Investor Sentiment and IPO Pricing during Pre-Market and Aftermarket Periods

Li Jianga, *, Gao Lia

a School of Accounting and Finance, Hong Kong Polytechnic University, Hong Kong, China

This version: January 14, 2012

Abstract

In this study, we separately measure pre-market and aftermarket investor sentiment and investigate their impact on IPO pricing in a two-stage framework. We find that compensation for institutional investors is associated with their fractional allocation and the risk they bear. This helps explain the partial adjustment of offer prices to pre-market sentiment. We also show that pre-market sentiment tends to spill over to the aftermarket period, and aftermarket sentiment causes a further price run-up in the secondary market. Overall, our findings suggest that institutional investors play an important role in re-distributing shares in the secondary market and the momentum in investor sentiment makes it possible for underwriters to implement the staged distribution strategy.

JEL Classification Code: G02; G14

Keywords: Pre-market sentiment; Aftermarket sentiment; IPO pricing; Institutional investors

* Corresponding author. Tel.: +852 2766 4672; Fax: +852 2330 9845

E-mail addresses: afljiang@inet.polyu.edu.hk (L. Jiang); gaidy.li@connect.polyu.hk (G. Li)

1

1. Introduction

IPO underpricing phenomenon has been studied extensively in finance literature. So far,

evidence has shown that investor sentiment is an important determinant for IPO pricing and

investor sentiment causes issuers to set offer price above intrinsic value (Derrien, 2005; Cornelli

et al., 2006; Ljungqvist et al., 2006; and Dorn, 2009). In this study, we show that it is important

to examine investor sentiment separately for pre-market and aftermarket periods.

A recent study by Chen and Wilhelm (2008) focuses on the transition from pre-market

bookbuilding to the secondary market trading, to account for the sequential arrival of informed

traders. Underwriters adopt a staged distribution strategy in order to circumvent the restriction of

a uniform offer price and to extract greater surplus from the informed investors over time. We

adopt the two-stage framework to address for the possibility that sentiment investors may arrive

sequentially and pre-market sentiment may spill over to the aftermarket period. The transition

from primary to secondary markets gives rise to an important role played by institutional

investors to re-distribute shares in the secondary market (Ljungqvist et al., 2006). To extract

greater surplus from sentiment investors over time, underwriters adopt a similar staged

distribution strategy. However, the fact that most of prior IPO studies on investor sentiment use

initial return (offer to close return) implies that they have not paid sufficient attention to the

transition from pre-market stage to aftermarket trading.

In this study, we fill this gap by separating pre-market sentiment from aftermarket

sentiment and examining their impacts on IPO pricing in the two-stage framework. By

considering pre-market and aftermarket investor sentiment, we can address two important

questions: First, how institutional investors are compensated? In their theoretical model,

Ljungqvist et al. (2006) show that underwriter partially adjusts for investor sentiment and leaves

2

room for institutional investors, in exchange for their role of re-distributing shares in the

secondary market. Only by partitioning the transition into pre-market and aftermarket periods,

we can empirically test their prediction. Second, why would investors (sentiment investors in

particular) pay an even higher price than the inflated offer price when trading starts in the

secondary market? If investor sentiment has a momentum over time, sentiment investors at an

earlier stage would be willing to chase the hot issues as long as more optimistic sentiment

investors are expected to arrive later.

Hong Kong provides a unique institutional setting to study investor sentiment and IPO

pricing. First, Hong Kong IPOs have two separate tranches for retail and institutional investors.

The oversubscription rate for retail tranche indicates retail demand, which is subject to investor

sentiment. Second, retail investors have a prominent presence in Hong Kong market, making it

ideal to study the impact of investor sentiment on IPO pricing. Prior research suggests that Asian

markets are more prone to investor overconfidence (Huang et. al, 2011). Third, the claw-back

clause in Hong Kong IPOs implies that the fractional allocation to institutional investors is

inversely related to retail demand. This generates a greater variation in the fractional allocation

to institutional investors (as compared with a more stable allocation of 70% in the US), which

allows us to examine the determinants of compensation for institutional investors.

Our study proceeds as follows. First, we posit that underwriter revises offer price to

incorporate pre-market sentiment. We find supporting evidence that offer price revision is

positively related to oversubscription rate and investor attention. Offer-to-open return remains

positively related to pre-market sentiment, suggesting a partial adjustment to pre-market

sentiment.

3

Second, we posit that institutional investors are compensated for their role in the staged

distribution in the secondary market. We show that compensation for institutional investors is

associated with the risk facing institutional investors. Our findings provide an explanation for the

observed partial adjustment phenomenon and confirm the prediction by Ljungqvist et al. (2006).

Third, we posit that investor sentiment has a momentum over the transition from primary

market to the secondary market periods. We use small trade order imbalance and turnover as

proxies for aftermarket sentiment and find that pre-market sentiment proxies positively affect

aftermarket investor sentiment.

Finally, we posit that aftermarket sentiment pushes stock price higher. While small trade

order imbalance captures investor overconfidence, the divergence of opinion captured by trading

volume in conjunction with short-sales constraints causes overpricing (Miller, 1977). Our

findings support our prediction for a positive impact of aftermarket sentiment on secondary

market returns.

Our study makes the following important contributions: First, we shed lights on the

impact of investor sentiment on IPO pricing during the transition from primary market to

secondary market. Second, we identify the sources of investor sentiment at different stages.

Investor attention appears to drive retail demand at the pre-market stage, whereas the divergence

of opinion in conjunction with short-sales constraints leads to overpricing in the secondary

market. Third, by linking pre-market and aftermarket sentiments, we provide reasons for why

sentiment investors are willing to participate in the hot IPOs. Finally, by separately considering

pre-market and aftermarket periods, we establish the link between compensation for institutional

investors and their risk. We offer an explanation for partial adjustment to pre-market investor

4

sentiment and shed lights on the role played by institutional investors in re-distributing new

shares in the secondary market.

The rest of paper proceeds as follows. Section 2 reviews related studies and develops the

hypotheses. Section 3 describes the data and discusses the measurement of investor sentiment.

Section 4 presents the empirical results. Section 5 conducts further analysis. Section 6 concludes.

2. Related research and hypothesis development

In this study, we adopt the two-stage framework of Chen and Wilhelm (2008) in order to

address for the dynamics of investor sentiment over the transition from pre-market to aftermarket

period. Chen and Wilhelm (2008) focus on the transition to account for the sequential arrival of

informed traders. The same framework allows us to investigate the impact of sequentially

arriving sentiment investors on IPO pricing during pre-market and aftermarket stages. By

reviewing related research on IPO pricing, we can generate empirical implications under the new

framework.

We first examine the impact of pre-market investor sentiment on IPO offer price. Derrien

(2005) and Ljungqvist et al. (2006) show that underwriter sets the offer price above a firm’s

intrinsic value to take advantage of investor sentiment. Using abnormal Google search volume

index (ASVI) to proxy investor attention, Da et al. (2011) find that ASVI is positively related to

IPO price. Using the number of news headlines as proxy for investor sentiment during the

roadshow period, Cook et al. (2006) find a positive relationship between numbers of news

headlines and offer price revision. We make a similar prediction for the offer price revision in

relation to pre-market sentiment.

H1: IPO offer price revision is positively related to pre-market sentiment.

5

The partial adjustment of offer price is a well-known phenomenon. Benveniste and

Spindt (1989) propose that underwriter reward the informed investors for revealing their private

information by partially incorporating the private information into the offer price. Hanley (1993)

confirms their partial updating hypothesis. Loughran and Ritter (2002) explain the partial

adjustment phenomenon based on prospect theory that investors’ preference for wealth depends

on whether it is long-held or recently acquired. Edelen and Kadlec (2005) argue that issuers

partially adjust the offer price in order to increase the probability of completion. When issuer’s

surplus is large and therefore the marginal benefit of increasing the probability of success is high,

he would set a more conservative offer price to ensure the success of the offer. Bradley and

Jordan (2002) and Lowry and Schwert (2004) find that even public information is not fully

incorporated into the offer price. In this study, we predict that the offer price partially adjusts to

(publicly observable) investor sentiment during the pre-market period.

H2: Offer price partially adjusts to pre-market investor sentiment: Offer-to-open return is

positively related to pre-market sentiment.

