Birth Order and IQ: Small Effect, Large Overlap

0 The Short Answer

Firstborns score very slightly higher on intelligence tests, by something on the order of one to one and a half points, and birth order has no detectable lasting effect on personality at all. Those two sentences come from the same paper, and together they overturn most of what people believe about siblings. The responsible eldest, the overlooked middle child, the rebellious baby of the family: none of that survives contact with large data. The intelligence gap does survive, but it is so small that it tells you nothing useful about any particular pair of brothers or sisters.

The evidence that settled this comes from Julia Rohrer, Boris Egloff and Stefan Schmukle, whose 2015 paper in the Proceedings of the National Academy of Sciences pooled three national household panels: the United States (n = 5,240), Great Britain (n = 4,489) and Germany (n = 10,457), more than 20,000 adults in total. They looked for birth-order effects on the Big Five personality traits and on measured intelligence, using both comparisons between families and comparisons between siblings inside the same family. Personality came back empty across every trait they tested. Intelligence came back with a small, consistent firstborn advantage.

What makes this worth a full article rather than a headline is the gap between those two results and how people use them. A one and a half point average difference is a genuine population fact and a worthless individual prediction, and almost nobody explains why. The rest of this page walks through the design choices that make the finding trustworthy, the confounds that ruined the older literature, the honest size of the effect, the proposed mechanisms and their evidentiary status, and the general skill this case teaches better than any other topic in psychometrics: how to read a real average that predicts nothing about you.

1 What the 2015 Study Actually Did

Birth order research had a credibility problem long before 2015. The literature was enormous, the effect sizes were all over the map, and the studies rarely agreed with each other. Rohrer and colleagues did not add another small sample to the pile. They attacked the design problem directly, and that is why their paper is the reference point today.

Three features of the design do the work. First, scale: pooling three national panels gave them enough statistical power to detect effects far smaller than anything a person could notice in daily life, which means a null result carries real weight rather than being a shrug about insufficient data. Second, cross-national replication: the United States, Great Britain and Germany have different family sizes, different school systems and different cultural scripts about eldest children, so a finding that appears in all three is unlikely to be a quirk of one country's sampling. Third, and most important, they ran the analysis two ways, comparing people from different families and comparing siblings raised inside the same family, then checked whether the two approaches agreed.

That last choice is the methodological heart of the paper, and section seven explains why in detail. For now, the short version: comparing a firstborn from one household with a thirdborn from another household is not a clean comparison, because thirdborns necessarily come from larger families, and family size travels with income, parental education and a dozen other things that predict test scores on their own. Comparing an eldest sibling with a younger sibling from the same household removes all of that in one stroke, because both children share parents, household income, neighborhood and genes to a large degree.

The measures matter too. The panels contained objectively administered cognitive tests as well as self-reported intellect, a Big Five facet covering things like vocabulary and comfort with abstract ideas. Splitting those two apart turned out to be informative in a way the authors did not have to plan for, and section four returns to it.

2 The Personality Result Is a Clean Zero

Across extraversion, emotional stability, agreeableness, conscientiousness and imagination, Rohrer and colleagues found nothing. Not a small effect, not an effect in some countries and not others, not an effect that appeared in one analysis and vanished in another. Their conclusion was that birth order does not have a lasting effect on broad personality traits outside the intellectual domain.

Null results are usually weak evidence, because failing to find something is what happens when you do not look hard enough. This one is different, and the reason is power. With more than 20,000 participants and multiple analytical strategies, the study could have detected effects small enough to be invisible in ordinary life. It detected none. When a well powered design goes looking for a widely believed effect and comes back empty across three countries and two study designs, the reasonable inference is that the effect is not there.

An independent line of evidence points the same way. Rodica Damian and Brent Roberts, working with Project Talent, a nationally representative sample of roughly 377,000 United States high school students, published their results in the Journal of Research in Personality the same year. They controlled for age, sex, sibship size, parental socioeconomic status and family structure. The average absolute association between birth order and personality traits came out at .02, and the association with intelligence at .04. Roberts put the practical meaning bluntly in the university's announcement of the work: you are not going to be able to sit two people down next to each other and see the differences between them.

