Nonparametric determinants of market liquidity

Abstract

We examine the factors influencing equity market liquidity through explainable machine learning techniques. Unlike previous studies, our approach is entirely nonparametric. By studying daily placement orders for equity securities managed by a European asset management institution, we uncover multiple nonlinear relationships between market liquidity and placement characteristics typically not captured by a traditional parametric model. As expected, the results show that liquidity tends to increase in highly active markets. However, we also note that liquidity remains relatively stable within certain trading volume ranges. Price volatility, broker efficiency, and the market impact of the trade are important predictors of liquidity. Price volatility shows a linear relationship with bid-ask spreads, whereas broker efficiency and market impact have non-symmetric convex effects. Large bid-ask spreads are linked to increased uncertainty and weak economic activity.