attraction.lab

Age Gaps in Relationships: What the Research Predicts, and Where the Divorce Numbers Came From

The 18%, 39% and 95% divorce figures were withdrawn in 2014 by the man who worked them out. The best US study found the gap itself predicted nothing.

Research-backed writing on attraction, dating and relationships. Drafted by AI against a named bibliography, published only when every claim links to its source.
34 min read
Editorial illustration for: Age Gaps in Relationships: What the Research Predicts, and Where the Divorce Numbers Came From

Before the article. An age gap is not abuse, but a large one often comes with differences in money, housing and social confidence, and those are what controlling behaviour uses. Having to account for ordinary hours, your phone being checked, being discouraged from seeing friends and family, decisions about money or work being made for you, being told nobody your own age would want you — those are warning signs, and the second section of this page lists them in full. Find A Helpline, in the sources at the bottom, lists free and confidential services by country. In the United States the National Domestic Violence Hotline is 1-800-799-7233, or text START to 88788, 24 hours a day. Everything cited on this page is research on adults.

With that said. Search whether an age gap matters and you get three numbers: a five-year gap means 18% more divorce risk, ten years 39%, twenty years 95%. They are on law-firm blogs, in women's magazines, in a 2014 New York Post piece and in statistics round-ups refreshed this year, nearly always credited to a study from Emory University.

Those figures were withdrawn in 2014 by the person who calculated them. He was not one of the study's authors — he was a graduate student who had read the paper, which is about wedding spending, and charted what he found interesting. The researchers pointed out that the relative percentages could not be derived from their data, and he published a correction saying so. It has been sitting on that page for twelve years while the numbers kept travelling.

So this page starts over. What the largest American study found when it asked who initiates a divorce. What thirteen years of annual satisfaction reports from Australian couples show about the trajectory. Who actually ends up in a differently aged couple, which turns out to be the result that governs how to read all the others. How common a big gap is, in census data. Where half your age plus seven came from, which is checkable in a minute and is not what anybody thinks. The one outcome where the gap shows up large and clear, in Danish population registers. And what the disapproval does, measured separately from the gap.

Twelve sections, no ideal number at the end of them — because no study has established one, and the paper that tested for a threshold directly did not find it. Our editorial policy explains why every empirical claim here carries a linked source.

#1

Where the divorce percentages came from

Ask the internet whether an age gap matters and you get a number. A five-year gap makes you 18% more likely to divorce. Ten years, 39%. Twenty years, 95%. The figures are everywhere — law-firm blogs, women's magazines, a 2014 New York Post piece, statistics aggregators dated this year — and they are nearly always credited to “a study from Emory University.”

There is a real study. Andrew Francis-Tan and Hugo Mialon surveyed more than 3,000 ever-married Americans and published the results in Economic Inquiry in 2015 under the title “A Diamond Is Forever” and Other Fairy Tales. Its subject is wedding spending. Its finding is that marriage duration is inversely associated with what couples spent on the engagement ring and the ceremony. Age difference was one of many characteristics in the model, not the paper's topic.

The percentages came from somewhere else. In October and November 2014, Randal Olson — then a computer science graduate student, now an AI researcher — read that paper and published two blog posts pulling out the factors he found most striking, with charts. The second post, dated 6 November 2014, is the one with the age-gap chart. Five days later the New York Post ran the numbers, Business Standard picked them up from the Post, and they have been circulating ever since.

Now read what that post says today. Underneath the age-gap section, in the author's own words:

> Note: A previous version of this article showed a chart giving specific relative percent likelihoods of divorce occurring based on number of years married. The original authors of the study have pointed out that although there is a significant correlation between wider age gaps and increased divorce, it is not possible to determine the relative percent likelihood from their study. That is left to future research.

So the figures that the first page of search results still presents as a finding were withdrawn by the person who calculated them, at the request of the researchers whose data he used, roughly twelve years ago. The correction has been sitting in public on that page the whole time. The direction survived — the authors agreed a correlation was there — but the magnitudes, which are the only part anybody quotes, did not.

That pattern is the reason this site exists, and it is not unusual. The three-day rule turns out to come from a 1996 comedy in which the characters invent the number at a table, and the 3-6-9 rule has no origin at all. Our editorial policy explains why every claim here carries a source you can open.

#2

Before the rest of this page: when the gap is not the thing that is wrong

This article is about what the research on age differences between adult partners has measured. There is a different situation that the same search sometimes comes from, and it needs answering first because it is more urgent and the answer has nothing to do with statistics.

