Price clustering of IPOs in the secondary market

Abstract

This article studies the integer price clustering of Initial Public Offerings (IPOs) in the secondary market trading during the first 240 trading days after their IPO dates. The results indicate the huge difference between the integer price frequency of IPOs in the primary market and that of matched stocks in the secondary market almost disappears on the first trading day after IPO. The integer price frequency of IPOs is still significantly higher than that of matched stocks during the first 240 trading days. However, after controlling for price level, trading characteristics and IPO price support, the integer price frequency of IPOs conforms to that of matched stocks and that those IPOs with integer offer prices have the same integer price frequency as IPOs without.

Notes

¹ Several studies provide evidences consistent with Harris' (1991) costly negotiation hypothesis (Harris, 1991; Chung et al., 2002, 2004, 2005; Cooney et al., 2003).

² This study focuses on NASDAQ IPOs because the majority of IPOs are listed on NASDAQ and most papers on aftermarket trading characteristics of IPOs focus on NASDAQ IPOs.

³ To ensure each close price is from an actual transaction, I recode the CRSP close prices that are less than or equal to 0 as missing since CRSP sets the close price negative if it is a bid–ask average and 0 if neither the close price nor the bid–ask average is available.

⁴ Unreported results on firm characteristics comparisons are available upon request.

⁵ Harris (1989, 1991) show that price clustering based on closing prices from CRSP is virtually identical to price clustering based on intraday transaction prices. Following Harris (1991) and Chiang and Harikumar (2004), I use CRSP closing prices to calculate integer frequency.

⁶ In unreported results, Augmented Dickey–Fuller (ADF) tests reject the null of existence of unit roots at 1% for the integer price frequency series of IPOs and matched stocks, independent of model specification on number of lags and trend. Following the Box–Jenkins procedure, I identify that both time series are autoregressive, AR(1) with trend. The time series of the integer price frequency difference between IPO and matched stocks is also stationary AR(1) with deterministic time trend.

⁷ The unreported results with 40-, 60-, 80-, 100- and 120-day sub-windows are similar.