Hysteresis Effects in Economics – Different Methods for Describing Economic Path-dependence
Relations between economic variables are often characterized by a situation where initial conditions and the past realizations of economic variables matter. I.e. past (transient) exogenous disturbances and past states of the economic system do have an influence on the current economic relations. Typical examples are the dynamics of (un)employment in business cycles and the dynamics of the nexus of exchange rate and exports. Since the standard characteristics of hysteresis apply – i.e. permanent effects of a temporary stimulus, resulting in path-dependent multiple equilibria – these economic phenomena are correctly titled as “hysteresis”. Empirical research in economics is using different methods in order to capture path-dependent effects. First econometric approaches tried to describe these effects by simple timeseries processes with unit- (or zero)-root dynamics. However, since unit-root-dynamics are not related to genuine multiple equilibria but on the order of integration of the time series, these first attempts were expanded by more sophisticated time-series models integrating structural breaks, threshold-cointegration or non-linear autoregressive distributed lag-models. Another branch of empirical studies tries to keep closer to the original concept of the macroloop, trying to apply an explicit Mayergoyz/ Preisach aggregation procedure for heterogeneous firms – if microeconomic information is available based on panel-data – or by using simple algorithms analogous to mechanical-play in order to apply simple OLS-regression methods on a filtered/transformed input-output relation. In this paper, we give an overview of the implementation of hysteresis in economics, with an emphasis on two aspects: (1) the differentiation between micro- and macroeconomic hysteresis including an outline of an adequate aggregation procedure, and (2) different methods applied in econometrics in order to capture economic path-dependency empirically.
