Fight Me, Psychologists: Birth Order Effects Exist And Are Very Strong

“Birth order” refers to whether a child is the oldest, second-oldest, youngest, etc. in their family. For a while, pop psychologists created a whole industry around telling people how their birth order affected their personality: oldest children are more conservative, youngest children are more creative, etc.

Then people got around to actually studying it and couldn’t find any of that. Wikipedia’s birth order article says:

Claims that birth order affects human psychology are prevalent in family literature, but studies find such effects to be vanishingly small….the largest multi-study research suggests zero or near-zero effects. Birth-order theory has the characteristics of a zombie theory, as despite disconfirmation, it continues to have a strong presence in pop psychology and popular culture.

I ought to be totally in favor of getting this debunked. After all, the replication crisis in psychology highlights the need to remain skeptical of poorly-supported theories. And some of the seminal work disproving birth order was done by Judith Rich Harris, an intellectual hero of mine who profoundly shaped my worldview with her book The Nurture Assumption.

So I regret to have to inform you that birth order effects are totally a real thing.

I first started thinking this at transhumanist meetups, when it would occasionally come up that everyone there was an oldest child. The pattern was noticeable enough that I included questions about birth order on the latest SSC survey. This blog deals with a lot of issues around transhumanism, futurology, rationality, et cetera, so I thought it would attract the same kind of people.

7,248 people gave me enough information to calculate their birth order, but I am very paranoid because previous studies have failed by failing to account for family size. That is, people of certain economic classes/religions/races/whatever tend to have larger family sizes, and if you’re in a large family, you’re more likely to be a later-born child. In order to be absolutely sure I wasn’t making this mistake, I concentrated on within-family-size analyses. For example, there were 2965 respondents with exactly one sibling…

…and a full 2118 of those were the older of the two. That’s 71.4%. p ≤ 0.00000001.

The same effect occurs in sibships of other sizes. Of the 1884 respondents from families with three children (n = 1884), 56.8% are the oldest, compared to predicted 33%. In families with four children (n = 765), 48.2% are the oldest, compared to predicted 25%.

Number of responses by birth order in sibships of different sizes; graph is by Emile and uses the public data only, which means exact numbers may be slightly different

This effect reaches the same scale as other effects people consider important. For example, the survey population drew heavily from STEM fields and was predictably very white; however, the birth order gap was larger in magnitude than the racial gap. It is bigger than gender gaps in some fields traditionally considered to have major gender gaps, like undergraduate economics. This can fairly be considered a large effect.

So what is going on here?

It’s unlikely that age alone is driving these results. In sibships of two, older siblings on average were only about one year older than younger siblings. That can’t explain why one group reads this blog so much more often than the other.

And all of the traditional pop psychology claims about birth order don’t seem to hold up. I didn’t find any effect on anything that could be reasonably considered conservativism or rebelliousness.

But there is at least one reputable study that did find a few personality differences. This is Rohrer et al (2015), which examined a battery of personality traits and found birth order effects only IQ and Openness to Experience, both very small.

I was only partly able to replicate this work. Rohrer et al found that eldest siblings had an advantage of about 1.5 IQ points. My study found the same: 1.3 to 1.7 IQ points depending on family size – but this did not reach significance. My other measure of intelligence was SAT, but SATs have been renormed and changed so many times over the past few decades that making apples-to-apples comparisons were really tough. I was able to get only a couple of weak and inconsistent effects: in sibships of two, eldest children had a slightly higher SAT1600 (1481 vs. 1458, p = 0.002) but not SAT2400; in sibships of 3+, eldest children had a slightly higher SAT2400 (2214 vs. 2248, p = 0.03), but not SAT1600. Overall this seems way too weak to say anything with certainty. Average SATs and IQs were already around the 99th percentile, so there may have been too much of a selection effect / ceiling effect to get good results.

The Openness results were clearer. Eldest children had significantly higher Openness (73rd %ile vs. 69th %ile, p = 0.001). Like Rohrer, I found no difference in any of the other Big Five traits.

