Using Swedish micro-level data, we trace the wage developments of annual cohorts in each city and, for each year of their early careers, we estimate the between-city scaling β among median earners (red markers) and tail earners (blue baseline). Differences in β for tail earners and median earners are highly significant at age 30 ( z = 4.432 and P < 0.001, two-sided test using Bonferroni correction for multiple comparisons) and at age 40 ( z = 7.839 and P < 0.001). Zeroing out the size-dependent cumulative advantage (cum. adv.) effect (blue squares) reduces tail differences by 33% ( z = 3.788 and P < 0.001). Once we factor in migration (see text), β increases by 33% ( z = 3.559 and P < 0.001). The inset shows that our estimated equivalents to the real-world wage trajectories (empty circles) reproduce the observed scaling exponents (full circles) by cohort age.
