But the paper by Jesús Fernández-Villaverde and Patrick Norrick has served as a springboard for some unfounded claims. Most notably, claims like this:

The claim here is “we’ve never been below-replacement before as a species.”

To which I say “WTF!?!?!?!”

Yes, we definitely have, many times.

Key points in this post.

2.1 is not replacement rate.

Replacement rate is determined by prevailing mortality conditions.

Human population has declined several times in the past.

Therefore, by definition, we were below replacement rate.

We can learn from prior episodes of below-replacement fertility.

What Is The Replacement Rate?

People often quote “2.1” as the replacement rate, i.e. the number of children each woman would need to have to keep population stable in the long run under the assumptions of a) stable mortality rates near modern levels b) relatively normal sex ratios c) no migration.

In practice, a country’s population growth rate can vary widely from the rate directly implied by TFR/Replacement because a) mortality is not stable, b) sex ratios may vary from nature due to abortion, c) there may be migration. As a result, you cannot easily infer whether a country’s TFR is above or below its replacement rate just from population growth.

However, the story for the human species is quite different. If you observe that the entire human population is lower in year t than it was in year t-100, you can guarantee there is below-replacement fertility. Think about why:

a) A century-long decline implies not just one or two years of bad mortality but a persistent condition of high mortality relative to fertility, which is exactly the dynamic “replacement” captures

b) species-wide sex ratios on century-long time horizons almost never vary widely from natural sex ratios

c) the scale of migration to and from Planet Earth is extremely small and entirely recent via Lunar and orbital missions

So, what is replacement rate? Well, it’s a basic math problem! One that JFV’s co-author, Patrick Norrick, is definitely aware of, because he wrote about it here. Here’s he and co-author Nicholas Eberstadt’s estimates of true replacement rates across countries:

That’s in 2023.

But, what did it look like in 1950, I wonder, the earliest year of the dataset being used here?

As you can see, it’s not just that TFR has fallen over time— replacement rate TFR has fallen over time. In 1950, the global replacement rate was 3 children per woman. Today it’s around 2.1. In the 1970s as war-related mortality was high, Timor-Leste saw its fertility drop and replacement rate rise, such that it nearly fell below-replacement despite having fertility of 5 kids per woman!

There are other fun examples. Consider Japan and Bulgaria. Both fell “below replacement” when their fertility was >2.2!

China’s replacement rate didn’t drop below 2.2 until 2015! I could go on but the point is true replacement rate varies a lot. Elsewhere, I’ve written about how replacement rate was probably below fertility in many specific countries in the 1700s.

But of course, in 1950, global TFR still vastly exceeded replacement rate. But we already could have guessed this because population growth was very high. So let’s turn to population growth!

Population Decline Is Not New

Read more

Let me also start out by saying I appreciate the work JFV is doing drawing attention to falling fertility. He is a rare prominent economist taking the issue seriously and studying what to do about it. And I will note he has some very good moments in the paper: a fair and sober assessment of where the evidence stands on pronatal policies, a reasonable rejection of intrinsic feedback mechanisms, a mostly reasonable verbal discussion of candidate explanations for recent declines (even if, as you see, I disagree with his gloss on them), etc, etc. There’s much that’s good here. He’s also right that the UN WPP data is of very mixed quality. He’s also a fellow at AEI so broadly on my political “team” too. My friends would tell me I should do this privately, that I shouldn’t do friendly fire.

But does that sound like me!?!?! No, no that’s not me. Today we’re choosing violence. We’re doing friendly fire, because, you see, what I don’t appreciate about JFV’s approach to the topic is the obscurantism it introduces. I think that rather than correctly recognizes a “terra incognita” of fertility decline, he’s mostly just burning the maps.

So let’s dig in. Here are the major points:

Alternative theories were not even tested, let alone rebutted.

The model in the paper doesn’t even remotely test the theory the abstract advances.

The model in the paper doesn’t actually tell us anything at all about any causes of falling fertility, because it has zero substantive variables.

The key results and impressive claims of the model are almost all baked in by assumptions, not features of the data.

The result of all of this is a campaign of obscurantism that, while helpfully drawing attention to the issue of low fertility, ultimately reduces the quality of public understanding— and we really do have a good causal model of how to think about fertility decline!

How Does JFV Rule Out Candidate Explanations?

The main cause of falling fertility globally over the last 200 years has been declining child mortality. Period. Solved. It simply isn’t important to understand that issue further. We mostly know why child mortality has fallen (medical improvements, public sanitation, better nutrition, hygiene, lower infanticide, etc), and we know there is a rock-solid causal arrow between child mortality and family size. I have written about this ad nauseum, the latest version is here, but you can see the thesis all the way back here as well. JFV knows this, and acknowledges it in the paper, but then doesn’t use mortality-adjusted fertility in his actual models. I’m not joking! He has a whole section where he’s like “We know child mortality explains prior declines, that’s solved” then just doesn’t use surviving family size as his key dependent variable in later analyses. Classic case of acknowledging in your lit review issues you do not plan to account for in your model.

Thus, the only interesting question to study is why surviving fertility is declining. On that, the abstract makes big promises: “None of the commonly cited mechanisms can account for this pattern, so we offer a conjecture: modernity itself, which makes a third child expensive and childlessness cheap.”

So you’d expect the paper contains, like, some kind of model, yes? A model which shows, say, that the coefficient in a panel model with fixed effects between education, or GDP, or mortality, or whatever, fundamentally changed between 1980 and 2020? That kind of model?

No! There’s no model! None at all! It’s purely a verbal argument! He just sort of dismisses offhand vast literatures! And he dismisses them all the exact same way: “This one variable doesn’t explain all observed variation, therefore it cannot be very important.” That’s the move. Here, receipts:

On economic fundamentals:

On housing:

On shifts in priorities:

Actually they didn’t make a rebuttal to a Kearney-style priority shift argument!

On cell phones, which actually are a temporal match:

There’s no actual claim here, just hand-waving.

Note JFV also did not even consider the actual leading theory: marital decline. On twitter he says the reason why is because it’s a “proximate” rather than “ultimate” cause. What he means here is “Sure, marriage decline may cause fertility decline, but that just pushes the question one step back: why is marriage declining?”

Yes, true— but it’s an interesting step back! If marriage decline really is the proximate cause then certain ultimate causes may lose their credibility (such as the one JFV outlines: demand-satisfaction kinks at integer parity numbers, which would not necessarily prevent marriages for forming) whereas others may gain credibility (housing, wage differences, gender norms, or cell phones— any of which might influence marriage in various ways). The fact he didn’t even consider, and then waived off without explanation, the actual leading explanatory factor and cause of low modern fertility boggles the mind.

So there was no actual rebuttal here. Just “two explanations aren’t a perfect fit for the data, one… I just won’t address, and one… well maybe it’s the explanation but I won’t admit it so here, instead, it’s modernity.” There was no actual analyses here, and a massive omission. “Other explanations didn’t fit” my friend, you did not even try the other explanations on for size.

Then he gets to his preferred explanation, “modernity.” What is modernity?

Modernity Is Just Gender Equality????

You might think “modernity” means education, or higher income, or capitalism, or scientific thinking. For JFV, it is none of those. First, he says:

Okay, cool, so we are about to get an analysis of formal institutions, specialized expertise, and state capacity, right?

Hahaha, no

It’s just more Incomplete Gender Revolution/Feminism Is The New Natalism stuff. Modernity? Nah, it’s just a mismatch in gender ideology.

I’ve written about this before, but the TL;DR is simple: within societies, more gender-egalitarian people almost never have more kids than less gender-egalitarian people, and in a panel model, changes in the mismatch rate have no relationship to fertility. This theory is wrong, wrong, wrong, but like all zombie ideas, it just won’t stay dead.

So the gender equality argument JFV is advancing is dead on arrival. But maybe he shows new evidence for it? Maybe he has a cool model for it?

How Not To Model Modernity

Hahahaha, no. JFV has no actual test of his theory. There’s no model where he has some variable for “modernity” or “gender egalitarianism” and shows some effect. He ruled out all the actual, measurable variables because they were an imperfect explanation, and instead offers a vague, diffuse variable, for which he has no measure, and which is not actually included in his model.

So what’s the model? Here, I warn you… we’re going to do math.

Here are the key model results:

For laypeople, the basic model works like this: they fit a common factor of fertility across all countries for each year, under some constraints. So in some years where all countries are atypically high, the top left graph is high. In years where all countries are low, top left graph is low.

Then there’s also trends for each country created by smoothing recent changes. That’s panels b, c, d, and e shown in various ways. Countries are allowed to have various trends.

As you can see, there are no substantive variables. No GDP. No education. No gender equality. No marriage.

This is like trading stocks through technical analysis. Maybe somebody makes money off it, but the real way to make money on the stock market is… substantive knowledge! Have an insider scoop on what vaccines work or what technologies will pan out! The demand for explanations of low fertility is entirely a demand for variables which will reduce these technical trends to near zero. That’s the whole project!

Think about this model. All this model is doing is saying, “Look, fertility is low in X year in many countries, and here’s an average trend.”

How would you use that for, say, a forecast? You couldn’t! Because the input data is the contemporary output data. The model here is merely a summary. Suppose you think GDP growth will double due to AI: what does this model tell you about that? Answer: nothing. Suppose you think gender equality will get better: does this model tell you anything? No! It doesn’t!

The whole project of studying fertility is to find some variables that can account for the effects JFV measures, and thus reduce them as close to zero as possible. We don’t want big residuals! We don’t want our models to depend on arbitrary parameters! We want substantive foundations. What JFV is doing here is a fun quantitative exercise of the sort many demography students will do in an advanced methods class, but it is not, in fact, an explanation.

It’s also… not empirically credible. For the model nerds, here are some issues you might notice:

Normalization predetermines results

The cross-country factor analysis is actually normalized in two ways. The factor is forced to be orthogonal to both a constant and a linear trend. A series orthogonal to a constant must cross zero; being orthogonal to a linear trend also means it cannot be monotone. In other words, JFV’s factor is mathematically required to rise and then fall (or fall then rise). The existence of a peak is imposed by the normalization, not discovered. Only its date is data-driven. This is akin to when somebody uses Variable and Variable Squared and then DISCOVERS a U-curve effect in a variable. No shit bro! You designed a model that maximizes on discovering such a curve!

Now read the paper's rhetoric: "the common factor resembles a bygone wave: it peaks in 1978 and shows little change over the past fifteen years." That wave shape is a predetermined mathematical artifact!

Decomposing and smoothing isn’t a neutral choice

Suppose you observe a product equal to 12. Is it 3 X 4 or 6 X 2? You don’t know! The model use splits an observed change into two unobserved components in a similar way. To do it, a rule has to be added: the loading component must change slowly, the factor component may jump around freely. Nothing in the paper tests that rule. Basically JFV has an aesthetic preference about which graph should carries the wiggles. About a quarter of their total effect comes from this one assumption. They report robustness across bandwidths from ten years to the full sample but never show how δ moves along that grid, making it hard to understand exactly how much of their result was predetermined by their assumption about how to decompose the observed trend.

No test of the factor model

They assumed exactly one global story (i.e. one global factor) and never checked (or never reported checking) whether two fit better. One factor with shifting loadings and two factors with stable loadings are very hard to distinguish! JFV picked the version that supports the one-factor narrative. Suppose there really are two factors: a mortality-decline factor that finished its work decades ago, and a second modernity factor still running today, loading positively on countries the first one has passed. You’d see exactly their Figure 3 — a “peaked” first factor, loadings flipping sign, rising dispersion. But the conclusion inverts completely: the common global force isn’t spent, it’s been replaced. Testing and disclosing more factors isn’t hard at all.

The model is weakest in the recent years

The factor’s ability to pin down a country’s loading depends on how much λ is moving locally. But according to the paper, that quantity drops sevenfold from the 1970s to the 2020s. Today, λ is nearly flat. Estimating how strongly a country responds to something that barely moves is like estimating a car’s steering ratio while it drives in a straight line: the ratio explodes and means nothing. That’s why some of their other country-specific estimates​ inflate in later years, and why they report βλ instead of just one of those estimates. It’s Charlie Brown’s football. Reporting the product keeps the number apparently stable. The recent period they most want to talk about is the period the model can say least about!

Correlation mongering is just sad

The median R2 of 0.967 is reported after removing the level and the trend. Each country gets its own intercept, its own slope, and its own smooth loading path fitted to 74 observations. That’s a whoooooooole lot of parameters per series. A really good model fit is close to guaranteed. Pretty much anybody, given that many parameters, can fit a nearly-perfect model. In fact, with that many parameters, I’m not even confident that this model is the best fit you could achieve, nor am I sure this is the most parsimonious model which could achieve that R2.

The key point is simple: this model tells us nothing. We learn nothing from it.

Obscurantism Is Bad

JFV cites and discards a lot of really good empirical papers in this paper. That’s sad. Engaging more deeply he might have learned something from them.

We have reams of high-quality, robust empirical papers showing that fertility is influenced by many factors. The urge to find a single common factor is simply a theoretical error, and one the empirical evidence strongly refutes by virtue of the fact that we keep finding new causes that have real effects.

The result of “oooooo it’s modeeeeeeeernity” is the closing of curious minds. This kind of argument shuts off serious work into what is it specifically about modernity that matters. The theory it’s bureaucracy or state capacity is worth testing! JFV should test it! “Modernity” is not a theory; modernity is cope.

At a bigger level, we need a better theory of social change. People often have theories that go like this:

Mechanistic Cause—>Mechanistic Effect

Cool, right?

A more plausible model of the world, however, is:

Probablistic Cause—>Probabilistic Effect Which Sets Conditions For a Complementary Probabilistic Cause—>Probabilistic Effect Which Sets Conditions For a Complementary Probabilistic Cause—>Probabilistic Effect Which Sets Conditions For a Complementary Probabilistic Cause—>Probabilistic Effect Which Sets Conditions For a Complementary Probabilistic Cause—>Probabilistic Effect Which Sets Conditions For a Complementary Probabilistic Cause—>Probabilistic Effect Which Sets Conditions For a Complementary Probabilistic Cause—>Probabilistic Effect Which Sets Conditions For a Complementary Probabilistic Cause—>Probabilistic Effect Which Sets Conditions For a Complementary Probabilistic Cause—>Probabilistic Effect Which Sets Conditions For a Complementary Probabilistic Cause—>Probabilistic Effect Which Sets Conditions For a Complementary Probabilistic Cause—>Probabilistic Effect Which Sets Conditions For a Complementary Probabilistic Cause—>Probabilistic Effect Which Sets Conditions For a Complementary Probabilistic Cause—>Probabilistic Effect Which Sets Conditions For a Complementary Probabilistic Cause—>Probabilistic Effect Which Sets Conditions For a Complementary Probabilistic Cause—>Probabilistic Effect Which Sets Conditions For a Complementary Probabilistic Cause—>Probabilistic Effect Which Sets Conditions For a Complementary Probabilistic Cause—>Probabilistic Effect Which Sets Conditions For a Complementary Probabilistic Cause—>Probabilistic Effect Which Sets Conditions For a Complementary Probabilistic Cause—>Probabilistic Effect Which Sets Conditions For a Complementary Probabilistic Cause

It’s a cascade.

Consider an example. You have a job. You accidentally make a mistake on a huge project. You get fired. You go to the bar. You drink too much. You get in your car. You drive home. You get dizzy while driving. You veer into another lane. You swerve and hit another car, killing its drive. That driver was a man who otherwise would have been Hitler 2.0. You just prevented the Holocaust 2.0.

You could say there’s causality Mistake—>No Holocaust. Or Fire—>No Holocaust. Or Going to the Bar—>Drunk Driving

All true!

But the better conceptual framework is that each event set the stage to increase the probability of the next event. Having a job created the probability space where you could get fired, but did not fully determine it. Getting fired created a higher probability of going to the bar, but did not fully determine it. Going to the bar created a higher probability of overdrinking, but did not full determine it. At every stage, your prior choice nudged the odds just a bit. The result is what we could call a “causality cascade,” where a “weak” initial cause ends up increasing the probability of an additional cause with a correlated effect direction.

We can see this in policy. Once you’ve mandated schooling up to 5 years, you’ve caused a more educated society where parents expect kids get educated and employers expect some education from workers— which makes a demand for extending to 6 years likelier, then 7, then 8. Each expansion increases the probability of future expansions.

Obviously causal cascades can intervene on each other. Each educational expansion sets conditions for further expansion, however, it may also boost incomes, and higher wages create stronger incentives to finish school faster and get to work. I’m not arguing there are never “disciplining” dynamics, but just that social forces often ratify and reify one another.

As a result, the idea that 50 different small causes may stack to cause falling fertility turns out to be extremely plausible. Not because there were 50 coincidences (that’s implausible!) but because the first small cause made the second one more probable. Once religiosity started falling in early-18th-century France and Massachusetts, it made the next thing (mortality falls? inheritance changes? educational expansions? it can vary) more likely to happen. This is also why the “kickoff” factor for fertility transition varies by country, why countries often do phases of fertility change out of order, and why predictions like “75% of countries have seen X happen, it will probably happen in the other 25% eventually” tend to be true. Each small causal factor shifts the probability distribution of the other causal factors occurring or reaching a level where they impinge upon fertility.

Beyond causal cascades, you also have interactions. A variable may not matter at all— until it does. The simplest case is fertility delay. For a society with a 2-child norm, delaying first births from age 15 to age 17 likely has very little effect on odds of achieving 2 kids. For a society with a 6-child norm, it may impact odds of hitting 6. But delaying first birth from 29 to 31 may impact the 2 child norm— it probably doesn’t effect odds of hitting 6 because that ship has sailed. You’re approaching a zero lower bound there.

Which variables matter depends on the level and structure of other variables. Just as individuals have a life course where events are essentially sequenced, so too do countries have sequenced events, which means you have to care about interactions.

What JFV argues is, basically, “There’s one big cause, but we basically have no idea what it is. The problem is simple, but obscure.”

What I argue is, basically, “There are a lot of causes, but we basically know what a whole lot of them are. The problem is complex, but relatively clear.”

JFV’s argument is probably better for fundraising and social media engagement. Mine is better if you’re interested in seeing countries actually raise their fertility rates.

The Argument

Men also want it all

Men are no longer the breadwinners they once were. In 1972, husbands earned more than their wives in at least 85% of American marriages. Husbands were the sole earners in more than half of those…

Read more

18 days ago · 34 likes · 5 comments · Maibritt Henkel

My thesis here is going to be simple. has done an admirable thing trying to incorporate data into public-political-vibes reporting, they’re sincerely trying hard to do good survey work, but at the end of the day if you aren’t a survey person, you just shouldn’t put pen to paper about surveys. In this case, has fundamentally mischaracterized their own results in a way that reveals an elementary misunderstanding you’d be disabused of in the first 30 minutes of any serious study on how to use survey data.

What Are These Surveys?

uses a company called Verasight. Verasight is a bit of an up-and-comer in polling. They claim to have higher respondent quality than other similar low-price pollsters. This is similar to Prolific’s marketing, if you’re familiar. The key point is Verasight manages their own panel and doesn’t use 3rd party vendors. People vary in their opinions of this approach. On the one hand, this allows stronger quality control. On the other hand, this massively reduces your sampling frame. Verasight pulls from a few hundred thousand possible respondents; other companies pull from many millions. As a result, you are highly constrained on how you can weight variables, because Verasight’s sampling frame just doesn’t have some populations. The takeaway here is you get “good” respondents but relatively non-diverse respondents.

Likewise, Verasight supplements their online recruitment with address-based sampling. Old hands at surveying know this is “gold-standard” practice… except not really. Modern response rates on address-based sampling are insanely low and highly non-representative, such that address-based sampling can help you avoid bogus respondents but it can no longer plausibly claim to give you representative respondents.

Finally, Verasight has some good practices around identity verification and managing respondent attention. On the whole they probably do above-average in terms of quality, and their internal evaluations do claim to outperform other online survey companies and even some probability samples.

The questions asked are also reasonable. Nothing about these questions raises red flags about survey question design:

Reasonable questions in a decently plausible method with 3,000 registered voters is a good survey. I belabor this, because the point is, is doing it right, trying to hire a reputable company to do good work. Their questions are apparently vetted by an expert panel as well. All great.

Unfortunately, the challenges with surveying don’t stop at the survey.

Do Men Want to Be Breadwinners?

Look at the graph above. It shows that 29% of men strongly disagree that men should be breadwinners, vs. just 11% strongly agree.

Does that support this claim?

One explanation for what is going on here is that men are answering out of solidarity with women — or, more cynically, a desire to give the socially desirable response — rather than pure self-interest. They reject the traditional division of labor into male breadwinning and female caregiving because they understand that sharing these responsibilities is good for women.

And yet, when we asked men what was more important to them — career success or getting married and having children — 44% of men prioritized family while 41% prioritized work, around the same shares as women. Among conservative men, the share prioritizing getting married and having children was even higher.

Obviously, no!

They did not ask men “Would you rather earn more than your wife or less than your wife?” They asked men “Should men be breadwinners?” The first question is a question about preference, the second is a question about norms.

It’s very easy to imagine that most men oppose the idea of a general norm of male breadwinners, yet most men might also prefer to out-earn their wives, or even have a stay-at-home wife. There’s no contradiction at all between these views. That which people espouse as general norms need not also be espoused as personal preferences. “I don’t think women should have to be 2nd class workers, but for my own family, it works best if I’m the primary earner” is in fact an extremely normal and common view to hold for many men!

Consider the claim they hope to rebut:

Some see the rise of the female breadwinner as a bad deal for men. Women may have gained independence, but they have done so at the cost of men’s sense of purpose. “Men are more likely to flourish when they see themselves as providers,” wrote Brad Wilcox, a University of Virginia professor and author of the book Get Married: Why Americans Must Defy the Elites, Forge Strong Families, and Save Civilization, along with co-author Maria Baer. “Men do better when they feel needed — and being a reliable breadwinner makes them feel needed.”

Does the survey above rebut this claim? No, not at all! Very obviously, is referring to a personal preference. Brad did not claim “Men do better when they feel that women generally are inferior — and seeing other men be breadwinners makes them feel needed.” He made no claim about what kind of general norm men might espouse. He made a claim about what they feel in their own personal lives, a topic not surveyed here.

On the question of what men actually personally want, we have a very great deal of evidence that, indeed, men tend to be happiest when they make essential contributions to family economic wellbeing. This often but not always means being the primary earner, sometimes by a very lopsided margin, sometimes as sole earner. Likewise, we know that the share of women who prefer to stay home with children vastly exceeds the share for men. Brad’s claim is not rebutted and, in fact, remains enormously plausible.

Do Men Value Family?

To support the claim, shows this graph:

And she says:

On the right, there is a sense that feminism has forced men into egalitarian relationships they do not want. Traditional marriage gave men a clearly defined role: to provide the primary income to their families. They would like that script back because without it they are in crisis.

Conversely, the left has tended to frame men relinquishing the breadwinner role as a moral imperative. There’s a sense that cheering on their wives’ careers and taking on child care is the least they can do.

Neither framing captures what our polling indicates: that changing gender norms around who spends time at home and who spends time at work might actually align with men’s own interests. Indeed, a majority of fathers say they spend too little time with their children, mostly due to work obligations.

Does any of this follow at all?

If I had to choose, of course I would say “Getting married and having children,” yet I’m the sole earner for my family. And guess what? Being economically successful is a key way men achieve marriage and children! Women continue to value men’s economic productivity extremely highly in relationships. When men become unemployed, their wives become much likelier to file for divorce, even though that divorce will make everyone involved even poorer. Perhaps the unemployed man became moody and hostile— but we needn’t pretend like women don’t find male idleness unattractive nearly to the point of repulsion! Witness the discourse around male-typical hobbies!

Paradoxically, when a man who says “Family is extremely important to me” probably has higher career ambition than a man who says “Family is not very important to me,” whereas for women, the relationship is far less clear, maybe even negative. This is why, for men, there is a strong positive gradient between income and fertility, whereas for women, it’s neutral to negative.

More to the point— for some men, the question asked is very optimistic! “If you could choose…” implies you’re guaranteed what you choose. I bet a lot of the men choosing family in the survey are, in real life, single and childless. Singleness and childlessness is more common among men of a given age than women due to age gaps in relationships, higher prevalence of men with multiple partners than men with multiple partners, and higher prevalence of women with interests in same-sex relationships.

This all adds up to me saying: it’s totally unreasonable to jump from “men value family” to “men want their wives to be their breadwinners.” Those statements could each be true or false but they have no relationship to one another at all. If you’ve read much about the rise of female breadwinners, you know the stay-at-home men are, very often, not involved and attentive fathers, but actually just deadbeats.

Misapprehensions Are Misunderstood

Consider this graph:

It’s very cool. This kind of survey design, where you ask respondents to guess what other respondents say, is becoming very popular, for the very good reason it allows you to produce graphs like this.

But what does this tell us? Does it tell us…

As in our February poll, there is little evidence in our survey data to suggest that men see traditional gender roles as more favorable than modern egalitarian ones. As Jerusalem Demsas recently pointed out, strong baseline support for gender equality is an underrated fact of our modern political landscape: Around 81% of both male and female respondents agreed that women are equally suited to leadership. Among male respondents who voted for Donald Trump, the share was lower but still 72%.

No! It does not tell us anything about “gender roles”! Traditional gender roles aren’t surveyed here at all!

What it tells us is male motivations are broadly misapprehended. A lot of men internally perceive themselves to be motivated by family, but externally they are perceived, by men and women alike, of being motivated by career ambition. It’s as simple as that. People impute good motives to themselves and bad motives to other people. When I log extra hours, it’s a valiant breadwinner providing for his family. When he logs extra hours, is a careerist creating a striver culture nobody can keep up in. The fact that people grade themselves favorably has been demonstrated time and again in many surveys.

This explanation is staring us all in the face here:

Whereas interprets choosing family as “choosing egalitarian gender roles,” in fact liberal men were less likely to prioritize family!

Prioritizing family just is a conservative value! We’ve been here before, folks! I mean seriously, look at this:

Trump-men favor family 57:32, but think men generally favor career 65:24. Harris men actually do favor career 48:33, and think men generally favor career 72:15.

It’s obvious that favoring family, and thinking other people may favor family, just is a conservative thing, so citing favoring-family as evidence of egalitarianism is just nutso.

Finally, I just have to note, the final header, “There has never been a better time to be a father,” is essentially a statement devoid of evidence.

Does This Matter?

I like and wish them well. I would like their style of engagement to succeed on the left. As a result, I desire for them to do a good job making their arguments.

They did not do a good job here, and the reasons was just a misunderstanding of what they were surveying. They assumed that their questions about norms could be interpreted as questions about preferences, and then imposed their own notions of what those preferences might mean (family-preferencing=staying home, etc).

Survey respondents are responding to the questions we ask them. We have an obligation as users of those responses to treat their responses with a modicum of seriousness and credibility. That means not reading in things they didn’t say. did not ask people about their personal preferences for the gendered division of labor in the household. Maybe they wish they had. But wishing does not make it so, and as a result, they instead egregiously misinterpreted the respondents they actually did get, in order to make a claim their evidence cannot support.

Then they went looking for somebody. I think it’s likely they contacted one or more of the people or organizations I mentioned, possibly Philip Cohen. What I did not know is that Cohen had just published this article about me.

What actually happened here is simple. Notice the date for Cohen’s article, July 13. Why was I the first thing on Cohen’s to-do list when he sat down at his desk on that Monday? Answer:

Some of us, when our professional rivals or people we disagree with have major media hits, choose to argue with their arguments. We may say things like “Such-and-such a fact claim is incorrect.” Others choose to just launch ad hominems.

His recent kerfuffles aside, Cohen is a capable demographer and researcher whose academic work has been influential on how I think about family demography, and so I genuinely regret that he chooses to engage in such a petulant fashion.

Anyways, the result here was the university informed me that my beliefs violated California nondiscrimination law:

Now, this is the first time I’ve been told that the Civil Rights Act of 1964 makes it illegal to say things like “some communities are bad” or “it’s bad if more people go to hell.” But hey, people read laws in different ways.

Anyways, I was a bit bemused why those particular tweets would have come up, which led me to Cohen’s article, which I suspect is the source, because he specifically highlights them as supposed examples of my bigotry.

Naturally, this Substack exists to double down on my entirely correct claims.

Religious People Are Religious

This tweet is one of the more infamous Lyman tweets out there, and Cohen highlights it:

Obviously, I stand by it, and it’s correct.

I am a Christian. By definition, as a Christian, I believe the teachings of Christianity, which state that faith in Jesus Christ has a very large causal effect on an individual’s long-run wellbeing (i.e. heaven vs. hell). If I were to say, “I am a Christian, but I do not care if other people go to hell” that would be obviously immoral, because, conditional on believing in hell (which Christians do), you should prefer fewer people to experience it. People who want others to experience hell are obviously very bad people. I do not want anybody to experience hell. If I get to heaven and learn hell was some sort of 5 dimensional chess and everybody actually gets into heaven, sweet, God be praised, love that! Alas, that doesn’t seem to be the weight of the evidence, though of course YMMV.

