teacher steve

fargm

IPOunderpricinginChinaa-sharemarket.pdf

Home >Computer Science homework help >teacher steve

An Empirical Examination of IPO Underpricing in Chinese

A-Share Market

YU TING

NATIONAL UNIVESITY OF SINGAPORE

2003

An Empirical Examination of IPO Underpricing in Chinese

A-Share Market

YU TING

(Master of Social Sciences), NUS

A THESIS SUBMITTED

FOR THE DEGREE OF MASTER OF SOCIAL SCIENCES

DEPARTMENT OF ECONOMICS

NATIONAL UNIVESITY OF SINGAPORE

2003

ACKNOWLEDGEMENTS I wish first of all to express my heartfelt gratitude and respect to my supervisor,

Professor Tse Yiu Kuen, who gave me untiring guidance, constructive suggestions

and valuable critique through all stages of this research.

Special thanks to my wonderful husband, Ni Houming, for his strong moral and

technical support. He was also a diligent proofreader. I am also grateful to my beloved

parents, my husband’s parents, and other family members for their encouragement,

care and love all the way.

Many other persons were helpful in the preparation stage of the study. Among them

are Huang Yizhi, who provided me part of the data needed, and Luo Lei who gave me

useful suggestions on data processing.

Finally the research fund and resources provided by the National University of

Singapore are highly appreciated.

Yu Ting

Singapore

December 2003

TABLE OF CONTENTS

ACKNOWLEDGEMENTS...................................................................... I

TABLE OF CONTENTS ........................................................................II

SUMMARY .......................................................................................... IV

LIST OF TABLES................................................................................. VI

LIST OF FIGURES .............................................................................VIII

CHAPTER 1 INTRODUCTION ............................................................. 1

1.1 MOTIVATION OF THE STUDY ................................................. 1

1.2 OBJECTIVES OF THE STUDY ...................................................... 3

1.3 CONTRIBUTION OF THE STUDY ............................................... 4

1.4 STRUCTURE OF THE STUDY. ..................................................... 5

CHAPTER 2 MODELS OF IPO UNDERPRICING AND A

SURVEY OF CHINESE PRIMARY MARKET .................................... 6

2.1 MODELS OF IPO UNDERPRICING .............................................. 6

2.2 FEATURES OF THE CHINESE PRIMARY MARKET .............. 16

2.3 PRIOR STUDIES OF THE CHINESE IPO UNDERPRICING .... 23

2.4 POSSIBLE EXPLANATIONS FOR CHINESE A-SHARE IPO

UNDERPRICING ................................................................................. 26

CHAPTER 3 HYPOTHESES AND METHODOLOGY ...................... 31

3.1 THE WINNER’S CURSE MODEL ............................................... 31

3.2 EX ANTE UNCERTAINTY .......................................................... 33

3.3 THE SIGNALING MODEL ........................................................... 39

CHAPTER 4 DATA AND EMPIRICAL RESULTS............................ 48

4.1 DATA AND UNDERPRICING ..................................................... 48

4.2 ALLOCATION AND ADVERSE SELECTION ........................... 54

4.3 EX ANTE UNCERTAINTY .......................................................... 56

4.4 THE SIGNALING MODEL ........................................................... 58

CHAPTER 5 CONCLUSIONS ........................................................... 67

APPENDIX A: OFFERING MECHANISM CHANGES IN CHINA .. 69

APPENDIX B CORRELATION MATRIX .......................................... 72

APPENDIX C TEST OF THE WINNER'S CURSE MODEL.............. 73

BIBLIOGRAPHY ................................................................................. 74

iii

SUMMARY

Much evidence suggests that initial public offerings (IPOs) of common stocks are

systematically priced at a discount to their subsequent initial trading price. The large

underpricing magnitude in the Chinese IPO market has attracted much attention. Mok

and Hui (1998) report an underpricing of 289% for a sample of 87 Shanghai IPOs

listed from 1990 to 1993. Su and Fleisher (1999) find the underpricing level as high as

948.6% for Chinese A-share IPOs before January 1, 1996. A more updated report is

from Tian (2003), who found an average of 267% of initial returns for IPOs from

1991 through 2000. These reported underpricing levels in the Chinese market are

much higher than the average level of 60% in the emerging markets (Jenkinson and

Ljungqvist, 2001). Despite many studies on the Chinese IPO underpricing, few studies

have been done to investigate the reasons in light of classical IPO underpricing

theories. Although previous studies such as Mok and Hui (1998), Su and Fleisher

(1999), and Chau et al (1999) have explored some reasons for the high IPO

underpricing, most of the studies examine a few aspects that may affect IPO

underpricing. For many markets, whether developed or emerging, IPO underpricing

may be explained in terms of some classical IPO underpricing models such as

asymmetric information models, institutional explanations and ownership and control

(see Jenkinson and Ljungqvist, 2001). Tests of the Chinese IPO underpricing against

classical IPO underpricing models are, however, far from comprehensive. This paper

attempts to shed light on this issue by examining some classical models of IPO

underpricing for the Chinese market, especially some hypotheses not studied before.

The classical IPO underpricing models examined in this study are the winner’s curse

model (Rock, 1986), ex ante uncertainty hypothesis (Ritter, 1984; Beatty and Ritter,

1986) and the signaling model (Allen and Faulhaber, 1989; Grinblatt and Hwang,

1989; Welch, 1989, 1996). Among those tested classical models, the winner’s curse

model has not been tested before. The ex ante uncertainty hypothesis was tested by

Mok and Hui (1998), but they test only one proxy of ex ante uncertainty, i.e. the

inverse of new funds raised. We use three proxies-the standard deviation of after-

market returns, the offer size and the age of firms, to examine the ex ante uncertainty

hypothesis. In examining the signaling model, we test eight key empirical implications

of the signaling model, some of which have been examined in Su and Fleisher (1999),

but the methodology adopted and the conclusion made are different.

Using data from November 1995 to December 1998, our results show that the

winner’s curse hypothesis is the main reason for the high IPO underpricing in China.

The signaling hypothesis does not stand in the Chinese market during the sample

period.

LIST OF TABLES Table 2.1 Main Underwriters and their performance (1991-2001).............................. 18

Table 2.2 Prior studies of the Chinese IPO underpricing ............................................ 25

Table 2.3 The Structure of Domestic Investors in 1998 .............................................. 27

Table 4.1Descriptive statistics on 343 IPOs in the 1996-1998 period and 215 SEOs in

the period 1996-2001 ........................................................................................... 49

Table 4.2 Distribution of 343 fixed pricing IPOs and 215 first seasoned equity

offerings (SEOs) by offering year, 1996-2001 .................................................... 50

Table 4.3 Initial returns in IPOs, with adjustment for allocation................................. 51

Table 4.4 Statistics of initial returns and PE ratios by years and by stock exchanges. 53

Table 4.5 Statistics of allocations in sample IPOs ....................................................... 54

Table 4.6 OLS regression Analysis Investigating Ex Ante Uncertainty and other

Significant Explanatory Variables of IPO Underpricing ..................................... 57

Table 4.7 OLS regression to test Leland and Pyle’s theoretical signaling model ....... 59

Table 4.8 First OLS regression to test Grinblatt and Hwang’s Bivariate Signaling

Model ................................................................................................................... 59

Table 4.9 Second OLS regression to test Grinblatte and Hwang’s Bivariate Signaling

Model ................................................................................................................... 60

Table 4.10 First regression of Housman test for the exogeneity of variable V ........... 60

Table 4.11 Logit Model to Test the relation between underpricing and the likelihood

of SEO.................................................................................................................. 62

Table 4.12 Tobit Regression to Examine the relationship between Time SEO and IPO

Unperpricing ........................................................................................................ 63

Table 4.13 Tobit Regression to Examine the relationship between SEO Size and IPO

Unperpricing ........................................................................................................ 64

Table 4.14 OLS Regression to Test the Price Reaction at the Announcement of SEO66

Table a: Statistics on the allocation methods adopted in the Chinese A-share market

from 1990 through 2000 ...................................................................................... 71

Table b: Correlation matrix of continuous explanatory variables in equation (4) ....... 72

Table c: Correlation matrix of continuous explanatory variables in equation (6) ....... 72

Table d: Correlation matrix of continuous explanatory variables in equation (10) ..... 72

Table e: Correlation matrix of continuous explanatory variables in equation (12) ..... 73

Table f: OLS Regression to Test the Winner's Curse Model....................................... 73

vii

LIST OF FIGURES

Figure 4.1 The distribution of the initial excess return in IPOs........................................... 52

Figure 4.2 The distribution of allocations to investors in IPOs ........................................... 55

viii

Chapter 1 Introduction

1.1 Motivation of the Study

Much evidence suggests that initial public offerings of common stock (IPOs)

are systematically priced at a discount to their subsequent trading price (for

review of international evidence, see Jenkinson and Ljunqvist (2001)). In

attempting to explain the puzzle, many academic researchers have formulated

different models. But no single explanation can account for the apparent

underpricing of new issues in all the stock markets. Even within one market,

one model on its own might not be sufficient to account for the full extent of

IPO underpricing.

The last decades have seen phenomenal growth in the Chinese stock market

both in the number o firms traded and dollar volume of shares traded,

especially after early 1990s when the two stock exchanges were established

(The Shanghai Stock Exchange (SHSE) in December 1990 and the Shenzhen

Stock Exchange (SZSE) in July 1991). ). As of December 2002, there are more

than one thousand companies listed on the two exchanges, with total market

capitalization equal to about 50 percent of China’s gross domestic product

(GDP). The combined market capitalization of the two stock exchanges has

reached RMB 1 3,832.9 billion and the negotiable share capital hits RMB

1,248.5 billion (more than that of Hong Kong, 1116.66 million HK$2). The

1 RMB is the abbreviation for Renminbi, which is the basic unit for Chinese currency. RMB has been pegged the US dollar at the exchange rate of about RMB 8 per US$1 during the sample period being studied. 2 The exchange rate from Hong Kong dollar to RMB is approximately 1.

size of the Chinese stock market has become comparable to those of the

industrialized countries and thus cannot be ignored (Allen and Gale 1995). In

addition, China joined the world trade organization (WTO) in November 2001.

Opening up its securities market has been put into the schedule of the Chinese

government. So an understanding of the characteristics and performance of the

Chinese IPO market would be of great value for investors and scholars at home

and abroad.

The large underpricing magnitude in the Chinese IPO market has also attracted

great attention. Mok and Hui (1998) report an underpricing3 of 289% for a

sample of 87 Shanghai IPOs listed from 1990 to 1993. Su and Fleisher (1999)

find the underpricing level as high as 948.6% for Chinese A-share IPOs before

January 1, 1996. A more updated report is from Tian (2003), who found an

average of 267% of initial returns for IPOs from 1991 through 2000. These

reported underpricing levels in the Chinese market are much higher than the

average level of 60% in the emerging markets (Jenkinson and Ljungqvist,

2001). Despite many studies on the Chinese IPO underpricing, few studies

have been done to investigate the reasons in light of classical IPO underpricing

theories. Although previous studies such as Mok and Hui (1998), Su and

Fleisher (1999), and Chau et al (1999) have explored some reasons for the high

IPO underpricing, most of the studies examine few aspects that may affect IPO

underpricing. For many markets, whether developed or emerging, IPO

underpricing may be explained in terms of some classical IPO underpricing 3 Underpricing is defined as the pricing of an IPO at less than its market value. A possible measure of the degree of underpricing is (MV−P0)/MV, where P0 is the offer price and MV is the firm’s per-share market value on the offering date. Since MV is unknown on the offering date, many researchers use the initial return, (P1−P0)/P0, where P1 is the first-day closing price, as a measure of underpricing. We shall adopt this terminology in this paper.

models such as asymmetric information models, institutional explanations and

ownership and control (see Jenkinson and Ljungqvist, 2001). Tests of the

Chinese IPO underpricing against classical IPO underpricing models are,

however, far from comprehensive. This paper attempts to shed some light on

this and examines a list of classical models of IPO underpricing for the

Chinese market using data from November 1995 to December 1998.

1.2 Objectives of the study

Given the above motivation, this study has two major objectives.

The first objective is to record the level of underpricing for IPOs in China over a

relatively more current period. With this in mind, IPOs are examined in this study

over period from November 1995 to December 1998. To filter out effects of offering

methods on underpricing, I examine only the most commonly used online fixed

pricing offerings (Shang Wang Ding Jia) in China. In total, 343 IPOs are analyzed

over the period of interest.

The second and perhaps more important objective of this study is to investigate

possible explanations for the level of underpricing recorded, across various issues.

This second area of study draws largely upon the existing models of IPO underpricing.

I review the theoretical IPO underpricing models and analyze possible model

explanations for the cross sectional difference in Chinese IPO underpricing. Based on

possible explanation models and related literature, hypotheses are formulated and

tested. The classical IPO underpricing models examined in this study are the winner’s

curse model (Rock, 1986), ex ante uncertainty hypothesis (Ritter, 1984; Beatty and

Ritter, 1986) and the signaling model (Allen and Faulhaber, 1989; Grinblatt and

Hwang, 1989; Welch, 1989, 1996).

By investigating the IPO phenomenon in the Chinese market, I hope to provide further

insights on international IPO underpricing.

1.3 Contribution of the study

Among those tested classical models, the winner’s curse model4 has not been

tested before. The ex ante uncertainty hypothesis was tested by Mok and Hui

(1998), but they test only one proxy of ex ante uncertainty, i.e. the inverse of

new funds raised. We use three proxies-the standard deviation of after-market

returns, the offer size and the age of firms, to examine the ex ante uncertainty

hypothesis. In examining the signaling model, we test eight empirical

implications of the signaling model, some of which have been examined in Su

and Fleisher (1999), but the methodology adopted and the conclusion made are

different.

My results show that investors’ high ex ante uncertainty about firm’s value and

the winner’s curse problems are the main reasons for the high IPO

underpricing in China. But the signaling hypothesis does not stand in the

Chinese market during the sample period.

Given the prominence of the Chinese stock market in the emerging markets,

the results should be able to shed some light on IPO underpricing for other

4 Wu (2001) finds a positive correlation between underpricing and allocation rate in China, which is in support of the winner’s curse model. However, other key implications of the winner curse model were not tested. Therefore it can not be considered as a complete test of the Winner’s Curse’s model in the Chinese market.

emerging markets and provide further insights on international IPO

underpricing. The results add more evidences on testing of the winner’s curse

and signaling model as well. This should be illuminating and of value to both

academicians and practitioners.

1.4 Structure of the study.

The rest of this paper is organized as follows. Chapter 2 summarizes the theoretical

literature on IPO underpricing and provides a survey on Chinese primary market.

Chapter 3 formulates the hypotheses to be examined and methodology adopted.

Chapter 4 describes data and reports empirical results. Chapter 5 summarizes and

concludes.

Chapter 2 Models of IPO Underpricing and

a Survey of Chinese Primary market

2.1 Models of IPO underpricing

It has been a well-know empirical regularity in the IPO market that companies

apparently underprice their shares when going public. Previous studies have shown a

phenomenon of underpricing in virtually every country. The first day premium that

investors experience is on average more than 15 percent in industrialized countries

and around 60 percent in emerging markets (Jenkinson and Ljunqvist, 2001).

In an efficient and perfect market, theory suggests, companies should not ‘leave

money on the table’, certainly not in such large quantities. In trying to explain why

firms are floated at too low a price, researchers have generated a large theoretical and

empirical literature. Jenkinson and Ljunqvist sum up most of the studies on IPO

underpricing in their book Going Public (Second Edition 2001). Briefly the IPO

underpricing models include the following (refer to the book for details of each

model):

2.1.1 Asymmetric Information

The most important modern theories of IPO underpricing arise from important

informational asymmetries between market participants, the issuing firm, the

underwriting distribution syndicate, the initial buyers and the larger set of investors in

the secondary market.

