationships.

As it turns out, our settling down technology is working just fine.

Many thanks to and for reading drafts of this, and for telling me to banish the math to an appendix.

Appendix

We assume that a woman born in 1958 has a life expectancy at birth of 73, and a woman born in 1998 79.14 The two women can have the following outcomes:

Never married
Married once (until death)
Married once and divorced; do not remarry
Married twice; the second marriage lasts until death
Married twice and divorced both times

For the sake of the model, I am choosing to ignore third or more marriages.15

For women born 1958, the likelihood of each category breaks down as follows:

Never married - about 10%
Married once (until death) - about 50%16
Married once and divorced; do not remarry - about 16%17
Married twice; the second marriage lasts until death - about 8%18
Married twice and divorced both times - 16%19

Calculating the length of each of these - and thus, the average marriage-years - is slightly trickier. We assume that a woman born 1958 has an average life expectancy of 73.

Never married - 0 marriage-years
Married once (until death) - married from age 22 to 73;20 51 marriage years
Married once and divorced; do not remarry - married about eight years;21 8 marriage-years
Married twice; the second marriage lasts until death - married for eight years in first marriage; four year gap between marriages; second marriage starts age 34 and ends at 73 and thus lasts 39 years; contributes a total of 47 marriage-years
Married twice and divorced both times - first marriage is eight years; second is seven; 15 marriage years

Therefore, the average number of marriage-years for a woman born 1958 is 33.22

Let’s consider a woman born 1998, marrying in 2026, at age 28.

We also know that the risk of divorce per year of marriage is now about two-thirds what it was in the 1980s. We know that about 45% of first marriages in the 1980s ended in divorce; therefore, I estimate that 30% of modern first marriages will end in divorce.23

Our categories are:

Never married - perhaps 30%
Married once (until death) - about 49%24
Married once and divorced; do not remarry - about 8%25
Married twice; the second marriage lasts until death - about 7%26
Married twice and divorced both times - about 6%

The expected marriage-years for each category are:

Never married - 0 marriage-years27
Married once (until death) - married from 28 to 79; 51 marriage-years. Note that this is the same number of expected marriage-years for the 1958 cohort, even though marriage is later; life expectancy gains completely offset this.
Married once and divorced; do not remarry - as above, eight marriage-years
Married twice; the second marriage lasts until death - eight marriage-years from first marriage; second marriage with similar divorce/remarriage timing will now occur at age 40, meaning the second marriage now lasts 39 years. Total of 47 marriage-years.
Married twice and divorced both times - fifteen marriage years

This woman’s expected marriage-years is therefore 30.28

So for all women, the number of years of marriage has declined about three years, or about 4% of her lifetime.

I’m not quite sure why it’s all addressed at women, since the men are also involved in these marriages, but nonetheless, they all are.

This is including both men and women, so presumably would be lower for just women.

This is at least partially because divorce is class-linked; the richer you are, the less likely you are to get divorced and the more likely you are to still be married if you have ever gotten married.

Most people do not enjoy getting divorced, or if they do, it is because their marriage was extremely bad.

Given this, one might expect selectivity of marriage and search time to increase, because you’re stuck with your partner for longer.

Overall divorce rates are more like 50% and 35%, but first marriage divorce rates are lower than overall rates.

Hi mom and dad! I’m glad it worked out for you, but if I tried that, you’d think I’d gone insane.

If you are wondering why only 71% of college-educated women are married in the first group and yet I assume 70% of all women will marry (even though college-educated women are a more marrying group) - note that the first graph includes only women who are still married at age 45, not those that have ever married.

This is no longer true; you are now more rather than less likely to get married with more completed education.

I’ve previously argued that the 1990 cohort will end up around 60% ever married; Gen Z is likely to be lower.

50.5% of adults were not married or civil partnered in 2024; 12.9% of all adults are in a cohabiting relationship but not married. Thus, around a quarter of unmarried people are currently cohabiting with a partner.

