過去研究探討高階主管薪酬、研究發展支出、股權結構與公司價值四者間關聯性,大多數採用多元或聯立迴歸估計方式,僅對各觀察變項逐項探討兩兩間的關聯性,忽略了具有「潛在」概念的研究變項是由多個觀察指標共同組合,且具有同時存在的現象及高階主管薪酬、研究發展支出、股權結構三者間透過彼此的中介作用,間接影響公司價值的關係,本研究以結構方程式(SEM)探討其直接與間接關係。以民國87年01月01日至92年12月31日持續在台灣上市上櫃公司為研究對象。 實證結果顯示: (1)直接效果: 研究發展支出、高階主管薪酬、股權結構與公司價值間均呈現非常顯著的正向關係。 (2)交互效果:(a)整體產業中高階主管薪酬與研究發展支出間具有非常顯著的正向交互作用。(b)高科技產業中股權結構與高階主管薪酬間具有顯著的負向交互作用。(3)中介效果:(a)研究發展支出對「高階主管薪酬與公司價值之關係」具有顯著的正向中介作用。(b)高階主管薪酬對「研究發展支出與公司價值之關係」具有顯著的正向中介作用。(c)高科技產業中高階主管薪酬對「股權結構與與公司價值之關係」具有顯著的負向中介作用。(4)在整體產業、高科技產業與非高科技產業中創造公司價值的最佳解釋模型皆有不同的要徑。(5) 並未發現不同的經濟景氣因素對高階主管薪酬、研究發展支出、股權結構三者與公司價值間的關係,會產生不同的效果。 In the prior researches study the associations among the compensation of executive officer (CEO), R&D expenditure (R&DE), ownership structure (OS), and corporation value (CV) that emphasizing the associations of both of them. The results of researches are not consistent. This thesis focuses the simultaneous phenomenon among the first three factors and their direct and indirect effects on the corporation value. By utilizing the structure equation modeling (SEM) instead of the multiple regression analysis, the aims are to realize the structured relationships among the above four factors under various industry and business cycle. The samples form the corporations listed continuously on the Taiwan Stock Exchange form January 1st, 1998 to December 31st, 2003. The results of research show as follows: (1) there are significantly positive direct effects of the associations between both of the first three factors on the corporation value. (2) There is a significantly interactive positive effect of the association between CEO and R&DE on the corporation value in all samples (industries) case, however , only in high-tech industry OS and CEO effect to be negative. (3) There are significantly positive intermediation effects of the association between CEO and R&DE on the corporation value in all samples case, however, only in high-tech industry OS and CEO effect to be negative. (4) There is the optimum model to achieve the best corporation value, but is different in all sample case and high-tech industry. (5) In the different business cycle, they don’t cause different effect.