Most models of IPO pricing typically assume one group has superior information on

firm value. Other agents know this and behave accordingly. Further everyone knows

that everyone knows this, and so on ad infinitum.

There are four informational assumptions one might make which accordingly lead to

the four underpricing models.

a) Assume a small group of investors has information superior to that of other

investors and the issuer

Rock’s (1986) asymmetric information model assumes that there are two groups of

potential investors in the IPO markets: (1) ‘informed’ investors, those prepared to

incur evaluation costs to assess the after-market performance of the offering and bid

only of attractively priced IPOs; (2) ‘uninformed’ investors do not commit resources

to acquire information and apply for every new issue coming into the market

indiscriminately. Thus uninformed investors face competition for good shares, but

have a higher probability of obtaining bad shares due to the rationing mechanism

applied to oversubscribed offerings. Rock argues that the bias in rationing produces an

equilibrium offer price with a finite discount sufficient to attract uninformed investors

to the issue (assuming that the primary market is dependent on the continued

participation of uninformed investors, in the sense that informed demand is

insufficient to take up all shares on offer even in attractive offerings). This does not

remove the allocation bias against the uninformed – they will still be crowded out by

informed investors in the most underpriced offerings – but they will no longer make

losses on average, even after adjusting for rationing. This gives rise to the “Winner’s

Curse” or “Adverse Selection” models. Implicit in the winner’s curse model is the

notion that, if properly adjusted for risk and rationing, uninformed investors’

abnormal returns are zero, on average – that is just enough to ensure their continued

participation in the market. Similarly, the informed investors’ conditional underpricing

return should just provide a normal return on their information production. While the

former is potentially testable, the latter is not, not least because informed and

uninformed investors cannot in practice be distinguished. Moreover, very few

markets publish enough allocation data to allow underpricing returns to be adjusted

for rationing. The evidence from countries use fixed price rather than book-building

mechanisms, mostly supports the presence of a winner’s curse: in Singapore, the UK,

and Finland initial returns do indeed tend to be zero when adjusted for rationing.

b) Assume the issuer has better information on securities value than do the

underwriter or investors.

If the issuing firm is better informed about the present value and risk of its future cash

flows than are investors or underwriters, underpricing may become a mean of

convincing potential buyers of the “true” high value of the firm, i.e. underpricing as a

signal of firm quality. Allen and Faulhaber (1989), Grinblatt and Hwang (1989), and

Welch (1989, 1996) have contributed theories of this underpricing signaling model.

They hypothesize that underpricing allows “good” firms to distinguish themselves

from “bad” firms and to improve terms of future external financing.

Under this assumption, good quality issuers are assumed to maximize the expected

proceeds of a two-stage sale: they sell a fraction of the firm at flotation and the

remainder in a seasoned equity offering, henceforth, SEO. In the words of Ibbotson

(1975), issuers underprice in order to ‘leave a good taste in investors’ mouths’. With

some positive probability, a firm’s true type is revealed before the post-IPO financing

stage, introducing the risk to low-quality issuers that any cheating on their part will be

detected before they can reap the benefit from the signal. This makes separation

possible, in that it decreases the expected benefit from signaling to low-value firms

and thus drives a wedge between high-value and low-value firms’ marginal signaling

cost. Signaling true value is beneficial to a high-value company as it allows a higher

price to be fetched at the second-stage sale if separation is achieved.

c) Assume underwriters /distributors possess information superior to the issuer

In the previous two models, underwriters do not have any particular role and thus

potential agency problems between the underwriter managing the floatation and the

issuing firm are ignored. Now if underwriters are better informed about investor

demand than issuers, underwriters may earn information rents in an imperfectly

competitive underwriter market, for instance in the form of sub-optimal selling effort.

When the underwriter has valuable private information on market demand, the issuer

will wish to learn this information. But the issuer must offer incentives to underwriter

to truly reveal it. In order to secure truthful revelation of private information and

encourage promotion efforts, the issuers may agree to a contract that leads to

underpricing. This leads to the Principal-Agent models of the IPO.

Baron and Holmstrom (1980) and Baron (1982) construct a screening model which

focuses on the lead manager’s benefit from underpricing. In a screening model, the

uninformed party offers a menu or schedule of contracts, from which the informed

party selects the one that is optimal given her unobserved type and/or hidden action.

The contract schedule is designed to optimize the uninformed party’s objective, which,

given his informational disadvantage, will not be first-best optimal.

To induce optimal use of the underwriter’s superior information about investor

demand, the issuer delegates the pricing decision to the bank. Given his information,

the underwriter self-selects a contract from a menu of combinations of IPO prices and

underpricing spreads; if likely demand is low, he selects a high spread and a low price,

and vice versa if demand is high5. This optimizes the underwriter’s unobservable

selling effort by making it dependent on market demand. Compared with the first-best

solution under symmetric information, the second-best incentive-compatible contract

involves underpricing in equilibrium, essentially since his informational advantage

allows the underwriter to capture positive rents in the form of below-first-best effort

costs.

d) Assume institutional investosrs know more than the issuer about the

prospects for the company’s competitors or the economy as a whole. Individual

investors know their own demand while the underwriter doesn’t know

Benveniste and Spindt (1989) suggest that a key function of the investment bank that

takes a company public is to elicit information from better-informed investors. In

signaling models, issuers have the informational advantage. However it seems

plausible that there is an information asymmetry running in the other direction.

Institutional investors may know more than the issuers about the prospects for the

company’s competitors or the economy as a whole because they are exposed to the

flow of IPOs on a continuous basis. And even the least well informed investor knows

something the issuer doesn’t: her own demand for the shares. So while the issuers

know their own particular pieces of the jigsaw better than anyone else, investors hold

other pieces of the same jigsaw which when put together give a clearer picture of the

5 There is empirical support for the notion of a menu of compensation contracts. Dunbar (1995) shows that issuers successfully offer underwriters a menu that minimizes offering costs by inducing self- selection.

value of the company. The task of the underwriter is then to acquire as many jigsaw

pieces as possible before setting the offer price.

Underwriters underprice the issue in this case to elicit information from better-

informed investors. This forms the information revelation theories. But why would

investors cooperate and reveal their information, especially when the information is

positive? Benveniste and Spindt (1989) show that book building is a mechanism that

induces investors to reveal their information truthfully. In book building, the book

contains investors’ indications of interest (which can take the form of price-quantity

bids, unlimited bids, or ‘soft’ information such as ‘give me what you’ve got’). These

indications of interest can communicate the various jigsaw pieces the investors hold.

To make sure that they do, the underwriter offers a stick and a carrot. The stick is that

any investor who claims that here jigsaw piece looks unfavorable is allocated no or

only very few shares. This mitigates the incentive to mispresent positive information.

The carrot is that any investor who claims her jigsaw piece looks favorable (for

instance via an aggressive indication of interest, such as bidding for a large quantity at

a high price) is rewarded with a disproportionately high allocation of shares. Taken

together, the stick and the carrot can ensure that an investor is never better off

claiming bad news when the news is in fact good.

To make sure this mechanism work, the underwriter underprices the issue. Effectively,

an investor’s monetary reward for truthful reporting equals the number of shares

allocated times dollar underpricing.

One common empirical implication among three of the above asymmetric information

models is that underpricing should increase in the ex ante uncertainty surrounding the

intrinsic value of an IPO (Beatty and Ritter 1986). Winner’s curse model explains that

an investor who decides to engage in information production implicitly invests in a

call option on the IPO, which she will exercise if the ‘true’ price exceeds the strike

price, the price at which the shares are offered to the public. The value of this option

increases in the extent of valuation uncertainty, so more investors will become

informed. This raises the required underpricing, since an increase in the number of

informed traders aggravates the winner’s curse problem. Signaling model says that a

noisier environment increases the extent of underpricing that is necessary to achieve

separation. And principal-Agent model implies the same because the more uncertain

the value of the firm, the greater the asymmetry of information between issuer and

underwriter, and thus the more valuable the underwriter’s services become, resulting

in greater underpricing.

2.1.2 Institutional Explanations

There are two main institutional underpricing models.

a) Legal insurance hypothesis of underpricing

The basic idea is that companies knowingly sell their stock at a discount to insure

against further lawsuits from shareholders disappointed with the performance of their

shares. In countries with stringent disclosure rules, underwriters and issuers are

exposed to considerable litigation risk. Lawsuits are costly to the defendants, not only

directly- legal fees, diversion of management time, etc.- but also in terms of the

potential damage to their reputation capital. Tinic (1988) and Hughes and Thakor

(1992) argue that intentional underpricing may act as insurance against such securities

litigation.

b) Price Support

Temporary price support in IPOs is legal in many countries including the USA, the

UK, France, Germany, Greece and the Netherlands. In these countries underwriters

can support prices by stimulating demand or by restricting supply in the after-market.

They can stimulate demand either by posting bids at or below the offer price

(stabilizing bids), or by actively buying back shares of weak offerings (stabilizing

trades). Although inventory positions expose underwriters to considerate risk, Ellis et

al. (2000) show that underwriters use their short positions to manage inventory risk.

Underwriters typically oversell IPOs, by allocating 115 percent of shares on offer. If

prices subsequently rise, they can cover their short position by exercising the over-

allotment option to buy another 15% of share from the issuer; if prices fall, they leave

the over-allotment option unexercised and close out their short position by open-

market purchases instead. Either way, they make a profit. Ellis et al find that price

support is not costly to underwriters because, as market makers, they earn trading

commissions that are large enough to offset any losses they might suffer on their

inventory. Trading profit, on the other hand, increase in initial underpricing, which

might give underwriters an incentive to increase underpricing.

Ruud (1991, 1993) argues that underwriters do not underprice deliberately and

underpricing is a byproduct of price support. His starting point is from the statistical

regularities of initial returns. Rather than forming a symmetric distribution around

some positive mean, initial returns typically peak sharply at zero, are highly positively

skewed, and include few negative observations. Ruud explains that underwriters price

IPOs at expected market value and support those offerings whose prices fall below the

offer price in after-market trading. Such behavior would tend to eliminate the left tail

of the distribution of initial returns, and thus lead to the observation of a positive

average price jump. According to Ruud, if price support suppresses the negative tail of

the initial return distribution, companies merely appear to be underpriced on average.

Asquith et al. (1998) investigate whether observed underpricing is a byproduct of

price support. If Ruud is correct in saying that there is no deliberate underpricing, then

the initial return distribution of unsupported offerings should have a mean of zero.

This, however, is not what Asquith et al find.

Ruud’s statistical view leaves many economic questions unanswered, such as why

underwriters would want to provide price support. A number of studies such as

Schultz and Zaman (1994), Prabhala and Puri (1999), Hanley et al (1993) and

Benveniste et al (1996) offer different explanations of why underwriters provide price

support, which will not be described in detail here.

2.1.3 Ownership and Control Going public is, in many cases, a step towards the separation of ownership and control.

Ownership matters for the effects it can have on the management’s incentives to make

optimal operating and investment decisions. In particular, where the separation of

ownership and control is incomplete, an agency problem (Jensen and Meckling 1976)

between non-managing and managing shareholders arises: managers may maximize

the expected private utility of their control benefits at the expense of outsiders, rather

than maximizing expected shareholder value.

Two principal models have sought to rationalize the underpricing phenomenon within

the bounds of an agency cost approach. Their predictions are diametrically opposed:

while Brennan and Franks (1997) view underpricing as a means to entrench

managerial control and the attendant agency cost by avoiding monitoring by a large

outsider shareholder, Stoughton and Zechner’s (1998) analysis instead suggests that

underpricing may be used to minimize agency costs by encouraging monitoring.

a) Underpricing as a Means to Retain Control

Brennan and Franks (1997) develop a model in which underpricing gives managers a

means to protect their private control benefit by allocating shares strategically when

taking their company public. Managers would wish to avoid a single investor

assembling a large stake for fear that their non-value-maximizing behavior would

receive unwelcome scrutiny. By deliberately underpricing the flotation, they can

ensure that the offer is over-subscribed and that investors will need to be rationed in

their allocations. Rationing allows managers to discriminate between applicants of

different sizes and so to reduce the block size of new shareholdings.

b) Underpricing as a Means to Reduce Agency Costs

There is one hidden assumption in Brennan-Franks’ model that managers try to

maximize their expected private utility by entrenching their control benefit. Actually

managers may wish to allocate the issue in a way that minimizes, rather than

maximizes their scope for discretion. Managers are part –owners of a company, they

bear some of the costs of their own non-profit maximizing behavior. Stoughton and

Zechner (1998) observe that, in contrast to Brennan and Franks, it may be value

enhancing to allocate shares to large outsider investor who is able to monitor

managerial actions. Monitoring is a public good. Since any large shareholder will

monitor only in so far as this is privately optimal, there will be underinvestment in

monitoring from the point of view of both shareholders and incumbent managers. To

encourage better monitoring, managers may try to allocate a particularly large stake to

an investor.

Different models explain different situations in different countries. Some models are

not possible explanations for IPO underpricing in a particular country because of the

country’s stock market characteristics. The Chinese stock market characteristics

determine that some IPO underpricing models do not apply in China, but the

characteristics do provide a unique situation where certain pmodels can be relatively

clearly examined. Section 2.2 describes features of the Chinese primary market;

section 2.3 summarizes previous studies on the Chinese IPO underpricing. In section

2.4, I will analyze comprehensively which classical IPO underpricing models are

possible explanation for Chinese IPO underpricing according to the Chinese market

features and previous findings.

2.2 Features of the Chinese Primary Market

2.2.1 The pre-offer process The IPO decision in China is made on the basis of political considerations as well as

profitability considerations. Every year, the Chinese authorities (the State Planning

Committee, the Central Bank, and the China Securities Regulatory Commission)

determine the total number of issues allowed and which firms can make issues.

The process begins when the State Planning Commission lays out its annual financing

plan for state enterprises. After the calculations and political balancing, a total figure

representing capital to be raised by all listings is entered into the State Plan regardless

of market conditions. The non-state sector has seldom been included in the thinking

about which enterprises should be enabled to finance through the sale of securities.

Thus equity financing for Chinese companies via either the domestic or the

international markets, has been mostly for State Owned Enterprises (SOEs). So the

listing process in China is also a process of restructuring and packaging SOEs into

shareholding companies and a process of privatization.

Once the China central authorities have set the overall quota, each province is

allocated a sub-quota. The stated criteria used for allocation of new issues among

provinces reflect the central security regulatory authorities’ perceived regional

development needs and provincial differences in production structure and industrial

base. Within each regional quota, the local security regulatory authorities invite

enterprises to apply a listing and make a selection based on some criteria6. These

criteria include the performance and sectoral development objectives of the enterprise.

Local government selection criteria take into account the profitability and

performance criteria of the exchanges.

Once approval for an issue has been obtained an investment syndicate will be formed

to draw up a detailed plan. Securities companies will perform the standard services of

providing advice, underwriting and distributing shares to the public, as well as

developing a secondary market in them. The underwriting market in China is

relatively competitive and there are 129 firms providing underwriting services. Table

2.1 shows that the top ten underwriters occupy 64.8 percent of the IPO market. All of

them are state owned companies and most of them are associated with the government

and are very well connected to the regulation authority (Tian L. G. 2003). A-share

issues are underwritten by domestic brokerage firms owned by the state, foreign share

6 A prerequisite for firms to get floated is to satisfy the conditions stipulated in the Company Law. The conditions stipulated in the Company Law enacted in December 1999 are:

i. Total share capital not less than RMB 50 million ii. With an operating history of at least three years and a continuous profitability over the past 3

years. iii. No less than 1000 shareholders holding shares paper-valued RMB1000 or above; More than

25 percent of total shares offered to the public. iv. No lawbreaking activities or dishonest accounting report in the past three years.

issues are underwritten by prestigious financial institutions with international

reputations.