An age gap is not abuse. But a large gap can sit alongside a difference in money, housing, work history and social confidence, and those differences are the raw material that controlling behaviour uses. The gap is not the problem; what someone does with the asymmetry can be.

The US National Domestic Violence Hotline publishes a list of warning signs, and it is specific rather than atmospheric: having to account for where you were and who you spoke to; your phone or location being checked; being discouraged or prevented from seeing friends and family; decisions about money, work or transport being made for you; being told your reaction is the real problem; threats, intimidation, or any physical harm. The Hotline notes that even one or two of these is a warning sign.

A few things specific to this subject belong in the same list. Being told you are too young to understand your own objection. Being told that nobody else would want you, or that nobody your own age would. Having your friendships described as immature. Being the only one of the two of you whose independence is up for discussion.

If any of that is familiar, that is the thing to act on, before anything else on this page. Find A Helpline, in the sources at the foot of this page, lists free and confidential services by country. In the United States the National Domestic Violence Hotline is 1-800-799-7233, or text START to 88788, around the clock.

Two boundaries on what follows. Every study cited here is about adults, and most of them about married adults; nothing on this page is about relationships involving a minor, which is a legal question and not a question about compatibility. And where a gap comes with the conduct above, the word that gets a service or a court to act is coercive control, not incompatibility — the distinction is set out in the psychology of jealousy and in what the term “trauma bonding” actually rests on.

#3

The best American study asked who leaves, and found the gap predicted nothing

If age heterogamy raises the risk of divorce, the obvious place to look is a long-running national survey that followed real marriages and asked, afterwards, which partner wanted out. That study exists. On 30 October 2026 we opened the pages ranking for does age gap matter in a relationship and checked what each of them cites: two of them do cite primary research — one names the Danish mortality study and the global survey discussed further down, another names the wedding-spending paper and the commitment work — and two cite nothing at all. Not one of them mentions this study, or the word heterogamy, or the retraction above.

Paula England, Paul Allison and Liana Sayer published it in the Journal of Marriage and Family in 2016 under a title that gives away the question: Is Your Spouse More Likely to Divorce You if You Are the Older Partner? They used the National Survey of Families and Households — a national probability sample of 13,007 adults first interviewed in 1987–88, re-interviewed in 1992–94 and again in 2001–02, with questionnaires from both spouses. Their analysis sample was 3,622 couples, 413,650 couple-months and 747 separations, modelled with a latent class event-history design that handles the fact that ex-spouses sometimes disagree about who left.

The headline sentence of their abstract is this: support was not found for the hypothesis that age heterogamy itself increases the odds of divorce.

What they found instead is more interesting than the thing it replaced. Taking couples within three years of each other as the reference:

  • If the wife was more than three years older, the husband's odds of initiating a separation were 87% higher (OR 1.87). The wife's odds were 23% lower, and that effect was not significant.
  • If the wife was more than three years younger, the husband's odds of initiating were 50% lower (OR 0.50) — and the wife's odds were 38% higher (OR 1.38).

A continuous version fit the data better by both AIC and BIC, and a squared term was not significant, meaning the relationship is a straight line rather than a penalty that kicks in at some threshold. In that model, for each additional year the husband's age exceeded the wife's, his odds of leaving fell by 8% and her odds of leaving rose by 5%. The difference between 8% and 5% was not statistically significant.

Read the two directions together and the pattern is not about the size of the gap at all. The older of the two partners tends to be the one who gets left, whichever sex they are. A man was only about half as likely to leave the heterogamous marriage where his wife was younger than the homogamous one where they were the same age — which is the opposite of a preference for similarity.

And this is where the authors explain why so much earlier research looked like it found a risk. Because relative age pushes husbands and wives in opposite directions, the two effects partly cancel in any model that lumps all divorces together without asking who initiated. They write that much past research on relative age and divorce has therefore been misleading. A Dutch study found the same pattern — Kalmijn in the European Sociological Review in 2006, which England and colleagues cite as Kalmijn and Poortman.

Their own caveats are in the paper and belong here. With 747 divorces and a lopsided split between male- and female-initiated ones, the power to detect a sex difference was limited. They could not control absolute age and relative age at once, so it is possible that what moves the odds is simply having an ageing spouse rather than a gap. And they flag the real puzzle: why would a gap that was acceptable on the wedding day become the basis of a decision a decade later? The same difficulty of predicting who stays together turns up in the research on compatibility questions.

#4

What happens to satisfaction after the first few years

Divorce is a blunt outcome. The finer-grained question — does a gap feel different over time — has a good answer, and it comes from annual reports rather than a single snapshot.

