신규공모주에대한수요예측조사, 공모가결정및초기수익률

SIRFE Working Paper Series 신규공모주에대한수요예측조사, 공모가결정및초기수익률 조성욱 ( 서울대학교 ) 31-October-2011 SIRFE Working Paper 11-A06 SNU Institute for Research in Finance and Economics Room 102, Bldg 83, 599 Gwanak-ro Gwanak-gu, Seoul, Korea 151-742 Tel: 82-2-880-5424, 5425 Fax: 82-2-888-4231 http://www.sirfe.com

Bookbuilding, Price Revision and Initial Returns of IPOs Sung Wook JOH and Yoo Hwan KIM August, 2011 ABSTRACT We study how public and private information revealed during bookbuilding is differentially reflected in price revisions, and how it affects IPO underpricing. Using a unique hand-collected data set on Korean IPOs, we find that the revision process appears to leave out underwriters private information, especially positive private information on IPO firms while more aggressively reflecting institutional investors demand for IPOs. This asymmetric partial adjustment in price revisions helps explain the initial returns of IPOs. Our results regarding the asymmetric nature of the price adjustment are robust when controlling for other IPO specific characteristics, market-related information and institutional regulations over time. JEL code: G24, G32, G14 Keywords: Bookbuilding, IPO underpricing, Offer price, Private information, Initial return, Joh and Kim are from Seoul National University. Financial support from the Institute for Research in Finance and Economics of Seoul National University is gratefully acknowledged. 2

When IPO stocks are publicly traded for the first time, they usually show a high icrease in stock prices compared to offer prices (Ibbotson, 1975; Ibbotson and Jaffe, 1975; Ritter, 1984). Since underwriters and issuing firms determine offer prices and allocate new issues according to the information obtained during the bookbuilding process (Benveniste and Spindt, 1989; Benveniste and Wilhelm, 1990), high initial returns imply that they do not adjust offer prices to fully reflect information that they obtained. Despite the theoretical importance of this issue, empirical studies have not fully examined the role that bookbuilding activities and offer price revisions play in IPO underpricing (Ritter and Welch, 2002). Until recently, Cornelli and Goldreich (2001, 2003) and Jenkinson and Jones (2004) are unique in showing how actual bookbuilding information is reflected in offer price revisions or how it is used in share allocation. However, their studies use a limited number of samples based on bookbuilding by a single underwriter. This study uses a unique hand-collected large dataset including subscription rate of institutional investors in bookbuilding processes and the band of filing price range for all IPOs in Korea. In general, the bookbuilding process in the Korean stock market is similar to that in the U.S. However, due to securities regulations, more information, including institutional investors demand revealed during bookbuilding processes, is 3

disclosed to the public. Using a large, hand-collected data set for all IPOs in Korea from April 2000 to December 2007, we empirically examine factors affecting offer price revisions compared to the filing prices. In addition, the study examines the effects of actual bookbuilding information on initial returns to see whether it is partially reflected in offer price revisions, just like other public information such as market run-ups of periods prior to IPO as shown in Hanley (1993) and Kutsuna, Smith and Smith (2009). Furthermore, we examine how private information revealed during bookbuilding is reflected in price revisions. Revision depends on bookbuilding information (i.e. institutional investors valuation), IPO event specific information, IPO firm specific financial information, market related information, and private information on IPO firms that underwriters have. Since we have access to the first four types of information, we are better able to disentangle private information than previous researchers who lack bookbuilding information. Our evidence supports the argument that underwriters reflect private and public information such as market conditions, bookbuilding, and IPO firm characteristics into price revisions differently. Based on regressions of actual price revision for each of 586 IPO firms, we decompose price revision into two components: price revision predicted 4

