Choosing between Different Time-Varying Volatility Models for Structural Vector Autoregressive Analysis

Abstract

Standard-Nutzungsbedingungen:Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden.Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen.Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in AbstractThe performance of information criteria and tests for residual heteroskedasticity for choosing between different models for time-varying volatility in the context of structural vector autoregressive analysis is investigated. Although it can be difficult to find the true volatility model with the selection criteria, using them is recommended because they can reduce the mean squared error of impulse response estimates substantially relative to a model that is chosen arbitrarily based on the personal preferences of a researcher. Heteroskedasticity tests are found to be useful tools for deciding whether time-varying volatility is present but do not discriminate well between different types of volatility changes. The selection methods are illustrated by specifying a model for the global market for crude oil.

Heteroskedastic Proxy Vector Autoregressions

“…Following standard conventions, we normalize the variances of the structural innovations in the first state to unity (Λ 1 = I 3 ). In line with comparable studies (Lütkepohl and Schlaak, 2018), we set the structural innovations' relative variances for the states m = 2, 3 such that a noticeable volatility pattern is generated. To achieve this, we choose Λ 2 = diag(4, 9, 12) and Λ 3 = diag(1, 4, 9).…”

Statistical Identification in Svars - Monte Carlo Experiments and a Comparative Assessment of the Role of Economic Uncertainties for the US Business Cycle

“…The numerical maximization of the log-likelihood is conditional on T sb . For the selection of a suitable and, subsequently, select the model with the highest log-likelihood as in Lütkepohl and Schlaak (2018). In this study we use the latter approach to obtain estimates denoted B UH .…”

“…Conditional on sample information {û t } T t=1 , we estimate the structural matrix B by means of the alternative procedures described in Section 2 to obtain a set of estimators B • , • ∈ {SR, UH, CH, nGML, dCov, CvM}. Similar to Lütkepohl and Schlaak (2018), we fit all identification schemes to all series regardless of the underlying DGP. Hence, we mimic the simplified case of an analyst who has no further knowledge about the properties of the data.…”

“…The literature on the comparative performance of statistical identification schemes is still scant. Lütkepohl and Netšunajev (2017a) review alternative heteroskedasticity-based models by pointing out their advantages and drawbacks, while Lütkepohl and Schlaak (2018) provide guidelines on how to choose between these models conditional on distinct forms of underlying covariance changes. Herwartz and Plödt (2016a) compare stylized sign restrictions and identification by means of covariance shifts.…”

“…Overall, the constraints are jointly maximized byε CH followed byε dCov . 20 20Lütkepohl and Schlaak (2018) argue that the AIC is an appropriate criterion to choose between alternative heteroskedasticity-based SVARs. In our case we obtain for UH and CH the diagnostic outcomes…”