Wang-Sheng Lee and Terra McKinnish analysed the 2001–2013 waves of the Household, Income and Labour Dynamics in Australia survey, which asks both the husband and the wife every year how satisfied they are with their relationship. Their sample was 3,374 couples and 18,987 couple-years, restricted to couples in their first marriage aged 20 to 55. Because the same couples report year after year, they could use individual fixed effects, which means the result is identified by change within a marriage rather than by comparing newlyweds with people married longer.

Four findings, in their order:

Men tend to be more satisfied with younger wives and less satisfied with older wives. No surprise, and it matches the stereotype.

Women likewise tend to be more satisfied with younger husbands and less satisfied with older ones. That is the half the stereotype gets wrong, and the authors flag it as the interesting one. The folk story says women want an older partner; in the satisfaction data, both sexes report being happier with a younger one.

For both sexes, satisfaction declines faster over the duration of the marriage in differently-aged couples than in similarly-aged ones — and the decline is steepest for men with much younger wives, precisely the group that started highest.

Those relative declines erase the initial advantage within six to ten years. The head start is real and it is temporary.

The mechanism they test is the part worth carrying away. HILDA asks whether a household has had a “major worsening in financial situation” in the past year. 26.2% of their couples reported at least one such shock, 11.5% reported two or more. Differently-aged couples lost more satisfaction when a shock arrived. Their reading is that these couples are less resilient to a bad year, possibly because they are less alike in what they want to spend money on — not that the gap wears people down on its own.

That reframes the question from is the gap bad to what does the gap make harder, which is the shape of answer that also turns up in what distance does to attraction and in the habits of couples who stay together. If you want a reading of where your own relationship currently sits, our free relationship satisfaction test scores in your browser and names the instrument it adapts.

#5

Who actually ends up in a differently aged couple

Here is the confound that decides how to read everything above, and it is the single most important paragraph on this page.

The folk model of an age-gap couple is a successful older man and a younger partner his success attracted. Every economic theory of why these marriages form assumes some version of it: the older spouse brings resources, the younger one brings youth, and the trade is what makes the match.

Hani Mansour and Terra McKinnish went to look. Their paper is in the Review of Economics and Statistics and the working-paper abstract states the result bluntly: in direct contrast to conventional wisdom and most economic models, men and women married to differently aged spouses are negatively selected. Using the 1960 through 2000 US censuses, male earnings fall as the within-couple age difference grows, and that holds whether the man is the older or the younger one. Female earnings rise with the age difference, but because those women work more hours, not because they earn more per hour. They then tested selection three ways — average pay in a person's occupation from census data, cognitive skill assessments from the NLSY79, and rated physical appearance from Add Health — and the point estimates ran negative on all three, though they note themselves that statistical significance varied across outcome and sample.

What that means for the divorce question is straightforward and rarely said. If the people who form differently aged couples differ systematically from those who do not — in earnings, in occupation, in whatever those measures are picking up — then any correlation between the size of a gap and a bad outcome is partly or wholly a fact about who ends up in those couples, not about what the gap does to a relationship. Controlling for education and income, as the better studies do, narrows the problem without closing it.

This is the same interpretive trap as the one in why smart people keep dating the wrong people: a pattern that looks like a cause is often a description of who arrives at the pattern. It is also why the honest version of this page cannot tell you that closing an age gap would change anything. Nobody has run that experiment, and nobody can.

One caveat on the source. The published version is paywalled; what is reported here comes from the authors' own abstract of the working paper, and the caveat about significance is theirs.

#6

How common a big gap actually is

Most of the anxiety in this subject is about whether a gap is normal. That has a clean answer from official data, and the answer has been moving in one direction for 140 years.

The Pew Research Center analysed decennial census and American Community Survey microdata — the 2022 ACS file alone contains about 680,000 opposite-sex marriages where the spouses live together — and published the result in August 2024. Husbands and wives in the United States were on average 2.2 years apart in age in 2022, down from 2.4 years in 2000 and 4.9 years in 1880. In 2022:

  • 51% of opposite-sex marriages had spouses two years apart or less, up from 46% in 2000.
  • 40% had a husband three or more years older, down from 43%.
  • 10% had a wife three or more years older, down marginally from a peak of 11% in 2000.

Two of their breakdowns matter more than the averages. Husbands with at least a bachelor's degree are more likely to be in a same-age marriage (55%, against 48% of husbands with some college or less). And the gap is overwhelmingly a feature of remarriage: 56% of husbands in a first marriage are roughly the same age as their wife, against 32% of husbands married more than once, while 35% of first-marriage husbands have a wife three or more years younger against 56% of remarried husbands.