We further examine the causes for the partial adjustment in the offer price. Prior research

suggests that the role played by regular investors in re-distributing shares in the secondary

market is a reasonable cause for the partial adjustment. Specifically, Ljungqvist et al. (2006)

suggest that regular investors are compensated for the risk of re-distributing shares in the

secondary market. Derrien (2005) argues that sentiment is only partially incorporated into the

offer price because underwriter intends to strike a balance between taking advantage of

sentiment investors and costly price support during the aftermarket period. To the extent that

6

partial adjustment in the offer price is intended to be the compensation for institutional investors,

there should be a connection between the compensation and the risk facing institutional investors.

H3: Partial adjustment is due to compensation for regular investors, which is positively

related to the fractional allocation and risk.

Next, we establish the link between pre-market and aftermarket sentiments. In their

theoretical model, Ljungqvist et al. (2006) show that pre-market demand by sentiment investors

exceeds their allocation and regular investors are allocated shares for them to re-distribute in the

secondary market. As a result, it is natural to expect that the unfulfilled demand may spill over to

aftermarket period. Kaustia and Knupfer (2008) show that investors overweight their personal

experiences in IPO and the reinforcement learning drives investor sentiment. Their findings

across IPOs can be extended to different stages of the same IPO, such that good experiences (in

terms of positive offer-to-open returns on the allocated shares) bolster investors’ confidence. In

addition, hot IPO issues and positive market reactions attract investors’ attention. Barber and

Odean (2008) argue that momentum investors would chase recent attention-grabbing stocks. We

therefore predict that pre-market sentiment tend to spill over to the aftermarket period.

H4: Pre-market sentiment is positively related to aftermarket sentiment.

Finally, we investigate whether aftermarket sentiment pushes up stock prices in the

secondary market trading. As informed and sentiment investors trade the newly listed shares, an

increase in divergence of opinion in conjunction with short sales constraints may lead to a

speculative premium (Miller, 1977). Using prices from pre-IPO grey market to proxy for retail

investors’ valuation, Cornelli et al. (2006) and Dorn (2009) show that overoptimistic retail

7

investors push the aftermarket IPO prices higher than its intrinsic value. Bradley et al. (2009)

find that information asymmetry and retail sentiment lead to an open-to-close return of 2.3% on

the first day of trading. Chan (2010) find that retail investors’ trading such as small trade order

imbalance is positively related to aftermarket returns. We make the following prediction:

H5: Aftermarket sentiment is positively related to open-to-close return.

3. Data and measurements

3.1. Data

We collect a sample of 567 IPOs in Hong Kong main board market from 1999 to 2009.

We limit our sample to IPOs using bookbuilding method and exclude IPOs of closed-end funds,

REITs, unit offering, companies switching from growth market. Our final sample contains 316

IPOs.

Firms in Hong Kong use dual-tranche mechanism in an IPO and reserve a fraction of IPO

shares for individual investors (“Public Tranche”). When applying for shares in an IPO, retail

investors are required to pay for the full cost of shares under their application. We download IPO

prospectus and the Factbook from the HKEx website to obtain offer price, price range,

oversubscription rate for the Public Tranche, proceeds for the Public Tranche and Placing

Tranche.

We obtain daily prices and trading volume from the Datastream. We obtain intraday

transaction prices and quotations from HKEx to calculate the return volatility. Since Loughran

and Ritter (2004) do not cover most of the Hong Kong underwriters in their ratings, we acquire

the annual market share of underwriters from Bloomberg.

3.2. Investor sentiment measures

8

In this study, we intend to capture investor sentiment for pre-market and aftermarket

stages. Previous studies use several indirect proxies to gauge investor sentiment:

oversubscription rate for public tranche (Derrien, 2005); the number of news headlines (Cook et

al., 2006); and price in grey market (Cornelli et al., 2006). We use two proxies for pre-market

sentiment.

First, the availability of oversubscription rate for retail tranche allows us to measure pre-

market retail demand for an IPO. Thus, we use oversubscription rate (SUBRATE) to proxy for

pre-market sentiment. Second, we use abnormal Google search volume index (ASVI) as our

second proxy for pre-market sentiment. Da et al. (2011) argue that Google search volume is a

direct and unambiguous measure of investor attention. We obtain the Google Insights1 Search

Volume Index (SVI) of company names in Chinese for our sample and compute the abnormal

SVI (ASVI) by calculating the percentage difference between the SVI in one week before the

listing day and the median SVI during the eight weeks earlier.

We use two proxies for aftermarket sentiment. First, we use turnover on the listing day to

proxy for aftermarket sentiment. Baker and Stein (2004) argue that liquidity is a proxy for

sentiment because sentiment investors systematically underestimate the information contained in

adverse trading decision, thus make the adverse price impact lower and result in more liquidity in

market. We obtain the turnover by dividing trading volume by the total number of shares

outstanding. Second, we use small trade order imbalance to proxy for aftermarket sentiment.

Following Lee and Ready (1991) procedure, we identify trades into buyer-initiated and seller-

initiated by comparing the transaction price with the midpoint of concurrent bid-ask quote. We

use dollar-value cut-off points at HK$ 50,000 and HK$ 500,000 to classify each of the trades

1 http://www.google.com/insights/search/#

9

into small, medium, and larger trades. Under the small trade category, we use the order

imbalance between buyer-initiated and seller-initiated trades to generate the net-buying measure

for small trade group and then scale it by the total dollar trading volume.

3.3. Descriptive statistics

Panel A of Table 1 provides the descriptive statistics of our sample. The average offer

price is HK$4.14 and the mean offer proceeds are approximately HK$4,371 million. More than

half of the firms (56%) on the main board use prestigious investment bank as their underwriter

(i.e., UWREP=1). The mean of initial return (offer-to-close return) on the first day is 13.47%.

However, once the initial return is decomposed into pre-market (offer-to-open) return and

aftermarket (open-to-close) return, offer to open return dominates the initial return (12.71%) and

open-to-close return has a mean close to zero (0.60%).

The IPO market in Hong Kong appears to be affected by investor sentiment. The average

oversubscription rate is 161 times higher than the number of shares assigned for retail tranche.

Although over-subscription is common in Hong Kong market, some firms are under-subscribed;

the lowest subscription rate is merely 0.02. The offer price revision has a mean of 3.48%,

indicating that underwriter typically increases offer price after soliciting private information and

gauging investor sentiment from bookbuilding process.

There is a significant variation in the price range for Hong Kong IPOs, ranging from 7%

to 67% of the mid-point offer price, and the average price range is 25%. This is in contrast with

the price range in the US, which is usually set at two dollars (Kutsuna et al., 2009). We use a

dummy variable, TOP, which equals one if the offer price reaches the upper bound of the price

range. About 40% of the IPOs are priced at the upper bound (i.e., TOP = 1).

10

Panel B provides year-by-year statistics. Our sample incorporates hot market periods like

2006-2007 and also the subprime crisis period in 2008. Although the first bookbuilding case was

introduced to Hong Kong in 1994, they were rare before 2003. Since 2003 the bookbuilding

method has prevailed for the IPOs in Hong Kong. The investor sentiment varies a great deal over

time. The IPO market in 2006 and 2007 was extremely hot in terms of oversubscription rates for

retail tranche; the median oversubscription rate was above 100 times. In contrast, IPOs in 2008

had a median of oversubscription rate of 2 times. In line with the oversubscription rate, the

average initial returns are about 22%-25% in 2006 and 2007, but merely 7% in 2008. Relatively,

the open-to-close return is less volatile than initial return. The mean and median 18-month

adjusted returns are negative for IPOs in 2006 and 2007, but positive for 2008. This suggests that

strong retail demand in hot IPO periods results in overpriced IPO shares, as compared with their

long-run prices.

Table 2 reports the correlation matrix that generates some preliminary results. The offer

price revision (REVISION) is positively correlated with each of the two pre-market sentiment

measures: abnormal Google search volume and oversubscription rate (ASVI and SUBRATE).

This suggests that underwriter incorporates pre-market sentiment in the offer price. However,

either ASVI or SUBRATE is positively correlated with offer-to-open returns (OTO), suggesting

that the offer price revision is only partial. The compensation for institutional investors is

positively correlated with either RANGE or VOLATILITY. The former indicates pre-market value

uncertainty, the latter indicates aftermarket value uncertainty. It is worth noting that

oversubscription rate (SUBRATE) is positively correlated with aftermarket volatility

(VOLATILITY). This reveals that those heavily over-subscribed IPOs tend to have greater

aftermarket value uncertainty. To the extent that the investor sentiment towards oversubscribed

11

stocks is more likely to change in aftermarket period, IPO shares held by regular investors

beyond their optimal inventory level would be risky.