Table 2.1 Main Underwriters and their performance (1991-2001)

Underwriters Under- written Firms

Underwritten shares (1,000)

Market Share

(%)

Main sponsors

1 Guotai Jun'an Securities Co. 238 1560.2 15.8 Central Bank 2 Shenyin Wan'guo Securities

Co. 182 930.1 9.42 Shanghai Govt.

3 Nangang Securities Co. 123 824.3 8.35 Central Bank 4 Huaxia Securities Co. 81 600.8 6.08 Central Bank 5 CITIC Securities Co. 62 499.5 5.06 State Council 6 Haitong Securities Co. 81 487.7 4.94 Shanghai Govt. 7 Guangfa Securities Co. 79 433.5 4.39 Guangdong Govt. 8 Everbright Securities Co. 54 394.4 3.99 State Council 9 Gousen Securities co. 63 338.9 3.43 Shenzhen Govt. 10 United Securities Co. 25 329.9 3.34 37 National SOEs

Source: Tian L. G. (2003)

2.2.2 Type of Shares in the Chinese stock market

In the privatization process, the Chinese government introduced 5 major categories of

shares to allow ownership of state-owned enterprises to be dispersed among the

government itself, state-owned enterprises, firm’s own employees, domestic public

and foreign investors.

(1) State shares, which are owned by the state and its various ministries, bureaus and

regional governments. They are not tradable.

(2) Legal entity shares, which can only be held by State-Owned Enterprises and/or the

foreign partner of a corporatized joint venture. These shares are highly illiquid. They

cannot be listed in the two official exchanges (SHSE and SZSE), but they can be sold

to other legal entities through the nation-wide computerized system called STAQS

(the Security Trading and Automatic Quote System), which was first introduced in

July 1992.

One distinguishing ‘Chinese characteristic’ is that the majority shareholding of

equities are non-negotiable government shares and legal entity shares. In my sample

of 343 IPOs from Nov 1995 through Dec 1998, 65% of A-shares outstanding are held

by the state and the legal entities. Individual investors own only 35% of shares. This

means that, on average, only about 35% of total shares outstanding are traded publicly

on either the SHSE or SZSE.

(3) Employee shares, which are those shares issued by the listed company and offered to

managers and employees prior to those offered to the public. These shares are initially

prohibited from trading for a certain period of time (typically 6 months or 3 years in

China). After that they become tradable A-shares.

(4) Ordinary domestic shares or A-shares designated only for private Chinese citizens and

traded on SHSE and SZSE. In terms of size and level of activity, the A-share market

dominates China’s equity markets. A-shares can only be bought and sold by individual

or legal persons within the PRC and are RMB dominated. Overseas investors are not

permitted to purchase A-shares unless they purchase authorized joint venture mutual

funds.

(5) Foreign shares, designated only for foreign investors to be traded on security

exchanges in China (B shares), in Hong Kong (H shares) or on the NYSE (N shares) 7.

2.2.3 The Issuing Mechanism

In established markets three methods are normally used in making initial public

offerings, fixed price, book building and auction. In China, the offering mechanism

adopted is mainly fixed price offerings, but it is quite complicated and different from

those of developed markets.

7 The B-share market was opened to the domestic residents on 19 February 2001.

The share offering mechanism in China has gone through several stages of reforms

(see Appendix A). The most commonly used method after 1995, however, is the

online fixed price offering methods called ‘Shang Wang Ding Jia’. This online fixed

price offering method 8 was first introduced in 1994, in which investors bid for

quantities, with pro-rata allocation in the event of oversubscription. Investors need to

pay a full subscription deposit but with repayment for unsuccessful applicants around

one week after subscription. This offering method has proved a more efficient

procedure and meets with the approval of investors. It has become the major offering

method from 1996 till the year 2002.

The offer price in the fixed price offering is chosen according to the formula of taking

the after tax profits per share multiplied by a price earning ratio (PE), the latter being

set in relation to the price earnings ratios of listed companies in the same locality and

industry. However, The PE ratio changes in accordance with the guidance of CSRC

(China Securities Regulatory Commission). Otherwise the IPOs will not get approval.

The CSRC often imposes a ceiling on the PE ratio, which prevents prices from being

set in relation to an individual firm’s characteristics and growth potential. Moreover

the ceiling changes over time. Before 1999, the ceiling level was controlled at 15. In

January 1999, the ceiling restriction was loosened and the PE ratio used in IPO pricing

raises to as high as 50 until the year 2002 when a ceiling of 20 was re-imposed. Not 8 In the online fixed price offering, the lead underwriter uses the exchange trading system to sell new stocks at a fix price and investors apply new stocks through the existing buy order at a designated time. Investors must have one stock account and one cash account and enough full payment funds deposited in their cash account prior to application. The number of shares applied must be in whole lots. A lot is 1,000 shares. The application movement is like this: 1st day Investors apply 2nd-3rd days Exchange validates investors’ deposit funds and allocates one serial number to each lot applied and investors affirm their application serial numbers. 4th day Lead underwriter organizes balloting 5th day Lead underwriter publicizes the balloted winning serial numbers on designated papers. Investors make payment for their successful applications and the rest part are refunded.

only the PE ratio affected by the authority’s policy, but also the after-tax profits per

share used in the IPO pricing. For example, from end 1995 to February 1997, the

regulated after-tax profit took the average level of one year before IPO and the IPO

year; from March 1997 to February 1998, the after-tax profit adopted the three year

average prior to IPO; while from February 1998 to March 1999, the after-tax profit per

share equals to the estimated after-tax profit per share for the IPO year.

In the fixed price offering, pro-rata allocation is used to allocate the overwhelming

applications. But the pro-rata allocation in China takes the form of random allocation

rules, where balloting chooses the investors. The ballot ratio used equals to the

number of shares publicly offered divided by the number of shares investors

subscribed. Therefore there is an inverse relation between the demand for new issues

and the allocation rate.

2.2.4 Supervision and Regulations

One requirement for a well-functioning market concerns supervision and this may be

provided through government legislation and/or internal regulation. Reliable and fair

trading procedures can increase investors’ confidence and help safeguard their

interests. During the period under study, the State Council Securities Commission (the

"SCSC") and the China Securities Regulatory Commission (the "CSRC") were

responsible for supervising and regulating the securities market. The SCSC,

established in October 1992, is the State authority responsible for exercising

centralized market regulation. The CSRC, also established in October 1992, is the

SCSC's executive branch responsible for conducting supervision and regulation of the

securities markets in accordance with the law and regularities. In November 1998, the

Central People's Government held the National Finance Conference and decided to

reform and reorganize the national securities regulatory mechanism. The local

securities regulatory departments will be supervised directly. Organizations engaged

in securities formerly supervised by the People's Bank of China were put under the

centralized supervision of the CSRC. But it was until July 1999 the first securities law

was enacted, which provided a consistent legal framework for the securities industry

and stock market in China. In general, before 1999, in terms of supervision and

regularities, the Chinese stock market is immature, compared with a fully-fledged

stock market in a developed market.

2.2.5 Other Characteristics Related to This Study

In China, almost all IPOs are oversubscribed (in my sample of fixed price offerings,

there is no under subscription) due to an extremely high demand relative to its limited

supply of new issues. Before the emergence of stock markets, Chinese households had

access to a very limited number of investment instruments, mainly savings deposits at

relatively low interest rates. Miurin and Sommariva (1993) describe how the lack of

consumer goods and financial instruments forced Chinese individuals to invest in

fixed-rate bank deposits that provide a negative real return during inflationary period

that started in the early 1980s. At the same time, China’s household savings rate was

one of the world’s highest, about 40 percent of total disposable income. Chen (1995)

reports the results of surveys of Chinese citizens indicating that about 80% of the

respondents desired to participate in the market but was unable to do so. Of those

investors in the market, about 83 percent indicated the intention to increase their stock

investment. On the other hand, the aggregate value of new shares to be issued is

limited by the national investment and credit plan. Therefore there has been a

persistent demand for new shares in China.

It is also noteworthy that seasoned equity offerings (SEOs) are very frequently

observed among Chinese issuers and that SEOs account for a substantial portion of

shares issued. About 91percent of the Chinese firms that went public before 1 July

1994 issued seasoned equities before 1 January 1996 (Su and Fleisher, 1999). Kim et

al (2000) also reports that in their sample IPOs from 1992 through 1995,

approximately 64 percent go back to the market to raise additional equity capital in the

three years after IPO.

Another characteristic of the Chinese stock market related to my study is that the

accounting report and market regulatory system in China are relatively primitive and

incomplete (Aharony et al., 2000; Xiang, 1998). The auditing standards in Chinese

stock market are also generally perceived to be low (Aharony et al 2000). There is far

less corporate disclosure in China than that of developed countries. Private investors’

major source of information is the IPO prospectuses, which unfortunately are not

reliable under the existing accounting and auditing standard. This makes individual

investors difficult to evaluate an IPO before investing. Furthermore A-shares are sold

to relatively unsophisticated private individual investors who usually do not commit

much time and recourses on IPO firm evaluation (while B shares are sold primarily to

international institutional investors such as foreign mutual funds). Therefore investors

lack information about the true quality of the firm going public and there are big ex

ante uncertainties about the issuing firm’s value.

2.3 Prior Studies of the Chinese IPO underpricing

There are some papers documenting the extraordinarily high underpricing of Chinese

IPOs. Table 2.2 presents a survey of these studies. Using different data sets, these

papers report the mean initial returns range from 127 percent to 949 percent and

present a number of determinants of underpricing, including time gap, offering size,

issuer’s fractional ownership etc. Most of the studies examine only a few determinant

factors instead of testing the classical IPO underpricing models comprehensively

except that Mok and Hui examined the ex ante risk and Su and Fleisher examined the

signaling model. Mok and Hui find that the high equity retention by the state, a long

time-lag between offering and listing and ex ante risk of new issues were key-

determinants of market adjusted IPO underpricing. Su and Fleisher examined the

signaling model comprehensively and find that the Chinese IPO underpricing is a

strategy for firms to signal their value to investors. They also investigate the effect of

offering mechanism on IPO underpricing and find that IPO underpricing is the largest

under the lottery with a fixed number of lottery-forms and is the smallest under the

auction mechanism. Wu (2001) reports that subscription rate in China is positively

related to IPO underpricing in support of winner’s curse model. She also finds that

there is no significant relationship between the underwriter’s reputation and the degree

of IPO underpricing. A more recent study by Tian (2003) argues that the listing quota

and pricing caps imposed by the government are major determinants of Chinese IPO

underpricing.

Table 2.2 Prior studies of the Chinese IPO underpricing

Papers Sample size

Sample period

Average Initial Return (%)

Findings pertaining to the explanations of the IPO underpricing

Mok and Hui (1998) 87 1990-1993 289 The high equity retention by the state, a long time-lag between offering and listing and ex ante risk of new issues were the key-determinants of the underpricing.

Kim et al (1998) 45 1993 594 IPOs for which a larger percentage of total shares are sold to individual investors are more underpriced and IPOs of firms that are expected to have larger increases in profitability are less underpriced , which is consistent with the political persuasion hypothesis.

Su and Fleisher (1999) 101 1987-1995 949 The signaling hypothesis explains the pattern of underpricing behavior rather well. In examining the effect of the offering mechanism on IPO underpricing they find the underpricing to be the largest under the lottery with a fixed number of lottery-forms and is the smallest under the auction mechanism.

Chau et al (1999) 102 1990-1993 546 Investors in previously centrally planned economies view agency costs as a consideration in investment; Initial returns are smaller when the government retains a large proportion of ownership and initial returns are negatively related to firm size. Investors rely on insider ownership to reduce agency costs.

Chen et al (2000) 277 1992-1995 350 The state underprices to ensure future seasoned equity issues to be successful; The long- time lag from the offering date to the first trading date explain the high underpricing; A- share IPOs that subsequently make rights issues are significantly more underpriced.

Wu (2001) 840 1990-2000 218 Subscription rate is positively related to IPO underpricing; there is no significant relationship between underwriter's reputation and underpricing;

Chi and Padgett (2002) 340 1996-1997 127 The quota system for new issues is the main reason for the underpricing Tian (2003) 1124 1991-2000 267 The listing quota and pricing caps imposed by the government account for more than

half of the severe underpricing. Information on the quality of the firm causes IPO underpricing, but it is not a major determinant. Besides the effects of the financial regulations, Chinese-specific investment risks also contribute to severe underpricing. The long time lag from the IPO date to the first trading date causes the underpricing.

Note: This table describes only studies on the Chinese IPO underpricing. Research on other aspects of the Chinese stock market such as the long term IPO after- market performance, the development of China’s privatization program, the price behavior of listed companies, or the relationships between company value and accounting earnings and book values is not included. Among the papers listed, only findings of relevant points are summarized.

2.4 Possible Explanations for Chinese A-share IPO Underpricing

The Chinese stock market characteristics determine that some IPO underpricing

models do not apply in China, but the characteristics do provide a unique situation

where certain models can be relatively clearly examined as well.

As shown above that the major offering mechanisms in China does not have any

pricing or rationing bias. This suggests that the first two models of ownership and

control will not apply since these two models need rationing discrimination as means

to realize the control ends.

In the Chinese IPO market, there is no book building offering mechanism until the

year 2001, therefore information revelation can’t possibly explain the high level of

underpricing, at least before 2001.The lawsuit idea is a US-centric model, which fails

in the international context: underpricing is a global phenomenon, while strict liability

laws are not. The risk of being sued is not economically significant in Australia (Lee

et al. 1996), Finland (Keloharju 1993b), Germany (Ljungqvist 1995a), Japan (Beller

et al.1992, Macey and Kanda 1990) Sweden (Rydqvist 1994), Switzerland (Kunz and

Aggarwal 1994), or the UK (Jenkinson 1990), all of which experience underpricing.

As an emerging stock market, China did not have a complete securities law in force

until July 1999; the risk of being sued is not economically significant. Therefore the

lawsuit hypothesis does not apply here. Price support is prohibited in Chinese stock

market; neither can price support underwritten underpricing explain the Chinese IPO

underpricing.

As to the principal-agent hypothesis, on one hand underwriters do not have much of

market power to seek the information rent because of the competition in the Chinese

underwriting line; on the other it is not a problem for underwriters to place all

available stock with investors due to the extremely high demand. In western countries,

securities companies underwrite stock issues at a price decided through negotiation

with the issuing company. There is a risk involved in the underwriting: when

securities companies fail to sell all the stocks they purchase, they have to lower the

selling price since according to the regulations in some countries they cannot hold

these unsold stocks. It is thus of vital importance for the underwriter to get the right

price for the stock they underwrite. In China things are different. Although like their

western counterparts, the Chinese issuing companies also preferred to have a high

premium price for their shares, the securities companies in China are, however,

willing to do so in order to attract more business. This is because the demand for

shares is always high. There is no risk of shares being unsold, and even if such a risk

does exist, the unsold shares could always be stored for future sale. So getting

underwriting contracts can almost guarantee profit for securities companies. Therefore,

without rent seeking or moral hazard problems, the principal-agent model cannot

possibly explain the Chinese IPO underpricing.

