Forecasting in dynamic factor models using Bayesian model averaging

G.M. Koop, S. Potter

Research output: Contribution to journalArticle

Abstract

This paper considers the problem of forecasting in dynamic factor models using Bayesian model averaging. Theoretical justifications for averaging across models, as opposed to selecting a single model, are given. Practical methods for implementing Bayesian model averaging with factor models are described. These methods involve algorithms which simulate from the space defined by all possible models. We discuss how these simulation algorithms can also be used to select the model with the highest marginal likelihood (or highest value of an information criterion) in an efficient manner. We apply these methods to the problem of forecasting GDP and inflation using quarterly U.S. data on 162 time series. For both GDP and inflation, we find that the models which contain factors do out-forecast an AR(p), but only by a relatively small amount and only at short horizons. We attribute these findings to the presence of structural instability and the fact that lags of dependent variable seem to contain most of the information relevant for forecasting. Relative to the small forecasting gains provided by including factors, the gains provided by using Bayesian model averaging over forecasting methods based on a single model are appreciable.
LanguageEnglish
Pages550-565
Number of pages15
JournalEconometrics Journal
Volume7
Issue number2
DOIs
Publication statusPublished - Dec 2004

Fingerprint

Bayesian model averaging
Dynamic factor model
Factors
Inflation
Marginal likelihood
Information criterion
Simulation
Forecasting method
Model averaging
Structural instability
Justification
Lag

Keywords

  • Bayesian model averaging
  • diffusion index
  • Markov chain
  • Monte Carlo model
  • composition
  • feference prior
  • econometrics

Cite this

@article{feb1c7b1b58543ae8a156ee80103f041,
title = "Forecasting in dynamic factor models using Bayesian model averaging",
abstract = "This paper considers the problem of forecasting in dynamic factor models using Bayesian model averaging. Theoretical justifications for averaging across models, as opposed to selecting a single model, are given. Practical methods for implementing Bayesian model averaging with factor models are described. These methods involve algorithms which simulate from the space defined by all possible models. We discuss how these simulation algorithms can also be used to select the model with the highest marginal likelihood (or highest value of an information criterion) in an efficient manner. We apply these methods to the problem of forecasting GDP and inflation using quarterly U.S. data on 162 time series. For both GDP and inflation, we find that the models which contain factors do out-forecast an AR(p), but only by a relatively small amount and only at short horizons. We attribute these findings to the presence of structural instability and the fact that lags of dependent variable seem to contain most of the information relevant for forecasting. Relative to the small forecasting gains provided by including factors, the gains provided by using Bayesian model averaging over forecasting methods based on a single model are appreciable.",
keywords = "Bayesian model averaging, diffusion index, Markov chain, Monte Carlo model, composition, feference prior, econometrics",
author = "G.M. Koop and S. Potter",
note = "Working paper version",
year = "2004",
month = "12",
doi = "10.1111/j.1368-423X.2004.00143.x",
language = "English",
volume = "7",
pages = "550--565",
journal = "Econometrics Journal",
issn = "1368-4221",
number = "2",

}

Forecasting in dynamic factor models using Bayesian model averaging. / Koop, G.M.; Potter, S.

In: Econometrics Journal, Vol. 7, No. 2, 12.2004, p. 550-565.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Forecasting in dynamic factor models using Bayesian model averaging

AU - Koop, G.M.

AU - Potter, S.

N1 - Working paper version

PY - 2004/12

Y1 - 2004/12

N2 - This paper considers the problem of forecasting in dynamic factor models using Bayesian model averaging. Theoretical justifications for averaging across models, as opposed to selecting a single model, are given. Practical methods for implementing Bayesian model averaging with factor models are described. These methods involve algorithms which simulate from the space defined by all possible models. We discuss how these simulation algorithms can also be used to select the model with the highest marginal likelihood (or highest value of an information criterion) in an efficient manner. We apply these methods to the problem of forecasting GDP and inflation using quarterly U.S. data on 162 time series. For both GDP and inflation, we find that the models which contain factors do out-forecast an AR(p), but only by a relatively small amount and only at short horizons. We attribute these findings to the presence of structural instability and the fact that lags of dependent variable seem to contain most of the information relevant for forecasting. Relative to the small forecasting gains provided by including factors, the gains provided by using Bayesian model averaging over forecasting methods based on a single model are appreciable.

AB - This paper considers the problem of forecasting in dynamic factor models using Bayesian model averaging. Theoretical justifications for averaging across models, as opposed to selecting a single model, are given. Practical methods for implementing Bayesian model averaging with factor models are described. These methods involve algorithms which simulate from the space defined by all possible models. We discuss how these simulation algorithms can also be used to select the model with the highest marginal likelihood (or highest value of an information criterion) in an efficient manner. We apply these methods to the problem of forecasting GDP and inflation using quarterly U.S. data on 162 time series. For both GDP and inflation, we find that the models which contain factors do out-forecast an AR(p), but only by a relatively small amount and only at short horizons. We attribute these findings to the presence of structural instability and the fact that lags of dependent variable seem to contain most of the information relevant for forecasting. Relative to the small forecasting gains provided by including factors, the gains provided by using Bayesian model averaging over forecasting methods based on a single model are appreciable.

KW - Bayesian model averaging

KW - diffusion index

KW - Markov chain

KW - Monte Carlo model

KW - composition

KW - feference prior

KW - econometrics

U2 - 10.1111/j.1368-423X.2004.00143.x

DO - 10.1111/j.1368-423X.2004.00143.x

M3 - Article

VL - 7

SP - 550

EP - 565

JO - Econometrics Journal

T2 - Econometrics Journal

JF - Econometrics Journal

SN - 1368-4221

IS - 2

ER -