Date of Award

1-1-2018

Language

English

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

College/School/Department

Department of Economics

Content Description

1 online resource (ii, xi, 156 pages) : illustrations (some color)

Dissertation/Thesis Chair

Kajal Lahiri

Committee Members

Zhongwen Liang, Koomla Ulrich Hounyo, Recai M Yucel

Keywords

asymmetry, density forecast, entropy, generalized beta distribution, Jensen-Shannon information, loss function, Economic forecasting, Entropy (Information theory), Uncertainty, Information asymmetry

Subject Categories

Economics

Abstract

In this thesis I estimate the asymmetry of forecasters’ loss functions and uncertainty of forecasts. Continuous distributions are fitted into histogram density forecasts. Of the two candidates of distributions used for the fitting process, generalized beta / triangular distribution is preferred to the normal distribution for its better fit and flexibility in accommodating excess skewness. Time varying loss function asymmetry is obtained by combining the fitted density distribution and the point forecasts of the same forecaster. Utilizing generalized beta distribution and triangular distribution, I find that forecasters tend to provide more favorable forecasts in their point forecasts than in their density forecasts for the same target variable. Forecasters systematically make more optimistic real point forecasts than the central tendency of underlying density forecast in good times and more pessimistic point forecasts than the density median in bad times. The level of asymmetry is significantly correlated with the level of the target variable. Time invariant asymmetry parameter is also obtained by two alternative methods, GMM proposed by Elliott et al (2005) and “combining density and the point forecasts” proposed by Lahiri and Liu (2008). The two methods produce distinct results. We estimate forecast uncertainty and disagreement using information framework, and compare these with moment- based estimates. We find these two sets of measures are largely analogous, except in cases where the underlying densities deviate significantly from normality. Even though the Shannon entropy is more inclusive of different facets of a forecast density, we find that with SPF forecasts it is largely driven by the variance of the densities. Using standard VAR analysis, we find information-based uncertainty affects the economy negatively. Jenson-Shannon Information is used to measure ex ante “news” or “uncertainty shocks” in real time. We find that this ‘news’ is countercyclical, closely related to revisions of forecast means and increases both output and inflation uncertainty.

Download

Included in

Economics Commons

Share

COinS