Winner’s curse problem is a possible explanation for Chinese IPO underpricing. There

are mainly two types of investors in the Chinese stock market: individual investors

and institutional investors. Table 2.2 shows the structure of investors as of 1998. The

vast majority of investors in the Chinese market are individuals who can be regarded

as uninformed investors. The small portion of institutional investors might function as

informed investors.

(In thousands) Table 2.3 The Structure of Domestic Investors in 1998

Shanghai Shenzhen Total

Institutional Investors 62.6 93.2 155.8

Individual Investors 19927.1 19024.1 38951.2

Total 19989.7 19117.3 39107

Source: China Security Year Book 1999

Rock’s winner’s curse model is examined only in countries where there are data on

allocation rate to subscribers (Koh and Walter, 1989; Levis, 1990; Keloharju, 1993;

Amihud et al, 2003). Fortunately we are able to have the allocation data in China. As

described before, In China, all applications of different sizes have an equal probability

of being accepted and the probability (ballot ratio) is publicly announced after IPOs.

This feature enables us to examine the adverse selection model in the Chinese market.

Due to weakness in disclosure and auditing standards, investors lack information

about the true quality of the firm going public. A relatively high degree of investor

uncertainty affects the IPO pricing. As mentioned before, the Winner’s Curse model,

the signaling model and Principal-Agent model all suggest a positive correlation

between ex ante uncertainty and underpricing. Mok and Hui (1998) argue that proxies

for ex ante uncertainty explains the pattern of A-share IPO returns for a sample of 87

Shanghai firms that went public during the years 1990-1993. Thus ex ante uncertainty

could also be one of the main reasons for Chinese IPO underpricing.

High degree of investor uncertainty also means that the information asymmetry

between the investors and the issuers is high. This provides incentives for good quality

issuers to underprice to signal their firm value. Moreover the frequent observation of

SEOs among Chinese issuers also proves that signaling might be a good explanation

for underpricing. Su and Fleisher (1999) find that the signaling hypothesis explains

the pattern of underpricing behavior among Chinese issuers rather well. Their findings

in support of the signaling hypothesis are: 1) the correlation between the degree of

IPO underpricing and initial offer price for the proportion of the firm offered to the

public is negative, given the issuer’s retained ownership. 2) the degree of IPO

underpricing is positively related to proxies for the issuer’s intrinsic value, the

variance o future returns, and the issuer’s fractional ownership. 3) issuers with larger

IPO underpricing are more likely to raise larger amounts of capital through SEOs and

to do so more quickly than issuers with a smaller degree of IPO underpricing,

although the latter relationship is weak. Mok and Hui (1998) also find a positive

relationship between the issuer’s ownership and IPO underpricing in support of the

signaling hypothesis.

In summary of chapter 2, the IPO underpricing literature has provided rich

explanations for the financial anomaly of IPO underpricing. These underpricing

models are mainly divided into three categories: asymmetric information, institutional

explanations and ownership and control explanations. Considering the situations in the

Chinese market, I eliminate all models under the last two categories. Among the

asymmetric information explanations, I narrow down the possible Chinese IPO

underpricing explanations to the winner’s curse model, ex ante uncertainty

explanation and the signaling model. In the rest of the study I am going to focus on

examining the winner’s curse model, ex ante uncertainty explanation and the signaling

model in the Chinese A-share market among which the winner’s curse model9 has not

been examined before. The ex ante uncertainty hypothesis was tested by Mok and Hui

(1998), but they tested only one proxy for ex ante uncertainty, i.e., the inverse of new

funds raised. We consider 3 proxies−the standard deviation of after-market returns, the

9 Wu (2001) finds a positive correlation between underpricing and allocation rate in China in support of the winner’s curse model. But other key implications of the winner curse model were not tested. Therefore it is not a complete test of the winner’s curse model.

offer size and the age of firms, to examine the ex ante uncertainty hypothesis. In

examining the signaling model, we test eight empirical implications of the signaling

model, some of which have been examined in Su and Fleisher (1999), but the

methodology adopted and the conclusion made are different.

Chapter 3 Hypotheses and Methodology

This chapter describes the hypotheses and methodologies adopted in this study to

examine the winner’s curse model, ex ante uncertainty explanation and the signaling

model.

3.1 The Winner’s Curse Model

Rock (1986) proposes that high positive returns observed in IPOs cannot be earned in

practice because of adverse selection. Uninformed investors are allocated greater

quantities in overpriced IPOs and smaller quantities in underpriced IPOs. This is

because investors who are informed about the issuing company’s value select to invest

in underpriced IPOs only. Underpricing is then needed to attract uninformed investors.

In equilibrium, “weighting the returns by the probabilities of obtaining an allocation

should leave the uninformed investor earning the riskless rate” (Rock 1986).

Therefore we expect:

H1: After ration-adjusted, uninformed investors tend to earn the riskless rate.

Rock’s model assumes that uninformed investors invest in IPOs indiscriminately.

Thus to test this proposition, we assume that uninformed investors subscribe a fixed

amount of shares for each and every IPO. And their allocation-weighted initial return

is given by10

01 * I

II BALLOT

P PP

AWIR −

− −

= (1)

where

AWIR is allocation-weighted initial return measured from the application-close

date to the initial-listing date

P1 is closing price on the first day of trading

P0 is IPO offer price

BALLOT is ballot ratio used in lottery, i.e. the allocation rate

(I1-I0) /I0 is A-share composite index return from IPO date to first trading date in

corresponding stock exchanges, my proxy for market return. I1 is the

corresponding stock exchange closing price of A-share composite index

on the first trading date; I0 is the corresponding stock exchange closing

price of A-share composite index on the IPO date.

The first part on the right hand side of equation (1) is the allocation-weighted initial

return without adjusting of the market factor. The second part is the A-share composite

index11 return from IPO date to first trading date. Interest cost associated with the

10 In the fixed price offering, unsuccessful parts of application deposit are refunded around one week after the IPO subscription date. However, since the interest rate is extremely low in the Chinese market (average one-week interest rate in the study period is close to zero, 0.039%) and there are few other investment opportunities, it does not make much a difference for investors to get the deposit money back. Therefore, here I treat the application deposit frozen until the first trading date.

11 Following other studies, I use the Shanghai A-share composite index and the Shenzhen A-share composite index as corresponding market benchmarks. They are capitalization-weighted indices using all A-shares listed on the respective stock exchanges.

application is ignored because of the extremely low interest rate. Hypothesis H1 states

that AWIR is approximately equal to the riskless rate of interest.

Rock’s (1986) winner’s curse model also implies a negative correlation between initial

returns and allocations to investors. Since informed investors avoid overpriced IPOs,

uninformed investors receive larger allocations of shares on which they earn negative

returns and smaller allocations in underpriced IPOs. Thus the joint participation by

both informed and uninformed investors in underpriced IPOs makes the demand for

underpriced IPOs high, thus allocation rate low. My second hypothesis to test Rock’s

model is to examine the relation between underpricing and allocation rate.

H2: IPO initial returns are inversely correlated to allocations with investors.

This relationship can be examined by the following simple OLS regression

εββ ++= BALLOTTIR 10 (2)

where IR is the initial returns and BALLOTT is the logistic transformation of the ballot

ratio12:

)1/()log( αα +−+= BALLOTBALLOTBALLOTT (3)

The logistic transformation is used here to accommodate the cases where BALLOT is

practically 0 or 1. We expect β1 in equation (2) to be negative and significant.

3.2 Ex ante uncertainty

Another key empirical implication of the winner’s curse model, pointed out by Ritter

(1984) and formalized in Beatty and Ritter (1986), is that underpricing should increase

12 I The same transformation is used as Amihud et al.(2003). The term a=0.5/T, where T is the sample size.

in the ex ante uncertainty surrounding an issue. The underpinning is that higher

uncertainty leads to proportionally more informed-investors, which deteriorates the

winner’s curse problem (see chapter 2). Other testable implications of winner’s curse

model are basically elicited from this relationship between ex ante uncertainty and

underpricing. For example, Carter and Manaster (1990), Johnson and Miller (1988),

James and Wier (1990) and many other researchers tested the relationship between the

underwriter’s reputation and initial returns as evidence of adverse selection since it is

argued that more prestigious underwriters can reduce the informational asymmetry

and thereby cut the underpricing cost. Another explanation is that hot issue periods are

characterized by a higher level of ex ante uncertainty, necessitating higher

underpricing (Ritter, 1984).

However these relationships are not unique to adverse selection model. As discussed

in section 2, the signaling model and principal-agent model imply the same. Therefore

I will test the ex ante uncertainty explanation separately.

Beatty and Ritter (1986) assert that the degree of underpricing is directly related to the

ex ante uncertainty about the value of the IPO. Their proposition supported by Rock

(1986) states that the greater the ex ante uncertainty about the value of IPO, the

greater the expected underpricing. In other words, more underpricing cost is needed

for firms with greater ex ante uncertainty. In the Chinese market, there are large ex

ante uncertainties about IPO values (see chapter 2). Therefore ex ante uncertainty

might be one of the main reasons for the high IPO underpricing observed from the

Chinese market.

Researchers have been using variance of the aftermarket returns of IPOs (Ritter 1984,

1987; Clarkson and Merkley (1994)), age of the firm at the time of offering (Ritter

1984, 1991; Megginson and Weiss 1991), offer size (Beatty and Ritter 1986;

McGuinness 1992) and underwriter’s reputation (Carter and Manaster, 1990; Johnson

and Miller, 1988; James and Wier, 1990) as proxies for measuring ex ante uncertainty

of the IPOs. It is easy to understand that the higher the standard deviation of

aftermarket returns the higher the ex ante uncertainty of the IPO firms. Large and old

companies should have less ex ante risk than small and young ones because there is

more information about them and because they are likely to be more closely

monitored by the government and regulatory authorities. I am not going to use

underwriter’s reputation as a proxy in this study because the Chinese A-share issues

are underwritten by domestic state-owned security companies and there are no

prestigious financial institutions with international reputations involved. Furthermore

Wu (2001) finds that underwriter’s reputation has no significant effect on the degree

of underpricing in the Chinese new issue’s market. The rest of three proxies for ex

ante uncertainties predict that the larger the variance of the aftermarket returns of IPOs,

the younger the age of the issuing firms and the smaller the offering size, the higher

the uncertainty about the value of IPO firms and therefore the more underpriced of

corresponding IPOs. Thus we would expect:

H3: The standard deviation of aftermarket returns of IPOs is positively related to IPO

underpricing.

H4: The offer size of the firm is inversely related to IPO underpricing.

H5: The age of the firm is inversely related to IPO underpricing.

I use a multiple linear regression model to examine the explanatory power of ex ante

uncertainty and control for other well-known determinants of IPO underpricing. The

dependent variable is the market adjusted initial return. The proxies for ex ante

uncertainty are SD, AGE and IPOSZ, where SD is standard deviation of returns over

days from 1 to 100 after IPO, AGE is the age of a firm in years from the establishment

date to the date of IPO, and IPOSZ is the number of shares offered at IPO times the

IPO offer price.

Other variables that might affect the level of ex ante uncertainty are also controlled for.

The first one is the market return before IPO. Ritter (1984) asserts that hot issue

period is characterized by a higher level of ex ante uncertainty, which necessitates

higher underpricing. There has been overwhelming evidence that underpricing is

higher in buoyant stock markets: Davis and Yeomans (1976) (UK), Reilly (1977)

(USA), McGuiness (1992) (Hong Kong) and Rydqvist (1990) (Sweden) all show that

initial returns tend to be higher following periods of high returns on the market index.

To test if Chinese IPOs are more heavily underpriced when the market is performing

well, I use BFMARTN, percentage change in the A-share composite index 3 months

prior to the issue, as one of the explanatory variables. Another control variable is the

issuers’ fractional ownership. In an emerging market with high information

asymmetry, the domestic investors interpret a high percentage of equity retention by

the state as government confidence and a business guaranty. That is high equity

retention by the state lowers ex ante uncertainty (Mok and Hui 1998). The time gap

elapsed between the IPO date and first trading date can also affect level of ex ante

uncertainty. Chowdry and Sherman (1996) demonstrate that an increasing lag between

the fixing of the offer price and the beginning of trading results in bigger ex ante

uncertainty and more IPO underpricing. Mok and Hui (1998) and Su and Fleisher

(1999) reported a very high time gap between offering and listing time in the Chinese

market. Therefore I add time lag from IPO date to first trading date as one of the

independent variables. Other control variables include year dummies, industry

dummies and the stock exchange dummy.

)4(64329796 lnln

13121110987

6543210

εβββββββ βββββββ

++++++++ ++++++=

STKCDSHININININYY LAGOWNSHPBFMARTNIPOSZAGESDIR

If H3, H4 and H5 hold, we expect β1 to be positive, and β2 and β3 to be negative. If the

ex ante uncertainty hypothesis stands, we also expect positive β4 and β6 and negative

β5.

For convenience of reference, I list all variables used below:

RAWIR (P1-P0)/P0

IR market adjusted initial return equals to RAWIR minus A-share index

return from IPO date to first trading date, (P1-P0)/P0-(I1-I0)/I0

AGE age of a firm in years from the establishment date to the date of IPO

IPOSZ number of shares offered at IPO times IPO offer price SD Standard

deviation of aftermarket stock returns over days from 1 to 100 after

IPO OWNSHP the proportion of shares owned by the government,

legal entities and employees after IPOs

LAG days elapsed between IPO date and the first trading date

YRD year dummies: Y96 equals to 1 for IPOs made in 1996 (including one

IPO in November and one in December 1995), Y97 equals to 1 for

IPOs made in 1997, Y98 equals to 1 for IPOs made in 1998.

IND industry dummies: IN2 utilities, IN3 properties, IN4 conglomerates,

IN5 industry, IN6 commerce

EXD exchange dummies, STKCDSH is a dummy for IPOs listed on

Shanghai Securities Exchange and STKCDSZ are IPOs listed on

Shenzhen Stock Exchange.

BFMARTN percentage change in the A share composite index 3 months prior to the

issue.

The following are variables that are going to be used in latter part of this study. I list

them all here for convenience of reference.

AFTRTN Abnormal return over the period from trading day 1 to trading day 400

after the IPO date. The abnormal return equals to [(P400-P1)/P1]-[(I400-

I1)/I1)] where P400 is the 400th day closing price of the stock and I400 is

the 400th day closing price of the corresponding exchange A-share

composite index.

SEOSZ number of shares offered at first SEO times SEO price

SEOSZ/MKT SEO size as a fraction of market value of equity one day before the

SEO announcement

TIMESEO number of days elapsed between the IPO and SEO first trading date

REACT Excess return around the date when the firm announces its SEO. The

excess return is estimated over the event days -1 through +4, where day

0 is the SEO announcement date, and equals to [(P4-P-1)/P-1]-[(I4-I-1)/I-

1], where P4 is the 4th day closing price of the stock and I4 is the 4th day

closing price of the corresponding exchange A-share composite index

after the SEO announcement (the SEO announcement date is taken as

the publishing date of the SEO prospectus).

V Firm’s intrinsic value equals to 400 trading days aftermarket abnormal

return (same as AFTRTN).

V1 firm’s market capitalization on the first trading day, equals to the first

day closing price times the number of shares offered at IPO

SEOPRC/TRDPRC SEO price over the closing price one day before the SEO

announcement date

NK Firm’s total net asset after the issue

Y2ROA firm’s return on assets two years after IPO

3.3 The signaling model

Group 1 - Correlations among IPO underpricing and issuer’s intrinsic value, fractional ownership and project variance

Leland and Pyle (1977) developed one of the first IPO valuation-signaling models.

