公司財務特性對宣告買回庫藏股之異常報酬影響：以美國科技業為例

Abstract

本研究以宣告買回自身公司股份之美國科技業公司為研究對象，並使用事件研究法以及累積平均異常報酬探討實施庫藏股宣告對短期股價的影響，並加入資產規模、市價淨值比、資本支出、現金持有狀況以及自由現金流量等公司特性進行分群以及迴歸分析。實證結果顯示，在實施庫藏股宣告前全體樣本皆呈現負向的累積平均異常報酬，而在宣告後，逐漸開始累積正向的累積平均異常報酬。而在分群及迴歸分析結果中顯示，當資產規模、市價淨值比越小、資本支出率越高，越能夠產生較高的正向異常報酬，而現金持有狀況及自由現金流量在建立之假說中與預期方向不符且不顯著。

而在長期衡量股價超額報酬的部分則以買進並持有之異常報酬作為衡量方法，並以市場指數調整法及控制公司法建構六種預期報酬的基準模型。在實證結果中，只有市場指數調整模型以及規模/市價淨值比模型在三年內的買進並持有之異常報酬為顯著正向，故長期而言，宣告實施庫藏股之科技業公司確實存在著正向異常報酬。而在其他的基準模型中，因和樣本公司所配對之非事件日公司過少，導致樣本的代表性較不足，皆為不顯著。

This study focuses on the U.S. technology industry that announced the share repurchase and uses the event study method and cumulative average abnormal return to explore the effects of the announcement of share repurchase on the stock price. Furthermore, this study adds company characteristics such as asset size, price-to-book ratio, capital expenditure, cash holdings, and free cash flow for clustering and regression analysis. The empirical results indicate that all the samples showed a negative cumulative average abnormal return before the announcement of share repurchase. On the other hand, all the samples gradually began to accumulate a positive cumulative average abnormal return after the announcement. The results of clustering and regression analysis show that the asset size, price-to-book ratio is smaller, and the capital expenditure rate is higher, the higher the positive abnormal return can be generated but the cash holdings and free cash flow are not accord with expected direction and hypotheses in this study.

In the long-term measurement of excess stock return, Buy and hold abnormal return is used as the measurement method, and the benchmark model of six expected returns is established by the Market Adjusted Returns Model and Control Firm Method. Empirical results show that only the BHAR calculated with Market Adjusted Returns Model and size/price-to-book ratio model are significant positive returns within three years. Therefore, technology companies that announced the share repurchase exist positive abnormal returns in the long run. In other benchmark models, there are too few non-sample companies matched with the sample companies resulting in the samples which is not significant and less representativeness.