Two independent research groups, two different data sources, two different countries of origin, same answer in the same year. That is about as close to a resolved question as personality research gets. Damian and Roberts went on to write the commentary that accompanied the Rohrer paper in PNAS, and the title they chose, on settling the debate, was not modest by accident.

3 Why the Stereotype Felt So Convincing

If the personality effect does not exist, the interesting question becomes why practically everyone believes it does. The belief has a long pedigree and several reinforcing sources, which is exactly the profile of an idea that survives without evidence.

It starts with Alfred Adler in the early twentieth century, who built sibling position into his account of personality development and gave the world the dethroned eldest and the pampered youngest as character types. It got a modern scientific push from Frank Sulloway's 1996 book Born to Rebel, which argued that laterborns occupy a different niche in the family and grow up more open to unconventional ideas, using historical figures and their positions on scientific revolutions as evidence. The thesis was elegant, well written and widely absorbed. It has not held up in the large modern samples.

Then there are the everyday mechanisms that keep the belief alive regardless of data. Family roles are real while children are children: an eight year old genuinely does supervise a four year old, and the four year old genuinely does get away with more. What the panels show is that those roles do not print themselves permanently onto adult personality. People also remember the confirmations and forget the exceptions, and the categories are loose enough that almost any adult can be fitted to their assigned label after the fact, the same way horoscopes work. Add the fact that a firstborn label is flattering and a youngest label is fun, and you have an idea with no natural enemies.

There is a cost to this beyond trivia. Birth-order beliefs shape how parents interpret their own children, how teachers read a class list, and how adults explain their own difficulties. An explanation that feels satisfying and is not true crowds out the ones that are, and the actual predictors of a child's cognitive development, which include family resources, health, schooling and language environment, are considerably less charming than a birth certificate.

4 The Intelligence Result: Real, Replicable, Tiny

The cognitive finding went the other way. Firstborns scored higher on objectively measured intelligence, the pattern held within families as well as between them, and the magnitude landed around a point and a half on the familiar scale where the population mean is 100 and the standard deviation is 15. Damian and Roberts, from completely different data, put their largest observed gap at approximately one point.

Two details are worth pulling out. The first is the shape of the effect: it is not a firstborn bonus followed by nothing, it is a gentle gradient. Scores decline slightly with each step down the birth order, so the gap between a first and a third child is larger than the gap between a first and a second, though every one of these gaps is small.

The second detail is the self-report result, which is where the paper gets subtle. Rohrer and colleagues found a decline of about a tenth of a standard deviation in self-reported intellect with increasing birth-order position, meaning later-born participants were somewhat less likely to describe themselves as having a rich vocabulary or as finding abstract concepts easy. That self-assessment gap persisted even after accounting for the objectively measured difference. In other words, the belief about the gap is bigger than the gap. Growing up second appears to leave a faint mark on how people rate their own intellect that measured performance does not fully justify, which is a nice reminder that self-estimates of ability are a different construct from ability, and a worse one.

Neither result licenses the conclusion people reach for. A gradient of roughly a point per position, on a scale where the middle 50 percent of the population spans about 20 points, is not something a school, an employer or a parent could ever act on. Schmukle described the effect precisely: it replicates very well in large samples, but it is barely meaningful on the individual level, because it is extremely small.

5 Putting One and a Half Points on the Scale

Numbers only mean something against a ruler. Here is the ruler, using the modern deviation scale with a mean of 100 and a standard deviation of 15, and here is where a birth-order gap sits on it. If you want the fuller version of this mapping, the percentile chart lays out the whole distribution.

Read those cards together and the practical verdict writes itself. The firstborn advantage is smaller than the noise in a single testing session. If you tested the same person twice on the same battery a month apart, the difference between their two scores would routinely exceed the entire effect that decades of birth-order research have been arguing about. That does not make the effect fake. It makes it a fact about populations that has no individual application, and confusing those two categories is the single most common error people make with cognitive statistics. For the broader question of what these scores represent in the first place, see what IQ tests actually measure.

6 Four Times Out of Ten, the Younger Sibling Wins

Schmukle offered the most useful sentence in the entire birth-order literature when describing the study's implications: in four out of ten cases, the later-born is still smarter than his or her older sibling. Sit with that number, because it is what a one and a half point average difference looks like when you translate it from population language into sibling language.