Not all; your cohabiting partner doesn’t automatically get medical power of attorney. But also: you don’t have to give them half your 401(k) if you split up.

Data on cohabitation is worse than data on marriage, largely because you are not required to tell the government much about your live-in partner.

This is life expectancy at birth, so realized life expectancy for the 1958 cohort is likely to be slightly higher, but it gives me an apples to apples comparison for the 1998 birth cohort.

In general, marriages beyond #2 are unlikely to last.

About 45% of unions in 1980 would end in divorce, so 55% would persist. I can’t find data on this cohort split by education; I use the overall rate because 1) fewer people in the 1980 marriage cohort were college-educated, 2) gaps in marriage dissolution rate by education level have appeared in more recent decades. Since 90% of women entered a first marriage, 55% * 90% would have a first marriage lasting until death.

90% marry * 45% divorce * 40% do not remarry

90% marry * 45% divorce * 60% remarry * 33% of those unions survive. Note that the last figure is estimated; in general, second (and third and fourth and…) marriages are less likely to last, with a dissolution rate about 50% higher than first marriages. Thus, I impute a divorce rate of about 67% for this cohort.

90% marry * 45% divorce * 60% remarry * 67% divorce again

Men have a lower life expectancy than women, and on average, a man in a heterosexual marriage is slightly older, so on average, women outlive their spouses.

I wasn’t able to find cohort-specific data on first marriage length or remarriage timing.

51 * 0.49 + 0.16 * 8 + 0.08 * 47 + 0.16 * 15

Overall likelihood is a bit higher, at perhaps 40% in the US and 35% in the UK, but first marriages are less likely to dissolve than subsequent ones.

70% marry * 70% stay married

70% marry * 30% divorce * 40% do not remarry

70% marry * 30% divorce * 60% remarry * 47.5% of those unions survive. Second marriages appear to be more similar between cohorts. Again, about 60% of those who divorce will remarry. I assume the percentage dissolution increase is similar between cohorts, so I estimate that likelihood of the second marriage ending is now 52.5% and survival rate is 47.5%.

Note that this number is quite likely to contain some non-zero number of cohabitation years, as cohabitation without marriage is much more common now than it was 40 years ago. But we’re discussing marriage-years, so zero it will remain.

0.49 * 51 + 0.08 * 8 + 0.07 * 47 + 0.06 * 15

weekly link roundup (April 6, 2026)

Lauren Gilbert — Mon, 06 Apr 2026 11:03:17 GMT

The Trump administration is once again proposing massive cuts to science.
says death to the policy report.
writes on how RCTs may not be enough to capture the effects of AI on development.
My friend and contributing editor Akib Khan is the International Economic Association’s featured economist for April.
My friend Oscar Sykes has a new article out about the history of instant coffee.
- He appears to have established a niche writing about things invented in New Zealand, so expect an article on Zorbing soon.
Coefficient Giving has open RFPs on biosecurity and effective careers.
IFP is hiring for an editorial director and operations coordinator.
is launching a new think tank on self-driving cars - launch event in London on Thursday. Come!
There’s a species of spiders with the scientific name Hotwheels sisyphus.

weekly link roundup (March 31, 2026)