by publicly available information and unpredicted price revision likely driven by private information. Initial return increases when the unpredicted component of price revisions increases, especially when the component is positive, implying that positive private information is not fully reflected in the offer price. On the other hand, publicly disclosed bidding ratio increases offer price but not initial return, suggesting that its full reflection into offer prices. Unlike previous studies that use information from only a single underwriter, our study is the first study to use a large data set on bookbuilding information from all underwriters in the market. We find that the revision process appears to fully reflect institutional investors subscription rates while private information is reflected partially. The difference hinges on the fact that bookbuilding information is publicly disclosed in Korea. This suggests that some credible public information is more likely to be incorporated into offer price than private information. Our results are in line with Löffler, Panther, and Theissen (2005) who argue that pre-ipo stock prices in the German OTC market is fully incorporated into offer prices. On the other hand, private information, particularly positive private information is partially left out in price revision process. This asymmetric adjustment of private and public information in price revisions is affected by the size of the stock issue. These results suggest that 5

partial and asymmetric adjustment of private information stems from the difficulty of conveying information in small firms. The rest of the paper is organized as follows. Section 1 briefly discusses institutional background information on IPO process in Korea. Section 2 introduces previous studies and hypotheses. The data and methodology used are described in Section 3. The main results are reported in Section 4. Robustness test results are discussed in Section 5. The summary and conclusion are presented in Section 6. I. Institutional Background in Korean IPO process Korean IPO market is unique in being very transparent. Much information in the filing and bookbuilding process is publicly disclosed. In this section, we briefly describe the IPO process in general. Furthermore, we describe two other distinct features of the Korean market that has direct influence on price revision and initial returns of IPOs: one about the bookbuilding process, and the other about the stock market in general. A. The Offering and Bookbuilding Process Once a firm decides to go public, the firm must apply for its listing eligibility from the listing committee of either the KOSPI or KOSDAQ market which can be 6

corresponded to NYSE and NASDAQ, respectively. If the committee approves the eligibility of a firm, the firm and its underwriter submit a securities registration statement 1 that is similar to the preliminary prospectus of the U.S. Like the preliminary prospectus, the Korean statement contains information regarding terms of offerings including the expected offer price or offer price range. The statement includes information on how many shares are allocated to different groups of investors and schedule for the bookbuilding. After the statement is filed, the underwriter presents the IPO firm and gathers information on the IPO case from the market ( road show ). As they are required to ration newly offered issues proportionally to the subscription rate, Korean underwriters have limited discretionary power in selling and distributing IPO stocks. In contrast, underwriters in the US or UK are known to strategically allocate shares to truth telling, frequent and aggressive institutional investors (Cornelli and Goldreich, 2001), or punish infrequent or institutions with flipping investors by giving them smaller allocations. Following the schedule in the preliminary prospectus, underwriters conduct a demand estimating survey in which all prospective institutional investors participate, 1 To avoid unnecessary confusion, we refer to this statement as the preliminary prospectus. 7

and their bids are equivalent to subscription. Due to the proportional share allocation practice, institutional investors might consider not revealing their true evaluation of firm values. Investors with high valuation for IPO stocks might not get more shares than investors with lower valuation when their quantity demanded is the same. However, their incentive not to tell the magnitude of their demand would be limited because under-reporting will result in fewer shares under the proportional share allocation practice. Therefore, participating institutional investors have an incentive to reveal their demand for the IPO shares truthfully or even overstate their demand. Subscription rate of the institutional investors participating in the demand estimating survey is disclosed to public 2 and indicates whether the market demand of the newly introduced shares is strong or not. Such information is only available to the public in Korea and Hong Kong to our knowledge. [Figure 1 (IPO Process in Korea) about here] B. Institutional Investors in Bookbuilding Process in Korea From 1999 to 2005, IPO firms are required to separately allocate their shares to four groups of (potential) investors: employees, two types of institutional investors, and 2 It is worth noting that due to several institutional changes in the Korean stock market, underwriters disclose less information, and this available window is slowly closing. 8

