Multidimensional Homophily: Why We Attract People Who Are Like Us on Many Levels

What Is Multidimensional Homophily?

People carry many attributes at once — gender, ethnicity, social status, age — and can match others on several of these dimensions simultaneously. Multidimensional homophily describes the phenomenon that these similarities do not affect friendship formation independently but interact with one another.

The key finding of a longitudinal study across eleven school classes in Scotland, England/Wales, and the USA (Grund & Densley): the well-known homophily effects hold — same gender, same ethnicity, or similar socioeconomic status each increase the probability of a friendship. Yet the interaction of these effects is negative. When two people are similar on multiple dimensions at once, the increase in friendship probability is less than the sum of the individual effects would suggest. In nine of the eleven schools studied, this interaction effect was statistically significant and substantial.

Diminishing Returns

A concrete example from the US data: same ethnicity makes a friendship 1.67 times more likely, and same gender increases the probability by a factor of 1.92. But both together — same ethnicity and same gender — increases the probability by only a factor of 2.03, far less than a multiplicative combination (≈ 3.21) would suggest.

Three explanations exist for these diminishing returns. First, there may be a utility ceiling: when bridging relationships with dissimilar people provide valuable new resources and ideas (Granovetter, 1973; Burt, 1992), too much similarity becomes redundant. Second, sharing multiple attributes may create additional opportunities for encounters, but those do not automatically translate into friendships — people who don’t get along in one context won’t in another. Third, Social Identity Theory (Tajfel & Turner, 1979; Hogg, 2006) suggests that in any given situation, only one dimension of similarity is salient — boys on a soccer field notice shared gender, not ethnicity or status.

Which Dimensions Count?

The study by Grund and Densley closes a research gap: although homophily has been demonstrated in over a hundred studies (McPherson et al., 2001), almost all examined only individual dimensions in isolation. As early as 1950, Simmel conceived of individuals as intersections of multiple “social circles” whose crossings constitute identity — a theoretical precursor to the multidimensional approach. Methodologically, the study uses stochastic actor-oriented network models (SAOMs) that control for endogenous network processes such as reciprocity and transitivity, allowing more causally robust claims than cross-sectional analyses. Replication across three different datasets from four countries strengthens generalizability.

Additionally, Gompers et al. (2016) show that an affinity score — an overall index of similarity-based attributes between venture capital investors — significantly increases the probability of collaboration but worsens investment outcomes. This supports the thesis that multidimensional similarity promotes cooperation but can generate costs through information redundancy and groupthink.

Network Effects

The study is limited to exogenous attributes (gender, ethnicity, socioeconomic status) that individuals can barely change. Whether the same diminishing returns also apply to endogenous, changeable attributes (political attitudes, musical taste) remains unclear — socialization effects could complicate the picture. Moreover, only homophily in adolescent friendships in a school context was studied; whether the findings transfer to adult friendships remains open. Schaefer (2010), using a different methodology (set-theoretic approach on adult ego-networks), found that people maintain significantly more relationships with those who resemble them on four or five sociodemographic dimensions — though without controlling for network dynamics or opportunity structures.

Practical Significance

The current state of research on this aspect is summarized below.

Frequently Asked Questions

What does multidimensional homophily mean?

People carry many attributes at once — gender, ethnicity, social status, age — and can match others on several of these dimensions simultaneously.

Is there a maximum of similarity?

Three explanations exist for these diminishing returns. First, there may be a utility ceiling: when bridging relationships with dissimilar people provide valuable new resources and ideas (Granovetter, 1973; Burt, 1992), too much similarity becomes redundant.

Which similarities matter most?

The key finding of a longitudinal study across eleven school classes in Scotland, England/Wales, and the USA (Grund & Densley): the well-known homophily effects hold — same gender, same ethnicity, or similar socioeconomic status each increase the probability.

Does too much similarity lead to boredom?

Additionally, Gompers et al. (2016) show that an affinity score — an overall index of similarity-based attributes between venture capital investors — significantly increases the probability of collaboration but worsens investment outcomes.