Lauren Gilbert — Tue, 31 Mar 2026 09:01:45 GMT

Are you a person with a master’s degree who wants to work in development? You should probably apply to be an ODI Fellow (by the end of today).
Claudia Goldin is a mensch: “after Claudia Goldin became the first woman to win a solo Nobel in economics in 2023, she received hundreds of invitations and requests. She accepted just three. One of them was advising the WNBA players union as the women prepared to negotiate a new labor deal with the league.”
launched “Fast Biology Bounties” to surface ideas to “speed up or reduce costs for wet-lab experiments”. He found the experiment more successful than anticipated, and says that “good ideas are cheap and can be surfaced for a small amount of money.”
- This makes me want to do something similar with at some point.
On a sadder note: Asimov Press is pausing for now (as goes to Astera and to Anthropic). This is a tremendous loss for the scientific community, even as I’m very excited to see what Niko and Xander do next.1
argues that we need both a macro- metascience and a micro-metascience. Andrew Gerard (who writes ) wrote up a few thoughts.
The average flat in Manila costs twenty times the median income. This is considerably worse than famously unaffordable London, where the median home is about 14x the median income, or San Francisco, where a home is 10x median income.
An Irish guy created an AI voice agent to track the price of a pint of Guinness across Ireland. Good use of AI.
- The national average price of a pint of Guinness is €5.95. The modal price is €5.50. But also: Guinness still isn’t tasty and you’re better off having a nice glass of wine.2
Via Matt Yglesias: “69%3 of all criminal cases in New Haven between 1710 and 1750 were for premarital sex.”
Recommend this thread of British nicknames. Some examples: Keth for a man named Keith missing an eye, an electrician who looks like Elton John nicknamed “Socket Man”, and a man nicknamed “Minty” because he always gets to work after eight.4

I know I’ve said this before, but , , and were all a big part of me deciding to launch ; Niko was also exceptionally helpful in thinking through the decision process and my ultimately deciding to Do The Thing.

It’s fine; despite The Hair, I am in no way Irish, I am allowed to say this.

Nice.

For the Americans: After Eight is a brand of mint chocolates.

weekly link roundup

Lauren Gilbert — Tue, 24 Mar 2026 10:02:43 GMT

and the Horizon Scanning Study Group1 have launched forecasting questions about the progress of science. Please go forecast!
Dave Evans and Amina Acosta have published a new meta-analysis of information interventions for education in LMICs (à la Jensen 2010). I’d love to see more work in this area - yes, they’re not exactly revolutionary, but they’re so cheap and they do appear to work.2
and wrote a piece I’ve been meaning to write: American immigration outcomes really are different (and better) than European ones.
- I’d broaden this a bit to include the Anglosphere; note this graph from Marie and Pinotti 2024:

Immigrants in the Anglosphere are particularly underrepresented in prisons.

argues that you will become like the people you work with, so choose them carefully.
points out in The Economist that LLMs are worse in languages other than English. This has real consequences as LLMs are increasingly used in medical care.
thinks we could be a little fucked on biosecurity (but it’s not hopeless).
Selection Rules Everything Around Me: shows Muslim immigrants in the US integrate even more than polls of Muslim-Americans suggest, because the most liberal often no longer identify as Muslim.
University students in sub-Saharan Africa are dangerously financially vulnerable, often verging on ultra-poor. Student finance and housing could help.
One in ten babies in San Francisco is born via IVF.
does not endorse nuking ourselves a new canal to bypass the Strait of Hormuz.
The average American home has about 900 sq ft per person living there. Britain could never.

My baby continues without me.

OK, I am a bit biased on this particular point, because I was part of commissioning this meta-analysis when I was at Coefficient Giving.

Updates From The Immigration Literature in 2025

Lauren Gilbert — Wed, 18 Mar 2026 10:02:54 GMT

Academic literature is not static. There are new papers published all the time, and some of those papers meaningfully update what we know (or think we know) about a subject. Since this is a living literature review, it too updates as events happen.

Below, I discuss what we’ve learned in the past year about topics previously covered in this living lit review.

US Immigration Updates

Updating Undocumented People In The United States

2025 and 2026 have seen increased immigration enforcement, with many more ICE arrests and interior deportations increasing 4.6x. This has resulted in a significant slowdown of migration to the United States, including of legal migration. Brookings estimates that for the first time in at least half a century, more people left the United States in 2025 than entered it.

Perhaps unsurprisingly, this does not look great for the American economy. The Dallas Fed estimates that declining immigration will knock a point off GDP growth, and the Penn Wharton Budget Model has produced a model of the effects of deporting 10% of undocumented people per year.

Over four years, Penn estimates this will increase the deficit by $350 billion and decrease overall GDP. The effects on GDP/capita are more ambiguous; there is a short-term increase (particularly for the remaining undocumented workers who manage not to be deported), followed by a longer-term decline.

