A new Markov regime‐switching count time series approach for forecasting initial public offering volumes and detecting issue cycles

Abstract

AbstractThis study proposes a novel Markov regime‐switching negative binomial generalized autoregressive conditional heteroskedasticity model for analyzing count data time series. We develop a likelihood‐based method for parameter estimation and give the one‐step‐ahead forecasting algorithms for the mean, variance, and quantiles. An empirical analysis of both the U.S. initial public offering (IPO) and Chinese A‐share IPO markets indicates that our method is very efficient in forecasting monthly IPO volumes and detecting hot/cold issue markets. The first‐day IPO return is positively correlated with the IPO volume in a hot issue market but negatively correlated with the IPO volume in a cold issue market, in both the U.S. and Chinese IPO markets. However, the average first‐day return in the previous hot issue market has a significant positive impact on the current IPO volume for only the U.S. IPO market. Our approach helps to more accurately model and understand the behavior of hot/cold IPO issue markets.