Aftermarket trading volume (TURNOVER) is positively correlated with either ASVI or

SUBRATE, confirming our conjecture that pre-market sentiment tends to spill over to aftermarket.

However, small trade order imbalance (SMALLNET) is weakly correlated with pre-market

sentiment measures. The market-adjusted open-to-close return on the first day is significantly

and positively correlated with either SMALLNET or TURNOVER, suggesting that aftermarket

sentiment leads to significant secondary market return. The long-run adjusted return is negatively

correlated with pre-market and especially aftermarket sentiment measures,

[Insert Tables 1 and 2 about here]

4. Empirical results

In this section, we conduct several tests that investigate the impact of pre-market and

aftermarket sentiment on IPO pricing. We first examine whether pre-market sentiment positively

affects offer price revision. After documenting offer price is only partially adjusted to pre-market

sentiment, we then examine whether compensation for institutional investors is related to the risk

for re-distributing new shares in the secondary market. We also examine whether pre-market

investor sentiment spills over to aftermarket period and test if aftermarket sentiment causes

further overpricing.

4.1. Pre-market sentiment and offer price revision

Hypothesis 1 predicts a positive relationship between pre-market sentiment and offer

price revision. In this section, we use the oversubscription rate (SUBRATE) and abnormal search

volume index (ASVI) to proxy for pre-market investor sentiment. We test their impact on the

12

offer price revision. In addition, we also include several control variables into our regressions to

control other IPO characteristics that would influence IPO price.

Table 3 presents the regression results. The coefficient of subscription rate (SUBRATE) is

positive and significant at the 1% level. Similarly, the coefficient of abnormal Google search

index (ASVI) is positive and significant at the 1% level. This suggests that pre-market sentiment

positively influences the offer price revision and Barber and Odean’s (2008) attention-induced

price pressure hypothesis holds for IPOs. This impact is also economically significant. On

average, a one standard deviation increase in subscription rate leads to a 61% higher price

relative to initial midpoint price. In addition, Derrien (2005) suggests that the bullishness of

investors towards an IPO is determined by current market condition and finds positive

association between market returns and offer price. Thus, we include the average initial return of

prior five IPOs (PRE_IPO_RTN) in the regression. Its coefficient is positive and significant at

the 1% level, consistent with the finding of Derrien (2005). Our findings suggest that both firm-

specific and market-wide pre-market sentiment cause the underwriter to revise the offer price

upward. The coefficient of price range (RANGE) is negative and significant at the 5% level,

suggesting that price range proxies for ex-ante value uncertainty. Economically, increasing price

range by one standard deviation will cause offer price to drop by nearly 17%-22% on average.

Our result is consistent with Kutsuna et al. (2009) that value uncertainty negatively influences

offer price.

Overall, the results in Table 3 confirm that underwriter takes into consideration of pre-

market investor sentiment, as reflected by oversubscription and abnormal Google search volume,

and revises the offer price upward to capitalize on investor sentiment.

[Insert Table 3 about here]

13

4.2. Partial adjustment to pre-market sentiment

In this section, we investigate whether underwriter adjusts the offer price fully to

incorporate information (including investor sentiment) gathered during the pre-market period.

Hanley (1993) finds that offer price revision is positively correlated with initial return,

suggesting a partial adjustment of offer price. In this regard, we regress offer-to-open returns on

pre-market sentiment measures to test our Hypothesis 2.

Table 4 presents the regression results. The coefficients of oversubscription rate

(SUBRATE) are positive and significant at the 1% level, suggesting that underwriter only

partially responds to pre-market sentiment when setting the offer price. On average, a one

standard deviation increase in subscription rate generates a return of 119% to 126% assuming

flipping new shares once the market trading starts. The result based on ASVI as the proxy for pre-

market sentiment is similar though the coefficients of ASVI are positive and significant at the 5%

level. The coefficients of PRE_IPO_RTN are positive and significant at the 5% or 1% level,

suggesting that the offer price does not fully incorporate market-wide investor sentiment. We

conclude that underwriter does not fully take advantage of investor sentiment when setting the

offer price, and leaves some profit to investors who acquire new shares at the offer price. These

results are consistent with previous findings that investor sentiment causes IPO partial

adjustment (Derrien, 2005; Cornelli et al., 2006).

[Insert Table 4 about here]

4.3. Compensation for institutional investors

So far, our findings indicate that underwriter partially adjusts the offer price in response

to pre-market investor sentiment. In this section, we attempt to provide an explanation for the

14

partial adjustment phenomenon by examining the compensation for institutional investors.

Specifically, we investigate whether compensation for institutional investors is positively related

to their fractional allocation and value uncertainty of an IPO.

In this study, we construct a measure for compensation for institutional investors, which

is defined as the offer-to-open return times the number of shares allocated to institutional

investors divided by total asset. This measure is more relevant than offer-to-open return itself

since it focuses directly on the reward for institutional investors. To remove the fractional

allocation to institutional investors that can be explained by pre-market sentiment due to the

Claw-back Clause in Hong Kong, we regress PLACING on SUBRATE and use the residual

(R_PLACING) in the regression. We use price range (RANGE) as a proxy for pre-market value

uncertainty and use the first-day return volatility (VOLATILITY) as proxy for aftermarket

uncertainty. To investigate whether compensation for institutional investors depends on the

“hotness” of IPO, we divide the full sample into two subsamples based on the sign of the offer

price revision. 237 IPOs are classified into “hot” IPO group and 79 into “cold” IPO group in our

subsample tests.

Table 5 presents the regression results. The coefficients of fractional allocation

(R_PLACING) are positive and significant at the 10% to 1% levels, for the full sample and hot-

IPO subsample, respectively. Specifically, in column 1, one standard deviation increase in

institutional placing results in an average 23% higher compensation for institutional investors. In

contrast, the coefficient of R_PLACING is not significant for the cold-IPO subsample. This

indicates that the positive relationship between fractional allocation to institutional investors and

their compensation holds only for hot IPOs.

15

Table 5 also reports the results on the impact of risk on the compensation for institutional

investors. The coefficients of price range (RANGE) are positive and significant at the 1% level,

for hot-IPO subsamples. For hot IPO subsample, a one standard deviation wider in price range

causes underwriter on average to give up 26% of the compensation to institutional investors for

bearing the risk. Similarly, the coefficients of VOLATILITY are positive and significant at the 1%

level, for both the full sample and hot-IPO subsamples (columns 2 and 4). However, the

coefficient of VOLATILITY for cold IPOs is insignificant. This result implies that the

compensation for aftermarket uncertainty depends largely on market conditions; underwriter

offers rewards to institutional investors for firm-specific uncertainty only when market is hot.

It is interesting to note that the coefficient of oversubscription rate (SUBRATE) is no

longer significant, while the sign of coefficients is even negative for hot IPO subsample

(columns 3 and 4). For hot IPOs, underwriter chooses (and the claw-back clause requires the

underwriter) to sell more shares directly to retail investors, rather than relies on institutional

investors. Therefore, underwriter may compensate institutional investors less when investor

sentiment is high.

We further investigate whether underwriter reputation affects the compensation for

institutional investors. We collect the yearly market share of underwriters in Hong Kong from

Bloomberg. We then construct a dummy variable UWREP which has a value of 1 for the top ten

underwriters in ranking and otherwise 0. As shown in Table 5, underwriter reputation is

negatively related to compensation for institutional investors at the 5% significance level.

Moreover, we find that the reputation effect is more pronounced in the hot market. As shown in

columns 3 to 6, high-reputation investment banks provide less compensation for hot IPOs than

16

for cold IPOs, consistent with Daniel (2002) that underwriter’s bargaining power is stronger in

good business period.

Overall, Table 5 provides empirical evidence supporting our hypothesis (H3) regarding

compensation for institutional investors. The results show institutional investor’s fractional

allocation and valuation uncertainty are important determinants of their compensation.

Furthermore, the observed partial adjustment to pre-market sentiment can be explained by the

compensation for institutional investors.

[Insert Table 5 about here]

4.4. Spill-over from pre-market sentiment to aftermarket sentiment

In this section, we investigate whether pre-market sentiment spills over to the aftermarket

period. Table 6 presents the regression results. In Panel A, the dependent variable is the first-day

turnover (TURNOVER), our first proxy for aftermarket sentiment. The coefficient of

oversubscription rate (SUBRATE) in columns 1 is positive and significant at the 5% level. In

Panel B, the dependant variable is small trade order imbalance (SMALLNET), our second proxy

for aftermarket sentiment. Chan (2010) proposes that small trade order imbalance increases

open-to-close return in the secondary market. The coefficient of SUBRATE is positive and

significant at the 10% level, providing similar evidence that pre-market and aftermarket

sentiments are positively linked.