They find that most of the high-skill population will net lose out under mass deportations. This includes both authorized immigrants and the native-born. Older high-skill workers are the exception, as their working lifetimes will not include the longer-term decline. The picture for low-skill authorized immigrants is more mixed, with most having net gains (particularly older workers or those with high wages), but younger workers ending up worse off.

Updating H-1B Visas and the American Economy

In September 2025, the US introduced a new $100,000 fee on H-1B visas. I wrote at the time that I did not think this was a good idea.

This fee was the brainchild of George Borjas, who served in the Trump administration. In a 2026 paper, he argued that H-1B visa holders earn 16% less than similarly credentialed Americans, suggesting firms might be willing to pay an additional fee to continue hiring these low-wage workers. He calculated the revenue-maximizing fee would be between $118,000 and $264,000, and would have no impact on the number of H-1B visas issued.

The Economic Innovation Group argues that Borjas’ paper has data errors, and the actual wage gap is much smaller, at around 5%. In this case, a $100,000 fee would be far more than an employer would be willing to pay for a H-1B visa.

Michael Clemens also argues that Borjas has made data errors, but he finds that H-1B visa recipients actually command a small wage premium relative to the native-born. In this case, it is extremely unlikely that an employer would wish to pay an additional fee.

I have not gone through the data in these papers in enough detail to litigate who is correct, though Ozimek and Clemens’ point that Borjas uses 2020-2023 salary data for immigrants and only 2023 for the native-born seems hard to argue with. Revealed preference also seems to suggest that the revenue-maximizing fee is considerably less than $100,000, because overall fees collected on H-1B visa applications have decreased substantially since the fee was introduced.

Brain Drain and Brain Gain Updates

Updating What happens when you send Ugandan students to Germany?

Malengo is a program that provides a loan for Ugandan students to attend university in Germany. The expectation is that many students will then get jobs in Germany, and earn perhaps 20-50x what they would in Uganda. Students that stay will also pay back some or all of that loan.

Last year, they updated their model. They now estimate that the program produces 28x the returns of a cash transfer program. The positive update largely comes from moving consumption gains forward in time - even while attending university, consumption is some 7x what it would be in Uganda.

That being said, I think the cost-effectiveness of Malengo remains fairly uncertain until their first couple of cohorts of students graduate. Since the goal is to increase income largely through students choosing to stay in Germany after graduation, we still need to know what their labor market placements will look like.

Updating Migration of Doctors and Nurses

I have previously posited that there are significantly diminishing returns to additional medical personnel in places that already have a reasonable supply of medical personnel. I still think this is true, but I was surprised to see that the effect of adding or subtracting physicians is still detectable even in high income countries.

For instance, Dodini, Lundborg, Løken and Willén 2025 estimates that losing a doctor in Sweden results in 0.21 deaths per year.1 Grabowski, Gruber and McGarry 2026 estimates that each additional immigrant health worker in the US prevents 0.07 deaths per year,2 though this may be higher than in other contexts because immigrant health workers appear to be of higher quality than US-trained physicians.3

There does not appear to be a saturation level at which point additional physicians are not helpful. This would seem to suggest that expanding the supply of health workers is important in almost all contexts.

This would seem to make the general equilibrium result found in Abarcar and Theoharides 2024 extremely important. Abarcar and Theoharides shows that you can end up with a net increase in the supply of health workers even in a country with significant emigration if the following two conditions are met:

High wages outside the country draw people into health professions.
Supply of health professionals is allowed to expand to meet demand.

Some of the newly-qualified health professionals will migrate, but not all will.

If supply is not allowed to expand, you will probably have a net brain drain effect.4 So: expanding supply is very important; without it, migration can be net negative.

UK Immigration

Updating Are recent immigrants a “ticking time bomb” for British public finances? / UK Immigration and Public Services

We have a couple new reports from the Migration Advisory Committee, focusing primarily on the 2022-2023 cohort of immigrant entries. (This is the peak of the “Boriswave”.)

