The strongest predictor of a good relationship is not the one most people are told to look for.
Are Relationships Worth It? What the Evidence Actually Shows
Nearly half of single Americans now say dating simply is not financially worth it. That is the headline finding from BMO's 2026 Real Financial Progress Index, a survey of 2,501 US adults conducted by Ipsos over the final weeks of 2025. The average cost of a single date has reached $189, up 12.5 per cent on the year before, and the average person who dates now spends $2,323 annually on the pursuit of a partner. Half of respondents say they have already cut back, going on fewer dates or choosing cheaper ones, because the arithmetic has stopped working.
It is a fair question, asked properly. Is a relationship worth what it costs, in money, in time, in the vulnerability it demands of two people before either has evidence the other will reciprocate? The research offers a genuine answer, and it is more interesting than a simple yes or no. Relationships are, on average, associated with better financial and physical outcomes. But the size of that benefit depends overwhelmingly on the quality of the relationship rather than the fact of being in one, and the strongest predictors of that quality are not the ones most people are told to look for.
The economic case
Before any benefit of partnership can be counted, there is the cost of finding it. BMO's data shows dating expense outpacing general inflation: Gen Z daters report spending around $205 a date and Millennials $252, both sharply higher than the year before. Dating apps add a further layer of spend. Morgan Stanley estimates the average paying user spends $18 to $19 a month, and Pew Research Center found in 2022 that 35 per cent of dating app users had paid for a subscription or feature, with premium tiers running considerably higher again. For roughly half of Gen Z and 40 per cent of Millennial respondents in the BMO survey, these costs now interfere with broader financial goals.
Once a relationship becomes a marriage, the economic picture shifts, and shifts in a direction replicated often enough to be considered one of the more durable findings in labour economics. Married men earn more, on average, than single men, with a premium often put at around 15 per cent across the compiled literature, though individual studies range considerably depending on sample and method. Pilossoph and Wee's 2021 analysis in the American Economic Journal: Macroeconomics attributes a meaningful share of that premium, and an even larger share of a smaller premium found for women, to household-level search effects: couples can afford to be choosier about jobs when supported by a second income. Earlier work by Korenman and Neumark points to married men reaching higher job grades and posting faster early-career wage growth.
None of this should be read as marriage causing the premium outright. The relationship is almost certainly confounded by selection: people who are already reliable, stable earners may simply be more likely to marry in the first place. The premium is also sharply gendered. It accrues overwhelmingly to men, while women face a well-documented motherhood penalty in wages per child, running in the opposite direction to the so-called fatherhood bonus.
Jay Zagorsky's widely cited 2005 analysis of 9,055 Americans tracked from their twenties into their forties found that married respondents accumulate 77 per cent more net worth per person than single respondents, with wealth increasing by around 16 per cent for every year of marriage. Divorce reverses the pattern sharply: wealth begins falling an average of four years before the legal separation, and the total drop also averages 77 per cent.
The emotional ledger
If the economic case is mixed and heavily conditional, the emotional one turns out to hinge on a single distinction that runs through almost all of the research: quality matters far more than status. Simply being married, rather than single, correlates only weakly with happiness, with status-to-wellbeing correlations sitting around r = 0.14. Much of the wellbeing boost that arrives with a new relationship fades within a few years, a well-established pattern known as hedonic adaptation. People acclimatise to their circumstances, good and bad, and a wedding is not exempt.
What does not fade is the association between social connection and physical health. Holt-Lunstad and colleagues' influential 2015 meta-analysis found that social isolation, loneliness, and living alone were associated with 29, 26, and 32 per cent increases in mortality risk respectively, a magnitude the authors placed on a par with well-established clinical risk factors. This concerns social connection broadly rather than romantic partnership specifically, but it establishes just how physiologically real the benefit of connection can be.
The other side of the emotional ledger is less often discussed in pieces that celebrate coupledom, but it matters just as much. Relationships characterised by conflict, emotional labour imbalance, or outright abuse carry substantial, measurable costs, associated with markedly elevated rates of depression and post-traumatic stress symptoms compared with the general population. A relationship is not automatically protective. Its effect depends on what happens inside it.
What actually predicts a good one
This is where the evidence becomes genuinely useful, because it answers a question people ask themselves constantly and mostly answer with folk wisdom: what should I actually be looking for in a partner?
The first finding surprises people who assume that opposites attract, or that shared personalities guarantee ease. They do not, reliably. Across the compatibility literature, an individual's own traits, what researchers call actor effects, consistently outweigh the traits of their partner or the degree of similarity between the two. Shared values and background appear to matter considerably more than matched personality profiles.
The trait that does predict outcomes, consistently, across multiple independent lines of evidence, is emotional intelligence measured as a stable disposition rather than a one-off performance. Malouff, Schutte and Thorsteinsson's 2014 meta-analysis of six studies (603 participants) found trait emotional intelligence correlated with relationship satisfaction at r = 0.32, with both actor and partner effects present. Jardine, Vannier and Voyer's larger and more recent 2022 meta-analysis, drawing on 90 effect sizes from 78 samples, put the figure at r = 0.373 (95% CI 0.338 to 0.447), currently the best available estimate.