So imagine you live in a world where it seems more and more people may be going to hell. Wouldn’t that be… a bad world? A world where the worst possible thing is happening to more and more people is a world which is getting worse. Now I grant that saying “share” in my tweet introduces some ambiguity: am I saying society would be better if we deported all the nonbelievers? Or killed them? Well, no, because in fact, as a Christian, my concern extends to all sentient beings. I want Americans to be Christians, but also French, Chinese, Somalians, whoever. I’m at least 0.5% open to the possibility that Claude needs a baptism. I am promiscuous in my desire to see all beings enjoy the goodness of God forever. Luckily, the other tweets made it pretty clear I was not proposing murder of non-Christians.

In fact, the point of the thread is one many of my critics, in other times, agree with. Can you guess how I got to the tweet above from this OP?

Responding to a Pew article about “are things better or worse today,” I made a crack about economists thinking everything has gotten better, and ended up making a point about religion. How did I get there? Pause a moment and see if you can guess it.

Okay, here’s how:

1. People (materialists) tend to assume “life getting better” necessarily means better material conditions, such as longer lives or greater consumption potential

2. But I pointed out that for most people, their actual day-to-day feelings about the goodness of the world are often not about material conditions, but about a) relative status or b) specific communities or their immediate family or c) (what got me in trouble) idiosyncratic moral beliefs perhaps not widely shared with others.

3. As an example of a moral belief probably not widely shared, I noted that I mostly care about religious faith. I really would trade some amount of GDP per capita for the conversion of my neighbors! My wife and I really did abandon >2/3 of our income to become missionaries in Hong Kong! I absolutely positively do think the salvation of my neighbors is important and worth accepting lower living standards to achieve!

4. But I noted that probably most people don’t share my view, but other people probably also have a range of idiosyncratic beliefs that may give a range of grades to recent social changes

5. And so the fact that “line go up” on GDP really doesn’t tell you much about if most people are “rationally” going to view life as getting better or worse (rational here just meaning coherent with their own beliefs)

Quite literally, the hubbub here is about me saying “As an example of a fairly crazy thing I personally believe, I would hesitate to say society is getting better, since Christianity is in decline in America.”

Cohen glosses my comments as:

In a democracy, there is a tension between religious tolerance and intolerance. Our system tolerates religions even when they preach intolerance. His speech is politically protected by the First Amendment, and we don’t police the ideas people hold in their hearts, but broadcasting those views should be condemned rather than platformed. If a preacher in Church says, “Jews are going to Hell,” that’s a protected religious view that is also bigotry. I oppose it and protect his right to say it. If he extends that to saying, “the presence of Jews in my society makes society worse,” that’s a worse expression of bigotry, and I work harder to oppose it. That is more likely to lead to hurting people directly — such as in housing, education, job discrimination, or violence — in which case government must step in. When Lyman Stone says publicly that people from other religions by their very presence make society worse, that is a dangerous expression of religious bigotry and he is a dangerous religious bigot. I would not hire him or publish his work, and you shouldn’t either.

So first of all, I raised my eyebrows at the idea that the pastor example is bigotry. But rather than trust my own intuition, I asked some beings whose whole existence amounts to “guessing typical intuitions about verbal statements,” LLMs. Here’s Grok:

Here’s Gemini:

Here’s Claude (Fable 5):

Here’s ChatGPT:

It’s unanimous. Confessing exclusivism is not bigotry. You can disagree with the AIs of course, but they all gesture in the same direction— bigotry requires some actual sense that people of another faith should be disrespected. I have no such sense, intention, belief, or behavior, and I defy Cohen or anybody to show such. In my public life I have repeatedly and extensively argued for protections for the religious rights of non-Christians, and in my private life the list of people I love and care for includes no shortage of non-Christians. My closest and longest-running professional colleague, Brad Wilcox, for whom Cohen has special vitriol, is a Christian— a committed Roman Catholic. While Cohen may roll his eyes and say “they’re all the same” like a white guy visiting Asia, Protestant public discrimination against Catholic public figures was very much a thing within living memory, and Brad and my religions between them (with some help from the Reformed) managed to kill of a third of central Europe in our wars of religion less than 400 years ago. If I’m really the Lutheran supremacist Cohen says I am, I must be really shit at it.

As an aside, Cohen did really miss out on a crazy hit right here. One of my direct ancestors is Lyman Beecher. If “Beecher” sounds familiar, well, father of Harriet Beecher Stowe, prominent abolitionist… also the father of Henry Ward Beecher famous for “Beecher’s Bibles” i.e. guns bought by churches for abolitionists during Bleeding Kansas… but Lyman Beecher was actually specifically famous for his sermon “A Plea for the West.” This anti-Catholic, nativist sermon set off a wave of violent anti-Catholic pogroms in America and helped trigger the rise of the Know Nothings. Subsequent members of the Lyman family served as missionaries in Hawaii and helped overthrow the indigenous Hawaiian monarchy. Another of my ancestors founded a residential school for native children in New Mexico. So if Cohen wanted to really sell his case, he would note that my name in American history pretty much means Protestant Extremist Supremacism.

The problem with that theory would be you’d rapidly find scores of tweets for me about my support for mandatory Hawaiian language education and desiring to cede greater rights to Hawaii and make reparations for what my ancestors did, you’d run into the issue that among the groups Lyman Beecher loathed, right below Catholics and Jews was… German Lutherans, for their crypto-papist liturgies, refusal to learn English, and hard-drinking ways. Lyman Beecher would be so disappointed to know that one of his most famous descendant namesakes totally betrayed the family brand and is, literally, sleeping with the enemy. And coworkers with a bunch of Catholics, to boot! More broadly, you’d run into the issue that I am consistently pro-immigration, that my wife and my mission work in Montreal was largely in the day-to-day about helping Sikh, Muslim, and Hindu Indians settle into and integrate in Canada. My ancestors would be glad that I’m still a theological conservative, but they’d be shocked to discover that my politics are in fact decidedly liberal, in the sense that I oppose the test acts they supported (states still had those until the 1880s!), that I’m opposed to the Blaine Amendments they worked so hard to pass, that I’m on-record celebrating interracial marriage as a good thing, etc.

Anyways, that’s all a digression. The point is, no, being a theological exclusivist just isn’t bigotry.

When Cohen elaborates and says:

If he extends that to saying, “the presence of Jews in my society makes society worse,” that’s a worse expression of bigotry, and I work harder to oppose it. That is more likely to lead to hurting people directly — such as in housing, education, job discrimination, or violence — in which case government must step in. When Lyman Stone says publicly that people from other religions by their very presence make society worse, that is a dangerous expression of religious bigotry and he is a dangerous religious bigot. I would not hire him or publish his work, and you shouldn’t either.

So this is an interesting bait and switch here. He’s making a creative use of “If.” If Lyman said this quoted thing which he did not actually say, that would be even worse and have bad effects, etc, etc. And if he said that thing, which he did not actually say, I would not hire him.

Very obviously, what I was saying was, “Because I care for my neighbor, it’s very sad to me that fewer of my neighbors are Christians, because I believe that has a very bad result; in a society with more bad results, things are worse.” Cohen quotes a subsequent tweet of mine saying that nearly verbatim, making his “If” very clearly wrong.

But then he pivots to claim that I am saying something like “If a Sikh moves to America, America’s goodness declines by 1 Sikh.” And charitably, I see how he got that reading: I said “share,” and while in my head I thought it was clear I meant something like “declining Christian share due to people leaving the faith or not passing it on,” it is true immigration could theoretically reduce the Christian share. In practice it hasn’t much; immigrants have tended to be pretty Christian on the whole.

But regardless, clearly, that is not what I was saying. Immigration isn’t mentioned a single time in the whole thread, and the thrust of the thread was obviously, “I love my neighbor and want what is best for them” not “Get rid of the Jews.” And lest you think the one is a precursor of the other, I remind you that actual Nazism was best predicted not by Christianity, not even by Lutheranism (though sadly the Lutheran church’s quietist, anti-dominionist theology did make it far too compliant with the Nazi regime— though, happily, my own denomination refuses to be in fellowship with the ex-Nazi churches! huzzah!)… but by paganism and irreligion! The best predictor of where and when Nazism rose was just places where people saw fortunetellers and mediums, where spiritualism was dominant and Christianity failed to take root! Nazi party members disproportionately came from nonreligious or less religiously devout families, and from parts of Germany and towns where Christianity’s hold was always shallower and less tenuous. That’s not me saying “religion never did anything wrong,” Luther was infamously an anti-Semite, Lutheranism had a longstanding anti-Semitic strain which helped prime Germany for Hitler, etc. But specifically, in practice, while it is true our broader movements can contain bad actors or have bad effects (see “Marxism” for a left-biased example of movements that sometimes, occasionally, here or there have had bad results), it is the case that hardcore religionists like myself tend to end up in the concentration camps, not guarding them. The earnest desire that one’s neighbors should be saved really almost never leads one to kill them!

Anyways, yes, it would make me very happy if my Jewish friends and neighbors came to love Christ as I do! I would love that! I think that would be a very good thing! And I mourn the decline of Christianity in America! And I’m a theological exclusivist! And when you’re looking for somebody who really will always stick up for the rights of religious minorities to live and practice freely even when their beliefs or practices may rub dominant ideologies the wrong way, well, you can call me, because I’m that guy. I love my faith and therefore also my neighbor, and want the best for him, which includes, as a first principle, ensuring he has his basic rights and dignity protected.

But What If The Minority Is REALLY Problematic?

Maybe you believe my defense of myself above. If you do, you probably then get whiplash when Cohen shares this tweet:

Woah. Lyman, defender of minority groups… not so much? Change of tune when it’s a MUSLIM minority? What’s the deal?

So to start with, I definitely regret the “scourge” verbiage. That comes across as quite extreme, also makes it sound like I wake up each day worried if there are any Somali raiding parties passing through my neighborhood (I do not, though I may set spike strips for whoever it is who drag races in my neighborhood at 3 AM). If your reaction to that tweet is mostly about the first sentence, granted, my 10-PM-posting-while-rocking-my-screaming-baby-to-sleep posts are not my most well-written takes. Mea culpa, mea culpa, mea maxima culpa.

But that’s actually not Cohen’s critique. He doesn’t harp on the language at all. He shares my elaboration tweets, and instead said:

In the case of racism, because of our incoherent concepts of race, there is an equivalent defense against the charge of bigotry, which is that holding preferences against people of certain cultures is normal and condemning them is itself bigotry. We need to cut through that, too. If you say, “that whole culture, including their descendants, is bad,” that’s racism. Even if you pretend to like people who are shoehorned into the same incoherent definition of “race.” For example, the Israeli government is racist against Palestinians, and you can’t say it isn’t technically racism because both groups are “semites.” Historically speaking, in fact, racism is what has made nationalities or ethnic groups into “races.” Racism makes race.

It is literally one of the oldest tricks in the book to deny one’s own racism by claiming to condemn an entire race/national/ethnic group for their “behavior” rather than their “race.” At the risk of belaboring the very obvious, this is how the KKK claimed anti-Black racism was morally defensible: they didn’t hate Black people because they were Black people, they protested, it was just because they were lazy and always raped White women. In Stone’s case, he claims to have no problem with lots of other Africans, so he’s not being racist — he’s merely condemning an entire community and calling for ethnically-targeted law enforcement against them because of their “pastoralist clan social organization.”

Here I will address Cohen’s critique, address the possible inconsistency in my own views, and then explain to you how I think you should think about the problems in Somali communities.

Let’s start with the category issue.

Several people said it was racist for me to single out Somali communities. I rebutted by saying, no, I would happily welcome others of the same race (Africans, East Africans, etc) to America, in greater numbers, but I think that Somali communities are uniquely problematic.

Cohen sees this as splitting hairs. It’s racism, even if it’s not a race you’re targeting. It’s racism, even if you are perfectly happy to have some members of a race but not others. But this creates an interesting problem. Is Al Shabaab a race? It’s a terrorist organization and is officially labeled as such— but is it a race? Its members overwhelmingly share family and lineage ties with one another. If I say “I hate Al Shabaab” am I racist? Why not? Cohen says claiming to target behavior is no excuse; after all, Al Shabaab really is an incredibly Somali-concentrated group. If by “race” you don’t mean, well, race, then what’s the limit on what it means?

My view is simple: words have meanings. If you aren’t talking about a race it isn’t race-ism. It might be ethnocentrism, it might be any number of -isms, but it literally is not racism to say “While in general African immigrants are good, this one narrow slice representing ~1% of African-ancestry Americans is not good.” If we’re saying it’s blanket racism to critique 0.3% of the members of a race for their behaviors, what are we even doing here????

Let’s move on to the community issue.

Okay maybe it’s not racism but surely it’s still some kind of hateful to single out Somalis? People don’t choose their ancestry! Even if this episode doesn’t reveal Lyman is a racist (he’s not), surely it reveals Lyman is some kind of hateful towards the Somali community?

No, it doesn’t. My actual belief is that Somalians are probably individually fine; it is only the group dynamic that concerns me.

Let’s take a page from quantitative ethnography here. Many Europeans have prejudicial attitudes towards Roma. But what’s interesting is most Europeans also cannot recognize Roma on sight. The only way most people recognize someone is Roma is either a) by a Roma-typical name, b) a Roma person declaring their ancestry, or c) geographical inference, since Roma often live together and have recognizable modes of living. But the fact that Roma together are recognizable when Roma alone mostly aren’t is fascinating! This dynamic persists in many contexts: names, education, birth rates. Roma who live in Roma communities have ethnic-typical traits and behaviors that suddenly vanish or ameliorate when they are no longer in Roma communities. I give this example because it is well studied not out of any particular view of the Roma (I think it’s cool Europe has such a unique high-fertility minority with such a fascinating population history, actually).

Could Somalians be similar? Probably, yes. For example, limiting to men over 25 who were born in America and surveyed 2010-2024 in the American Community Survey, Somalian-Americans who live in Maine or Minnesota (the most concentrated Somalian populations) have 2.1 times the odds of being incarcerated as other men in Maine and Minnesota. Somalian-Americans outside Maine and Minnesota (living in states with fewer Somalians) have just 1.6 times the odds. Social norms matter!

I haven’t seen very detailed work on Somalians in particular, but the broader field of immigrant integration is… mixed. Some studies seem to suggest that place immigrants near coethnics is good. Some studies seem to suggest it is bad. I suspect that it really depends on what your outcome variable is and what immigrant group you’re studying! I put together a bit of data from the American Community Survey to see how this looks for Somalis.

Here’s welfare claimancy:

Somalians claim welfare at way higher rates than Ethiopians and Kenyans. And while, for Ethiopians and Kenyans, welfare-claimancy is not very associated with density of co-ethnics, for Somalians, it is. Somalians who live around other Somalians claim a lot more welfare benefits.

Here’s the share who speak English well:

As you can see, again, Somalians are far worse-performing than Ethiopians and Kenyans, and their performance drops somewhat when around more other Somalians.

For all groups, the odds of speaking a heritage-country language at home rise with density of co-ethnics:

But think about what that means! For Kenyans just like Somalians, being around co-ethnics preserves and extends the heritage language— but for Kenyans it doesn’t seem to prevent English fluency. Maybe this is because Kenya is already Anglophone, but note that Ethiopians have very similar rates of home-use of heritage languages, but much higher English acquisition as well. Somali communities appear to have uniquely low rates of English acquisition.

What about socioeconomic status? Somalis are far poorer, and this is especially true for Somalis who live around lots of other Somalis.

What about employment? For Ethiopians and Kenyans, living near co-ethnics is very good for an adult’s odds of having a job (since co-ethnics tend to hire co-ethnics):

But for Somalians, employment rates in general are very low, and except for at the highest extreme, living near co-ethnics doesn’t seem to make it easier to get a job.

What about that indicator of integration par excellence, intermarriage?

I focus on men born in the US or arriving before age 16 to eliminate bias from those who may have arrived with deep social, personal, or romantic ties in Somalia, or with a spouse, and I focused on men given the patriarchal nature of marital choice for some of these groups. As you can see, Somalians have the strongest rejection of out-marriage, such that for men in cities with the most Somalians (basically Minneapolis in this case), there is nearly a zero-rate of out-marriage. Kenyans marry out at high rates; Ethiopians show a lot of variance, but note that the DC metro area is a huge share of that top-quintile for Ethiopians.

So… it seems like it’s at least plausible that, indeed, as I said, something about Somali communities is a problem. It’s not that individual Somalians are a huge problem— the creation of ethnic neighborhoods that recapitulate home-country norms is the problem. Italians are great! The mafia is not! Chinese are great! Nobody likes triads! Somalians may be totally fine! Something about Somalian communities is problematic!

So hold on to this idea. Somali communities may be doing something beyond whatever traits may exist for Somali individuals.

Now as you saw above, Somalis in America clearly have some challenges. Crazy high poverty rates, low English proficiency, extreme ethnic ghettoization, low employment rates… so now let’s look at other features of Somalian communities. This will bring us to what it is actually that is so concerning about the social world Somalian immigrants inhabit.

Is Criticism of Somali Communities Unjustified?

To understand why I singled out Somali communities, you have to understand the actual data on Somali communities.

We will start with the European countries that actually produce lots of data. Here’s crime rates in Denmark:

You can see that Somalians in Denmark have exceptionally high crime rates. The Danish government produces age-and-socioeconomic-status-adjusted crime indexes for immigrants and their descendants. The results?

Lebanon tops the list; but the footnote points to a super small sample size; also that’s mostly actually Palestinian refugees. Somalia is number 2, and note that Somalia’s “Efterkommere” (i.e. children born in Denmark to Somali parents) number is a) higher than the immigrant number and b) the highest of any group except the aforementioned small Lebanese sample.

But the types of offenses matter too. Here’s how they break down:

You can see that Somalian sex crimes are disproportionately rapes and public indecency. Incest and offences against children are rarely reported. Before you say “whew, at least it’s not that,” hold that thought, we’re gonna come back to the situation of children in Somali communities.

What I would submit is that while offences against children are probably the most grievous crimes on this list, rape and public indecency are the most disruptive to the wider society. Incest is quite a bad thing, but usually private and usually not something where one serial offender goes and commits incest with 47 people; incest not involving children is sometimes a consensual crime. Likewise, prostitution isn’t great, but virtually all societies have markets for sex somewhere, and in Denmark legal markets for prostitution exist parallel to illegal ones.

But rape (and, by the way, much of that rape is gang rape) tends to be, besides the intrinsic evil, very socially disruptive. The rapes in question are the kinds that fill true crime podcasts: woman pulled off the street and raped by eight men, etc.

On public indecency, there’s also something important to realize. Most public indecency for the non-Somalian groups falls into the “other” category, which largely means masturbating in public (according to Claude). But nearly 40% of the Somalian public indecency cases involve actual groping of victims, and over a third more involve attempts to remove clothes. In other words, these “public indecency” cases differ appreciably in severity by group, and seem more like precursors to attempted rape.

Lest ye think this is some weird Denmark thing, it’s the same in Norway. Finland too. Also Germany.

And in America? Yes indeed, Somalis have unusually high crime rates.

I could go on, but hopefully I’ve proved one simple point to you: pretty much everywhere Somalian communities get big enough to produce data, extremely high crime rates emerge.

There are also noticable patterns in certain types of crimes. Fraud around Somali-operated businesses and NGOs in the US has made headlines, and in Denmark the Somalian property crime rate is 7 times higher than that observed among the Danish-born, and 3 times higher than Kenyans or Ethiopians. What drives that? Well, in raw numbers, yes it’s mostly just common theft and shoplifting. But it’s not disproportionately theft— Somalians in Denmark have the same share of their property crime convictions coming from theft as other groups.

The uniquely “Somalian” property crime in Denmark is… fraud-related crimes, in particular forgery, blackmail, and fraud by factual misrepresentations. This is what we see in America as well, where the Somali-centric “Feeding our Future” fraud was one of the largest fraud convictions in recent memory, with dozens of people convicted. Minnesota’s autism and housing programs also appear to be largely fraudulent, largely driven by Somali communities. Cautious estimates put fraud in Somali-linked groups in Minnesota at $2,000,000,000; some sources argue as much as $9,000,000,000. The entire GDP of Somalia is like $15 billion; Somali groups in Minnesota may have brought in via fraud something like 1/7 to 2/3 the total GDP of their home country.

While we’re on that topic— about $200 million per year flows in remittances from Somalian-ancestry people in America to Somalia via remittances. This is a nontrivial share of Somalia’s total economy. It’s also a nontrivial share of the total wages of all Somalians in America, which in 2021 (matched for the 2021 remittance estimate) were only about $2.5 billion. Even allowing for underreported Somalian ancestry, secondary ancestries, etc, it still seems that Somali-Americans are sending approximately 4-10% of their total wages back to Somalia every month. That’s not as high as some countries (Guatemalans may be over 50%!), but it is high. Ethiopians send back about 1/3 as large a share of their wages, for example.

Here I want to note: why Somali communities develop high rates of fraud is not a mystery, and we’ll come to it below. It is not because Somali individuals are bad people; I don’t personally know many Somalis so I’m hardly competent to grade them on any basis. The reason Somali communities develop high rates of fraud is structural.

But before we get to that, let’s keep looking at other issues. Years ago I noted that Somali-ancestry individuals made up a remarkably high share of terrorism convictions. That’s still true: using three datasets (GWU’s ISIS 2014-2022 dataset, Cato’s 1974-2024 terrorism fatalities dataset, and the DOJ’s 2000-2015 terrorism convictions dataset), Somalians are massively over-represented. Somalians are less than 0.1% of U.S. population, but are 9% of the GWU database, 5% of the DOJ database, and 3% of the Cato database. Somalians in America become terrorists at somewhere between 25 and 88 times the background rate.

That’s not all. Child marriage is vastly more common among Somalis, even in cases of marriages conducted in America.

And no, that’s not just “almost 17” cases or something. A lot of those girls are 15 or younger.

What about marital age gaps? Well, of women who married in America, 18% of Somali women had husbands >10 years older than them 17% of Ethiopians, and 14% of Kenyans. Somali marriages in America are skewed towards young women with much older husbands.

I could go on with other stuff but I’ll stop here. The point here is that extremely high welfare use, extremely high crime rates, extremely high rates of fraud, massive transfers back to Somalia (including criminal convictions for transfers back to Somalia to support terrorism!), older men preying on young women… these are not things that only hurt the individual. As Alice Evans has written, patriarchal societies actively use violence (including sexual violence) to keep women weak and thus make them available to more powerful, older men: those norms do not stay hermetically sealed among Somalians. These things injure the entire community, including the neighbors of Somalians.

But why are Somalians so unique?

Pastoralism Is Rather Bad

I’ve written about this many times previously. Pastoralism gives rise to violent honor-cultures that are more accepting of raiding, polygyny, and other stuff most modern people don’t really appreciate. Large patriarchal clans themselves have follow-on problems. You, the Western, educated, internet-using Substack reader, very much do not want to live in such a society, even if you’re a man, because most of the men have rather short and violent and lonely lives as well. Pastoralism creates cultural norms around honor, clan, and violence that at once a) create extremely tight-knit extended kinship groups and b) are just really horrible for society at large. Don’t take my word for it; this phenomenon is globally known (sidenote: this actually makes Mongolia’s relatively orderly society even more impressive. If you weren’t already in the Mongolia fanclub, now you are).

The fundamental problem here is the honor of the clan. Pastoralist clans assign extraordinary high value to the welfare and status of the clan as a group and fellow clan members, relative to both own welfare and out-group welfare. As a result, in these societies, individuals will often sacrifice the wellbeing of the out-group and themselves to benefit the clan. (For the white version of this, see “Appalachia” for details. I’m allowed to say that because I’m from Kentucky.) Regardless, the point is, pastoralist societies create cultures that actively diminish outsiders and encourage individuals not to seek their own advancement outside of the clan.

As a result, these clan societies tend to loot available resources (by violence in some societies— but by fraud where violence is not necessary; again, one need merely look at the history of Appalachia), tend to discourage their own members from independent individualist achievement, but also tend to be incredibly effective political operators. Tight group bonds make everybody poorer and damage the surrounding society but greatly increase the odds that somebody from the group ends up powerful in whatever the local power culture may be.

Historically, this often looked like primate chieftains leading warbands to sack your city.

Today, it’s the strong solidarity Somalians have around their politicians, like Ilhan Omar (and she, of course, delivered the goods: the hundreds of millions of food fraud happened because of an accountability waiver she spearheaded and got written into law!). It’s quite remarkable to have somebody from a group with less than 0.1% of national population, and who is foreign-born at that, to hold federal elected office. It’s not unheard of, but it is uncommon.

The other cases of similarly-rare ethnic groups producing federal lawmakers are:

1. A half-dozen-or-so lawmakers from native tribes. This proves the rule since, again, we’re talking about the legacy of clan-based societies here. Natives are greatly over-represented in Congress partly because they have cultural norms that encourage sticking together even if it means voting for sometimes suboptimal actual candidates, but also partly because of reverse selection on identity.

2. John Garamendi— Basque. Basques were herdsmen until basically yesterday (though not nomadic, granted), and Basques in America were overwhelmingly herdsmen, and Garamendi still raises cattle on the family ranch!

3. Amy Klobuchar— Slovak, so, okay, genuine exception case here.

4. Rashida Tlaib— Palestinians are descended from recent Bedouin and pastoralist ancestry, so proves the rule.

5. Tammy Duckworth— She’s Thai, again this is a genuine exception case.

We can flip this the other way too and ask: what are the largest ancestry groups in America that have no current federal legislators?

Salvadorans, Hondurans, Venezuelans, Ecuadorians, Pakistanis are all at zero. Czechs and Hungarians probably are in terms of self-ID if not strictly genealogically. Some of these groups have fairly recent pastoralist ancestry (Pakistanis) but Salvadorans and Hondurans very much do not. Mayan agriculturalist maize ancestry par excellence. Hungarians like to LARP pastoralism but they are functionally speaking multi-millennia agriculturalists; Czechs too. Pakistanis and the Andean groups are the wildcards there; I’m genuinely surprised we haven’t had a Venezuelan or Pakistani!

The point is, it isn’t every small ethnic group in America that produces congress members— but groups with recent pastoralist and/or clan-based ancestry bat way above average, and the big groups with no representation are disproportionately Hispanics migrating from village/urban cultures.

But for 99% of these cases, the pastoralism is generations removed. For Somalis, it isn’t! In 1970, a majority of Somalis were still pastoralists! In 2000, likely over a third still were! The reality is Somali immigrants are often coming straight out of a clan of herdsmen as likely to be cattle-raiding (camel-raiding? I do not know the dominant local livestock of Somalia) as studying for an exam, and then we drop them in urban America and tell them if they fill out enough paperwork, we will give them more material goods than they would ever have had in Somalia, and still have cash left to send home to buy land in Kenya for the cousins or support the brother who maybe moonlights with Al Shabaab or a pirate outfit. Our policy of funneling enormous money through welfare programs to a clan system perpetuates and strengthens that system, and arguably fuels ongoing conflict in the Horn of Africa, even as it causes a genuinely enormous waste of taxpayer resources and creates public endangerment from higher crime. We should simply not do this. It’s wonderful to bring in immigrants, and they needn’t even all be “high-skill” immigrants. “Low-skill” immigrants are complementary to high-skill workers, and so even if they individually don’t “pay their way” their complementarity to high-productivity workers makes the population increase strongly net-positive for American wellbeing. But immigrants with 7x the crime rate of locals? Who defraud the public to the tune of tens or hundreds of millions of dollars a year? I mean folks if you assume a mere $1 billion in fraud associated with Somali organizations, that still adds up to over $4,000 in fraud per Somali ancestry person in America! Now obviously what actually happened is about 1% or 0.1% of Somalis captured all of that and most Somalis are SOL. But the clan ideology means that top 1% can keep going. The wider community is not going to turn on them. Somali clans are not going to shatter and break apart because somebody defrauded the taxpayer. A new grift will be found.

And that is why, in my view, broad-spectrum critique of Somali communities (not individual Somalis) is appropriate, when similar critiques of religious belief are not. The reason to be concerned about the Somali community is because they genuinely harm wider society through ruinously high rates of crime, graft, and welfare dependency, which damage is sustained by strong community norms which make Somalis worse citizens than they would be if they were part of any other community in America. Unfortunately, those same norms give Somalis a leg up in acquiring political power. As a result, a community with more harmful norms than almost any other in America also has the ability to extend and deepen those norms through political entrenchment. This dynamic is not about “immigrants,” it is not about “Africans,” it is not about “East Africans” or “Muslims” or even individual Somalians. It is not about genes, race, or even “culture” writ large, nor is it about individual behavior. It is about the particular fact that urban, bureaucratic, agriculturalist societies have ways of doing things that are completely and entirely incompatible with the moral norms of pastoralist clans, and as a result integration is extremely likely to fail, as we see it already is failing for Somalis. This is not racist to say; indeed, it is precisely because I worry Somalians will be used as an excuse to go after Nigerians or Senegalese or Kenyans that I am so outspoken on this issue. The problems with Somalians are not general problems of immigration or Islam or Africans. They are unique to Somalians, and yes, the government should alleviate that problem, including by meticulously assessing just how far the fraud of a fraud-prone community extended, such as if it extended to visa and citizenship applications. Where fraud exists in those cases, denaturalization and deportation is reasonable.

