Factor-based forecasting in the presence of outliers: Are factors better selected and estimated by the median than by the mean?

Abstract

Macroeconomic forecasting using factor models estimated by principal components has become a popular research topic with many both theoretical and applied contributions in the literature. In this paper we attempt to address an often neglected issue in these models: The problem of outliers in the data. Most papers take an ad-hoc approach to this problem and simply screen datasets prior to estimation and remove anomalous observations. We investigate whether forecasting performance can be improved by using the original unscreened dataset and replacing principal components with a robust alternative. We propose to use an estimator based on least absolute deviations (LAD) as this alternative and establish a tractable method for computing the estimator. In addition to this we demonstrate the robustness features of the estimator through a number of Monte Carlo simulation studies. Finally, we apply the estimator in a simulated real-time forecasting exercise to test its merits. We use a newly compiled dataset of US macroeconomic series spanning the period 1971:2-2012:10. Our findings suggest that the chosen treatment of outliers does affect forecasting performance and that in many cases improvements can be made using a robust estimator such as the proposed LAD estimator.

Robustness and the general dynamic factor model with infinite-dimensional space: Identification, estimation, and forecasting

“…These results are in concordance with those obtained by Kristensen (2014) and Trucíos et al (2019a) who, in a different but related context, found that the number of principal components (Peason, 1901;Hotelling, 1933) and principal volatility components (Hu and Tsay, 2014;Li et al, 2016) also tend to be over-identified when the series are contaminated by additive outliers. for the impulse-response functions is given by…”

“…In particular, the criterion of Hallin and Liška tends to overestimate the number of common shocks. These results are in agreement, for instance, with those obtained by Kristensen (2014), who finds that the commonly used information criteria of Bai and Ng (2002) (estimating the number of static factors) is severely inflated by outliers. Second, we propose robust procedures for the identification, estimation, and prediction of the GDFM.…”

Robustness and the General Dynamic Factor Model With Infinite-Dimensional Space: Identification, Estimation, and Forecasting

“…In all cases, outliers of size 10 times the standard deviation of the univariate uncontaminated processes were considered. These results are in concordance with those obtained by Kristensen (2014) and Trucíos et al (2019a) who, in a different but related context, found that the number of principal components (Peason, 1901;Hotelling, 1933) and principal volatility components (Hu and Tsay, 2014;Li et al, 2016) also tend to be over-identified when the series are contaminated by additive outliers.…”

“…A method for detecting and estimating the size of outliers in the dynamic factor model is proposed by Baragona et al (2007), based on linear transformations of the observed data. Kristensen (2014) shows that the performance of predictors in static factor models can be improved by replacing principal components with a robust alternative based on least absolute deviations. A similar idea has been investigated previously by Croux and Exterkate (2011).…”

Dynamic factor models

“…The alternative approach of not screening for outliers requires the adoption of robust estimation methods that are not fully developed at least for maximum likelihood estimation. Recently, Kristensen (2014) proposed the use of a robust alternative to principal components, an estimator based on least absolute deviations (LAD) and evaluated its performance in forecasting.…”