Expanding the Technological Frontier of Macroeconomic Modeling

20  January 2020    5:00 pm CET

Michael Miess, Vienna University of Economics and Business (WU Wien)


wiiw, Rahlgasse 3, 1060 Vienna, library (2nd floor)


The potential of agent-based models (ABMs) to complement or even gradually replace existing economic models has been pointed out for more than three decades. However, the success of ABMs in entering the heartland of the economic profession so far has been limited. In this context, it might be useful to view ABMs as a disruptive innovation (or methodology) that still has to assert itself against existing technologies, i.e. the methods and models used by the majority of economists. I will present a novel macroeconomic ABM that surpasses the state of the art in AB modelling along several dimensions, showing that ABMs have matured and are ready to enter the toolbox of economic models for policy analysis and forecasting. This ABM is validated by comparing its forecasting performance to that of more conventional techniques for economic modelling, i.e. simple time series models and a standard DSGE model. The macroeconomic ABM introduces several innovations to the field of AB modelling. Firstly, it achieves a realistic depiction of an actual economy (Austria) by estimation to detailed data from national accounts, input-output tables, government statistics, census data, and business demography data. Furthermore, and also unique in comparable literature, this model incorporates all economic entities of a national economy, i.e. all households (more than 8 million) and firms (more than 600,000) of the Austrian economy are represented by agents within the model. Moreover, several open issues in ABM literature are addressed by adhering closely to these data sources, respecting identities of the system of national accounts, and by adopting simple formulations to structure agents' behavior. In particular, the model proposes solutions to several long-standing and well-known problems of ABMs, including ad-hoc and arbitrary assumptions, comparability between different models, transparency and reproducibility, over-parameterization, estimation and calibration, empirical validation, as well as interpretation and generalization of results. Especially, our ABM approach allows reducing the number of free parameters - i.e. parameters that cannot directly be set according to empirical data and have to be calibrated using techniques that focus on replicating 'stylized facts' of actual economies - to zero. Therefore, this ABM is not subject to the parameter identification problem, which relates to multiple sets of admissible parameters that produce the same model result. Additionally, this ABM is able to abstract from what is known as a transient or 'burn-in' phase in ABM literature, i.e. possibly long periods of time where these models usually exhibit volatile and irregular behavior and which usually are discarded for analysis of model results. Contrary to such a burn-in phase, this ABM offers a stable and clearly interpretable depiction of an actual economy starting from the first quarter of the simulation period. These features turn this ABM into a viable tool for economic forecasting and policy analysis.

Michael Gregor Miess (*1983 in Innsbruck, Austria) is actively researching in the fields of macroeconomics, macroeconomic modelling (agent-based, stock flow consistent, and computable general equilibrium models), financial markets and their link to the real economy, as well as ecological economics. He is currently affiliated with the Institute for Ecological Economics at the Vienna University of Economics and Business (WU Wien), and with the Institute for Advanced Studies (IHS) Vienna. Michael is a graduate in economics from the University of Vienna, and has obtained his PhD in economics at WU Wien in November 2019.
Michael's work for his PhD thesis focused on the development and methodology of empirical agent based models (ABMs), conducted in cooperation the International Institute for Applied Systems Analysis (IIASA) in Laxenburg. He has participated in building an ABM for the Austrian economy incorporating a wide range of macroeconomic data sources (national accounts, input-output tables, government statistics, census data and business surveys). This model is able to outperform standard time series models (ARMA, VAR models) and a standard DSGE model in short- to medium-term forecasting of major macroeconomic variables, and has a large potential for scenario and policy analysis.
Currently, at the Institute for Ecological Economics at WU Wien, Michael is involved in projects funded by the Austrian Central Bank. His focus here is on the development of stock-flow consistent (SFC) macroeconomic models. One aspect of his work with SFC models is to explore harmonized policy responses to mitigate financial-risks related to climate change and thus to smoothing the low-carbon transition. Moreover, Michael co-developed an SFC model that is able to depict credit-fueled financial cycles due to credit creation by the financial sector, causing asset price inflation and influencing investment behavior by firms (financial vs. real investment).
At IHS, Michael has co-developed an innovative hybrid top-down bottom-up computable general equilibrium (CGE) model with focus on energy provision and environmental effects. This model has repeatedly been used for policy-related studies for Austrian ministries. Michael has coordinated and researched for a trans-national project (DEFINE) co-funded by the European Commission that featured the extension and application of this CGE model to analyze a large-scale shift to electromobility in Austria, Germany and Poland.

Powerpoint presentation, as far as available, will be posted on this page after the seminar.