The mechanics are worth understanding rather than memorizing. Two overlapping distributions with means separated by a tenth of a standard deviation are, visually, almost the same distribution. If you drew them, you would need a careful eye to see that one was shifted at all. Picking one person at random from each and asking which scores higher is close to a coin flip, tilted a few percentage points toward the higher-mean group. Forty percent versus sixty percent is that tilt. It is a real tilt, detectable with 20,000 people, and it is nowhere near enough to predict a single case.

Compare that with the intuition the stereotype produces. People do not hear about a firstborn advantage and picture a 40/60 split. They picture something more like a rule with occasional exceptions, and they apply it to families they know, which contain two or three data points. With two siblings, the observed pattern is essentially random relative to the effect. You will see plenty of families where the youngest is clearly the sharpest, and that observation tells you nothing about whether the population effect exists, in either direction.

This is the transferable skill. Every reported average difference between groups, on any trait, comes with an overlap that the headline never mentions, and the overlap is usually enormous. When you next read that group A scores higher than group B on something, the question that turns the claim into information is not whether the difference is statistically significant, it is how often a randomly chosen member of B beats a randomly chosen member of A. For most published psychological differences the answer is close to half the time. Birth order happens to be the cleanest teaching example available, because the study authors did the translation for us.

7 Between Families and Within Families Are Different Questions

Here is where a great deal of older birth-order research went wrong, and why the distinction deserves a section of its own rather than a footnote.

Between-family designs take a large sample of individuals, record each person's birth position, and compare the average scores of firstborns, secondborns and thirdborns across the whole sample. This sounds reasonable and contains a trap. Thirdborns can only exist in families with at least three children. Firstborns exist in families of every size, including one-child families. So when you compare the firstborn group with the thirdborn group, you are also comparing smaller families with larger families, and family size is correlated with parental income, parental education, maternal age at birth and everything those variables carry with them. Any score difference you find is a mixture of birth position and family background, and the design cannot separate them.

Within-family designs compare siblings to each other. The eldest and the youngest child of the same household share their parents, their household income bracket, their neighborhood, their schools in most cases, and about half their genes. Comparing them holds all of that roughly constant by construction, so whatever difference remains has a much better claim to being about position in the sibling sequence itself.

The critique is not hypothetical. When the Norwegian conscript results made news in 2007, the psychologist Joseph Rodgers argued in press coverage that between-family birth-order studies simply are not valid for this question, because of exactly this confounding, and he has made versions of that argument in the academic literature for years. His position is that much of the classic birth-order literature measured family size and social class while believing it was measuring sibling rank.

What makes the modern evidence persuasive is that the within-family analyses did not make the effect disappear. They shrank it and cleaned it up. The cognitive gradient shows up when you compare siblings raised in the same house, which is the comparison that matters, while the personality effects vanished under both designs. That combination, one effect surviving the strict test and the other failing it, is what a real finding and a real null look like sitting next to each other.

8 The Norwegian Data and a Grim Natural Experiment

The largest and most cited data on this question comes from Norway, where male conscripts took a standardized cognitive test at 18 or 19 as part of compulsory military service, generating a national register with over a quarter of a million scores recorded between 1984 and 2004.

Petter Kristensen, Tor Bjerkedal and colleagues published two analyses of that register in 2007. The paper in the journal Intelligence reported the basic gradient in the raw units of the conscript scale, a nine point scale: among adjacent sibling pairs, the elder brother averaged 5.18 and the younger 4.93, a difference of about a quarter of a point. On a scale whose standard deviation is roughly two points, that works out to somewhere near two points once converted to the 15 point IQ metric, consistent with the size reported everywhere else. The authors also noted that the patterns and magnitudes were rather similar between families and within families, and that score differences tracked maternal education, marital status, paternal income, sibship size and birth spacing.

The companion paper in Science is the one that gets quoted, because it exploited a natural experiment nobody would design on purpose. Among 241,310 conscripts, the register identified families in which an older sibling had died in infancy or early childhood. In those families, a biological secondborn grew up as the social eldest. Their scores looked like firstborn scores. The same held one step further down: thirdborns who lost both older siblings scored like firstborns. The interpretation the authors drew is that what matters is rank within the family as it is actually lived, not the order of birth as a biological fact.

