Dissertation Information for John Duffy Jr.NAME: - John Duffy Jr.
DEGREE:
- Ph.D.
DISCIPLINE:
- Economics
SCHOOL:
- University of California, Los Angeles (USA) (1992)
ADVISORS: - Roger E.A. Farmer
COMMITTEE MEMBERS: - None
MPACT Status: Incomplete - Not_Inspected
Title: Essays on learning rational expectations
Abstract: This dissertation consists of three essays which consider the process by which individuals might learn to form rational expectations forecasts.
The first essay (Chapter 2) examines learning behavior in an overlapping generations model. I consider whether individuals will learn to believe in the locally unique, stationary rational expectations equilibrium of the model or learn to believe in one of the many non-unique, non-stationary rational expectations equilibria of the model. Several researchers have argued that the incorporation of disequilibrium, adaptive learning behavior into rational expectations models may be useful as a selection criterion, allowing the researcher to isolate a locally unique, stationary equilibrium in models with multiple rational expectations equilibria. In this essay, I provide an example of an adaptive learning rule which leads individuals to believe in one of the non-stationary, non-unique equilibria of the model. I conclude that adaptive learning behavior cannot be relied upon to select a locally unique, stationary rational expectations equilibrium.
The second essay (Chapter 3) examines learning behavior in a Cagan-type hyperinflation model with multiple equilibria. Individuals in this model use a slightly misspecified learning rule: they forecast inflation using a model that is underparameterized with respect to the government's policy rule. In this case, simulations reveal that individuals are more likely to learn to believe in one of the many non-unique, hyperinflationary equilibria of the model, and less likely to learn to believe in the locally unique, stationary equilibrium of the model.
The third essay (Chapter 4) is critical of least squares learning models. These models do not take account of the feedback that exists between the learning process and the data generating process. I imagine that individuals do take account of this feedback and choose to learn using a time-varying parameter model instead of a least squares learning model. While both models lead individuals to form rational expectation forecasts, simulations reveal that the time-varying parameter model converges more quickly to rational expectations and allows individuals to cope more effectively with policy regime changes than the least squares model.
| |
MPACT Scores for John Duffy Jr.A = 0
C = 1
A+C = 1
T = 0
G = 0
W = 0
TD = 0
TA = 0
calculated 2010-03-04 08:52:24
Advisors and Advisees Graph
generating graph, please reload |