Richard J. Smith, University of Cambridge
Jaap Abbring, Tilburg University
Massachusetts Institute of Technology
Michael Jansson, Berkeley
Andrew Patton, Duke University
This workshop is a joint initiative between The Econometrics Journal and Cambridge-INET and will take place in the Cambridge City Hotel on ... Read more
The Econometrics Journal has a new Replication Policy for accepted papers
To benefit readers of accepted papers a new facility for supporting material linked to the online published manuscript is available.
Special Issue 2015
The papers in this Special Issue on Econometrics of Forecasting arise out of the invited presentations given in The Econometrics Journal Special Session on this topic at the Royal Economic Society Annual Conference held 26-28 March 2012 at the University of Surrey. The organization of Special Sessions on subjects of current interest and importance at Royal Economic Society Annual Conferences is an initiative of the Editorial Board of The Econometrics Journal to enhance further the profile and reputation of the journal. The Editorial Board is responsible for the choice of topic and organization of the Special Session. The intention is by judicious choice of topics and speakers to encourage further a higher standard of submissions to The Econometrics Journal. It is hoped that the specific topics discussed in these papers although necessarily restrictive in scope provide an indication of the current frontiers of the econometrics of forecasting.
The paper by Borus Jungbacker and Siem Jan Koopman reconsiders likelihood-based analysis of the dynamic factor model. The latent factors are modeled as linear dynamic stochastic processes with the idiosyncratic components linear autoregressive processes. The particular focus is a high-dimensional panel with a relatively small number of common dynamic factors. A common approach to estimation is quasi-maximum likelihood treating the idiosyncratic components and the common factors as if they are Gaussian and exploiting the Kalman filter for evaluation of the quasi-likelihood. Many recent applications of the dynamic factor model concern a high-dimensional panel of time series with a consequential large number of parameters which make such an approach infeasible. The key insight of this paper is that the observed time series can be split into a low dimensional and a high-dimensional vector series with the computationally intensive Kalman filter applied to the former with simple regression-style calculations for the latter yielding large computational gains. The resulting methods allow real-time estimation of the dynamic factors, the estimation of the past factors as well as the prediction of the factors and observations with the Kalman filter providing mean squared errors of the factor estimates.
The paper authored by Raffaella Giacomini addresses a number of important questions for economic forecasters. In particular can economic theory play a helpful role in formulating accurate forecasts? Similar fundamental questions have frequently been posed and have inspired serious debate in the profession affecting forecasting practice for many years. Most recently partly in response to the 2007 crisis the usefulness of estimated DSGE models has been called into question leading to a preference for richer and larger models able to account for aspects necessarily excluded by these models and possibly indicating a repetition of similar past debates. The paper is a summary of a few key contributions rather than a comprehensive review of the literature with its main focus on econometric methodology and some discussion of empirical findings. Three themes form the basis of the paper. The first stresses the importance of using appropriate econometric tools. A number of examples are presented of theory-based forecasting that have not proven useful, e.g., theory-driven variable selection and some popular DSGE models. The paper then discusses various forms of theoretical restrictions that have shown some usefulness in forecasting, e.g., accounting identities, disaggregation and spatial restrictions, and cointegrating relationships. The paper is concluded by the suggestion that economic theory might help to deal with the widespread instability that affects forecasting performance by acting as a guide in the search for stable relationships thereby yielding potentially more accurate forecasts.
To access these Special Issue papers please click on the links below:
Likelihood-based dynamic factor analysis for measurement and forecasting, by Borus Jungbacker and Siem Jan Koopman.
Economic theory and forecasting: lessons from the literature, by Raffaella Giacomini
By Peter C. B. Phillips. Read More
The 2013 journal impact factor is 1.128. This represents an increase from 1.000 in 2012. Read more.
Permission information on reproducing Journal content.