They modeled IPO firm current value as a positive function of the proportionate share

ownership of the entrepreneurs who bring the company to listing. The intuition behind

this model is that entrepreneurs who retain a large fraction of shares only do so if they

are very confident about the firm’s prospects. Investors recognize this commitment by

the entrepreneur and place a higher valuation on the IPO.

H6: Positive relationship between the issuer’s fractional ownership and firm value

Downes and Heinkel (1982) design an empirical test of Leland and Pyle’s theoretical

signaling model in the form of the following regression:

εαβββ +++= ˆ1 210 IPOSZV (5)

Where V1 is the firm’s market capitalization on the first trading date, equals to the

first day closing price times the number of shares outstanding, α is the issuer’s

proportional ownership and α̂ = α +ln(1- α), the Leland and Pyle’s signal of a firm’s

future cash flow13, as a function of α. Based on this construction, 2β is expected to

have a negative sign. ε is a disturbance term. Downes and Heinkel’s regression model

assumed constant risk across all firms. To account for different risks across firms, I

incorporate a risk term in the above equation, using the ex post variability in the stock

market returns as a proxy for ex ante risk:

εβαβββ ++++= SDIPOSZ 3210 ˆ1V (6)

where α̂ = α +ln(1- α) and α equals to OWNSHP. So if Leland and Pyle’s ownership

signaling model stands, we expect a negative 2β . Higher risk should be compensated

with more returns. Grinblatt and Hwang (1989)’s signaling model also implies that,

keeping the issuer’s fractional holding constant, the value of the firm is positively

related to the variance of its cash flows. Therefore we expect 3β to be positive and

significant.

After Leland and Pyle’s study using issuer’s ownership as a signal of firm quality,

Allen and Faulhaber (1989) and Welch (1989) extend the signaling model by

13 Leland and Pyle show that in the context of the CAPM, the value of the firm will be given by:

KrbZV +−++−= )]1log())[1/(()( ααα where r =the riskless interest rate; b =the risk aversion parameter in the entrepreneur’s mean-variance utility function; K =the amount of capital raised; and

222 )]~,~[cov(

mx mxZ σ

σσ − =

~ where x~ and m are the returns on the firm and the market, respectively.

introducing the notion that IPO underpricing is another signal of firm quality. In the

same year, Grinblatt and Hwang construct a model of bivariate signaling which is

closest to Leland and Pyle’s model. Leland and Pyle analyze only one signal; only one

parameter can be unknown, implying that the variance of the cash flows of the firm’s

projects must be observable. In Grinblatt and Hwang’s model, the variance as well as

the mean of the project’s cash flows is unknown, so that a second signal, the level of

underpricing, is needed to convey the firm’s value to the market.14 In the model’s

separating equilibrium, a firm’s intrinsic value is positively related to the degree its

new issue is underpriced. There are eight empirical implications of Grinblatt and

Hwang’s model, the following three are unique (Among the remained five

implications, four are consistent with Leland and Pyle’s model and one is consistent

with Rock’s model):

H7: The degree of underpricing is positively related to the issuer’s fractional

ownership given the variance of the firm’s cash flow

H8: Firm value is positively related to the degree of underpricing given the issuer’s

fractional ownership

H9: Firm value and the degree of underpricing are positively related given the

variance of the firm’s cash flows

H7 can be tested with the following simple OLS regression:

εβββ +++= SDOWNSHPIR 210 (7)

H8 and H9 can be examined as follows

εββββ ++++= SDOWNSHPVIR 3210 (8)

14 Grinblatt and Hwang’s model explained IPO underpricing, while Leland and Pyle’s did not.

where V is a proxy for firm’s intrinsic value-the 400 trading days aftermarket

abnormal return (this is more than one and half years of aftermarket return; 400 days

are chosen because of the data availability). We expect 1β to be positive if H8 and H9

hold. However there might be an error in variable problem in equation (8). In

Grinblatt and Hwang (1989)’s model, if the expected value and the variance of future

cash flow for a risk-averse firms is unknown to investors, then there exists an

equilibrium signaling schedule relating two signals, fractional holding and IPO

underpricing corresponding to the probability of the firm’s intrinsic value to be

revealed. So IR and OWNSHP are determined endogenously. Moreover, since both IR

and V include the first day trading price, they might be spuriously negatively

correlated. To check these possible spurious correlation and simultaneous bias

problems, I use Housman test to examine the exogeneity of the two variables, V and

OWNSHP. Exogenous variables AGE, Y2ROA, NK, year dummies, industry

dummies and the stock exchange dummies are used as instrumental variables.

Different proxies have been used for IPO firm’s intrinsic value in previous studies

such as initial market valuation15 of the firm’s quality (Firth and Liau-Tan 1997;

McGuiness 1992) and two year aftermarket excess returns on equity (Michaely and

Shaw 1994). I choose Michaely and Shaw’s method because, firstly, initial market

valuation does not necessarily reflect firm’s true value, it sometimes can be a result of

over reaction or fads. Su and Fleisher (1999) use the market capitalization in 1996 as a

proxy of firm’s intrinsic value for his entire sample IPOs from 1987 to 1995. I do not

regard this as a suitable measure since different IPOs have different aftermarket

15 Initial market valuation of the firm is estimated using the price at the end of the first day of trading in Firth and Liau-Tan (1997) and using the price at the close of the tenth day of trading in McGuiness (1992).

information revelation time and there is likely to be a bias for firms that issue IPOs

later.

Group 2 – Linkages between IPOs and Seasoned Equity Offerings

The central result of the theoretical models of Allen and Faulhaber (1989), Grinblatt

and Hwang (1989) and Welch (1989) is that high quality firms underprice IPOs with

the expectation that the loss can be recouped through subsequent equity offerings after

investors have had the opportunity to recognize the firm’s true potential. Low quality

firms cannot mimic high-quality firms as they are denied the opportunity to sell

seasoned issues at attractive prices and capture the potential benefits of IPO

underpricing. In other words, good firms underprice to “leave a good taste in

investors’ mouths” and they will be rewarded at the time of the seasoned issue by a

higher price for the shares. Therefore, the signaling model leads to the empirical

predictions:

H10: Firms with more underpriced IPOs are more likely to issue seasoned equity than

firms with less underpriced IPOs

A direct implication of the signaling model is that, in their eagerness to capitalize on

the favorable news, high-quality firms will return to the capital market as soon as the

opportunity comes, and to maximize the benefit. Namely high-quality firms are more

likely to reissue. Another reason is that the costs of raising fund are higher for firms

underprice more, so they are more likely to come to the seasoned equity market to

recoup their lost. This also implies that:

H11: Firms with more underpriced IPOs are likely to issue seasoned equity more

promptly than firms with less underpriced IPOs

Another intuition behind is that it is more costly for high quality firms to defer their

investments in new projects than for firms of low quality (Jegadeesh et al, 1993).

The signaling model implies empirically a positive association between underpricing

and the success of the SEO. The success of SEO can be measured in terms of the SEO

size relative to its IPO size and the market reaction to the seasoned issue. This leads

to our H12 and H13.

H12: Firms with more underpriced IPOs are likely to issue larger amount of seasoned

equity than firms with less underpriced IPOs

H13: Firms with more underpriced IPOs are likely to experience a less unfavorable

price reaction to SEO announcement than firms with less underpriced IPOs

H13 follows the notion that firms with higher IPO underpricing are more likely to

return with seasoned equity issue and hence investors are more prepared for or less

surprised by their SEOs. In other words, SEOs from firms with more underpriced

IPOs are better received by investors because of their superior quality.

There is, however, an alternative explanation for the existence of the above relations

between IPO underpricing and SEO activity. In fact market feedback hypotheses

posits that the market is better informed than the issuer and hence a high return on the

IPO date or a very good aftermarket performance of stocks implies that the issuer has

underestimated the marginal return to the project. The issuer uses this information and

increases the scale of the project by raising additional capital through seasoned

offerings. To explore whether the relations between IPO underpricing and SEO

activity can be explained by market feedback hypotheses, I examine whether the

returns in 400 trading day after the IPOs are related to subsequent offering. I choose a

400 trading day post-IPO window to measure the aftermarket returns because the

cross-sectional standard deviation of the aftermarket returns in the 400 day window is

about the same as the cross-sectional standard deviation of the IPO date returns, which

suggests that the same amount of information is revealed to the market during these

two periods. This follows the logic in Jegadeesh (1993). Whereas Su and Fleisher

(1999) use only the 10-day after-market returns to test the market feedback hypothesis.

There is comparatively too little of information revealed in such a short time than that

revealed on the initial trading date. Therefore it is not appropriate to compare the

effect of the two variables on SEO activities even after adjusting for standard

deviation.

Following Jegadeesh et al. (1993), I test the H10 using a logit model.

[ ] )9(643

29796ln)1/(ln

10987

6543210

εββββ βββββββ

++++ +++++++=−

STKCDSHINININ INYYIPOSZAFTRTNIRpP seoseo

Where Pseo is the probability that a firm issues seasoned equity after the initial offering.

The first two independent variables are market adjusted initial return (underpricing)

and the aftermarket abnormal eturn over the period from trading day 1 to trading day

400 after the IPO date. The after-market abnormal return equals to market-adjusted

return over the same period. Since firms with a small IPO size are more likely to come

to seasoned equity offering, I include the natural logarithm of the IPO size as an

additional explanatory variable. Finally, I also control for potential differences in SEO

activity across years, industries and exchanges. We expect a positive β1 if H10 is true

and a positive β2 if market feedback hypothesis is true.

To examine the relationship between the time elapsed between IPO and SEO,

TIMESEO, and IPO underpricing, I use a tobit model with right censoring. For firms

with no SEOs over the years from 1996 through 2001, I assume that the time it takes

for their re-issuance is infinity. For firms that issue their first SEOs during that period,

the maximum time elapsed between IPO and SEO in our sample is 1394 days.

Therefore, I take ln(1400) as the right censoring value. The explanatory variables are

the same as those in the previous logit model. Su and Fleisher also use a Tobit model

to test the same hypothesis. But for IPOs with no seasoned equity offerings, they take

TIMESEO value as zero and use a left censoring test, which is inaccurate.



  

∞ <+++

++++++++ =

)10( 1400ln64

329796ln ln

1098

76543210

otherwise RHSifSTKCDSHININ

ININYYIPOSZAFTRTNIR TIMESEO εβββ

ββββββββ

We expect a negative β1 if H11 is true and a negative β2 if market feedback hypothesis

is true.

To test H12 we use a Tobit model similar to Jegadeesh et al. (1993). The Tobit model

specifies the relation between the relative size of seasoned offering and the

explanatory variables as follows:



  

>+++ ++++++++

= )11(0

064 329796ln

/ 1098

76543210

otherwise RHSifSTKCDSHININ

ININYYIPOSZAFTRTNIR IPOSZSEOSZ εβββ

ββββββββ

where SEOSZ/IPOSZ is the relative size of Seasoned Equity Offerings. The

independent variables are the same as those in the logit model. Similarly, we expect a

positive β1 if H12 is true and a positive β2 if the market feedback hypothesis holds.

To examine the excess return around the date when the firm announces its SEO, I

estimate the excess return over the event days -1 through +4, whereday 0 is the SEO

announcement date. The excess return, REACT, equals to [(P4-P-1)/P-1]-[(I4-I-1)/I-1],

where P4 is the 4th day closing price of the stock and I4 is the 4th day closing price of

the corresponding exchange A-share composite index after the SEO announcement

(the SEO announcement date is taken as the publishing date of the SEO prospectus).

P−1 and I−1 are the stock price and index price 1 day before the SEO announcement.

Moreover I include a variable TIMESEO, which measures the number of days between

the IPO date and the SEO date. The longer the time between these events, the greater

the volume of public information released about the firm, thus reducing the

uncertainty about the firm value. Additional independent variables are SEOSZ/MKT,

which is the SEO size over the stock market value 1 day before the SEO

announcement and SEOPRC/TRDPRC, which is the SEO price over the closing price

1 day before the SEO announcement. These variables are included to control for

possible differences in the extent to which the market is surprised by the SEO

announcements that are not related to the initial returns of their IPOs or their after-

market returns. For firms with SEOs, we do the following regression to examine H13:

)12(// lnln64

329796ln

1413

12111098

76543210

εββ βββββ

ββββββββ

+++ +++++

+++++++=

TRDPRCSEOPRCMKTSEOSZ SEOSZTIMESEOSTKCDSHININ

ININYYIPOSZAFTRTNIRREACT

Similarly, we expect a positive β1 if H13 is true and a positive β2 if the market

feedback hypothesis is true.

Chapter 4 Data and Empirical Results 4.1 Data and underpricing

To pursue the objectives of this study, I examine all online fixed price (Shang Wang

Ding Jia) and firm commitment A-Share IPOs over the period November 1995 -

December 1998. Financial institutions and close-end funds are excluded because

previous work, reviewed in Smith (1986) indicates that regulation of these entities

affects securities issuance phenomena. Online fixed price offering is the most

commonly used offering method in Chinese A-share IPOs. The study of the online

fixed price IPOs can represent the general IPO market in China. I exclude the IPOs

after 1999 to obtain sufficient after-market data for testing the signaling model. To do

this, I need at least 3 years’ time for the listed firms to issue their first seasoned equity

offerings. The sample period ends in 1998, which also helps control for government

intervention in the pricing of IPOs since after 1998 there was a policy change in the

ceiling for the PE ratio (Before January 1999, a ceiling of 15 was imposed; after that,

the restriction was loosened and the PE ratio used in IPO pricing rises dramatically.

Tian (2003) finds that after the deregulation, initial returns decrease by 133.5 percent

in China.) The data comes from several sources including the trading database from

GTA (Guo Tai An Information Technology Co.), IPO database from Haitong

Securities and the panorama network website, www.p5w.net. Finally a sample of 343

IPOs is collected, representing a broad spectrum of industries such as utilities,

properties, conglomerates, industry and commerce. Descriptive statistics are reported

in panel A of table 4.1. The mean and median offering sizes of the IPOs (i.e., gross

proceeds), are RMB 304 million and RMB 220 million respectively. The average

proportion of shares retained by the state, legal entities and employees in sample IPOs

is 71.04%, indicating that the majority of shareholding of equities are non-negotiable

government shares and legal entities shares. The average age of IPO firms is 3 years

old. The mean of PE ratio and offering price are 14.85 and 6.19 respectively.

Table 4.1Descriptive statistics on 343 IPOs in the 1996-1998 period and 215 SEOs in the period 1996-2001

Variable No. Obs.

Mean Median Maximum Minimum Std. dev.

Skewne ss

Panel A IPO characteristics IPOSZ (million RMB) 343 304.00 220.50 2625.00 33.00 310.00 3.60

OWNSHP 343 0.7104 0.7353 0.8649 0.3670 0.0721 -1.2751 AGE ( years) 343 3.06 2.03 40.99 0.10 3.27 6.08

PE 343 14.85 14.57 32.52 8.80 2.51 0.76 LAG ( days) 343 32.43 21.00 377.00 9.00 35.09 6.02

NK (million RMB) 343 751.44 466.27 7363.60 103.97 950.01 3.97 P0 343 6.19 5.99 15.70 2.45 1.83 1.38 P1 343 13.71 12.74 53.57 4.41 6.01 1.90

BFMARRTN 343 0.1583 0.1076 0.8649 -0.2859 0.2407 0.6655 AFTRTN 343 0.1482 -0.0210 3.9045 -3.0145 0.7575 1.3241

Panel B SEO characteristics SEOSZ (million RMB) 215 248.27 188.10 1395.00 2.08 206.07 2.49 SEOPRC 215 8.79 8.00 26.00 3.30 3.50 1.60 TIMESEO (days) 215 805 805 1394 441 219.44 0.55 REACT 215 -0.0109 -0.0144 0.2672 -0.1308 0.0472 1.2076

215 out of the 343 IPOs issue their first SEOs in the period 1996 to 2001. In other

words, over 60% of the sample IPOs issues SEOs. All the 215 SEOs included in our

study are rights offers (SEOs to the existing shareholders). Public seasoned offering

(SEOs to the general public investors) are rarely seen in China because public

seasoned offerings are not permitted until 1997. Even after the restriction was lifted,

very few firms use public offerings in their re-issuances. Some characteristics of the

SEO data are reported in panel B of table 4.1. The average SEO price is RMB 8.79

and the average SEO size is RMB 248.27 million. The mean and minimum time it

takes from IPO to SEO is 805 days and 441 days, respectively.

