Can gender inequality be created without inter-group discrimination?

Abstract

Understanding human societies requires knowing how they develop gender hierarchies, which are ubiquitous. We test whether a simple agent-based dynamic process could create gender inequality. Relying on evidence of gendered status concerns, self-construals, and cognitive habits, our model included a gender difference in how responsive male-like and female-like agents are to others' opinions about the level of esteem for someone. We simulate a population who interact in pairs of randomly selected agents to influence each other about their esteem judgments of self and others. Half the agents are more influenced by their relative status rank during the interaction than the others. Without prejudice, stereotypes, segregation, or categorization, our model produces inter-group inequality of self-esteem and status that is stable, consensual, and exhibits characteristics of glass ceiling effects. Outcomes are not affected by relative group size. We discuss implications for group orientation to dominance and individuals' motivations to exchange.

Fig 1. On the left, perceived dissimilarity with someone thinking that it is important to listen to people however, for different sexes and in different countries.

On the right, perceived dissimilarity with someone thinking that it is important to treat people equally. Source: ESS8-2016 data, ed.1.0.

Fig 2. Examples of credibilities (i.e. strength of the influence) given by an agent i to an agent interlocutor j depending on the absolute perceived distance i between j and him-herself (on the x-axis), and the fact i hold j in lower esteem (plain lines) or higher esteem (dotted lines), for σ = 0.1 (in blue), 0.5 (in orange), 1 (in green) and 3 (in red).

We observe that the higher the perceived distance, the lower the credibility for j when held in lower esteem, or the higher credibility for j when held in higher esteem. Moreover, the larger σ, the smaller the difference of given credibility to j if he/she is held in low compared to high esteem.

Fig 3

Upper plot: average group reputation for different values of σ_S. Lower left plot: average self-esteem of the groups for different values of σ_S. Lower right plot: Intra-group variability—variation coefficient for different values of σ_S. The black baseline represents a single group with open–mindedness parameter σ_S. In the other cases the population is divided into two equal-size groups with open–mindedness parameter σ_S (blue lines) and σ_L = f(σ_S) (magenta lines). The empty symbols correspond to the function: σ_L = σ_S + 0.1 the others to the case σ_L = 1.5σ_S.

Fig 4. Group S (in magenta) is less open-minded than Group L (in blue), and then occupies on average higher ranks (1 is the best rank).

Plot A: Boxplot of status rank values for different populations having different values of the open–mindedness parameter σ_S for the group S, and the corresponding value of (σ_L = 1.5 x σ_S) for the group L. Dashed lines indicate group means for the lowest σ_S 0.05, across levels of σ_S: we observe the difference of average ranking for group S and group L increases with the value of σ_S. Plot B: Probability for a rank to be occupied by an agent in a group at each level of the open-mindedness parameters (see legend, still with σ_L = 1.5 x σ_S). Plot C: Probability of moving to a lower status position (higher in rank) given the agent’s initial rank, by group and open-mindedness parameter. The three plots represent averages computed from 100 replicas of each tested parameter set.

Fig 5. Z-scores, indicating how different from the expected probability from the fraction of L in the population, is the probability of the more open-minded Group L being in the 5 most prestigious ranks (green points, lower than the expected z-score 0) and of being in the 5 least prestigious ranks (orange points, higher from the expected z-score 0) plotted by their proportion of the population of agents.

Each small point represents a simulation setup replicated 50 times, varying the combined values of (σ_S, (σ_L, k)—see the method section for details.

Fig 6. Credibility given to an interlocutor for whom our esteem varies around our self-esteem by 0.1 for different values of the open-mindedness σ; for a variation of: -0.1 implying the interlocutor is held in lower esteem (red square), or +0.1 implying the interlocutor is held is higher esteem (blue triangle).