Real-Time Forecasting with a (Standard) Mixed-Frequency VAR During a Pandemic

Abstract

In this paper we resuscitate the mixed-frequency vector autoregression (MF-VAR) developed in Schorfheide and Song (2015) to generate real-time macroeconomic forecasts for the U.S. during the COVID-19 pandemic. The model combines eleven time series observed at two frequencies: quarterly and monthly. We deliberately do not modify the model specification in view of the recession induced by the COVID-19 outbreak.We find that forecasts based on a pre-crisis estimate of the VAR using data up until the end of 2019 appear to be more stable and reasonable than forecasts based on a sequence of recursive estimates that include the most recent observations. Overall, the MF-VAR outlook is quite pessimistic. The estimated MF-VAR implies that level variables are highly persistent, which means that the COVID-19 shock generates a long-lasting reduction in real activity. Regularly updated forecasts are available at www.donghosong.com/.

Large mixed-frequency VARs with a parsimonious time-varying parameter structure

“…In Schorfheide and Song ( 2020 ), the authors update their MF-VAR from the 2015 paper to provide a sequence of real-time forecasts, starting at the last pre-COVID quarter (2019:Q4) until the summer of 2020. We do the same exercise with our models for GDP.…”

“… 15 We have extended the data set until July 31 (downloaded on August 5, 2020). We have also adapted the new specification for the hyperparameters from Schorfheide and Song ( 2020 ). Specifically, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\lambda _1=0.01$\end{document} , \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\lambda _2=1$\end{document} , \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\lambda _3=1$\end{document} , \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\lambda _4=3$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\lambda _5=1$\end{document} .…”

A comparison of monthly global indicators for forecasting growth

“…As the health crisis unfolded, one of the most pressing tasks for policymakers around the world was to gauge the evolution of the state of the economy in a timely fashion. There are a number of recent studies that provide nowcasts and forecasts of real GDP growth for the U.S. economy (see, e.g., Diebold, 2020;Schorfheide and Song, 2020) and the G7 countries (see, e.g., Foroni, Marcellino, and Stevanović, 2020) for 2020. Existing growth assessments for a particular country tend to rely on domestic variables only.…”

“…This is relevant when deciding whether to update the estimates of the model parameters by including the most recent observations or whether to …x the parameters at pre-crisis estimates. Schorfheide and Song (2020) note that if the pandemic shock was a one-time aberration that did not change the structure of the economy, then it is best to omit the crisis observations from estimation since they might distort the parameter estimates; but, if the pandemic did a¤ect how economic variables interact, then parameter estimates should be updated using the latest information. We examine the role of the estimation sample by comparing the nowcasts for 2020Q2 derived from a model with updated estimates and estimates kept at their 2019Q4 values.…”

“…These e¤orts have been supported by a large academic literature that has developed nowcasting and forecasting approaches geared toward reliably gauging the underlying state of the economy before the release of o¢ cial real GDP numbers based on high-frequency indicators. Popular approaches include factor models (e.g., Stock and Watson, 2002;Giannone, Reichlin, and Small, 2008;Schumacher and Breitung, 2008;Chernis and Sekkel, 2017), bridge equations (e.g., Ba¢ gi, Golinelli, and Parigi, 2004;Foroni and Marcellino, 2014;Golinelli and Parigi, 2014), mixed-frequency models (e.g., Andreou, Ghysels, and Kourtellos, 2010;Galvão, 2008, 2009;Kuzin, Marcellino, and Schumacher, 2011;Schorfheide and Song, 2020), and combinations thereof (e.g., Marcellino and Schumacher, 2010;Schumacher, 2016). This literature has concluded that exploiting the information content of high-frequency variables improves the accuracy of macroeconomic nowcasts and forecasts.…”