Slow Expectation-Maximization Convergence in Low-Noise Dynamic Factor Models

Arbeitspapier

This paper addresses the poor performance of the Expectation-Maximization (EM) algorithm in the estimation of low-noise dynamic factor models, commonly used in macroeconomic forecasting and nowcasting. We show analytically and in Monte Carlo simulations how the EM algorithm stagnates in a low-noise environment, leading to inaccurate estimates of factor loadings and latent factors. An adaptive version of EM considerably speeds up convergence, producing substantial improvements in estimation accuracy. Modestly increasing the noise level also accelerates convergence. A nowcasting exercise of euro area GDP growth shows gains up to 34% by using adaptive EM relative to the usual EM.

Object Details

• Language: Englisch

References and Relationships

• Bibliographic citation: Series: Tinbergen Institute Discussion Paper ; No. TI 2023-018/III

Classification and Topics

• Classification: Wirtschaft ; Multiple or Simultaneous Equation Models: Time-Series Models; Dynamic Quantile Regressions; Dynamic Treatment Effect Models; Diffusion Processes; State Space Models ; Model Construction and Estimation ; Forecasting Models; Simulation Methods ; Prices, Business Fluctuations, and Cycles: Forecasting and Simulation: Models and Applications
• Subject: Dynamic factor models ; EM algorithm ; artificial noise ; convergence speed ; nowcasting

Contributors, Places and Time

• Event: Geistige Schöpfung
• (who): Opschoor, Daan ; van Dijk, Dick
• Event: Veröffentlichung
• (who): Tinbergen Institute
• (where): Amsterdam and Rotterdam
• (when): 2023

Further information

• Handle: http://hdl.handle.net/10419/273829
• Last update: 10.03.2025, 11:43 AM CET