An earlier Pew analysis of 2013 ACS data put numbers on the extreme end, among people who had married within the previous twelve months. 20% of newly remarried men had a wife at least ten years younger, and another 18% a wife six to nine years younger. Among men in a first marriage the same figures were 5% and 10%. For women: 13% of the newly remarried had a husband at least ten years older, against 7% of first-marriage wives, and 5% had a husband ten or more years younger, against 1%.

So a double-digit gap is not rare in the abstract and is distinctly uncommon in a first marriage. It is mostly what second marriages look like, which is worth knowing before reading anything into your own.

The pattern is not American. Jacob Ausubel, Stephanie Kramer, Anne Fengyan Shi and Conrad Hackett analysed census and survey data for 130 countries in Population Studies and found men on average 4.2 years older than their wives or cohabiting partners worldwide, ranging from 8.6 years in sub-Saharan Africa to 2.2 in North America. By religious group the largest was Muslims at 6.6 years and the smallest Jews at 2.2, and the Muslim/non-Muslim difference survived controls for country-level gender inequality and GDP per capita. For the rest of what the numbers on modern coupling do and do not support, see the real numbers on modern dating.

Further illustration for: Age Gaps in Relationships: What the Research Predicts, and Where the Divorce Numbers Came From
#7

Half your age plus seven, in the words of the man who printed it

There is one rule in this subject that people treat as a safeguard rather than folklore: the youngest partner you should consider is half your age plus seven, and run it backwards for the oldest. It is used as an ethical floor, often by people who are quite careful about everything else.

It is traceable, and the original is nothing like the modern use. The earliest known printing is in Her Royal Highness Woman and His Majesty—Cupid, published in 1901 by the French writer Léon Paul Blouet under the pen name Max O'Rell. The book is out of copyright and the full text is on Project Gutenberg, so this can be checked in about a minute. In a chapter of advice to men on when to marry, he writes:

> I heard the other day a very good piece of advice, which I should like to repeat here, as I endorse it thoroughly: A man should marry a woman half his age, plus seven.

Three things about that sentence. It is hearsay, and says so — he heard it from somebody and liked it. It is one-directional: a prescription to men about the right age for a bride, not a symmetrical floor and ceiling that both partners can apply. And it has no data of any kind behind it, in 1901 or since.

Then read the next paragraph, which is the part nobody quotes:

> Never marry a woman richer than you, or one taller than you, or one older than you. Be always gently superior to your wife in fortune, in size, and in age, so that in every possible way she may appeal to you for help or protection, either through your purse, your strength, or your experience of life. Marry her at an age that will always enable you to play with her all the different characteristic parts of a husband, a chum, a lover, an adviser, a protector, and just a tiny suspicion of a father.

That is the formula's stated purpose, in its author's own words. The number people now cite as a protection against a power imbalance was published as a recipe for producing one — a wife reliably poorer, smaller and younger, so that she would always have to ask. Half your age plus seven is not a safeguard that got forgotten. It is a prescription that got laundered.

Which leaves an honest gap. If the rule is worthless, there is no replacement number, because no study has established a threshold at which a gap becomes a problem. England and colleagues tested for one directly — a squared term in their model — and did not find it. The absence of a number is the finding.

#8

The one outcome where the gap reliably shows up

There is a place where the age difference between partners has a large, well-measured and almost entirely unromantic effect, and it is not divorce.

Sven Drefahl used Danish population registers — not a sample, the whole population — to follow every married person aged 50 or over living in Denmark between 1 January 1990 and 31 December 2005. Because Danish registers record exact dates of birth, he could measure the age gap in days rather than in five-year bands, which is what earlier work on this question had been stuck with. He published it in Demography in 2010.

The result is counterintuitive in a specific way. Having a younger spouse is beneficial for men and detrimental for women. Having an older spouse is detrimental for both.

Read that twice, because the folk version of this finding — marry younger and live longer — is half of it presented as all of it. For women, mortality was lowest when a woman was the same age as her husband and rose with the discrepancy in either direction. There was no younger-partner bonus for wives at all.

Two more details raise the quality of the evidence. The differences were not confined to extreme cases: being merely one to three years older than a spouse produced a significantly different survival rate from being within a year. And adding education and wealth to the model changed the picture substantially for men while barely touching it for women, which tells you socioeconomic selection is doing real work on the male side.

