Some finished research projects

On this page members of the Chair of Macroeconomics present to you some finished research projects.

Applications of agent-based models to policy

In an empirical agent-based model we study the migration behavior of a generation of Mexican migrants. In particular, the role of networks is analyzed and quantified, and it is taken into consideration that many migrants move back and forth between Mexico and the US over the years. Distributions of migrants across cities and of the number of trips a migrant takes are determined in both the empirical sample and in the simulation. We use the model for policy analyses, to see for example the effect an increase in Mexican wages would have on the number of migrants in the US, or whether tightening immigration policy increases or decreases the number of migrants within the country.

The model is an example of an agent-based model whose parameters are completely empirically determined and which could be used directly for policy advice.

In a second project we analyze the impact of investors’ behavior in a stylized business angel setting. In particular, we study the role of trust and distinguish those kinds of decisions that have to be based on trust from those that make it possible to rely entirely on rational profit-maximization. In a series of papers we study the properties of a market of investors and entrepreneurs where the initial investment decision is based on trust, but where trust is updated taking into consideration only received returns. Both investors’ decisions (collectively and from an individual perspective) as well as entrepreneurs’ reactions to investor behavior are related to market outcomes. An important next step will be to determine how such a market can be made more efficient by distinguishing unproductive and productive entrepreneurs early on.

Project participants: Anna Klabunde

Heterogeneity of price-setting and consequences for monetary policy: Experimental evidence

In modern monetary macroeconomic models monetary policy has an effect on output because of the New Keynesian Phillips curve (NKPC). In contrast to the Phillips curve older macro models the NKPC is derived by modelling price-setting frictions at the firm level. Because empirical evidence is not in line with theoretical predictions, we tested the two most popular models in laboratory experiments.

In a first study we focussed on the model by Calvo (1983). In this model firms cannot reset their price in every period but receive a random signal with a known probability when they can reset their price. Otherwise their price keeps unchanged. Rational agents in this context anticipate this friction and put a weight close to unity on future expected inflation. However, empirical estimations of the New Keynesian Phillips curve support a-theoretical versions with a positive weight on lagged inflation and a weight less than one on expected inflation.

We argue that myopic price-setting of some agents explains the low weight on expected inflation. The lagged term can be explained by trend extrapolation if information about the future is costly. In a laboratory experiment we implement the Calvo (1983) microfoundations of the Phillips curve. Our hypotheses are supported by the experimental data. About half of the subjects set optimal Calvo prices while about a third is myopic. The Phillips curves we estimate with our data are in line with empirical studies.

In a second study we examine the quadratic price adjustment model by Rotemberg (1982). Firms can reset their prices in every period but face costs when they change their price over time. We also implement this model in a lab experiment.

By constructing categories and a quantitative measure that compare price-setting with optimum we find heterogeneous price‐setting behaviour by degree of information acquired about the future. Subjects rarely use past information, but overweight their own past price. We study the impact of heterogeneous price‐setting behaviour on estimated theoretically derived and hybrid Phillips curves. We find support for features of both NKPCs in our findings at the micro level. But the hybrid NKPC has a superior fit.

Project participants: Andreas Orland