Demand-Led Growth In A Multi-Commodity Model With Learning: Some Preliminary Results

The paper represents a preliminary attempt to shed light on the following question: in the context of demand-led growth, how does learning by agents about the economic system's structure and the determinants of long-run growth affect the long-run dynamics of the economy? Analysis is conducted in terms of an extension of the simplified two-sector model with autonomous demands in White (2008). The focus of the analysis is on the impact of learning about two mechanisms in particular: about how the growth of autonomous demand influences growth of the economy as a whole; and about how expectations about growth affect the dynamics of growth. The mechanics of learning are twofold: first, a simple gradient-descent rule, whereby key coefficients in the investment function relating producers expectations about growth to past growth in their own sector and in the economy are modified in a way which aims to minimize forecast errors; and, second, a more ambitious mechanism whereby producers attempt to uncover aspects of the true relation between past growth rates and expected growth rates. Analysis of the system's dynamics is primarily by means of computer simulation.