Alexandra Rottenkolber

Interactive · Agent-based model

Why can workplace segregation make
occupations less segregated?

Imagine two different worlds. The workers, their preferences, and the occupations are identical — the only difference is whether workplaces within each occupation are gender-mixed or gender-sorted. Below two simulated labour markets run side by side. Watch how occupational segregation changes. That gap is the shielding effect.

How strongly workers prefer same-gender colleagues
η > 1: workplace matters more · η < 1: occupation matters more
Iterations per second
Iteration 0 / 180

Integrated workplaces

Within each occupation, workplaces are gender-mixed  (Dʷ ≈ 0.1)

Occupational segregation Dᵒ 0.00

Segregated workplaces

Within each occupation, workplaces are gender-sorted  (Dʷ ≈ 0.9)

Occupational segregation Dᵒ 0.00

Occupational segregation over time

Dissimilarity index of the two worlds. The shaded band is the shielding effect.

Integrated workplaces Segregated workplaces Shielding effect (the gap) Women Men
Press Play and watch the two curves separate.

The intuition

Workers tend to prefer colleagues of their own gender. This is what is called gender homophily. When someone becomes a minority in their occupation, that pressure can push them to leave — which makes the occupation even more lopsided, and so on.

But people don't experience an occupation in the abstract: they experience their own workplace. If that workplace still has plenty of same-gender colleagues, the worker is shielded from the wider imbalance at the occupational level and stays put.

Why the surprise

Counter-intuitively, this means workplace gender-sorting can dampen occupational segregation. Sorted workplaces give minority workers local pockets where their preferences are met, so they don't exit the occupation entirely.

The effect is structurally asymmetric: it works mainly by keeping occupational minorities in place, with little offsetting effect on the majority — so it doesn't cancel out in the aggregate.

A simplified, in-browser illustration of the agent-based model in Rottenkolber, Arvidsson, Brandén & Takács — "The Shielding Effect: How workplace segregation can countervail occupational segregation." Agents choose an occupation×workplace combination by multinomial logit over the paper's utility function (SI‑1), with inertia coefficients β_SW=5, β_SO=2, β_SWO=1.5. Population and grid are scaled down for live animation; the qualitative dynamics match.