This has an interesting analysis on the fiscal impacts of each visa type, with a deeper dive into care worker fiscal impacts here.

The average skilled worker main applicant will pay £689,000 more in taxes than they receive in benefits (in net present value terms). The average health and care worker main applicant will pay £54,000 more in taxes than they receive in benefits.5 Care workers are net negative (-£36,000) while higher-skilled professionals are net positive.

Dependents are somewhat less fiscally positive; the adult dependents of skilled workers (e.g. spouses) are just about break-even, while the adult dependents of health and care workers cost £67,000 more than they contribute in taxes.

People who enter on partner visas are the most fiscally negative, at NPV -£109,000. This is consistent with the low labor force participation rate for this group (<50%). It’s not overly surprising that a group that is only 50% in work is fiscally negative.6

However, when considering the types of visas that the MAC analyzes, the cohort as a whole (skilled workers, their dependents, health and care workers, partners) is highly fiscally positive. They estimate that the 2022-2023 cohort of these visas will contribute £44.4bn more in taxes than they receive in benefits.

This is not the overall fiscal effect of the 2022-2023 immigrants, though, as we don’t know the overall fiscal cost of those entering on humanitarian visas.7 2022 was the peak of the influx from Ukraine; about 40% of immigrants that year entered on a humanitarian visa.

Given how positive those who entered on work visas are, though, these entrants can be significantly fiscally negative and the whole cohort will be positive. All other visa entrants can have a fiscal impact of -£122,781 (that is, more negative than any other visa type) and the whole cohort would still be fiscally neutral.8

3.74 additional physicians leaving results in 17.6 additional deaths.

Specifically, they find that each additional 1,000 immigrants to the US produces 143 additional health workers. These health workers prevent 9.8 deaths, or 0.07 deaths per health worker.

Since the US has some of the highest healthcare salaries in the world, I could imagine that it draws in some of the most talented doctors from around the world.

Given that condition 1 - there are high wages in health professions outside of low-income countries - is unlikely to change.

This is, of course, ignoring any other positive impacts from increasing the number of health workers in the country, as discussed above.

Increasing the income threshold for sponsoring partners is intended to address this, but this seems like such an odd policy choice. Is one only allowed to fall in love with a non-Brit, marry them, and live with them in Britain if you make greater than median wage? Surely the restriction here falls on your own citizens as much as migrants. And generally, citizens are allowed to make choices that cost their government a lot of money, particularly in countries where medical costs are covered by the government. No one prevents you from smoking a pack a day, even though it will cost the government more than bringing a foreign partner to the UK.

I don’t think I can really use previous numbers on refugee earnings, as the Boriswave cohort was largely from Ukraine and Hong Kong - not previously countries that had sent many refugees to the UK, and with a significantly different skill distribution and likely earnings potential.

Note that this is the “worst case” for immigration in the UK; 2022 had significantly more dependents and refugees than any other year, so if it’s fiscally neutral, every other year definitely is.

weekly link roundup

Lauren Gilbert — Mon, 16 Mar 2026 11:03:38 GMT

shows that people are more likely to support legal immigration when they know how difficult it actually is to legally immigrate.
writes a bear case for AI in Britain. Unfortunately, I worry Archie is right: Britain’s labour market does move very slowly and it is very exposed to services. This doesn’t look great for the future.
Astera is having an essay contest on systemic bottlenecks to science.
A point-counterpoint pair on AI and automation: on why doing tasks does not mean doing a job, and on why the ATM didn’t kill the bank teller, but the iPhone did.
India’s recent growth may have been less impressive than stated.
1Day Sooner is hiring for a policy lead for clinical trial abundance.
China is not a young country - the median age is now the same as the UK, and will shortly overtake it.
The US Post Office has simply defined the Pentagon (geographically very much in Virginia) to be in DC.
One in four gay and bisexual men 1 intend to vote Reform. has written a very funny taxonomy of them.