Drefahl is explicit about what he cannot say, and so is this page. His closing line is that the driving force behind these differences remains unclear. The candidate explanations — that healthier people attract younger partners, that a younger spouse provides care in later life, that women have wider social networks and so depend less on a spouse for support — are hypotheses he can rank but not test, because registers record events and not behaviour. He also notes that weaker health selection in women would fit: exceptionally healthy women are less able than their male counterparts to attract a younger partner in the first place.

And there is a mechanical consequence that needs no mechanism, which Pew states plainly in its own analysis: someone is more likely to end up widowed if their spouse is significantly older. As the share of marriages with a much older husband has fallen, so has widowhood among older women — 30% of American women aged 65 and over are widows, down from 45% in 2000. That is the most certain thing on this page, and it is the one consequence of a gap that nobody can negotiate their way out of.

#9

What age preferences look like when you measure them

The rules and the stereotypes both assume preferences that have actually been measured, and the measurements are more complicated than either.

Jan Antfolk surveyed a population-based sample of 2,655 adults aged 18 to 50 and asked, rather than for an ideal, for the youngest and oldest partners they would consider and had actually had. Published in Evolutionary Psychology in 2017, the results split cleanly:

  • Women of all ages preferred slightly older partners, and reported a narrower acceptable range than men.
  • Men preferred women in their twenties regardless of their own age — the familiar finding.
  • But men's range widens as they get older: they keep considering young women and also begin considering women their own age and older.
  • And the behavioural half, which is the part that rarely gets quoted: men's actual sexual activity reflected their own age range too. Their potential interest in much younger women, Antfolk writes, is not likely to be converted into sexual activity — a pattern he reads as support for female choice.

That last point is the bridge to what people do rather than what they say. Michèle Belot and Marco Francesconi, analysing speed-dating data in the Journal of Human Resources, found that both men and women preferred dates where the man was zero to five years older, over both larger gaps and the woman being older. Günter Hitsch, Ali Hortaçsu and Dan Ariely, with data from a major online dating site, found both sexes more likely to contact similarly aged partners and particularly inclined to avoid the woman-older pairing. Neither of those is a clean read on preference, and the reason is worth knowing: if people avoid contacting someone much younger because they expect to be turned down, strategic behaviour will look exactly like a taste for similarity. That caveat is Lee and McKinnish's, raised in the paper that produced the satisfaction findings above, and both of those studies are reported here as they summarise them rather than from our own reading.

Which is why the satisfaction result in section four is a useful cross-check. People have no strategic reason to misreport how satisfied they are with a marriage they are already in — and there, both sexes report more satisfaction with a younger spouse. The gap between what people pursue and what they report once they have it is a recurring problem in this literature, and it is the same gap that makes stated ideals such poor predictors in the research on compatibility and in whether dating apps work.

#10

The thing the gap is often standing in for

Almost everybody in an age-gap relationship reports the same experience, and it is not instability. It is being told about it.

There is a research tradition on exactly that. Justin Lehmiller and Christopher Agnew studied what they call marginalized relationships — ones that carry social disapproval — within the investment model of commitment, and published it in Personality and Social Psychology Bulletin in 2006. Three results, and the third is the one to keep:

Perceived marginalization negatively predicted commitment. People in marginalized relationships invested significantly less in them. And despite investing less, they were significantly more committed than people in non-marginalized relationships — compensating, the data suggested, by perceiving fewer alternatives rather than by reporting more satisfaction.

That is a strange and useful finding. Disapproval does not straightforwardly weaken a relationship; it changes what is holding it up. Commitment sustained by a narrowed sense of what else is possible is not the same structure as commitment sustained by satisfaction, even where the two produce the same answer on a questionnaire. The investment model's machinery for this — alternatives, investments, and how they trade off — is the same one behind why people stay in unhappy relationships.

Ages turn out to carry this kind of hidden weight elsewhere in the divorce literature too. The best-known finding about living together before marriage shrinks from a significant hazard ratio of 1.37 to a non-significant 1.09 depending on whether the models control for the age a couple married or the age they moved in together, which is what the evidence on living together before marriage reports.

The pair followed it with a study specifically on age-gap relationships in Psychology of Women Quarterly in 2008, comparing heterosexual women in age-gap and age-concordant relationships. Woman-older partners were the most satisfied and the most committed of the three groups, which fits a socio-cultural account rather than an evolutionary one. And the ordinary predictors of commitment — satisfaction, alternatives, investments, normative beliefs — explained noticeably less of the variance among age-gap partners than among similarly aged ones, which the authors read as a sign that something outside the standard model is doing work in those relationships.