That result is genuinely clever and should still be held loosely. Rodgers raised two objections worth carrying: families that lose an infant may differ systematically from families that do not, in health, nutrition or other exposures that also affect the surviving children, and he suspected a problem in how the scores had been standardized over two decades of rising raw performance, the Flynn effect, which could favor earlier-born brothers who were tested in earlier years. Neither objection demolishes the finding. Both are the kind of thing that should stop you from treating a single striking study as settled mechanism.

9 What Family Size and Income Are Doing in the Background

The variables that contaminate between-family comparisons are not statistical nuisances to be waved away. They are large effects in their own right, considerably larger than the thing being studied, which is precisely why they are dangerous. The same confounding structure dominates the breastfeeding and IQ literature, where adjusting for maternal IQ alone absorbs a third of the raw effect.

• Family size. Average measured scores decline as sibship size grows, and this pattern is visible in the Norwegian register and in most national datasets. Because the number of children a family has is bound up with parental education, income, region and generation, this decline is not evidence that having siblings costs you points. It is mostly evidence that families of different sizes differ in many other ways.
• Parental socioeconomic status. Parental education and income predict children's test performance robustly, through nutrition, health care, language environment, book access, school quality and stability. In a between-family comparison, any correlation between family size and social class leaks directly into the apparent birth-order effect.
• Maternal age and birth spacing. Later children are born to older mothers by definition, and spacing between births varies with family circumstances. The Norwegian analyses found spacing associated with scores, which means it can act as a mediator, a confound, or both, depending on what caused the spacing.

Notice what this list does to the interpretation of any single number you might read. A study reporting a three point birth-order gap in a between-family sample and a study reporting a one point gap in a within-family sample are not contradicting each other. They are measuring different mixtures, and the larger number is the more contaminated one. When you meet a birth-order claim in the wild, the first question is not how big the effect is, it is which siblings were compared with which.

This is also why the topic connects to broader questions about score interpretation. Population averages shift with the composition of who is being measured, which is the same reason discussions of what counts as an average score have to specify the reference group before the number means anything.

10 The Proposed Explanations, Clearly Labeled as Hypotheses

Assume the small cognitive gradient is real, since the within-family evidence supports it. What causes it? Several accounts have been proposed, all of them plausible, none of them established. The honest state of play is that researchers have a well measured effect and competing stories about its origin. Comparing siblings holds the household constant but not their genetic differences, and the size of that share does not stay put: twin work on heritability reports estimates for general cognitive ability rising from about 41 percent in childhood to 55 percent in adolescence and 66 percent in young adulthood.

Two things are worth saying about this set. First, they are not mutually exclusive, and the true picture may combine them with additional factors nobody has isolated yet, such as differences in parental expectations or in how parents behave with a first child versus a fourth. Second, and more usefully, the mechanism question is close to unanswerable with observational family data, because everything that would distinguish the accounts moves together in real households. You cannot randomize birth order. The Norwegian sibling-death analysis is about as close to an experiment as this field will ever get, and it comes with the caveats already noted.

What you should not do is repeat any of these as established causes. The frequent claim that firstborns are smarter because they get more attention states a hypothesis as a finding. The measured gradient is solid. The explanation for it is open, and a page that pretends otherwise is doing storytelling rather than reporting.

11 The General Skill This Case Teaches

Birth order is the best available training exercise for a habit that pays off everywhere in psychology and health reporting: separating a population average from an individual prediction. The two get written with the same words and mean almost opposite things. The 0.16 correlation between height and measured IQ is another case this skill handles well: real, replicable, and useless for judging individuals.

A population average is a claim about what happens when you aggregate thousands of people. It can be small, real and scientifically interesting all at once, and its usefulness lies in what it reveals about a mechanism or a distribution, not in what it lets you say about a person. An individual prediction is a claim about a specific case, and it requires the effect to be large relative to the spread of the outcome. Almost nothing in behavioral science clears that bar. The birth-order effect misses it by an order of magnitude.

