Why Country Divorce Rate Rankings Are Measuring the Wrong Thing

If you've ever seen a chart ranking countries by divorce rate, the number on it was almost certainly the wrong one for that purpose. Not fabricated — wrong in the sense that it doesn't measure what the chart implies it measures. This is a methods problem, not a data problem, and it's worth understanding before you cite, share, or build on any country-level divorce statistic.

The Number You've Seen Is the Crude Rate

Every major free source for cross-country divorce data — the OECD Family Database (table SF3.1), Eurostat's `demo_ndivind`, Our World in Data's divorce pages, and the UN Demographic Yearbook (Table 25) — publishes the same underlying statistic: the crude divorce rate, defined as divorces per 1,000 total population in a given year.

That definition has two confounders baked in, and both push in the same direction.

First, the denominator is the *entire population*, including children, the unmarried, and people who will never marry. A country where marriage is rare or delayed will show a low crude divorce rate almost no matter how unstable the marriages that do happen are — because most of its population was never eligible to contribute a divorce in the first place.

Second, age structure matters independently. A country with a younger population, or one where marriage happens later, will mechanically show fewer divorces per 1,000 people, again regardless of how marriages inside it are actually holding up.

The methodological citation for this is Stevenson and Wolfers (2007, *Journal of Economic Perspectives*), who lay out why the crude rate is unsuitable for cross-country comparisons of marital stability and why economists studying divorce trends use age-standardized or cohort-based measures instead. A country's crude divorce rate tells you as much about its marriage rate as its divorce rate — you cannot separate the two from that one number.

This is why ranking countries by crude divorce rate — the kind of chart that circulates as "most/least divorce-prone nations" — measures something closer to marriage prevalence than marital instability. A society with a low marriage rate can produce an artificially low crude divorce rate while having a high proportion of marriages that eventually fail.

The Trap Inside Eurostat: DIVMARPCT

Eurostat also publishes a second, more tempting number: divorces per 100 marriages in the same calendar year (sometimes labeled DIVMARPCT in older releases, now folded into the demography tables). It looks like exactly the "percentage of marriages that end in divorce" figure everyone wants.

It isn't. This ratio compares two different cohorts: the marriages registered *this year* against the divorces finalized *this year*, which mostly dissolve marriages that started years or decades earlier. If marriage rates are falling — which they are across nearly every OECD country — the denominator shrinks over time while the divorce count reflects a larger, older pool of marriages. The ratio inflates for a reason that has nothing to do with any single marriage's odds of survival.

This is the identical logic behind the "50% of marriages end in divorce" claim in the United States: it came from dividing one year's marriage count into that same year's divorce count, comparing groups that were never the same population. The CDC/NCHS and Kennedy and Ruggles (2014, *Demography*) both trace the error to this same-year-ratio mistake. Eurostat's DIVMARPCT is the same trap wearing a different flag.

The Number That Actually Answers the Question

If what you want to know is "what share of marriages that started in year X eventually end in divorce," you need a cohort study — one that follows an actual group of marriages forward through time and tracks how many of them dissolve by year 10, year 20, year 30. For more details, see our guide on follower tracker guide. For more details, see our guide on see who unfollowed you.

The best public version of this is UK Office for National Statistics data, republished by Our World in Data as the dataset "marriages-uk-ended-in-divorce." It follows real marriage cohorts (grouped by the year the marriage began) and reports the cumulative percentage that had ended in divorce by a given anniversary. This is structurally different from both the crude rate and DIVMARPCT — it can't be inflated or deflated by a falling marriage rate in some unrelated later year, because it's tracking one group of marriages against itself over time.

Cohort-based estimates like this are also the source of the more defensible range you'll see in careful reporting: roughly 35-45% of marriages ending in divorce, depending on the cohort and country, rather than a flat "50%." Kennedy and Ruggles (2014) is again the relevant citation here, and their finding cuts against another common assumption: they show age-standardized divorce actually *rose* among Americans over 35 in recent decades, which complicates any simple "divorce is steadily declining" narrative built on crude rates alone.

