Abstract
A second-moment, regime-switching model with not only a switching intercept and a switching slope, but also a switching error variance, is applied to examine the impacts of exchange rate volatility (ERV) on corporate values (CV) for the 10 industries investigated in Taiwan. Two different regimes categorized as strong-impact and weak-impact are identified. The dominant power varies from one industry to another. The Wald statistics for the null of equality are ambiguous, which show that if the Markov-switching (MS) model is plausible, then the ERV might not be one major factor, but another factor that could switch the CV of Taiwan's industries. For the model's volatility influence, the data of 8 out of 10 industries are shown to fit a two-state model when the volatility is stimulated. A two-state, first-order MS model is appropriate for the ‘goodness of fit’ analysis at the 10% significant level.
Notes
1 The elaborated work of Hamilton's MS model adopts the spirit of the probability switching mechanism by Goldfelt and Quandt's (1973) for heteroscedasticity, which categories two unobserved regimes (states), each with a fixed probability, by the observed data.
2 The original MS model focuses on the mean behaviour of variables. We take into account the probability of conditional heteroscedasticity of the disturbance term in order to examine the time-varying, two-state unobserved variances, which emerged from the impulse of CVs to ERV.
3 The applications of the extended MS model accommodating the pattern of conditional volatility can be found in Hamilton and Lin (Citation1996), Dueker (Citation1997) and Ramchand and Susmel (Citation1998).
4 All series are integrated of order one. In order to save space, we do not present these ADF results, but these are available from the authors upon request.
5 The result of an existing GARCH effect for the ERV of NTD/USD is consistent with those findings in Pozo (Citation1992) and Arize (Citation1995, Citation1997), which measure the ERV of the US and the G-7 countries.
6 The nonstationary property is obtained from the ADF (Dickey and Fuller, Citation1981) unit-root test, which is omitted in this article; however, those references will be available upon request.
7 The Markov property argues that the process of s t depends on the past realizations only through s t −1.
8Garcia and Perron (Citation1996) employ Hamilton's (Citation1989) MS model to explicitly account for regime shifts in an autoregressive model with three-state MS mean and variance.
9 For simplicity, the following analysis is based only on one industry. We thus omit the symbol.