DOCX

IPO Waves, Information Spillovers, and Analyst Biases

By Tiffany Dixon,2014-03-28 00:54
11 views 0
IPO Waves, Information Spillovers, and Analyst Biasesand,IPO,Waves

    IPO Waves, Information Spillovers, and Analyst Biases*

    a,b,cSusan Christoffersen

    b Amrita Nain

    dYa Tang

    September 2010

_________________________________

    *Previously circulated under the title ―IPO Cycles, Firm Characteristics, and the Role of Underwriters‖ a Rotman School of Management, University of Toronto, 105 St. George St., Toronto, ON, M5S 3E6 b Desautels Faculty of Management, McGill University, 1001 Sherbrooke St.West, Montreal, QC H3A 1G5 c Copenhagen Business School, Solbjerg Plads, DK-2000 Frederiksberg d Guanghua School of Management, Peking University, 5 Summer Place Road, Beijing, 100871, P.R. China

    Contact author: Amrita Nain, amrita.nain@mcgill.ca, 514-398-8440.

    Susan Christoffersen is grateful for support from SSHRC, IFM2, and the Leibovitch Award. Amrita Nain is grateful for support from SSHRC and FQRSC. The authors are grateful for helpful comments from Michelle Lowry, J. Ari Pandes, and participants of the Northern Finance Association Meetings, 2010. All

    remaining errors are our own.

    IPO Waves, Information spillovers, and Analyst Biases

    Abstract

    Existing research suggests the presence of learning and information spillovers in IPO waves. Consistent with information spillover models, we find that the level of information asymmetry surrounding IPO firms is significantly higher in the early stages of an IPO wave than in the late stages. Thus, the early stages of an IPO wave are characterized by high valuation levels as well as high valuation uncertainty. According to current thinking, private firms learn about IPO valuations by observing the price outcomes of recent IPOs. We propose a more direct method for underwriters to convey this information to potential issuers: affiliated analysts. We show that affiliated analysts issue excessively optimistic recommendations early in the wave, but not later in the wave. Institutional holdings are unaffected by the early-stage bias of affiliated analysts and the stock market appears to discount affiliated analysts in the early stages of a wave. We suggest that the affiliated-analyst bias in the early stages of an IPO wave is an attempt to convey positive demand information to private firms.

1. Introduction

    It is a well-established fact that there are pronounced cycles in IPO volume and

    1initial returns. Recent theoretical and empirical work indicates the existence of learning and information spillovers during IPO waves. Lowry and Schwert (2002) find that more companies file IPOs following periods of high underpricing because high initial returns reflect positive information learned during the registration period. Evidence of information spillovers can be found in Benveniste, Ljungqvist, Wilhelm, and Yu (2003) (henceforth, Benveniste et. al. (2003)) who show that the decision to complete an IPO and the terms of the IPO depend on the experience of contemporaneous IPOs in the same industry. More recently, Alti (2005) presents a model of endogenous information spillovers in which firms with higher growth opportunities go public earlier in the wave when information asymmetry is high.

    A central tenet of the information spillover argument is that information asymmetry declines as an IPO wave progresses. The price outcome of each additional IPO reveals information about market demand and common valuation factors. Although several existing papers discuss the link between underpricing and information asymmetry, there is a dearth of evidence on the dynamics of information asymmetry within an IPO

    2wave. Thus, the first objective of this paper is to examine a key feature of the information spillover argument does uncertainty surrounding IPO firms decline during

    the course of an IPO wave? Using different methods for identifying spurts in IPO volume

    1 See, for example, Ibbotson and Jaffe (1975), Ibbotson, Sindelar, and Ritter (1988, 1994), Lowry (2003). 2 The positive correlation between underpricing and information asymmetry is discussed in Beatty and Ritter (1986), Rock (1986), Grinblatt and Hwang (1989), Benveniste and Spindt (1989), Michaely and Shaw (1994), and Lowry, Officer, and Schwert (2010).

     1

    and several different proxies of valuation uncertainty, we show that valuation uncertainty is higher in the early stages of an IPO wave than in the late stages.

     Having provided evidence that the informational environment is opaque at the beginning of an IPO wave, we examine the mechanism of information spillovers. Like other papers, we find extensive empirical evidence that price realizations early on in the wave provide information to the market about valuations and demand. However, we explore an alternative and more direct method for investment banks to communicate their private information to firms: affiliated analysts. During the road show, issuing firms and underwriters obtain information about investor demand and valuations. Underwriters would like to convey positive information gleaned during the registration period to private firms in order to build up the supply of IPOs. Investment banks can signal high market valuations by encouraging their analysts to provide favorable views of IPO performance. Accordingly, we expect the views of affiliated analysts to be biased early in the IPO wave when investment banks are responding to new demand information and building up the supply of IPO offerings.