Gay and bisexual women are apparently holding up the Left.

weekly link roundup

Lauren Gilbert — Mon, 09 Mar 2026 11:02:35 GMT

Doug Rand and friends have launched a new immigration podcast: The Melting Pod. A must-listen.
My friend Josh Martin is launching Peace Per Dollar - a GiveWell for conflict reduction.1
New York is considering banning chatbots from giving medical advice. Look, ChatGPT is already vastly better than my GP.2
Being tough on crime… works? And lowers all-cause mortality.
Abortion bans reduce housing prices.
Residents of Istanbul are 2.4x likelier to search “Tinder” than “prayer times”.
Ezra Klein makes the case for Just Doing The Intellectual Work. Just sit down and read all the things. Do The Work.
People are more likely to apply to a particular college if they visit when the weather is good. This is basically UCSD’s entire recruitment strategy and honestly, it works.

And to close: a tweet from Ava I’ve been thinking about:

Text within this block will maintain its original spacing when published

my problem w the popular conception of agency is that most people are trapped not by an inability to act but rather by an inability to conceive of a wider range of things to use their free will on. their actions are constrained by the aperture of their desire

May the aperture of your desire be wide enough for you to use your free will to its fullest extent.

I would like to take, like, 0.0001% credit for this initiative, because Josh and I have talked about an EA approach to peacebuilding a lot.

My GP’s website turns off at night. I am not joking.

weekly link roundup

Lauren Gilbert — Tue, 03 Mar 2026 10:03:13 GMT

has an op-ed in the FT arguing for universal lead screening. has put together a tool to examine lead exposure levels in your community.
This feels weird to put in a link list, but if you identify as an effective altruist, or spend time around the EA community, you should read ’s account of trying to report sexual harassment.
is starting a program for Fast Biology Bounties - $10k for good ideas to speed up wet lab experiments.
The Centre for British Progress has outlined what an actual contributory migration system would look like.
is writing about living in Doha, watching the war happen.
Chris Blattman has shared his workflow with AI agents.
Ryan Briggs argues selection on precision would be better than selection on significance.
Telework doesn’t appear to make people lonely.
Dario Amodei is right: “Disagreeing with the government is the most American thing in the world.”
Every year, Paris holds a blind taste test for the best baguettes. The winner gets to supply the Élysée Palace for a year. The last two winners have been Sri Lankan immigrants. That’s integration!

weekly link roundup

Lauren Gilbert — Wed, 25 Feb 2026 10:02:26 GMT

Aishwarya Khanduja, , and Analogue Group have launched the Research Revival Fund - focused on reviving research that has been neglected due to funding loss or other similar situations.
- I’m particularly keen on their work on translating forgotten-but-still-relevant Soviet work. I’ve had more than one conversation with people in the nuclear industry where someone told me that their technology was reliant on an obscure paper written by a Soviet physicist in 1970.
George Borjas has a new paper where he lays out the empirics behind the $100,000 H-1B fee. The Economic Innovation Group says he’s made data errors.
disagrees with the Cato study that finds strongly positive fiscal impacts from immigration. (Cato has responded.)
argues that it should be easier to fire workers in Europe.
The UK graduate premium has declined. This is not true in other rich countries.
- This appears to be driven by two things: the UK’s real GDP/capita has only now recovered from the financial crisis and relative wages for non-graduates have risen (probably due to a rising minimum wage). Mechanically, if overall wages are not rising, but wages for non-graduates are rising, the graduate premium will decline.

Some journals require paper “highlights” in addition to the abstract. Dean Karlan is having fun with the form:

Text within this block will maintain its original spacing when published

In a Blank Space of rural finance, new platforms quietly entered.
Fearless farmers invest when digital credit is carefully tailored.
But Out of the Woods we are not — lending seems best when swift."
So Shake It Off, Begin Again: more innovation would be a gift.

Car accidents increase on days with major music releases, because people are more likely to be using Spotify while driving.
Bald eagles are suffering lead poisoning from accidentally ingesting bullet fragments. This feels… symbolic.

