Visualize the landscape of sexual interests
Based on a dataset of 1.3M.
The Big Kink Survey is a long-running online survey. It covers childhood, sexuality, and almost every fetish. Over 1.3 million people have taken it; 590 of its sex-related items are charted here.
Free, no account needed. Adults only (18+).
But wait, is this representative? What about sampling bias?
The survey is disproportionately popular among young, western, liberal women, and also kinky people. This mostly affects questions like 'how common do we think this fetish is?'
To help this problem, I did two things - demographic weighting, and kink-adjacent balancing.
It goes like this - we simply deflate the younger liberal women in the sample, and inflate older conservative men.
But we still have another issue. Kinks are likely not evenly distributed. For example: it's pretty normal for a young liberal women to take this survey - surveys in general tend to be popular among young liberal women. But an older conservative man who takes this survey is probably abnormal among older conservative men, because those guys are usually fixing cars or something, not taking sex surveys online. Thus you might see inflated fetish rates in the data for older men, compared to their actual rates in the population.
We can't perfectly fix this. It's very hard to get accurate fetish data in surveys in general - even proper randomized surveys are affected by nonresponse rates and dishonesty. You're much less likely to admit a taboo fetish if you feel less anonymous!
But all is not lost. Hence the second goal - kink-adjacent balancing, which hits demographics differently. It works like this.
We don't have data about true fetish rates in the population, but we do have data about stuff that fetishes correlate with - primarily being non-cis, being non-straight, having unusually high or low BMI, and mental illness, plus childhood and adult sexual-assault history, religious upbringing, and lifetime partner count. This shows up in the Big Kink Survey - we have high fetish rates, and we also have very high non-straight, non-cis, mental illness, etc. rates. And best of all, we have good estimates of what those rates are in the general population, from other rigorous research!
So we weight for those factors. (The calibrated numbers cover ages 14–50 in the US, Canada, UK and Western Europe, the populations with reliable benchmarks to weight against.) And this, predictably, brings down total fetish rates by quite a lot - mostly around 20%, but ranging from 15-50%, depending on the subtype and the rarity, and it affects the prevalence differently in different demographics.
This is the hardweight dataset, an effective sample of ~42k, that is aggressively balanced with the goal of getting more accurate fetish prevalence rates. I use this dataset for the Fetish Map. The tree is about how kinks cluster together rather than how common they are, so it uses the unweighted responses.
This isn't perfect, but it's currently the best I've figured out to do with the data available to me.