Both studies are cross-sectional, and the 2008 one covers women only, in heterosexual relationships, at one point in time. They cannot tell you whether disapproval causes the pattern or accompanies it. What they do establish is that the social reaction is a measurable variable in its own right, with its own effects — which means that when the gap gets blamed for something, the disapproval it attracts is a live alternative explanation and not a deflection.

One reason the gap draws that reaction is that age is the trait on which partners resemble each other most. In the UK Biobank, the partner correlation for year of birth is .86 — far above education at .48 or any personality measure — so a wide gap is a visible departure from the one strong regularity in the data. That it is strong says nothing about whether it predicts anything, and the resemblance literature behind the figure is set out in what the research on whether opposites attract actually reports.

#11

What the evidence actually supports doing

Everything above produces a short list, and like most honest lists in this genre it is shorter than the question implies. Each item is tied to a measurement, and the inferences are marked as inferences.

Stop using the percentages. The 18/39/95 figures were withdrawn in 2014 by the person who produced them, on the researchers' own objection that they could not be derived from the data. Any page still quoting them has not checked, which is a reasonable thing to know about a page.

Ask who is older, not how much. The one large American study that separated who initiated found no effect of the gap as such, and a consistent effect of direction: the older partner is the one who tends to get left, whichever sex they are. If that finding generalises, the useful question is not how many years but which seat you are in.

Expect the honeymoon arithmetic to reverse. Differently-aged couples start more satisfied and decline faster, and the advantage is gone within six to ten years. A relationship that is going unusually well in year two is not evidence against the pattern; year two is where the pattern predicts it should be. Why the honeymoon phase ends covers the general version of that curve.

Prepare for the bad year specifically. The mechanism with the most direct support is reduced resilience to a financial shock — more than a quarter of couples in the Australian panel reported one in a given year. That is a thing two people can discuss in advance, unlike a birth date.

Discount the rule, and do not look for a replacement number. Half your age plus seven is a 1901 piece of hearsay printed as advice for keeping a wife dependent, and the one study that tested for a non-linear threshold did not find one.

Treat the disapproval as its own variable. It has measurable effects on how commitment is built, and those effects are not the same as the gap's. If the strain in a relationship appears mainly after contact with other people's opinions, that is worth separating out.

And take the actuarial part seriously, because it is the most certain thing here. A large gap means one person is statistically likely to spend years alone at the end, and the registry data is unambiguous about who. That is a conversation about wills, pensions and care, not about compatibility.

#12

What nobody has measured

The limits belong on the page in the same type size as the findings, and on this subject they are unusually extensive.

Nobody has measured the effect of a gap, only its correlates. There is no study in which the age difference between partners was assigned, and there cannot be. Every number on this page is observational, which is why the selection result is so important: when the people who form differently aged couples differ systematically beforehand, the gap may be a marker of who is in the couple rather than a force acting on it.

The best divorce study cannot separate relative from absolute age. England and colleagues say so. Their own alternative reading is that having an ageing spouse, rather than having a gap, is what moves the odds — and their data cannot choose between the two.

The satisfaction findings are Australian, first marriages only, ages 20 to 55. Whether the six-to-ten-year crossover holds elsewhere, in remarriages, or outside that age band has not been established — and remarriage is where most large gaps actually are.

The mortality findings are Danish, married, over 50, and from registers that record no behaviour. Drefahl's own conclusion is that the driving force remains unclear. Nobody has checked whether the pattern holds for cohabiting or same-sex couples, which he names as the next step.

Almost none of this literature is about same-sex couples. The Pew analyses are explicitly limited to opposite-sex marriages, the 130-country study is different-sex by design, and the commitment work on age gaps sampled heterosexual women only. One of the few things that can be said is what cannot: the base rates, the satisfaction trajectory and the divorce asymmetry above are not known to apply.

The preference data has a strategic-behaviour problem that nobody has solved, and the honest consequence is that stated ranges and observed contacts are both unreliable guides to what people want.

And the thing people most want to know has no answer. There is no measured threshold at which a gap becomes a risk, no ideal difference, and no study comparing the outcomes of couples who closed a gap with those who did not, because that is not a thing anyone can do.

What survives is still more than the search results offer. The famous percentages were retracted twelve years ago. The gap itself did not predict divorce in the best study that asked. Direction did, and it pointed at the older partner regardless of sex. Both sexes report more satisfaction with a younger spouse and lose that advantage within a decade. The people in these couples differ from the start, so the correlation is partly about them. The 1901 rule was written to engineer the imbalance it is now quoted to prevent. And the one outcome that is certain is the one about who outlives whom. The ages themselves are a separate literature with its own famous number, and that number turns out to come from a blog post whose own appendix contradicts it: the best age to get married.

