Main Research Areas of the Research Group

It is nowadays widely acknowledged, at least in the behavioral finance literature, that agent-based models where behavioral heterogenous expectation formation processes are explicitly taken into account (see e. g. Day and Huang 1990, Kirman 1991, De Grauwe et al. 1993, Lux 1995, Brock and Hommes 1998, LeBaron et al. 1999 and Farmer and Joshi 2002), can better explain developments on the financial market than "traditional" financial market models with solely rational agents (for surveys of this literature see inter alia Chiarella et al. 2009, Hommes and Wagener 2009, Lux 2009 or Westerhoff 2009).

Besides an adequate description of the behavior of individual agents, which is often based on empirical evidence about forecast- and investment strategies, the advantage of agent-based models is their possibility to also include in a realistic manner direct interacting effects such as herding behavior within social networks, as well as institutional aspects of real world markets (see e. g. Alfarano and Milakovic, 2009; Alfarano, Milakovic and Raddant, 2013). Moreover, agend-based financial market models are increasingly being used as "artificial laboratories" due to their efficiency, in order to improve the institutional frame conditions of financial markets. By means of agent-based financial market models for example Westerhoff (2003a) investigates for example the effects of transaction taxes, Westerhoff (2003b) the effects of trading interruptions, Wieland and Westerhoff (2005) the effect of central bank interventions, He and Westerhoff (2005) the effect of price controls as well as Thurner et al. (2012) and Anufriev and Tuinstra (2013) the effect of the restriction of leverage transactions and short sales.

A shortcoming of many existing agent-based financial markets models is, however, that they only refer to one speculative market. In other words: market interactions hardly play any part in the previous models. An exception within this field is the model of Schmitt and Westerhoff (2014), where interactions between various European stock markets are investigated. By means of such a model it can be explained for example why speculative bubbles and/or phases of high volatility normally occur simultaneously across markets. Dieci and Westerhoff (2010) develop a model where the stock markets of two countries are connected by the exchange market. In this model crises in one country can have direct consequences for the exchange rate and the stock market of the other country. Another open flank of this approach is that interaction between financial and real markets within this theoretical framework has hardly been investigated so far. Some exceptions, where financial and real markets have been linked, are Proa?o (2011, 2012) and Westerhoff (2012). In these models it is be investigated how real economy alterations are transferred to the financial markets and vice versa. Unfortunately, only limited research efforts in this direction have been made so far. The few existing approaches in this area urgently need to be placed on a broad base. Furthermore, it is necessary to include additional interacting markets in the analysis, such as the real estate market, the crude oil market or credit markets.

The assumed expectation formation process within an economic system is not only of key importance to its dynamics and stability properties (see e. g. Samuelson 1939), but it constraints defines the effect and scope of economic policy. While economic policy influences the real economy under adaptive (backward looking) expectations through "traditional macroeconomic transmission mechanisms", these effects run under the assumption of "rational expectations" mainly through the expectations formation processes of the market participants. Consequently, the main task of economic policy in the latter case is to influence the rational expectation formation process by means of significant and predictable actions, in order to lead the economic system in this way to a certain equilibrium (using technical jargon it means that monetary policy provides for the determinacy of a certain - indeed required -general equilibrium). So the efficiency and even the adequacy of economic policy is based on an extreme and in reality most unlikely situation where a) all agents perfectly understand the "whole model" and its functionality and accordingly form their expectations, and b) the state also perfectly understands the agents' expectation formation process, in order to finally take advantage of this knowledge for its steering funcion.

However, since it is in no way ensured that rational - or better: model-consistent - expectations can indeed be formed endogenously a discrepancy between the true data-generating process, the real expectation formation process of the economic subjects and the agents' expectation formation process assumed by the state can lead to macroeconomic instability, as the learning literature has shown (see e. g. Bullard and Mitra 2002, Evans and Honkapohja 2003a, 2003b as well as lately Bask and Proa?o 2012). In the learning approach  of Evans and Honkapohja (2001) rational expectations are not assumed the applied expectation formation process of the agents but they can be the result of a learning process, where the agents pull information on the real data generating process from the observed dynamics of the economic system and that way make their expectation formation more model-consistent, which at best allows them after a certain time to form rational expectations - under the condition of an adequate conduct of the economic policy. This is, however, not always guaranteed, especially not if the actions of the state and the economic subjects are not consistent with one another.

Against this background we want to investigate within this research program how the design of monetary and fiscal rules shall be performed in such an insecure macroeconomic frame without the compelling assumption of rational expectations. In this regard we want to further develop already existing work of our own research (see Westerhoff 2006, Proa?o 2011, 2012, Bask and Proa?o 2012), as well as to refine new approaches, such as e. g. De Grauwe (2011, 2012), especially in regard to the designs of economic rules which are robust against alternative expectation formation mechanisms.