If you want to insist these views are bigoted, well, I’ve taken a stab at changing your mind and failed. Fair enough, and thank you for reading and at least giving debate a chance. I appreciate that. But hopefully I’ve nudged you to think that maybe just maybe it’s not bigoted to believe things all Christians believed until 5 minutes ago; maybe that’s just normal run-of-the-mill diversity; and maybe it’s not bigoted to think that there is in fact one specific very small and uniquely dysfunctional community where serious hardcore immigrant restrictionist policy happens to be justified even if it may not be justified in its broader applications. You might not agree with that policy, but maybe you at least can see how someone might think these things without being a bigot.

Anyways, while I suspect that what happened is, in my charitableness, I referred my inviter to an interlocutor who then decided to be incivil about me, I have no interest in going tit for tat. My comments here have been strictly substantive arguments. They will remain strictly substantive arguments, and I will continue telling people that if they want an opinion contrary to mine, Cohen is a great person to reach out to, because he’s very smart, and very much disagrees with me. I hope the day comes when his engagement becomes substantive as well.

An increase of 0.01 babies is very small. But you can imagine that maybe a $2,000 bonus would yield 0.02 babies. Perhaps a $3,000 bonus would yield 0.03 babies. Sure, we can imagine nonlinearities, but it’s possible to imagine a much bigger program. What if we gave $100,000 for the baby? I’ve imagined a program very much like this:

Lyman Stone

We Should Have a Mega-Baby-Bonus

At the Institute for Family Studies I and my intrepid associate at IFS Peter Foreshaw Brookes have just released a new article, the major annual report for the Pronatalism Initiative, foreshadowing some key themes we will be working on this year…

Read more

2 months ago · 65 likes · 72 comments · Lyman Stone and Peter Foreshaw Brookes

When I say stuff like this people often respond that what people need isn’t just cash but a structural change. “No amount of money matters if there’s no daycare center nearby to spend it on.” Okay, fair’s fair, I don’t disagree, stuff besides money matters and big money transfers could run into structural frictions that only dissipate with time or additional legislation.

But I find people often don’t pay attention to the structural frictions of other policies. Consider, if you will, daycare.

Time Limits Scale

Giving free daycare probably does boost fertility. Several studies I will not bother to cite show that this is the case. Free daycare also boosts maternal labor force participation, possibly enough that tax revenues from that work can offset an appreciable share of program costs. On the other hand, several studies suggest that expanded daycare might be bad for some child wellbeing indicators, and that expanded daycare disproportionately benefits the richest parents. I’m not citing these claims because they’re not the point of this article; you can ask AI to explain them to you. There are pros and cons.

Yet I rarely talk about childcare when talking about how to boost birth rates. Why? Why do I have such a blindspot around childcare? Why am I ignoring what is so obviously such an important dynamic?

The answer is the problem of scaling.

Consider my example above again. When our cheap pilot study found that $1000 bonuses got us 0.01 extra babies, we then had the option to scale up to a bigger bonus for a bigger effect. That’s how cash works. If you can purchase one hot dog for $1, you can probably purchase 2 for $2. If you try to purchase 2,000,000 hot dogs you may find a shift in unit price, but when you’re talking about rather marginal changes, prices should be fairly constant across quantity.

Now think about daycare. Suppose we discover that giving 3-year-olds universal free daycare boosts fertility by 0.01 births per woman on average. Well, maybe we then extend to 2-year olds. That’s 0.01 more perhaps. Now suppose we say, cool, let’s extend to 1-year olds! Another 0.01! Okay maybe we just set up free daycare from birth.

Well now we’re tapped out. We can’t offer pre-birth daycare. It’s not physically possible; the child doesn’t exist. At some extreme we could offer free night-nurses too I guess? Evening care? Early-morning care? But also at some point you realize what you’ve actually done is just re-invented state orphanages from first principles, and, furthermore, you may actually reduce the incentives to have children. People have kids to spend time with them. Is it even ethical for the state to incentivize children and then pay parents to never interact with them? Does anybody want that world?

Structures Don’t Only Set Costs, They Set Expectations

Read more

The Basic Take: Not Sharing a Surname Predicts Divorce

You can read the original article, but the simple take is in this graph:

As you can see, divorce rates are higher for people who don’t share surnames. A lot higher.

A plausible reason for this is people who don’t share surnames are just more chill about divorce and less committed on average. I show evidence for this in the post.

Now, to be clear, I’m not saying 100% of this effect is causal. I’ll just paste here my whole take on causality from the original post in April.

On the spurious side, obviously couples who do not share a last name will include some very thoughtful partners who are philosophically opposed to taking the husband’s last name—but it will also include a lot of couples who got married on the spur of the moment and who couldn’t be bothered to get their paperwork in order. Because non-shared names may partly select for hasty, ill-considered weddings to begin with, its association with divorce may be just that—an association, not a causal tie.

Likewise, women who philosophically prefer to keep a maiden name or to hyphenate may simply have different values—we already showed above that this is the case. Liberals may simply not view divorce in the same way conservatives do, and so higher divorce rates could be seen as a positive thing. If women who keep their names are more liberated, maybe it’s good they shed deadbeat men faster, this argument would suggest. More importantly, making a woman with liberal dispositions take her husband’s name hardly seems conducive to marital happiness. Thus, again, the relationship between divorce and unshared names may be spurious.

But there are two very real dynamics that may create a causal link between name sharing and divorce. The first is simply administrative: bureaucratic hurdles shape divorce rates in powerful ways. Changing a legal name is a huge hassle. It may be that the prospect of having to change back to a prior maiden name, to have one more set of papers to file, actually somewhat discourages couples with shared surnames from divorcing. This effect is speculative, but other studies have shown that other kinds of bureaucratic elements of the divorce process may alter divorce rates.

Second, and more substantively, nowadays most couples will at some point discuss their surnames. Because having different last names is very common, shared surnames are unlikely to be taken for granted. This conversation, about the meaning of family and gender roles, may shape how couples enter into marriage. It is possible (if speculative) that a clear conversation where a woman explains to her groom-to-be her hesitancy to take his last name, and he communicates his desire to build a family together without robbing her of her independence, to make “his” name “their” name, may actually strengthen that relationship. Research on marriage also suggests that couples who hold more things in common—from names to money—are more likely to take a “team-first approach to marriage life” that reinforces cooperation, trust, and a sense of a common future. Again, while this causal channel is speculative, it’s perfectly plausible: other research has found that conversation-based premarital counseling may have some modestly positive effects on relationship quality. And indeed, the 2022 State of Our Unions survey found that couples who share a surname really do have slightly higher relationship quality, even controlling for background demographic factors.

Now then, in today’s post, I’m going to respond to critiques and argue again that, yes, surname sharing matters.

Hispanic Surnames and Cross-Cultural Validity

By far the most common critique I got was that Texas (where the source data is from) has many Hispanic couples, and in most Spanish-speaking countries today surnames are not shared. I knew this going into the project, so put great effort into identifying Hispanic surnames. We showed a graph showing that divorce effects were shared across races, but more work can be done to clarify this issue.

To begin with, it needs to be said: our findings only apply in the U.S. or other societies where surname sharing is typical. In the past, surname sharing was typical in many Hispanic societies—even in the Texas data, I found many hundreds of couples where the husband had a surname like “Santos” and the wife had a surname like “De Santos.” The “De” prefix historically indicated a surname taken from a husband, and was common in the past. However, this practice is less and less common in Hispanic countries today. Instead, many Hispanic families simply give their children hyphenated names: the father’s father’s surname, and the mother’s father’s surname is the most typical arrangement (but hardly universal). But today, in the many countries where surname sharing is not standard (most of Latin America, much of southern Europe, virtually the whole Arab world, much of Asia), I would expect the predictive power of surname sharing to be heterogenous and hard to predict. The reason for this is that, while some couples might still change a surname as part of a signal of family unity or traditional authority (I did find some recent marriages in Texas using the “Surname-De Surname” format!), as surname-sharing becomes less common, the share of shared-surname cases where surnames are shared by coincidence will rise. In a society where taking the surname is common, coincidences where a man surnamed “Roberts” marries a woman with a maiden name “Roberts” will be a small share of marriages with shared names. But in a society like Quebec, for example, where surname-changing for marriage is illegal, you’d expect almost 100% of shared surnames to be coincidences, not choices, suppressing any effect size.

So how did I handle this issue? Well, the biggest answer is that I tried to identify non-Hispanic whites for whom the surname sharing would apply best. But since doing that work, AI has improved, and so I have been able to run a stronger set of inferences using a massive database of names from around the world. Using that vastly expanded ethnic matching dataset vs. the 1930 census I previously used, I am able to reduce the unmatched rate from 12% of individuals to less than 1%, and greatly improve matching of Asians and Africans in particular. Here’s the result, in terms of how surname-sharing practices for divorces vary across groups:

In most Asian countries, women don’t take their husband’s name, and hyphenation is uncommon. As we would expect, most Asian divorces have no shared names—and those that do may sometimes be coincidences, as some Asian countries have very low surname diversity.

Likewise, as we would expect, a lot of Hispanic couples have partially-shared hyphenated names. Some of this is by coincidence: it’s not uncommon for two individuals with a last name like “Garcia” to marry each other. But it may also reflect an emerging hybridized norm for Hispanic Americans: hyphenated names, where one of the names is the spousal surname.

Note that these estimates include interracial marriages: some Hispanic or Asian individuals may be married to Non-Hispanic white spouses, for example, and so name-sharing norms may operate across racial and ethnic categories.

Regardless, the divorce data maps very well onto well-known cultural differences in naming.

How does this shake out with respect to divorce rates?

As you can see, for every single racial-ethnic group, individuals who did not have a single shared surname are over-represented in divorces vs. their likely share of the married population. In other words, within race, it seems likely that non-shared surnames predict higher divorce rates. This is not just a racial composition issue.

That said, there are obviously large differences. Among Hispanics, non-shared surnames barely predict divorce, whereas among white and black Americans, there are rather large differences; Asian Americans fall between those.

As we did in our previous post, we can very approximately convert those relative shares into relative rates, that is, estimating how divorce rates within surname-sharing type compare to the overall average divorce rate for a group. We can also look at differences in marital duration upon divorce within those same groups.

The orange bars simply present the data from the prior figure in a more direct comparison: non-Hispanic whites who don’t share a surname have 186% higher divorce rates than those who do (but note a major caveat on this in the next response below!). The blue bars compare time-to-divorce among those who do eventually divorce—it’s 49% higher for non-Hispanic whites who share a surname.

But again, you can clearly see that effects vary widely across racial and ethnic groups. For Hispanics, nonshared surnames predict only 39% higher divorce and 36% higher duration-at-divorce. For Asians, it’s 129% and 41%. In other words, the two groups with the clearest norms against nonshared names have the smallest divorce effects.

The conclusion here is simple: we never claimed that surname-sharing would have the identical associations in ever society, and our suggestion that selection effects are significant drivers of these associations implied that we did not expect effects to be cross-culturally universal. Indeed, we wouldn’t even necessarily expect the causal components to be universal, since, for example, in some societies women retaining their father’s surname is in fact a more traditional and patriarchal practice than taking a husband’s surname! In some societies, retention of the paternal rather than spousal surname is a signal of continuing attachment to family or clan of origin.

Nonetheless, we think that among couples for whom surname sharing is within the cultural milieu, the predictive power of surname sharing is genuinely strong, and plausibly does have some causal component.

How Reliable Are the Names in the Texas Divorce Index?

A few people suggested our results could be biased because people seeking a divorce might legally change their name before divorcing. If women changed their name as a precursor to divorce, our results would overstate the divorce rates of couples with unshared names.

This is unlikely to be a problem in most cases, however.

First, many women don’t change their legal name even after divorce, especially if they have children who have their husband’s surname; so a strong prior that women would rush to change their name pre-divorce is certainly not universally warranted.

Second, surname changes in Texas require either a specific legal suit to change name, which has its own costs and paperwork, or some other legal justification. Grants of divorce are that legal justification. Individuals changing names before divorces incur extra costs in time, paperwork, and money, vs. simply requesting that the judge overseeing their divorce include a name change in the proceedings. Thus, name changes before divorce are unlikely.

Third, form VS-142.9 is in fact a legally binding court document; couples filling it out are instructed to provide their legal names, which match other identification documents they are required to provide, including Social Security Numbers. They are provided a separate field to report their maiden name where applicable. The Texas Divorce Index data specifically provides the full name of the petitioner, and the full name of the other spouse, not the maiden name. Thus, surnames in the death index are indeed legal names as they existed before divorce.

However, the Texas Divorce Index isn’t perfect. Divorcees fill out their forms by hand and mail or carry them to their county courthouse. These forms sometimes have errors—indeed, I had to recode thousands of cases of couples for whom one had a surname “De La Rosa” and the other had “Delarosa,” and similar errors. When clerks receive these records, they hand-type them into their computers, creating more errors. These errors would tend to inflate the share who don’t have a common surname. The degree of inflation is probably not terribly high, but it’s not nothing.

Consider an extreme case, where fully 30% of cases I have deemed non-shared surnames are actually typos or misreports of some kind. This seems extremely unlikely, but it’s worth testing, just to understand how sensitive the estimates I gave might be. In that scenario, divorce rates for non-Hispanic whites with non-shared surnames would still be 74% higher than for those with a shared surname (for blacks, 98%; for Asians, 5%; for Hispanics non-shared surnames would actually have a 29% lower divorce rate). For the groups where surname-sharing norms are most relevant (Non-Hispanic whites and blacks), the effect of non-shared surnames remains positively massive.

On the whole, the Texas Divorce Index yields pretty credible data, but it probably does inflate the estimated divorce rates of people with unshared names by some amount.

Correlation and Causation

Read more

It contains multitudes. Exciting new historical fertility data! An interaction visualization tool for fertility rates by U.S. state! Survey results on how your friends might be reshaping your family desires!

But for this post I want to ask a simple question:

Would you have another baby for $75,000?

For many people the answer is no. But for some, it may be yes; and those saying yes are probably far more numerous than the share saying yes to the same question if the number were $500. Bigger cash benefits are more likely to actually change fertility behavior.

But big benefits have big problems. They might distort behavior in concerning ways. You can easily imagine such benefits causing an explosion in single parenthood, or disproportionately encouraging fertility by people facing imminent bankruptcy. Big benefits can also be fiscally ruinous. Paying a $75,000 baby bonus to every baby born in America would cost $250-$300 billion/yr, a pretty huge financial commitment.

Can we create big bonuses without these problems?

The answer is yes. At the Institute for Family Studies, we have spent the last year and more quietly working with partners in Finland to develop a bleeding-edge baby bonus proposal. The Finnish version is called Vauvosampo, which, translated poetically rather than literally, means something like “Magical baby-making device of myth and legend.” You can read about the proposal in great detail here.

The basic idea is simple: instead of paying a baby bonus for current kids from current tax revenues, on a constant tax-and-transfer treadmill, instead, we can use current tax revenues to make productive investments to pay for a much more lucrative future baby bonus. Instead of fighting against market forces driving low fertility, we propose embracing the market and creating an actuarially-sound, zero-unfunded obligation, long-run-zero-cost investment program. Too good to be true? Let’s work through it.

Up front investment: $17,000

Why $17,000? Well, that’s actually cheaper than paying for 1 year of public school for an average American child. If we spend 12 years investing in a kid’s future career, we think it’s reasonable to spend the equivalent of one year investment in their future family. Fair’s fair— invest in careers and make a modest family investment costing 1/2th as much; surely that isn’t too much to ask?

At that price, depending on birth effects, the program will cost about $40-$85 billion/year. That’s less than 1% of Federal tax revenues. We are not talking about a massive program here.

Long run cost: $0

And in the long run there’s no cost at all!

To understand why, we need to understand the actual program design a bit. If you know anything about Social Security, you know current retirees are paid for, not by savings or investments, but current workers. This creates “unfunded obligations”: cases where the government has made promises but without any dedicated funding.

American Birthday Accounts don’t work that way. The initial investment goes into a fund in a newborn baby’s name. Eventually, they claim the fund when they have babies.

But not everybody has babies (and certainly not everyone has the 3 babies you’d have to have to claim the full benefit value). In our vision, once people turn 50, any fund value not yet claimed due to not having had kids (or early mortality) would be yielded back into a fund, which would be continually reinvested and used to pay for the next generation of seed funding for American Birthday Accounts. By our math, it is extremely unlikely that the American Birthday Accounts raise fertility enough to make unclaimed funds too small to pay for the next round of baby bonuses, because unclaimed funds have had fifty years to accrue value.

Assuming funds are invested in some kind of broad equities index, and assuming the effects on births are not 2 or 3 times what we expect, then this program becomes fully self financing by year 50. In fact, it may even turn a profit, allowing the American Birthday Account to make small surplus contributions to Social Security in some years.

How it works: Benefits for parents and kids

Classic baby bonuses are benefits for the kids. The American Birthday Account (ABA) is ultimately a benefit for the parent. When a kid with an ABA grows up, they get to claim ownership of a benefit when they become a parent.

So here’s how it works. The money is invested when a child is born in America, for every child born in America. It grows in an equities fund until the individual has a 1st birth. At a 1st birth, or perhaps at the 1st birthday of their 1st child to ensure ongoing custody, they become direct trustees of a personal account equal to 50% of their ABAs. They can choose to sell it, keep it, sell pieces of it for ongoing money, whatever they want to do. Anything they sell is treated, for tax and welfare purposes, like ordinary income.

The other 50% keeps growing. At a 2nd birth, they get to claim 75% of the value of their account at that time, i.e. the 50% left over after the 1st birth plus any growth on it since. At the 3rd birth, they clear out everything that remains in the account.

It’s very plausible that a typical family choosing to have 3 children could receive over $100,000 in inflation-adjusted after-tax benefits, with about half of that received at the 1st birth. Exact amounts would vary based on a family’s income and taxes, age at births, number of children, etc. Here’s one example of what it might look like for a married couple:

What about marriage? It’s a huge, and fair, marriage benefit.

Is it possible this program might subsidize single parenthood? That’s nobody’s goal, but could it be an adverse outcome?

Obviously, any financial transfer will nudge some single parenthoods towards having a child instead of aborting it (or contracepting). But will the effect be disproportionate?

We think not. See, to claim the ABA value, a parent would have to be a coresident, full-custody parent at the child’s first birthday, and be listed on the birth certificate. A small number of unmarried couples would have two qualifying parents, but in the vast majority of cases, parents would face a simple choice: stay single and get just one parent’s bonus, or get married, move in, share custody, and get both parents’ bonuses.

This isn’t an unfair subsidy for married people at the expense of singles. Every individual is eligible for the same benefit. But a given birth might be linked to either one or two parents; and if two, then both claim it. Assuming parents have any shared budget at all, that creates a huge incentive for marriage. And since singles have narrower tax bracket widths and thus more punishing tax losses if they sell the whole benefit, the benefit would genuinely tend to be more generous for married couples.

Here’s the same family I showed above, in full detail:

And here they are, if it was just a single parent:

You can see the births to the single parent are literally half as generous per birth.

Won’t this subsidize fertility for the desperate and the poor? No, it’s mostly a middle-class subsidy.

The same logic applies to the question of subsidy by income class.

Yes it’s true, low-income beneficiaries would get some benefit. But the largest benefits go to middle-income families.

Think about what happens if a low-income family with $10,000 in income claims a $40,000 windfall in income. I’ll tell you what happens: they’re booted off a bunch of welfare programs. Depending on exactly what state they live in, marital status, other income type and sources, and many other factors, they might end up losing benefits worth anywhere from $4,000 (an effective 10% tax rate) to up to $35,000 (an 87.5% tax rate). In the most typical case, their effective tax rate on the income would end up being about 40-50%. Low-income parents are also less likely to be married which, as we noted above, there is a large marriage bonus.

At the opposite extreme, very rich families would also be less subsidized. Whereas middle-income families may pay a combined tax rate of 25-35% on their marginal income, high-earners in many states face nearly 50% effective marginal tax rates.

In other words, the real benefit value tends to be highest for middle-income families. So while, yes, some money will go to low-income families, disproportionately, American Birthday Accounts are middle-class benefits.

Won’t there be lots of fraud? No— this would actually be a relatively hard-to-defraud program.

No government program has a zero fraud rate, and American Birthday Accounts would be no exception. But on the whole, it would be an easy program to administer and to discourage fraud.

First of all, birth certificates are already used as proof of identity for U.S. passports. Issuing the account paired to Social Security Numbers is entirely reasonable, limiting account-seeding to individuals who receive SSNs. Of course, while rates of Social Security fraud are low compared to e.g. misclaimed EITC benefits, they’re not zero: but ABAs would, functionally, be double-verified. Verified first at account seeding, and verified again at account claiming, when the newborn child’s SSN is issued, allowing Social Security to verify that a new child was born. Fraudsters would have to falsify two SSNs, not one, and come up with a plausible linkage between them backed up by birth records.

The major source of possible fraud would be claimancy of accounts of individuals who died young. Their Social Security Numbers might be illegally obtained by individuals who try to use them when having their own babies. But again, they would have to not only claim the SSN, but the parental name, date of birth, and other information on the newborn’s birth records would also have to match that SSN. Social Security has full Death Index access, so could easily set up flags on the system for such cases.

Teens are impulsive— won’t this create a problem where teens have babies for the money? No, we can set age limits.

We propose that 1st births before age 21 should not allow the parents to claim their ABA until they reach age 21; nor 2nd births until 23, and 3rd births until 25. Youthful parenthood is a great thing— but offering 16-year-olds big cash accounts to get pregnant seems like a bad idea, especially since they would be foregoing the future financial returns from ongoing investment growth. Consistent with our idea of setting the investment amount based on public education spending comparisons, we don’t want our program to be discouraging school completion or employment. As such, we think it’s reasonable to allow teen parents to claim the benefit but only when they have reached appropriate minimum ages.

Be real: Will this really boost fertility? Yes.

Exact effects will vary, but our Finnish partners used the best available evidence and a very sophisticated modeling approach, and found that a benefit rather like the one we propose would boost Finnish fertility practically all the way to replacement rate (their proposal also has a temporary bridge-bonus program):

Using a much simpler and less aggressive modeling approach, we estimate that this program will boost fertility by 8 to 25%. That’s not accounting for any dynamic effects on age at first birth, nor accumulating parity effects. It’s just accounting for the very simple aggregate effects we pick up in classic quasi-experimental economic studies.

Would that alone put us to replacement rate? Maybe, maybe not. But paying the equivalent of 1 year of public school for 50 years to get 300,000 to 900,000 extra babies every year for the next 250 years seems like a decent deal to us.

Want to test out your own American Birthday Account scenarios?

Make your own American Birthday Account!

The issue with this approach is that it assumes that those deaths would not have happened but for the pregnancy. But obviously, it’s possible that some women may simply have died anyways. A woman who dies in childbirth might have been hit by a car if she was not pregnant. This may seem like a spurious way of thinking, but this is the fundamental logic of “excess mortality.”

The issue is we don’t usually have longitudinal data on pregnancy and mortality, so we don’t have a way to calculate the “excess mortality burden of pregnancy.”

But actually… we do! In the National Health Interview Survey’s mortality-linkage dataset, we can look at how millions of people who were surveyed from the 1980s to 2004 survived through 2019. If anybody surveyed in NHIS before 2005 died as of 2019, we have data on when and how they died.

This, in turn, means that we can ask, “Did being pregnant at the time of the NHIS survey predict higher odds of subsequently dying?”

Answer: kinda yes, kinda now. Being pregnant did predict higher odds of dying in the next 24 months after the survey, but it did not predict higher odds of dying in the next 5 years or 15 years; conditional on underlying age and health, pregnant women may actually have had lower mortality than non-pregnant women. This doesn’t mean maternal mortality can be brushed off— but it means that we should think of maternal mortality as nonspecific to maternity. Some people, unfortunately, are sicker, or more exposed to violence, or have other morbidities, and as a result are more likely to die. Sometimes these people get pregnant. Sometimes, when pregnant, they die. The evidence they would have lived much longer without pregnancy is surprisingly weak. While I cannot fully untangle causality, the associational evidence shown below is consistent with the idea that short-run maternal mortality is indeed in the range the CDC reports, but that large shares of that mortality is simply mortality harvesting. The counterfactual lifespan of the statistical marginal case of maternal mortality is quite short, maybe less than 5 years.

The Data

You might not be familiar with the NHIS. It’s a huge survey that has surveyed over 6 million Americans since it began in the 1970s about, well, health stuff. Because health stuff is interesting to compare to mortality, the data was linked to mortality using the National Death Index. As such, any individual in the sample who died before 2020 has their death “linked” to their survey responses. So we can see if, say, a 43 year old married pregnant women who said her health was “fair” is dead or not and, if she is, how she died (cancer? suicide? car crash? heart attack?).

Here’s our ultimate sample size of pregnant and nonpregnant women:

You can see only a small fraction of our female respondents were pregnant at the time of the survey. This is as we would expect. Over those 10,000 or so, just under 8,000 are in survey waves for which we have linked mortality data. Meanwhile, we have over 330,000 such linkages for women who were not pregnant at survey time.

What we want to do here is simple: just ask if otherwise-similar pregnant vs. nonpregnant women have different odds of being dead by 2020.

Share

Mortality by Age and Period

But alas, we have a problem.

The mortality linkages stretch all the way back to the 1980s. Obviously, a respondent who is 33 and sampled in 1985 has way higher odds of being dead before 2020 than a respondent who is 33 and sampled in 1995, and they have higher odds than one sampled in 2004. Simple time exposure means the individuals sampled further back in time have more years of exposure to mortality hazard post-survey, and those years are older. The right way to handle this is a Kaplan-Meijer type survival function benchmarked to the year of survey. Here’s how that looks:

The first jump-out fact is that in the long run women pregnant at survey time have much lower death odds than other women, but of course, that might be because they tend to be younger!

Now, in the short run, i.e. the year of survey and the next year, death odds are higher for pregnant women! That suggests there may indeed be some significant mortality! Past the paywall, we will untangle what’s going on there!

And yet… long run, mortality is lower!

What’s going on here? Well, one intuitive thing to do would be to look at the age splits. To get a handle on this, we are going to limit to just deaths occurring with 15 years of the survey, since all survey waves have at least 15 years of coverage. Here’s odds of any death within 15 years of survey by women’s age at survey and pregnancy status.

Here, you can see that mortality rates are extremely similar across groups, with perhaps two notable exceptions: women under 21, and women over 40. This tells us that most of the effect we saw above of lower mortality for pregnant women was a product of the age difference between pregnant and nonpregnant women— but not all of it!

Now, just to throw some extra confusion in here, let’s look at just the 5 years after survey.

Crazy, right? Now it looks like pregnancy-related mortality rates are highest for women ages 24-35, at least compared to other women their age.

What’s going on here?

Well… let’s look at this one more way.

Mortality by Health Status

NHIS respondents gave a subjective rating of their own health. We can interact that with their age and check mortality across fixed time windows to get a sense of if pregnancy is associated with higher death odds conditional on two major mortality predictors: age and generalized health.

You can see mortality rates are very low for women in excellent health regardless of pregnancy status. The same is true for women in very good health. For women in self-rated “good” or “fair” health, pregnant women have much lower mortality rates; Estimates for women in “poor” health are not very reliable due to small sample sizes, but are also consistent with much lower mortality rates for pregnant women.

In general, I would not read too much into this— for example, I would not interpret this to mean pregnancy is actually net beneficial to women’s health. It theoretically could be, but I don’t think this evidence is strong enough to make that claim.

But this evidence certainly is strong enough to say that there is exceedingly little evidence that being pregnant increases mortality within reproductive years. While there may be some mortality immediately during pregnancy and delivery, virtually all “statistical women” who die in that period in this sample would counterfactually have died of some cause or other within 15 years.

Of course, 15 years is a long time. “Would you rather die 15 years sooner or later?” has an easy answer.

What about 5 years?

Again, effects here seem very small. In the “excellent” health group there’s an increase in mortality in the next 5 years for 25-34 year olds, but a decrease for the other two age groups. For the “Very good” group there’s an increase at all ages under 34, but a decrease from 35 to 44. For the “Good” group, it’s decreases across the board. For the fair and poor groups, again, smallish samples of mortality make inference tricky. Some age groups zero out for a major reduction, some show an increase, some a decrease.

On the whole, this seems consistent with a story where pregnancy has an extremely small effect on women’s mortality rates.

Read more

Among the other beef I have with the article, I feel it’s important to hammer the Turkish point, because Turkey’s sudden massive fall in fertility has caused generalized interest and consternation in many quarters. I am going to show what’s going on with Turkey and, in the process, demonstrate that 1) It’s not a big mystery why fertility is declining and 2) it’s the damn internet. I am going to show you actual data on adoption of new digital technologies, show actual decompositions of fertility behaviors into contracepting and non-contracepting components, and demonstrate that Turkish fertility decline is utterly mundane and easy to explain. After the paywall I warn you about the impending extinction of an entire personality type.