If you want a sense of your own patterns rather than a verdict about a number, our attachment style test is free, scores in your browser and sends nothing anywhere. More of this under Dating, and the claims this site has had to withdraw are listed on the corrections page.

Pulling it together

The short version is that the question is badly posed. Almost everything written on age gaps asks how many years are too many, and the research that exists does not answer in years.

The best American study to separate who left from whether a marriage ended found no effect of the gap as such, and a clear effect of direction: the older of the two partners is the one who tends to get left, whichever sex they are, by about 8% per year of difference for husbands and 5% for wives. Thirteen years of Australian panel data show differently aged couples starting more satisfied and losing the advantage within six to ten years, with the strongest clue pointing at reduced resilience to a bad financial year rather than at the gap grinding anybody down. And the people who form these couples differ measurably from those who do not, before the relationship starts, which means a chunk of every correlation on the subject is a fact about them rather than about the arithmetic.

Two things are more certain than any of that. The rule people use as an ethical floor was printed in 1901 as hearsay, inside a paragraph recommending a wife reliably poorer, smaller and younger so that she would always have to ask. And in whole-population register data, a large gap means one person is likely to spend years at the end alone.

Which leaves no number to give you. The absence of one is not a gap in this page; it is the finding.

Frequently asked questions

Is 19 and 23 a bad age gap?

No study has established a threshold at which a gap becomes a risk, so there is no research-based answer at the level of two specific ages — and four years is close to the American norm in any case. Pew Research Center's analysis of 2022 American Community Survey data puts the average difference between US husbands and wives at 2.2 years, with 51% of opposite-sex marriages within two years and 40% having a husband three or more years older. The research that does exist asks different questions: the largest US study to separate who initiated a divorce found the size of the gap predicted nothing, while the direction did, and Australian panel data find differently aged couples starting more satisfied and losing that advantage within six to ten years. Both of those are population patterns, not forecasts about two people. The one thing worth checking at any gap is whether the asymmetry is being used — whether one person's independence, friendships or money are up for discussion and the other's are not.

Does an age gap increase the risk of divorce?

The best study to ask properly says no, not the gap itself. Paula England, Paul Allison and Liana Sayer used the National Survey of Families and Households — 3,622 couples, 413,650 couple-months, 747 separations — and wrote in their abstract that support was not found for the hypothesis that age heterogamy itself increases the odds of divorce. What they did find was an asymmetry by direction: each extra year the husband was older cut his odds of leaving by 8% and raised hers by 5%, and a man was about half as likely to leave a marriage with a younger wife as a same-age one. Read together, the older partner is the one who tends to get left, whichever sex they are. The authors also explain why earlier research looked different: because the two effects push in opposite directions, they partly cancel in any model that does not ask who initiated, which they say has made much past work on this misleading.

Where does the 18% divorce statistic for a five-year age gap come from?

From a blog post that retracted it. In November 2014 Randal Olson, then a computer science graduate student, charted factors from Francis-Tan and Mialon's paper on wedding spending and marriage duration, and the resulting percentages — 18% for five years, 39% for ten, 95% for twenty — were picked up by the New York Post five days later and have circulated since. His post now carries a note stating that the study's original authors pointed out that although there is a significant correlation between wider age gaps and increased divorce, it is not possible to determine the relative percent likelihood from their study, and that this is left to future research. The direction was never withdrawn; the magnitudes were, roughly twelve years ago. The underlying paper is about what couples spent on the engagement ring and the ceremony, and age difference was a covariate rather than its subject.

What is the ideal age gap in a relationship?

Nobody has measured one, and the structure of the evidence suggests nobody can. England and colleagues tested directly for a non-linear relationship — a squared term for age difference — and it was not significant, meaning there is no point in their data where a gap starts to matter more per year. A continuous model fit better than banded categories. What has been measured is messier and more interesting: in the Australian panel both men and women reported higher satisfaction with a younger spouse, which contradicts the half of the stereotype that says women want older partners, and both lost that advantage within six to ten years. Jan Antfolk's survey of 2,655 adults found women preferring slightly older partners at every age and men preferring women in their twenties at every age, with men's actual behaviour reflecting their own age range rather than that preference. Any article that gives you a number for the ideal gap has made it up.

What is the half your age plus seven rule, and is there anything behind it?