individual investors. Investors in one group cannot compete for shares allocated to other investor groups. One of the two types of institutional investors is called General Institutional Investors which consists of institutional investors in the traditional sense including corporations, financial institutions and pension funds. The other is a handful of government-designated High Return Institutional Investors which were troubled from losses incurred from the bankruptcy of Daewoo, the second largest business group in Korea. 3 Implicitly allowed to pursue high returns in various ways to compensate their earlier losses, (the limited number of) participants in the latter group show bidding behavior that is often contradictory to that of traditional investors. Such separation was abolished in 2005 as criticism rises on this regulatory practice, and all institutional investors participate in equal bookbuilding process since. For the purpose of the study, we focus on the bidding ratio given from the General Institutional Investors. C. Daily Price Limits Daily Price Limit is an upper and lower bound to which the price of each stock can move on 3 The Economist, 1999, South Korea: The death of Daewoo, August 19. 9

a given day. Thus, any investors or member firms can only place orders or quotations that are within the daily upper or lower price limits. There have been a couple of changes in the regulation on this daily price limit on stocks traded in the stock markets. The price limit was 12% ~ 15% depending on time periods and sections of stock exchange, and currently is at 15%. Furthermore, there have been regulations on the first price day limit on IPO stocks. 4 These daily price limits weaken the assumption that the closing price of the first trading day fully marks the market evaluation. In fact, a significant portion of our IPO samples hits the ceiling for some consecutive days after trading has begun. Hence in addition to basic tests with first day return as the dependent variable, we conduct tests with subsequent day s returns to fully capture the notion of initial return, or underpricing. We also include regime variables to reflect changes in regulatory limits on both daily prices and IPO specific returns. II. Hypotheses on IPO offer price revision and initial returns The IPO literature shows that firms and underwriters determine IPO offer price reflecting public information as well as bookbuilding information (Benveniste and Spindt, 1989). For example, Hanley (1993), Lowry and Schwert (2004) and Kutsuna et al. (2009) find that initial returns of IPOs are positively correlated with market run-ups. 4 For more detailed information, see the Korea Exchange Commission, http://eng.krx.co.kr. 10

Ritter (1984), Bradley and Jordan (2002), Lowry and Schwert (2004), and Derrien (2005) show that initial returns can be explained by previous IPO returns that may be independent of the market returns. However, underwriters and issuing companies have various reasons not to fully reflect the information that is revealed during road-shows and bookbuilding process, contributing to high underpricing of IPOs (Ibbotson, 1975). Underwriters might not fully use information to compensate truth telling investors from whom they extract information (Benveniste and Spindt, 1989; Benveniste and Wilhelm, 1990; Benveniste and Wilhelm, 1997), to reduce a burden of market making (Benveniste, Busaba, and Wilhelm, 1996; Chowdhry and Nanda, 1996), to reduce the risk of law suits (Lowry and Shu, 2002), to protect their reputation (Beatty and Ritter, 1986; Carter and Manaster, 1990), to increase their payoffs as the prospect theory suggests (Loughran and Ritter, 2004), or to allocate shares to corporate executives (Loughran and Ritter, 2004), etc. Without using actual bookbuilding information, Hanley (1993) argues that offer price is revised, but only partially to the information from bookbuilding process. Hanley (1993) and Kutsuna et al. (2009) argue that public information is not fully incorporated into the offer price, leading to significant amounts of first-day returns. 11

In this paper, we argue that the offer price of an IPO firm reflects bookbuilding information along with other information such as market run-ups. We also test whether types of information affect the degree of information reflection in the offer price revision. Information can be classified into different types depending on whether it contains negative or positive content, and whether it is private or public information to prospective investors. As an attempt to reduce the downside risk, underwriters and issuers tend to incorporate negative information more aggressively than they do positive information into the offer price (Bradley and Jordan, 2002; Edelen and Kadlec, 2005; and Hanley, 1993). Underwriters and firms are less likely to reflect all private information into the offer price, especially positive private information. Private information is difficult to convey to prospective investors as it is not subject to a disclosure rule nor explained by publicly disclosed information. On the other hand, credible public information is more likely to be incorporated into offer price than private information. For example, Löffler, Panther, and Theissen (2005) show that pre-ipo stock prices in the German OTC market are fully incorporated into offer prices. The degree of information reflection can be partial or full. It can be measured through its effect on initial returns. If offer price fully reflects certain information, such 12