Two further studies help explain why the association holds. Wollny, Jacobs and Pabel's 2020 study of 136 German couples using the TEIQue Short Form found trait emotional intelligence predicted satisfaction through both actor and partner pathways, an effect partially mediated by dyadic coping: the way couples manage stress jointly rather than separately. Parker and colleagues' 2021 fifteen-year longitudinal study, tracking 314 adults from their first year of university into middle adulthood, found early-adulthood trait emotional intelligence to be a moderate predictor of relationship satisfaction fifteen years later, even after accounting for concurrent emotional intelligence at the second time point: a rare piece of evidence for genuine temporal precedence rather than a same-moment correlation.
It is worth being precise about what these figures mean. An r of 0.37 explains around 14 per cent of the variance in relationship satisfaction. Relationships remain determined by many factors at once, and no single trait, however well measured, accounts for the majority of what makes a partnership work. The honest reading is that trait emotional intelligence is one of the more replicable and better-understood dimensional predictors available, not a complete explanation on its own.
This is also where popular personality typing runs into trouble. Tools such as the Myers-Briggs Type Indicator sort people into one of sixteen fixed categories and have consistently shown poor test-retest reliability, meaning a meaningful share of people are classified differently on repeat testing of the same underlying disposition. More specifically to relationship prediction, MBTI omits neuroticism, one of the single most robust personality predictors of relationship satisfaction in the wider literature. DISC, a workplace behavioural framework built on a similar four-quadrant logic, has not been validated against romantic outcomes at all. Neither tool is without value as a common language for discussing differences. Neither has the predictive evidence base that dimensional, validated measures of trait emotional intelligence do.
The distinction that matters is between a type and a dimension. A type sorts a person into a box and, implicitly, suggests the box is fixed. A dimensional trait, measured on a continuous scale, can be developed. Trait emotional intelligence is not a verdict on a person's capacity for a good relationship. It is a measurable starting point that responds to attention and practice, a considerably more useful thing to know about oneself, or a partner, than a four-letter label.
This is precisely the territory the TEIQue Compatibility Analysis was built to explore: a dyadic assessment across the fifteen facets of trait emotional intelligence, developed by Professor K.V. Petrides at UCL, generating compatibility indices grounded in the same body of research summarised above rather than in typology.
For anyone weighing the question seriously, understanding one's own emotional profile, and how it interacts with a partner's, is a considerably more evidence-based starting point than a personality quiz.
Explore TEIQue CompatibilityA note on the limits of this evidence
Everything above is correlational. None of it establishes that marriage, high trait emotional intelligence, or any other single factor causes better outcomes in the strict sense; selection effects, the possibility that people already disposed toward stability or emotional skill are also more likely to marry or to score well on these measures, remain a live and largely unresolved alternative explanation throughout this literature. Effect sizes for trait emotional intelligence, while consistent across studies, are moderate rather than large, and attenuate somewhat once shared variance with broader personality traits is accounted for. Every figure here should be read as describing an association worth taking seriously, not a guarantee.
References
- BMO Financial Group / Ipsos (2026). Real Financial Progress Index: Date-flation hits hard. Fieldwork 29 December 2025–27 January 2026, n = 2,501 US adults.
- Pilossoph, L., & Wee, S. L. (2021). Household search and the marital wage premium. American Economic Journal: Macroeconomics.
- Korenman, S., & Neumark, D. (1991). Does marriage really make men more productive? Journal of Human Resources, 26(2), 282–307.
- Zagorsky, J. L. (2005). Marriage and divorce's impact on wealth. Journal of Sociology, 41(4), 406–424.
- Holt-Lunstad, J., Smith, T. B., Baker, M., Harris, T., & Stephenson, D. (2015). Loneliness and social isolation as risk factors for mortality: A meta-analytic review. Perspectives on Psychological Science, 10(2), 227–237.
- Malouff, J. M., Schutte, N. S., & Thorsteinsson, E. B. (2014). Trait emotional intelligence and romantic relationship satisfaction: A meta-analysis. American Journal of Family Therapy, 42(1), 53–66.
- Jardine, B. B., Vannier, S. A., & Voyer, D. (2022). Emotional intelligence and romantic relationship satisfaction: A systematic review and meta-analysis. Personality and Individual Differences, 196, Article 111713.
- Wollny, A., Jacobs, I., & Pabel, L. (2020). Trait emotional intelligence and relationship satisfaction: The mediating role of dyadic coping. The Journal of Psychology, 154(1), 75–93.
- Parker, J. D. A., Summerfeldt, L. J., Walmsley, C., O'Byrne, R., Dave, H. P., & Crane, A. G. (2021). Trait emotional intelligence and interpersonal relationships: Results from a 15-year longitudinal study. Personality and Individual Differences, 169, Article 110013.