In addition, offer price revision may serve as a signal that stimulates aftermarket

sentiment2. To test the signaling effect, we include in the regression price revision (REVISION)

and a dummy variable (TOP) to indicate offerings at the upper bound of the price range. As

2 Cornelli et al. (2006) constructs a dummy variable indicating whether grey market price exceeds midpoint price of price range.

Using it as their pre-market sentiment proxy, they find that this dummy variable effectively explains aftermarket turnover.

17

shown in columns 1 and 4, price revision shows positive but insignificant relation to our two

aftermarket sentiment proxies. However, the coefficients of TOP in columns 1 and 2 are positive

and significant at the 5% or 1% level. The fact that the revised offer price is set at the upper

bound of the price range attracts investors’ attention and generates aftermarket trading.

Several other characteristics also exert significant impact on aftermarket sentiment. SIZE

is negatively related to turnover at the 5% significance level, implying sentiment investors are

more likely to speculate with small-cap firms, which are more subject to information asymmetry.

This is consistent with prior studies (Ritter and Welch, 2002; Ofek and Richardson, 2003; and

Kumar and Lee, 2006). The positive coefficients of UWREP are also consistent with the claim of

Cook et al. (2006) that underwriter’s marketing effort enhances investor sentiment.

In sum, pre-market investor sentiment tends to spill over to the aftermarket stage. It

appears that the unfulfilled demand by sentiment investors in pre-market period continue to

affect aftermarket trading, and the initial market response to the IPO attract additional sentiment

investors to participate in the aftermarket trading. The fact that investor sentiment exhibits a

momentum over time can help explain why sentiment investors would be willing to chase the hot

issues even though pre-market sentiment is likely to produce an overpriced offer price.

[Insert Table 6 about here]

4.5. Aftermarket sentiment and secondary market returns

In this section, we investigate whether aftermarket sentiment further drives IPO

overpricing in the secondary market. Table 7 reports the regression that examines the

relationship between aftermarket sentiment and open-to-close return. Through columns 3 to 6,

the coefficients of TURNOVER and SMALLNET are positive and significant at the 1% level.

18

Based on the parameter estimates in column 3, a one standard deviation change in turnover ratio

leads to a 33% change in open-to-close return in the same direction. In column 4, a one standard

deviation increase in the small trade order imbalance leads to a 61% increase in the open-to-close

return. Thus, aftermarket sentiment affects open-to-close return with both statistical and

economical significance. In columns 5 and 6, we include both TURNOVER and SMALLNET in

the regressions. Their coefficients remain positive and significant at the 1% level. These results

provide strong empirical evidence that aftermarket sentiment further increases the IPO

overpricing in the secondary market.

Interestingly, we find that our pre-market sentiment measures, oversubscription rate

(SUBRATE) and abnormal search volume index (ASVI), as well as prior IPO returns

(PRE_IPO_RTN), lose their power in explaining the open-to-close returns. This is different from

previous studies with the offer-to-close return as the dependant variable, and implies that

investors have fully incorporated pre-market sentiment into the opening price when trading starts

in the secondary market. Another worth-noting variable is VOLATILITY. Falconieri et al. (2009)

use intraday volatility as the proxy for aftermarket uncertainty and find it positively related to

open-to-close return. The coefficients of VOLATILITY through columns 3 to 6 are positive and

significant at the 1% level, suggesting that aftermarket uncertainty contributes to secondary

market return.

In columns 6, we incorporate additional control variables, including large trade order

imbalance (LARGENET) and institutional fractional allocation (R_PLACING). Only coefficient

of LARGENET is positive and significant at the 1% level. Chan (2010) finds that medium and

large trades explain aftermarket price movement in cold or neutral market, while small trades can

only predict share price in hot market. Our finding indicates that net buying by large investors

19

also contributes to IPO price run-up in the aftermarket. Overall, Table 7 shows that aftermarket

sentiment results in a further price run up once the secondary market trading starts. In contrast,

pre-market sentiment no longer has any impact on open-to-close return in the secondary market.

[Insert Table 7 about here]

5. Further Analysis

5.1. Investor attention and retail demand

In this section, we further investigate the potential source of pre-market retail demand for

IPOs. Investor attention is a major factor in determining the purchase decision by individual

investors (Barber and Odean, 2008 and Barber et al., 2009). Da et al. (2011) show that Google

search volume (ASVI) is a direct and timely measure of investor attention. Using it to proxy for

individual investor attention, they find increased attention leads to high initial return and

subsequent long-run underperformance of IPOs.

Table 8 presents the regression results to test whether investor attention drives

oversubscription rate (SUBRATE). As expected, the coefficient of ASVI is positive and

significant at the 1% level. This suggests that retail investors’ attention indeed increases the

oversubscription rate of the IPO. Similarly, the coefficient of prior IPO returns (PRE_IPO_RTN)

is also positive and significant at the 1% level. In contrast, the coefficient of price range (RANGE)

is negative and significant at the 5% level. Overall, our findings confirm that those attention-

grabbing IPOs are more subject to the demand by sentiment investors.

[Insert Table 8 about here]

5.2. Long-run underperformance

20

In this section, we examine whether pre-market and aftermarket sentiment measures can

explain the long-run underperformance of IPOs. A consensus view is that as sentiment fades

away and short-sale constraints are lifted, the overpriced share price is expected to revert to its

normal level in the long run. Previous research provides both theoretical argument (Ritter and

Welch, 2002; Ljungqvist et al. 2006) and empirical evidences (Derrien, 2005; Cornelli et al.,

2006; Agarwal et al., 2008; Dorn, 2009; Da et al., 2011) of IPO long-run underperformance. In

this study, we adopt similar methods to test whether such phenomenon exists.

First, we calculate the buy-and-hold returns of IPOs in the 6-month, 1-year, and 18-

month horizons from the first-day closing price. Table 9 presents the reports on mean and

median values in each holding period. The long-run median market-adjusted returns are -

5.58%�-11.52%, and -16.90%, respectively, indicating an increasing underperformance relative

to the benchmark market index (Hang Seng Index). To further investigate the cause of long-run

underperformance, we run regressions with the long-run market adjusted returns as the

dependent variables.

Table 10 reports the estimation results of 1-year market-adjusted returns. Various

sentiment measures are our focus. The coefficients of aftermarket sentiment measures,

TURNOVER and SMALLNET, are negative and significant at the 5% or 10% level. This is in line

with Barber et al. (2009) that purchase by small investors pushes up price and results in a

subsequent price reversal. In contrast, the coefficients of pre-market sentiment measures,

SUBRATE and ASVI, are not significant. This is different from Agarwal et al. (2008) that pre-

market sentiment is negatively related to long-run performance. The difference might be due to

our incorporation of aftermarket sentiment. The coefficients of SIZE are both positive and

21

significant. Since big firms are often less opaque, they might be less influenced by investor

sentiment. Overall, Table 10 confirms that aftermarket sentiment causes IPO overpricing.

[Insert Tables 9 and 10 about here]

6. Conclusion

In this study, we examine the impact of investor sentiment over pre-market and

aftermarket stages on IPO pricing. By allowing sentiment to evolve over the transition from

primary market to secondary market trading, we are able to show empirically how institutional

investors are compensated for re-distributing new shares and who are leaving the money on the

table. Our findings can be summarized as follows: Underwriter capitalizes on pre-market

investor sentiment by revising the offer price upward. Investor attention appears to drive pre-

market retail demand for IPO. The offer price revision is only partial in the sense that pre-market

sentiment still affects offer-to-open return. We empirically confirm the prediction by Ljungqvist

et al. (2006) that compensation for regular investors is one of the main reasons for partial

adjustment.

We also shed lights on why sentiment investors are willing to participate in the IPOs by

paying an even higher price than the offer price in the secondary market. We show that pre-

market sentiment tends to spill over to aftermarket period, suggesting that additional sentiment

investors may arrive in the aftermarket stage. The aftermarket sentiment pushes up the stock

price further, making those sentiment investors who participate in the early stage benefit from

the sequential arrival of sentiment investors. The long-term underperformance confirms that

sentiment eventually fades away and overpricing is corrected over time. However, the presence

22

of investor sentiment and its momentum make it possible for underwriter to successfully

implement the staged distribution strategy.

23

References

Agarwal, S., Liu C., Rhee S. G., 2008, Investor demand for IPOs and aftermarket performance: Evidence from the Hong Kong stock market, International Financial Market, Institutions & Money 18, 176-190.

