Structural VARs, deterministic and stochastic trends: how much detrending matters for shock identification

Varang Wiriyawit, Benjamin Wong

Research output: Contribution to journalArticleResearchpeer-review

Abstract

Detrending within structural vector autoregressions (SVAR) is directly linked to the shock identification. We investigate the consequences of trend misspecification in an SVAR using both standard real business cycle models and bi-variate SVARs as data generating processes. Our bias decomposition reveals biases arising directly from trend misspecification are not trivial when compared to other widely studied misspecifications. Misspecifying the trend also distorts impulse response functions of even the correctly detrended variable within the SVAR system. Pretesting for unit roots mitigates trend misspecification to some extent. We also find that while practitioners can specify high lag order VARs to mitigate trend misspecification, relying on common information criterion such as the Akaike information criterion (AIC) or Bayesian information criterion (BIC) may choose a lag order that is too low.

Original languageEnglish
Pages (from-to)141-157
Number of pages17
JournalStudies in Nonlinear Dynamics and Econometrics
Volume20
Issue number2
DOIs
Publication statusPublished - 1 Apr 2016
Externally publishedYes

Keywords

  • bias
  • detrending
  • identification
  • structural VAR

Cite this

@article{501b1589b29c452cb94a43c3222f0d05,
title = "Structural VARs, deterministic and stochastic trends: how much detrending matters for shock identification",
abstract = "Detrending within structural vector autoregressions (SVAR) is directly linked to the shock identification. We investigate the consequences of trend misspecification in an SVAR using both standard real business cycle models and bi-variate SVARs as data generating processes. Our bias decomposition reveals biases arising directly from trend misspecification are not trivial when compared to other widely studied misspecifications. Misspecifying the trend also distorts impulse response functions of even the correctly detrended variable within the SVAR system. Pretesting for unit roots mitigates trend misspecification to some extent. We also find that while practitioners can specify high lag order VARs to mitigate trend misspecification, relying on common information criterion such as the Akaike information criterion (AIC) or Bayesian information criterion (BIC) may choose a lag order that is too low.",
keywords = "bias, detrending, identification, structural VAR",
author = "Varang Wiriyawit and Benjamin Wong",
year = "2016",
month = "4",
day = "1",
doi = "10.1515/snde-2015-0030",
language = "English",
volume = "20",
pages = "141--157",
journal = "Studies in Nonlinear Dynamics and Econometrics",
issn = "1081-1826",
number = "2",

}

Structural VARs, deterministic and stochastic trends : how much detrending matters for shock identification. / Wiriyawit, Varang; Wong, Benjamin.

In: Studies in Nonlinear Dynamics and Econometrics, Vol. 20, No. 2, 01.04.2016, p. 141-157.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Structural VARs, deterministic and stochastic trends

T2 - how much detrending matters for shock identification

AU - Wiriyawit, Varang

AU - Wong, Benjamin

PY - 2016/4/1

Y1 - 2016/4/1

N2 - Detrending within structural vector autoregressions (SVAR) is directly linked to the shock identification. We investigate the consequences of trend misspecification in an SVAR using both standard real business cycle models and bi-variate SVARs as data generating processes. Our bias decomposition reveals biases arising directly from trend misspecification are not trivial when compared to other widely studied misspecifications. Misspecifying the trend also distorts impulse response functions of even the correctly detrended variable within the SVAR system. Pretesting for unit roots mitigates trend misspecification to some extent. We also find that while practitioners can specify high lag order VARs to mitigate trend misspecification, relying on common information criterion such as the Akaike information criterion (AIC) or Bayesian information criterion (BIC) may choose a lag order that is too low.

AB - Detrending within structural vector autoregressions (SVAR) is directly linked to the shock identification. We investigate the consequences of trend misspecification in an SVAR using both standard real business cycle models and bi-variate SVARs as data generating processes. Our bias decomposition reveals biases arising directly from trend misspecification are not trivial when compared to other widely studied misspecifications. Misspecifying the trend also distorts impulse response functions of even the correctly detrended variable within the SVAR system. Pretesting for unit roots mitigates trend misspecification to some extent. We also find that while practitioners can specify high lag order VARs to mitigate trend misspecification, relying on common information criterion such as the Akaike information criterion (AIC) or Bayesian information criterion (BIC) may choose a lag order that is too low.

KW - bias

KW - detrending

KW - identification

KW - structural VAR

UR - http://www.scopus.com/inward/record.url?scp=84964721821&partnerID=8YFLogxK

U2 - 10.1515/snde-2015-0030

DO - 10.1515/snde-2015-0030

M3 - Article

VL - 20

SP - 141

EP - 157

JO - Studies in Nonlinear Dynamics and Econometrics

JF - Studies in Nonlinear Dynamics and Econometrics

SN - 1081-1826

IS - 2

ER -