Why are the Beveridge-Nelson and unobserved-components decompositions of GDP so different?

• 成果类型: Article
• 署名作者: Morley, JC; Nelson, CR; Zivot, E
• 署名单位: Washington University (WUSTL); University of Washington; University of Washington Seattle
• 刊物名称: REVIEW OF ECONOMICS AND STATISTICS
• ISSN/ISSBN: 0034-6535
• DOI: 10.1162/003465303765299765
• 发表日期: 2003-05
• 页码: 235-243
• 关键词: nonstationary time-series; BUSINESS; TRENDS; permanent; cycles

摘要

This paper reconciles two widely used decompositions of GDP into trend and cycle that yield starkly different results. The Beveridge-Nelson (BN) decomposition implies that a stochastic trend accounts for most of the variation in output, whereas the unobserved-components (UC) implies cyclical variation is dominant. Which is correct has broad implications for the relative importance of real versus nominal shocks. We show the difference arises from the restriction imposed in UC that trend and cycle innovations are uncorrelated. When this restriction is relaxed, the UC decomposition is identical to the BN decomposition. Furthermore, the zero-correlation restriction can be rejected for U.S. quarterly GDP, with the estimated correlation being -0.9.