Backwash and Spread: Effects of Trade Networks in a Space of Agents who Learn by Doing

This study focuses on the widely observed phenomenon of leading and lagging areas in economic development. Trade may be a source of spread effects, which promote convergence where dichotomization could otherwise occur, or of backwash effects, which could exacerbate or even independently cause dichotomization. Beginning from a cellular model of growth with learning by doing, in which long-lasting dichotomization is observed, trade is introduced in order to investigate the implications of spread and backwash effects. The trade network is initialized as a random Boolean network. Each "agent" shops among the established trading partners for the best deal and trades accordingly, but also may, partly at random, drop or add a trading relationship, so that the trading network becomes endogenous. The probability of dropping a trading partner is inversely related to the time since the last trade, while the probability of adding a new trading partner is inversely related to distance, and hence presumably to transportation cost. Comparative simulations are undertaken to determine, first, whether trade increases or reduces dichotomization relative to the no-trade case, and second, to what extent the answer to this question is sensitive to specific parameters of the model.