A dynamic factor model approach to incorporate Big Data in state space models for official statistics

TL;DR

This paper introduces a dynamic factor model to enhance state space models with high-dimensional auxiliary data, enabling real-time unemployment nowcasting using Google Trends and claimant counts.

Contribution

It extends traditional state space models by integrating high-dimensional data sources through a dynamic factor approach for improved real-time estimation.

Findings

Effective incorporation of Google Trends and claimant counts improves unemployment nowcasting.

The model provides reliable real-time unemployment estimates before official survey data are available.

Demonstrated application to Dutch unemployment data with enhanced accuracy.

Abstract

In this paper we consider estimation of unobserved components in state space models using a dynamic factor approach to incorporate auxiliary information from high-dimensional data sources. We apply the methodology to unemployment estimation as done by Statistics Netherlands, who uses a multivariate state space model to produce monthly figures for the unemployment using series observed with the labour force survey (LFS). We extend the model by including auxiliary series of Google Trends about job-search and economic uncertainty, and claimant counts, partially observed at higher frequencies. Our factor model allows for nowcasting the variable of interest, providing reliable unemployment estimates in real-time before LFS data become available.