On forecasting recessions via neural nets

Kiani, Khurshid (2008) On forecasting recessions via neural nets. Economics Bulletin, 3 (13). pp. 1-15. ISSN 1545-2921

PDF - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
Download (214kB) | Preview


In this research, we employ artificial neural networks in conjunction with selected economic and financial variables to forecast recessions in Canada, France, Germany, Italy, Japan, UK, and USA. We model the relationship between selected economic and financial (indicator) variables and recessions 1-10 periods in future out-of-sample recursively. The out-of-sample forecasts from neural network models show that among the 10 models constructed from 7 indicator variables and their combinations that we investigate, the stock price index (index) and spread between bank rates and risk free rates (BRTB) are most likely candidate variables for possible forecasts of recessions 1-10 periods ahead for most countries.

Item Type: Article
Keywords: Business cycles; Neural network; Out-of-sample forecasts; Recession; Real GDP
Schools/Departments: University of Nottingham Ningbo China > Faculty of Business > Nottingham University Business School China
Depositing User: CHEN, Jiaorong
Date Deposited: 22 Nov 2017 09:05
Last Modified: 24 Nov 2017 12:02
URI: http://eprints.nottingham.ac.uk/id/eprint/48257

Actions (Archive Staff Only)

Edit View Edit View