information would not raise or lower the stock price and no significant relationship between initial returns and information should be observed. On the other hand, if offer price is partially adjusted, initial return should be related positively with such information as stock price should reflect the information left out. In this case, positive (negative) information should be reflected and price would increase (decrease) once trading begins. Although it might rarely occur, if offer price is excessively adjusted to reflect information, initial return should be related negatively with price revision or information used in price revision. III. Data and Methodology We have hand-collected information on 786 Korean IPOs from April 2000 to December 2007 from the Data Analysis, Retrieval and Transfer System (DART) 5 operated by the Korean Financial Supervisory Service (FSS). All listed firms in the Korean stock market submit their disclosures to the DART. 6 The DART reports preliminary and revised prospects that include estimated firm value for 390 IPOs and low and high of filing price range for 674 IPOs. IPOs that report only estimated firm 5 DART is an internet based electronic disclosure system that allows companies to submit disclosures online, where it becomes immediately available to investors and other users. 6 Though convenient, DART stores and displays but yet does not compile disclosures in a systematic way. Thus despite the risk of input errors, direct hand-collection was inevitable. 13

values without price ranges are concentrated in earlier periods. In addition, this study uses filing price ranges for the following reason. Previous literature on offer price revisions uses price bands for their analyses (Hanley, 1993; Kutsuna et al., 2009). Kutsuna et al. (2009) point out that the offer price range itself signals the underwriter s commitment. Along with the band midpoint, we include the band width and percent band width which normalizes the band width in respect to the band midpoint. A renewed prospectus includes information on bidding rates and/or a weighted average price of the bids made by the institutional investors including the High Return Institutional Investors mentioned in Section 2. We utilize only the bidding ratio 7 of General Institutional Investors as its disclosure has been subject to consistent regulation and it is not subject to collusive behavior of a few participants. Subscription rate or bidding ratio is the number of shares subscribed by investors over total number of shares allocated to them. Table I shows the time trends of all 586 IPOs that have all information used in the study such as offer price range and subscription ratios from April 2000 to the end of 2007. Issue size (bil) is the total amount of proceeds collected in the IPO process in 7 We use the term subscription rates, and bidding ratios throughout the article interchangeably. 14

billion won; Bidding Ratio is the subscription rate by institutional investors. We can observe many IPO cases during the IT booming in the early 2000s and a sharp decline in the following period. Furthermore, the mean and median of issue sizes steadily increase while the bidding ratios slowly decrease from year 2000 to year 2007. This observation is also due to the IT booming when many small IT-based venture firms went public. [Table I (Trends of IPOs) about here] Underwriters and IPO firms revise their offer prices after bookbuilding process is over. Offer Revision 8 measures the percentage difference of the finalized offer price from the midpoint of the filing offer range in the early prospectus following Hanley (1993) and Cornelli and Goldreich (2001). Offer Revision = (Offer price midpoint of offer range) / midpoint of offer range (1) We examine factors affecting offer revision and whether such factors are fully reflected in the offer price revision or not. We test whether IPO firm specific characteristic, 8 We use the term offer revision and price revision throughout the article interchangeably. 15

