How much did selection bias impact my cheating rates?
Furiously googling other studies
In my relationships survey, I asked people about cheating. The results were pretty pessimistic - long term relationships, but especially men, and especially sexually, were not looking so good.
Lots of things I look at are things I strongly suspect aren’t much impacted by sampling bias, but this one is a bit riskier. Although only ~18% of my main twitter audience reports being from primarily a porn source, I still am quite vocal about being a sex worker, and presumably people who follow me are somewhat selected towards… I don’t know, the kind of relationships that either are chill with that, or are doing it secretly.
I looked at a few things besides cheating, but I want to pick one metric to focus on here, and cheating seems the most concrete/easy to check other sources about. And if I do find that cheating rates are similar, this would be a point in favor that the rest of my results probably aren’t that skewed either.
I’m also gonna focus strictly on males, since that’s who I mainly wrote about in my last post, and who seem to be more vulnerable to the effects of selection bias (me being a thirst-trap female). You’re welcome go go through the same data and check to see how closely it matches for females, though!
So let’s look at the existing research to see how far off I am! I’m going to write this essay as I go, and commit to reporting on the results of each study before I see the results of the study, so that I’m not accidentally doing a bias where I’m only pointing out things that agree with me.
So to summarize my data, I found that cheating rises over time, and for men in relationships 22+ years, the rate is over 40%.
My first stop: Wikipedia
When I check research on a thing, I usually use the list of sources that Wikipedia cites as the primary go-to list. So let’s see.
States that GSS in 1991 surveyed 1212, a national probability sample, anonymous responses, of people who are or had been married. 11% of women and 21% of men reported engaging in sex with someone other than their spouse, while married. It notes that religious people cheat no more than 25%, and non-religious at 35%, which is confusing to me because I thought they said the original cheating rates were 11% and 21%?
It includes this graph, which… I guess is from the GSS? I’m not sure if I’m missing something in the article, but it doesn’t seem to offer another source. I’ve overlapped my own data on top of this
I don’t really know what to make of this. (I didn’t bother to pry confidence intervals for this, but the last two bins for females are quite low sample size). My guess is this looks roughly consistent with generational aging? Theirs was done 30 years ago; maybe they’re observing a cheating-wave that was peaking at 40-49, and I’m observing a cheating wave that’s peaking at 60+? This doesn’t seem obviously consistent with “men who follow Aella, tend to cheat more”, unless for some reason this is true but only for men aged 60+? It might just be noise - my sample size for older bins wasn’t huge, down to ~220 respondents in the male 60+ category.
They seemed to use the term ‘infidelity’, and I used ‘cheating’; not sure if this impacts anything. I did not specify sex; my guess is that reported rates of cheating will be a bit higher in my data than in others due to people interpreting some non-sex activities as cheating.
Source #2 is the New York Times, which I don’t want to pay for, but
*Wikipedia reports the link claims that 12% of men and 7% of women have had an extramarital relationship, again from the GSS.
I’m not sure which GSS report they’re citing, but want to also note the average % of people in relationships in my data who reported cheating was 14.6%; the real high numbers comes in at specifying long-term relationships.
The NYT then apparently cites a University of Washington study: “In that study which involved 19,065 people during a 15-year period, rates of infidelity among men were found to have risen from 20% to 28%, and rates for women ranging from 5% to 15%.”
Source #3 is the Extramarital Sex and HIV Risk Behavior among US Adults, Results from the National AIDS Behavioral Survey, in 1994.
Looks like they interviewed 13,786 people via calling random numbers on the telephone.
And… oh, they only asked about the past year, which was around 2.2%. I don’t think I’d be able to do a good estimate about what this means for “have you ever cheated in your relationship” so I’m going to stop here.
Wikipedia says:
Studies suggest around 30–40% of unmarried relationships and 18–20% of marriages see at least one incident of sexual infidelity.
Rates of infidelity among women are thought to increase with age. In one study, rates were higher in more recent marriages, compared with previous generations.
Source #4: Infidelity In Committed Relationships II: A Substantive Review (2005)
Yay, a review! Maybe I should have just looked at this first.
