ETF 對台灣股票市場定價資訊效率之影響

Abstract

近年來ETF規模在全球及台灣市場快速成長，已成為資本市場核心交易工具。既有國外文獻（如Israeli, Lee, & Sridharan (2017)）指出，ETF持股增加可能因吸引噪音交易者離開個股市場，導致成分股交易成本上升並損害長期定價資訊效率。然而，台灣股市具備淺碟性質、設有漲跌幅限制，且外資法人影響甚鉅，這些特性使得外資進出易造成股價鎖停與流動性衝擊。因此，本研究動機在於驗證在台灣特殊的制度環境下，ETF的一籃子交易與套利機制是否能反向扮演提供流動性的角色，進而提升市場效率。 本研究旨在探討台灣市場ETF持股變動對成分股「交易成本」與「價格資訊性」的長期影響。具體目標包括衡量流動性代理變數之變化、驗證交易成本對資訊效率的傳導機制，並分析不同股票類型（如散戶偏好股）是否存在異質性效果，最後提供監理機關與投資人實務啟示。 研究樣本涵蓋2006年至2024年台灣上市櫃公司，總計17,593筆觀測值。本研究採用線性迴歸模型，以「買賣價差」（Corwin & Schultz指標）與「價格衝擊」（Amihud分解指標）衡量交易成本，並以「股價報酬同步性」作為價格資訊性的代理變數。模型中嚴格控制了機構持股變化、市值、周轉率、帳面市值比等控制變數，並納入產業與年份固定效果。 實證結果顯示，ETF持股每上升1個百分點，個股買賣價差顯著下降約3.5個基點，價格衝擊亦顯著下降1.6個基點，支持ETF能降低台股交易成本之假說。資訊效率方面，同步性雖呈下降趨勢（代表資訊性提升），但在統計上未達顯著水準。此外，研究發現機構持股增加反而會顯著提高交易成本，與ETF的效果截然不同。分成不同期間的子樣本分析進一步證實，隨著近年ETF成交比重提升，其改善流動性的正面效果在2015年後更加顯著。異質性分析則顯示，影響差異主要來自ETF成分股本身的特徵（如高價格、高市值、低本益比），而非散戶偏好導致的傳導機制差。

In recent years, the scale of ETFs has grown rapidly in both global and Taiwanese markets, making ETFs a core trading instrument in capital markets. Existing international literature, such as Israeli, Lee, and Sridharan (2017), suggests that an increase in ETF ownership may draw noise traders away from the individual stock market, thereby increasing the trading costs of constituent stocks and impairing long-term pricing efficiency. However, Taiwan’s stock market is characterized by a relatively shallow market structure, daily price limit rules, and substantial influence from foreign institutional investors. These features make foreign capital flows more likely to trigger limit-up or limit-down price movements and liquidity shocks. Accordingly, the motivation of this study is to examine whether, under Taiwan’s unique institutional environment, ETFs’ basket trading and arbitrage mechanisms may instead play a liquidity-providing role and thereby enhance market efficiency. This study aims to investigate the long-term effects of changes in ETF ownership on the “trading costs” and “price informativeness” of constituent stocks in the Taiwanese market. Specifically, this study seeks to measure changes in liquidity proxies, examine the transmission mechanism through which trading costs affect informational efficiency, analyze whether heterogeneous effects exist across different types of stocks, such as retail-investor-preferred stocks, and provide practical implications for regulators and investors. The sample covers listed and over-the-counter companies in Taiwan from 2006 to 2024, with a total of 17,593 observations. This study adopts a linear regression model. The bid-ask spread, measured by the Corwin and Schultz indicator, and price impact, measured by the decomposed Amihud indicator, are used as proxies for trading costs. Stock return synchronicity is used as a proxy for price informativeness. The model rigorously controls for changes in institutional ownership, market capitalization, turnover rate, book-to-market ratio, and other control variables, while also incorporating industry and year fixed effects. The empirical results show that a one-percentage-point increase in ETF ownership significantly reduces the bid-ask spread of individual stocks by approximately 3.5 basis points and also significantly lowers price impact by 1.6 basis points. These findings support the hypothesis that ETFs can reduce trading costs in the Taiwanese stock market. In terms of informational efficiency, stock return synchronicity exhibits a downward trend, indicating improved price informativeness, but the result is not statistically significant. In addition, this study finds that an increase in institutional ownership significantly raises trading costs, in sharp contrast to the effect of ETFs. Subsample analyses across different periods further confirm that, as the trading share of ETFs has increased in recent years, their positive effect on liquidity improvement has become more pronounced after 2015. The heterogeneity analysis indicates that the differences in effects mainly stem from the characteristics of ETF constituent stocks themselves, such as high stock prices, large market capitalization, and low price-to-earnings ratios, rather than from differences in transmission mechanisms driven by retail investor preferences.