Top Down

## Overview

Our long-run Platonic ideal would be a general purpose tool in which any well-defined model could be described using a carefully thought out modeling syntax, and then the provided tools would be able to solve and simulate and estimate that model.

A rough approximation of that ideal already exists in some carefully-delimited domains: for models whose dynamics evolution is characterized by small perturbations around a steady-state, the Dynare language can be used to describe and solve a very wide variety of models. For (non-smooth) single agents models, the dolo toolbox also proposes a simple model description syntax based on yaml files. The natural first step toward the more expansive vision would be to think about how these to languages can be extended to rigorously describe models with heterogeneous agents, and how generic algorithms can be used to solve these models.

Presentation

Presentations by Pablo and Michel can be found here.

## Important Questions

• How can users build out a large model incrementally under such a framework? Many large models do not spring fully formed from a users head, but rather are constructed organically as the user works out each aspect slowly.
• Researchers need to be able to "lift the hood" and look at the code driving the model easily.