I am developing a synthetic economy with millions of firms interacting on buyer-seller networks to explain numerous macroeconomic phenomena. A synthetic economy is a system of autonomous agents who make decisions on prices, quantities, and other economic variables. A synthetic economy is useful to study economic problems from a theoretical point of view and as a test-bed for policy questions. The synthetic economy is begin developed over four stages. At each stage the experiments on the synthetic economy will shed light on numerous economic questions. The first in the series of papers is titled "Price effects of monetary shocks in a network economy". The paper is available here. In what follows, I describe each of the four stages. Stage 0 is nearly complete. Python version of the model is available here. C++ version of the model will be available before end of 2017.
Stage 0 An economy with millions of firms interacting on a network. Firms make decisions on prices, quantities, and network weights (the proportions in which to combine inputs). Households supply fixed quantity of labor. Firm productivity and labor are stochastic. Prices and network weights can be sticky, where the degree of stickiness is a parameter. The synthetic economy works with several theoretical network like scale-free and random. It can also take an empirical network as an input into the model. The paper linked above presented some results using the empirical buyer-seller network of Japan. Sage 0 is based on the Gauldi-Mandel model.
Stage 1 Firms enter and exit the network based on economic decisions. The dynamics of the entry and exist of firms produces empirically plausible topology of buyer-seller networks.
Stage 2 Firms make capital investments. Households save. Banks collect savings and make loans. Central bank interacts with the network of banks.
Stage 3 Firms are endogenous. Individuals work in firms, leave one firm to work another firm, and leave firms to form other firms. Stage 3 places Axtell's firm formation model within a buyer-seller network model.