Determinants of Price Discovery in Option Markets: An Interpretable Machine Learning Perspective
用可解释机器学习方法，基于1秒分辨率数据，发现上证50ETF期权市场比现货市场更具信息优势，交易成本是主因，杠杆、做市商风险和投机次之，流动性和持仓量影响较小。
ABSTRACT This paper empirically demonstrates that the SSE 50 ETF option market has the informational advantage compared to the underlying market, and evaluates the relative importance of option characteristics in price discovery using interpretable machine learning methods. Estimating the Information Leadership Share using 1‐s resolution price data as a measure of price discovery indicates that price discovery occurs in the SSE 50 ETF option market more, less in the underlying market. The feature importance analysis reveals that trading cost is the primary factor contributing to the informational advantage of option markets, followed by leverage, market maker risk, and speculation, while liquidity and open interest have less impact. Extensive robustness tests are also conducted to assess the stability of the feature importance.