Is Liquidity Risk Negligible?

Liquidity risk and its pricing have always been an important topic in financial research.It shows that liquidity affects asset returns through two different channels: liquidity level effect and liquidity risk effect.The level effect shows that illiquid stocks tend to earn higher average returns than liquid ones.The risk effect reveals that stocks with high exposure to shocks in market liquidity command a premium relative to stocks with low exposure,inline with the argument about the importance of liquidity risk in asset pricing.However,recent development in asset pricing has largely overlooked liquidity risk.For instance,the developments of the Fama-French five-factor model and the Hou-Xue-Zhang four-factor model do not consider liquidity risk.This is because they use the liquidity factor which constructed based on the price impact dimension as a representative and show that it yields little improvement in model performance.Nevertheless,due to the multi-dimensional characteristics of liquidity,there are other liquidity risk factors and the corresponding factor models proposed in the literature.Therefore,this thesis addresses the question:Is liquidity risk negligible for asset pricing if also considering other liquidity risk factors/models?Specifically,this study assesses the important role of liquidity risk in asset pricing through examining whether different liquidity risk factors or liquidity-risk-based models proposed in the literature perform differently,and whether the liquidity-risk-based models well explain the cross-section of stock returns in comparison with non-liquidity-risk-based pricing models.Using the latest model evaluation methods such as combination analysis,two-stage crosssectional regression analysis that allows for model misspecification,time series forecasting regressions,and model comparisons with the Sharpe ratio metrics,this paper involves the following works including examining whether the construction of the factor models conforms to the Merton's intertemporal capital asset pricing theory(ICAPM),testing the explanatory power of factor models for anomalous portfolios,comparing the Sharpe ratio metrics between the liquidity-riskbased models and the non-liquidity-risk-based models,and evaluating the additional explanatory power of the liquidity factor relative to the competitive factor models.The main conclusions are as follows:(1)According to the ICAPM theory,liquidity factor models based on different construction methods have different performances.The ICAPM theory requires that the proposed factors should price the cross-sectional asset returns,also the corresponding state variables should predict future investment opportunities with the consistent directions as the factor risk premiums.For the liquidity factor models examined,the liquidity factor constructed based on the trading discontinuity measure meets the ICAPM theory.For example,under the assumption that the model is misspecified,it produces a significant positive risk premium no matter which methods and test assets are used.Also,the related state variable positively predicts the future investment opportunities represented by the macroeconomic and the stock market variables with consistent signs,indicating the systemic nature of liquidity risk.In contrast,the risk premiums of liquidity factors based on different price influence dimensions are not significant in the cross-sectional tests,and the corresponding state variables do not have a robust forecasting ability for future investment opportunities,which are not consistent with the ICAPM criteria.(2)Liquidity factor models based on different constructions perform differently in capturing liquidity risk and other anomalous portfolios.By examining the portfolios grouped by different liquidity measures and various anomalies,we find that the liquidity factor constructed based on the price impact dimension shows limit contributions to improve the model's explanations by producing nearly zero asset loadings for portfolio returns,consistenting with the findings in the literature.On the other hand,the trading-discontinuity-based factor model can capture the liquidity premiums produced by different sorts,well explains the liquidity risk.Furthermore,the trading-discontinuity-based factor model shows good explanations for anomalous portfolios related to momentum,investment,profit,market ?,variance,and industry,and especially for small size portfolios.Therefore,the unimportance of liquidity risk the recent studies reveal might be an overgeneralization from the use of the price-impact-based liquidity factors in the model performance evaluation.(3)According to the maximum square Sharpe ratio of the models,the trading-discontinuitybased liquidity model shows a comparable performance with the popular asset pricing models.Different from the common comparison methods,the model comparison based on the maximum square Sharpe ratio metric does not depend on the test assets,and therefore obtains a consistent conclusion.According to the method,the trading-discontinuity-based factor model performs similarly with the popular Fama-French five-factor model and the Hou-Xue-Zhang four-factor model,and the corresponding liquidity factor produces the largest marginal contribution among all the competitive factors,implying that the liquidity risk is an important source of asset pricing.(4)Compared with the benchmark factor models,the liquidity factor shows additional explanatory power.The impact of liquidity risk on asset prices is related to the economic cycle.During periods of market pressure and turbulence,liquidity risk has a greater impact on asset prices.Consistent with economic intuitions,after controlling for the benchmark factors,the factor constructed based on the trading discontinuity of liquidity generates a significant premium during the period of market downturn and market turmoil.Furthermore,in various market conditions and different sub-periods,the information contained in the trading-discontinuity-based liquidity factor cannot be explained by all benchmark factor models.This evidence confirms that liquidity risk is different from existing risk sources and cannot be captured by other factor models.