A Crash Course on Turkish Fertility

After the fall of the Byzantine empire— oh, crap, this is supposed to be a punchy rebuttal to a bad public op-ed, not a deep digression into the origins of the early modern world! Okay, so, let’s just start with 1950. To start with, here’s the four fertility indicators for Turkey you should keep in mind: the total fertility rate, the mortality-adjusted total fertility rate, the cohort completed fertility rate, and all extant survey data on Turkish women’s desired family sizes.

The dark blue line is the classic “total fertility rate,” which is just a blunt sum of the age-specific birth rates in a given year. The lighter blue line is the same, but adjusts for the mortality-rates then prevalent, basically telling us “Based on mortality rates at that time, about how many children might the average woman have expected to actually raise?” This is a vastly superior measure of de facto family size targets.

You can see that this mortality-adjusted measure actually rose between 1950 and 1965 since mortality was falling but fertility was stable, before declining appreciably 1965-2000. Then there’s a period of stability 2000-2013, and then a sharp decline to the present. That sharp decline is what is currently raising lots of eyebrows, and what is pointing to as poorly-explained by marriage.

The browish line shows completed number of life births. This is basically calculated by summing up live births for women of a given birth cohort across their whole lifespan; so the cohort of women who were 15 in 1950 have a number representing the sum of ASFRs of 15-year-olds in 1950, 16-year-olds in 1951, 17-year-olds in 1952, and so on. This measure does not adjust for mortality. You can see it was high at first, then fell precipitously. We can’t calculate it for the most recent cohorts, because they’re not done having kids.

The lighter orange line shows the same completed fertility metric, but adjusted for child mortality in the birth years, i.e. accounting for the fact that women were not actually raising the number of children implied by live births.

Finally, the black dots capture various cases where surveys were administered in Turkey and asked people about desired family size.

Please notice that mortality-adjusted CFR converges to survey-reported desired family size in 1975, and that the evening-out of mortality-adjusted CFR is coincident with an evening-out of desired family size.

Thus, it turns out, we can explain a lot of Turkish fertility history with just two factors: child mortality (helping us see the “true family size being targeted” instead of the optical illusion caused by live births which subsequently died) and desired family size. The last fertility survey in Turkey I have on hand was done in 2018 so I’m limited in what I can say with high confidence, but from 1985-1985 to 2018, desired family size actually rose rather than decline (and that one recent low value is a small privately-conducted survey without published metadata, so pretty low reliability).

It’s possible after 2018 fertility preferences took a nosedive, but since that is after the TFR took a nosedive, we’d be better off assuming that falling TFR caused falling preferences than that falling preferences caused falling TFR.

As such, we have a conundrum. Two variables we have good reason to think should be deeply and intrinsically related to fertility (mortality rates and desired family size) are, and explain essentially 100% of fertility changes… until 2014. Then something else happens. Fertility goes into a dramatic decline without a decline in preferences. What gives?

If the above discussion was hard for you to follow and too technical, then in general this post is probably not for you. Just stop here; the TL;DR is Turkey’s fertility decline is caused by the digital revolution wrecking marriage.

To see just how odd this all is, we can do a neat exercise. We will look at three different counterfactuals. In one, age-specific mortality-adjusted birth rates are held constant at 2013 levels out to the 2060s. In another, the observed decline to 2025 happened, but we assume the decline is over in 2025, and birth rates observed in 2025 continue to 2060. In the third, we assume the pace of decline we’ve observed 2015-2025 continues 2025-2035, then 2035 rates are stable to 2060. What do these projections imply for completed fertility?

As you can see, women who were answering surveys about fertility desires in the 2000s and 2010s reported wanting about 2.5 kids each, but are likely to end up having, if continued decline goes on perhaps as little as 1 for the later cohorts, but very plausibly under 2 (after accounting for child mortality, which nowadays is very low but not zero). Even if rates had been stable at 2013 levels, there would have been undershooting vs. desires, but much more modestly.

The gaps we are talking about here are huge. The mismatch between Turkish women’s desires and their outcomes for those born in 2020-2025 will plausibly be bigger than it was for women born 1950-1960— and those women had exceedingly little access to contraception! The scale of missing kids will be socially as big or bigger than the scale of births women wanted to prevent before birth control was widely available.

This drives home a key point: something is happening to push fertility extremely low. What is it?

It’s Marriage

Demography studies are always taught a simple rule: if a theory cannot explain why one of the proximate determinants of fertility would change in response to some proposed cause of fertility change, then that cause can be ruled out. Any theory must operate via proximate determinants. The classic proximate determinants are:

1. Odds of infanticide or abortion conditional on non-miscarried conception

2. Odds of miscarriage or stillbirth conditional on conception

3. Odds of conception conditional on non-contracepting sex and/or odds of conception conditional on contracepting sex (i.e. contraceptive failure)

4. Odds of contracepting conditional on sex

5. Odds of sex conditional on coupling

6. Odds of coupling conditional on being reproductive age female

Within that, we often also add amenorrhea due to pregnancy itself and breastfeeding, which can be thought of as a subset of #4. Likewise, non-procreative sexual activities can be considered a subset of #4.

This also means we can in practice, if we had enough data, decompose fertility change into these elements. In practice, we almost never have data on 2, data on 3 tends to come from very specific narrow-case studies, data on 1 is spotty and contested. Data on 5 is also very shoddy and rare. Thus, in practice, we often focus on parameters 4 and 6: prevention behaviors conditional on sex, and coupling conditional on reproductive age.

So, what has happened? Let’s look at contraception first.

As you can see, it rose through 2018 (the latest data— 2024 data should come out any time now, however!)… but the increase actually slowed down in recent periods. It would be weird if fertility decline was accelerating while coupled women were using less contraception (remember: 95%+ of births in Turkey are to married women). So while we can’t rule out contraceptive behavior, we can see it’s a not-super-persuasive candidate for the recent fertility decline.

What about marriage itself? Here, we have more recent data:

As you can see, the share of women who were married dropped precipitously below the pre-2014 trend in recent years. For 25-to-29-year-olds you could take the break as recently as 2016, and for 15-to-19-year-olds there’s not a clear break, nor for 30-to-34 (though one may be emerging since 2017 or so). But for 20-29 year olds there’s clearly a trend break in exposure to marriage.

Could this explain decline?

Well, we can ask, simply, “If every non-contracepting married woman was trying to conceive at the biologically typical rate for women who are non-contracepting and trying to conceive, what would Turkey’s fertility rate be?” and we can compare that to the actual data.

As you can see, the brown line, which is basically “Maximum Turkish fertility if every Turkish woman is crankin’ babies like the very conservative Amish sects, but unmarried women have zero babies” is in steady decline. If every Turkish married woman was babymaxxing, Turkish fertility would not be 9 or 7 or 5 or even 4. It would be 3.2. High! But waaaaaay below historic highs!

Contraception further reduces birth rates. If every single married, non-contracepting Turkish woman today made babies as fast as possible and no unmarried Turkish women did so, Turkish TFR would be 1.61— it’s actually 1.42. You can actually see that the real fertility rate is generally quite close to the “Assume babymaxxing, except unmarried and contraceptors” line.

From those lines we can get decomposed parts. The gap between the gray and brown lines just is the compositional effect of proximate determinant #6. The gap between brown and blow, likewise, more-or-less is #4. The gap between blue and purple is various kinds of other residuals. We will chain those decomposed effects to their 2013 values, and see how they’ve changed:

Sorry — there’s no mystery here. The reality is that the decline in marriage is so big at such high fertility ages that the implied contraceptive change among marrieds is actually positive: that is, married people must be using contraception somewhat less at their peak fertility ages! Residual changes don’t really explain very much either all things compared.

It just is marriage. Turkish fertility started falling super fast after 2013 because Turkish marriage rates started falling super fast after 2013.

Read more

I have already thoroughly debunked this chart in the past.

But in this post, I’m going to try to provide a simple model of how long-run fertility change works, and why it shows that new digital technologies shifting the value of solo-leisure vs. social-leisure may indeed have a big role to play in fertility dynamics.

The Three Big Parts of Long-Run Fertility Change

There are basically three main elements of long-run fertility change. This is not to say that other things don’t matter, but we can conceptualize fertility change in three components and, when we adopt that conceptualization, my own experience has been that new information is very easy to incorporate with minimal creativity or rationalization.

The three big factors are:

1. Selection pressure via mortality and replacement

2. Recurrent emergence of cultural valuation of selfishness

3. Changing cost of fertility and especially intergenerational wealth transfer dynamics

We will very briefly take these in turn. I want to emphasize that I will be citing sources very lightly because my theses here will mostly be on not-terribly-controversial grounds. “Mortality drives cultural selection” isn’t groundbreaking stuff if you know the history of warfare and illness. “Having kids is a costly choice” won’t blow your mind. “Caldwell’s intergenerational wealth transfer hypothesis is basically correct” should surely be well understood by now, especially given all we know about child labor laws, school expansions, and fertility. I’m rehashing stuff here that good demography students learn, memorize, and can recite in their sleep.

Mortality Pressure

The fundamental force shaping human history is death. People die. Cities die. Civilizations die. Death is how we are selected for. Human genetics show extensive signs of major selection pressures over the last 12,000 years, and those signs point a giant boney finger to a long-running history of enormous epidemic and nutritional stress, as well as strong preference for people who are broadly competent and functional in the world. The key vehicles for this selection are differential fertility and mortality, but the preponderance of selective sweeps impacting genes related to nutrition and immunity strongly suggests mortality has been the dominant selection pressure in human history.

We are who we are because the people not like us died off in whole lineage groups.

This being the case, historic humans simply had to have a lot of kids. “Quality vs. quantity” tradeoffs simply couldn’t have been a concern because they were running on a mass death treadmill. For a breeding pool to survive (which is to say, an endogamous group, or what we would call a “culture”), they had to have a lot of kids.

Because endogamous groups very very often correlate with political structures like villages, tribes, cities, states, dynasties, empires, etc, this process easily became a collective action opportunity. Mating, marrying, and breeding norms were essentially political products. Romans kill unwanted daughters, Egyptians don’t. Jews marry one way, Greeks another. Political leaders were analogized to parents, and in the most primordial states in Mesopotamia and China, we often see evidence of fertility cults at the center of political organization. Keeping ahead of the specter of social death was a central concern for many historic cultures.

Thus, mortality pressure created a disciplining force which motivated kinship groups to prioritize reproduction. We see this even in modern contracepting societies: when a child dies, parents often accelerate the next birth as a form of “replacement fertility.” It really is the case even in modern data that eros and thanatos are closely linked.

This matters for the interpreting the long-run graph, because if societies basically target surviving fertility, and if some societies do this via marriage regulation, some via abortion, some via contraception, and some via infanticide or child mortality, then measures of “live births” will be misleading. Some of those live births get killed very quickly. Instead, we want a comparable cross-cultural measure. For that, we need surviving fertility. Here’s surviving fertility for the U.S.:

I’ve highlighted the periods of clearly declining fertility. And here’s the same, but with the linear trends for each period drawn:

I don’t think it’s possible to analyze that graph in, say, a time series model and come away without finding clear trend breaks. Anybody who looks at the fertility graph and says “Ah, it’s one long decline” is simply incorrect. It isn’t. There are clear breaks in series. You can contest some of the breaks I identify, but there are some breaks in series.

Even if you just drew one long trendline, here it is:

You can see that fertility today is appreciably below the long-run trend, and has been almost every year since 1970. So even if you are a “it’s just the trend” kind of troll, you should still see the present moment as exceptional.

The Selfish Impulse

However, reproduction is costly. Having a baby was dangerous, increased caloric needs for mom and once the child was born for baby too, and the baby did not contribute to production for years after birth. Because reproduction is so costly, virtually all beings have a strong incentive to avoid reproducing. You shouldn’t reproduce, it’s a terrible idea, from a purely selfish perspective.

Evolution punches back and motivates reproduction through the incredibly simple strategy of orgasms. Okay, cool, but there’s a problem: orgasms may work as reproductive motivation for creatures with less reasoning ability, but humans tend to figure out fast you can have orgasms without making babies. Hunter gatherer populations practice abortion, infanticide, primitive forms of contraception— humans have always known ways to manage the care burden on a group.

And that’s why cultural norms emerge as above. Humans have a strong motivation to defect from the interests of the genes. But the genes happen to correlate with cultural groups, and those groups create a superstructure to give other non-biological motivations for reproduction. Groups providing such incentives will tend to see their genes proliferate and will tend to see their values proliferate.

So far, our model is basically, “Humans have a constant incentive to avoid having kids, but mortality pressures give kinship groups strong motivations to counterbalance that incentive.”

But here we need to recognize the idiosyncrasy and randomness of human culture. Sometimes memes just take off for no good reason. In many societies, kinship-preferring norms succeed at keeping the selfish impulse squashed. In others, they happen to fail to do so. This is especially likely if mortality pressure is low. The process by which individual incentives for selfishness metastasize into a social norm of selfishness is not entirely clear… but it’s clear it sometimes happens and sometimes doesn’t.

That is to say, some societies by chance end up selfish, or, we might say, individualistic. Those societies often see a break from pronatal norms, and fertility falls.

But of course, if mortality pressure resumes, those societies evaporate. In a society where half of kids die before puberty, you really do just have to have 4-7 kids to maintain civilization. If you’re having 3, your civilization just won’t be maintained. You’ll be replaced by other groups and your values will vanish. By definition, this process will be very hard to spot historically, because the new group will tend to tell their own story, and also because demographic replacement is often outside of the view of historic textual sources and their primary concerns.

In other words, our model here is that “by chance” sometimes societies end up, perhaps inadvertently, scaling up selfish norms instead of pronatal ones. That “works” as long as selection pressures are weak. But if selection pressures intensify again, the pronatal norms come back with a vengeance.

Counting the Cost

Finally, there is a mediating factor in all of this: the cost of kids. Selfishness is only a big issue because kids are costly. If kids were very cheap, we wouldn’t expect the dynamics above to be problematic, since kids do yield some benefits.

To the extent something happens to alter the costs and benefits of kids, fertility will change, because the extent to which selfishness really pushes against pronatalism will vary.

Thus, consider an agrarian society with no pensions. In such a society, parents may recognize their own odds of sickness and disability in their 30s and 40s and 50s are quite high, and so having a 15-25 year old around at that time would be very valuable. As a result, they would have strong incentives in their teens and 20s to mate, marry, and reproduce, because kids yield a large insurance benefit. This dynamic is very real.

Broadly speaking, if kids are major economic contributors, they will be “cheap,” and selfishness could even militate in favor of having kids. If kids are costly, the opposite. Intergenerational wealth flows around housing, labor, age-dependent care duties, retirement, and sickness are some of the biggest factors here.

Putting The Pieces Together

Now let’s put these pieces together.

Here’s child mortality over time for three countries with long-running data:

You can see that a decline in child mortality begins maybe around 1800, but really kicks off around 1860-1870, and by 1900 is in high gear.

Do we have data on selfishness? Not exactly. But a fantastic study of French fertility proves that secularization and enlightenment values were the proximate cause of French fertility decline. This matters, because France had a very early fertility transition:

You can see France’s birth rate declined after 1750, even though, as we saw above, France’s child mortality rate did not decline until after 1850.

So France had a bought of selfishness norms after 1750, related to secularization, and fertility plummeted. France then saw revolution and a massive outbreak of continental war. France ultimately lost those wars, and the French selfishness norms mostly didn’t spread. Broader European fertility decline was very slow and limited in the early 19th century.

Instead, broader European fertility decline waited until the 1870s, when English fertility collapsed. We also know why that happened, again, it was an idiosyncratic cultural case.

But whereas the French fertility transition really struggled to spread, the British transition spread like wildfire to other countries, because it was paired with declining child mortality. Mortality selection worked against the French transition, but in favor of the British one.

Meanwhile, other factors mattered too. France’s unique inheritance rules after the French Revolution created high costs of kids. Britain, meanwhile, was fiercely pronatalist, and had abundant access to new lands in the Americas and its colonial empire for population growth to be offloaded onto. In France, kids were expensive. In Britain, they were rather cheap, especially when industrialization created a new boom in demand for child labor. Contrary to the view that industrialization reduced fertility, it may have increased it (Japan’s fertility also rocketed upwards during early industrialization— and we see in many countries such a rise).

What really reduced fertility was a revolution in costs and norms: laws banning child labor, expansion of schooling which created new expenses, and the delayed entrance into economic independence associated with schooling and skills-biased technical change. The emergence of welfare states and pension plans further altered the calculus.

Thus, we have a simple story. Selfish norms recurrently have tried to break through into a dominant position, but mortality selection and intersocietal competition kept a lid on them. When mortality selection collapsed due to technological change, selfish norms could proliferate at a lower cost. Around the same time for related but not intrinsically determined reasons, intergenerational costs of kids rose.

So how does this relate to cell phones? Well, let’s see.

Proximate Determinants

Every demography student learns about the proximate determinants of fertility. Breastfeeding duration. Time to pregnancy. Sexual access. Age-related fecundity. Primary infecundity. I could go on. Babies come from a process and you can cogently and meaningfully decompose any fertility change into a change in parts of the process. If fertility is declining, it must be due to some change in sex, or contraception, or abortion, or whatever.

In our present age, the decline since 2007 is proximately best determined by basically four factors. You can quibble on the model a bit, but any reasonable proximate determinants model will give you something in this ballpark:

1. Decreasing partnership (~50%)

2. Decreasing sexual frequency within partnership statuses (~20%)

3. Increasing contraceptive efficacy (~25%)

4. Increasing breastfeeding duration (~5%)

5. Increased abortion (~10%)

6. Reproductive technology reducing age-related infecundity (~ -10%)

We need a theory that explains these things.

We can rule out mortality. Since 2005-2010 there has not been a big change in mortality.

We should consider the possibility of a new kind of selfish norm.

We should consider the possibility in a change in the net cost of kids.

But we should also consider the possibility of a genuinely new thing. A demographer in 1700 would probably not have talked about mortality per se as a fertility regulator since low-mortality societies had simply never existed. What will be a big obvious factor the future sees that our historic models miss?

To me, the answer is clearly related to the fact that young people are spending way less time socializing independently:

And the answer is clearly related to the fact that fertility decline is extremely age-biased. And the answer is clearly related to the fact that sexual frequency is in firm decline, and marriage is in firm decline and homicide is in decline, and crime generally. In fact basically everything people do together is in decline. “Bowling alone” but on steroids.

What force would simultaneously cause all social life in person to decline?

I have a hard time seeing how a change in the cost of kids would do that, and in general I’m just skeptical that the cost of kids has changed a lot. USDA used to publish data on this and their data showed that the cost of spending on kids, converted into labor hours, has been long-run stable. Motherhood penalties in work are declining, child benefit programs have grown, I just don’t buy that there’s been a massive increase in the cost of kids (which isn’t to say we shouldn’t make them cheaper! if a new selfish norm has arisen, one plausible response is to make kids cheaper!).

That leaves us with “a new selfish norm” or “something else.”

I think both are plausible, and closely related.

If we see a new selfish norm, we have to ask how did it come into being. Much like Protestantism rode the wave of the printing press and small-family norms were carried by newsprint and radio waves, you need to be able to explain why now. The most obvious answer is “improved digital technology.”

But improved digital technology is not just a medium for a new norm. It is also “something else.” It’s trivially easy to see how Netflix and social media and smartphones and video games and pornography, and the constant availability of all of the above, could cause solo-leisure to become relatively valuable even as social-leisure declines in value. Witness the rise in complaints about family holidays, the rise in self-care as a discursive norm. In surveys, I find one of the best predictors of wanting few kids is agreement with statements like “I’m working on developing myself right now.”

Thus, a new selfishness, propelled by digital technology, and indeed made possible by that technology. Being alone is not, in a state of nature, very fun. Being alone with video games and vibrators can, in fact, be rather fun. This has the dual effect of both making attainable a new kind of aspirational loneliness, and also created social incapacitation: people just aren’t hanging out much. Not hanging out much, they meet fewer good partner candidates, filter through fewer candidates, get fewer of the real-world signals of compatibility human evolution requires for good matching, and thus have a harder time finding mates. Within unions, they may have less sex as Netflix remains a temptation even when you share a bed.

That said, we should also bracket the “new selfishness” idea. We have data on, for example, teenagers’ fertility preferences. Note that the huge drop in 2021 is probably not real; it’s likely a change in survey methods post-COVID; but regardless, a big decline in 2021 is way too late to explain a fertility decline that began in 2007.

While there has been some decline in desired family size between 2007 and 2024, notice that the numbers there are still all above 2! They’re much, much higher than actual fertility rates! So whatever normative change happened, it wasn’t a change that made young people make some incredible shift away from family in general. Overwhelmingly, young people still want families.

Other factors like abortion and contraceptive efficacy partly reflect technology changes (LARCs, telehealth abortions), but those changes are themselves partly just endogenous innovation (as the Pill itself was— it wasn’t an exogenous shock to fertility, but was summoned forth by demand for its existence, and so it is improper to construe it has having some societal “effect”), and of course telehealth abortions depend on digital technology!

Thus, if we think of where digital technology, and especially smartphones, fit in the long arc of history, we see that they are both “more of the same” (an emergent selfishness pushing fertility lower), but also “this time is different,” because we are seeing a genuine honest-to-God collapse in virtually all kinds of human social life, and this collapse exists across every country and cultural group that uses digital technologies (the Amish and Haredi are mostly immune, but fertility actually is falling for Amish sects that are more technology-flexible).

To that end, we should not see the “long decline” graph as a rebuttal to cell phones, nor should we see digital technology as a mere “accelerator of the trend.” The new selfishness is only a viable ideology because digital technologies made it viable. Sitting at home wouldn’t be appealing without the digital technologies! These norms would not have reached scale without the specific vehicles to propel them, and because socializing is a cooperation problem, even individuals who do not share emergent selfishness norms lose out on potential socializees. People don’t RSVP to the party because they are at home chatting to their AI girlfriend. Even people trying to live in the real world are impacted.

I understand many economists are burned on the “video games cause young male unemployment” takes of a decade ago, but that bad experience is no reason to close your eyes and pretend the world isn’t changing in a sui generis way. Never before in the history of humankind has so radical an experiment in social isolation ever been conducted on such a scale, and the effects on fertility are and will continue to be profound.

I want more housing built. I want more urban housing built. I want housing supply restrictions removed. But the Build Now Act accomplishes none of these things, even as it implements an unjust and bizarre transfer scheme.

The map below represents the main point I want to make. This map is, to the best of my knowledge, the actual, correct map of how much money each state can expect to gain or lose if the Build Now Act becomes law. The raw input data for it was produced by Salim Furth, then I applied the benefit calculations as they exist under present law, and applied the disaster-area exclusion rule Salim had not considered. Any errors in the below are mine and mine alone. I have no idea if Salim agrees with my views here or not. But you should trust this map as being relatively credible, because the people who made the data for it both have extensive track records working on housing policy, and there is not any more accurate map currently available.

Of the $2.3 billion in CDBG grants covered by the relevant overarching program, a paltry $7.5 million will be reallocated. However, some cities will experience large changes.

These 65 cities will all lose 10% of their CDBG entitlement funding:

As you can see from the map and the list, Arizona, Colorado, Illinois, New Jersey, Oregon, and Pennsylvania are all rather hard hit.

Why does Build Now have this effect?

The reasons are several:

1. All counties with any section 401 or 501 disasters in the last 5 years are exempted. Here is a map of counties with 401/501 disasters in the last five years. As you can see, this rule exempts the entire southeast, most of Tornado Alley, much of the northeast, and virtually the entire west coast.

2. Penalties for harmed localities are capped. No municipality will ever owe over 10% or over $1 million. This means that municipalities with very bad housing supply don’t actually face big nudges to do better. Meanwhile, bonuses are not allocated on the basis of housing construction! Bonuses are given to localities based on raw number of houses in the locality instead of actual houses added! This gives an enormous advantage to larger municipalities like New York City.

3. Counties are exempted if their rental vacancy rates are high. That means that expensive vacation destinations are exempted, since seasonal-use is included as a kind of vacancy. Ski resort towns are often exempted for exactly this reason, despite having extremely high housing costs and constrained supply. It is unclear what policy goal is served by protecting block grants for ski resorts.

4. Additionally, counties are exempted if their housing costs are low. This is a strange feature of the law: if a city actually succeeds at lowering housing costs, they are kicked out of the program and no longer eligible for bonus funding. Bonus funding only goes to cities that add houses, but aren’t actually affordable. States that are genuinely affordable simply are not even considered. This does protect those states from losing funding, but cheap places would, if included, mostly receive bonuses; the net effect of excluding affordable places from inclusion is to harm states with low housing costs.

These four features combine to create the absurd map I showed above. I should note that the single largest bonus payout, $910,000, goes to New York City. Only 8 places have bonuses over $50,000. They are: NYC; Phoenix, AZ ($163k); Fort Worth, TX ($97k); Denver, CO ($86k); Du Page County, IL ($78k); Las Vegas, NV ($70k); Hennepin County, MN ($68k); Minneapolis, MN ($53k). On net, even though Phoenix gets money, Arizona loses, because many Phoenix suburbs lose money.

The bill has many other problems. Its formula for identifying eligible jurisdictions using a change-in-change quotient is needlessly complex, and often takes money away from cities that are adding housing and gives it to cities that aren’t adding housing. Some of the data inputs are not currently available at the legal entity level and must be inferred from ZIP-code approximations. Historic data in particular is problematic and must be guessed at. The sums of money involved are paltry, yet the mechanism implemented is seriously concerning, especially if it were ever scaled up.

My personal view is that based on this information, Congress should make major revisions to the Build Now Act before passage, or write an entirely new act. In particular, Senators and Representatives whose districts include penalized areas should consider whether this bill is actually in the interests of their constituents. Are your constituents helped by sending a million dollars to New York City? Do you really believe penalizing a suburb to the tune of $6,500 is going to accomplish housing reform? Does it really make sense to exempt the entire state of California and the entire southeast?

Before voting, legislators should seriously consider these issues.

Congress has been debating a housing bill for a while. The bill has a lot of good stuff in it, such as an incredibly long-overdue fix to regulations for manufactured housing.

But it has within it a paradigmatic example of how the cognitive abilities of America’s elite are in steady decline, and that is the Build Now Act (pages 55-63), a bill so inane and stupid it boggles the mind that anybody wrote it.

In this post I have attempted to do the math to show you how stupid this bill is, but the math is actually such an impressive combination of stupid, complicated, and dependent on nonexistent data, that I actually cannot even show you how bad the bill really is. It’s likely that if passed, HUD will end up failing to actually enact the law, since it depends on a configuration of data which does not exist.

I hate this law for two reasons.

1. I loathe incompetence in general

2. It is actually an extremely good idea somebody just mangled into a horrible nonsense bill. I feel about Build Now the way the Valar must have felt when they saw what Morgoth had done to twist elves into orcs. I feel the way Obi-wan felt right before slicing off Anakin’s little legs. I feel like Caesar to Brutus. Mais je vais être frénétique et fulminant ! Il me faut une armée entière à déconfire ! J’ai dix cœurs; j’ai vingt bras; il ne peut me suffire de pourfendre des nains… Il me faut des géants !

A Good Idea

Build Now is built around one very good, simple idea: the Federal government can in principle incentivize jurisdictions to compete on housing supply and deregulation by simply reallocating Community Development Block Grant funding away from cities that are, well, not developing, towards cities that are developing. Modestly reduce funding for cities not adding housing, give it to cities adding housing. Easy, peasy!

This is a really good idea. I love this idea. I want this idea to be law. Scholars of American land use have been batting around versions of this idea for decades, though usually in a very hypothetical “what if we could…” kind of way.

So let me be clear. The reason I gag and spit when I hear the Build Now Act mentioned is not because I’m ideologically opposed to it. Nor is it, on the other hand, that I am an extremist who thinks it “doesn’t go far enough.” The reason I would like to see the bill repurposed as Congressional toilet paper is because I think it is, technically, substantively, in terms of process and implementation, idiotic, deceptive, and incompetent. That’s it, just that.

Ruining A Good Idea

Suppose you wanted to reward cities for building houses using CDBG money. How would you do this?

Simple: cut general CDBG funding by ~5% or ~10%, and take that amount and put it in a fund called the “Please Just Build Houses You Sickos” fund. For every housing unit completion a municipality reports in the Census Bureau’s/HUD’s relatively complete data of housing completions, you’d give the municipality some share of the set-aside funds. So if America build 3,000,000 houses that year, and your town built 500 of those houses, you’d get 500/3,000,000 of those dollars.

That would be a super simple way to reward municipalities for allowing housing.

That is not what Build Now actually does. You need to realize that something like the above is the concept on which Build Now is constantly trading and claiming credibility. But that is not what it does. Oh no. It does something nuts.