Let’s end with a poem by Su Tung-Po:

Text within this block will maintain its original spacing when published

"Families when a child is born
Hope it will turn out intelligent.
I, through intelligence
Having wrecked my whole life,
Only hope that the baby will prove
Ignorant and stupid.
Then he’ll be happy all his days
And grow into a cabinet minister."

The UK is dangerously ignorant about lead exposure.

Lauren Gilbert — Mon, 23 Feb 2026 08:30:18 GMT

About 170,000 UK children are exposed to clinically significant levels of lead. Only about 250 of them 1 will receive any intervention to reduce that exposure.

This morning, the FT covered an open letter from , and me calling for universal lead screening in the UK. It’s been signed by six MPs and a Lord across five parties. We also have a new briefing with the Social Market Foundation expanding on why this is important.

The National Screening Committee has responded that they are “working on plans for a review [of screening requirements]”. It’s a start.

Per year.

weekly link roundup

Lauren Gilbert — Mon, 16 Feb 2026 10:00:29 GMT

India has similar residential segregation to the US.
Rural electrification does not increase employment or incomes, even in the medium-term (~10 years).
- I do wonder how much of this has to do with poor reliability. Like much of the region, Rwanda has poor power reliability, limiting the usefulness of power for commercial applications.
London housebuilding has collapsed to a degree not seen elsewhere in the developed world.
Thomas Malthus’ students were less supportive in fiscal emergencies than students of other economists. (h/t )
The British public believe Nigel Farage would be a terrible date who would make you pay and then ghost you.

The UK should screen children for lead exposure.

Lauren Gilbert — Fri, 13 Feb 2026 12:04:28 GMT

, and I think that the UK should have universal screening for lead exposure, and have written an open letter to that effect. Seven MPs - including one Tory, one Labour, one Lib Dem, one SNP, and one Green - agree with us.

If you would like to add your name to the open letter, please email me at lagilbert@gmail.com.

weekly link roundup

Lauren Gilbert — Mon, 09 Feb 2026 10:02:11 GMT

The Cato Institute finds that immigration has reduced the US deficit by a third since 1994.
DIV is baaaaack!
There’s a new database of econ PhD placements.
Merely having choices is a source of utility.
This is old, but: Isaac Asimov was a serial harasser of women. As a woman who grew up on Asimov, I just… it’s bleak.
Apply for the NBER Innovation Research Boot Camp!
In 2014, Mark Carney had contraband cat food confiscated after he tried to smuggle his cat’s favorite food into the UK. We love a devoted cat owner who also loves econometrics.
Considering going on a pilgrimage to the site of the invention of ibuprofen.

weekly link roundup

Lauren Gilbert — Tue, 03 Feb 2026 10:03:18 GMT

Recommend the New Statesman on the other UK tax trap1 - unpaid carers face a cliff edge losing £4,331 the second they make over £10,192 (per year).
- This obviously disincentivizes people from working! Also… we know how to create benefit tapers; we don’t need to have cliff edges.
London is the startup capital for non-Americans.2
The modal degree for a cabinet minister is economics.
There’s a class gap in academia. PhD graduates whose parents don’t have PhDs place less well and make less.
Dina Pomeranz lets students from low- and middle-income countries join her global policy analysis class. Ask to join here.
writes about cities.
- She describes why I love cities perfectly: “This is what cities are for: to create bigger and better things, to experiment, to aspire, to push boundaries.”
The share of life spent in retirement has been stable since 2000 - improvements in life expectancy have been offset by later retirements.
- It still seems wild that people spend ~50% of their lifespan not working - the first 20 years in education and then another 20 years being retired.
Here’s a website that turns architectural renders into what they will actually look like on a random Tuesday in November.
Greenland sharks live to 450 and have a gestation period of perhaps 18 years.

There is a more famous tax trap at £100,000. All income under £100,000 is taxed at 40% (marginal), all income between £100,000 and £125,140 is taxed at 60% (marginal), and income over £125,140 goes back down to a 45% marginal tax rate.

But if you can move to SF to raise, you’re still probably better off.