bookbuilding information, market information and past initial returns of IPOs are reflected in offer price revisions. We include information on underwriters (Carter and Manaster, 1990; Carter, Dark and Singh, 1998) and venture capitals (Megginson and Weiss, 1990; Lee and Wahal, 2004) and band width, percent band width which signals the lack of the commitment of underwriters and issuing firms according to Kutsuna et al. (2009) Offer Revision =β 0 + β 1 *Size + β 2 *Ln(Age) + β 3 *Band Width + β 4 *Band Width (%)+ β 5 *Ln(Bidding ratio) + β 6 *UW Incentive + β 7 *UW mkt share + β 8 *Number of VC + β 9 *ROA + β 10 *DtoA + β 11 *Market Dummy + β 12 *Pre-issue risk + β 13 *Pre-issue runup + β 14 *Past IPO Returns+ ε t (2) In addition, we examine whether certain information is partially reflected in price revision by examining its effects on initial returns. We measure the initial return through the percentage difference of the stock price of the seventh trading day from the offer price. Initial Returns = (first trading day end price offer price) / offer price (3) As discussed earlier in Section 2, there are regulatory limits on price for IPO firms. In order to ensure the robustness of our tests, we also conduct additional empirical tests 16

using different time intervals two, three, seven, 14 and 21 trading day intervals. The latter two are approximately 18 and 29 day initial returns in calendar days, respectively. Furthermore, we also include regime variables in regressions to note different regulatory regimes over time and over market divisions. Table II shows the summary statistics for 586 IPOs that have all necessary information used in the study. Mean bidding ratio is 101, which means on average institutional investors subscription is 101 times larger than shares allocated to the institutional investors. Mean initial return is over 50%, briefly confirming severe underpricing. As the first day returns might not reflect the magnitudes of underpricing due to the regulations, it also reports second day accumulated returns. The mean of second day accumulated returns is slightly higher than that of the first day, but the standard deviation, maximum and minimum levels of the returns indicate a large magnitude of underpricing. Also, the maximum accumulated returns for the first two trading days is 164.5%, reaching the upper price limit any IPO firm can have on the second trading day. This indicates that even the second day closing price may be below the fair market valuation for a portion of the IPO firms. Size is the natural log of equity scaled by thousand won; Age is the natural log of the firm age in months; Band width is the absolute difference between the high and low endpoints of the price band of IPO; 17

Band width (%) is Band width scaled by the midpoint of the band; Bidding ratio is the subscription rate of institutional investors; UW incentive is a dummy variable that is 1 if the underwriter fee is contingent of proceeds and 0 if not; UW mkt share is the underwriter s market share of IPO proceeds for the past 6 months; Number of VCs is the number of venture capitals participating in the firm at IPO; ROA is the ratio of return on assets; DtoA is the ratio of debt to assets; Pre-issue period is the days between IPO registration and issuance; Pre-issue run-up is the average rate of returns of the IPO firm s matching group during the Pre-issue period; Pre-issue risk is the standard deviation of Pre-issue run-up of the IPO firm s matching group; Pre-trade period is the days between issuance and first trade date; Pre-trade run-up is the average rate of returns of the IPO firm s matching group during the Pre-trade period; Pre-trade risk is the standard deviation of Pre-trade run-up; Past IPO Returns is the average initial return of IPO stocks for the past 6 months. Panel B shows the simple summary statistics and the significance of mean differences of two subsets depending on whether the offer price was higher than the midpoint of the price band or not. Bidding ratio is significantly larger in the subgroup of IPOs with upward revision than with downward revision. Initial returns are also higher in IPOs with upward revision than those with downward revision. However, 18

IPO firm size, age, and filing price band width are not significantly different in these two groups. Panel C shows the difference of simple summary statistics depending on the level of bidding ratio. The data set is divided into two subsets based on whether institutional bidding ratio is higher than the mean/median of the sample. Offer price is revised upward in 56% of IPOs with high bidding ratios while only it is 34% of IPO firms with low bidding ratios. Revision is greater in firms with high bidding ratios than in those with lower bidding ratios. In addition, first day return is higher in the group with high bidding ratios than that in the group with low bidding ratios. Firms in the group with high bidding ratios had higher ROAs, lower Debt to Asset ratios, and waited a shorter time to share allocation after registration. [Table II (Summary Statistics) about here] The variables we have collected might be correlated with each other. For example, Pre-issue risk and Pre-trade risk are highly correlated with a Pearson and Spearman coefficients of 0.593 and 0.623, respectively. So in the following analyses, we use only Pre-issue risks. Table III presents the correlation matrix among variables. 19

