Do arranged marriages last longer than love marriages?

Start with the famous statistic, because it is worth knowing that it is not one. The four percent figure appears in infographics, listicles and wedding-industry copy, and tracing it leads back only to other infographics. No study anyone can name produces it. Treat it as folklore that acquired a decimal point.

The broader comparison it stands in for is real, and it is confounded. India, where arranged marriage remains the norm, records among the lowest divorce rates in the world. It also has divorce law segmented by religious community with different requirements for each, heavy and asymmetric social cost attached to being divorced, low female labour force participation, and family structures in which a separated woman often has no independent household to move into. Every one of those raises the price of the exit without telling you anything about what is happening inside the marriage.

Economists who have tracked divorce across a century and a half find that recorded rates move with legal regime, women's economic position and the changing gains from marriage rather than tracking marital quality in any simple way. The effect of specific reforms is itself argued over, with some estimates of unilateral divorce showing a sharp rise that later dissipates rather than a permanent shift[9]. The direction of the logic is what matters here. Where divorce is difficult, expensive and shaming, a low divorce rate is a fact about the door, not about the room.

This has a practical edge for anyone reading in order to decide something. The sentence "these marriages rarely end" is only reassuring if ending was available in the first place.

The study behind almost every confident claim in this area is Gupta and Singh, published in 1982 in the Indian Journal of Applied Psychology[1]. They surveyed fifty couples in Jaipur using Rubin's love scale and reported that couples who married for love scored high early and declined, while couples in arranged marriages started low and rose, with the two crossing at roughly the five year mark. That single result is the source of the widely repeated line that love fades in one and grows in the other.

Fifty couples, one city, one measure, four decades ago, comparing different couples at different stages rather than following the same couples through time. It is a reasonable exploratory study. It is not a foundation, and it has never been replicated at anything like the scale the claim implies.

The closest thing to a replication cut the other way. Regan, Lakhanpal and Anguiano surveyed fifty-eight Indian adults living in the United States, twenty-eight in arranged marriages and thirty in love-based ones, and found no significant differences in marital satisfaction, commitment, passionate love or companionate love[2]. Myers, Madathil and Tingle had reached a similar place earlier with forty-five people in arranged marriages in India compared against existing American data: differences in which qualities people said mattered, no difference in reported satisfaction[4].

Those null results deserve their own caveat, and it was published. Johnson and Bachan responded in the same journal arguing that a sample of fifty-eight is far too small to support a claim that no difference exists, since a study that size would miss even a fairly large real effect most of the time[3]. That correction is the most useful thing in this literature. Nobody has shown arranged marriages are more satisfying. Nobody has shown they are less. The studies are too small to have shown either, and the three most-cited ones together involve fewer than two hundred people.

Where the research gets more solid is when it stops treating arranged and love as two boxes and starts asking how much say the person marrying actually had.

Allendorf and Ghimire examined marital quality in the Chitwan Valley of Nepal, a society where arranged marriage is standard, using a survey large enough to support the analysis. Across five dimensions of marital quality, the strongest determinants were gender, education and spouse choice: people who had participated in choosing their own spouse reported higher quality. Caste, occupation, age at marriage, marital duration and number of children showed little or no association[5].

Flicker and colleagues tested the point directly with a hundred and sixteen Bangladeshi women who were engaged or in the first three years of marriage. They compared three models: marriage type alone, perceived influence over partner selection alone, and both together. Influence over partner selection predicted intimacy, passion, commitment and positive marital quality better than the label did, and the label added little once influence was accounted for. Their own conclusion was that the arranged and couple-initiated categories are better handled as a continuum than a dichotomy[6].

The largest study in this area points the same way with an uncomfortable detail attached. Imamoglu, Ads and Weisfeld surveyed four hundred and fifty-six Turkish couples, about a third of them in arranged marriages. Spouses in self-selected marriages reported more love. Wives reported more problems than their husbands whatever the marriage type. And wives in arranged marriages reported significantly less sense of partnership than both the husbands and the wives in self-selected marriages[7]. That is a gendered finding, and it is invisible to any comparison that only counts divorces.

Two arguments are usually presented as conclusions when they are actually views held by people, and it is worth naming them as such.

The first is the stability argument, made by defenders of family involvement in matchmaking and by a good deal of conservative commentary in South Asia and the Middle East. It holds that families evaluate the things that predict whether a marriage will work, compatible finances, aligned expectations, comparable family situations, at a moment when romantic feeling makes the couple themselves poor judges, and that two invested families supply mediation and practical support that a couple who married alone will not have. That is a coherent argument. It is not a research finding, and the evidence above does not confirm it.

The second is the modernity argument, common in Western commentary and among younger people in the same societies. It holds that choosing your own partner is simply the more evolved arrangement and that arranged marriage is a survival on its way out. Setting aside whether progress is the right frame, the trend data does not cooperate. Allendorf and Pandian examined Indian marriage between 1970 and 2012 and found women increasingly active in choosing their husbands and spouses meeting before the wedding day more often, but the changes were modest and substantial majorities of recent marriages still carried the hallmarks of arranged marriage. What was actually happening was not young people displacing parents. It was young women working with their parents to choose collectively, a hybrid rather than a handover[8].

The useful move is to stop asking which category wins and start asking what varies inside them. On that question the evidence, thin as it is, points consistently at one thing: whether the person getting married had a real say.

What decides this for you depends on what you are actually asking. If the question is which arrangement is more likely to produce a marriage that continues, the answer in most arranged-marriage societies is probably yes, and a serious part of that is barriers to exit rather than merit. If the question is which produces a better marriage to be inside, no one has demonstrated an advantage in either direction, and the studies are too small to have detected a moderate one. Those are different questions and the evidence separates them cleanly.

There is also a comparison problem that no amount of data solves. An arranged marriage in Hyderabad and a love marriage in Manchester differ in legal system, economy, housing market, extended family structure and the cost of being a divorced woman, all at once. Anyone comparing the two is comparing societies at least as much as marriages. The comparisons worth trusting are the ones made inside a single society, which is exactly what the Nepali, Bangladeshi and Turkish studies did, and those find the arranged and love labels doing less work than the amount of choice involved.

One last thing worth saying plainly, because it is where the argument usually goes wrong on both sides. A marriage that cannot be exited is not thereby a marriage that is working, and treating duration as the score builds that confusion into the question before anyone starts answering it.