Because I only had one blunt measure of Openness, I couldn’t do as detailed an analysis as Rohrer’s team. But they went on to subdivide Openness into two subcomponents, Intellect and Imagination, and found birth order only affected Intellect. They sort of blew Intellect off as just “self-estimated IQ”, but I don’t think this is right. Looking at it more broadly, it seems to be a measure of intellectual curiosity – for example, one of the questions they asked was, “I am someone who is eager for knowledge”. Educational Testing Service describes it as “liking complex problems”, and its opposite as “avoiding philosophical discussion”.

This seems promising. If older siblings were more likely to enjoy complex philosophical discussion, that would help explain why they are so much more likely to read a blog about science and current events. Unfortunately, the scale is completely wrong. Rohrer et al’s effects are tiny – going from a firstborn to a secondborn has an effect size of 0.1 SD on Intellect. In order to contain 71.6% firstborns, this blog would have to select for people above the 99.99999999th percentile in Intellect. There are only 0.8 people at that level in the world, so no existing group is that heavily selected.

I think the most likely explanation is that tests for Openness have limited validity, which makes the correlation look smaller than it really is. If being an eldest sibling increases true underlying Openness by a lot, but your score on psychometric tests for Openness only correlates modestly with true underlying Openness, that would look like being an eldest sibling only increasing test-measured-Openness a little bit.

(cf. Riemann and Kandler (2010), which finds that the heritability of Openness shoots way up if you do a better job assessing it)

If we suppose that birth order has a moderate effect size on intellectual curiosity of 0.5 SD, that would imply that science blogs select for people in the top 3% or so of intellectual curiosity, a much more reasonable number. Positing higher (but still within the range of plausibility) effect sizes would decrease the necessary filtering even further.

If this is right, it suggests Rohrer et al undersold their conclusion. Their bottom line was something like “birth order effects may exist for a few traits, but are too small to matter”. I agree they may only exist for a few traits, but they can be strong enough to skew ratios in some heavily-selected communities like this one.

When I asked around about this, a couple of people brought up further evidence. Liam Clegg pointed out that philosophy professor Michael Sandel asks his students to raise their hand if they’re the oldest in their family, and usually gets about 80% of the class. And Julia Rohrer herself was kind enough to add her voice and say that:

I’m not up to fight you because I think you might be onto something real here. Just to throw in my own anecdotal data: The topic of birth order effect comes up quite frequently when I chat with people in academic contexts, and more often than not (~80% of the time), the other person turns out to be firstborn. Of course, this could be biased by firstborns being more comfortable bringing up the topic given that they’re supposedly smarter, and it’s only anecdotes. Nonetheless, it sometimes makes me wonder whether we are missing something about the whole birth order story.

But why would eldest siblings have more intellectual curiosity? There are many good just-so stories, like parents having more time to read to them as children. But these demand strong effects of parenting on children’s later life outcomes, of exactly the sort that behavioral genetic studies consistently find not to exist. An alternate hypothesis could bring in weird immune stuff, like that thing where people with more older brothers are more likely to be gay because of maternal immunoreactivity to the Y chromosome (which my survey replicates, by the way). But this is a huge stretch and I don’t even know if people are sure this explains the homosexuality results, let alone the birth order ones.

If mainstream psychology becomes convinced this effect exists, I hope they’ll start doing the necessary next steps. This would involve seeing if biological siblings matter more or less than adopted siblings, whether there’s a difference between paternal and maternal half-siblings, how sibling age gaps work into this, and whether only children are more like oldests or youngests. Their reward would be finding some variable affecting children’s inherent intellectual curiosity – one that might offer opportunities for intervention.

If you want to double-check these results or analyze them further, you can download the data as .xlsx or .csv. Some people have complained of weird problems in the csv format and I recommend the xlsx if at all possible. I have removed the data of a few people who did not want their answers to be public, so you may not get exactly the same numbers I did, but they should be pretty close. If you think this could be turned into a paper and are interested in making it happen, please get in contact with me.