A comparative analysis of independent director appointments on stock price behaviour: an empirical investigation based on the Markov regime-switching model

Abstract

In 2002, the Taiwanese government regulated that new companies should appoint independent directors for good corporate governance. The purpose of this article is to compare appointing independent directors' motivation of structural changes in stock behaviour. The empirical results showed that the companies with mandatory or voluntary independent director appointments significantly reduced both stock returns and risks over 2000–2006.

Keywords:

• independent directors
• Markov regime-switching model
• stock price behaviours

JEL Classification::

• G38
• G12

Notes

¹ Hansen (1992) proposed the use of LR tests to assess quantity. When the null hypothesis is established, however, indeterminate problems occur. Hence, Garcia (1998) proposed the use of the LR principle to establish an LR test to infer asymptotic distribution and generate a critical value table, to eliminate the presence of indeterminate problems in null hypotheses.

² Filtering probability derives the probability value of based on data obtained from the first to the tth periods; smoothing probability, conversely, derives results based on data from all the samples, and regards probability based on hindsight. Filtering probability is thus a real-time viewpoint; smoothing probability presents an overall observation of the event in hindsight, and therefore, has a superior smoothing effect. See Hamilton (1994) for more information regarding the calculation of filtering probability and smoothing probability.

³ Eliminated companies were those underwent capital reduction or full delivery stocks, which led to incomplete stock prices.

⁴ The 5% threshold limit value was derived from Hamilton's (1994) ‘Time Series Analysis’, Table B.6 Case 2, Case 4. The optimal lag number p of the unit root test was derived from Schwert (1987).

⁵ Due to length restrictions, this study did not include the results of the unit root test and the SUR assessment.

⁶ The total probability rate of the regimes 0 and 1 is 1. Therefore, only presenting the filtering probability and smoothing probability of a regime is necessary to derive the probability values of the other. Thus, Figs 3 and 4 only presented the filtering probability and smoothing probability of the regime 1.