ChiNext-Based Iterative Refinement for No-Short-Sale Constrained MV Portfolios

Against the backdrop of increased volatility in global capital markets, the traditional mean-variance model faces challenges due to conflicts between its generation of negative weights and the institutional environment of Chinas A-share market, which imposes a no short-selling constraint. How to balance the return objectives with the no short-selling constraint has become an urgent issue to address. Moreover, the strong homogeneity and high volatility of the global emerging technology sector pose new tests to the applicability of the traditional mean-variance model. This study focuses on highly positively correlated stock portfolios on the ChiNext board and integrates an optimization approach that iteratively removes assets with negative weights: by gradually excluding negatively weighted assets and dynamically updating the covariance matrix, it solves for the efficient frontier under realistic constraints. The empirical analysis uses five years of monthly return data and finds that the optimized portfolio significantly concentrates on a small number of stocks with high Sharpe ratios, achieving fully positive weights when the target return is set above the market average; meanwhile, the strong sector linkage effect weakens the traditional diversification benefits, causing the portfolio to naturally tend toward concentrated holdings. This method provides a practical asset allocation approach for assets with strong sector linkages and holds practical significance for institutional investors seeking to balance returns and risks.