基于资金流数据的筹码结构因子构建——投资者分层视角下的信息增量

Key Points - The article discusses the construction and application of chip structure in stock selection, emphasizing the importance of understanding investor behavior and the distribution of funds at different price levels [2][6][19]. - It highlights the dynamic nature of chip costs, which are influenced by both buying and selling behaviors, and the necessity of detailed micro-trading data for accurate calculations [4][7]. - The analysis of chip structure allows for the identification of potential support and resistance levels in stock prices, providing a richer set of information compared to traditional volume-price factors [6][18]. Chip Average Cost Construction and Application - The construction of chip average cost involves calculating the weighted average cost of historical chips to depict the current market's average holding cost [8][9]. - A higher indicator value indicates that the historical average cost is above the current stock price, suggesting a floating loss for the market, while a lower value indicates a floating profit [9][10]. Improvement of Chip Structure Based on Fund Classification - Traditional methods of chip construction do not differentiate between different types of investors, which can obscure important behavioral information [19][20]. - By introducing fund flow data from institutions and retail investors, the article proposes a more nuanced approach to constructing chip distribution and weighted cost indicators for different investor types [19][21]. - The method shows slight improvements in factor IC and monotonicity of grouped returns, although the overall enhancement is limited [20][22]. Factor Synthesis and Performance Analysis - The article discusses the complementary relationship between the institutional-retail chip cost difference factor and traditional chip cost factors, suggesting that one contributes to predictive strength while the other enhances ranking stability [31][32]. - The synthesized chip factor achieves an IC mean of 4.35%, outperforming traditional volume-based chip factors and showing improved stability in multi-group performance [32][35]. - The performance of the synthesized factor varies significantly across different market capitalization segments, with better results observed in small to mid-cap stocks compared to large-cap stocks [41][42]. Market Capitalization Domain Results - The synthesized factor demonstrates stronger predictive capabilities in small-cap stocks, where investor behavior is more aligned with trading strategies that involve high turnover and profit-taking [45][46]. - Adjustments to the factor application in the mid-cap segment have led to improved performance, indicating the importance of tailoring strategies to specific market conditions [43][44].