An Overview of the Factor-augmented Error-Correction Model

Abstract

The Factor-augmented Error Correction Model (FECM) generalizes the factoraugmented VAR (FAVAR) and the Error Correction Model (ECM), combining errorcorrection, cointegration and dynamic factor models. It uses a larger set of variables compared to the ECM and incorporates the long-run information lacking from the FAVAR because of the latter's specification in differences. In this paper we review the specification and estimation of the FECM, and illustrate its use for forecasting and structural analysis by means of empirical applications based on Euro Area and US data.

Modelling non-stationary ‘Big Data’

“…The factor-augmented error correction model (FECM) introduced by Banerjee and Marcellino (2009) combines equilibrium-correction, cointegration and dynamic factor models, see Banerjee, Marcellino, and Masten (2016) for a discussion. In the FECM approach cointegration is captured between the variables and the factors.…”

Evaluating restricted common factor models for non-stationary data

“…A general survey of the more popular methods for stationary variables is provided by Stock and Watson (2011). Methods suitable for the empirically important case of non-stationary data, still scarce, are reviewed in Barigozzi, Lippi and Luciani (2016) and Banerjee, Marcellino and Masten (2016). Examples of applications for forecasting and the construction of cyclical indicators are respectively given by, e.g., Giannone, Reichlin and Small (2008) and Altissimo, Cristadoro, Forni, Lippi and Veronese (2010).…”

High-dimensional time series analysis

“…A noteworthy exception is that the time series that are considered I(1) in the FRED-MD classification are now kept in levels, whereas those that are considered as I(2) are differenced once. The methods included in the comparison are: (i) the factor error correction model (FECM) by Banerjee et al (2014bBanerjee et al ( , 2016Banerjee et al ( , 2017, (ii) the non-stationary dynamic factor model (N-DFM) by Barigozzi et al (2017, (iii) the maximumlikelihood procedure (ML) by Johansen (1995a), (iv) the QR-decomposed VECM (QR-VECM) by Liang and Schienle (2019), (v) the penalized maximum-likelihood (PML) by Wilms and Croux (2016), (vi) the single-equation penalized error correction selector (SPECS) and (vii) a factor-augmented SPECS (FASPECS). The latter method is simply the single-equation model derived from the FECM, based on the same principles as the FAPADL from the previous section.…”

“…Recent proposals are brought forward in the literature that allow for application of these techniques on non-stationary and possibly cointegrated datasets. We sequentially discuss the dynamic factor model proposed by Barigozzi et al (2017 and the factor-augmented error correction model by Banerjee et al (2014bBanerjee et al ( , 2016.…”

“…The factor-augmented error correction model of Banerjee et al (2014bBanerjee et al ( , 2016 adopts the estimation strategy proposed by Bai and Ng (2004), which extracts the factors from the data in levels. While the number of factors may still be determined based on the differenced dataset, Bai (2004) proposes a set of information criteria that allows for estimation of the number of non-stationary factors without differencing the data.…”