|Abstract: ||名稱：多次購買資產的經理人有學習效果、過度自信或代理問題嗎？ Do managers frequently buying assets experience learning effect, overconfidence, or agency problem? 1. 研究動機與目的 購買資產是公司重要的投資活動，類似的投資是公司併購(M&A)，過去文獻常從 CEO過度自信、經驗學習及代理問題等理論解釋多次M＆A的行為，但實證結論不一致。 但購買資產與併購在範圍、複雜度及整合難易程度上均有所差異，故多次執行購買資產 的動機及其績效，未必與M&A文獻的結果一致，爰本研究即針對資產購買活動，從CEO 而非公司角度，檢視其任職時的購買資產事件的績效符合何種假說或理論之內容。 為了解購買資產績效與CEO過度自信可能的動態變化的關係，本研究擬直接衡量 CEO過度自信程度與代理問題嚴重性，俾利觀察純粹由過度自信或代理問題所產生的效 果及其交互影響，以及是否有經驗學習存在。 2. 研究方法 樣本是美國公司的購買資產事件，估計宣告報酬和長期績效，再以分組差異檢定(t 檢定與無母數)及複迴歸分析，觀察過度自信、代理問題、事件次序對事件績效的影響。 最後也進行多項穩健性檢查以確保結果有效性。 3. 預期貢獻 本研究可補充行為財務學在過度自信方面的證據，及透過同時考慮CEO過度自信、 代理問題和學習效果，對購買資產的決策有更深的認識。特點在檢視動態過度自信對購 買資產績效的影響、觀察長期績效來檢定組織學習成效，結果可與過去文獻互為比較。
Topic: Do managers frequently buying assets experience learning effect, overconfidence, or agency problem? 1. Motives and objectives Assert purchase is firms’ important investment activities. A similar investment is firm merger and acquisition (M&A). Past studies often explain serial M&A behaviors using theories such as CEO overconfidence, experience learning, and agency problem, but empirical results are much inconsistent. Asset purchase and M&A are considerably different in activity area, complication, and difficulty of integration, so the incentives and performance for serial purchases of assets may be inconsistent with those from past studies’ findings. Therefore, this study intends to investigate the performance of asset purchases from CEO level, not firm level, to test which financial theories have explanatory power for incentives and performance of serial activities. To understand the relationship between performance of asset purchases and dynamic overconfidence, this study will directly measure the degree of CEO overconfidence and the seriousness of agency problem. This procedure is helpful for examining the pure effects and interaction effects of overconfidence and agency problem as well as learning effects. 2. Methodology This study uses company asset purchase events in US as samples and estimates announcement returns and long-term performance. These performance results are employed to perform difference tests and multiple regression analyses for examining the effects of overconfidence, agency problem, and deal order on event performance. Finally, some robustness checks are performed to ensure the validity of conclusions. 3. Expected contributions This study can fill the evidence of overconfidence in behavioral finance. This research can give an insight into the asset purchase decisions by simultaneously considering overconfidence, agency problem, and learning effects of CEO. The special feature of this study is that I examine the effect of dynamic overconfidence on asset purchase performance and the organization learning via observing long-run performance. The results can be comparable to the past studies’ those.