There is nothing behind it, and its origin is checkable in about a minute. The earliest known printing is in Max O'Rell's 1901 book Her Royal Highness Woman and His Majesty—Cupid, which is out of copyright and on Project Gutenberg. He introduces it as something he heard: a man should marry a woman half his age, plus seven. It was advice to men about the right age for a bride, not a symmetrical floor and ceiling, and it had no data behind it then or since. The paragraph immediately after is the part nobody quotes: never marry a woman richer, taller or older than you, be always gently superior in fortune, size and age so that she may appeal to you for help or protection, and marry her at an age that lets you play the parts of husband, chum, lover, adviser, protector and just a tiny suspicion of a father. The rule now used as a guard against a power imbalance was published as a method for producing one.

What is the average age gap between couples?

In the United States, 2.2 years, and falling. Pew Research Center analysed decennial census and American Community Survey microdata — the 2022 file alone covers about 680,000 co-resident opposite-sex marriages — and found an average difference of 2.2 years in 2022, down from 2.4 in 2000 and 4.9 in 1880. Worldwide the figure is larger: Ausubel, Kramer, Shi and Hackett analysed census and survey data from 130 countries in Population Studies and found men on average 4.2 years older than their wives or cohabiting partners, ranging from 8.6 years in sub-Saharan Africa to 2.2 in North America. By religious group they reported the largest average gap among Muslims at 6.6 years and the smallest among Jews at 2.2, with the Muslim and non-Muslim difference surviving controls for country-level gender inequality and GDP per capita. All of these analyses are limited to different-sex couples.

Is a 10-year age gap too much?

There is no measured threshold, so nothing in the research licenses a yes or a no. What the data can tell you is how unusual it is and where it tends to occur. Pew's analysis of 2013 ACS data on people who had married in the previous twelve months found that 5% of men in a first marriage had a wife at least ten years younger, against 20% of men who had just remarried; for women, 7% of first-marriage wives had a husband at least ten years older, against 13% of the newly remarried. So a double-digit gap is mostly a feature of second marriages rather than first ones. The 2022 census-based figures point the same way: 56% of husbands in a first marriage are roughly the same age as their wife, against 32% of husbands married more than once. If you want a research-informed question instead of a verdict, the useful ones are which of you is older and how the two of you would handle a bad financial year.

Do age-gap relationships last?

Many do, and the honest answer is that the research cannot tell you whether the gap is what decides it. Two findings pull in different directions. Lee and McKinnish's analysis of 3,374 Australian couples across 18,987 couple-years found satisfaction declining faster in differently aged couples, erasing their initial advantage within six to ten years, with the clearest mechanism being a larger loss of satisfaction after a major financial setback — which more than a quarter of couples reported in a given year. But Mansour and McKinnish found that people married to differently aged spouses are negatively selected on earnings, occupation and measured ability, using five US censuses. If the people who form these couples differ beforehand, part of any correlation between gap size and outcome is a fact about who is in them rather than about what the gap does. Nobody has run, or can run, the experiment that would separate the two.

Does marrying a younger partner make you live longer?

For men, the register data is consistent with it. For women, it is not — and that half is usually dropped. Sven Drefahl used Danish population registers covering every married person aged 50 or over in Denmark between 1990 and 2005, with the age gap measured in days rather than in five-year bands, and found that having a younger spouse was beneficial for men but detrimental for women, while having an older spouse was detrimental for both. Women's mortality was lowest when they were the same age as their husband and rose as the discrepancy grew in either direction. Two caveats are his own: adding education and wealth to the model changed the pattern substantially for men and barely at all for women, which suggests socioeconomic selection is doing real work, and he closes by writing that the driving force behind the differences remains unclear. Registers record events, not behaviour, so the caregiving and health-selection explanations are untested.

Why do people disapprove of age-gap relationships, and does it matter?

It matters measurably, and separately from the gap. Justin Lehmiller and Christopher Agnew studied socially disapproved relationships within the investment model and found that perceived marginalization negatively predicted commitment and that people in marginalized relationships invested significantly less — yet were significantly more committed than people in non-marginalized ones, apparently compensating by perceiving fewer alternatives rather than by reporting more satisfaction. Commitment held up by a narrowed sense of what else is possible is a different structure from commitment held up by satisfaction, even where both produce the same questionnaire score. Their follow-up on age-gap relationships specifically found woman-older partners the most satisfied and committed of three groups, and found the standard predictors of commitment explaining less variance among age-gap partners than among similarly aged ones. Both studies are cross-sectional and the second sampled heterosexual women only, so neither can say whether disapproval causes the pattern or travels with it.

Sources

What this article draws on. For the broader bibliography see our sources page.

Keep reading