Each step is one generation. Lines connect a family's rung with its child's rung (300 families shown).
Each family has a standardised log-income position z (mean 0, sd 1). The child's position is an AR(1) step:
zchild = ρeff zparent + √(1-ρeff2) ξ, ξ ~ N(0,1), which keeps the spread of the population constant. ρeff = ρ (1 - 0.75 a): inherited advantage ρ (wealth, networks, schools) is reduced by open-access policy a (the 0.75 factor is illustrative).
Quintiles (fifths of the population) are the rungs. The exact transition probability is P(j|i) = 5 ∫quintile i φ(z) [Φ((cj+1-ρz)/s) - Φ((cj-ρz)/s)] dz, with s = √(1-ρ2) and edges c = Φ-1(i/5). The check panel compares this Simpson-rule integral with 200,000 simulated parent-child pairs.
With income = exp(σ z) (log-normal) the Gini coefficient is G = 2Φ(σ/√2) - 1; the check also compares this with the Gini of the sampled families.
Simplifications: a single Gaussian persistence parameter; real mobility depends on geography, race, gender, migration and shocks. The 0.75 "opportunity effect" is made up for illustration - real estimates of income persistence vary by country and measurement, so treat the presets as stylised (typical published rank-rank slopes are roughly 0.2-0.5, flagged as a typical range, not a lookup).