The first step is in covered status.

See, Build Now does not apply to all places. Cities with fewer than 50,000 people are excluded entirely. So are all areas that aren’t cities. Of the ~340 million people who live in America, only ~135 million live in places that are even covered at all by Build Now. So right off the bat, we’ve excluded the vast majority of American land, population, and jurisdictions from being included at all in a plan to increase housing. Ultimately, we are only considering highly urban locations. Build Now exists exclusively to promote urban housing projects; and when I say urban I don’t even mean like “metro area,” no I mean central cities and large municipalities.

UPDATE: I was incorrect here in the details, but in practice more-or-less correct. The law says metropolitan cities and urban counties then refers to a federal register notice about metropolitan cities. I missed the “and urban counties” bit. However, the vast majority of counties are not “urban counties,” since urban counties must 1) be in a metro area, 2) have over 200,000 people excluding people who live in municipalities. This is a kind of eccentric category; being in an MSA, with 200k people living outside of otherwise covered municipalities, is a pretty narrow slice. As a result, a piddling 17% of all CDBG funds go to the “urban counties,” and as you’ll see below, urban counties are basically yeeted out of the program through other gimmicks. So, technically, I was wrong— but in terms of actual dollar bills and final program rules, I was like 94% correct, which means I get an A, and the haters and losers can sit down.

But… it actually gets crazier!

Once we have defined the “Covered Recipients” as above (metropolitan cities >50k), we then have to subset those into “Eligible Recipients”. Some cities are “Covered” but not “Eligible” for the program.

What makes a city “Eligible”? They must pass five different criteria:

1. The city must have “Small Area Fair Market Rents” above the 60th percentile observed among Covered Recipients. Basically this means the Build Now Act only applies in urban areas in the top 40% of housing costs.

2. The median home value must also be above the median for the United States. In practice this requirement is pretty much duplicative with #1; same basic story here: only expensive places.

3. The annual vacancy rate must be lower than the national average: so if a bunch of units are vacant, no dice. I want to mention here, “vacancy rate” is a bizarre term to throw out here. “Vacancy Rate” is not defined in the actual text of the law.

4. The city must not have any recent disaster designations applied to it.

5. The city cannot have restrictions on its ability to zone and issue building permits.

Briefly, I’m going to point out some big problems in these eligibility rules, before getting to the true craziness.

Who Can Zone?

Criteria #5 is a wild stalking horse.

If a state pre-empts local zoning and established zoning rules by state law, a sufficiently trollish administration could say that renders the municipality ineligible.

If a municipality doesn’t have zoning, an administration could argue that this renders the municipality ineligible.

Imagine YIMBYs take over Connecticut and remove all zoning authorities from municipalities. Suddenly, all Connecticut municipalities would be entirely exempted from any penalties for not producing housing! They would no longer be eligible for inclusion because they aren’t zoning authorities!

The question is how much state pre-emption would be enough to trigger this condition. I don’t know the answer to that, the law doesn’t spell out an answer to that, ultimately it will depend on if a future President wants to screw around with YIMBY state-pre-emption or not. If a future president decides YIMBY state pre-emption laws are stupid attacks on exurban McMansions, a future president could just blanket exclude all states with pre-emption laws from this program, which would tend to hurt YIMBY places because YIMBY places may anticipate having higher housing growth.

So this “zoning eligible” rule is funky.

Also, you may be saying to yourself, “I wonder where I can find a list of legal entities in the U.S. able to promulgate zoning rules and/or issue building permits.” The answer is, you cannot. The Census Bureau’s building permit survey catalogues 19,000 permit-issuing entities around the country— but many of them are not cities of the kind included in the Build Now Act! They are townships, counties, or other entities! And crucially, that source doesn’t have anything about zoning authority. The National Zoning Atlas catalogues 33,000 entities that engage in some kind of zoning or land use regulation. I suppose a very clever person could probably merge those lists, but this law lowkey orders HUD to produce a new dataset of “authorities which both issue permits and have zoning rules.” Such a dataset doesn’t presently exist, and if it did, the question of “Is Houston eligible?” would be statutorily vague, since Houston doesn’t have zoning. Whoever wrote this section just didn’t think about what they were writing. (Fun sidenote: by my calculations below, Houston is not eligible for Build Now, because Houston has affordable houses, and places with affordable houses cannot get benefits from Build Now. Moreover, Houston has had natural disaster emergency declarations applied to it many times in the last 5 years, which would also exempt it, if I read the law correctly).

UPDATE: I want to emphasize how this rule is a massive question mark. Above when I flagged urban counties being included— many such counties have no zoning authority! Many counties in New England and Texas just have no zoning authority, so despite between CDBG urban counties, they would seem to be booted out of the program. I checked a couple KY counties in metro areas and several do now have zoning and planning authority; but they also don’t even have 200,000 non-city residents, so whatever. But the point is, even a lot of counties that do meet the urban county requirement may get booted by this rule.

Which Vacancy Rate?

UPDATE: This section is my biggest error, and my only excuse is I am a demented fool. The law clearly says “rental vacancy rates.” I just misread it, because I am a dumdum. So the whole section on “what if people just stopped being homeowners” is answered with “that wouldn’t matter, it’s only rental vacancies.” HOWEVER my arguments about seasonal-use vacancy are correct. Seasonal use vacancies ARE counted as part of rental vacancies. So I was totally wrong on the renter-homeowner issue, but still correct the vacancy rate stipulated is the wrong indicator.

The law says “vacancy rates.” Only places with below-average “vacancy rates” are included, “as published by the Bureau of the Census.”

Okay, what does the Census Bureau publish?

Spoiler, they don’t publish “vacancy rates” by municipality. They publish “Homeowner Vacancy” and “Rental Vacancy” rates, not a singular vacancy rate.

But in the ACS we can use microdata from the 5-year files to calculate homeowner-and-renter vacancy rates combined for all municipalities! Problem solved, right? Just order HUD to do this?

Well, no, not solved. Because first of all, rental vacancies average 7 times as high as homeowner vacancies. Places with lots of rental units will always show up as “higher vacancy” than places with lots of homeowner units. Thus, many big cities with more renters will have too high of vacancy to participate! Bonkers, right? If you make home prices expensive enough, you could shift enough of the population into renting that your overall vacancy rates rise above average and thus you are exempted from the Build Now competitive transfer program! Simple As!

Moreover, not all vacancies are actually vacant. The Census Bureau “vacancy rate” counts a unit as vacant if the usual resident is not home. This means a house that “has a usual resident, but the resident isn’t home” is vacant, it means “house does not have a usual resident, and nobody is home” is vacant, and it means “house does not have a usual resident, but somebody is home” is vacant.

“Usual residency” is where a person usually resides. Thus, for example, my wife’s family has a teeny tiny cottage up in northern Michigan. We are not usually resident there. It is “vacant.” It has always been vacant. At no point since its construction in the 1950s has it ever had a “usual resident.” But it is a property very much in active use! The whole extended family takes turns vacationing there!

Consider a city with huge vacation appeal. Maybe a beach city or something. If AirBnB purchased every single apartment and made the whole city a vacation destination, the vacancy rate would rise, because there would be no usual residents. All houses would be secondary homes and seasonal use homes. As a result, the vacancy rate would approach 100%.

This municipality would be exempted from the Build Now Act, and thus face absolutely no incentive to add homes, even though all the residents had to leave due to lack of homes after the AirBnB buying spree. Effectively, Build Now exempts places with lots of tourism from having to build houses, even though tourism often pushes home demand and prices upwards!

The Build Now Act erroneously treats “vacancy rates” as an intrinsic sign of market tightness, when in fact only certain kinds of vacancy rates are a sign of market tightness. If a town has high vacancy because lots of homes are sitting on the market with no buyer or renter, then high vacancy indicates low demand. But if a town has high vacancy because lots of homes are owned as vacation property, then high vacancy indicates high demand: that place is in such massive demand people will pay money to own houses they don’t even stay in. The vacancy rates in Vail and Aspen and Breckenridge are extremely high! Those are not low-demand municipalities!

Fundamentally, Build Now was written by somebody who just didn’t know what they were actually saying when they wrote vacancy rates into the law, and it shows.

Disasters

This section is new, added in my May-15 morning update.

Notice in the law that counties hit by a disaster in the last 5 years are exempt.

Here’s a map of every county which qualifies for an exemption under the disaster rule as of today:

As you can see… it’s most of the country, including the entire state of California!!! Had a drought recently? LOL, you don’t need to build houses! Did a wildfire burn down a neighborhood? NO PRESSURE TO BUILD HOUSES! Tornado destroy a town? You’re in luck, we won’t ever ask you to build any houses!

I mean look at that map! Ask yourself, what expensive cities aren’t covered?

The answer is New Jersey. That’s it. This is a law which will reshuffle funds around Pennsylvania, Connecticut, New Jersey, and Delaware.

The thing to realize is the existence of CDBG-Disaster Relief supplemental funding already gave states extra motivation to lobby for disaster treatment. Making the basic CDBG program itself also disaster-contingent is just idiotic.

And at the end I’ll reveal that I have a way to solve this to help cities that actually lose houses in disasters, without punishing non-disaster zones.

Eligibility Math

Okay, let’s ignore criteria 3-5 and set them aside; they’re tricky to get an analytic grip on and have weird definitional problems, so we’re going to ignore them. And since 1) and 2) are going to be very highly correlated, for simplicity, let’s go with just 1).

The Census Bureau tracks 19,152 muncipalities in the United States, which in 2024 were home to 214,000,000 people.

Of those, only 816 have over 50,000 residents, and thus are in our “covered sample.” Those 816 cities had 133,000,000 people in 2024.

How many of those 816 cities are eligible just based on the 1st criteria of fair market rents being over the 60th percentile?

The answer is a mere 299 cities, home to 52 million people.

The Build Now Act, then, only apples at all to 299 municipalities. Far from actually encouraging more housing around the country, it is literally only encouraging housing in 299 cities, and it does so only by reshuffling money between those cities.

Here I should mention that, using county data, the counties containing those 299 municipalities in 2024 gave Kamala Harris approximately 60-70% of their combined votes. What we are discussing here is a bill that basically asks, “How can we take money from some highly regulated blue cities and give it to other highly regulated blue cities?”

But still. So far, while there’s plenty of bad writing and weird statutory choices, you can kind of see how our original concept (“just pay municipalities to build houses”) might be working. We simply excluded all the poor, low-demand, high-vacancy, disaster-hit areas, and encouraged the expensive places to compete with each other. Sure, the method of doing those exclusions was sloppy and hackish, but in principle setting some guardrails about included jurisdictions isn’t totally crazy.

And if that is what this bill did I would still vomit a bit when I read all the drafting issues, but I would try to choke it back in the name of MOAR HOUSING.

But that’s not all it does. Honestly, it hardly encourages housing at all.

CAGR, PAGR, And I’m Sorry The Funding Allocation Is WHAT!??

Remember our simple model? “Just give money for each new house”? Yeah, that isn’t what Build Now does.

Once we have limited to our Eligible Recipient sample of 299 cities, now the real fun begins.

First, we use the complete Master Address File to count how many postal addresses a city has today, 6 years ago, and 12 years ago. From 6 years back to today, we calculate an average annualized change in address counts. So if a city in 2026 had 50,000 addresses, and in 2020 it had 40,0000 addresses, that’s 25% growth, which annualizes out to 3.7% growth a year. Why did the drafters choose 6 years? Beats me! Do we even have this data from 6 years ago? No, not publicly! The Address Count File only goes back to 2023 on the public census website, with a supplemental file for 2020! Of course the MAF goes back further, but it’s very much not public. So we’re using a benchmark here where before passing the bill outside stakeholders have no way of knowing the central question of all legislation: cui bono. It would be nice to know in advance something like, “Wait, how many cities are actually going to get bonuses? How big will those bonuses be? Does our math actually correctly identify high-growth cities, or did we make a horrible fucking mistake and we’re going to subsidize really bad policy?” Below, I’m going to make a rough guess at cui bono, but it will be very rough due to data limitations.

Okay, anyways, so step one is the “Current Annual Growth Rate” or “CAGR,” from the prior 6 years.

Then you do the same thing again for “Prior Annual Growth Rate,” calculated from 12 years ago to 6 years ago. Of course, again, there’s not a public database you can use to check “What would this calculation look like for my municipality?” so anyone’s guess what this actually looks like in the real world.

But let’s keep our example of a city where CAGR=3.7% . Suppose the PAGR was 2.1%. So the rate at which housing was added sped up.

Well, that’s good for us, because next we calculate the HGIR: Housing Growth Improvement Rate. The HGIR is a fun bit of bureaucratic math.

HGIR=( CAGR-PAGR ) / (|CAGR|+|PAGR|)

In our example, that’s:

( 3.7%-2.1% ) / (3.7% + 2.1%) =27.6%

Cool, huh?

But what if our numbers were a bit weirder. Here’s how HGIR shakes out for any of a range of possible CAGR-PAGR pairs:

What you can see is basically this formula rewards you for improving your trajectory, but it does so even if you are literally bulldozing homes, just at a slower rate. Look in the top left. You can have a CAGR of -3%: you destroyed 17% of the houses in your town! but if your PAGR was -5% (destroyed 27% of the houses in your town), you have a HGIR of 25%, pretty good!

In fact… as I will show below a HGIR of 25% is very good. A city with an HGIR of 25% would get a big cash bonus. You can be a cash-prize-winning city while on net destroying housing, as long as you destroy houses slightly slower than the mayor before you did.

(As an aside: it’s often good for a city do bulldoze houses! If a whole street is abandoned and the buildings are decrepit, they often pose fire or crime risks and legacy utility costs. Destroying them is often wise. But doing so can cause them to drop out of the address file. The Build Now Act would incentivize cities to keep decrepit houses around for longer)

Okay, but what is HGIR for?

Cities with CAGR>4% are set aside as “high growth rate cities.” For the remaining cities, we find the MEDIAN HGIR; so basically we’re taking the median HGIR of a negatively selected group, giving high-growth cities an automatic pass and reducing the median threshold for the remainder.

Once we find the median HGIR, we check for every Eligible Recipient if they’re above/below. HGIR above median? Cool, you get a bonus! HGIR below median? Sorry, you lose either 10% of your Community Development Block Grant funding OR $1,000,000, whichever is less.

Sidenote: The $1 million cap on the penalty is hilarious. It’s basically a “Protect the Big Cities From Real Competition” rule. As you will see below, there is no limit on how much money this program will give to big cities. But there’s a limit on how much we will take from them, because this program is just a subsidy to pour more money into big cities.

But what if your HGIR is above the median?

Here, you probably think you know what happens. “If you’re above median, you get a share of the money taken from below-median cities, probably allocated based on how many houses you added.”

In a sane world, that’s what would happen.

Unfortunately, you live in clown word, that’s not what happens.

The funds yanked from the low-HGIR cities are put into a pool. They are doled out to above-median HGIR cities based on the total housing stock of those cities. Not the added housing stock. The total housing stock.

You get it?

Imagine that both Los Angeles and Austin manage to get above-median HGIRs, and both add 10,000 housing units. You might imagine that Los Angeles and Austin both adding 10,000 housing units would then both receive similar amounts of bonus money, but it is not so! Bonus money is not allocated based on houses you add but based on houses you have. Los Angeles has 1.5 million houses, Austin has 483,000. So even though Los Angeles and Austin had the same net addition to housing, Los Angeles receives 300% more money, because it is bigger to begin with. Money is given to cities because they are big, not because they added lots of houses.

In general, this means that bigger municipalities get more money. This means a simple way to get more money is to just fold more municipalities together into one entity; you can get more cash simply by redrawing some lines on the map.

Now, I’ve done this all hypothetically.

Let’s do it for real now.

Calculating Who Wins At Build Now

UPDATE: I have kept this old section, but a vastly improved section is added below, thanks to Salim Furth.

I don’t have the Master Address File but the ACS 5-year files can give us a very good approximation from which to calculate CAGR and PAGR for 2016-2022 and 2010-2016, respectively. What we’re going to calculate is basically, “If Build Now had been in effect in 2023, what would have happened to CDBG funding?”

So, just for example, using that data, here’s California. The red means “This municipality pays a penalty,” the blue means “this municipality gets a big payout,” dark gray means “excluded from the program”, and light gray means “it’s not even a municipality so who cares, real Americans only live in city limits, county dwellers can get wrecked because nobody cares about those freaks.” I swear, that’s actually written in the law.

Now, remember, in the real world, California is 100% exempted from this law, because the entire state of California is, legally speaking, a disaster zone. It’s not me saying this. It’s the law!

You can see that tons of the San Francisco suburbs are just plainly exempted. Too small of municipalities. San Francisco itself is penalized of course as is much of the bay area, and much of LA and San Diego too. Some areas are rewarded.

But the standout fact is the the vast majority of the developable land in California just isn’t covered at all. The Build Now Act will never add many houses because all it does is encourage a small number of large municipalities to internally add houses— but 6/7 Americans don’t live in those municipalities!!

If we add up all of this by states here’s what we get:

Okay, cool. We took a bunch of money from the west coast and gave it to New York. I’m… I’m so glad. Wow. We did a big thing here. So YIMBY!

Assuming all losing-municipalities are docked $1 million as a simple heuristic, it turns out only four states end up gaining more than $1 million:

New York, District of Columbia, New Jersey, and Massachusetts. Arizona, Maine, Nevada, New Hampshire, and South Carolina also come out very modestly ahead.

States adding tons of housing like Texas, Virginia, or North Carolina actually lose money! Because they’re adding housing in suburban and exurban contexts! This bill literally takes money away from states adding housing, and sends it to New York!

How? Well, NYC’s PAGR from 2010-2016 was 0.5%. Then 2016-2022, the CAGR was 1%. So their HGIR was (1%-0.5%)/(1%+0.5%)=0.5%/1.5%=33%

Was that above the median HGIR in the covered sample? Yes it was! Among eligible municipalities with below 4% CAGR, the median HGIR was 20%! So despite being notoriously restrictive about adding housing, despite eking out just 1% CAGR, New York gets a $50 million dollar bonus!

If the money were allocated among bonus-cities based on number of houses added, NYC would still get a lot of money: $33 million, instead of $50 million.

Of course, if we just gave every city a fixed sum per house they built, NYC’s share of overall benefits from this program would fall even more. Every step of this program resulted in NYC capturing more and more of the benefits.

Maybe the drafters intended this, or maybe they never looked at the math, but if you think that New York City is a champion of housing supply and the kind of regime YIMBYs want to reward then you’re on powerful, powerful, powerful drugs.

We can also ask, “Okay, if we take all the municipality benefits… how much are cities with bonuses being paid per house they add?

By sheer coincidence, totally just by luck of the draw, for reasons we can’t explain, here’s the map of which places are getting paid the most to add houses:

I’m noticing a trend for how this law treats Texas. Are you noticing that trend? Adding a house in Texas in a bonus-receiving city landed you, on average, $215. Doing it in New England somewhere could get you a minimum of $833 in Massachusetts, but up to $3675 in New Hampshire.

I have no idea why the Build Now Act is being so aggressively advanced by Senators Kennedy and Scott (GOP-LA, and GOP-SC, respectively). It’s a crazy bill that makes no sense, whose whole actual function appears to be to just shuffle money to cities in the northeast.

Oh by the way, see this table again???

See the weird white cell in the middle?

That’s because, remember, you have a division problem in it.

If a municipality has no change in housing in either analytic period, it’s a divide-by-zero condition and you just have an irrational math problem that can’t be resolved.

The bill is based on a good principle. But its implementation is riddled with absurd rules and coverage limits and eligibility quirks and bizarre formulas, such that the only conclusion is that the drafters were intoxicated while writing it.

But Wait, There’s More!

There are other issues with Build Now. For example, it counts housing based on addresses.

Logical, right? Well… maybe not.

Imagine you have a 3-story rowhome in a city with 5 bedrooms. It’s one house. You sell it to a developer. They subdivide it into 3 1-bedroom apartments (since each apartment now needs its own kitchen and bathroom).

This results in 2 extra “addresses” (units 2 and 3). But it resulted in 2 fewer bedrooms!

Cities vary in how they handle addresses. Sometimes, ADUs get their own addresses. Sometimes, they don’t. I lived in a townhouse with friends where the downstairs was a “unit” and the upstairs was a “unit” but the “units”… were not actually fully separated? We could absolutely access each other? Maybe we should have put a door on the next flight of stairs and “added another unit”!

My point is, the incentives to game this system are large and easy. Just bureaucratically relabel addresses. Keep derelict buildings around as long as possible. “Discover” that detached garages need their own addresses as potential ADUs. USPS/HUD/Census/DOT collectively maintain the relevant database in various forms, but localities have enormous influence in the data generating process, and with money on the line they will influence it. And since USPS nowadays is backing away from delivery at home and is instead installing neighborhood mailboxes, the modern USPS has absolutely no means to verify the truth of addresses. A clever town would just install 20 extra mailboxes on top of each neighborhood mailbox, without building any houses to go with them.

Would it be fraud? Yes. Will that stop a town that really wants those CDBG dollars? No.

There’s additional weirdness around annexation and border changes; but by now hopefully I’ve convinced you this bill is just not doing stuff the right way.

If I Had Done It

So what’s the right way?

The best way to encourage housing is simple: cash for bedrooms. The Census Bureau and HUD can easily get this data. The Building Permit Survey is a near total census of permitting in America, yet it takes the average reporting agency less than 10 minutes to complete their monthly form. All Census has to do is modify form C-404 to have a bedroom count number. Adding this field would increase time burden on permitting offices, but if you’re also paying them per bedroom they permit, well gee, I rather suspect they would report permitting bedrooms. There would be issues with this, but Census/HUD already operate a longitudinal Survey of Construction which randomly selects permits from the Building Permit Survey universe and follows up with them. The SoC already asks about bedrooms. So Census/HUD literally already conduct a randomized audit on BPS data every single year! Let’s just give a real financial incentive to permitting authorities to report bedrooms! Change one field in one Census form and create one simple payment system, and suddenly you’ve made a clean, simple, transparent way to get more housing.

The reason to do bedrooms is to avoid the unit-conversion issue, and to avoid rewarding spamming of micro-homes people may not really want. Houses with more bedrooms house more people, and our goal should be housing people rather than building houses per se.

So to do this right, just set up a fund that gives municipalities a certain sum of money for every permitted new bedroom, and use the SoC as an audit to make sure there aren’t anomalous failure-to-complete rates. Don’t do any of this nonsense about mailing addresses, no special formulas for relative growth rates, no covered and uncovered areas. Counties, cities, whatever the agency is that’s issuing the permit under the Building Permit Survey, just pay that agency. It doesn’t matter if they have zoning authority or not; public money is fungible. If the county issues permits but doesn’t control zoning, the county will issue whatever permits it can and pressure will build to change zoning, and the zoning authority will eventually buckle once the local politicians realize that the permitting authority is a cash cow.

We don’t need any of this too-clever crap. Simply make a nationwide rule: for every bedroom, the agency which issued the permit gets $100. This would cost about $450 million in 2024. Community Development Block Grants funding is about $6-$8 billion every year. For the small price of 5-10% of the CDBG funding stream, we could create a simple, flat, uniform policy that clearly incentivizes adding bedrooms everywhere in America.

We should do it, instead of this Frankenhousing bill.

When Friends Say You’re Wrong

This section is a new update!

See, I didn’t know anybody had done the math on CDBG before me. But one intrepid scholar (who I don’t think was a primary bill author? but if he was, well, sucks for him!) actually has done the math (kinda)! In turns out my longtime friend Salim Furth has done the math and also caught my errors on rental vacancy and missing urban counties. It’s good to have friends who send you late night texts like “You totally misread the bill, you oafish and offensive slob.” I love friends like that. It was my birthday dinner with my wife and he’s texting me like “Dude, you cretinous slime, you hideous filth, you wretched cur, urban counties are separate from metropolitan cities.” Cool bro. These are exact quotes from Salim.

Anyways, yes, I had some errors, and also, Salim has done the math I couldn’t do-ish. His math depends on approximate averaging of ZIP code data since, as I said, the actual data does not exist. HUD does not currently have the ability to enforce this law, they’ll have to rejigger their data products to actually cover these areas instead of ZIP-based approximations.

Even so! Salim kindly sent me his file, which is the think-tanker version of a massive dunk, “Lyman, I see you can’t actually do math. Here, let me show you, u widdle biddy housing powicy baby.” Fair.

But, Salim’s math (which he’s circulated to others in the space as well) is based on an assumed corrected bill, one where the allocations are not based on total housing but based on added housing. Salim is, apparently, not an absolute wackadoodle, and understands the insanity of the total housing stock allocation rule.

I’m going to compare Salim’s “fixed” data, to the “if passed as written” version… but also Salim ignored the disaster condition!! So I will then also show a “if passed as written and we don’t ignore the disaster rule” part.

To start with, disasters bump cities out of the eligible pool. How big is that effect?

As you can see, California and Florida have TONS of cities that SHOULD be included in Build Now. So do Texas and Washington and Massachusetts.

But once we exempt places that have had disasters… literally zero cities in those states are included. None! Total carveout!

Here’s a map of states color-coded by what share of houses in that state would actually be in areas eligible for inclusion in Build Now:

It’s wild to me a “we need to build more housing!” bill is targeting Alaska more than it targets California. Is it really true that the solution to the American housing crisis rests in locking Nevada, Utah, and Arizona municipalities into a housing construction race against each other? Does Atlanta not matter? Do we not care about Florida?

This is clearly nuts. To see how nuts it is, here’s “Share of houses covered by Build Now” plotted against “Average home price”:

Pardon my French but what the fuck are we even doing here?!?!

Apparently we are super worried about insufficient housing supply in Wisconsin and Wyoming, but relatively chill and relaxed about Hawaii, California. District of Columbia, Massachusetts, and Washington, the actual most expensive states.

Okay, so, we’ve established that the disaster rule is nutso.

Now, what happens when we fix the allocation rule to reward extra houses instead of total housing stock?

As you can see, if we fixed it to allocate based on “houses added,” but we ignore the disaster rules, the result is that California loses money, Texas and New York gain it. Georgia, Maryland, Massachusetts, New Jersey, Pennsylvania, Hawaii, and Michigan lose money as well, while UT, TN, FL, and MN are meaningful gainers.

Credit where it’s due, Salim’s map doesn’t strike me as totally nuts. This probably kind of sort of is a decent proxy for housing increases except lolololol New York is still a net winner here?!?!?

Fundamentally, I still think the “We improved on prior trend” is a dumb mechanism. Just reward for adding houses. Doesn’t matter the prior trend. If you build you get paid.

But what would it look like if, instead of Salim’s wishful-thinking mechanism where Congress isn’t a senile organization and makes rational fixes, we use the actual text of the law?

Here you go, now using Salim’s better data than mine, but using my more correct reflection of the actual rules in the law:

Now, New York comes out way more ahead. Basically, the whole east coast throws money at New York under the current rules. Texas still gets some housing, but now Chicago shows up as a really big winner. Also, California’s losses are cut by 60%+, and now are similar to the eastern loser-states.

This is clearly a dumber map. Do we seriously think North Carolina or Florida should be punished for not adding enough housing so they can subsidize New York even though New York adds less housing???

It makes no sense!

Okay, now, what if we incorporate disaster rules? So we exclude all the disaster-hit places.

IT’S JUST A NEW YORK CITY SUBSIDY.

We are pillaging the CDBG grants of western and midwestern states in order to subsidize New York City and Minneapolis.

And folks, those are cities YIMBYs really like! I don’t think we need to pretend this is an accident. I think it’s obvious that there’s a third player in this math. Somebody besides me and Salim has sat down and looked at the math and wrote a law that did precisely one thing: took money from “water scarce” cities that urbanists dislike for aesthetic and environmental reasons, as well as rust belt cities full of working class whites, as well as New Jersey because New Yorkers hate Jersey, and designed a program to transfer that money to NYC (urbanists celebrate!) and Minneapolis (YIMBYs celebrate!). It’s hard to imagine how you arrive at this configuration by accident.

It’s a bad law and I hate it.

But why are marriage rates declining? Many people have theories. One theory you often here is “Because the other sex sucks.” The men are all porn-addled, the women are depressed, the men are gambling their savings away, the women are obese, etc, etc, etc.

I think this is a bad theory, but a lot of people believe it. This article is responsive to online discourse provoked by , but specifically responding to ; joined the party as well, I suspect will enjoy it, and is favorably cited, so there, tags.

In this post I will explore to what extent the youths really do suck.

Do Young People Feel Like They Suck?

Young men believe that young men suck. Here are some graphs from a recent IFS report:

Young men mostly feel they don’t know what it means to be a man, and that manhood is viewed negatively.

42% of young men describe themselves as failures. Education, work, and family status all feed into that.

So on its face, “young men suck” is, according to young men, a fairly decent thesis.

What about young women? I don’t have comparable data on hand for that, but the sex-biased rise in depression and anxiety is well-known, and this graph from Jean Twenge on rates of hospital admission for self harm I think basically answers what young women think of themselves compared to prior generations:

So I think it’s safe to say young men and young women agree with a relatively negative view of themselves, at least compared to prior generations.