Baker, M., Stein, J. C., 2004. Market liquidity as a sentiment indicator. Journal of Financial Markets 7, 271-299.

Barber, B. M., Odean T., 2008, All that glitters: The effect of attention and news on the buying behavior of individual and institutional investors, Review of Financial Studies 21, 785 – 818.

Benveniste, L. M., Spindt P. A., 1989, How investment bankers determine the offer price and allocation of new issues? Journal of Financial Economics 24, 343-361.

Bradley, D. J., Gonas, J. S., Highfield, M. J., Roskelley, K. D., 2009. An examination of IPO secondary market returns. Journal of Corporate Finance 15, 316-330.

Bradley, D. J., Jordan, B. D., 2002. Partial adjustment to public information and IPO underpricing. Journal of Financial and Quantitative Analysis 37, 595 – 616.

Chan, Y. C., 2010. Retail trading and IPO returns in the aftermarket, Financial Management 1475-1495.

Chen, Z., Wilhelm, W., 2008. A theory of the transition to secondary market trading of IPOs. Journal of Financial Economics 90, 219-236.

Cook, D., Kieschnick, R., Van Ness, R., 2006. On the marketing of IPOs. Journal of Financial Economics 82, 35-61.

Cornelli, F., Goldreich, D., Ljungqvist, A., 2006. Investor sentiment and pre-IPO markets. Journal of Finance 61, 1187-1216.

Da, Z., Engelberg, J., Gao, P., 2011. In search of attention. Journal of Finance 66, 1461-1499.

Daniel, K., 2002. Discussion of “Why don’t issuers get upset about leaving money on the table in IPOs?”. Review of Financial Studies 15, 445-454.

Derrien, F., 2005. IPO pricing in “hot” market conditions: Who leaves money on the table? Journal of Finance 60, 487-520.

Dorn, D., 2009, Does sentiment drive the retail demand for IPOs? Journal of Financial and Quantitative Analysis 44, 85-108.

24

Edelen, R. M., Kadlec, G. B., 2005, Issuer surplus and the partial adjustment of IPO prices to public information. Journal of Financial Economics 77, 347-373.

Espinasee, P., 2011, IPO: A Global Guide. The Hong Kong University Press.

Falconieri, S., Murphy, A., Weaver, D., 2009. Underpricing and ex post value uncertainty. Financial Management 285-300.

Hanley, K. W., 1993. The underpricing of initial public offerings and the partial adjustment phenomenon, Journal of Financial Economics 34, 231-250.

Huang, Z., Heian, J.B., Zhang, T., 2011. Differences of opinions, overconfidence, and the high-volume premium. Journal of Financial Research 34, 1-25.

Kaustia, M., Knupfer, S., 2008. Do investors overweight personal experience? Evidence from IPO subscriptions. Journal of Finance 63, 2679 – 2702.

Kumar, A., Lee, M.C., 2006. Retail investor sentiment and return comovements. Journal of Finance 61, 2451-2486.

Kutsuna, K., Smith, J. K., Smith, R. L., 2009. Public information, IPO price formation, and long-run returns: Japanese evidence. Journal of Finance 64, 505-546.

Lee, M. C., Ready, M., 1991. Inferring trade direction from intraday data. Journal of Finance 2, 733-746.

Ljungqvist, A., Nanda, V., Singh, R., 2006. Hot markets, investor sentiment and IPO pricing. Journal of Business 79, 1667-2702.

Loughran, T., Ritter, J. R., 2002. Why don’t issuers get upset about leaving money on the table in IPOs? Review of Financial Studies 15, 75-109.

Lowry, M., Schwert, G. W., 2004. Is the IPO pricing process efficient? Journal of Financial Economics 71, 3-26.

Miller, E. M., 1977. Risk, uncertainty, and divergence of opinion. Journal of Finance 32, 1151-1168.

Ofek, E., Richardson, M., 2003. Dot-com mania: The rise and fall of Internet stock prices. Journal of Finance 58, 1113-1138.

Ritter, J. R., Welch, I., 2002. A review of IPO activity, pricing and allocations. Journal of Finance 57, 1795-1828.

25

Table 1 Descriptive Statistics

The full sample consists of 316 IPOs in Hong Kong from 1999 to 2009. The sample is restricted to IPOs using bookbuilding method. We exclude IPOs of closed-end funds, REITs, unit offerings, and companies switching from growth market. OFFER_PRICE is the offer price after price revision. PLACING is the portion of shares allocated to the Placing Tranche in an IPO. SUBRATE is the number of shares subscribed by individual investors divided by the number of shares assigned to the Public tranche. PRE_IPO_RTN is the average initial return of five latest IPOs before the IPO. RANGE is the price range announced in its prospectus divided by its midpoint price. SIZE is the logarithm of total asset. PROCEEDS is the amount raised in millions of HK dollars. UWREP is a dummy variable which equals one if one of sponsors is among the top ten based on the underwriting market share ranking. REVISION is the offer price divided by the midpoint of initial price range minus one. TOP is a dummy variable equals 1 when offer price is set at the upper bound of the price range. OTO is the offer to open return, that is, the open price on the first trading day divided by the offer price minus one. SMALLNET is the buyer-initiated small trades minus seller-initiated small trades divided by the total dollar trading volume on the first trading day. TURNOVER is the total trading volume divided by the number of shares outstanding on the first-trading day. LARGENET is the buyer-initiated large trades minus seller-initiated large trades divided by the total dollar trading volume on the first trading day. ADJOTC is the market-adjusted open-to-close return. VOLATILITY is the standard deviation of intraday prices on first trading day normalized by the offer price. COMPEN is the compensation for institutional investors and calculated as the offer to open return times the number of shares allocated to institutional investors scaled by total asset. ASVI is the abnormal Google search volume, defined as the search volume index during the bookbuilding week minus the median of search volume index in the previous eight weeks. 6MADJRTN, 1YADJRTN, and 18NADJRTN are 6-month, 1-year and 18-month buy-and-hold market-adjusted returns from close price on the first trading day. Panel A: Full Sample Variable N Mean Median Maximum Minimum Std Dev

OFFER_PRICE 316 4.14 2.70 84.00 0.34 5.99

PLACING 316 69.32 69.78 99.34 0.00 18.94

SUBRATE 316 161.81 54.85 1703 0.07 242.44

PRE_IPO_RTN 316 12.97 8.84 61.92 -13.59 14.12

RANGE 316 24.82 24.61 66.67 7.00 8.32

SIZE 316 22.03 21.73 29.65 18.01 2.04

PROCEEDS 316 4371.77 1260.83 124947.90 2.75 10919.99

UWREP 316 0.56 1 1 0 0.50

REVISION 316 3.48 6.51 20.42 -33.33 10.02

TOP 316 0.40 0 1 0 0.49

OTO 316 12.71 5.95 122.22 -27.62 20.60

SMALLNET 316 -0.15 -0.15 36.35 -51.58 5.26

TURNOVER 316 0.48 0.33 5.84 0.01 0.59

LARGENET 316 -2.09 -0.13 51.75 -61.85 12.39

ADJOTC 316 0.60 -0.03 76.02 -27.99 10.61

VOLATILITY 316 3.27 2.40 23.45 0.24 2.84

ASVI 158 75.02 41.67 900.00 -52.84 116.55

COMPEN 316 3.68 0.94 53.78 -27.68 9.19

6MADJRTN 316 -2.09 -5.09 220.19 -100.15 46.01

1YADJRTN 316 -1.10 -10.54 381.49 -108.91 66.75

18MADJRTN 316 1.91 -16.60 977.46 -132.33 101.58

26

Table 1 Descriptive Statistics (Continued)

Thi

s ta

ble

prov

ides

yea

r-by

-yea

r de

script

ive

stat

istic

s fo

r ou

r sa

mpl

e. N

is

the

num

ber

of I

POs. O

FFER_P

RIC

E i

s th

e of

fer

pric

e af

ter

pric

e re

visi

on. SU

BRATE

is

the

ratio

be

twee

n th

e to

tal d

eman

d su

bmitt

ed b

y in

divi

dual

inve

stor

s du

ring

the

appl

icat

ion

period

, and

the

amou

nt o

f sha

res as

sign

ed to

pub

lic tr

anch

e. P

RO

CEED

S is th

e am

ount

rai

sed

in

mill

ions

of

HK

dol

lars

. U

WREP is

a du

mm

y va

riab

le w

hich

equ

als

one

if o

ne o

f sp

onso

rs is

amon

g th

e to

p te

n ba

sed

on the

und

erw

ritin

g m

arke

t sh

are

rank

ing.

