A multifactor approach to forecasting Romanian gross domestic product (GDP) in the short run

PLOS ONE, Dec 2019

The purpose of this paper is to investigate the application of a generalized dynamic factor model (GDFM) based on dynamic principal components analysis to forecasting short-term economic growth in Romania. We have used a generalized principal components approach to estimate a dynamic model based on a dataset comprising 86 economic and non-economic variables that are linked to economic output. The model exploits the dynamic correlations between these variables and uses three common components that account for roughly 72% of the information contained in the original space. We show that it is possible to generate reliable forecasts of quarterly real gross domestic product (GDP) using just the common components while also assessing the contribution of the individual variables to the dynamics of real GDP. In order to assess the relative performance of the GDFM to standard models based on principal components analysis, we have also estimated two Stock-Watson (SW) models that were used to perform the same out-of-sample forecasts as the GDFM. The results indicate significantly better performance of the GDFM compared with the competing SW models, which empirically confirms our expectations that the GDFM produces more accurate forecasts when dealing with large datasets.

A PDF file should load here. If you do not see its contents the file may be temporarily unavailable at the journal website or you do not have a PDF plug-in installed and enabled in your browser.

Alternatively, you can download the file locally and open with any standalone PDF reader:

https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0181379&type=printable

A multifactor approach to forecasting Romanian gross domestic product (GDP) in the short run

July A multifactor approach to forecasting Romanian gross domestic product (GDP) in the short run Daniel Armeanu 0 1 Jean Vasile Andrei 1 Leonard Lache 0 1 Mirela Panait 1 0 Department of Finance, The Bucharest University of Economic Studies , Bucharest , Romania , 2 Business Administration Department, Petroleum-Gas University of Ploiesti , Ploiesti, Prahova, Romania, 3 Cybernetics, Economic Informatics , Finance and Accounting Department, Petroleum-Gas University of Ploiesti , Ploiesti, Prahova , Romania 1 Editor: Wei-Xing Zhou, East China University of Science and Technology , CHINA The purpose of this paper is to investigate the application of a generalized dynamic factor model (GDFM) based on dynamic principal components analysis to forecasting short-term economic growth in Romania. We have used a generalized principal components approach to estimate a dynamic model based on a dataset comprising 86 economic and non-economic variables that are linked to economic output. The model exploits the dynamic correlations between these variables and uses three common components that account for roughly 72% of the information contained in the original space. We show that it is possible to generate reliable forecasts of quarterly real gross domestic product (GDP) using just the common components while also assessing the contribution of the individual variables to the dynamics of real GDP. In order to assess the relative performance of the GDFM to standard models based on principal components analysis, we have also estimated two Stock-Watson (SW) models that were used to perform the same out-of-sample forecasts as the GDFM. The results indicate significantly better performance of the GDFM compared with the competing SW models, which empirically confirms our expectations that the GDFM produces more accurate forecasts when dealing with large datasets. Data Availability Statement; All relevant data are in the paper Introduction Modelling the short-term dynamics of real GDP is paramount to economic policy. Of the many statistical approaches that have been used in this area recently, principal components analysis stands out as a preferred choice because it integrates large sets of variables in frameworks that rely on only a few common factors used to produce nowcasts and forecasts of economic output. Forecasting economic output and understanding the main drivers of output dynamics are paramount to economic policy making. In today's complex and highly connected economies we are faced with a discouraging amount of data linked to the processes that determine the fluctuations of GDP. While it is not possible to determine a priori whether this huge amount of data should be entirely used in the decision making process, it is also obvious that one cannot afford to ignore the complex macroeconomic relationships that are based on the correlations between output and at least some of these variables. As highlighted in [ 1 ], most empirical analyses of monetary policy are based on the (somewhat inappropriate) assumptions that policy makers use limited amounts of information, despite being clear that policy makers exploit overwhelming numbers of data series. The mere fact that policy makers choose to not ignore variables that are not obviously relevant for the purpose of modelling the processes of interest implies that there is value in the data that may improve significantly the forecasting exercise. The same reasoning applies to the endeavor of forecasting economic growth, in which overreliance on superficial but easily understandable methods (i.e.,vector autoregressions) could turn out to be a risky business due to the very large number of economic and noneconomic variables that influence economic growth. In this paper we aim to develop a dynamic multifactor model that can be used to forecast the short term dynamic of Romania's real GDP using a large dataset of predictors. The model is based on the modelling framework developed by [ 2 ];[ 3 ]. As we shall discuss later, the dynamic multifactor model uses spectral density analysis to assess the dependence of GDP on several economic and noneconomic variables and subsequently produces forecasts (or, if needed, nowcasts) of GDP based on the generalized principal components technique. Afterwards, we propose a alternative model building on the work of [ 4 ];[ 5 ] and based on the mechanics of standard (static) principal components analysis. This model is similar in concept to the dynamic multifactor model but, as we shall see later, it has some limitations. Nevertheless, we provide it in order to assess empirically whether resorting to complex dynamic data analysis leads to an improvement in performance compared with conceptually simpler models. This manuscript is structured as follows. In the ensuing section it is provided a brief review of the literature that deals with factor models and their applications in the process of estimating economic variables and then we (...truncated)


This is a preview of a remote PDF: https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0181379&type=printable

Daniel Armeanu, Jean Vasile Andrei, Leonard Lache, Mirela Panait. A multifactor approach to forecasting Romanian gross domestic product (GDP) in the short run, PLOS ONE, 2017, Volume 12, Issue 7, DOI: 10.1371/journal.pone.0181379