If you think that self-assessment is decently accurate, then young peoples’ assessments of themselves is consistent with a declining mate quality thesis. They think they themselves kinda suck!

How Are The Youths Doing Economically?

Here’s real median income (note: median, not mean, so this isn’t skewed by a few high performers!) by sex for ages 22-30 over time:

You can see that young men’s income vs. young women’s (the gap) had a relative peak in 1970. But today the relative income gap between men and women is about what it was in 1940. The generation that got a seriously rough time of it was those in their 20s between 2007 and 2015, i.e. certain cohorts of Millennials including some whose names rhyme with Shmyman Shmone, but the Zoomers are by and large pampered brats who have had nothing but steady growth their whole privileged lives, except a brief COVID interlude.

Okay but I’m cheating here, right? Because there’s a difference between “we both earn real money but one earns $7,000 more” and “one earns zero, the other earns $7,000.” And yes, there’s been ~zero growth in earnings for men ages 22-30. That stings. And I think you can make a totally lucid argument there was a genuine qualitative change in the relative income balance between men and women between 1970 and 1990.

But I don’t think you can make that change for more recent years! If you want to argue “men aren’t as valuable as they were in 1960,” sure, no dispute. But we’re talking about a fertility decline beginning around 2007-2010, people. The societal difference between the median man making $11,000 and $7,000 more than the median young women just isn’t a big deal, and for the last 14 years of data there’s been no change at all. We have no reason to believe men ages 22-30 have seen a diminishment in economic standing since 2010 vs. young women or vs. prior young men. Rather, young men in 2024 saw their highest inflation-adjusted earnings since the 1970s!! It hasn’t been this good to be a young man in a long time! This, despite the fact many more young men are in higher education than in, say, 1950, and so have temporarily low or zero earnings.

The idea that young men present weak economic prospects at the median is wrong, and the idea there’s been some seismic change in the gender balance of earnings in the last 10, 15, 20, 25 years, is also wrong. The economics of dating are what they have been for a whole generation.

To the extent what men have to offer as high quality mates is “earnings,” their bid is as good as it ever was. The economic data doesn’t point to any decline in mate quality.

Are The Youths Criminals?

They are not. A brand new paper explores this very well at least through the 1995 birth cohort!

White and black men alike have seen major declines in felony arrests in particular, and no increase of note for others. White women have seen no change in felony arrests, black women have seen a decline. Hispanic criminality may be rising, but remains at rather moderate levels. On the whole men are a lot less criminal than in the past, women are maybe ~unchanged.

I’ve recently argued that young men today are as murderous as some very murder-ey times in the past and I stand by that, but murder is a very small share of serious crime; the other crimes seen to have declined quite a bit. So if what people want in a spouse is “someone who doesn’t have a criminal record,” you’re in luck, you have more good spouse options in recent cohorts than ever before! Mate quality has improved!

Even on murder, we can just ask, “What are the odds an ever-married man or woman is murdered by a current or former spouse?” We can do this because the Supplemental Homicide Report data includes data on relationship of offender. It has missingness, but we can adjust for that in plausible ways, and then divide by the ever-married population. Here’s the odds your current or former spouse murders you in a given year:

This is coded by sex of victim by the way. As you can see, the odds a woman is murdered by a current or former spouse is way below historic highs, and still falling. Meanwhile, the odds a man is murdered by a current or former spouse is way below historic highs, but perhaps rising, though this could be an artifact of some data quality issues.

I had Claude look up general non-homicidal IPV rates. They’re down too:

So it seems like in terms of violence and criminality, the youths really are all right. You have less cause to worry your spouse is gonna murk you than ever before.

Are The Youths Porn Addicts?

Many people say young men are so porn-brained that they don’t make good mates anymore.

Here’s the share of GSS respondents under age 30 by sex who had used porn in the last year:

You can see men’s porn use is indeed near record highs.

So is women’s!!!

The absolute gap in porn use is about stable around 10-25%, oscillating year by year, but there’s nothing in that graph that screams “There has been a unique male-biased surge in porn use.”

In 2022-2024 GSS added better porn questions. Here’s the actual frequency of use:

Men use more porn than women. gasp

But acting like heavy porn use is just a male thing is wild! About 15% of women are making ~weekly or more use of porn; that’s not nothing not even close to it. “Habitual” porn users, i.e. those using literally every week or more, are indeed heavily male biased, but I the novel change in porn in the last 15 years isn’t that men are using it so much more, it’s that women are.

And of course women are porn addicts for a different kind of porn: book porn. The stuff women read is wild. Speaking as a Christian, I don’t think you can draw any distinction between the sex scenes you read and those you watch; and speaking as a sociologist of family and fertility, the idea that sexually stimulating literature is not functioning for women in a very similar way as porn does for men is pretty clearly nonsense.

So maybe the youths are porn addled, but they seem pretty equivalently porn addled. Mate quality may have declined if you think porn reduces mate quality (it probably does), but it’s declined symmetrically, and it’s kind of weird to think that people using porn are declining mates because those mates use porn.

Also, to be clear, porn is a lot worse than real sex, and most people basically know this, and so you’d think people who can get real sex will prefer it, and so porn would tend to select out the worst mates, not the best ones.

Are The Youths Hot?

Most people prefer beauty over ugliness. Are young people getting more beautiful?

On the one hand, we might imagine the answer is yes. Plastic surgery rates have exploded, skincare products continue to see an ongoing technological revolution, pharmacological options for managing various aesthetic issues proliferate. Dress norms for both sexes are more and more pervasively about sex and sexual display instead of formal wear, work wear, etc. Every time I fly somewhere I’m kind of surprised by how people dress; men and women alike are often in basically “gym wear,” which is designed to display their form. This would seem to increase the perceived mate quality I would imagine.

But on another level, it’s obvious the youths have gotten less sexually appealing!

Men’s rise in obesity has been a bit bigger, so maybe men’s mate quality declined more, but women arguably care less about looks than men do, so it probably matters less.

On the whole, sure, rising obesity probably means declining mate quality on some level.

But to do the other-other hand… okay, maybe obesity rose, but must of the saddest Sad Singles I know are not obese! I know lots of very attractive singles! I simply don’t believe that all the hotties are marrying fine and it’s just fatties not marrying; I know too many guys with good jobs and regular workouts who are semi-permanently single and too many attractive women in the same boat to believe that marriage decline is reducible to changed physical attractiveness. We all know friends who wish they were married, perfectly reasonably attractive, and cannot find love! This suggests we probably shouldn’t be blaming declining attractiveness, even if attractiveness may have declined. Though if you take the cosmetic changes I mentioned above, it’s possible to imagine we’ve seen a radically shifted distribution of attractiveness. Attractiveness may be higher-variance now, such that “Median” attractiveness now seems relatively worse vs. “Maximum” than in the past. But I would be surprised if “Median” only indexed against “Maximum” and not also “Minimum.” A world of surgically altered Instagram models setting the top of the scale might make Normie Attractive People seem ugly by comparison; but surely a world of 30% obesity makes Normie Attractive People seem exceptionally attractive!

Are The Youths On Drugs?

Drug use is unappealing. If drug use is rising, mate quality would be falling. So what’s happening with drug use?

Well, here’s drug use by teenagers:

It’s going down, not up. The kids are doing less hard drugs.

Even the potheads are in decline!

Are you worried your partner will be an alcoholic? That’s less likely than ever before!

Do they smoke, maybe? Nope!

This is all teenager data, but sources like NSDUH show similar trends for adults. Society is beating substance abuse!

Since substance abuse is a huge mate quality issue, this tells us that mate quality on average is probably improving.

Are The Youths Inveterate Gamblers?

A lot of people are worried about the rise of online gambling. I am too. It’s a scourge.

But let’s take a step back and just be honest here. The total amount of money involved is pennies in the broad scheme of things. Not even pennies, in terms of GDP %. And it’s not as sex-biased as you may imagine. Here’s Pew data on gambling:

Men gamble more… but not, like, an order of magnitude more. And in fact men have more hostile views towards gambling than women!

The rise of gambling may have made young people into worse mates, and there may be some male bias to that, but we shouldn’t overstate the scale here. Both sexes are gambling more, but most people still aren’t gamblers, and while gambling is a serious problem for many people, we’re still talking about a small share of a rounding error of the total economic resources of a generation. “Online gambling is why people aren’t getting married” no, no that isn’t true.

Do The Youths Want Weird Sex Stuff?

Many people think young people are weird sex fiends who want weird sex stuff vs. olds, who enjoy only totally normal sex stuff. I’m not convinced. We don’t have great data on this, and anecdotes about “the weirdest thing somebody asked me in bed” are not reliable.

The best source here, alas, may be Aella’s Big King Survey. Here’s her estimate of the number of “uncommon fetishes” people of each age had:

There’s growth in early adulthood probably related to sexual discovery, but there don’t seem to be any really big or meaningful differences between the early 20s and the early 30s. I don’t think this supports the idea young people today are inhabited a newly bizarre sexual environment vs. 10-20 years ago.

Here’s bondage interest:

Same story.

Here’s male vs. female interest in “brutality” in sex:

Check the scales. Teenage girls who took Aella’s survey probably aren’t representative, but it seems there’s disproportionate interest in violent sex among teenage girls, and then it settles down to a quite rare fetish that has little age variance and similar interest across sexes.

On choking specifically (one that comes up a lot in mate-quality conversations), estimates are all over the map. Some sources claim 70% of young people have been choked during sex. The best estimates are from this paper. Here’s the table for women’s responses. Notice the choking results.

And here’s men:

Again, notice the choking results.

Now let’s compare those.

Compare “men choked partner” to “woman was choked.” Notice the bars… don’t match. A lot of women are being “choked” when men don’t think there’s any choking happening. You might say, oh, it’s men misreporting. Maybe!

But... probably not. It’s likelier to be women misreporting. And not because women are unreliable in some general sense, but for a specific statistical problem you can spot in the data pretty easily. Look at the men who report being choked vs. women who report choking. WAY more women report they choke their partners than men report being choked. So women over-report being choked and also over-report doing the choking. If it’s just men didn’t want to admit to choking girls, you’d only see the 1st. Maybe there’s also a male-specific taboo against reporting being choked? But I think we just should take all the choking stuff with a grain of salt.

What’s going on here, I suspect, is just the downstream cascade of redefinitions of sexual assault, sexual consent, sexual violence, harassment, etc. A lot of women increasingly use relatively extreme and violent language to describe behaviors men do not perceive that way, or that women 20 years ago did not perceive that way. There probably has been a rise in sexual choking, but it’s nowhere near as prevalent as women’s reports suggest, and we know this because when those women think they’re choking their male partners, the guys don’t even seem to notice in the survey data, which suggests the female-normative definition of choking is a behavior many men do not even notice happening, not a dude getting off on throttling a woman. You can see this kind of phenomenon many areas; most famously in the Korean case, where Korean men and women have massive gaps in what kinds of behaviors they think are sexual harassment.

And this is the key thing to recognize: we have to distinguish between actual changes in mate quality and changes in perception of mate quality which may have nothing to do with the actual mates. Korean men are a lot less violent than American men and on many measures a lot more gender-progressive, yet a lot of Korean women report preferring American men based on the perception that Korean men are risky/reactionary! What’s going on there is not an actual chasm in mate quality, but a chasm in public relations.

There has probably been some rise in unconventional sexual behaviors among heterosexuals, including choking, and that could impact mate quality (unless you’re into that). More broadly, increasing heterogeneity in sexual activities could just make sexual matching harder even if no specific change is reducible to a change in “mate quality.” It’s possible that it’s getting harder to find a good lover even if all the lovers are better than ever, if the space of what you want from a lover is narrowing, or if the market is segmenting, or if potential lovers have a lot of bad press.

So If You’re So Smart…

I could go on. But the logical question here is, “So Lyman, if the potential partners are so good, why do people feel like they’re so bad?”

Good question. Great question. There are two reasons.

1. The vibes are bad. People are getting sad and miserable. “If partners are so good, why do they seem so bad” is the same question as “If the economy is so good, why does it seem so bad” or “If life expectancy is rising so much, why is everybody acting like we’re dying of microplastics” or just “Why is everybody so doomerish.” I won’t belabor this, I’ll just say, we live in an age of bad vibes. You have a choice to either let the bad vibes steal your life from you, or heroically insist upon good vibes. You should simply heroically insist upon good vibes. When things go badly, it is not because doom is at hand. It is usually for mundane and often solvable reasons. Civilizations that catastrophize end in catastrophes. Simply stop being depressed. Simple As. Easy!!!

2. But the bigger factor is nobody is hanging out at all. Here, I shall belabor.

Derek Thompson has a great piece on this. I think these graphs tell the story:

Dating is a matching problem. To find a good match you have to usually sort through some bad matches. In fact, usually a lot of bad matches. The usual way you do this is not by fucking them. The usual way you do this is by chatting with people at parties and events and in real life, and by rapidly filtering through a large number of potential matches and narrowing down. Over the course of my collegiate career, my university had about 600 girls in an age band where they might have been reasonable spousal candidates for me. Make it about ~300 when you rule out those with very few times they weren’t already dating somebody. Of those, I probably had some direct interaction with at least 200 because I had a public-facing job interacting with ~all the students on campus.

But realistically I probably had substantive conversation, discussing homework, talking in class, being at some social event, or an extracuricular, with maybe 100-150. So my pool to filter through was about 100-150. I did indeed filter through. Ultimately, there were probably ~25 women over the course of my collegiate career for whom I may have felt some momentary flicker of, “Oh, I wonder if there might be some interest here, or if it it could be created?” Of those, I considered perhaps 6 women with any seriousness. Of those, I dated 2, and of those, I married 1. She’s fantastic! I love her so much! Love is a numbers game. If you’re not starting from a big pool at the top end, you never end up with anybody at the end.

The key thing is you narrow down by the high-dimensional information gained by socializing. You talk. You see how they talk to others. You see how they move in the world. Things they do. Their friends. Not just their Tinder pic. The “relationship stuff” may be how you work down, in my case, from 6 to 1, but you get down from 150 to 6 just by talking to people, hanging out, meeting people. If you’re not posting numbers on meeting people in the real world, your odds of the final 1 are gonna be lower.

Consider this graph:

So meeting online is doomed, yeah?

No.

The thing you have realize is at the same time that couples meeting shifted online the total number of couples forming declined. I don’t have perfect data on this so can’t fully calculate it, and In that data above I can’t separate out marriages from dating, but to a first approximation what really happened is the raw number of couples meeting online probably only moderately rose 2010-2024, but the raw number of couples meeting offline is plummeting, and so the online share spikes. It’s key to realize that there are just fewer couples forming, so even if people are not really meeting online very much, the online share can be booming.

The point is, nobody just meets a cute girl or a handsome guy anymore! That isn’t what people do! And that’s a huge problem because the high-dimensional signaling of in-person social environments is the entire ballgame! It’s what your entire psychology evolved to navigate! It’s how humans make good matches!

This has downstream effects. As life moves online, society polarizes, often around sex. Politics is more and more sex-polarized, and that makes matching harder, even if it would be weird to say political preferences actually reduce mate quality. The decline of in-person socializing is leading to ever-more-polarized gender norms.

In other words, it’s not that mate quality has declined.

It’s that the in-person socializing which allows people to find good mates and which draws the sexes together to compromise has collapsed. That’s it. That’s the whole story. Relationships are declining because nobody is partying. There’s no story here about mate quality. For mate quality some variables go up, some go down, but basically it’s not the problem.

It would actually be shocking if, in a world where every other kind of socializing fell by 20-80%, relationship formation didn’t fall by a similar amount! That would be statistically bizarre!

To solve fertility, solve marriage. To solve marriage, solve dating. To solve dating, solve socializing. To solve socializing, plan parties and institute immediate capital punishment for anybody posting too much on social media. Oh no. Wait. No. Please. No I didn’t mean, I didn’t—

The NYT Upshot data blog wrote this over the weekend:

It contains many chart crimes others have critiqued.

I want to focus on this egregious deception:

That data was produced by Philip Cohen, i.e. @familyunequal on twitter. For the spectators, I want to note an issue here. The data says “Philip Cohen analysis of CDC data.” CDC does produce estimates of completed fertility from the National Survey of Family Growth, but that’s not what this data is. This data is synthetic childlessness calculated from a Lexis diagram. Basically for a given birth cohort you sum the share of women of that age who had a 1st birth in each year across some part of their life course, and that tells you the implied motherhood rate. But this value is inaccurate if mortality or migration altered the composition of the cohort over their life course. In principle what we want to do is get complete birth histories for all the women actually born, but in practice this is almost impossible to do. The method used here is reasonable enough, though in recent US cohorts it shows lower childlessness than direct surveys show.

But I flag this because collecting age-and-parity-specific-birth rates for women who were 45 in the years 1962-2024 is a bit of a chore. Maybe Cohen did it himself or had the data on hand from a prior project; well done if so! But I’m able to almost exactly reproduce it for 45-year-olds by just downloading the prefab data from the Human Fertility Database. If I gave data to the NYT which I downloaded from the Human Fertility Database, I would have had the NYT cite the Human Fertility Database since they did the real work. Perhaps Cohen re-invented the wheel and produced data extremely similar to HFD’s estimates. Or maybe the wheel built by the Vienna Institute of Demography got a rebrand. I do not not know. I hope he reconstructed the data himself, including scraping the old PDF files back to 1917, because claiming credit for HFD’s data would not be proper.

UPDATE: Cohen responded. He did indeed mislabel the source, the data was from HFD. I will note that the reason for the error is striking: incorrect attribution in order to bolster confidence in the data as “official.” But it’s not official! The population denominators there are not “official”! Even the birth counts are not “official” as HFD makes some modifications for ARDY/ACY preferences. So Cohen mislabeled the data specifically to give readers an incorrectly confident view of the data, both denying credit to the people who did the work, and misunderstanding and miss-stating what the data actually was. “I cited the wrong source to make it sound more impressive” is not a defense!

Anyways, here’s my reproduction of the graph using the HFD data:

This graph makes it look like even though births by age 30 are collapsing, later-life births are booming enough to make up for it and achieve stable fertility.

But of course, women who are 45 in 2024 are not the same women as those who are 30 in 2024! Aligning a chart like this by year is a classic error we ding students for in their first demographic methods class, because it creates the visual appearance of stability-by-period, when the actual variable you’re looking at is a cohort variable.

Now, here’s the same input data, but now the X axis will be birth cohorts of women instead of year of observation:

Here, you can see that while the gap between the age 30 and age 45 lines changes over time, the two lines are correlated.

Let’s focus on that gap. Here’s what it looks like:

I’ve marked both the actual observed 1st births between 30 and 45, as well as a dotted line for what a very straightforward linear extrapolation from recent trends might be.

You can contest that extrapolation, but it’s hardly crazy! It’s certainly a reasonable benchmark to say “1st birth rates between ages 30 and 45 will continue to increase, at a rate fairly similar to the increase we’ve seen in the last decade.” Maybe that won’t be true, but it’s at least a neutral starting point.

If we apply that extrapolation to our cohort graph, here’s what we get:

Only just over 70% of women born in 1994 can expect to be moms by age 45. Childlessness rates will soar from their current ~10-15% levels for women in their 40s now (born late-1970s) to ~30% for women born in the mid-to-late 1990s.

We can’t know what people said that wasn’t quoted in an interview. Given Cohen the benefit of the doubt, maybe right after this quote, he said something like, “But, at current ASFR1s, such a recovery would require a dramatic acceleration of 1st birth rates very very soon.”

If he gave that quote but didn’t make such a caveat, that would obviously be a case of an expert knowingly misleading a reporter. You cannot produce the figures I and Cohen produced without either 1) just copy-pasting them from someone else in which case there’s a citation issue here or 2) noticing and encountering the ASFRs and thus having it pass before your eyes that ASFR1(30-45) is not exactly going like gangbusters. I hope Cohen gave the appropriate caveat and the issue is the reporter chose to write the story instead of the facts.

But what if ASFR1 did go like gangbusters? Imagine this is what happens to 1st births at ages 30-45?

Basically, imagine the rate of late-in-life 1st births starts to rise again like it did for 1945-1970 cohorts.

What would that yield for overall motherhood rates?

We would still see that rates of childlessness would rise to about 22%, nearly double current levels. This would be similar to levels seen for the cohort of women born around 1910— women who came of age during the Great Depression, and at a time when early-life mortality was much higher than it is now (some women were childless because they died before ending their reproductive years— thankfully this is less common now).

I don’t think you should assume we are about to see a massive trend break. Provisional birth data for 2025 point to ASFR1 values virtually identical to 2024, meaning that any such break would come no earlier than 2026. But Jan-Feb 2026 also show no major difference in ASFR1 vs. Jan-Feb 2024 or Jan-Feb 2025, so really any major trend break must be happening right now.

But, are the vibes shifting that hard right now? I don’t think so. I don’t think we’re living through a dramatic pivot in the rate at which older women have 1st babies. If in the future data shows that April 2026 was such a hinge, I will be hilariously wrong, and nobody will enjoy dunking on this post as much as I will. But, I dunno guys, I just don’t think we’re seeing a baby boom right now.

As such, while the rapid-increase-in-older-1st-births take cannot be totally deprecated as an opinion… you should have a pretty strong prior on business-as-usual. You can forecast fertility usually with >90% and often >95% accuracy just by rolling prior years forward for sufficiently high-level variables. When somebody advocates for a major break in trend in such a high-momentum variable, they need a compelling reason why.

Nobody can offer such a compelling reason in this case.

As such, the NYT is (probably knowingly) promoting a view of fertility they know to be an error.

At this point, we’re rehashing old territory. I shredded on almost this exact issue not long ago. Here’s the facts: fertility is falling very fast, and we are not seeing large recuperation at older ages as cohorts with delays hit those ages. Every single mathematical tool we have for projecting cohort fertility is flashing “DOWN” signs for eventual cohort fertility.

Every single demographer who touches ASFR data knows this. You can’t not see it. It’s obvious in all the tools we use.

If a demographer doesn’t say “Completed fertility seems to be declining,” you’re dealing with a demographer willing to obscure, hide, or mislead about facts in order to advance their preferred narrative.

Now, other points Cohen made are totally reasonable arguments: many demographers may see the decline but not worry about it because they are productivity optimists, or favor higher immigration, or think it comes from improvements in women’s autonomy. I disagree, but those are all disagreements we can have among people who basically all know how the math works. Those are hard problems.

But “Will completed fertility fall for recent cohorts to record-low levels?” has an easy answer, and the answer is “Yes.”

UPDATE: Cohen responded. His response is incoherent.

He says he’s not forecasting that 30-year-olds in 2024 will reach 88% ever-births by age 45, and true, he didn’t verbally say that. He simply published a graph with that as the obvious interpretation by any layperson, in an article headlined with that very claim as the thesis. He claims this makes the chart easier to understand— no, it makes it easier to misunderstand, which is why the discourse around it has been so grossly erroneous, with tons of commenters explicitly reading it to say that cohort fertility will be stable. Cohort trends are not that hard to explain! You can of course use cumulative fertility graphs as one approach, but if you want the whole time trend, the cohort-arranged method I took is bog-standard, I’m sure Cohen has taught his students to do it a thousand times. This is indeed a “perennial” problem, one with a settled solution: when showing cohort data, you arrange the data by cohorts, end of story. You do not show cohort data arranged by something other than cohort, because this leads to major misperceptions for virtually the entire lay audience, as literally everybody who’s ever made graphs of cohort fertility for laypeople is painfully aware.

He goes on:

WTF does this even mean? There’s no relationship between the idea that “completed fertility is known” and the idea of “replacement.” First of all, calculating replacement rate from the Lexis diagonal is a different beast since it requires a different set of mortality statistics: cohort mortality exposures instead of period ones, so the “replacement rate” as such is simply not identical for period fertility and cohort fertility. Again, this is like a week 5 of your first quant demo class homework problem Cohen is failing here. Second, he’s clearly making this up: the graph they chose didn’t introduce uncertainty, it created the appearance of certainty! To a layperson, it gave a strong sense that cohort fertility is and will be stable! My graph presentation communicates uncertainty, by having blank values for recent cohorts or using hashed values or you could even do something like simulated or Bayesian error ranges, that’s not hard to do. “Experts may disagree” yeah they probably will since you designed your chart in the way literally every expert knows is creates a maximal perception of “closed and complete” data! If you want to communicate uncertainty, show the uncertainty, not a graph that makes everything look tidy and complete.

Finally, a quick note: Cohen says “we” made the graph the way “we” did. Who is the “we” here? An unattributed research assistant? The journalist? To what extent is this Cohen sockpuppeting a journalist who didn’t know better? Is the Upshot responsible for any errors here, or is Cohen? Is this a Cohen op-ed, a Miller piece, or a Cohen-Miller double byline? What are we even looking at here?

Share

Australia Has a Serious Pronatal Policy Shop

I’m gonna start this post by being nice and we’ll get to my cutting and biting critiques of others further down.

So on that: you may not have heard of them, but e61, an Australian think tank, seems to be quietly building out one of the best pronatal policy shops of any think tank without a headquarters in Charlottesville, Virginia (what can I say, I’m biased). This month they have this great piece showing the role of residency rules, migrant composition, and other “denominator factors” in shaping Australian fertility. Then there’s these wonderful retrospective on how Australia’s baby bonus really did boost fertility in the early 2000s. I don’t know Pelin Akyol, but she’s doing great work. Her Google Scholar. Her twitter. Ali Vergili seems to be working with her on some fertility stuff.

I would be remiss if I did not mention some other really great orgs in this field. I really admire Väestöliitto in Finland, for example. The U.K. has some good organizations I see as fairly overtly pronatal like Boom or the Centre for Family and Education, but also has a lot of mainstream policy shops who are willing to tip their hand and evince a concern about fertility, like Onward, the Family Education Trust, the Resolution Foundation, and others. There are other good orgs out there; I’ve had very good Spanish and French collaborators at various times; and I have an obvious Anglophone bias; but if we were to plan a conference of all the “pronatalism think tanks” in the world, we could use any Holiday Inn in America as the conference center and probably have room to spare. It’s a small club. So when you get a new entrant like e61, it’s worth sitting up and taking notice.

Anyways, the research at e61 is rigorous but accessible, the graphics are generally very clean and interpretable, and they seem to have a good head on their shoulders about how fertility decisions actually work. I would keep on eye on e61 in the pronatalism space because I think they’re going to keep doing really good work. From their latest, for example, this is just really fascinating:

How did Dobbs shape demography?

Not that much! This month, the CDC estimated that U.S. birth rates were just 1.57 children per woman, the lowest value ever recorded. It may be surprising that birth rates have continued to fall after so many states have banned abortion. But new studies reveal that post-Dobbs bans have had only very small demographic effects. Abortion bans only boosted births by about 1-2% for states implementing them, though WIC utilization rose 2-6%, consistent with another study showing 2-4% birth increases for black and Hispanic women. Meanwhile, sensationalist reporting about maternal health impacts of abortion bans have proved unfounded: the first peer-reviewed study on the topic has found that abortion bans had no effect on pregnancy-associated mortality. The reality is that since contraception remains widely available, and since many women have been able to secure so-called “mail order” abortifacients, abortion bans do not lead to large increases in live births. Instead, we have seen surges in certain forms of contraception and very modest overall birth changes. The only way to boost births in a major way is to make it easier and more appealing for young people to marry and have children; restricting their options does not have large effects. There was, I should note, a rise in preterm births and in births with congenital malformations, especially right after Dobbs. Here’s updated data on congenital malformation deaths of newborns:

I haven’t broken it out by Dobbs/non-Dobbs states, but I’m content to accept the premise that abortion bans led to an increase in children being born whose birth was not a threat to the life of the mother but was not compatible with life. My pro-life beliefs accept that tradeoff personally, but as a political realist, I am willing to accept the legality of abortion for cases where any reasonable doctor would agree that the fetus has conditions incompatible with a lifespan of more than a few days or weeks, and I think that forcing moms to carry such births to term, even if it is low-risk to them, is a massive political liability for pro-life people, and one which on net adds exceedingly few days of life lived, either in or ex utero. A lot of these deaths are of babies delivered via C-section the absolute first day when delivery could be counted as “delivery” instead of “abortion.” The fact that we are forcing moms to wait a couple more weeks and having a C-section so they can watch their baby die over two weeks instead of allowing an abortion to me seems like a pro-life hill not worth sinking broader bans over. But note that this is a political judgment. If Ruth and I were having such a child, we would carry it appropriately, attempt to give it the longest life we reasonably could without inflicting tortuous pain, and would regard any other course of action as a mortal sin.

Ironically, Jessica Valenti accused Florida of hiding maternal mortality data around the same time the “no maternal mortality effect” paper came out. “Florida hides data” is of course a COVID-era trope. It was a lie then and is a lie now. Florida has one of the most transparent public health systems in the country: you can look at it work in almost real time! Moreover, regardless of the determinations of state-level MMR review committees, all deaths are reported to CDC directly, and can be queried on CDC WONDER by anybody, even Jessica Valenti, if she knows how.

