- The study analyzed 43 longitudinal couples studies from 29 labs using machine learning (PNAS, 2020).
- The top 5 predictors: perceived partner commitment, appreciation, sexual satisfaction, perceived partner satisfaction and conflict.
- Relationship variables explained up to 45% of initial satisfaction; individual traits, up to 21%.
- Bottom line: how two people experience and choose each other matters more than a list of traits.
Tall, a professional, loves to travel, doesn’t smoke. We all have a list. But how well does that list actually predict whether a relationship will work? A large team of scientists set out to find out, using the biggest couples dataset ever assembled at the time.
The study
In 2020, psychologist Samantha Joel and an international team published in PNAS, one of the world’s most prestigious scientific journals, an analysis of 43 longitudinal couples studies conducted by 29 labs. They used machine-learning algorithms to identify which of hundreds of variables best predict satisfaction and commitment in a relationship.
The 5 most powerful predictors
The variables that best predicted relationship quality were all about the relationship itself, not about the individuals separately:
- Perceived partner commitment: feeling that the other person is serious about you.
- Appreciation: feeling valued and valuing the other person.
- Sexual satisfaction.
- Perceived partner satisfaction: feeling that the other person is happy with you.
- Conflict: how much you argue, and how.
What about the person’s own traits?
They matter too, just less. Among individual traits, the most predictive were satisfaction with one’s own life, negative affect, depression and avoidant and anxious attachment styles. We explain them in attachment styles.
A revealing detail: once researchers knew how each person experienced their relationship, individual traits and the partner’s opinions added no further predictive power. What matters most is the shared experience.
The most humbling finding
The study found something else: changes in satisfaction over time were, to a large extent, unpredictable. No combination of variables could fully anticipate how a relationship would evolve. Relationships aren’t a formula: they’re built day by day.
What this means for you if you’re looking for a partner
- Shorten your list of superficial requirements and lengthen your list of relational ones: does this person commit? Do they appreciate me? How do they handle conflict?
- Look for shared intention. If the number-one predictor is perceived commitment, meeting people who want the same thing from the outset is your best bet.
- Work on yourself: your personal wellbeing and the way you attach do make a difference.
- Give it time: satisfaction is built, not found ready-made.
We’re not looking for someone perfect for you. We’re looking for someone who chooses you as much as you choose them.
TovLov philosophy
Frequently asked questions
What predicts a happy relationship?
According to the analysis of 43 studies published in PNAS (2020), the strongest predictors are perceived partner commitment, appreciation, sexual satisfaction, perceived partner satisfaction and the level of conflict.
What matters more, the person or the relationship?
The relationship. Variables about how the relationship is experienced explained up to 45% of initial satisfaction, compared with up to 21% for individual traits.
Can you predict whether a relationship will last?
Only partly. The study found that changes in satisfaction over time were largely unpredictable: relationships aren’t a formula, they’re built day by day.
- Joel et al. (2020) — Machine learning uncovers the most robust self-report predictors of relationship quality across 43 longitudinal couples studies, PNAS
TovLov TeamPersonalized matchmaking in Mexico. We research and write about relationships using official data.
