Birth Order: Mutational Load or Environment?

Responding to Cremieux

Cremieux writes:

Birth order: older children tend to be more successful than younger ones when they reach adulthood. The question is why, and one answer is "mutational load"—later-born siblings have more mutations due to older fathers’ gametes having had time to accumulate additional mutations.

Another answer is phlogistonal preference—in some phlogistones, earlier-born children receive greater shares of bequests, favored roles in the family, etc. If resources given to kids matter for their development, then even simple resource dilution could explain birth order effects.

There are additional, more specific theories to be sure, but I want to contrast biological and nonbiological theories. I suspect that when it comes to birth order effects, biology isn’t dominant.

Ignoring the epistemic issues with phlogiston¹, there is certainly evidence for environment causing some component of some birth order effects. However, contrary to Cremieux, I suspect that biology is dominant.

Here’s why. Most traits have a null or near null shared environmentability component. This component is essentially a measure of how much differences in parental behavior between families effect a trait. If parental behavior between families doesn’t have impact on a trait, then parental behavior varying by child within a family will probably not affect that child’s trait.

Cremieux basically repeats Inquisitive Bird’s recent error. Most traits do not respond to differences in parental behavior, but both of these writers focus on rare exceptions to that general rule, and from exception they theorize that the rule does not exist.

Cremieux’s conclusion in his article is:

There may be some birth order effects that are attributable to mutational load, but whatever they are isn’t clear to me. From the vantage point of designs leveraging phlogistonal differences, adoptions, and taking on different spouses, birth order effects on educational attainment and achievement look to be both real, replicable, and not attributable to mutational load. … So far, mutational load doesn’t seem like it explains any birth order effects of note.

Bolding is mine, and it highlights his reasoning. Cremieux generalized from outlier traits to everything. The shared environmentability of educational attainment is 40%. This is much higher than the shared environmentability of IQ, leftism, ADHD, schizophrenia, and so on. For all of these traits, there are documented birth order effects, and varying degrees of evidence that they are driven by paternal age, i.e. mutational load.

ADHD

Source

The heritability of ADHD is 100% when adjusted for measurement noise (h^2 = 2(.7 - .2) = 1, but MZ twins correlate at .7 instead of 1), and the shared environmentability is 0.

Source

Yet there is a paternal age effect on ADHD, similar in magnitude to the ones measured for IQ and leftism (r ~= .10). It’s extremely unlikely this could come from parental behavior. The estimate above uses family fixed effects, so it doesn’t come from differences between younger fathers and older fathers (it would seem from the chart that ADHD fathers breed younger, obscuring the effect in a simple design).

From the same study:

This is the only set of results from the study where the family fixed effects deviates from the linearity that mutational load theory predicts.

I think there is a mutational load effect on education, but it’s covered up by parental compensation. This theory is supported by Cremiuex’s first citation in his article, and explains the rest of the results he presents, while also allowing for there to be a paternal age effect on IQ, which correlates with education measures weakly to moderately.

Whether there is a mutational load effect component on education or not, you can clearly see that Cremieux picked the least representative trait to generalize from. Of 7 or 8 different traits from the study I’m citing, Cremieux picked the 1 or 2 that deviate from mutational load theory, and then generalized from this. (I doubt he literally plucked education from this study, but you get my point — the broader picture is missing from his article).

Leftism and IQ

The p values are bad, but for leftism and IQ , independent paternal age effects haven’t been falsified. They also aren’t fully verified due to massive understudying of how these traits evolve. This understudying is insane because it’s easy to argue that these are the two most important traits.

But there do seem to be paternal age effects on IQ and leftism. It stands to reason that this is more likely to be due to mutational load the lower the shared environmentality of these traits are.

Source

Source

IQ’s shared environmentality is much less than that of educational attainment (15% vs. 40%). Conservatism’s is 0 when your IQ is greater than 100 and you’re an adult. As IQ decreases, parental influence goes up, but not the the levels of educational attainment. Overall, it was estimated at 5.4% for the population.

So a paternal age effect on IQ is unlikely to be due to environmental birth order effects, and a paternal age effect on leftism is even more unlikely to be due to environmental birth order effects.

Under reasonable priors, we must disagree with Cremieux. Mutational load DOES seem like it explains MANY birth order effects of note.

SSSM Priors

So what drove Cremieux to make what I consider to be mistakes?

Source

If you look at Cremieux’s work, he likes to talk about phlogiston (“culture”) a lot! In fact, I thought he would take this position if he wrote on this topic because of his positions on stuff like “the importance of culture”. Clearly this goes beyond the evidence we’ve been passing around — frankly a lot of research on this topic is still ambiguous to a degree and therefore has to be informed by priors much more than is the case in the race and IQ field.

And what priors are those? See the table above. Cremieux believes, without any firm falsifying evidence, that with respect to birth order effects and other areas of human behavior, that “Biology is relatively unimportant in understanding behavior.”

