A Forward Projection of the Cross-Country Income Distribution

This paper proposes and implements a method to predict evolution of the crosscountry income distribution from a nonconvex growth model with unbounded productivity shocks, fitted to panel data by threshold autoregresion. We estimate the stochastic kernel of the process, and define inducively all future distributions as a norm-convergent sequence in the function space L1. Elements of the sequence are calculated using Monte Carlo simulation. Our results suggest that nonlinearities in the growth process are responsible for emerging bimodality in the distribution of income, but that such bimodality eventually peaks and declines. In the long run we predict convergence.