- 期刊
臺灣上市櫃電子資訊業撤離大陸的決策：檢驗地主國績效與選擇偏誤的影響
Divestiture from China by Taiwanese Listed Electronic Information Firms: Effects of Host Country Performance and Selection Bias
摘要
過去文獻多採地主國財務績效解釋撤離決策，主張海外子廠商或部門業績不佳，讓母公司無法承受進一步虧損而撤離地主國。然而，決策者有時出現不承認投資失敗的承諾升高現象，在地主國績效不佳時反而加強投資。此外，廠商亦可能因規避不確定性，而放棄尚有利潤的事業部門，導致地主國績效與撤離決策之關係並未收斂。本研究以廠商行為理論和前景理論為基礎，重新檢驗大陸投資績效對撤離決策的影響，透過臺灣經濟新報資料庫所搜尋赴大陸投資之上市（櫃）之電子資訊業計506家，研究期間2016至2020年共獲致2,530個觀測值進行實證，實證時採用Heckman（1979）的兩階段研究法。研究發現大陸投資獲利的情況下，大陸投資績效與撤離決策呈現倒U型關係，但大陸投資虧損時，廠商愈不可能撤離大陸。另外，我們發現撤離大陸與市場績效呈正向關係。最後，修正因子及決策錯誤均達顯著水準，可達到控制選擇偏誤的目的。
並列摘要
Past literature has often explained withdrawal decisions from the host country regarding financial performance, suggesting that poor performance of foreign subsidiaries or divisions forces the parent company to withdraw to avoid further losses. However, decision-makers sometimes exhibit an escalation of commitment, where they increase investment despite poor performance in the host country, refusing to acknowledge investment failure. Additionally, companies might abandon profitable business units to avoid uncertainty, resulting in a lack of convergence between host country performance and withdrawal decisions. This study re-examines the impact of investment performance in mainland China on withdrawal decisions based on firm behavior and prospect theory theories. Using data from the Taiwan Economic Journal (TEJ) database, the study focuses on 506 publicly listed Taiwanese electronic information companies that invested in the Chinese market from 2016 to 2020, yielding 2,530 observations for empirical analysis. The analysis was conducted using two-stage research method (Heckman, 1979). The findings reveal an inverted U-shaped relationship between investment performance in China and withdrawal decisions when the investments are profitable. Conversely, firms are less likely to withdraw from China when investments are unprofitable. Additionally, we found a positive relationship between withdrawal from China and market performance. Finally, correction factors and decision errors are statistically significant, effectively controlling for selection bias.