AD

JIR a

nd

AD

JOTC

are

mar

ket-ad

just

ed o

ffer

-to-

clos

e an

d op

en-to-

clos

e re

turn

s, res

pect

ivel

y. 1

8MAD

JRTN

is 1

8-m

onth

buy

-and

-hol

d m

arke

t-ad

just

ed ret

urns

fro

m c

lose

price

on

the

firs

t trad

ing

day.

Panel B: B

y Year

Yea

r N

OFFERPRIC

E

PRO

CEED

S (H

K$

Mil)

S

UBRATE

AD

JIR

AD

JTO

C

18M

AD

JOTC

Mea

n M

edia

n M

ean

Med

ian

Mea

n M

edia

n M

ean

Med

ian

Mea

n M

edia

n M

ean

Med

ian

1999

1 1.

70

1.70

71

4.68

71

4.68

32

.50

32.5

0 -1

7.22

-1

7.22

-2

.03

-2.0

3 10

7.30

10

7.36

2000

7 4.

61

1.59

15

249.

40

1078

7.00

14

.48

4.06

0.

97

-2.2

7 -5

.76

-4.4

5 45

.57

28.0

2

2001

5 4.

20

4.10

35

84.6

9 12

74.5

0 22

.64

7.14

6.

90

8.95

3.

40

7.78

26

.49

36.7

5

2002

9 12

.45

1.68

45

30.3

4 16

37.0

3 9.

43

3.70

2.

87

4.74

0.

88

1.50

-0

.54

7.26

2003

20

3.44

2.

04

2745

.98

1045

.41

130.

61

64.8

0 16

.47

10.9

1 2.

45

4.08

2.

20

-4.0

8

2004

31

4.34

2.

50

2351

.82

1040

.37

157.

80

55.7

0 4.

40

1.73

-1

.40

-0.9

6 5.

60

-17.

16

2005

45

2.36

1.

70

3593

.18

517.

50

95.6

0 14

.00

4.40

1.

95

-0.4

4 -0

.38

37.7

5 16

.18

2006

47

3.16

2.

28

7009

.63

1367

.21

257.

54

142.

70

25.3

2 19

.91

2.26

1.

39

-17.

17

-43.

73

2007

72

4.64

3.

84

3951

.32

1866

.45

234.

64

169.

00

22.4

7 9.

51

1.38

-1

.06

-28.

77

-29.

03

2008

22

3.15

2.

77

2924

.92

2115

.20

32.1

1 2.

07

7.17

-0

.11

2.61

0.

51

18.7

1 2.

66

2009

57

4.91

3.

45

4342

.23

1620

.00

163.

02

57.7

0 10

.14

6.91

-0

.47

-0.5

9 11

.92

-4.9

8

All

316

4.14

2.

70

4371

.77

1260

.83

161.

82

54.8

5 13

.47

6.08

0.

60

-0.0

3 1.

91

-16.

86

27

Table 2 Correlation M

atrix

PLA

CIN

G is

the

portio

n of

sha

res

allo

cate

d to

the

Pla

cing

Tra

nche

in

an I

PO. SU

BRATE

is

the

num

ber

of s

hare

s su

bscr

ibed

by

indi

vidu

al inv

esto

rs d

ivid

ed b

y th

e nu

mbe

r of

sh

ares

ass

igne

d to

the

Publ

ic tr

anch

e. P

RE_I

PO

_RTN

is th

e av

erag

e in

itial

ret

urn

of fiv

e la

test

IPO

s be

fore

the

IPO

.RAN

GE is

the

pric

e ra

nge

anno

unce

d in

its pr

ospe

ctus

div

ided

by

its

mid

poin

t price

. SIZ

E is

the

loga

rith

m o

f to

tal a

sset

. PRO

CEED

S is

the

loga

rith

m o

f th

e am

ount

rai

sed

in m

illio

ns o

f H

K d

olla

rs. R

EVIS

ION

is th

e of

fer pr

ice

divi

ded

by th

e m

idpo

int o

f in

itial

price

ran

ge m

inus

one

. OTO

is th

e of

fer to

ope

n re

turn

, tha

t is, th

e op

en p

rice

on

the

firs

t tra

ding

day

div

ided

by

the

offe

r pr

ice

min

us o

ne. S

MALL

NET

is th

e bu

yer-

initi

ated

sm

all t

rade

s m

inus

sel

ler-

initi

ated

sm

all t

rade

s di

vide

d by

the

tota

l dol

lar t

radi

ng v

olum

e on

the

firs

t tra

ding

day

. TU

RN

OVER is

the

tota

l tra

ding

vol

ume

divi

ded

by

the

num

ber

of s

hare

s ou

tsta

ndin

g on

the

first

-tra

ding

day

. LA

RG

EN

ET

is the

buy

er-ini

tiate

d la

rge

trad

es m

inus

sel

ler-

initi

ated

lar

ge t

rade

s di

vide

d by

the

tot

al d

olla

r trad

ing

volu

me

on th

e firs

t tra

ding

day

. AD

JOTC

is th

e m

arke

t-ad

just

ed o

pen-

to-c

lose

ret

urn.

VO

LATI

LITY

is th

e st

anda

rd d

evia

tion

of in

trad

ay p

rice

s on

first

trad

ing

day

norm

aliz

ed b

y th

e of

fer pr

ice.

ASV

I is

the

abno

rmal

Goo

gle

sear

ch v

olum

e, d

efin

ed a

s th

e se

arch

vol

ume

inde

x du

ring

the

book

build

ing

wee

k m

inus

the

med

ian

of s

earc

h vo

lum

e in

dex

in the

pr

evio

us e

ight

wee

ks.C

OM

PEN

is

the

com

pens

atio

n fo

r in

stitu

tiona

l in

vest

ors

and

calc

ulat

ed a

s th

e of

fer

to o

pen

retu

rn tim

es t

he n

umbe

r of

sha

res

allo

cate

d to

ins

titut

iona

l in

vest

ors sc

aled

by

tota

l ass

et. 1

YAD

JRTN

is 1

-yea

r buy

-and

-hol

d m

arke

t-ad

just

ed re

turn

s from

clo

se p

rice

on

the

firs

t tra

ding

day

.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

� 8

�

� 9

�

� 10

� � 11�

(12)

(13)

(1

4)

(15)

(1

6)

PLA

CIN

G

(1)

1

SUBRATE

(2)

-0.4

873

1

PRE_I

PO

_RTN

(3)

-0.3

859

0.32

18

1

RAN

GE

(4)

0.17

87

-0.1

344

-0.1

762

1

SI

ZE

(5)

0.17

49

-0.0

049

0.04

32

-0.1

449

1

PRO

CEED

S (6

) -0

.037

4 0.

0588

0.

1622

-0

.192

4 0.

8232

1

REVIS

ION

(7)

-0.5

831

0.40

83

0.34

49

-0.1

492

0.01

1 0.

0506

1

O

TO

(8)

-0.4

450

0.57

34

0.36

26

-0.1

439

0.02

08

0.14

35

0.39

69

1

SM

ALL

NET

(9)

-0.1

476

-0.0

515

-0.1

449

-0.0

263

-0.0

947

-0.1

496

-0.0

583

0.00

54

1

TURN

OVER

(10)

-0

.483

1 0.

3628

0.

2488

-0

.166

9 -0

.083

2 0.

1292

0.

3876

0.

6172

0.

2307

1

LARG

EN

ET

(11)

-0

.207

1 0.

2094

0.

0641

-0

.153

6 -0

.114

1 -0

.104

6 0.

2843

0.

2655

0.

4081

0.

3132

1

AD

JOTC

(12)

-0

.073

2 -0

.057

1 -0

.060

5 -0

.026

5 -0

.020

3 -0

.119

3 0.

0389

0.

0695

0.

6500

0.

2673

0.

4920

1

VO

LATI

LITY

(13)

-0

.244

9 0.

0612

0.

1241

-0

.192

2 -0

.020

0 0.

1206

0.

0557

0.

4427

0.

1844

0.

3189

0.

2164

0.

3272

1

ASV

I (1

4)

-0.2

098

0.36

72

0.04

29

0.14

62

0.01

45

0.01

24

0.29

49

0.25

17

0.04

19

0.16

44

0.08

07

-0.0

061

0.00

29

1

C

OM

PEN

(1

5)

-0.2

913

0.13

53

0.27

77

0.00

20

-0.1

100

0.10

20

0.34

34

0.53

02

0.07

41

0.20

62

0.13

29

0.04

38

0.30

66

0.06

33

1

1YRTN

(16)

0.