Table 4.2 presents the distribution of new and seasoned equity offerings through time.

Fixed priced new issues from 1996 to 1998 distributed quite evenly and each year

accounts for around 30% of all sample IPOs. Most SEOs occur during the time from

1998 to 2000. 85% of the SEOs in my sample are from the year 1998 to 2000.

Table 4.2 Distribution of 343 fixed pricing IPOs and 215 first seasoned equity offerings (SEOs) by offering year, 1996-2001

Initial Public Offering Seasoned Offerings

Year Number Percentage (%) Number Percentage (%) 1996 121 35.28 0 0.00 1997 125 36.44 19 8.84 1998 97 28.28 66 30.70 1999 52 24.19 2000 65 30.23 2001 13 6.05 Total 343 100.00 215 100.00

To measure the level of IPO underpricing, I use the market adjusted initial return i.e.

the raw initial return after taking into account of the overall market effect.

The raw initial return (RAWIR) is calculated as:

001 /)( PPPRAWIR −=

P0 is the offer price and P1 is the first day closing price.

Market adjusted initial return equals to RAWIR minus A-share composite index return

from the IPO date to its first trading date.

001 /)( IIIRAWIRIR −−=

where I1 is the closing price of the SHSE A-share composite index or SZSE A-share

composite index on the first trading day of the new issue, and I0 is the closing price of

the SHSE A-share composite index or SZSE A-share composite index on the IPO date.

To examine a longer-term IPO after-market performance, I calculate initial returns

over 10 and 100 trading days after the IPO as

00100010 /)(/)(10 IIIPPPIR −−−=

and 0010000100 /)(/)(100 IIIPPPIR −−−=

Table 4.3 Initial returns in IPOs, with adjustment for allocation

Variabl e

Mean Median Maximum Minimum Std. Dev. Skewness

Initial return (underpricing)

1 IR 1.2359 1.1123 8.2050 -0.1211 0.8479 2.4129 (27.00) 2 IR10 1.1927 1.0619 5.6548 -0.1886 0.8372 1.5264 (26.38) 3 IR100 1.2379 1.0192 5.4331 -0.8410 0.9145 1.1342 (25.07) Allocation-weighted initial return 4 AWIR -0.0033 -0.0050 0.3652 -0.3621 0.0955 -0.1360 (-0.63) The number in parenthesis is the t-statistics test that the mean is different from zero.

Some summary statistics for the initial excess returns are presented in table 4.3 and the

distribution of IR is depicted in figure 1. The average IR is positive and significant: the

mean is 123.59% with t = 27.00. Only 7 out of 343 IPOs have negative initial returns.

Nearly 98% of the IPOs have positive initial returns. The average 10-day and 100-day

initial excess returns, IR10 and IR100, are 119.27% and 123.79%, respectively. This is

much lower than the Chinese IPO underpricing level in early 1990s reported in

previous studies, showing that the Chinese new issue market has indeed improved its

efficiency. IR10 is slightly lower than IR, and IR100 is slightly higher than IR.

Notably, the mean initial return from day +1 to day +100 is not significantly different

from zero (mean = −1.72%, t = −0.93). If the price of a new issue at the opening of

trade represents an overreaction or speculative bubble rather than the true economic or

fundamental price, we should witness a significant decline in the stock return in the

after-market. However, we do not see a statistical difference between IR and IR100,

which suggests that there is no momentum effect in pricing. The correlation between

IR and the subsequent initial returns from day +1 to day +100 is 0.114, indicating that

the price of the IPO stocks adjusts efficiently after the IPO. All 3 initial return

distributions are positively skewed, reflecting the very high returns obtained in a few

cases (see figure 4.1).

Figure 4.1 The distribution of the initial excess return in IPOs

2.0

5.0 5.0

11.4

12.8 13.7 13.4

7.3

5.5 5.2

1.2

3.5

1.2 0.6 0.9 0.6 0.6

1.5 2.3

6.4

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

0 & be

low

0.0 1:

0.2 0

0.2 1:

0.4 0

0.4 1:

0.6 0

0.6 1:

0.8 0

0.8 1:

1.0 0

1.0 1:

1.2 0

1.2 1:

1.4 0

1.4 1:

1.6 0

1.6 1:

1.8 0

1.8 1:

2.0 0

2.0 1:

2.2 0

2.2 1:

2.4 0

2.4 1:

2.6 0

2.6 1:

2.8 0

2.8 1:

3.0 0

3.0 1:

3.2 0

3.2 1:

3.4 0

3.4 1:

3.6 0

3.6 1&

ab ov

e IR

F re

qu en

cy (

% )

Table 4.4 presents the summary statistics of the initial returns and the PE ratios used

in IPO pricing by year and by stock exchange. There are only 2 observations in 1995

and I include these into the data for 1996. The average initial returns in 1996, 1997

and 1998 are 95.87%, 144.96% and 130.57%, respectively. The significant difference

in underpricing across years is mainly caused by the changes in the IPO pricing policy

over time. As discussed before, the offer price in the Chinese fixed price offering is

determined by the multiplication of PE ratio of the same industry and the issuing

firm’s after-tax profit per share. The supervisory authorities often impose a ceiling to

the PE ratio used and the ceiling level changes over time. Table 4.4 shows that the PE

ratios used in 1996 are significantly higher than that in 1997 and 1998. I will not

analyze in detail why the initial return in 1996 is lower than that of 1997 and 1998 or

why the IPOs in 1997 are more underpriced than IPOs in 1998 since the policy

changes are complicated. There is not much difference in the initial returns and PE

ratios across the two stock exchanges.

Table 4.4 Statistics of initial returns and PE ratios by years and by stock exchanges

Years Mean Median Maximum Minimum Skewness 1996 (121) IR 0.96 0.95 3.37 -0.12 0.90 PE 15.34 14.90 32.52 9.70 0.26 1997 (125) IR 1.45 1.29 4.64 0.01 1.46 PE 14.67 14.90 18.00 10.00 -0.16 1998 (97) IR 1.31 1.13 8.20 -0.05 3.25 PE 14.48 14.50 18.00 8.80 0.20 Shanghai (170) IR 1.20 1.11 4.31 -0.12 1.05 PE 14.86 14.57 32.52 8.80 3.13 Shenzhen (173) IR 1.27 1.13 8.20 -0.11 2.86 PE 14.84 14.57 27.96 11.35 -0.59 The figure in the parenthesis is the number of sample IPOs in that year or at that stock exchange.

4.2 Allocation and Adverse selection The pro rata allocation rate in China is the ballot ratio used in the lottery and equals to

the ratio of the number of shares publicly offered in the IPO to the number of shares

subscribed by investors. There is no under-subscription in the sample. BALLOT

denotes the ballot ratio (allocation rate). Some summary statistics of BALLOT are

presented in table 4.5, and the pattern of its distribution is depicted in figure 4.2.

Table 4.5 Statistics of allocations in sample IPOs

Mean Median Maximum Minimum Obs. BALLOT 0.0218 0.0065 0.9540 0.0013 343 Ballot classified by initial return For IR<0; Ballot 0.3432 0.0851 0.954 0.0057 7 For IR>0; Ballot 0.0151 0.0065 0.7315 0.0013 336

The ballot ratio in most IPOs is extremely small due to the overwhelming

oversubscription. The overwhelming oversubscription is mainly caused by surplus

demand for the limited supply of negotiable shares. Moreover, the vast majority of the

primary market investors are relatively unsophisticated private individual investors.

The characteristics of Chinese individual investors and the weakness in information

disclosure result in a very big proportion of uninformed investors in the Chinese

market. The distribution shows that the allocation rate in most IPOs (95%) is below

5% and there are only a few cases with ballot ratio greater than that. The mean for

BALLOT in our sample is 2.18% and the median is much lower, 0.65%. The average

allocation rate for overpriced IPOs is 34%, which is much higher than that of the

underpriced IPOs (1.51%). This is consistent with the winner’s curse theory that the

uninformed investors have a higher probability of obtaining overpriced IPOs.

However, this is weak evidence of the presence of winner’s curse problem since we

have only 7 overpriced IPOs.

Figure 4.2 The distribution of allocations to investors in IPOs

3.5

25.7

18.4

11.1

7.9

5.2 5.5

3.2 3.2 2.3 1.7

0.9 2.3

1.2 0.3 0.0 0.6 0.3 0.0 0.6

6.1

0.0

5.0

10.0

15.0

20.0

25.0

30.0

0.0 00

0:0 .00

0.0 04

1:0 .00

0.0 08

1:0 .01

0.0 12

1:0 .01

0.0 16

1:0 .01

0.0 20

1:0 .02

0.0 24

1:0 .02

0.0 28

1:0 .03

0.0 32

1:0 .03

0.0 36

0:0 .03

0.0 40

1& ab

ov e Ballot Ratio

F re

qu en

cy (

% )

If H1 is true, the allocation-weighted initial returns minus the riskless rate should be

approximately zero. The statistics for AWIR are presented in table 4.2. The mean of

AWIR is negative (−0.33%), and is not statistically different from zero (t = −0.63).

This suggests that, despite the seemingly high initial returns, uninformed IPO

investors essentially break even.

The OLS regression result for equation (2) is as follows:16

1535.0)94.7()74.2( 41.067.0

2 =−− −−=

R BALLOTTIR

The estimated coefficient for BALLOTT is −0.41, with t = −7.94. The strong inverse

relationship between initial returns and allocations to investors is again consistent with

Rock’s hypothesis of adverse selection.

The above two empirical results confirm the major empirical implications of Rock’s

theory. I conclude that individual investors in China face the winner’s curse problem.

However, without data on application sizes and other details, it is not clear who the

informed investors in the Chinese IPO market are.

4.3 Ex ante uncertainty

Model 1 of table 4.6 presents the regression results for equation (4). Consistent with

H3, the coefficient for the standard deviation of the after-market returns is positive and

strongly significant. The coefficients for lnAGE and lnIPOSZ are both negative and

significant, which supports the ex ante uncertainty hypotheses H4 and H5, namely, the

age and offer size of the issuing firm are inversely related to IPO underpricing in the

Chinese IPO market.

The coefficient of BFMARTN is positive and significant at the 5% level, which means

that the IPOs are more underpriced in hot market. This is consistent with previous

studies (Davis and Yeomans, 1976 (UK), Reilly, 1977 (USA), and McGuiness, 1992

(Hong Kong)). The coefficient of OWNSHP is negative and significant, consistent 16To control for other relevant independent variables, another regression has been done (see appendix C). As shown in the table, the conclusion is of no big difference from equation (2). Therefore only regression result of equation (2) is reported here.

with Mok and Hui (1998). This shows that Chinese investors interpret high state and

legal entity retention as government support and business guaranty. That is, high

equity retention lowers the ex ante uncertainty about firm value, thereby lowers the

required level of underpricing. The time lag between the IPO date and the first trading

date is insignificant in explaining IPO underpricing in the regression. Different from

Mok and Hui (1998) and Su and Fleisher (1999)’s sample, the time lag after 1996 has

been dramatically shortened, which removes previous uncertain factors caused by the

extreme long time lag.17

Table 4.6 OLS regression Analysis Investigating Ex Ante Uncertainty and other Significant Explanatory Variables of IPO Underpricing

Dependent Variable: IR Model 1 Model 2 Explanatory Variables: Coeff. t stat. Coeff. t stat. Constant 0.8331 2.64

*** 0.7767 2.51

SD 9.9021 78.61 ***

9.9022 79.57 ***

LNAGE -0.0312 -3.23 ***

-0.0304 -3.20 ***

LNIPOSZ -0.034 -2.23 **

-0.0315 -2.10 **

BFMARRTN 0.0985 2.45 **

0.1003 2.50 **

OWNSHP -0.2875 -2.44 **

-0.2775 -2.36 **

LAG 0.0003 0.9 0.0002 0.74 Y96 -0.1937 -6.56

*** -0.1968 -6.69

***

Y97 -0.0091 -0.39 -0.0109 -0.48 IN2 -0.0031 -0.09 IN3 0.0029 0.03 IN4 0.0109 0.47 IN6 -0.0546 -1.71 STKCDSH 0.0517 2.93

*** 0.0531 3.04

***

Adjusted R² 0.9671 0.9671 ** Significant t statistics at the 5 percent level *** Significant t statistics at the 1 percent level 17 The average lag time in our sample is only 32 days, which is much shorter than the average of 260 days reported in Su and Fleisher’s (1999) study. The much shorter lag time from the IPO date to the first listing date in our sample shows that the online fixed pricing offering method is more efficient than previously used offering methods.

The positive and significant coefficient for the dummy variable Y96 shows that the

IPOs made in 1996 are less underpriced than the IPOs in 1998. This might be affected

by the changes of the PE ratio used in IPO pricing. There is no statistical difference in

underpricing across industries. The IPO underpricing in SHSE is significantly higher

than that in SZSE. As we have seen in table 4.4, there is not much difference in the PE

ratio used in IPO pricing across the two exchanges. Thus, this cannot be caused by the

difference in the PEs. One explanation is that many firms at SZSE are joint ventures,

while those listed at SHSE are mostly SOEs. There are relatively more disclosure and

less uncertainty in joint venture firms. That is why IPOs listed on SZSE are less

underpriced. The model explains 96.7% of the variability in initial returns of the

sample of A-share IPOs, which shows the strong explanatory power of ex ante

uncertainty. This supports my hypothesis that the high ex ante uncertainty in IPO

value is the main reason for the high level of IPO underpricing observed in the

Chinese market. Model 2 in table 4.6 presents a regression excluding 4 insignificant

industry dummy variables. The conclusion is the same as those from model 1.

4.4 The signaling model

Group 1- Correlations among IPO underpricing and issuer’s intrinsic value, fractional ownership and project variance Table 4.7 presents my regression result for equation (6). Control for risks and IPOSZ,

the slope coefficient estimate (t-statistics) of α̂ is 79.85 (1.06). Not only it is not

statistically significant but also the sign is opposite to Leland and Pyle’s prediction.

This indicates that the ownership signaling model does not stand in the Chinese A-

share market. As expected, the coefficient for SD is positive and significant.

Table 4.7 OLS regression to test Leland and Pyle’s theoretical signaling model

Dependent Variable: V1 Explanatory Variables: Coeff. t stat. Constant -103.81 -2.12 ** IPOSZ 1.75 41.91 *** AHAT 79.85 1.06 SD 1828.72 11.65 *** Adjusted R² 0.8389 ** Significant t statistics at the 5 percent level *** Significant t statistics at the 1 percent level

If H7 is true, we expect a positive sign on OWNSHP. But our OLS regression result in

table 4.8 reports, on the contrary, a significant (at 5% level) and negative coefficient

for OWNSHP, our proxy for issuer’s fractional ownership at IPOs. This indicates that

Grinblatt and Hwang’s bivariate signaling model does not stand in the Chinese A-

share market either.