It says 1994 GSS says 22% of men and 12% of women report having extramarital sex.
Says 1991-1996 GSS says 13% reported extramarital sex.
1981 survey of women says 10% of women in relationships reported a secondary sex partner.
Laumann et al (1994) says 25% of married men and 15% of married women have had extramarital sex.
It goes on to cite some surveys that measure cheating rates in the last year or so, so I’m skipping
“In general, based on the above data, we can conclude that over the course of married, heterosexual relationships in the United States, EM sex occurs in less than 25% of committed relationships, and more men than women appear to be engaging in infidelity (Laumann et al., 1994; Wiederman, 1997).”
“It is important to note that prevalence data vary greatly in studies with broader definitions and populations of interest.”
Source #5: Understanding the relationship between gender and extradyadic relations: The mediating role of sensation seeking on intentions to engage in sexual infidelity (2011)
This says nationally representative data from the National Health and Social Life Survey analyzed by Waite and Gallagher (2000) found that “4% of married men, 16% of cohabitating men, and 37% of men in dating relationships had sex with someone other than their primary partner in the previous year. In comparison, 1% of married women, 8% of cohabitating women, and 17% of women in dating relationships reported engaging in extradyadic sex.”
Ooh, what does my data say?
Sorry it’s not pretty, just wanted to glance. Looks like not a huge difference - maybe a 5% gap? This is a far cry from the gap of like… 4% to 37% from the Waite and Gallagher survey. Their survey does kinda feel weird to me - it feels very counterintuitive that getting married cuts cheating rates by 9x? I would guess marriage probably correlates with greater cheating rates, because cheating increases by length of relationship, and the longer a relationship, the more likely people are to get married.
Let’s check their study directly. I’m actually not sure this is it - I can’t find any mention of cheating rates in there. The original paper seems to cite their book, and when I look at the book it seems like it’s by a sociologist and a journalist and purports to “dramatically contradict anti-marriage myths”, so ???
Moving away from wikipedia; let’s google!
Source #6: Institute for Family Studies (based on the GSS, 2010-2016)
They provide this figure, which I like a bit more because it’s much more recent than the above figure from 30 years ago.
Let’s overlay mine on top (I’m excluding the last female bin now because n=28 and that’s tiny)
Mine is definitely higher - around a 10% gap, particularly starting age years 40+. This is pretty significant.
I’m not totally sure how to interpret this gap. They specify extramarital sex (much like many of the other studies), and I didn’t specify sex - I just said cheating. But still, I’m not sure that would account for a 10% gap! GSS data seems to be pretty good, so I defer to them on this.
Source #7: Once a Cheater, Always a Cheater? Serial Infidelity Across Subsequent Relationships
They had a baby sample size at 484, and consisted of people they pulled from a bigger study, which was itself pulled from a nationwide sample of young people in unmarried relationships. People were recruited with telephone sampling, and completed surveys by mail, and were paid $40. The sample is small but the randomization seems pretty decent. They found “Overall rates of infidelity in this sample were toward the high end of the range of previous estimates, with 44% of participants reporting engaging in infidelity themselves during the relationships captured by this study”
(They do point to this other study about married vs. unmarried by Treas & Giesen, 2000, which, glancing at, I think says there’s a 4% gap between married and unmarried).
Okay so what does this mean
It looks like, in general, my numbers are on the high end - male cheating ranges seem to range from 20-25%, whereas mine range more like 30-35%.
Most concerning is the gap in GSS data, which seems to be roughly 10%. This might be explained by a few things, but I’m not convinced.
As in:
Most studies seem to say married people cheat less than unmarried people. Maybe I’m disproportionately surveying unmarried people, thus getting unusually high numbers? But my data seems to indicate unmarried people cheat at only roughly 3-5% more than married, adjusted for length of relationship, which isn’t a huge difference and I don’t think can fully account for my higher numbers.
Some of the studies said that nonreligious people cheat more than religious people. Maybe I’m disproportionately surveying nonreligious people? But I asked respondents if they followed a religion; people who said no cheated at 14%; people who said loosely/culturally cheated at 16%, people who said yes cheated at 15%, and people who said devoutly/heavily cheated at 10%. This indicates a gap of maybe roughly 5%; again, mostly small.
