Three essays in dynamic macroeconomics
This thesis presents three papers within the field of dynamic macroeconomics.
The first paper, entitled “Medium-frequency cycles and the remarkable near trend-stationarity of output”, presents a dynamic stochastic general equilibrium model with endogenous growth, capable of reconciling the observed large medium-frequency fluctuations in output, with its long run (near) trend-stationarity. This requires a model in which standard business cycle shocks lead to highly persistent movements around trend, without significantly altering the trend itself. The robustness of the trend also requires that scale effects are eliminated both in the long and short runs. In an estimated version of the model, a financial-type shock to the stock of ideas emerges as the key driver of the medium frequency cycle.
The second paper, entitled “Learning from learners”, is an intervention into two long running debates: the first, on whether learnability may be used to rule out explosive paths for inflation in New Keynesian models, and the second, into whether Taylor rule parameters may be identified from observing the data. We find that in an economy populated with traditional macroeconomic learners, Taylor rule parameters can always be identified by sophisticated econometric techniques. Furthermore, when all agents in the economy use such sophisticated techniques, stationary sunspot solutions are readily learnable, and there is no guarantee of convergence to a stationary solution even in the “determinate” case. This implies that learnability cannot be used for equilibrium selection.
Finally, in the third paper, “Efficient simulation of DSGE models with inequality constraints” (joint with Michael Paetz), we present a new algorithm for the simulation of models subject to inequality constraints, such as the zero lower bound on nominal interest rates. Our algorithm is shown to deliver higher accuracy than all other non-global algorithms, and leading speed. We go on to provide a number of applications of our algorithm.
Supervisor: Simon Wren-Lewis