我国创业板IPO首日高频量价分位相关的变点分析

Change point analysis to high-frequency price-volume quantile dependence of IPOs on listing date in China's growth enterprise market

提出含有结构变点的面板数据分位数回归模型,给出基于贝叶斯推理和MCMC模拟的参数估计方法,对我国创业板IPO上市首日的收益率与交易量、大盘走势之间的日内高频变结构相关性进行实证研究发现:在中高收益率分位点上结构变点均出现在开盘阶段附近,而在较低的分位点上变点位置向后推延至中盘附近;IPO交易量对收益率在变点之前的影响要大于在变点后的影响,而深证成分指数交易量对收益率在变点后的影响要大于在变点前的影响;IPO交易量对收益率的影响在变点前的较低分位点、变点后的任意分位点上均呈现出"量利分离"特殊现象;深证成分指数交易量与新股收益率在变点前后任意分位点上均呈现正相关,但影响系数呈现变点后大于变点前、极端分位大于中间分位的特征.

This paper proposes a new panel quantile regression model with change points at unknown positions, which parameters are estimated by Bayesian inference and MCMC sampling. By an empirical study to explore structural changes in the high-frequency dependence from trading volume and market trend to IPO＇s initial return in China＇s growth enterprise market, it is found that： structural change points are near the market opening for higher quantiles of returns, and close to the median trading time for lower quantiles; the IPO＇s trading volume gives a greater impact on IPO＇s return before change points than after change points on listing date, while the volume of Shenzhen composite index contributes larger impact on IPO＇s return after change points than before change points; the separation of return and volume of IPOs is detected both at lower quantiles before the change points and at any quantile after the change points; the volume of Shenzhen composite index is positively correlated to IPO＇s return through all quantiles before or after change points, moreover, the correlation coefficients after change points are stronger than ones before change points, and the same is true at extreme quantiles than at median quantiles.