In this paper we integrate heterogeneous inflation expectations into a simple monetary
model. Guided by empirical evidence we assume that boundedly rational agents, selecting
between extrapolative and regressive forecasting rules to predict the future inflation rate,
prefer rules that have produced low prediction errors in the past. We show that integrating
this behavioral expectation formation process into the monetary model leads to the
possibility of endogenous macroeconomic dynamics. For instance, our model replicates
certain empirical regularities such as irregular growth cycles or inflation persistence.
Moreover, we observe multi-stability via a Chenciner bifurcation.
We develop a novel financial market model in which the stock markets of two countries
are linked via and with the foreign exchange market. To be precise, there are domestic
and foreign speculators in each of the two stock markets which rely either on linear
technical or linear fundamental trading strategies to determine their orders. Since
foreign stock market speculators require foreign currency to conduct their trades, all
three markets are connected. Our setup entails a natural nonlinearity which may cause
persistent endogenous price dynamics. Moreover, we analytically show that market
interactions can destabilize the model’s fundamental steady state.
In the framework of small-scale agent-based financial market models, the paper starts
out from the concept of structural stochastic volatility, which derives from different noise
levels in the demand of fundamentalists and chartists and the time-varying market shares
of the two groups. It advances several different specifications of the endogenous switching
between the trading strategies and then estimates these models by the method of simulated
moments (MSM), where the choice of the moments reflects the basic stylized facts
of the daily returns of a stock market index. In addition to the standard version of MSM
with a quadratic loss function, we also take into account how often a great number of
Monte Carlo simulation runs happen to yield moments that are all contained within their
empirical confidence intervals. The model contest along these lines reveals a strong role
for a (tamed) herding component. The quantitative performance of the winner model is
so good that it may provide a standard for future research.