2435

-0

.062

8 -0

.075

3 -0

.033

0 0.

1591

0.

0523

-0

.144

7 -0

.116

8 -0

.198

7 -0

.270

1 -0

.178

9 -0

.066

5 -0

.099

1 -0

.103

7 -0

.016

4 1

28

Table 3 Pre-Market Sentiment and Offer Price Revision

This table tests of Hypothesis 1. The dependant variable, REVISION, is the offer price divided by the midpoint of initial price range minus one. Explanatory variables: SUBRATE is the number of shares subscribed by individual investors divided by the number of shares assigned to the Public tranche. ASVI is the abnormal Google search volume, defined as the search volume index during the bookbuilding week minus the median of search volume index in the previous eight weeks. PRE_IPO_RTN is the average initial return of five latest IPOs before the IPO. RANGE is the price range announced in its prospectus divided by its midpoint price. SIZE is the logarithm of total asset. UWREP is a dummy variable which equals one if one of sponsors is among the top ten based on the underwriting market share ranking. Robust t-statistics are in parentheses. *, ** and *** represent significance at the 10%, 5% and 1% level.

Dependent Variable: Offer price revision (REVISION) Variables Model 1 Model 2

SUBRATE 0.015***

(6.10) ASVI 0.028*** (2.88)

PRE_IPO_RTN 0.154*** 0.201*** (4.31) (3.96) RANGE -0.193** -0.215*

(-2.23) (-1.66) SIZE -0.170 -0.368

(-0.62) (-0.98)

UWREP 1.894 0.743 (1.52) (0.39) Constant 6.652 12.330

(1.09) (1.45) Observations 316 158

R-squared 0.268 0.206

29

Table 4 Partial Adjustment of Offer Price to Pre-Market Sentiment

This table tests Hypothesis 2. The dependant variable, OTO is the offer to open return, that is, the open price on the first trading day divided by the offer price minus one. Explanatory variables: SUBRATE is the number of shares subscribed by individual investors divided by the number of shares assigned to the Public tranche. ASVI is the abnormal Google search volume, defined as the search volume index during the bookbuilding week minus the median of search volume index in the previous eight weeks. PRE_IPO_RTN is the average initial return of five latest IPOs before the IPO. RANGE is the price range announced in its prospectus divided by its midpoint price. SIZE is the logarithm of total asset. UWREP is a dummy variable which equals one if one of sponsors is among the top ten based on the underwriting market share ranking. REVISION is the offer price divided by the midpoint of initial price range minus one. TOP is a dummy variable equals 1 when offer price is set at the upper bound of the price range. Robust t-statistics are in parentheses. *, ** and *** represent significance at the 10%, 5% and 1% level.

Dependent Variable : Offer-to-Open Return (OTO) Variables Model 1 Model 2 Model 3 Model 4 SUBRATE 0.041*** 0.039*** (6.14) (5.78) ASVI 0.035** 0.032** (2.59) (2.40) PRE_IPO_RTN 0.240*** 0.211** 0.516*** 0.406*** (2.79) (2.45) (4.08) (3.30) RANGE 0.138 0.147 -0.176 -0.148 (1.11) (1.20) (-0.81) (-0.74) SIZE -0.348 -0.308 -1.023 -0.788 (-0.74) (-0.66) (-1.40) (-1.10) UWREP 1.850 1.391 -0.138 -2.227 (0.77) (0.57) (-0.03) (-0.52) REVISION 0.262** 0.174 0.529*** 0.192 (2.27) (1.41) (3.41) (1.06) TOP 3.725 11.680*** (1.41) (2.86) Constant 5.311 3.905 29.020* 21.890 (0.49) (0.36) (1.69) (1.33) Observations 316 316 158 158 R-squared 0.376 0.380 0.289 0.316

30

Table 5 Compensation for Institutional Investors

This table tests Hypothesis 3. In models 3 to 6, the full sample is divided into “Cold” and “Hot” IPO subsamples based on the sign of REVISION. The dependent variable, COMPEN, is the compensation for institutional investors, defined as the offer to open return times the number of shares allocated to institutional investors scaled by total asset. Explanatory variables: SUBRATE is the number of shares subscribed by individual investors divided by the number of shares assigned to the Public tranche. R_PLACING is the residuals from regressing PLACING on SUBRATE. RANGE is the price range announced in its prospectus divided by its midpoint price. VOLATILITY is the standard deviation of intraday prices on first trading day normalized by the offer price. UWREP is a dummy variable which equals one if one of sponsors is among the top ten based on the underwriting market share ranking. SIZE is the logarithm of total asset. PRE_IPO_RTN is the average initial return of five latest IPOs before the IPO. REVISION is the offer price divided by the midpoint of initial price range minus one. TOP is a dummy variable equals 1 when offer price is set at the upper bound of the price range. Robust t-statistics are in parentheses. *, ** and *** represent significance at the 10%, 5% and 1% level. Dependent Variable: Compensation for institutional investors (COMPEN) Full Sample Hot IPOs s Cold IPOs s

Variables Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 SUBRATE 0.005 0.005 -0.003 -0.001 0.007 0.009 (1.28) (1.04) (-1.07) (-0.36) (1.22) (1.03) R_PLACING 0.090** 0.074** 0.075* 0.119*** 0.018 0.025 (2.53) (2.14) (1.84) (2.75) (0.60) (0.83) RANGE 0.106* 0.219*** -0.052 (1.69) (2.85) (-0.99) VOLATILITY 0.764*** 0.837*** -0.163 (4.21) (4.06) (-0.61) UWREP -2.276** -3.455** -3.538** -4.295** -2.680** -2.594** (-2.22) (-2.55) (-2.32) (-2.59) (-2.17) (-2.09) SIZE 0.843*** 0.833** 0.933* 1.095** 0.432 0.389 (2.62) (2.15) (1.81) (2.02) (1.15) (1.00) PRE_IPO_RTN 0.049 0.005 0.001 -0.026 0.063* 0.068** (1.23) (0.12) (0.012) (-0.50) (1.96) (2.05) REVISION 0.175** 0.145** 0.149 0.160* -0.021 -0.014 (2.52) (2.10) (1.60) (1.73) (-0.37) (-0.25) TOP 2.202 1.877 3.127* 2.671 -1.211 -1.183 (1.52) (1.26) (1.81) (1.56) (-0.77) (-0.75) Constant 16.460** -4.957* -7.205* -4.484 -2.874 -3.670 (2.29) (-1.84) (-1.79) (-1.34) (-1.03) (-1.39) Observations 316 316 237 237 79 79 R-squared 0.161 0.181 0.174 0.196 0.234 0.229

31

Table 6 Pre-Market and Aftermarket Sentiment

This table tests Hypothesis 4. The dependant variables, TURNOVER and SMALLNET, are defined as follows: TURNOVER is the total trading volume divided by the number of shares outstanding on the first-trading day. SMALLNET is the buyer-initiated small trades minus seller-initiated small trades divided by the total dollar trading volume on the first trading day. Explanatory variables: SUBRATE is the number of shares subscribed by individual investors divided by the number of shares assigned to the Public tranche. ASVI is the abnormal Google search volume, defined as the search volume index during the bookbuilding week minus the median of search volume index in the previous eight weeks. RANGE is the price range announced in its prospectus divided by its midpoint price. SIZE is the logarithm of total asset. UWREP is a dummy variable which equals one if one of sponsors is among the top ten based on the underwriting market share ranking. PRE_IPO_RTN is the average initial return of five latest IPOs before the IPO. .REVISION is the offer price divided by the midpoint of initial price range minus one. TOP is a dummy variable equals 1 when offer price is set at the upper bound of the price range. OTO is the offer to open return, that is, the open price on the first trading day divided by the offer price minus one. Robust t-statistics are in parentheses. *, ** and *** represent significance at the 10%, 5% and 1% level.

