A Simple Agent-based Macroeconomic Model
This simple model is built by students taking the MSc Computational Economics, Financial Markets and Policy course at the University of Essex. This is a lab-based exercise in EC910, based on Lengnick (2013).
The only economic agents this model consists of are households and firms, both of which are described through simple, adaptive rules of behavior. Although it is simple, the model can already imitate many empirical facts: from the interaction, reasonable level of unemployment, the cyclical fluctuations in the total production level as experienced rules like the Phillips or Beveridge curves, right-skewed firm size distribution, median price change frequency between 9% and 12%, dynamic correlation between output and inflation, long-term neutrality and short-term non-neutrality of currency. He has shown that one can change the business cycle forever and only need to reduce demand by 5% within one day. Finally, he proves that equilibrium is a strong result of the ACE economy in all markets. If the price or salary adjustment is completely closed, it will crash.
The model is built using the object-oriented programming language Java utilising the external library JAS, which simplifies the programming needed to set up multi-agent economic models as well as dealing with big data and visualization. The model consists of 4 classes:
- The model class is a superclass within the macro model, it is responsible for building the model and the events that are run over the course of each month.
- The observer class within the macro model outputs tools to help track how variables change over time.
- The firm class establishes the parameters for each firm and returns and sets their values within Model.
- The household class establishes the parameters for each household and returns and sets their values within Model.
The study of the economy by means of ACME and network analysis is a relatively new field. It also represents a suitable approach to respond to the criticisms raised on the methodological foundations of the macroeconomic theory. Because of the failure of the existing macroeconomic models in predicting the great recession of 2008-09 and evaluating the consequences of such a recession, these models and their usage in policy analysis have received severe criticism (Wieland (2010)). Macroeconomists have been accused of a heavy dependence on dynamic stochastic general equilibrium (DSGE) models that are built around special cases where market inefficiencies are not possible (Stiglitz (2011)) and institutional details and financial interconnections in the provision of liquidity, capital adequacy and solvency are ignored (Markose (2013)).
Buiter (2009) points out that “… the typical graduate macroeconomics and monetary economics training received at Anglo-American universities during the past 30 years or so, may have set back by decades of serious investigations of aggregate economic behaviour and economic policy-relevant understanding”. This view is supported by Krugman -in the Economist, June 2010- who indicates that “most work in macro-economics in the past 30 years has been useless at best and harmful at worst”. In particular, critics of the standard macro models have used the aggregation of individual economic units, the perfect rationality of these units, and the assumption of equilibrium as a ground to attack these models (Lengnick (2011)).
First, the majority of macro models link the macro movements directly to the individual units’ behaviours -either through equating the aggregates to the representative units or by adding up the individual decisions to find the aggregates- to provide a proper microfoundation. However, several experiments have indicated that the aggregate behaviour of big groups usually differs considerably from the behaviours of the individual units. For instance, Schelling’s (1969) analysis of racial segregation models points out that interaction between individuals may create significant segregation in big cities even if individual preferences for residing in areas dominated by people of the same race are slight. More recently, in the context of the 2007 financial crisis, many authors have noted the pitfall of the macroeconomic models where extrapolation of the behaviour of the representative optimizing agents can result in fallacy of composition. Specifically, with microprudential policies where the risk is specified at the level of individual units and the implications of their interaction with each other are ignored, system-wide risks and instabilities are not modelled or managed (For example, Markose (2013) and Goodhart et al (2009)). Moreover, the assumption of perfect rational utility-maximizing agents has proven to be very unrealistic. Rather than complex utility optimization approach which requires everyone to have perfect information, individuals tend to use relatively simple behavioural rules to make decisions (Akerlof (2002)).
Lastly, most of macro models are built around the assumption of a stable state once reached there will be no incentive for any further changes (i.e. equilibrium), and if the economy for some reason deviates for that state, it returns to it through quick adjustment processes. Yet, it has been frequently proven that such adjustment processes barely exist (Gaffeo et al. (2008), Kirman (2006), Ackerman (2002)) and that real markets are often characterized by multiple equilibria, volatility, and coordination problems (Arthur (2006)).