I also had a brief question regarding demographics and political conservatism, as I didn't see ethicnity/race in the demographics and wondered if social equity may be a confounding or mediating variable here.
The short answer about why I did not include ethnicity/race here is that it's a little more complicated as there are lots of categories, and to display it in this format, I'd create multiple columns where I "dummy-code" the variable so, it would essentially show the correlation between white, non-Hispanic vs. nonwhite and each outcome, black, non-Hispanic vs. non-black and each outcome, Asian, non-Hispanic vs. non-Asian and each outcome, Hispanic vs. non-Hispanic and each outcome, etc. The other concern is that in doing that, the error term is many times larger for some of the nonwhite groups, and I don't want people to overinterpret nonsignificant differences.
We also have household income that would speak to economic equity (correlated with, but distinct from social equity).
I'd be happy to run additional quick analyses and share, if it'd be helpful.
Great work, Matt! I do have a question about measurement: what did you use to measure conservatism?
I ask because I did some back-of-envelope calculations on the correlations between the demographic correlations in the last table presented. They all made sense, except for r(gender, conservatism):
r of r's
gender, age -0.12
gender, edu -0.33
gender, cons 0.56
edu, age 0.29
edu, cons -0.45
cons, age 0.49
There's a million different reasons a high r(gender, cons) could exist, but I'm always interested in looking at the instrument before I look at the band, so to speak!
The exact wording of that question is: "On a scale from 0 to 100 where 0 is the most liberal and 100 is the most conservative, what number would you give to yourself?"
That said, the actual correlations are far different from your back of the envelope calculations. Gender (where 1 = women) and conservatism (where 0 = most liberal and 100 = most conservative) correlate at -.059. I'll fill in the rest of the actual correlations you mentioned below:
This was wonderfully interesting.
I also had a brief question regarding demographics and political conservatism, as I didn't see ethicnity/race in the demographics and wondered if social equity may be a confounding or mediating variable here.
Thank you! I'm glad you liked it.
The short answer about why I did not include ethnicity/race here is that it's a little more complicated as there are lots of categories, and to display it in this format, I'd create multiple columns where I "dummy-code" the variable so, it would essentially show the correlation between white, non-Hispanic vs. nonwhite and each outcome, black, non-Hispanic vs. non-black and each outcome, Asian, non-Hispanic vs. non-Asian and each outcome, Hispanic vs. non-Hispanic and each outcome, etc. The other concern is that in doing that, the error term is many times larger for some of the nonwhite groups, and I don't want people to overinterpret nonsignificant differences.
We also have household income that would speak to economic equity (correlated with, but distinct from social equity).
I'd be happy to run additional quick analyses and share, if it'd be helpful.
Great work, Matt! I do have a question about measurement: what did you use to measure conservatism?
I ask because I did some back-of-envelope calculations on the correlations between the demographic correlations in the last table presented. They all made sense, except for r(gender, conservatism):
r of r's
gender, age -0.12
gender, edu -0.33
gender, cons 0.56
edu, age 0.29
edu, cons -0.45
cons, age 0.49
There's a million different reasons a high r(gender, cons) could exist, but I'm always interested in looking at the instrument before I look at the band, so to speak!
Thanks, Ian!
The exact wording of that question is: "On a scale from 0 to 100 where 0 is the most liberal and 100 is the most conservative, what number would you give to yourself?"
That said, the actual correlations are far different from your back of the envelope calculations. Gender (where 1 = women) and conservatism (where 0 = most liberal and 100 = most conservative) correlate at -.059. I'll fill in the rest of the actual correlations you mentioned below:
gender, age ~ +.13
gender, edu ~ -.04
gender, cons ~ -.059
edu, age ~ +.022
edu, cons ~ -.14
cons, age ~ +.20
You might also find a recent post I made on political attitudes across social media of interest. Check it out here: https://psychoftech.substack.com/p/political-attitudes-on-social-media
Backs of envelopes are just that; no substitute for the glory of rich point-biserial data 😄
Thanks for such a thorough answer, and for helping feed a little curious joy in me today!
(And I'll check out that other post, too)
Of course! These data are incredibly rich, and there is so much to be learned from them.