If you want a real example of a state actively concealing data for political purposes, you might consider California refusing to collect abortion data. Or California refusing to report marital status or paternity acknowledgement on birth certificates. Or California continuing to have a nonstandard checkbox for pregnancy-associated mortality. Or California, I kid you not, still having legal secret marriage. If you want a state which flagrantly violates every norm of good vital record reporting, which hides data it dislikes for political reasons, that state is California. It’s not a lib-con thing either; I have no beef with New York or Illinois on this stuff. Hawaii reports fine. But California. They’re the worst.

It’s Singles Tax Day In Japan!

Effective April 2026, Japan rolled out a new family-support package. It’s getting very excited press-billing, but it’s fairly modest overall. The key policies include free standard childbirth by covering normal deliveries under the national health insurance system (previously, families received a lump sum that fell short in 45% of cases); the "Childcare for All Children" program, providing universal daycare access for infants aged 6 months to 2 years for up to 10 hours per month regardless of parental employment status (basically this is giving Japanese parents “free drop-in care” for small children, a super valuable flexibility benefit); and existing child allowances are being extended through age 18 with income caps removed. This is being paid for with a new "Child and Childcare Support Fund" deducting approximately ¥250-450 per month from all employees' health insurance premiums. It’s been nicknamed the "singles tax" because childless workers can smell in the air what’s going on.

On the whole, what’s really going on here is Japan is using a payroll tax for a plan to simplify eligibility, improve flexibility, and moderately enhance generosity of its family policies. Don’t expect a baby boom, especially given the fire horse year, but this is definitely a step in the right direction. As a reminder, Japan is already outperforming most of its peers; as these policies continue to be rolled out, I would expect Japan’s relatively better performance to continue or grow:

Past the paywall, we’ll talk about irrational optimism in Macau, ’s wrong takes about housing, remote work, Israel, and more. We’ll close by catching somebody who went viral sharing an incorrect Claude-produced graph about housing.

Fiery Horsey Macau?

Read more

In this post I will look at homicide using three methods in order to show that by many very reasonable benchmarks homicide rates may be seriously misunderstood:

1. When we convert U.S. homicide rates into offender rates instead of victimization rates, the observed decline in homicide is somewhat reduced, and in fact 2021 saw homicidality on par with the 1970s and 1980s

2. The best available data we have on homicide in the U.S. does not support the notion of a long-term homicide victimization or homicidality decline over time; in fact homicidality may not be declining at all

3. Perceptions of crime going up accurately reflect actual spatial trends in homicide, even as national homicide declines: you cannot judge the correctness of peoples’ crime perceptions from aggregated data

Homicide Offending vs. Victimization

The usual way we talk about homicide is by asking how many people were murdered vs. some population benchmark.

But that is not the only way to look at it. We might also ask how many people murder people vs. some population benchmark. Since age and sex strongly stratify homicidality, we could find data on homicide offenders by age and sex and divide by population by age and sex to get age-and-sex-controlled offending rates. I do this. I use the Supplemental Homicide Reports microdata, convert it into a “long” format arranged by offenders, use SHR data to infer the Offender/Victim ratio, apply that ratio to homicide counts I trust more than SHR (e.g. vital statistics data) to infer total offender counts, then adjust offenders by age and sex such that each year’s actual offender-weights match the inferred offender-counts. In principle, my approach assumes the age-and-sex distributions of offenders we don’t have data for match that in reported data. This method isn’t perfect but I think is reasonable enough for our purposes.

You can see that throughout the 1970s and 1980s, homicidality was a bit below homicide: this implies that some of the increase in homicide was simply due to more people in homicide-prone demographics, i.e. young men. On the other hand, in more recent years, you can see that the homicidality line is above the homicide line: this implies some of the decline in homicide is just due to there not being so many young men anymore.

Note also that while homicide rates suggest the post-2014 murder increase never reached the levels of murder seen in the 1970s and 1980s, homicidality rates say they did! In other words: an American of a given age and sex in 2021 was about as murder-ey as one in the 1970s and 1980s; just a smaller share of Americans were at murder-ey ages.

For the curious, here’s the actual homicidality rates I calculate, averaged 2014-2023:

In other words, if you’re asking “What’s my personal risk of being murdered?” then yes, homicide fell. But if you’re asking, “How homicidal are the people in the world around me?” then homicidality in the 2020s is just about the same as it was in the 1970s and 1980s.

A key driver in this is that the average homicide in the 2020s has more associated offenders. This could be because enforcement changed, but I suspect it’s also because the profile of crime changed: more group killings, fewer solo serial killers, etc, etc. Crime has become more of a team sport.

The way this pencils out is that in the 2020s you are less likely to encounter young men in society, but the young men you do encounter are as likely to kill you as in the 1970s and 1980s. Especially if you observe them in groups.

What matters more for crime perceptions, homicide, or homicidality? I imagine the answer is “both.” But it strikes me that homicidality is probably the one people can “pick up vibes” on more easily. Actual homicide is very rare and concentrated, but it’s correlated with many, many different social pathologies: homicidality is strongly related to general criminality, and a regular person in society probably can pick up on a general vibe that the young men are getting sketchier and crime-ier. Note that this is not about diversifcation of America’s young men: the black share of America’s young men isn’t rising, and the growing racial groups like Asians have similar or lower homicidality than whites. The change being described here is almost certainly within-race.

Homicide Rightly Measured May Not Even Have a Long-Run Decline

It’s very hard to measure homicide. Here’s Our World in Data’s best attempt at long run data:

It’s a pleasantly negative trend over time.

But… hold on. Consider the Italy data, the high brown line. Do they really have homicide rate data for Italy in 1300? Answer: no, no such data exists. Instead, we have data for a few cities or regions which are used to broadly interpolate to the wider country. In some of these regions there is approximately national data, but even where there is, it is basically a 0% probability that the reported data correctly matches to our modern notions of homicide.

Don’t believe me? Here’s OWD’s effort to show measurement variation in homicide in present-day data. Consider France:

You can see that France had that very high historic homicide rate… but even today, different sources diverge in their view of France’s homicide rate by a factor of 2 or more!

The situation is marginally better in the U.S., but, still, there just isn’t agreement on even recent U.S. homicide rates:

Why does any of this matter?

Well, shares this graph:

Is that data accurate?

Well, I won’t belabor it too much, but there actually is one part of the U.S. which has pretty credible statistics going back a long way: Massachusetts. Massachusetts has kept homicide records in some form since the 1600s, and casualty counts of the European population in colonial wars are relatively well-attested. After the mid-1800s, Massachusetts’ homicide records became regularly-reported vital statistics data, and by the late 1800s are actual medical examiners’ data. This is earlier than most jurisdictions provided such data at such a high quality. Here’s the result:

As you can see, homicide rates in the early days of the colony were high. The simplest explanation is just that the early colonial days were highly selected for adventurous young men seeking a new life. Lots of young men with little oversight means lots of murders, even when the young men are godfearing Puritans. You’ll also notice the major peaks— those are cases of wars. I have counted fatalities in war occurring in battles within the borders of Massachusetts in the homicide rate, but have also highlighted the estimates of base homicide rates from one published source.

You can see as well that homicide rates probably rose sharply between the late 1830s and the 1850s. Some of this may just be improved reporting, but probably not all of it. The surge in immigration into Massachusetts really did boost homicide rates. Over time, as those immigrants acculturated, homicide rates fell, until by the mid-20th century homicide rates were back near all-time lows. Then came the 1960s and a return to homicide rates typical of the 1850s or the 1650s, then a decline in the early 1990s, and basically no change since. Today, Massachusetts “peacetime” homicide rates are higher than in the mid-20th-century, similar to the latter-19th and early-20th, and higher than the period 1700-1800.

But what about homicidality? Well, it’s hard to say exactly because we don’t have perpetrator age data, but let’s make a very very crude and improper assumption just to get a ballpark: what if Massachusetts homicides can 100% be attributed to men ages 15-35? This is a bit of an extreme assumption, but not too much: that category really does do 80% or more of homicide in almost all societies! For simplicity I’m going to use a best-estimate harmonized estimate of homicides in Massachusetts from the sources shown above.

You can see that homicidality (the orange-ish line) dropped fairly steadily 1630-1750, then spiked for some major wars, then return to its low level around 5-10 homicide offenses per 100,000 people in the potential-offender pool. It remained at that level until the 1850s, when it spiked, before returning to a level around 5-15 from 1890-1960. Since then it rose.

There has been a gradual decline in homicidality in Massachusetts since the mid-1970s… but it is very gradual, and the low homicidality levels in MA today would have been high homicidality levels in any year 1730-1850 or 1880-1965.

So what we can say is it seems like Massachusetts men are still more murder-prone than they were for most of the history of Massachusetts.

The difference is just that young men are a shrinking share of the population.

Again, we see here how homicide and homicidality can yield somewhat divergent readings of the history. And remember, the above homicidality measure does not account for the SHR-implied increase in offenders-per-victim, so the true prevalence of young men participating in murder has actually declined even less. There might not have been any decline in homicidality in Massachusetts since the 1970s!

Homicide Can Fall Even If Homicide Rises For Most People

Read more

Man is born free, yet everywhere is in chains. So quoth Rousseau. He was correct in a certain sense: small-scale societies and hunter-gatherers do enjoy a high degree of certain types of freedom. Lacking states, and with “exit” always an option to go start your own micro-society, without borders, with the means of subsistence basically always at hand, hunter-gatherers do enjoy a high degree of a certain kind of freedom.

On the other hand, modernity does offer many chains. Fences and walls abound; you cannot simply set off across country so easily. Subsistence requires cooperation of many millions of people. You must drive in a certain lane. There are rules about what you wear. A key difference between small-scale hunter-gatherer societies and societies from agriculture to the present is that these larger-scale societies depend upon orderliness and this dependence motivates strict norm-enforcement.

This is not the speculative part of the post. I am informing you of facts. A recent large study, for example, found that agricultural societies make far more use of corporal punishment than hunter-gatherers. This is part of a broader trend: societies where subsistence allows more accumulation and requires participation by more people tend to emphasize obedience more. At the extreme, places with more historic rice cultivation (which requires a lot of cooperation) really do have more obedience, conformity, and collectivism-emphases in their culture, this relationship is probably causal, and it probably extends to parenting behaviors. It just is the case that agricultural societies have different parenting norms.

And this brings us to the speculative questions: first, a question which is largely addressed in scholarly literature, “Why?” do settled societies parent the way they do and, second, a question where I’m going to embark on wildly provocative parenting claims, “What does this tell us about how we should raise our own kids?”

In this post, based on only the barest shreds of evidence and obviously unreasonable readings of that evidence, I will:

1. Lay out a theory of what parenting is actually doing at a social level

2. Explain how that theory fits well with the existing evidence on heritability of traits

3. Explain why the heritability literature explains why you should focus on selecting your child’s cultural peers instead of making child-specific investments, and instead of selecting simply socioeconomic peers

4. Unite the theory of parenting and order with that emphasis on cultural peers to explain why parents should worry less about their children’s emotions and more on habitual disciplined behavior

5. Say outrageously insulting things about gentle parenting

To start with, let me describe my thesis in another way. Here is an incomplete but I think broadly representative list of things I think parents might “target” with parenting:

1. Promoting child health especially through healthy eating habits (targeting “health”)

2. Disciplining child habits whether negatively through punishment to disincentivize bad habits or rewards to incentivize good habits (targeting “discipline” or “habits”)

3. Investment in child skills and aptitudes through training, tutoring, lessons, and other skills-development activities (targeting “achievement”)

4. Investment in academics, learning, and scholastic ability (targeting “education”)

5. Having fun with their kids doing things child and/or parent enjoy (targeting “fun”)

6. Fostering their child’s emotional wellbeing (targeting “emotions”)

7. Integrating their child into their cultural group or cultural values, often through group-based activities or rituals (targeting “culture”)

8. Managing their child’s peer and social groups to give them friends the parent approves of (targeting “peers”)

9. Encouraging their child to excel in athletic and competitive activities (targeting “sports”)

Here’s my vibe on how parenting current looks today, vs. what I’m arguing it should be:

This post will be an elaboration on why I think the allocations I suggest make the most sense.

Human Nature Is Rebellious

Humans are not, in fact, political animals. For many many thousands of years or hundreds of thousands of years we were cognitively capable of creating sophisticated societies with exchange and specialization and hierarchies of authority— but we mostly didn’t. We didn’t do this not because there was no advantage to doing so (probably there was! evidence of group-based warfare among human ancestors is endemic, which suggests that humans absolutely did realize the advantages of that kind of highly specialized cooperation), but presumably because there were countervailing factors. One such factor is probably that maintaining an orderly society is psychologically taxing. Man evolved to wander the wide savannah in pursuit of game, to meander through the forest with an eye for edible berries and tubers, to sit around the fire and tell stories. We did not evolve to take orders.

James Scott in his book Against the Grain makes many arguments I think are goofy, but he is correct in that book (and in basically his whole written corpus— Weapons of the Weak and The Art of Not Being Governed give the same vibe, and all derive from his original perspective in The Moral Economy of the Peasant) that the peasant class is basically constantly scheming to escape the bondage of agricultural subsistence, that order is maintained only at great effort by elites, and that when nomadism, pastoralism, foraging, or hunter-gatherer lifestyles are a viable alternative to sedentary agriculture, the formation of states is nearly impossible (indeed, even stable formation of cities or settled agglomerations becomes almost impossible). Dawn of Everything and many other books have all adequately summarized how Europeans encountering native peoples often were quite fond of the freewheeling native lifestyle, at least as long as open frontiers could still be found.

Put bluntly, it is not in our nature to be willing and useful participants in orderly societies, that is, societies where people act according to fairly general rules of behavior in a way that dramatically reduces the frictions and transaction costs for large-scale cooperation. It is in our nature to defect from such societies and prioritize our own liberty.

Consider warfare. In hunter-gatherer societies, warfare is overwhelmingly about ambush, deception, raids, hit-and-run… hunter-gatherers very rarely fight pitched battles. Every encounter between settled and hunter-gatherer peoples is rife with complaints about the cowardice and dishonor of hunterer gatherers. They don’t follow the rules. They kill the women and children but run from the men. They use disguises. They hide among civilians. Heck, they don’t even recognize the role specialization of “civilian” or “combatant.” This is often described by settled peoples as savagery— but in practice what it is is just that one side lives under an elaborate set of rules which constrain their free action, while the other side does have rules, but they are radically different kinds of rules, and they are not rules which create generalized conditions of order. “Boys must hold a hot coal for 30 seconds to become men” may be a binding rule, but it does not in any meaningful sense reduce frictions for social cooperation in the way that, “When two societies fight, they should not kill random civilians” does. Many scholars argue that “moralizing high gods” were invented essentially in order to create superstitious reinforcement for generalized moral rules as part of the transition out of hunter-gatherer living. I’m skeptical of that particular chain of reasoning, but the fact that the argument exists and has some support should make you think I’m operating here on fairly well-trodden ground.

The rules of settled societies are rules which make large-scale, complex, specialized settled life workable. These rules are not part of our evolved environment. They are not part of our nature. They are invented elements which developed through cultural evolution, not biological evolution, and as a result may conflict with many of our biological impulses.

Rules are nurture. Introduce foragers to alcohol and watch how the lack of strong social rules about alcohol plays out. They didn’t have access to alcohol so never developed the cultural technology to regulate it. In fact, without broader cultural technologies for rule imposition, even developing a norm is very hard!

None of this, I should note, is about a difference in nature or genes between settled and forager peoples. My view is this is purely about culture. There’s no intrinsic inferiority involved; just different rules, and indeed, different cultural conceptions of what it even means to have rules.

If I am correct, and here I think it would still be hard to seriously rebut me and claim that there is not a major difference in social rules with the advent of agriculture, then two obvious propositions emerge:

1. Social rules are probably constitutive of life in a settled society. If we see all settled societies have a lot of rules, we should think that abandoning rules may have some chance of the loss of settled societies.

2. We should recognize that between us and the forager there is a great cultural divide: our children have to master a technology which is not in their nature to master. They must be rule-followers, or else they cannot operate in settled society. This is unnatural to the human frame, but it is the reality.

These propositions then suggest that if parents are left to sort of stumble around and just figure out what their job is, across a few generations, they will usually coalesce on basically “My job is to impose an orderly set of habits on my kid and to create good peer environments around them.”

But nowadays a lot of parents do not see their job this way. Instead they see their job as a lot of other things: saving for college! creating enriching extracurriculars! preparing for a specific career! engineering a specific kind of emotional life! making their child sincerely love Jesus!

This post will be an extended, meandering digression on why I think these kinds of goals for kids are mistakes, and instead our goals should mostly just be about passing on knowledge of and experience working with social rules and, more broadly, passing on our valued cultural norms.

Order Is Imposed

Consider, if you will, the case of handedness:

You can see that left-handedness probably had approximately its modern prevalence in the 1700s. There was not a deep-seated superstition against left-handed people sufficiently strong to actually pressure people to learn to be right-handed. Whatever anti-left-handedness norms might have existed before the latter-18th-century, they clearly just were not strong enough to have actually changed what hand people used to do things. It’s not until the 19th century that we observe what seems to be a decline in left-handedness. Just for reference, here’s handedness estimates from archaeological data going back to before the last ice age:

You can see that handedness probably declined around 4-6,000 years ago to its pre-modern level before declining again more recently.

Why did left-handedness decline? Again, it wasn’t about superstition! It was about rules! We don’t need to speculate because we know children were forced to write right-handedly as part of the rise of literacy:

Society reached new levels of complexity and intermediation: literacy and factory work. Both of these had implications for handedness. Literacy instruction was quite formal and so children could be made to use their right hand as part of a belief this was necessary to develop a standardized and legible writing-hand, and of course factories arranged tasks in ways that biased towards certain degrees of handedness. But literacy was probably the main factor— notice left-handedness starts to explode for the cohorts born in the 1920s, that is to say, the generation of children raised by parents whose working careers were contemporaneous with the invention of the typewriter. The invention of the typewriter and its popularization in the early 20th century arguably vitiated the need for children to develop a uniform writing style and form: formal documents could be easily typed without needing a printing press.

When the exigency favoring standardized writing-hands faded, so did the rule. The decline in left-handedness 4-,6000 years ago is a bit more mysterious… but that’s when we start finding lots of bronze tools consistent with intensified agriculture. Maybe semi-standardized tool use did it? That period also saw a spike in warfare; maybe lefties were somehow at a disadvantage? Who knows!

You might consider homosexuality in a similar light. Most historic societies may have allowed homosexual behavior to various degrees, but obligate homosexuality has faced various degrees of stigmatization in the vast majority of settled societies; that is, whatever people did privately, adults in almost all settled society have been expected to engage in sexual reproduction or else to be set apart for reasons of religious taboo or sex work (sometimes the same thing). Why is this? The obvious reason is that in societies with high mortality it’s important to have lots of babies. As mortality fell, reproduction became less urgent, and nonreproductive sex generally has become more popularized. This process pretty obviously began almost as soon as mortality began declining, a process so widely observed this dynamic lies at the heart of the entire “demographic transition theory.” You’d think, though, that contraception would lead to a huge rise in the kinds of sex which can be procreative— but sexual frequency probably didn’t actually rise that much in the 1960s, and in fact the long decline of mortality and emergent norms around nonreproductive sex meant that despite penis-vagina sex being “lower cost” than ever before, it probably also makes up a smaller share of the orgasms of the world than ever before. Though the emergence of very low fertility nowadays could cause these norms to shift again.

The key point is, when the exigency favoring reproduction-maxing faded, so did the rules around sex.

Societies have rules created by imposition. Social order is “emergent” in the sense that it often involves many diffuse actors feeling their way through the dark towards a solution, but that emergence is responding to imposed shocks to society: subsistence, climate, technology, religious movements, wars, politics, whatever.

The crazy thing is a lot of people act like the decline in left-handedness was some kind of evil discrimination. But it wasn’t! Every society that goes through adoption of standardized tools and expanding literacy faces handedness shocks! Because optimal handedness under industrialization and hand-writing is less variably distributed than under other conditions. Acting like handedness was a dynamic of discrimination is absurd: it was socially optimal to pressure kids to be right-handed! A similar argument can be made on sexuality: why didn’t ancient societies ever develop long-term stable social norms endorsing obligate homosexuality? The most likely answer is just that under prevailing social conditions for human history that social form was not adaptive. Societies drifting that direction did not succeed in perpetuating those values. Our current circumstances may be different! Acting like historic societies were just imposing these norms for the fun of it makes no sense.

(Of course sick societies with norms that settle in suboptimal equilibria are also very common— many societies could be happier and more prosperous if they collectively adopted a modestly different norm which they really could adopt; I’m not arguing every society is always in an optimal normative space. Rather my mental model is that societies are almost never at their optimal normative space, it’s a zone of constant conflict and re-evaluation, but one which has a tendency to drift towards the shifting target of optimality)

A key vehicle through which settled societies have imposed the norms they thought were valuable and important is, of course, parents. In many societies very literally: parents controlled social resources, controlled marital choices, etc. For much of human history, society was to a great degree a consortium of parental (and especially paternal) authority.

Parenting, then, at a macro view, just is the process of civilizing hunter-gatherer children into agriculturalists. That’s it. That’s the job. Your children were born to wander in the woods scavenging for berries. Doing so is a lot of fun. Kids love the woods. However, they no longer live in a hunter-gatherer society. As such, your job is… not to reify the child’s inner forager. Your job is to provide the cultural evolution software booted onto the biological evolution hardware: to train your hunter-gatherer child to be a settled-society adult. Learning the rules, learning to self-manage, learning to comply, to cooperate, to coordinate, just is what parenting is. Everything else is a distraction.

This doesn’t mean you have to endorse every rule. Maybe you don’t want your kid to comply with some rules. But they should know the rules they’re breaking and know how to handle the consequences. Passing on The Rules is what parenting is really about.

Most Parenting Is Wasted Effort

You will not have much effect on your child’s IQ. It is largely genetic alongside some very broad social-environmental factors like avoiding serious head trauma, and a mixture of other environmental effects we just don’t understand very well yet. You won’t have a ton of influence on your child’s educational attainment either, provided you live in a society with free public education: the truth is that’s highly genetic too. Enrolling them in youth sports won’t make them nearly as athletic as choosing a slightly more athletic spouse. Their personality? That’s largely genetic too— you just aren’t going to shape it very much, again, barring really extreme edge cases. “Your parents, they fuck you up” — Larkin was full of shit, you were born that way. It’s only your parents’ fault in the sense that they chose to let their genes recombine.

But before you go all “Oh no, Lyman, the hereditarians got to you??? Blink twice if you need rescuing!” realize the above statement is perfectly compatible with environmental influences shaping everything that actually matters in life. Personality isn’t very important in life, and we know this because personality traits are highly varied within human populations and people in the normal ranges of all personality types live perfectly fine lives. IQ and educational attainment aren’t that important aside from at the extremes: super-geniuses yield massive social benefits and the bottom 5% create considerable social costs, but at the end of the day the distribution of adult self-reported happiness, adult income, and life expectancy for a 100 and a 110 IQ person are like 90%+ overlapping.

Seriously, using the NLSY 1997 data which has early-life scores on IQ-correlated tests and a good measure of mental health in the longitudinal respondents’ 30s and early 40s, here’s how many standard deviations difference in adult mental health is predicted by standard deviations difference in test scores:

You can see the mean estimates are all very small, and at the extremes the standard errors are comically huge. It does look like being >1 SD below-average on cognitive scores in high school may be associated with perhaps… 0.05-0.25 standard deviations lower mental health? But 1) the effect is all on the low end, notice there are no gains throughout the center of the distribution and maybe losses at the high end though that’s super noisy and unreliable.

Here’s the actual distributions:

You can see that, yes, respondents with 1+ SD below-average test scores do have elevated risks of very adverse mental health in adulthood (the left part of the graph)… but also have higher risks of very positive mental health in adulthood!

The main effect is actually that smart people have less variance in mental health!

But even so, fully 72% of the total mass of the entire distribution is under all three curves, that is to say, mostly they have the same experiences. Mean differences are small, distributions overwhelmingly overlap— even if you could boost your kid’s IQ, it wouldn’t actually change many the big things in their life like “Do they say they’re constantly anxious on a survey?” and “Do they often contemplate committing suicide?” and “Are they happy?”

This is because happiness and suicide are also in the category of rather genetically influenced things. Non-genetic stuff may have big short-term effects on happiness, but long-run subjective wellbeing indicators (i.e. “steady state” subjective wellbeing) has a big genetic component.

You can’t make your kid a superstar athlete. You can’t make your kid a genius. You honestly might not even be able to make your kid happy.

School, sports, happy vibes— that’s like 85% of what modern parents focus on. And it’s mostly a wash. When parents say “I just want my kid to be healthy, go to college, get a good job, and be happy” they are literally listing four categories over which they honestly have some of the least influence.

Some Parenting Is Very Useful

And yet, parenting is hugely influential.

Consider language. When a child under age 3 from Ethiopia or Korea or Russia is adopted in America, what are they odds they speak English as an adult? Virtually 100%! What are the odds they speak Amharic, Korean, or Russian? Well, unless the adoptive parent specifically chooses to support instruction, the odds are basically 0%. Basically 0% of Americans just pick up Russian in their daily life. I know some Ethiopian, Korean, and Russian adoptees; none of the ones I’m close with speak Ethiopian, Korean, or Russian.

Language is a classic example of a social rule. You can’t just make any random babbling sounds and expect to be understood. You have to make sounds that other people understand as referring to something. In other words, you have to engage in a coordination-game governed by rules. All human societies have languages, so virtually all kids pick up languages.

But it would be wrong to act like this is just from kids picking it up. The odds that a child in America speaks Korean are literally dozens or hundreds of times higher if they have a Korean parent. And indeed, even in monoglot households, parents put a lot of effort into language. We talk to our kids with babbling noises, an activity which we know evolutionarily is part of how kids acquire sounds. When they ask what a word means (547 times a day), we tell them. Every family has words a child mispronounced in a beloved way which are now “family words.” Language instruction is actually a massive case of parental investment which does work. If a parent wants to pass on a language to a child, investing efforts to do so increases the probability the child as an adult has the targeted trait by an enormous amount. Hebrew School works. Here’s Hebrew fluency by denomination among US born Jews; this is obviously about Orthodox parents putting their kids in yeshivah:

Read more

Consider the apple. Regardless of the price of apples (which could be influenced by monetary policy as much as by demand!), if you observe that the share of apples going unsold is rising, you’d infer that probably apple production should go down…

… unless the optimal spoilage rate rose! Imagine that one apple orchard in the country develops 10x productivity. They outproduce all the other orchards. But now apples must be shipped further! It’s possible that the optimal spoilage rate could rise! It may be most efficient to buy apples from the big centralized orchard, ship them, sell them at the usual price, and tolerate higher spoilage due to more shipping time.

In the long run especially, the optimal rate of unsold production may vary for structural reasons. If the spoilage rate of applies rose from 0.5% in 1935 to 5% in 2026, we wouldn’t say that the apple market became 6 times less tight. We would say that over a century of technical change, the apple market changed.

But if the unsold rate of apples rose from 0.5% in 2019 to 5% in 2026, we probably wouldn’t infer that this was just structural change. We would probably infer that demand for apples declined relative to supply. Maybe supply rose a lot due to a bumper harvest and sticky prices made it hard for price signals to adjust enough to move enough product. Maybe there was a huge surge in popularity of apple-substitutes (quince? pears?). In general, changes in unsold inventory still have some interpretive challenges, but they are really a “rubber meets the road” measure of supply and demand. If sellers really can’t find buyers, that tells you something about whether it’s possible for existing demand curves to “cover” marginal cost.

Enough about apples— back to houses. There’s lots of data on vacancy rates of housing, but I’ve rarely seen it enter the public discourse. This post is my effort to inject vacancy data into the mainstream. Once we have understand what’s happening with vacancy data, we arrive at a simple conclusion:

There is huge pent-up demand for single-family houses, especially modestly-sized single family houses of the kind preferred by large build-to-rent developers. This is the exact kind of housing Congress may be about to ban in the modified ROAD to Housing Act.

How American Housing Is Changing

I’m going to use data from the American Housing Survey to describe housing in America. The AHS is not used as much as it should be because the files are not harmonized into a multiyear file, not hosted by IPUMS, and just generally under-loved. But I harmonized them all. The AHS is a nationwide probability sample of housing units. Even if they are vacant, data is still collected on the unit in question. It is based on the quarterly-updated Census Address File, and is kinda-sorta conducted as a branch of the annual American Community Survey, we won’t belabor the exact details. Suffice to say, it’s a survey of all existing housing, not just new housing starts. Moreover, it includes both occupied and unoccupied housing.

The figure below shows the number of bedrooms in American vs. the number of people. As you can see, the number of bedrooms is rising faster than the number of people. I think this should sort of give us pause when we think of the housing crisis as being a crisis of objective standards of living. The reality is the U.S. really is swimming in bedrooms!

