Abstract
This paper develops a dynamic approximate factor model in which returns are time-series heteroskedastic. The heteroskedasticity has three components: a factor-related component, a common asset-specific component, and a purely asset-specific component. We develop a new multivariate GARCH model for the factor-related component. We develop a univariate stochastic volatility model linked to a cross-sectional series of individual GARCH models for the common asset-specific component and the purely asset-specific component. We apply the analysis to monthly US equity returns for the period January 1926 to December 2000. We find that all three components contribute to the heteroskedasticity of individual equity returns. Factor volatility and the common component in asset-specific volatility have long-term secular trends as well as short-term autocorrelation. Factor volatility has correlation with interest rates and the business cycle.
| Original language | English (US) |
|---|---|
| Pages (from-to) | 231-255 |
| Number of pages | 25 |
| Journal | Journal of Econometrics |
| Volume | 132 |
| Issue number | 1 |
| DOIs | |
| State | Published - May 2006 |
Funding
We would like to thank Heather Anderson, John Campbell, Frank Diebold, Serena Ng, and two referees as well as seminar participants at the University of Cambridge, the London School of Economics, ECARES, Bruxelles, and the Common Features in Rio conference for helpful comments. We also thank Christian Huse for research assistance and the Financial Markets Group for financial support.
Keywords
- APT
- ARCH
- Factor models
- Principal components
- Volatility
ASJC Scopus subject areas
- Applied Mathematics
- Economics and Econometrics