    Consistent with this idea, we find strong evidence of an affiliated analyst bias early in the wave, but not later on. Specifically, we show that in the early stages of an IPO wave, when valuations are high but information quality is poor, analysts affiliated with underwriters provide disproportionately more positive recommendations to issuing firms as compared with unaffiliated analysts. Later in the wave, when valuations and information asymmetry are much lower, the recommendations of affiliated and unaffiliated analysts are similar. In a related test, we divided IPOs into two mutually exclusive groups - IPOs that receive a stronger recommendation from affiliated analysts

     2

than from unaffiliated analysts (termed Affiliated Favored IPOs), and IPOs that receive a

    stronger recommendation from unaffiliated analysts (termed Unaffiliated Favored IPOs).

    We find that there are significantly more Affiliated Favored IPOs in the early stages of an IPO wave than in the late stages.

     It is well accepted that analysts affiliated with underwriters issue more favorable

    3 This bias is generally viewed in a negative light as an attempt by recommendations.

    underwriters to appease issuing clients and institutional investors who are long the stock. In contrast, we suggest that the early-stage analyst bias is an attempt by underwriters to convey positive demand information to potential issuers during periods of high information asymmetry. However, we consider two potential alternative explanations for the affiliated-analyst bias in the early stages of an IPO wave. First, we explore whether affiliated analysts provide more favorable recommendations to early issuers than unaffiliated analysts because they have private information about the higher quality of early issuers. However, we find no significant differences in profitability, sales growth, cash holdings, liquidity, and abnormal stock returns of early and late issuers two years after issue date. Thus, there are no notable quality differences that would justify the more favorable outlook of affiliated analysts early in an IPO wave.

    The second alternative explanation we explore is that the analyst bias early in the wave is a mechanism to attract institutional demand rather than an attempt to signal positive valuations to private firms. To test this alternative explanation, we investigate how institutional holdings respond to analyst recommendations. Institutional investors need to trade-off two features of affiliated analysts. On the one hand, affiliated analysts

     3 See, for example, Michaely and Womack (1999) and Bradley, Jordan, and Ritter (2008).

     3

    are likely to have an informational advantage over unaffiliated analysts. On the other hand, their recommendations may be biased due to pressure from the investment banking divisions. Our tests reveal that institutional investors optimally balance these competing factors and do not respond to the affiliated analysts bias early in the wave.

    Specifically, we find that in the early stages of an IPO wave, when affiliated analysts provide biased recommendations, institutional holdings of Affiliated and Unaffiliated Favored IPOs are similar. Later in the wave, when the analyst bias is absent, institutional holdings are significantly higher in Affiliated Favored IPOs (38.9%) than in Unaffiliated Favored IPOs (30.69%). Thus, institutions pay more attention to affiliated analysts later in the wave, when they are not biased and the recommendations reflect private information about the firms. We also find that within the subset of Unaffiliated Favored IPOs, institutions hold 39.3% of early issuers and 30.69% of late issuers. Institutional investors are, therefore, more likely to follow the recommendations of unaffiliated analysts early in the wave because affiliated analysts are overly optimistic at that time.

    Since institutional investors appear to be unmoved by the affiliated-analyst bias early in the wave, we also explore the possibility that the analyst bias is targeted to influence uninformed investors. We follow existing literature and calculate cumulative abnormal returns (CARs) at the announcement of strong-buy recommendations. We find that the market reacts positively to strong-buy recommendations from both affiliated and unaffiliated analysts. However, the CARs of early issuers are higher when unaffiliated analysts issue strong-buy recommendations than when affiliated analysts issue strong buy recommendations. In contrast, CARs of late issuers are higher when affiliated analysts

     4

    issue strong-buy recommendations than when unaffiliated analysts issue strong buy recommendations. These stock market reactions indicate that the market pays more attention to affiliated analysts‘ opinions of late issuers and to unaffiliated analysts‘ opinions of early issuers.

    Overall, the affiliated analyst bias early on in the wave is not successful in convincing institutions to increase their holdings of early issuers and appears to be discounted by the market as a whole. Since investors appear to be unmoved by the early-stage bias of affiliated analysts, the bias is possibly targeted at a different audience. We propose that the affiliated analyst bias serves as a mechanism of conveying information learned by underwriters during the book building process to the pool of private firms that are considering an IPO but are uninformed about IPO demand and valuations.

    Our paper makes several contributions to existing literature. Recent theory suggests that firms issuing early in a wave are likely to be different from firms that issue late in a wave. Ours is one of the first papers to show empirically that issuer characteristics proxying for valuation certainty change during the course of an IPO wave. We are also the first to demonstrate that the bias of affiliated analysts dissipates as an IPO wave progresses and that sophisticated investors like institutions know when to ignore affiliated analysts and when to follow their advice. Our results suggest that the analyst bias plays a positive role of disseminating information to the supply side of the IPO market. Finally, given the high correlation between IPO volume and IPO initial returns, existing literature tends to use high volume and high initial returns as substitute measures of IPO heat. Our results show that IPO wave periods are not uniformly hot and that

     5

    treating all issuers in an IPO wave as a homogenous group can result in a significant loss of information.

