特性差異對美國存託憑證與其標的股間報酬傳遞關係之研究

Abstract

本研究以中國、日本、南韓、及台灣四個國家共45個有發行ADR之公司為樣本，而以其美國存託憑證與標的股之報酬為研究標的，以檢視其在不同特性的分類下，其互動及資訊傳遞過程是否會有差異。分類特性選用「美股交易量大小」、「美股交易量占標的股當地交易量比重」、「媒體上報次數」、「總分析師家數」、以及「美國分析師家數占總分析師家數比重」，並將樣本依分類分成高中低組來進行比較。 因果檢定顯示，「美股交易量大小」、「美股交易量占標的股當地交易量比重」、及「總分析師家數」都是高組別的ADR領先標的股之報酬，而低組別則是雙向回饋；「媒體上報次數」則是高次數組的兩者間具有雙向回饋，低次數組的則是ADR之領先標的股之報酬；「美國分析師家數占總分析師家數比重」因為高低組皆為ADR與標的股報酬雙向回饋，表示分類無法造成兩組的因果差異。 從衝擊反應分析結果顯示，在各分組下，美國存託憑證受自身外生衝擊之反應幅度大於標的股受自身外生衝擊的反應幅度。在標的股之報酬和ADR之報酬的相互衝擊影響上，短期內高組別皆大於低組別的影響係數，以「美股交易量大小」及「媒體上報次數」的分組其高低組間的影響係數差異最明顯，而「美國分析師家數占總分析師家數比重」的高低組間差異最小。 解釋波動變異的變異數分解顯示，標的股報酬對自身變異的解釋能力比ADR 之報酬高。在「美股交易量大小」、「美股交易量占標的股當地交易量比重」、及「媒體上報次數」三組中， ADR報酬在高組別比較有影響力，而標的股報酬則是在低組別比較有影響力；在「總分析師家數」之ADR報酬和標的股報酬都是在低家數組有較大的影響力；而「美國分析師家數占總分析師家數比重」則剛好和「總分析師家數」的結論相反。

This thesis aims to explore whether and how information characteristics help explaining the differences in international market return transmissions between ADRs and the underlying securities (hereafter also denoted as UND) in China, Japan, South Korea, and Taiwan. The information factors investigated in this study include ADR trading volume, ratio of ADR trading volume to UND trading volume, media coverage, number of research firms covering the security, and proportion of U.S. research firms covering the security. From the causality test, ADRs index in the group with greater ADR trading volume, greater ratio of ADR trading volume to UND trading volume, and larger number of research firms covering the security appear to lead the UNDs index. In contrast, ADRs index in the groups with smaller ADR trading volume, smaller ratio of ADR trading volume to UND trading volume, and smaller number of research firms covering the security are in a mutual feedback relationship with UNDs index. ADRs index in the group with high media coverage is in a mutual feedback relationship with UNDs, but ADRs index in the low group leads the UNDs index. Moreover, ADRs index in both high and low groups with respect to proportion of U.S. research firms covering the security are in a mutual feedback relationship with UNDs. The empirical results of impulse response functions support the notion of stronger sudden impulse to the exogenous event for the ADRs than the UNDs. Moreover, the difference between high versus low group is significant when ADR trading volume and/or media coverage is used to partition the sample but is insignificant when proportion of U.S. research firms covering the security is adopted. The estimates from variance decomposition function indicates that the UNDs index in the group with higher ADR trading volume, higher ratio of ADR trading volume to UND trading volume and/or higher media coverage are affected more by ADRs index, and ADRs index in the low groups are affected more by UNDs. The UNDs index and ADRs index are affected more by each other in the groups with lower number of research firms covering the security and in the group with greater proportion of U.S. research firms covering the security.