Regime-Switching Factor Models and Nowcasting with Big Data
Electronic Access:
Free Download. Use the free Adobe Acrobat Reader to view this PDF file
Disclaimer: IMF Working Papers describe research in progress by the author(s) and are published to elicit comments and to encourage debate. The views expressed in IMF Working Papers are those of the author(s) and do not necessarily represent the views of the IMF, its Executive Board, or IMF management.
Summary:
This paper is concerned with nowcasting in a setting involving a large dataset with mixed frequencies, missing data problems, and regime-switching over time. I show that the Expectation-Maximization (EM) algorithm provides satisfactory performance relative to other estimation methods and delivers a good tradeoff between accuracy and speed, which makes it especially useful for real-time applications involving large dimensional data. The methodology solves the closed-form solutions to the parameter estimators instead of relying on numerical maximization routines. In an application to vintage US data, I show that the regime-switching modification promises improved forecasting performance. The regime-switching model demonstrates superior nowcasting performance, particularly when key economic indicators are released. In addition, the model is able to closely match the recession dating of the NBER despite having less information than actual committee meetings. In a backcasting exercise, I show that the fit between actual NBER recession dates and model recession estimates becomes more apparent with the additional available information. Finally, I discuss the policy implications of regime-switching nowcasting model, by allowing central bankers to evaluate whether the economy is subject to an upcoming recession or experiencing a persistent or transitory inflationary regime, gives the opportunity to act preemptively, more hawkish or dovish if necessary.
Series:
Working Paper No. 2024/190
Subject:
Econometric analysis Economic forecasting Economic growth Economic recession Factor models
Frequency:
regular
English
Publication Date:
September 6, 2024
ISBN/ISSN:
9798400286407/1018-5941
Stock No:
WPIEA2024190
Format:
Paper
Pages:
29
Please address any questions about this title to publications@imf.org