DMCA
Nonlinear Dynamic Structures (1993)
| Venue: | Econometrica |
| Citations: | 130 - 10 self |
BibTeX
@ARTICLE{Gallant93nonlineardynamic,
author = {A. Ronald Gallant and Peter E. Rossi and George Tauchen and A. Ronald Gallant and Peter E. Rossi and George Tauchen},
title = {Nonlinear Dynamic Structures},
journal = {Econometrica},
year = {1993},
pages = {871--907}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
We describe three methods for analyzing the dynamics of a nonlinear time series that is represented by a nonparametric estimate of its one-step ahead conditional density. These strategies are based on examination of conditional moment profiles corresponding to certain shocks; a conditional moment profile is the conditional expectation evaluated at time t of a time invariant function evaluated at time t + j regarded as a function of j. The first method, which compares conditional moment profiles to baseline profiles, is the nonlinear analog of conventional impulse-response analysis. The second assesses the significance of a profile by comparing its sup-norm confidence band to a null profile. The third examines profile bundles for evidence of damping or persistence. Experimental designs for choosing an appropriate set of shocks are discussed. These methods are applied to a bivariate series comprised of daily changes in the Standard and Poor's composite price index and daily NYSE transactions volume from 1928 to 1987. The findings from these data are: (i) The multi-step ahead conditional volatility profile exhibits a symmetric response to both positive and negative price shocks. In contrast, the conditional volatility profile of the univariate price change process exhibits an asymmetric response. (ii) The one-step ahead response of volume to price shocks is different than the multi-step ahead response. Price shocks produce an increase in volume one-step ahead but decrease it in subsequent steps. (iii) There is little evidence for long-term persistence in either the conditional mean or volatility of the bivariate process. o 1
Keyphrases
conditional moment profile nonlinear dynamic structure conditional volatility profile price shock nonlinear analog conditional mean sup-norm confidence band time invariant function daily change bivariate series null profile third examines symmetric response composite price index first method negative price shock experimental design conditional density long-term persistence daily nyse transaction volume nonlinear time series appropriate set univariate price change process asymmetric response nonparametric estimate certain shock little evidence conditional expectation conventional impulse-response analysis bivariate process subsequent step