Table 4.8 First OLS regression to test Grinblatt and Hwang’s Bivariate Signaling Model

Dependent Variable: IR Explanatory Variables: Coeff. t stat. C 0.11 1.15 OWNSHP -0.31 -2.34 **

SD 10.18 87.56 ***

Adjusted R² 0.9573 ** Significant t statistics at the 5 percent level *** Significant t statistics at the 1 percent level Model 1 in Table 4.9 gives the OLS regression results for equation (8). We can see

that V is very insignificant in explaining IR. This again proves that Grinblatt and

Hwang’s signaling model does not stand in the Chinese market during the sample

period. Table 4.10 presents the first regression of Hausman test for the exogeneity of

variable V. We take the residual of the regression in table 4.10 as VRESID and add it

as one more explanatory variable into equation (8). Model 2 in table 4.9 shows the

result. The insignificant coefficient for VRESID indicates that V is exogenous in

equation (8). Using the same Hausman test, I find that OWNSHP is exogenous too

(results not reported here). So the spurious correlation and simultaneous bias problem

has been eliminated.

Table 4.9 Second OLS regression to test Grinblatte and Hwang’s Bivariate Signaling Model

Model 1 Model 2 Dependent Variable: IR Explanatory Variables: Coeff. t stat. Coeff. t stat. C 0.11 1.13 0.11 1.11 V 0.01 0.37 -0.04 -0.78 OWNSHP -0.31 -2.33 ** -0.30 -2.26 ** SD 10.18 87.12 *** 10.20 86.21 *** VRESID 0.05 0.90 Adjusted R² 0.9572 0.9577 ** Significant t statistics at the 5 percent level *** Significant t statistics at the 1 percent level

Table 4.10 First regression of Housman test for the exogeneity of variable V

Dependent Variable: V Explanatory Variables: Coeff. t stat. Constant 0.17 1.45 LNAGE -0.01 -0.13 Y2ROA 0.00 0.73 NK -0.16 -3.34 *** STKCDSH 0.12 1.51 IN0002 0.18 1.09 IN0003 -0.62 -1.39 IN0004 0.06 0.49 IN0006 0.17 1.11 Y96 0.00 0.02 Y97 -0.10 -0.98 Adjusted R² 0.0350 ** Significant t statistics at the 5 percent level *** Significant t statistics at the 1 percent level

Group 2 - Linkages between IPOs and Seasoned Equity Offerings

Table 4.11 presents the logit regression test for the relation between IPO underpricing

and the probability of seasoned equity issue (equation (9)). The slope coefficient (t-

statistics) on the variable IR is −0.07 (−0.40). The slope coefficient for AFTRTN (t-

statistics) is 0.86 (3.97). For H10, the signaling hypothesis expects a positive and

significant role of IPO initial return in explaining the likelihood of issuing subsequent

equity offerings. However, I find a negative and insignificant coefficient for the initial

returns. This suggests that the signaling model does not stand. At the same time, the

estimates show a strong relation between the after-market price appreciation and the

likelihood of SEOs. In other words, the coefficient for AFTRTN suggests that the

higher the after-market returns, the more likely the listed firm re-issue. This is

consistent with the market feedback hypothesis. Other two significant variables are

Y96 and Y97, which means that IPOs in 1996 and 1997 are more likely to issue SEOs

than those in 1998. This is probably because IPOs in 1996 and 1997 have longer time

for SEOs than those in 1998. The rest of the dummy variables are insignificant and

their coefficients are jointly not different from zero. Therefore, I report a second logit

regression excluding those insignificant dummy variables in Model 2 of table 4.11.

Model 2 reflect almost the same result as that of model 1 except that the significance

level for dummy variable Y96 decreases.

Table 4.11 Logit Model to Test the relation between underpricing and the likelihood of SEO

Dependent Variable: SEOD Model 1 Model 2

Explanatory Variables: Coeff. t stat. Coeff. t stat. Constant 4.05 0.88 3.02 0.67 IR -0.07 -0.40 -0.07 -0.40 AFTRTN 0.86 3.97 *** 0.81 3.89 ***

LOG(IPOSZ) -0.23 -1.00 -0.17 -0.77 Y96 1.02 2.61 *** 0.98 2.53 **

Y97 1.46 4.78 *** 1.39 4.69 ***

IN2 -0.12 -0.25 IN3 1.52 1.20 IN4 0.16 0.45 IN6 -0.90 -1.95 STKCDSH 0.16 0.67 Note: 1. 215 observations with Dep=1, total observations is 343. 2. ** Significant t statistics at the 5 percent level *** Significant t statistics at the 1 percent level

The right censoring Tobit regression examining H11 is presented in table 4.12. The

slope coefficient estimate for IR is 0.04, with t = 0.83. The sign of IR is opposite to

our expectation and the t-statistics is insignificant. This again shows that the signaling

model does not stand in the Chinese market. The coefficient for AFTRTN is negative

(−0.17) and statistically different from zero at the 1% level (t = −4.42). This result

indicates that firms that experience large price appreciation after the IPOs are likely to

raise larger amounts of capital through seasoned equity issues. This is again consistent

with the market feedback hypothesis.

Table 4.12 Tobit Regression to Examine the relationship between Time SEO and IPO Unperpricing

Dependent Variable: LNTIMESEO Explanatory Variables: Coeff. t stat. Constant 7.44 6.98 ***

IR 0.04 0.83 AFTRTN -0.17 -4.42 ***

LOG(IPOSZ) -0.01 -0.21 Y96 -0.38 -4.18 ***

Y97 -0.35 -4.85 ***

IN2 0.09 0.83 IN3 -0.40 -1.34 IN4 -0.05 -0.62 IN6 0.28 2.49 **

STKCDSH 0.00 0.07 Adjusted R² 0.1248

Note: 1. Total observations are 343 and right-censored observations are 128. 2. ** Significant t statistics at the 5 percent level *** Significant t statistics at the 1 percent level

Model 1 of table 4.13 reports the Tobit regression estimates examining the relation

between the size of the seasoned offerings and the explanatory variables (equation

(11)). The estimate (t-statistic) of the slope coefficient for the variable IR is 0.02

(0.15), which indicates that the excess initial returns in the IPOs are weak in

explaining the relative SEO size and H12 is rejected. Same as the previous logit

regression, I find a positive (0.46) and significant (t = 4.84) coefficient for AFTRTN.

Consistent with my previous findings, the market feedback hypothesis is verified for

the Chinese A-share market. The tobit regression also shows that IPOSZ is negative

(−0.37) and significant (t = −2.85) in explaining the relative SEO size. The

coefficients (t-statistics) for year dummy variables Y96 and Y97 are 0.6 (2.67) and

0.72 (3.97), respectively, indicating that the IPOs in 1996 and 1997 raise higher

amount of capital through seasoned equity issues than the IPOs in the year 1998.

Another significant dummy variable is IN6 suggesting that commercial firms raise

smaller amount of capital in SEOs than industrial firms. There is no statistical

difference in the two stock exchanges. Therefore I report a second Tobit regression

excluding STKCDSH in model 2 of table 4.13, which shows almost the same results as

those of model 1.

Table 4.13 Tobit Regression to Examine the relationship between SEO Size and IPO Unperpricing

Dependent Variable: SEOSZ/IPOSZ Model 1 Model 2 Explanatory Variables: Coeff. t stat. Coeff. t stat. Constant 7.05 2.68 *** 7.08 2.69 ***

IR 0.02 0.15 0.02 0.15 AFTRTN 0.46 4.84 *** 0.46 4.82 ***

LOG(IPOSZ) -0.37 -2.85 *** -0.38 -2.87 ***

Y96 0.60 2.67 *** 0.60 2.67 ***

Y97 0.72 3.97 *** 0.72 4.01 ***

IN2 0.25 0.88 0.23 0.83 IN3 1.04 1.41 1.06 1.43 IN4 -0.05 -0.25 -0.06 -0.32 IN6 -0.54 -2.00 ** -0.54 -1.99 **

STKCDSH -0.07 -0.49 Adjusted R² 0.1440 0.1434 Note: 1.Total observations are 343 and left censored observations are 128. 2. ** Significant t statistics at the 5 percent level *** Significant t statistics at the 1 percent level

To examine the relation among the stock-price response to the announcement of

seasoned equity offerings, underpricing and after-market returns, I first use a sub-

sample of 215 IPOs with subsequent offerings to run the OLS regression. The results

are presented in model 1 of table 4.14. The estimated coefficient for IR is positive, as

expected, but statistically insignificant (t = 1.69)18. This indicates that underpricing the

IPO does not significantly mitigate the negative share-price response to a first

seasoned equity offering. The estimate of the coefficient for the variable AFTRTN is

18 Another regression excluding insignificant dummies from model 1 has been done; the result is of no difference from model 1. Therefore it is not reported here.

also not significantly different from zero. The rest of the explanatory variables are

insignificant. The adjusted R2 is 0.0089, which shows that the regression has very

weak explanatory power. This is not surprising in the Chinese market because the

seasoned equity offering news is normally leaked out long time before the publication

of the SEO announcement. Usually months before a re-issuance, a board meeting is

held to discuss the re-issuance decision and the meeting resolution is published the

next day after the meeting. Therefore, by the time of SEO prospectus publication, the

SEO news is not new and the stock price has already adjusted.

Model 1 of table 4.14 examines only 215 firms with their first SEOs within 3 years of

IPO. This is only a subset of our larger population. The decision to make subsequent

offerings is endogenous, which is not reflected in the cross-sectional estimates of

model 1. Therefore the estimator may be inconsistent as a result of truncation bias.

Eckbo et al. (1990) derive consistent estimators using a latent variable model. These

estimators account for the presence of the potential truncation bias. Michaley and

Shaw (1994) use this method to detect the dividend announcement effect. We also

adopt the same model to further examine H13.

Firstly, a probit regression is estimated as followings:

εγ += ZSEOD

where Z denotes the independent variables, which are IR, AFTRTN, Ln(IPOSZ), Y96

and Y97. They are related to the likelihood that a SEO will be issued. Then we

calculate the Mill’s ratio MILLSRATIO as φ(Zγ)/Ф(Zγ), where φ is the normal density

function and Ф is the normal cumulative distribution function.

By adding MILLSRATIO as one more explanatory variable into equation (12),

consistent parameters can be obtained:

)13(// lnln643

29796ln

151413

121110987

6543210

εβββ ββββββ

βββββββ

+++ ++++++ +++++++=

MILLSRATIOTRDPRCSEOPRCMKTSEOSZ SEOSZTIMESEOSTKCDSHINININ

INYYIPOSZAFTRTNIRREACT

The estimation result of equation (13) is presented in model 2 of table 4.14. Same as

our regression in model 1, the slope coefficient for IR and AFTRTN are still

insignificant. This verifies the fact that more underpriced IPOs do not experience a

less unfavorable price reaction to SEO announcement than firms with less underpriced

IPOs. Thus H13 is rejected.

In summary, the relations between IPOs and SEOs activities in the Chinese market are

mainly caused by the after-market performances of stocks instead of the issuer’s

signaling behavior. The signaling hypothesis does not stand in the Chinese A-share

market, while the market feedback hypothesis is supported.

Table 4.14 OLS Regression to Test the Price Reaction at the Announcement of SEO

Dependent Variable: REACT Model 1 Model 2 Explanatory Variables: Coeff. t stat. Coeff. t stat. Constant -0.35 -1.98 ** -0.09 -1.04 IR 0.01 1.69 0.00 1.59 AFTRTN 0.00 0.12 0.00 0.41 LNIPOSZ 0.00 0.51 0.00 0.94 Y96 0.02 1.61 0.01 0.82 Y97 0.00 -0.08 0.00 0.08 IN2 -0.01 -0.76 0.00 -0.40 IN3 0.04 1.14 0.03 1.12 IN4 -0.01 -1.06 -0.01 -0.99 IN6 -0.01 -1.03 -0.01 -0.86 STKCDSH 0.00 -0.19 0.00 -0.37 LNTIMESEO 0.02 0.81 0.00 -0.03 LNSEOSZ 0.02 1.54 0.00 0.44 SEOSZ/MKT -0.03 -1.25 -0.01 -0.85 SEOPRC/TRDPRC -0.04 -1.71 -0.04 -2.29 **

MILLS RATIO 0.00 0.22 Adjusted R² 0.0089 0.0177 Note: 1. Number of observations for model 1 is 215 and model 2 is 343. 2. ** Significant t statistics at the 5 percent level *** Significant t statistics at the 1 percent level

Chapter 5 Conclusions

This study examines the degree of underpricing for 343 online fixed price offerings

from November 1995 to December 1998. The initial return is on average 123.59%,

much lower than the level in early 1990s reported in previous studies. This indicates

that the efficiency in the primary market has improved. However, it is still larger than

what is found in most emerging markets.

I investigate possible explanations for the level of underpricing. I analyze possible

explanations for the Chinese market according to the characteristics of the Chinese

market and examine all major models, i.e., the winner’s curse model, the ex ante

uncertainty explanation and the signaling model.

Consistent with the winner’s curse model, after adjusting for rationing, uninformed

investors in the Chinese market essentially break even. The negative relation between

the initial returns and the allocation rates to investors also suggest that Chinese

individual investors face the winner’s curse problem. Using several proxies for ex ante

uncertainty, I find ex ante uncertainty has very high explanatory power in explaining

the Chinese IPO underpricing. This is consistent with Mok and Hui (1998)’s assertion.

After an extensive examination of 8 hypotheses of the signaling model, I conclude that

the signaling model does not stand in the Chinese market. Evidence shows that the

relations between IPO underpricing and SEO activities are caused by the market

feedback information. This is contrary to Su and Fleisher (1999)’s findings.

In all, the main reasons for the Chinese A-share IPO underpricing are investor’s high

level of ex ante uncertainty about IPO value and the winner’s curse problem. As I

have eliminated the possibility of the principal-agent and the signaling explanation, I

conclude that the positive relation between ex ante uncertainty and underpricing is

evidence in support of the winner’s curse problem. This suggests that reducing issuing

firms’ ex ante uncertainty, such as through more information disclosure from IPO

firms, will help to ameliorate the winner’s curse problem and thereby lower the level

of underpricing.

Given the prominence of the Chinese stock market in the emerging markets, the

results in this paper should be able to shed some light on explanations of IPO

underpricing in other emerging markets. The results add more evidences on testing of

winner’s curse model, the ex ante uncertainty explanation and the signaling model as

well. This should be illuminating and of value to both academicians and practitioners.

Appendix A: Offering Mechanism Changes in China In China, almost all new issue offerings before 2001 are fixed price offerings (The

offer price in the fixed price offering is chosen according to the formula of taking the

after tax profits per share multiplied by a price earning ratio). But the fixed price

offerings are not the same in their share allocations. There has been an overwhelming

demand 19 of new issue stocks in the Chinese market due to the few investment

opportunities and the high saving rate for the public. Therefore, how to distribute a

fixed amount of shares is a problem from the outset. As the Chinese stock market

develops, share allocation methods have gone through many stages of reforms. I will

discuss five of the most commonly used methods in the past ten years.

1. Limited lottery forms In 1991 and 1992, there was a lottery system with a pre-announced fixed number of

lottery forms. The maximum number of lottery forms each individual investor could

purchase was also fixed. Winners of the lottery could purchase a designated number of

shares per form. Thus, investors knew the odds of winning the lottery in advance. But

the limited lottery forms relative to an overwhelming demand of IPO and widespread

corruption caused a social chaos in Shenzhen in August 1992.

