Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
Using un-truncated settlement size data from eight countries, we show that the "double Pareto lognormal" (DPLN) distribution provides a better fit to actual city sizes than the simple lognormal (LN) distribution. The DPLN has a lognormal body and features a power law in both the lower and the upper tail. It emerges in the steady-state of a stochastic urban growth process with random city formation. Our findings reconcile a recent debate on the Zipfian rank-size rule for city sizes.