On the parameter identifiability problem in Agent Based economical models

TL;DR

This paper examines the fundamental issue of parameter identifiability in agent-based economic models, comparing it with DSGE models and discussing numerical methods for testing identifiability in complex systems.

Contribution

It provides a comparative analysis of structural identifiability in agent-based and DSGE models and discusses numerical protocols for testing global identifiability in economic systems.

Findings

Identifiability is crucial for model validation.

Comparison between DSGE and ABM in terms of parameter identifiability.

Discussion on numerical methods for testing identifiability.

Abstract

Identifiability of parameters is a fundamental prerequisite for model identification. It concerns uniqueness of the model parameters determined from experimental or simulated observations. This dissertation specifically deals with structural or a priori identifiability: whether or not parameters can be identified from a given model structure and experimental measurements. We briefly present the identifiability problem in linear and non linear dynamical model. We compare DSGE and Agent Based model (ABM) in terms of identifiability of the structural parameters and we finally discuss limits and perspective of numerical protocols to test global identifiability in case of ergodic and markovian economical systems.