19 Over 95% of IPOs in China are oversubscribed

2. Unlimited lottery forms In 1993, when Tsingtao Brewery got listed, a method of selling unlimited number of

lottery forms was first used. Investors in this case could purchase as many lottery

forms as they desired. This is a fairer method than that of selling limited forms, yet

many shortcomings still exist such as high cost and low efficiency. A big amount of

money is spent on the printing of application forms while the issuer cannot raise more

funds from it.

3. Unlimited number of deposit certificates In August 1993, the authorities announce that issuing firms can use a new offering

method called unlimited number of deposit certificates. The deposit certificate here is

a fixed maturity, fixed amount and specially designed deposit certificate. The ballot

ratio in this case is decided by number of certificates sold, number of shares each

deposit certificates are entitled to and number of shares publicly offered. Although this

method saves social cost in printing the unlimited lottery forms, it causes a large

amount of social deposit savings move around the country. Large amount of cash

flows into banks where a new issue is offering. When the new issue is over, investors

must go to the banks to get back their defrozen funds. This is extremely inconvenient.

4. Full payment and pro rata allocation In this method, investors were required to deposit a certain quantity of funds into a

special saving account when submitting an application for shares, which could not be

withdrawn until the lottery was completed. These special saving accounts were given

relatively low interest. And losers have to wait a long time for funds to be repaid.

5. Online fixed price offerings (Shang Wang Ding Jia) In 1994, another share allocation mechanism was introduced in which investors bid

for quantities, with pro-rata allocation in the event of oversubscribed shares. This

method is the same as the full payment and pro rata allocation except that all

transactions happen here use the existing stock exchange trading system. Investors

again need to pay a deposit but with prompt repayment for unsuccessful applicants. It

has proved a more efficient procedure and meets with the approval of investors. This

is also the most commonly used offering method in the past 10 years.

Table A provides a summary statistics on the allocation methods adopted for A-share

IPOs from 1990 through 2000 (financial firms and closed end funds are excluded).

Table a: Statistics on the allocation methods adopted in the Chinese A-share market from 1990 through 2000 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 Total 1 & 2 0 7 49 53 13 2 0 0 0 0 0 124 3 0 0 0 3 6 1 4 2 0 0 0 16 4 0 0 0 1 1 1 49 59 5 0 0 116 5 0 0 0 0 0 6 120 125 97 91 73 512 Other methods 8 1 22 3 3 2 0 0 0 2 64 105 Not disclosed 2 4 9 63 13 0 0 0 0 0 1 92 Total 10 13 80 123 36 12 173 186 102 93 138 965 Note: 1 &2 Limited or unlimited lottery forms; 3 Unlimited number of deposit certificates; 4 Full payment and pro rata allocation; 5 Online fixed price offering (Shang Wang Ding Jia) Source: Haitong Securities

Appendix B Correlation Matrix

Table b: Correlation matrix of continuous explanatory variables in equation (4) IR SD LNAGE LNIPOSZ BFMARRTN OWNSHP LAG IR 1.00 SD 0.98 1.00 LNAGE 0.01 0.07 1.00 LNIPOSZ -0.28 -0.33 -0.28 1.00 BFMARRTN 0.11 0.13 0.14 -0.30 1.00 OWNSHP 0.00 0.03 -0.07 0.03 -0.07 1.00 LAG 0.20 0.18 -0.12 0.09 -0.23 0.02 1.00

Table c: Correlation matrix of continuous explanatory variables in equation (6) V1 IPOSZ AHAT SD V1 1.00 IPOSZ 0.88 1.00 AHAT -0.11 -0.14 1.00 SD 0.03 -0.25 -0.01 1.00

Table d: Correlation matrix of continuous explanatory variables in equation (10) IR AFTRTN LNIPOSZ

IR 1.00 AFTRTN -0.09 1.00 LNIPOSZ -0.28 -0.19 1.00

Table e: Correlation matrix of continuous explanatory variables in equation (12) (1) (2) (3) (4) (5) (6) (7)

(1) IR 1.00 (2) AFTRTN -0.14 1.00 (3) LNIPOSZ -0.22 -0.23 1.00 (4) LNTIMESEO 0.26 -0.08 -0.06 1.00 (5) LNSEOSZ -0.18 0.05 0.51 -0.01 1.00 (6) SEOSZ/MKT -0.19 -0.17 -0.03 -0.15 0.48 1.00 (7) SEOPRC/TRDPRC -0.04 -0.32 0.37 0.19 0.21 0.13 1.00

Appendix C Test of the Winner's Curse Model

Table f: OLS Regression to Test the Winner's Curse Model

Dependent Variable: IR Explanatory Variables: Coeff. t stat. Constant -0.79 -2.89 *** BALLOTT -0.06 -5.10 *** SD 10.11 82.83 *** LNAGE -0.03 -3.36 *** LNIPOSZ 0.04 2.85 *** BFMARRTN 0.00 -0.06 OWNSHP -0.41 -3.36 *** LAG 0.00 1.13 STKCDSH 0.04 2.00 ** Number of observations 343 Adjusted R² 0.9644

Bibliography

1. Aharony J., Lee C. J. and Wong T. J., 2000. Financial Packaging of IPO Firms in

China. Journal of Accounting Research 38, 103-125.

2. Allen /f. and Faulhaber G. R., 1989. Signaling by Underpricing in the IPO Market.

Journal of Financial Economics 23, 303-323.

3. Allen F. and Gale D., 1995. A Welfare Comparison of Intermediaries and

Financial Markets in Germany and the United States. European Economic Review

39, 179-209.

4. Amihud Y., Hauser S. and Kirsh A., 2003. Allocations, adverse selection, and

cascades in IPOs: Evidence from the Tel Aviv Stock Exchange. Journal of

Financial Economics 68, 137-158.

5. Amihud Y., Hauser S. and Kirsh A., 2003. Allocations, adverse selection, and

cascades in IPOs: Evidence from the Tel Aviv Stock Exchange. Journal of

Financial Economics 68, 137-158.

6. Beatty R. P. and Ritter J. R., 1986. Investment Banking Reputation, and the

Underpricing of Initial Public Offerings. Journal of Financial Economics 15, 213-

232.

7. Benveniste L. M. and Spindt P. A., 1989. How Investment Bankers determine the

Offer Price and Allocation of New Issues. Journal of Financial Economics 24,

343-361.

8. Brennan M. J. and Franks J., 1997. Underpricing, Ownership and Control in Initial

Public Offerings of Equity Securities in the UK. Journal of Financial Economics

45, 391-413.

9. Carter R. B. and Manaster S., 1990. Initial Public Offerings and Underwriter

Reputation, Journal of Finance 45, 1045-1067.

10. Chau C. T., Ciccotello C. S. and Grant C. T., 1999. Role of Ownership in Chinese

Privatization: Empirical EvidenceFrom Returns in IPOs of Chinese A-Shares,

1990-1993. Advances in Financial Economics 4, 51-78.

11. Chi J. and Padgett C., 2002. The Performance and Long-run Characteristic of the

Chinese IPO Market. ISMA discussion papers in Finance 2002-09, ISMA Center,

University of Reading, UK.

12. Chowdry B. and Sherman A., 1996. International Differences in Oversubscription

and Underpricing of IPOs. Journal of Corporate Finance. 2, 359-381.

13. Clarkson P. M. and Merkley J., 1994. Ex Ante Uncertainty and the Underpricing

of Initial Public Offerings: Further Canadian Evidence. Canadian Journal of

Administrative Sciences 11, 54-67.

14. Davis E. W., and Yeomans K. A., 1976. Market Discount on New Issues of

Equity: The Influence of Firm Size, Method o fIssue and Market Volatility.

Journal of Business Finance and Accounting 3 27-42.

15. Downes D.H. and Heinkel R., 1982. Signaling and the valuation of unseasoned

new issues. Journal of Finance 37, 1-10.

16. Eckbo B.E., Maksimovic V. and Williams J., 1990. Consistent Estimation of

Cross-Sectional Models in Event Studies. The Review of Financial Studies 3, 343-

365.

17. Ellis K., Michaely R. and O’Hara M., 2000. When the Underwriter is the Market

Maker: An Examination of Trading in the IPO After-market. Journal of Finance

55, 1039-1074.

18. Grinblatt M. and Hwang C. Y., 1989. Signalling and the pricing of New Issues.

Journal of Finance 44, 393-420.

19. Hanley K., Kumar A. A. and Geguin P. J., 1993. Price Stabilization in the Market

for New Issues. Journal of Financial Economics 34, 177-197.

20. Hughes P. J. and Thakor A. V., 1992. Litigation Risk, Intermediation, and the

Underpricing of Initial Public Offerings. Review of financial Studies 5, 709-742.

21. Ibbotson R. G., 1975. Price Performance of Common Stock New Issues. Journal

of Financial Economics 2, 235-272.

22. James C. and Wier P., 1990. Borrowing Relationships, Intermediation and the

Cost of Issuing Public Securities. Journal of Financial Economics 28, 149-171.

23. Jarque C. and Bera A., 1980. Efficient tests for normality, homoskedasticity and

serial independence of regression residuals. Economics Letters 6, 255-259.

24. Jegadeesh N., Weinstein M. and Welch I., 1993. An Empirical Investigation of

IPO Returns and Subsequuent Equity Offerings. Journal o Financial Economics

34 153-175.

25. Jenkinson T. and Ljungqvist A., 2001. Going Public-the theory and evidence on

how companies raise equity finance. Oxford University Press. Second Edition.

26. Jenson M. and Meckling W., 1976. Theory of the Firm: Managerial Behaviour,

Agency Costs and Ownership Structure. Journal of Financial Economics 3, 306-

360.

27. Johnson J. M. and Miller R. E., 1988. Investment Banker Prestige and the

Underpricing of Initial Public Offerings. Financial Management 17, 19-29.

28. Keloharju M., 1993. The Winner’s Curse, Legal Liability, and the Long-Run Price

Performance of Initial Public Offerings in Finland. Journal of Financial

Economics 34 251-277.

29. Kim S., Rui M. and Xu P., 1998. An Empirical Analysis on IPO Underpricing and

Performance of Newly Privatized Firms. Review of Pacific Basin Financial

Markets and Policies 1, 461-479.

30. Koh F. and Walter T., 1989. A Direct Test of Rock’s Model of the Pricing of

Unseasoned Issues. Journal of Financial Economics 23, 251-272.

31. Leland H. and Pyle D., 1977. Informational Asymmetries, Financial Structure, and

Finanical Intermediation. Journal of finance 32 371-387.

32. Levis M., 1990. The Winner’s Curse Problem, Interest Costs, and the

Underpricing of Initial Public Offerings. Economic Journal 100, 76-89.

33. McGuinness P., 1992. An Examination of the Underpricing of Initial Public

Offerings in Hong Kong: 1980-90. Journal of Business Finance and Accounting

19, 165-186.

34. Megginson W. and Weiss K. A., 1991. Venture Capitalist Certification in Initial

Public Offerings. Journal of Finance 46 879-903.

35. Michaely R. and Shaw W. H., 1994. the Pricing of Initial Public Offerings: Tests

of Adverse-Selection and Signaling Theories. Review of Financial Studies 7, 279-

319.

36. Mok M. K. and Hui Y. V., 1998. Underpricing and aftermarket performance of

IPOs in Shanghai, China. Pacific-Basin Finance Journal 6, 453-474.

37. Reilly F.K., 1977. New Issues Revisited. Financial Management 6 28-42.

38. Ritter J. R., 1984. The Hot Issue Market of 1980, Journal of Business 57, 215-240.

39. Ritter J. R., 1987. The Costs of Going Public. Journal of Financial Economics 19,

269-282.

40. Ritter J.R., 1991. The Long-Run Performance of Initial Public Offerings. Journal

of Finance 46 3-27.

41. Rock K., 1986. Why new issues are underpriced? Journal of Financial Economics

15, 187-212.

42. Ruud J. S., 1991. Another View of the Underpricing of Initial Public Offering.

Federal Reserve Bank of New York Quarterly Review 16, 83-85.

43. Ruud J. S., 1993. Underwriter Price Support and the IPO Underpricing Puzzle.

Journal of Financial Economics 34, 135-151.

44. Schultz P. H. and Zaman M. A., 1994. After-market Support and Underpricing of

Initial Public Offerings. Journal of Financial Economics 35, 199-219.

45. Smith C., 1986. Investment banking and the capital acquisition process, Journal of

Financial Economics 15, 3-30.

46. Stoughton N. M. and Zechner J., 1998. IPO Mechamisms, Monitoring and

Ownership Structure. Journal of Financial Economics 49, 45-78.

47. Su D. W. and Fleisher B. M., 1999. An empirical investigation of underpricing in

Chinese IPOs. Pacific-Basin Finance Journal 7, 173-202.

48. Tian L. G., 2003. Financial regulations, Investment risks, and Determinants of

Chinese IPO Underpricing. Working paper, Peking University Management

School and London Business School.

49. Tinic S. M., 1988.Anatomy of Initial Public Offerings of Common Stock. Journal

of Finance 43, 789-822.

50. Welch I., 1989. Seasoned Offerings, Imitation Costs, and the Underpricing of

Initial Public Offerings. Journal of Finance 44, 421-449.

51. Welch I., 1996. Equity Offerings Following the IPO: Theory and Evidence.

Journal of Corporate Finance 2, 227-259.

52. Wu J., 2001. Short-run performance and Valuation of IPO in China. Master

dissertation. National University of Singapore.

53. Xiang B., 1998. Institutional factors influence china’s accounting reforms and

standards. Accounting Horizons, 12, 105-119.

54. Yau H. M. O. and Steele H., 2000. China Business: Challenges in the 21st Century.

The Chinese University Press, 95-111.

ACKNOWLEDGEMENTS
TABLE OF CONTENTS
SUMMARY
LIST OF TABLES
LIST OF FIGURES
Chapter 1 Introduction

Motivation of the Study
1.2 Objectives of the study
1.3 Contribution of the study
1.4 Structure of the study.

Chapter 2 Models of IPO Underpricing and a Survey of Chinese Primary market

2.1 Models of IPO underpricing

2.1.1 Asymmetric Information
2.1.2 Institutional Explanations
2.1.3 Ownership and Control

2.2 Features of the Chinese Primary Market

2.2.1 The pre-offer process
2.2.2 Type of Shares in the Chinese stock market
2.2.3 The Issuing Mechanism
2.2.4 Supervision and Regulations
2.2.5 Other Characteristics Related to This Study

2.3 Prior Studies of the Chinese IPO underpricing
2.4 Possible Explanations for Chinese A-share IPO Underpricing

Chapter 3 Hypotheses and Methodology

3.1 The Winner’s Curse Model
3.2 Ex ante uncertainty
3.3 The signaling model

Group 1 - Correlations among IPO underpricing and
Group 2 – Linkages between IPOs and Seasoned Equi

Chapter 4 Data and Empirical Results

4.1 Data and underpricing
4.2 Allocation and Adverse selection
4.3 Ex ante uncertainty
4.4 The signaling model

Group 1- Correlations among IPO underpricing and
Group 2 - Linkages between IPOs and Seasoned Equity Offerings

Chapter 5 Conclusions
Appendix A: Offering Mechanism Changes in China

- 1. Limited lottery forms
- 2. Unlimited lottery forms
- 3. Unlimited number of deposit certificates
- 4. Full payment and pro rata allocation
- 5. Online fixed price offerings (Shang Wang Ding Jia)

Appendix B Correlation Matrix
Appendix C Test of the Winner's Curse Model
Bibliography