Three habits convert this from an abstraction into something usable:

• Ask for the effect in the units of the outcome. A correlation of .04 means nothing intuitively. One point on a 15 point standard deviation scale means something immediately. Any source that will not translate its effect into the scale you care about is hiding the size.
• Ask what fraction of the comparison goes the other way. Forty percent of younger siblings out-scoring their elder sibling is the sentence that makes the finding honest. Demand that number, or estimate it, before you let an average shape your expectations about a person.
• Ask which comparison produced it. Between-family, within-family, cross-sectional, longitudinal. In this literature, the design determined the answer more than the data did, and that is true of many other contested effects.

Run those three questions on any group difference you encounter and most of the alarming ones deflate immediately. It is also the right frame for reading your own results on any cognitive assessment, where the relevant question is never a single number in isolation but where your profile sits across separate domains, with intervals around each. That is why the difference between fluid and crystallized ability tells you more about a person than any composite, and why a serious report shows a shape rather than a point.

12 The Verdict, and What It Leaves You With

Birth order does not shape your personality. The evidence against that idea is now strong enough that the burden has flipped: anyone claiming a firstborn temperament or a lastborn rebelliousness needs to explain why two independent, well powered 2015 studies found nothing where the effect was supposed to be. Birth order does move measured cognitive scores by a small amount, roughly a point to a point and a half favoring the eldest, with a gentle decline down the sequence, and that result holds up in the strict within-family comparisons that killed most of the older literature.

The practical consequence for anyone reading this about their own family is close to zero, and that is the point rather than a disappointment. If you are the youngest of four, nothing in this research predicts that you will score below your siblings, and in something like four families out of ten you would out-score the eldest. If you are a firstborn, the research grants you a fraction of a measurement error, which is not a personality and not an identity. Whatever explains the differences you actually observe among your siblings, it is not the order you arrived in.

The more interesting question the research leaves open is the one it cannot answer for you: what your own cognitive profile actually looks like across domains. Verbal comprehension, fluid reasoning, visual spatial processing, working memory, processing speed and quantitative reasoning are related but distinct, and people are routinely uneven across them in ways no family fact predicts. That unevenness is the informative part, and it is measurable. If the topic interests you further, the questions of what counts as a good score and how scores relate to work outcomes both hinge on the same distinction between averages and individuals that this page has been making throughout.

13 Frequently Asked Questions

14 Sources Behind This Page

The relationship discussed here comes from published research, and the honest reading includes its limits. These are the primary sources behind the numbers on this page.

• Plomin, R. & Deary, I.J. (2015). Genetics and intelligence differences: five special findings. Molecular Psychiatry. Open access. The standard modern review of what twin and DNA evidence does and does not show.
• Nisbett, R.E. et al. (2012). Intelligence: new findings and theoretical developments. American Psychologist, 67(2). The broad APA review of what moves measured intelligence and what does not.
• Deary, I.J., Strand, S., Smith, P. & Fernandes, C. (2007). Intelligence and educational achievement. Intelligence, 35(1). A five year cohort of over 70,000 English schoolchildren.
• Schmidt, F.L. & Hunter, J.E. (1998). The validity and utility of selection methods in personnel psychology. Psychological Bulletin, 124(2). The meta-analysis behind general ability as the strongest single predictor of job performance.
• Strenze, T. (2007). Intelligence and socioeconomic success: a meta-analytic review of longitudinal research. Intelligence, 35(5).
• Kuncel, N.R., Hezlett, S.A. & Ones, D.S. (2004). Academic performance, career potential, creativity, and job performance: can one construct predict them all? Journal of Personality and Social Psychology, 86(1).
• Whalley, L.J. & Deary, I.J. (2001). Longitudinal cohort study of childhood IQ and survival up to age 76. BMJ, 322(7280). Open access.
• Batty, G.D. et al. (2007). IQ and mortality: a population based cohort study. International Journal of Epidemiology, 36(3), 538-553.
• Zagorsky, J.L. (2007). Do you have to be smart to be rich? The impact of IQ on wealth, income and financial distress. Intelligence, 35(5).
• Judge, T.A., Colbert, A.E. & Ilies, R. (2004). Intelligence and leadership: a quantitative review and test of theoretical propositions. Journal of Applied Psychology, 89(3).