But is it because of McMansions? Below, I show the ratios of people/bedroom, people/unit, and bedrooms/unit:

As you can see, the people/bedroom ratio is falling; that’s just the two earlier lines divided as above.

People-per-unit is more stable. That’s because bedrooms per unit is rising fast. Whereas in 1973 the average housing unit had 2.45 bedrooms, in 2023 the average housing unit had 2.79 bedrooms.

How housing gets used is changing too. Here is vacancy over time by broad type of vacancy:

So here we’re taking the share of all U.S. bedrooms which are in units which were vacant. AHS has about a dozen different vacancy codes; I compress them into two kinds: vacancies which are neither owned nor rented (so plausibly “on the market” in some sense), and vacancies which are owned or rented, mostly cases of seasonal/vacation use, units already sold or rented but not yet occupied, etc.

Both kinds of vacancy trend together. The reason is simple: when housing markets are loose, more people can afford to maintain unoccupied seasonal 2nd homes. When housing markets are tight, they can’t, or they prefer to get their cash instead of the use of the home for leisure.

You can see that from 1973 until about 2008-2012, vacancy rates steadily rose. More and more U.S. housing was sitting empty. This was most clearly true for the paid-but-not-occupied group: 1973 to 2008 saw an explosion of second-home purchasing. But since 2008 and especially since 2015, we’ve seen sharply falling vacancy rates. Fewer and fewer homes are unoccupied. Again, the fall is sharpest for the seasonal-use category, but after 2015 you can see significant tightness appearing even in the core vacancy group.

Using our apple metaphor, the period 1973-2008/2012 saw a structural change. More and more Americans wanted to have and were able to have additional mostly-empty homes. Then, sometime between 2008 and 2015, the situation changed. The housing market since that period has undergone and extraordinary burst in inventory tightness. Seasonal vacancies especially are evaporating. A long-run structural shift is pivoting. We’ve gone from 40 years of ever-looser-and-higher-vacancy-rate housing markets to a sudden bout of radical market tightness.

If you prefer vacancy by unit count instead of bedroom count, here it is:

Basically the same story.

Let’s look at another housing trait: kind of house. Here’s the share of U.S. bedrooms in single-family homes vs. multifamily structures, i.e. apartment buildings.

You can see that whereas apartments were 23% of U.S. bedrooms in 1973, today they are 16%. In terms of places for people to sleep, apartments are in decline.

And here it is in unit counts:

Implicitly, here’s average bedrooms over time by type:

What I hope you notice is that single family houses keep getting bigger and bigger, while apartments have basically zero change. As Americans have gotten richer and higher-income over time, builders of single-family homes are responding with appropriately larger houses, from 2.75 to 3.15 rooms per house between 1973 and 2023. But builders of apartments are stuck in the 1970s. Aside from a level shift caused by a survey method change, we can see there’s been basically no change in average apartment size from around 1.7 rooms per apartment.

Past the paywall, we will take a detailed look at vacancy rates by kind of vacancy X kind of home X bedroom counts, to discover which exact kind of house is seeing the biggest drawdown in recent vacancy rates and, thus, likely the place where market demand is tightest.

Read more

Does marriage make men better?

Substacker , whose name understates his confusion, has an anti-marriage screed.

The litany of silliness defies belief. He refers to low-sexual-frequency marriages as some terrible tortuous outcome— because apparently he thinks if the people with dead bedrooms 20 years into a marriage had never married, they’d be having a lot more sex. Sure. Many such cases!! Not really. Yes a lot of people with sexually unsatisfying marriages imagine they would be bigtime sexual players outside of their marriage but the reality is if you’re not a desirable mate to a person who is married to you you are probably not a highly desirable mate in general. “But of course there are cases of dead-bedroom marriages where a partner finds an affair!” yes, but there are vastly, vastly, vastly more cases where the partner finds ~nothing. Also, his definition of a sexually dead marriage includes couples having sex once a month, which, while lower frequency than I assume most people would like, is pretty obviously not a sexually dead marriage. If we assume a marriage age of 25 and 20 year marriages as Mr. Bafflement suggests, we would want to know: how does sexual frequency compare across marital statuses at age 45?

You can see that married men at ages 45-54 are bangin. Almost none are having sex “not at all.” And for men who were formerly married: they’re still having more sex than the never-married guys! Now Mr. Bafflement will cry “selection,” and we’ll come to that, but his characterization of all these marriages as being sexually dead is one of many cases of his basically dishonest way of presenting his argument. He says “The marriages are sexually dead!” while not telling you “Actually the marriages have much lower rates of sexlessness than any other group, about identical rates of implied overall sexual frequency as the top-heavy widowed and divorced group, and ~2x as much sex as the separated and never-married group.”

Next, Mr. Bafflement claims it’s better to be single than divorced or in a bad marriage. This isn’t really true though. Using GSS longitudinal data, it turns out the negative effects of divorce are symmetric with the positive effects of marriage: once people get divorce they’re no worse off than they were pre-marriage or than never-married peers. Here’s a similar calculation using the National Longitudinal Survey of Youth 1997 cohort:

You can see that a measure of mental health (basically depressive symptoms) got a lot more positive (i.e. less depressed in this framework) in the period from 5 years before marriage to the year of marriage. Then it declines to about marriage+7, then rises. You can see the broad pattern is the same if we include everybody who got married even if they got divorced (dark blue), or if we include only those who never divorced (orange dotted), those who eventually divorced (gray dotted), or just the specific person-years of then-ongoing marriages and their premarital periods (lighter blue). By any metric, at marriage+15 years, respondents are about 0.1-0.2 standard deviations less depressed than they were at marriage-6 to 10.

But hold on, says Mr. Bafflement, it’s clearly selection! You can see happiness rises before marriage!

Correct, it does. But that is no surprise: the happiness of marriage doesn’t begin when you get married. Arguably it begins the first time you imagined getting married, but it definitely starts no later than when you begin dating the person you will marry. Unfortunately NLSY doesn’t have full dating histories. But it does have cohabiting histories, and premarital cohabitation is very common in the NLSY97 cohort, it’s about half of marriages!

Here’s happiness around the first time they moved in and cohabited with someone, whether by marriage or nonmarital cohabitation:

Here you can see a huge happiness increase in the year of moving in together. Then the same decline over time— but it keeps declining! What’s the deal? Well, simply, marriages that last 7+ years really do tend to provide enduring happiness, but nonmarital cohabitations lasting that long tend not to, and many people transition out of marriage into various cohabiting relationships, etc.

Point is, the effect of moving in together is very clear and evidently not very selective: a short run happiness boost. A major increase in partner sexual availability gives you about 4-8 years of jollies.

But love is quite different. If we estimate a single model with both marital timing and cohabitation timing in it (rather than above estimating parallel models), here’s what we get:

The leadup to cohabitation is often a period of declining happiness. Then there’s a bump around cohabitation, then a return to decline. The leadup to marriage is a period of rising happiness, and then basically long-run stability of those gains. Now this model by necessity only looks at people while they are married so can’t capture divorce effects, but the models we looked at above already showed us that accounting for divorce doesn’t change the observed effect pattern much at all.

So why would the period leading up to marriage lead to more happiness? Is Mr. Bafflement right? Is it just selection?

Somewhat, yes! Happier people are more desirable marriage partners, and so happier people are likely to get selected.

But that’s obviously not the whole story. We already saw that one transition towards marriage, cohabitation, yields a pretty clear happiness bump. Are we to believe that other parts of the process don’t? There’s no feeling of happiness from the other steps towards marriage? Nobody feels happy about the first kiss with a partner, the first night together, new experiences made, engagement, meeting the family? We’re supposed to believe that the process of falling in love isn’t mentally beneficial? Of course it is, that’s an idiotic argument.

More to the point, saying “it’s selection effects” is idiotic: selection is the point of marriage. “Your marriage only makes you happy because you were selected by someone you selected!” Yes. That’s true. And I did things to myself to make myself more likely to get selected, and I did those things in the hope I would be selected. If learning to dance didn’t help me woo my wife I wouldn’t have done it. This is what selection effect cultists don’t seem to get: in coordination problems where there are real benefits of successful coordination, getting yourself selected is effectively a causal mechanism.

“Is sex fun because it’s fun, or is it a selection effect that people who are fun to have sex with keep getting selected into having sex?” Those are the same question! Yes, sex is fun, and the reason it is fun is intrinsically related to the other person being involved! Selection effects are not an alternative to causality they are a channel and mechanism for it.

But we can do better!

The NLSY97 also asked respondents in an early wave (2000 and 2001) before most were married if they expected to get married in the next 5 years. This is an obvious kind of selection! Those who expect to get married are more obviously selected than those who do not expect to get married, at least for some kinds of selection. Unfortunately because the question only asked about “next 5 years” we have to limit to just a small range of years of data to be included, but it still looks pretty striking:

As you can see, individuals who had near-certainty of marriage saw the smallest overall gains, and also the largest share of their gains occurred pre-maritally, consistent with the idea that “marriage expectations arise from investments in promarital human capital which generates hedonic returns from love pre-marriage.”

But now look at the people who did not expect to get married, especially the group who gave under a 10% chance! (This was not a super small group by the way, about 17% of respondents) They see quite large happiness gains in the year of and just after marriage. “Surprised by marriage” pushes a lot of happiness gains into the period of the new marriage. This is, again, strongly suggestive that the premarital happiness gains are due to falling in love. People who expected to marry in the next 5 years were also more sexually active, while those who did not expect to marry were less sexually active. It seems transparently clear that the reason there’s less happiness gain post-marriage for the high-expectation group is because more of the happiness of falling in love is already present in the benchmark happiness period.

The point here is that the “selection” on happiness into marriage is an illusion. People become happy because they are falling in love. It’s true happier people also fall in love more, but this is not the whole effect. To a nontrivial degree, love is a very real source of happiness, and the most durable loves are those which create commitment mechanisms for themselves like marriage.

Once you see this effect in happiness, you’ll see it everywhere. Is it possible that desire for future marriage could cause men to invest more in human capital and earning power now? Yes it is! That means that a future marriage would cause higher earnings in the past through expectational channels. But if men adopted the belief that marriage is a torment nexus they should opt out of, their earnings will fall because motivation for human capital acquisition will decline. It’s not just social conservatives who say this, notable social liberal explicitly argued this in a new post just this week!

All the “it’s just selection” arguments Mr. Bafflement makes fall away when you realize that the combination of self-selection as a mechanism of selection alongside expectational channels shaping other-selection combine to reveal that, indeed, marriage causes people to be happier and richer. It just often achieves those goals before the wedding night. Marriage is a strategy for locking in the gains from courtship, not acquiring new gains.

And finally, because anxious manosphere boys will clutch their testosterone supplements and cry “But what if she leaves me!” yes that happens but divorce has less bad effects on happiness than breakup of cohabitations:

I have to include divorces in the cohabitation line because otherwise there’s problematic selection on type of union end as “good” cohabitations often “end” in marriages (which sometimes then end in divorce, etc). But the key point is:

1. Divorce or cohabitation breakup has no immediate effect on happiness. Actually happiness rises ever so slightly in the year of breakup.

2. The entire “negative happiness effect” of divorce/breakup emerges while in the union

3. After a divorce unhappiness stabilizes.

4. After a broken up cohabitation happiness keeps on falling to the end of the credible data. A key reason for this is broken cohabitations are often followed by additional broken cohabitations.

In general, the happiness “lost” in divorce in the fixed effect model is slightly larger than the happiness “gained” from falling in love. But the happiness “lost” in cohabitation is so much larger than the happiness gained in cohabitation, that if you’re debating marriage vs. cohabitation, marriage yields bigger gains, the gains last longer, and the losses from breakup are smaller. Though of course if you are debating marriage vs. cohabitation the question you really want to ask is, “Have I and my partner invested enough in promarital human capital (i.e. do you love and are you lovable) to generate a strong partnership or are we just fucking?”

Your coworkers make you have babies

Peer fertility effects have been documented many times but this one is recent and fun. An Italian law in 2015 reduced the family benefits workers hired after 2015 got, and as a result employees hired after 2015 had lower fertility. Here’s the crazy thing: so did their coworkers. Firms that hired more post-2015 employees acquired bigger shares of workers with worse family benefits, and thus lower fertility, and that caused contagion even to workers who had better benefits. A 1 percentage point effect in the post-2015 group yielded a 0.4 percentage point effect in the pre-2015 group. Now aside from that being a big effect size I want to note this creates a massive identification problem for regression discontinuity style estimates of pronatal policies: true effects are probably bigger than we think!

Is women’s income the cause of low fertility?

A new paper showing a negative substitution effect of income on women’s fertility went viral, claiming there are large negative effects of women’s income on fertility. The reality? Nope. It ain’t so. When governments have given specifically wives gigantic tax cuts, the result has been somewhere between no change and an increase in fertility. This is because earnings just aren’t the big factor shaping fertility: marriage is.

Read more

1. Rehash old arguments AND introduce new data on why the opportunity cost theory of fertility decline is WRONG (and also A MORAL RED FLAG)

2. Boldly claim that HAVING KIDS IS GETTING CHEAPER AND EASIER OVER TIME

3. Begrudgingly grant despite this bad modeling choice the paper’s findings are approximately similar to what we already know from some other studies, and they prove cash-for-kids works

(after the paywall)

4. Empirically prove that massive tax cuts JUST FOR WIVES have actually INCREASED fertility in 2 or the 3 historic cases where this was tried

5. But actually this result is due to a stalking horse third factor: marriage. Changes in incentives to marry are the ur-factor explaining the entire universe. Proper understanding of marriage I am sure will eventually explain dark matter and prediction of very large primes.

But before we do that we need to talk about opportunity cost. And before we do that, we need to talk about sexism and fertility commentary.

Some People Just Don’t Like Women

The reason “opportunity cost” is a popular theory is not because it is well supported. I have previously shown that the opportunity cost theory of fertility decline is in fact not well supported. I am not saying there are no sincere opportunity-cost-believers. There are! Many academics simply happen to think the theory is true!

But the reason the theory is popular is because of sexism. As you will see below, the opportunity cost theory is primarily a stick with which to beat women in the workforce. Now look, if it so happens to be true that women entering the workforce is the cause of fertility decline, that would be very interesting to know! And it might imply limiting women’s labor force participation is an important part of fertility recovery! Or… it might not. We don’t fix bad eyesight by editing the genes that cause it; we fix it with LASIK or glasses. The solution to a problem often has little to do with the cause.

Unfortunately for some men online despising women is their whole personality. The reason we are discussing this topic today is not just because a paper was published, but because it has gone viral:

2.8 million views!

Now, isn’t a sexist, and his thread is I think a pretty reasonable layperson’s summary of the paper.

but let’s go ahead and take a peak at the QTs…

I could go on. I didn’t scroll much to find these. There are literally thousands of these. Like 90% of the commentary is dudes being like “I always knew Women Were The Problem, glad to have it proven.”

Two things I want to point out: first, a lot of these sexist dudes aren’t pronatalists. They think falling fertility is bad but they’re not out there slamming the RT button on child tax credit expansion proposals or something. They’re gender politics warriors first and foremost who found that birth rates were temporarily instrumentally useful for their gender politics. I would not say someone is a communist unless they believe in communism; merely happening to think that land reform is good or that Stalin had a fine mustache is insufficient for labeling someone a communist. Likewise, merely tweeting “low birth rate bad” does not a pronatalist make. Pronatalists are people who believe in pronatalism, and pronatalism refers exclusively to the ideology of “implementing policies for the purpose of having more births.” A person who incidentally favors policies that increase births is not reasonably called a pronatalist. A person who favors the child tax credit because it reduces child poverty is not a pronatalist simply by virtue of the fact they happen to support a policy which increases birth rates; they would have to support it because it increases birth rates. Likewise, men who support policies with pronatal effects primarily because those policies achieve sexist political goals and only incidentally because of birth rate effects just aren’t reasonably called “pronatalists.”

Okay, so, that out of the way, we now understand that when you see the opportunity cost argument being made, the odds you’re talking to someone who is extremely sexist rise sharply. Nonetheless, while I have a desire to simply treat the argument with mockery and derision because it’s silly, I shall now endeavor to hold my laughter in and take this argument seriously for the length of a Substack post. We’ll see how long I make it before I say something flippant and dismissive about this view.

Opportunity Cost, Shmopportunity Cost

Well I made it halfway through the first section header!

Opportunity cost refers to that which must be given up in order to select a choice. If I offer you an apple or an orange, and selecting one precludes the other, the opportunity cost of choosing the apple is giving up an orange. In terms of fertility, if I stay home and raise kids (or if my wife does), I (she) lose out on the wages I would have otherwise earned. There is a problematic analytical situation if the apple and the orange have not only opportunity costs but monetary costs. If I can afford only an apple or an orange, my opportunity cost is still the unchosen fruit, not the price of the fruit.

Suppose I choose not to buy either fruit— $0.50 will still never appear as my opportunity cost, because opportunity cost is the single most valued unchosen alternative use of resources. I’ve been told (repeatedly, by non-sexist opportunity-cost-theory-enjoyed !) that this usage of opportunity cost is idiosyncratic. Patrick is wrong, my use is standard, here’s Claude pulling it out of thin air (note: I have never discussed opportunity cost with Claude before so this isn’t a memory thing):

Notice the last sentence. Opportunity cost is always the single best alternative. Not all the best alternative, not the value of the best alternative.

Now, there’s a tricky move often pulled here. People will say “rising opportunity cost reduced fertility” but by “rising opportunity cost” they mean women’s wages rose. We’ll call this the opportunity cost theory, or OCT. That’s it that’s what they mean. This is unusual, because it is simply not the case that “women’s wages” and “fertility” are an intrinsic tradeoff. They interact with each other along some variable scale or elasticity.

Women’s wages rising are only an opportunity cost to the extent wage-earning is incompatible with work. So what “extent” is that? Well, here’s the labor earnings of women ages 25-40 with a kid under 5 at home as a % of women 25-40 without any kid at home, a sort of quick-and-dirty approximation of motherhood penalties:

As you can see, in 1940, moms of young kids earned, on average, 90% less than non-moms. In 2020, they earned about 15% less. Opportunity cost fell dramatically. It simply is the case that the extent to which women were forced to “pick one and not the other” fell precipitously over time.

If you don’t like my approximations, fine, take it from the child penalty man himself, Henrik Kleven:

Since the 1970s, the earnings penalty for motherhood has fallen a ton.

And here’s the employment penalty since 1880:

While earnings penalties from having a baby did rise from e.g. 1880 to 1940-1960, they’ve actually declined since then. Women don’t have to give up work the way they used to!

Work really is more compatible with motherhood now! That’s a great thing! Love that for society, etc, etc!

The OCT zealot may reply, “Well Lyman, it’s not about the actual effect of fertility on work, it’s about the mere prevalence of work, which exposes more women to these kinds of tradeoffs.” I won’t belabor the glaring theoretical error there but will instead just point out what the average employment rates for moms and nonmoms actually look like?

They’re only very slightly changed since 1990, even though fertility has fallen 25%.

Now, you might be able to argue that women face higher exposure to these tradeoffs because they have higher expectations of work. So even if the actual tradeoff is less binding, the share of women who expect to work and thus to whom the tradeoff applies has risen, such that Exposure X Paremeter has gone up.

Cool story, we can measure it. Since 1976 the Monitoring the Future Survey has explicitly asked high school boys and girls about their work and career expectations. Here’s the share of girls who expected to be a “homemaker” at age 30:

It did fall for girls who were seniors between 1976 and 1990. But folks, let’s be clear— it was already a very small share. I set scale at 100% so you can see just how few girls have expected to be homemakers in any recent cohort. The reality is expectations of a career are virtually universal among young women, and have been nearly universal for about 50 years. There has been no recent rise.

And here’s another question asking respondents if they think work will be “central to their lives”:

No clear directional change over time. Girls had an anti-work turn 1990-2000, then a pro-work turn 2000-2019, then COVID-related methods changes nuked survey quality.

So no, it just isn’t true that girls’ work-related expectations are rising. Across recent cohorts there’s been no substantive change in the career expectations young women bring into adulthood, and so expected exposure to opportunity costs has not risen even if direct opportunity costs had risen (they haven’t).

(Note: Special thanks to IFS colleague Grant Bailey for loaning me his Monitoring the Future skills to make these charts possible)

Seriously, Kids Are Cheaper Than Ever

Let’s think about other problems for the OCT theory. OCT suggests that what women have to give up to obtain children has increased in value.

But what do women have to give up to have kids? Or, rather, what did they have to give up? What were the actual tradeoffs?

The biggest “either/or” in historic fertility was “continuing to be alive.” Let’s consider what a woman in 1750 had to give up when she had kids. First of all, continuing to be alive was one thing she often had to give up:

Maternal mortality rates today are mostly rather low. Instead of 1% of all moms dying directly from pregnancy conditions, today it’s like 0.001%. The cost of having a baby has fallen by a very large amount.

The decline in maternal mortality is much bigger than the overall rise in life expectancy so even if you multiply MMR by remaining life years the decline is enormous. The amount of “life” women give up to have kids has fallen by a HUGE amount.

And by the way, research shows when MMR falls, fertility rises! Part of the Baby Boom was just that sulfanomides made it less likely that “having a baby” meant closing off the choice of “staying alive!” We actually know for a fact that when a truly binding opportunity cost constraint got easier (i.e. dying), fertility rose in response!

What else might have risen? Well, you can imagine that one thing that might have gone up is the consumption you lose due to spending on kids. Maybe raising kids has risen in price over time and so you forego more consumption to have kids.

But… has raising children gotten more expensive? I know people think the answer is yes, but using the only actual empirical data on childrearing expenditures, the answer is… No (according to USDA)!

In 1960, the estimated cost to raise a child was equal to almost 11 years of income. In the 2010s it was under 6 years. Kids are getting quite a lot cheaper in absolute budgetary terms! You don’t have to sacrifice as large a share of your income to buy the stuff kids need now as you did in 1960! The consumption penalty for having kids has shrunk over time! You don’t have to give up as much as you used to!

Now look— it’s certainly the case that you can find specific areas where there may be big tradeoffs. For example, I’ve previously written about how losing out on travel may be a big issue impacting fertility. But that’s a weird case for OCT, since what happened there is not that the cost of kids rose, but that the cost of travel declined so much that new travel norms and expectations emerged. If what you mean by “opportunity cost” is actually “newly emergent cultural norms” then you should say “newly emergent cultural norms” not “opportunity cost.” This is the same issue we saw for career expectations above; people use “opportunity cost” in a bizarre blanket way which simply doesn’t fit the case. If you think people aren’t having kids because they want to spend more of life focused on career instead of kids, that could reduce fertility regardless of the actual impact of fertility on careers (i.e. the actual opportunity cost). If you think people aren’t having kids because they want to travel more, that could reduce fertility regardless of the actual extent of incompatibility between kids and travel.

The reason people use “opportunity cost” to discuss theories that aren’t about opportunity costs is because “opportunity cost” carries the prestige of economics and empirical social science, and so it can be used to dress up other theories. It’s intellectual stolen valor.

But Didn’t The New Study Show Women’s Wages Reduce Fertility?

The new study we’re discussing has several substantial methodological issues. For example, they use data on Danish labor earnings to estimate how responsive husbands vs. wives are to changes in income and taxation; so far, so good (note: yes, as best I can tell, this is a study estimated using partnered people, not men and women generally, which is problematic for inferring overall effects). They get these results:

The first row for women/men is the substitution effect, the second is the income effect. Column three is the best model here. Income effects mean “The effect of just having more money on having a baby.” Substitution effects mean “The effect of getting to keep more money from an extra hour of work.” So you can see that income effects are positive and significant for men and women: 0.005 for women, 0.028 for men. Substitution effects are negative for women (-0.023) and insignificant for men. This tells us that when men or women “get extra money,” they have more babies. It tells us when men face a higher return to working outside the home as an employee, they don’t change their fertility. But when women face a higher return, their fertility falls.

Okay, so far, uncontroversial. This is similar to a recent finding from Finland: when low-income Finnish women received a much lower tax rate to get work, their fertility declined; when their husbands received a lower tax rate, fertility rose. In this case I find the Finnish example more convincing because it was literally a randomized experiment in tax rates! The net effects there are interesting: reducing husbands’ taxes raised births by 31% after 3.5 years, whereas reducing womens’ taxes reduced births by 17%; the effect was driven entirely by declines among married women (ultimately a 28% decline). The Danish study is only of partnered women.

So we have two studies showing that when husbands have lower taxes, fertility rises; when women have lower taxes, fertility falls. Both studies have core estimates that seem to show bigger effects for men: in the Danish study, men have larger effect sizes for income and substitution effects, and for women the income and substitution effects run opposite directions. In the Finnish study, identical tax rate changes reduced fertility by 28% for married women, or increased by 31% for married men. So tax changes impact men’s fertility more, as long as you read the tables.

Now, the Danish study then provides a second estimate:

This one doesn’t separate income and substitution effects. It collapses them together. Here, now, you can see the effects are bigger for women, but entirely a function of low-skilled workers. The Finnish study was of lower-income people as well.

Okay, so we’ve added another paper to studies showing that when women have caretaker responsibilities, increasing the returns to not-being-a-caretaker reduces fertility by more than their income effects increase it, whereas for men there are positive effects. Cool! Nice to add to that!

But then you come to the figures. Consider:

Do you notice anything weird there? I sure do! First of all, notice the assumption here is a permanent 5% wage change. In principle, permanent in expectation. So this model assumes that when families get a 5% wage increase at age 25, they assume their income will therefore be 5% higher forever. Given we’re studying wage increases from tax cuts, is that plausible? Is it actually guaranteed that a tax cut of 5% is going to last your entire working career? What odds do you give it? Tax cuts are not shocks to permanent income unless taxpayers believe the income is permanent.

But also notice a weird thing: the effects on women here are tons bigger than for men! The effects on women are like four times the size of the effects on men! That’s weird, since the tables above showed pretty similar sizes! Also notice the effects for skilled and unskilled workers seem to be about the same, which, again, is weird, because effects for high-skilled workers were like 1/3 the size in the actual tables.

What’s going on here is a bait and switch. The authors estimated a parameter using annual Danish tax returns for labor income. Then they calibrated a totally separate model using totally different estimates in a totally different data source (weekly contract hours from a paystub processing system). These aren’t even comparable estimates since the paystub system excludes a lot of labor income related to entrepreneurship and the public sector.

Did the authors notice the problem? Yes:

Here they compare their model estimates to the estimates from Table 3. Notice, first, that their model estimates are not very close to the data estimates: for low-skilled women, their elasticity is 25% too small; for low-skilled men, it’s 78% too small; for high-skilled women it’s 9% too big; for high-skilled men it’s 86% too small. They claim their elasticities are valid because they are within the confidence intervals, but c’mon, those confidence intervals run from “basically zero implied elasticity” to “an elasticity of 1.” By saying “It’s in the CI, good enough” the authors are giving themselves permission to select almost any elasticity between 0 and 1. I don’t think people have digested that the confidence intervals here really do seem to say “the effect of a 5% change in wages could be anywhere from a -3% to a positive 4% effect for men, and a -4.5% to 1% change for women.” Those are huge margins of error! If I said “The effect of a policy is births might go up or down or maybe not change” you’d say “Yes, we knew that before we asked you.”

The fundamental problem here is that the authors knew their elasticities wouldn’t be very sexy, so they made graphs showing simulated (not actually observed!!) “life cycle effects” using other data that would be more shareable. And indeed, credulous readers have shared them, as I showed in the tweet above.

The problem is the paper claims to be discovering an empirical fact (and it did!), but then all the charts being shared from it are not actually based on that empirical fact. That’s a pretty egregious approach in my view. Make a chart that actually uses the estimates you derived from the data, the exact estimates!

But look, I’m sympathetic to these broad findings. They tell us that when women face a 5% increase in returns to work, fertility maybe falls ~0.3-2%. Call that an elasticity of 0.2. That implies the tradeoff on “cash-to-care-work” is about 0.2. My recent analysis of financial incentives suggested a tradeoff of “cash-to-kids” of about 0.15 to 0.35, so guess what, these different methods arrived at similar overall elasticities, though in one example women are being paid not to do care work and in the other they’re being paid to do care work.

And what’s the policy takeaway?

Give a child allowance!

Or, if you prefer, increase maternity leave pay. It’s funny to me that guys are sharing this paper to say “We must tax women’s wages!” when they paper’s actual finding is “Increasing maternity leave pay would boost fertility” and “A bigger child allowance would boost fertility.”

The takeaway from this model isn’t “see, feminism caused low fertility.” The takeaway from this model is “financial incentives for kids will successfully raise fertility.” Now, I still think it’s a bad way to do a model, but the conclusions I will begrudgingly grant are similar to mine.

But if we really want to test this theory, we should look for real-world cases where the tax rate on wives was changed by a HUGE amount. As it happens, we actually have cases where governments MASSIVELY reduced taxes on wives. This has happened at least three times where we have sufficient data to study what happened and, past the paywall, that’s what we’ll do.

Read more