I didn’t specify cheating as sex; it’s possible this resulted in more people selecting ‘yes’ to cheating, assuming they were also including some non-sex things. I don’t have any good evidence for the degree this is true, but I very much doubt it’s 10%. Maybe 3-5%?
It’s possible people felt my survey was more anonymous (an informal internet survey) than official things administered by scientists and formal interviews, and thus were more likely to admit to cheating? I don’t have any good evidence either way for the degree this is true, here.
I think it’s possible that the above four things combined - I’m looking at unmarried, nonreligious people with cheating more loosely defined, with possibly greater feelings of anonymity - is resulting in my higher rate of self-reported cheaters.
But I’m not sure. We also have the popular hypothesis:
People who follow Aella are disproportionately selected to be unhappy in their relationships (because people in happy relationships are less likely to follow sex workers online), or are disproportionately selected to be unhappy with monogamy (because Aella often speaks strongly about polyamory and sexual freedom).
I do think I would be surprised if this didn’t impact my data at all. The question is, to what degree? And, eyeballing all the hypotheses together, if I had to make a guess, I’d say the Aella-specific selection bias probably accounts for 2-4% of the results. I’d guess I’m probably encountering greater selection bias by measuring a more modern/progressive subpopulation.
Still though - let’s adjust! I find older men cheat at around 30-35%. When binning for relationship length, long relationships jump to around 41% cheating - around a 5-10% increase.
This might mean that, if in the normal population (including religious, more married people), extramarital sex in older men is actually at around 25%, this would probably suggest that actual extramarital sex in long-term relationships is at around 30-35%.
In general, I think my data falls mostly in line with existing research on populations that match the population I’m looking at.
But actually though, if ‘people who follow people like me online are more likely to cheat’ is a real phenomenon, this in itself would be fascinating to study! I’d like to pair up with someone who has a very different demographic so that we can measure this more specifically. What kinds of personality figures online are more likely to have cheating followers? What predicts this and why?
Note that in general, though, I’ve found that things mostly don’t correlate with each other, or if they do it’s usually through obvious confounds. In my kink survey, surprisingly little correlated that I thought would. With cheating, neither income nor amount of kids seemed to predict cheating rates, both things I might intuitively expect to impact it to a similar degree that “following me online” would impact it.
If I had to bet money on the amount selection bias (beyond the typical unmarried/nonreligious population sampling) impacts my cheating rates among men in long-term relationships, I would make a squinty intuitive guess at probably 3%.
Let me know if you think my estimation here is accurate! I don’t have a ton of experience at deeply predicting and estimating things based off existing studies, let me know if I missed anything in my reasoning!
Two major confounding factors jump to mind here:
The first is that questions of how and how often people are caught cheating, or disclose their cheating, and how those conversations play out is a really big deal.
Consider this, if every case of cheating was immediately discovered, and every case resulted in the immediate dissolution of the relationship. The cheating rate for ongoing relationships in a survey like this would be zero. So, in that sense, what you are actually measuring is not only cheating, but also the successful hiding of cheating, and the successful working through and past of discovered cheating. So when you see a spike in cheating in 20+ year relationships, one theory is that people are cheating more. But an equally valid theory is that people are becoming more likely to continue in a relationship where cheating has occurred.
(EDIT: This is literally survivorship bias. Asking people currently in relationships whether they have cheated and then drawing conclusions about how often cheating occurs in relationships is roughly analogous to asking pilots returning from bombing runs if they took a direct hit to the engine and then concluding that anti-aircraft guns pretty much never hit engines.)
The other confounding factor has to do with societal shift and technology. The accessibility and risk of cheating has changed dramatically in the age of tinder/grindr/etc. And this accessibility has disproportionately skewed younger. I think this becomes especially relevant when comparing against older data.
Some evolutionary scientists, like David Buss, endorse the Mate-switching-hypothesis of cheating, which states that women cheat to find a new partner to replace their current partner. If this is true, it would open up an interesting question: Do women cheat less or is their cheating just more likely to end the relationship (which would mean it doesn't show up in as many statistics)?