Panel A Panel B Dependant Variables: TURNOVER TURNOVER SMALLNET SMALLNET Model 1 Model 2 Model 3 Model 4

SUBRATE 0.002** 0.002* (2.27) (1.79) ASVI 0.003* 0.002 (1.72) (1.60) RANGE 0.007* 0.006 -0.018 0.033 (1.87) (0.73) (-0.44) (0.61) SIZE -0.041** -0.056** -0.050 0.231 (-2.29) (-2.20) (-0.28) (0.90) UWREP 0.239*** 0.194** 0.304 0.085 (4.15) (2.31) (0.55) (0.10) PRE_IPO_RTN -0.003 -0.005* -0.029 -0.055** (-1.24) (-1.87) (-1.52) (-2.52) REVISION 0.004 0.005 0.071* 0.060 (1.32) (1.02) (1.82) (0.97) TOP 0.197*** 0.247** 0.782 0.785* (2.88) (2.35) (0.88) (1.77) OTO 0.010*** 0.012*** 0.008 0.011 (3.96) (4.31) (0.52) (0.65) Constant 0.834** 1.322** 1.316 -6.278 (2.50) (2.57) (0.26) (-0.83) Observations 316 158 316 158 R-squared 0.335 0.298 0.031 0.041

32

Table 7 Aftermarket Sentiment and Open-to-Close Return

This table tests Hypothesis 5. The dependant variable, ADJOTC, is the market-adjusted open-to-close return. Explanatory variables: TURNOVER is the total trading volume divided by the number of shares outstanding on the first-trading day. SMALLNET is the buyer-initiated small trades minus seller-initiated small trades divided by the total dollar trading volume on the first trading day. VOLATILITY is the standard deviation of intraday prices on first trading day normalized by the offer price. R_PLACING is the residual from regressing PLACING on SUBRATE. LARGENET is the buyer-initiated large trades minus seller-initiated large trades divided by the total dollar trading volume on the first trading day. SUBRATE is the number of shares subscribed by individual investors divided by the number of shares assigned to the Public tranche. ASVI is the abnormal Google search volume, defined as the search volume index during the bookbuilding week minus the median of search volume index in the previous eight weeks. PRE_IPO_RTN is the average initial return of five latest IPOs before the IPO. SIZE is the logarithm of total asset. RANGE is the price range announced in its prospectus divided by its midpoint price. UWREP is a dummy variable which equals one if one of sponsors is among the top ten based on the underwriting market share ranking. REVISION is the offer price divided by the midpoint of initial price range minus one. TOP is a dummy variable equals 1 when offer price is set at the upper bound of the price range. OTO is the offer to open return, that is, the open price on the first trading day divided by the offer price minus one. Robust t-statistics are in parentheses. *, ** and *** represent significance at the 10%, 5% and 1% level.

Dependent Variable: Market-adjusted open-to-close returns (ADJOTC) Variables Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 TURNOVER 3.788*** 2.989*** 2.940*** (3.15) (2.86) (2.93) SMALLNET 0.618*** 0.817*** 0.759*** (5.75) (7.40) (7.11) VOLATILITY 1.482*** 1.257*** 1.350*** 1.225*** (6.06) (5.21) (6.11) (5.73) R_PLACING 0.032 (0.90) LARGENET 0.206*** (4.90) SUBRATE -0.002 -0.000 0.003 -0.000 -0.001 (-0.56) (-0.08) (0.901) (-0.00) (-0.38) ASVI -0.005 (-0.88) PRE_IPO_RTN -0.065 -0.086 -0.074 -0.071 -0.057 -0.043 (-1.44) (-1.55) (-1.58) (-1.57) (-1.38) (-1.07) SIZE -0.458 -0.348 0.195 0.185 0.130 0.288 (-1.22) (-0.74) (0.65) (0.50) (0.37) (0.85) RANGE -0.040 -0.009 -0.780 -0.036 -0.101 -0.093 (-0.49) (-0.10) (-1.07) (-0.52) (-1.56) (-1.50) UWREP 2.668 0.935 0.340 1.046 0.049 -0.207 (1.25) (0.47) (0.23) (0.75) (0.03) (-0.17) REVISION -0.038 0.090 -0.039 -0.094 -0.090 -0.090 (-0.39) (1.07) (-0.47) (-1.19) (-1.25) (-1.30) TOP 0.990 0.650 0.060 0.213 0.798 0.518 (0.53) (0.28) (-0.03) (0.14) (0.53) (0.36) OTO 0.040 0.039 -0.097* -0.071* -0.112*** -0.112*** (0.73) (0.69) (-1.89) (-1.84) (-3.13) (-3.26) Constant 11.186 8.716 -17.309** -7.450 -13.969* -15.649** (1.31) (0.78) (-2.39) (-1.11) (-1.95) (-2.09) Observations 316 158 316 316 316 316 R-squared 0.021 0.038 0.165 0.238 0.361 0.410

33

Table 8 Investor Attention and Pre-Market Retail Demand

This table examines the relation between oversubscription rate and investor attention during the pre-market period. The dependant variable, SUBRATE is the number of shares subscribed by individual investors divided by the number of shares assigned to the Public tranche. Explanatory variables: ASVI is the abnormal Google search volume, defined as the search volume index during the bookbuilding week minus the median of search volume index in the previous eight weeks. PRE_IPO_RTN is the average initial return of five latest IPOs before the IPO. RANGE is the price range announced in its prospectus divided by its midpoint price. SIZE is the logarithm of total asset. UWREP is a dummy variable which equals one if one of sponsors is among the top ten based on the underwriting market share ranking. Robust t-statistics are in parentheses. *, ** and *** represent significance at the 10%, 5% and 1% level.

Dependant Variables: Oversubscription Rate (SUBRATE) Variables ASVI 0.825*** (4.08) PRE_IPO_RTN 6.147*** (3.59) RANGE -3.971** (-2.57) SIZE -9.950 (-1.23) UWREP 1.179 (0.023) Constant 347.800* (1.90) Observations 158 R-squared 0.283

34

Table 9 Long-run Returns

This table shows the IPO long-run returns. Raw Return is calculated as the percentage difference between the close prices of 6 months, 12 months, and 18 months after the listing, and the close price of the listing day. Benchmark Return is the contemporaneous market return. The Adjusted Return is the difference between the corresponding Raw Return and the Benchmark Return. Variables 6 Months 12 Months 18 Months Raw Return (%) Mean 2.14 6.87 9.83

Median -2.71 -5.11 -11.29

Benchmark Return (%) Mean 4.40 8.25 10.08

Median 5.07 11.34 13.61

Adjusted Return (%) Mean -2.13 -1.10 0.35

Median -5.58 -11.52 -16.90

Observations 316 316 316

35

Table 10 Investor Sentiment and Long-run Underperformance

This table examines the determinants of long-run IPO underperformance. The dependant variable, 1YADJRTN, is the 1-year buy-and-hold market-adjusted return from the closing price on the first trading day. Explanatory variables: TURNOVER is the total trading volume divided by the number of shares outstanding on the first-trading day. SMALLNET is the buyer-initiated small trades minus seller-initiated small trades divided by the total dollar trading volume on the first trading day. SUBRATE is the number of shares subscribed by individual investors divided by the number of shares assigned to the Public tranche. ASVI is the abnormal Google search volume, defined as the search volume index during the bookbuilding week minus the median of search volume index in the previous eight weeks. PRE_IPO_RTN is the average initial return of five latest IPOs before the IPO. RANGE is the price range announced in its prospectus divided by its midpoint price. UWREP is a dummy variable which equals one if one of sponsors is among the top ten based on the underwriting market share ranking. SIZE is the logarithm of total asset. VOLATILITY is the standard deviation of intraday prices on first trading day normalized by the offer price. Robust t-statistics are in parentheses. *, ** and *** represent significance at the 10%, 5% and 1% level. Dependant Variable: One-year market-adjusted buy-and-hold return (1YADJRTN) Variables Model 1 Model 2 Model 3 Model 4 TURNOVER -11.850** -11.950** (-2.20) (-2.23) SMALLNET -0.196** -0.174* (-2.01) (-1.79) SUBRATE -0.016 -0.026 (-0.75) (-1.30) ASVI -0.050 -0.061 (-1.33) (-1.55) PRE_IPO_RTN -0.543* -0.504 -0.548* -0.503 (-1.92) (-1.11) (-1.94) (-1.11) RANGE -0.415 -0.599 -0.499 -0.647 (-1.04) (-1.01) (-1.24) (-1.07) UWREP -12.34 -17.10 -13.33 -16.46 (-1.19) (-0.93) (-1.28) (-0.90) SIZE 7.182*** 8.554** 7.423*** 8.852** (2.80) (2.24) (2.90) (2.30) VOLATILITY -0.108 -1.275 -0.719 -2.501 (-0.06) (-0.73) (-0.43) (-1.40) Constant -8.733 -13.17 -3.485 -7.751 (-0.34) (-0.32) (-0.14) (-0.18) Observations 316 157 316 157 R-squared 0.057 0.080 0.054 0.057