    Our paper adds to a small set of empirical studies that take into account the order of moves within corporate event waves. Benveniste, et. al. (2003) provide evidence that information spillovers influence issuers‘ decision on whether to complete an IPO and how to price it. They examine industry IPO waves and show that information spillovers are stronger among pioneers (first movers) and early followers in an industry than among late followers. Çolak and Günay (2010) find that the average quality of issuing firms is lower in the early stages of rising IPO cycle. Bouwman, Fuller, and Nain (2009) examine mergers and acquisitions and show that acquirers that buy early in a merger wave perform worse than those that purchase later in a merger wave.

    The rest of the paper is organized as follows. Section 2 describes the data, Section 3 discusses IPO characteristics and valuation uncertainty, Section 4 examines the analyst bias, and Section 5 studies issuer quality. In Section 6, we present an analysis of institutional holdings and market reactions. Section 7 discusses robustness and Section 8 concludes.

2. Data Description

2.1. Sample

    To identify IPO waves, we use the 12,648 initial public offerings between January stst41, 1970 and December 31, 2005 provided by Jay Ritter. To obtain offer-specific and

     4 The data are available on Jay Ritter’s website at http://bear.warrington.ufl.edu/ritter/ipodata.htm

     6

    firm-specific information, we rely on Securities Data Company (SDC). This dataset include 11,490 firm-commitment IPOs between 1970 and 2005. We exclude unit offerings, real estate investment trusts (REITs), closed-end funds, and ADRs, and further restrict our analysis to stocks with information available on CRSP. If a firm‘s stock price

    appears on CRSP more than 14 days after the IPO date, we drop it from the sample. This leaves us with a final sample of 7,043 IPOs. All the analyses in this paper are conducted on this sample of 7,043 IPOs.

    2.2. Classification of IPO waves

    Our primary method of identifying IPO waves follows the simulation

    5methodology of Harford (2005). Since each of the four decades in our sample period is characterized as a distinct era, we do the simulation separately for the 1970s, 1980s, 1990s and 2000s. From the sample of 12,648 IPOs, we take the total number of IPOs in each decade and create a simulated distribution by randomly assigning each actual IPO to a quarter within the decade. The probability of assignment is 1/40 for each quarter. We repeat this process 1000 times to generate 1,000 randomly drawn series of quarterly IPO volume. Any quarter where the actual number of IPOs exceeds the 95th percentile from the simulated distribution is designated as a high-volume quarter. Quarters where the

    thactual number of IPOs lies below the 5 percentile are designated low-volume quarters.

     5 As robustness check, we also identify IPO waves using a methodology based on Helwege and Liang (2004). Details of this method and the robustness of our results are provided in Section 7.

     7

    This method gives us 38 high-volume quarters and 28 low-volume quarters. Since it takes time for a wave to form, peak, and disappear, we define an IPO wave as a period of three or more consecutive high-volume quarters. This method results in 6 IPO waves

    ststbetween January 1, 1970 and December 31, 2005. Out of the 38 high-volume quarters

    identified above, 5 are not included in our waves because the high volume period did not last long enough. Table 1 provides summary statistics of the 6 IPO waves. Panel A shows the start and end date of all six waves along with the total number of IPOs and the number of IPOs per quarter in each wave. A total of 3,052 IPOs occurred during IPO wave periods with an average of 99 IPOs per quarter. Panel B presents the number of IPOs, dollar volume of IPOs, initial returns and oversubscription in the first and last month within in each IPO wave. Initial returns, IR, are calculated as the first-day closing

    price less the offer price divided by the first-day closing price. First-day closing prices are collected from CRSP daily file and offer prices are obtained from SDC. The percentage of oversubscription, OVERSUBPT, is calculated as shares offered and

    oversold minus shares filed divided by shares filed. Panel B shows that IPO volume as measured by number of IPOs and dollar volume of IPOs is high in both the first and the last months of the six waves. This is to be expected since our classification of wave periods is based on the number of IPOs. However, underpricing and oversubscription is noticeably lower in the last month of the IPO waves than in the first month. Thus, by the end of a wave period, volume may still be high but IPO demand and underpricing have dropped significantly. This is consistent with Lowry and Schwert‘s finding that initial returns lead volume by a few months. The low initial returns at the end of a wave period are indicative of the imminent drop in IPO activity. The difference in initial returns and

     8

Report this document

For any questions or suggestions please email
cust-service@docsford.com