I didn’t do this myself because I didn’t trust my statistical ability enough, and I forgot to mention it on the original post, but...
Can someone check for birth order effects? Whether Less Wrongers are more likely to be first-borns than average? Preferably someone who’s read Judith Rich Harris’ critique of why most birth order effect analyses are hopelessly wrong? Or Gwern? I would trust Gwern on this.
I don’t know Harris’s critique, but here are some numbers.
Out of survey respondents who reported that they have 1 sibling (n=453), 76% said that they were the oldest (i.e., 0 older siblings). By chance, you’d expect 50% to be oldest.
Of those with 2 siblings, 50% are the oldest (vs. 33% expected by chance), n=240.
Of those with 3 siblings, 45% are the oldest (vs. 25% expected by chance), n=120.
Of those with 4 or more siblings, 50% are the oldest (vs. under 20% expected by chance), n=58.
Of those with 0 siblings, 100% are the oldest (vs. 100% expected by chance), n=163.
Overall, 69% of those who answered the “number of older siblings” question are the oldest.
Those look like big effects, unlikely to be explained by whatever artifacts Harris has found.
There are a handful of people who left the number of older siblings blank but did report a total number of siblings), or who reported a non-integer number of siblings (half-siblings), but they are too few to make much difference in the numbers.
This doesn’t seem to vary by degree of involvement in LW; overall 71% of those in the top third of LW exposure (based on sequence-reading, karma, etc.) are the oldest. Here is a little table with the breakdown for them; it shows the percent of people who are the oldest, by number of siblings, for all respondents vs. the highest third in LW exposure.
n all high-LW 0 100 100 1 76 80 2 50 45 3 45 51 4+ 50 62
There seems to be a pretty big potential confounder: age. Many respondents’ younger siblings are too young to be contributing to this site, while no one’s older siblings are too old (unless they’re dead, but since ~98% of the community is under age 60 that’s not a significant concern).
You’re saying that if we randomly picked 22-31 year-olds, a disproportionate member would be eldest children? For that to work, there’d have to be more eldest children in that age-range than youngest. Given the increase in population, that is certainly plausible. You would expect more younger families than older families, which means that within an age range there would be a disproportionate number of older siblings (unless it’s so young that not all of the younger siblings have been born yet) but it doesn’t seem like it would be nearly that significant.
Many respondents’ younger siblings are too young to be contributing to this site, while no one’s older siblings are too old
The fact that most of the respondents are eldest children is a confounder for this.
(unless they’re dead, but since ~98% of the community is under age 60 that’s not a significant concern).
In that case, wouldn’t people over 60 also be too old?
I didn’t do this myself because I didn’t trust my statistical ability enough, and I forgot to mention it on the original post, but...
Can someone check for birth order effects? Whether Less Wrongers are more likely to be first-borns than average? Preferably someone who’s read Judith Rich Harris’ critique of why most birth order effect analyses are hopelessly wrong? Or Gwern? I would trust Gwern on this.
I don’t know Harris’s critique, but here are some numbers.
Out of survey respondents who reported that they have 1 sibling (n=453), 76% said that they were the oldest (i.e., 0 older siblings). By chance, you’d expect 50% to be oldest.
Of those with 2 siblings, 50% are the oldest (vs. 33% expected by chance), n=240.
Of those with 3 siblings, 45% are the oldest (vs. 25% expected by chance), n=120.
Of those with 4 or more siblings, 50% are the oldest (vs. under 20% expected by chance), n=58.
Of those with 0 siblings, 100% are the oldest (vs. 100% expected by chance), n=163.
Overall, 69% of those who answered the “number of older siblings” question are the oldest.
Those look like big effects, unlikely to be explained by whatever artifacts Harris has found.
There are a handful of people who left the number of older siblings blank but did report a total number of siblings), or who reported a non-integer number of siblings (half-siblings), but they are too few to make much difference in the numbers.
This doesn’t seem to vary by degree of involvement in LW; overall 71% of those in the top third of LW exposure (based on sequence-reading, karma, etc.) are the oldest. Here is a little table with the breakdown for them; it shows the percent of people who are the oldest, by number of siblings, for all respondents vs. the highest third in LW exposure.
n all high-LW
0 100 100
1 76 80
2 50 45
3 45 51
4+ 50 62
That 62% is 8⁄13, so not very meaningful.
There seems to be a pretty big potential confounder: age. Many respondents’ younger siblings are too young to be contributing to this site, while no one’s older siblings are too old (unless they’re dead, but since ~98% of the community is under age 60 that’s not a significant concern).
You’re saying that if we randomly picked 22-31 year-olds, a disproportionate member would be eldest children? For that to work, there’d have to be more eldest children in that age-range than youngest. Given the increase in population, that is certainly plausible. You would expect more younger families than older families, which means that within an age range there would be a disproportionate number of older siblings (unless it’s so young that not all of the younger siblings have been born yet) but it doesn’t seem like it would be nearly that significant.
The fact that most of the respondents are eldest children is a confounder for this.
In that case, wouldn’t people over 60 also be too old?
Can somebody redo the analysis by controlling for age?
I don’t know anything about birth order effects, sorry.