We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Statistics > Methodology

Title: Dynamic Mixed Frequency Synthesis for Economic Nowcasting

Abstract: We develop a novel Bayesian framework for dynamic modeling of mixed frequency data to nowcast quarterly U.S. GDP growth. The introduced framework utilizes foundational Bayesian theory and treats data sampled at different frequencies as latent factors that are later synthesized, allowing flexible methodological specifications based on interests and utility. Time-varying inter-dependencies between the mixed frequency data are learnt and effectively mapped onto easily interpretable parameters. A macroeconomic study of nowcasting quarterly U.S. GDP growth using a number of monthly economic variables demonstrates improvements in terms of nowcast performance and interpretability compared to the standard in the literature. The study further shows that incorporating information during a quarter markedly improves the performance in terms of both point and density nowcasts.
Comments: arXiv admin note: text overlap with arXiv:1601.07463
Subjects: Methodology (stat.ME)
Cite as: arXiv:1712.03646 [stat.ME]
  (or arXiv:1712.03646v3 [stat.ME] for this version)

Submission history

From: Kenichiro McAlinn [view email]
[v1] Mon, 11 Dec 2017 05:01:50 GMT (2871kb,D)
[v2] Tue, 16 Jan 2018 17:09:16 GMT (2872kb,D)
[v3] Thu, 7 Jun 2018 23:48:56 GMT (3959kb,D)

Link back to: arXiv, form interface, contact.