Obviously, Cremieux is not as SSSM as non-HBD people. But he’s more SSSM than me. SSSM is a spectrum. And within the HBD scene, there are differences in general SSSM priors, there’s just less variance than in the general population. I believe these differences within HBD drive within group disagreements on relatively niche topics like mutational load.

I believe that biology is always the null hypothesis, and that good falsifying evidence must be provided for any special exceptions to this general rule. Therefore, I could never say something like “Needlessly Invoking Biology”, which is what Cremiuex writes of mutational load theory with respect to birth order effects.

It’s near certain mutational load explains birth order effects like ADHD. Invoking biology is never needless — it’s the default that research on birth order effects should revolve around. My priors go even beyond the integrated model — I would say “Understanding genetic effects is absolutely critical to understanding nearly all human behavior and its variation.” With this in mind, let’s review his evidence as hereditarians.

He begins by mentioning that parents tend to compensate for their children’s flaws by putting more effort into less abled children. If this is the case, then we should expect mutational load effects to be attenuated by greater parental effort any time a trait responds to such effort. This is important to note because now we have a model where there are competing pressures on birth order effects. They may cancel each other out partially or totally.

With this in mind, Cremieux seems to post data that shows the opposite — older parents environmentally persuade kids to go to college less. Perhaps going to college is bad? Bryan Caplan would argue this. This chart also shows that there is indeed no need to invoke biology. But biology isn’t falsified — if higher paternal age drives equal paternal compensation toward college, but older mothers are less active in their kid’s lives due to declining energy and so on, we would expect to see this. 3 effects, 1 affects adoptees and the other two cancel each other out in line with Cremieux’s parental compensation data and the broader research on mutational load effects.

My theory here also explains this chart. Older fathers fully compensate behaviorally for their later children’s lower IQs, but older mothers actually decline in effort leading to their later children attending college less.

Finally, race differences (which are probably genetic) in compensatory behavior and mutation rate (this book discusses how mutation rate is evolving. If it’s evolving it’s heritable, and if it’s heritable, there will likely be genetic differences between races) can explain this chart. Have you heard of East Asian tiger moms? Remember my theory above about white mothers losing effort? Maybe East Asian mothers don’t do that. Maybe Indian mothers increase is parenting behavior as they age.

Occam’s razor

At first glance, what I just did violates Occam’s razor. Cremieux appears to have a theory with exactly 2 less moving parts. But my theory is more logically coherent and fits the broader body of evidence better — “theories shall be as simple as possible, but no simpler than that.” Cremieux’s theory is not fit for explaining everything that my priors are built on, so we need more complexity.

This will sound unconvincing to people who don’t share my priors. Maintaining my moderate usage of their lingo in this article, the Bay Area Rationalists blamed this on something called inferential distance. My priors are informed by everything I know, plus my genetics, so we run into an issue. People who would respond well to knowing more are probably light readers, unless they’re very young or new to the space. Therefore, they won’t want to read everything I’ve ever written and a few other books. The distance remains. Everyone else who still disagrees probably does so because their inherited starting point is a lot different than mine, because they’ve been exposed to all of the same information as me.

More data

If elites like Cremieux are information saturated, meaning evaluations have extremely high heritability, then more data is the only way to achieve a consensus on a topic. More data doesn’t have to work. In general, if someone is convinced by the black-white IQ data, then I think they will be convinced of mutational load by more data. But if they are still denying black-white IQ gap stuff, don’t count on them accepting mutational load even with excellent data. Cremieux, I believe, will be convinced by better data if my predictions are right — Kevin Bird, probably not.

As I said earlier, it’s easy to argue IQ and leftism are the two most important traits. Proving mutation driven paternal age effects for these two traits will therefore do a lot more to show that mutational load is important.

To do this, we need family level data — because of the low c^2 of leftism among the top 50% of the population, this should be enough for leftism is we do something like taking college educated families.

More than this, if we could use family level gene sequencing to measure the number of de novo mutations in offspring (paternal age only reflects the expected number), then we could correlate this variable controlled for parental age, birth order, and so on. We could also check the effect of the mutations on PGSs, but this is likely not possible to do currently due to missing heritability.

If you would like to give your energy to me to one day get such a study done, please subscribe:

1. This is me standing by my stance to not use the concept of “culture”, which I believe to be an anti-concept with over 150 different definitions. When I would use it, I replace it with “behavior”, “population”, “race”, “information”, “meme”, “memome”, or “genetics” most of the time. It can also be replaced with “art”, “extended phenotype”, “technology”, and “tools” in some cases, etc up to over 150 cases. When someone else uses it, instead of doing their work for them, I just replace it with phlogiston, the unseen substance people used to claim caused fire. This is fitting because “culture” is just as vague and just as mystical, with many weakly implying it exists somehow metaphysically as an unmeasurable and underlying determinant of whatever more specific word they’re using it in place of.