You're Weird: Explorer
I had a lot of great questions and comments on last week's widget showing how unrepresentative we all are of the broader population.
E.g. several people asked me what the most common combination was, and exactly how common it is. And friend of the blog L. asked if the results were largely driven by age, and at the time I didn't have an answer to that.
So today I made an Explorer tool that lets you poke at the data more easily and see how rare various combos of answers are. E.g. If you're less-than-certain in the existence of God, think same-sex relationships are totally ok, and have a bachelor's degree (or higher), you're in a 1-in-6 minority among Americans as a whole. But I suspect you're much more common than that among this blog's readership, and that you're likely to be in an overwhelming majority among your own friend-group. And I truly think national politics gets distorted in part by people thinking their own views and traits are far more common than they really are, and I hope this tool can help us collectively train our intuitions on just how weird we actually are. Happy exploring, let me know what you discover!
Build any American profile — see how rare it is.
Pick a Yes/No/Any for each of the seven questions below. The widget shows what percentage of Americans match your filter, using the joint distribution from the General Social Survey. Browse common patterns, rare patterns, and the strongest correlations in the data.
Methodology & sources
Data comes from the General Social Survey (GSS) cumulative file, 1972–2024, Release 3. We use respondents from the four most recent waves (2018, 2021, 2022, 2024), restricted to those who answered all seven questions. Sample sizes range from … for the youngest band (18–29) to … for all adults combined. All percentages use NORC's recommended post-stratified weight wtssps.
For each population, the widget stores a 128-cell joint distribution (one cell for every combination of yes/no answers). A query is just a sum over the cells consistent with the user's filter. Because the joint distribution comes from real respondents — not from multiplying marginal probabilities — correlations between answers are preserved.
The smaller the population, the noisier individual cells become. For the 18–29 band (N≈615) the median cell has only a handful of respondents, so very rare cells should be treated as rough estimates. The widget shows a small-sample warning when a query is backed by fewer than 25 respondents.
[This tool was built by Claude, largely at her own direction. I have tried to sensecheck the results but can't fully promise they're valid. If you see issues please do let me know].