Research On Equity Duration And Market Risk Based On Machine Learning And Information Theory

Equity duration and equity term structure is an emerging financial research field in recent years.They have an important impact on asset pricing and risk management.They can be used as macro indicators to observe the economic and financial crisis situation,and can also be used as micro indicators to measure the future cash flow of firms and investors’ preference for the term structure of cash flow.However,few articles have studied the equity duration or the equity term structure in the Chinese market.In addition,we find that the existing literature methods on measuring equity duration do not perform well in the Chinese market.To solve this problem,by applying machine learning model and analyzing the characteristics of algorithm and data,we propose to predict firms’ future cash flow by applying a one-dimensional convolutional neural network model,so as to calculate the equity duration.This neural network method does not have the problem of existing methods.As the degree of underfitting decreases,we find that the return of the duration long short portfolio will increase,which indicates that the return of the duration long short portfolio is positively correlated with the accuracy of the equity duration forecasting.Then,we select the optimal hyperparameter combination through grid search method,and make an empirical study on the duration under this parameter combination.After obtaining the equity duration,we further test the performance of the duration portfolio empirically.The empirical result is that the long-short portfolio constructed in China’s A-share market based on this equity duration has significant returns.And after size and risk adjustments including the capital asset pricing model,Fama French threefactor model and China three-factor model are controlled for,the long-short portfolio returns are still significant.In both raw and risk-adjusted returns,the implied equity term structure is on averagedownward sloping,in line with the term structure characteristics of other markets such as the United States.Moreover,according to the change of beta value in risk adjustment along with the duration,it is found that the characteristics of the duration portfolio are consistent with those of growth stocks and value stocks documented in the existing literature.The longer the duration of the stock,the greater the beta of market excess return,and to some extent can be explained by book-to-market ratio factor.Based on the good orthogonality between the duration portfolio and the market capitalization portfolio,we construct a new value factor named SML by using equity duration and the factor construction method in the China three-factor model.We then find that,SML outperforms the book-to-market ratio factor of Fama French three-factor model and the price-to-earnings ratio factor of China three-factor model.Through the equity duration proposed in this paper,we study the impact of COVID-19 on Chinese equity market.It is found that in terms of the changes in equity duration,the reaction of China’s stock market to COVID-19 is not significant,because there is no obvious reversal of the excess return of long-short portfolio,and the trend of short duration portfolio and long duration portfolio is roughly the same.