There is a longlasting debate on the role of empirical analysis in the development of macroeconomic theories. This debate takes place between two parties: the one which promotes the primacy of theory and leaves only a subordinate role for empirical analysis ("theory first") and the one which considers the empirical analysis as the starting point for the constant improvement of an existing theoretical frame ("reality first") (see Juselius, 2009 and Spanos, 2009).

The "theory first“ approach is mainly characterized by the work of Kydland and Prescott (1982) who introduced the calibration method into macroeconomics. However, the aim of the calibration approach is not to deliver an empirically-based specification of theoretical models by a congruent representation of data but instead to bring the simulated characteristics of a theoretical model according to specific momentums (expected values, variances and co-variances) as closely as possible to the empirical momentums of the analyzed variables, without wanting to explain the real development of these variables. This approach is diametrically related to the "reality-first“ approach, which still is mainly present in continental Europe, see e. g. Hendry (1995), B?rdsen et al. (2005) and Juselius (2007). According to the "reality-first" approach the relation between theory and empiricism is of interactive nature with the empiric-econometric validation as a medium for the further development of theoretical models. The failure of former macroeconometric large models of the 60s and 70s is explained by the "reality-first" approach by the fact that at that time statistic characteristics of the investigated variables were not take sufficiently into account. Accordingly, for the "reality-first" approach faith in the constructive role of empirical analysis remains unbroken, as long as the statistic characteristics of the investigated variables in the empirical validation of a theoretical frame are treated adequately.

Against this background the work of all three chairs/professorships involved in this doctoral research program is assigned to the European "reality-first" approach, not only because their research shows a strong empirical motivation but also because they frequently undertook the empirical validation of their theoretical models by means of various parametric and non-parametric econometric methods (see e. g. Proa?o, 2009, 2011, 2012, Manzan and Westerhoff 2007, Franke and Westerhoff 2012). Within the framework of the doctoral research program the methodologic approaches shall be further developed and refined. An example for that would be a systematic combination of the dissequilibrium macrodynamics approach (see e.g. Chiarella et al. 2005) with the cointegrated VAR approach (Juselius, 2007), which models the dynamics of many macro variables as the result of a short-term situation of disequilibrium and its gradually adaption to long-term equilibrium relationships.

Although each respectively functional form of economical regularity of distribution shows a remarkable stability and only variation in parameters of the distributions as well in the lapse of time as in cross section these regularities resp. their explanation approaches are doomed to a shadowy existence within macroeconomics. One of the main reasons for this surely lies within the methodology of the representative agents, because such an approach excludes per definition distributional issues - it makes little sense to consider the "distribution" via one single resp. a few respresentative agents. Aggravating this situation the representative agent unifies in personal union all forms of economic income and thus a priori leaves (little to) no room for distributional conflicts in the sense of classical economics. Worded even stroger the central macroeconomic problem of aggregation will be negated already in advance by the assumption of a representative agent. Thereby the general theory of equilibrium already realized decades ago that aggregation beyond heterogenous agents is burdened with substantial formal problems and is only permitted in extremely pathological cases (see also the initially mentioned works of Gorman, Mantel, Sonnenschein and Debreu).

In contrast, the basic idea of a statistic equilibrium involves the fact that the interaction of heterogenous "parts" (e. g. of molecules in statistical physics or of agents of an open economy) only because of combinatorial considerations leads to statistic distribution regularities, which are mainly independent of the individual characteristic of just those parts (see e. g. Jaynes, 1978; Foley, 1994). Formally the concept of a statistical equilibrium is to be found already for a long time in econometric workbooks on time series analyses (beginning with Box and Jenkins, 1976), where it is defined by the characteristics of ergodicity and stationarity of time series. However, statistical mechanics emphasises that in a statistical equilibrium the cross-sectional distribution of a system is identical with the longitudinal distribution of individual conditions or fates and that the parameters of this distribution remain stable. From a macroeconomic point of view this perspective appears methodologically interesting because it can combine our everyday experiences of a certain macroscopic stability with in parts significantly fluctuating individual fates.

The distribution of income and wealth may here serve as an illustrative example within an economic context: the basic equipment of economic subjects, e. g. in the sense of their "intelligence" or "abilities" are normally distributed and therefore hold a characteristic scale (see e. g. Herrnstein and Murray, 1994), whereas the distribution of income or wealth follows a leptokurtic exponential form or even a scale-free power law (see e. g. Castaldi and Milakovic, 2007; Milakovic, 2003). So it becomes clear that the interaction of agents on markets or of institutions leads to a more unequal distribution than it is (or can be) in the least in initial setups (or educational successes or the like). Following this, the still not sufficiently answered question arises, which structural interactions beyond individual differences are mainly responsible for the observed distribution regularity? Fortunately such questions are once again of current interest, particularly with regard to Piketty's (2013) relevant article to "capital in the 21st century", and not least due to that fact will be investigated intensely within the proposed doctoral research program.