A Correction Worth Making Explicitly

The popular belief is that you can look up "the divorce rate" for a country and compare it to another country's the way you'd compare life expectancy. The data doesn't support treating the crude rate that way. What it actually shows, credibly, is a per-1,000-population event count that mixes marriage prevalence, age structure, and divorce incidence into one indistinguishable figure. Two countries can have identical marital stability and wildly different crude rates just because one marries more people, younger.

To be fair to the crude rate: it isn't useless. It's a legitimate measure of divorce *events per capita* in a given year, which matters for things like family court caseload planning or year-over-year trend-watching *within* the same country, where the population's marriage patterns aren't shifting drastically year to year. The problem is specifically cross-country ranking and any claim that it measures how likely a marriage is to fail.

One more correction: the World Bank is frequently cited as a divorce-rate source, including in casual reference lists and some AI-generated summaries. It publishes none. If you see a "World Bank divorce rate," the citation is wrong regardless of what number sits next to it.

Method

Nothing here required new data collection — this is a methodological explainer, and every figure and definitional claim above is checkable directly against the primary sources:

• OECD Family Database, table SF3.1 ("Marriage and divorce rates"): https://www.oecd.org/en/data/datasets/oecd-family-database.html
• Eurostat, `demo_ndivind` (divorce indicators): https://ec.europa.eu/eurostat/databrowser/view/demo_ndivind/default/table
• Our World in Data, divorce pages (aggregating the above with crude-rate caveats): https://ourworldindata.org/marriages-and-divorces
• UN Demographic Yearbook, Table 25 (divorces and crude divorce rates): https://unstats.un.org/unsd/demographic-social/products/dyb/dybsets/2022.pdf
• Our World in Data, cohort dataset "marriages-uk-ended-in-divorce" (UK ONS-derived): https://ourworldindata.org/grapher/marriages-uk-ended-in-divorce
• Stevenson, B. and Wolfers, J. (2007). "Marriage and Divorce: Changes and their Driving Forces." *Journal of Economic Perspectives*, 21(2). The methodological case against crude cross-country comparison.
• Kennedy, S. and Ruggles, S. (2014). "Breaking Up Is Hard to Count: The Rise of Divorce in the United States, 1980-2010." *Demography*, 51(2). Source for the 35-45% cohort range and the over-35 age-standardized rise.

To reproduce any comparison in this piece: pull the OECD or Eurostat table, check whether the metric is labeled "crude divorce rate" (per 1,000 population) or a same-year ratio like DIVMARPCT, and treat either as a marriage-prevalence-confounded number rather than a marital-stability score. For an actual "percentage of marriages that end in divorce," use the cohort series, not the annual ratio.

This piece is part of a research series on storiesfly.com auditing commonly cited relationship statistics for sourcing. The raw datasets behind other pieces in the series are public: relationship-seasonality.json, relationship-seasonality.csv, and relationship-vocabulary.json.

FAQ

Does a country with a high crude divorce rate have less stable marriages?

Not necessarily. The crude rate divides divorces by total population, so it's pulled around by how many people marry and the country's age structure. A country with a high marriage rate and a moderate divorce rate can show a higher crude number than a country where marriage is rare, even if the second country's marriages are objectively less stable.

What percentage of marriages actually end in divorce?

Cohort studies following real marriage groups over decades — not same-year ratios — put the figure roughly in the 35-45% range depending on country and marriage cohort, per Kennedy and Ruggles (2014). There is no single global number, and "50%" specifically comes from a same-year ratio error, not a cohort study.

Does the World Bank publish country divorce rates?

No. Despite being cited as a source in some secondary summaries, the World Bank does not publish a divorce-rate indicator. For legitimate cross-country divorce data, the OECD Family Database, Eurostat, and the UN Demographic Yearbook are the primary free sources.

Is Eurostat's divorces-per-100-marriages number reliable?

It's real data, but it compares two different marriage cohorts (this year's new marriages against this year's finalized divorces, which mostly involve older marriages), so it isn't a valid "percentage of marriages that end in divorce." Use a cohort-based series, like the UK ONS data on Our World in Data, for that question instead.