[Table III (Correlation Matrix) about here] IV. Empirical Results A. Price revision Table IV reports outcomes of the basic regression results on offer price revisions shown in Equation (2) including control variables and publicly known market information. We incorporate Hanley (1993) by testing with bookbuilding information only in Model (1), and adding firm specific information, market condition information, and IPO specific information into the regression equation Models (2), (3) and (4), respectively. For market information, we include pure public information such as market run-up (Pre-Issue run-up), initial returns of IPOs in the past 6 months (Past IPO returns), and market risk (Pre-issue risk). Table IV shows that coefficient of bidding ratio is positive and significant. Effects of bidding ratio remain robust even after controlling for firm specific financial conditions, IPO event specific conditions, market conditions, and underwriters in models (2), (3) and (4), respectively. In all models, coefficients of bidding ratio are positive and significant. Firm size has a positive, though insignificant coefficient. Debt to asset ratio of IPO firms has a negative impact on price revision, suggesting that offer price of high debt-ridden IPO 20

firms was revised downward. Negative coefficients of Pre-issue risk show that offer price is revised downward when there is a greater risk in stock returns of matched firms. Coefficients for Pre-issue run-up are consistent with previous studies such as Edelen and Kadlec (2005). Positive coefficients for Past IPO returns show that high initial returns of previous IPOs raise offer price, which is in accordance with Ljungqvist and Wilhelm (2002). [Table IV (Price Revision) around here] B. Initial returns Table V summarizes the results of the basic regressions on initial returns which are measured as the percentage change from the offer price to close price on the first trading day. It shows results of tests that first include bidding ratio, magnitude of offer price revision and firm specific financial conditions, then adds IPO event specific conditions, and market conditions, respectively. In Table V, firm size is negatively correlated with initial returns, consistent with results in previous studies (Hanley, 1993; Ritter, 1987). Large stocks tend to suffer less from the information asymmetry issue and as a result, yield smaller initial returns. The 21

significant and positive coefficient of Offer Revision suggests that firms with more revision experience higher returns. This means that revision process partially reflects information, though the significance level is marginal. As shown in the next section however, when we use a longer interval to resolve problems associated with the price limit regulation, the revision process has a positive and significant coefficient. Considering that a significant portion of IPO firms hit a price ceiling on their first trading day, we may have to examine the outcomes with a longer interval as well. Coefficients of bidding ratio are not significant, suggesting that price revision fully reflects information conveyed through institutional investors demand. This outcome can stem from that bidding ratio is publicly disclosed and credible information in Korea because institutional investors have to commit themselves to their bidding. Therefore, IPO firms and underwriters cannot easily manipulate information on bidding ratio. Because of these properties of bidding ratio, investors can use bidding information in IPO firm valuation once trading begins. We conjecture that underwriters, knowing how investors will react to the bidding information, can reflect its bidding information into the offer price. Table V also shows that coefficients of market run-ups (run-ups during the timeperiods between filing and trading, including the bookbuilding period) are positive, 22

raising the initial returns. This positive relation between public information revealed during registration period and underpricing is in accordance with Loughran and Ritter (2002) and Edelen and Kadlec (2005). Pre-issue risk is also positively correlated with initial returns while it affects price revision negatively. These results show that market conditions affect the returns of IPO stocks even though they were reflected in price revision shown in Table IV. These results of pre-issue risk and return confirm previous studies of Hanley (1993) and Kutsuna et al. (2009) in the sense that market run-ups were left on the table. 9 Past IPO returns retain positive relationship with initial returns although not significant. IPO firms risk measured through percentage band width is significant when some of IPO event specific information is not included. [Table V (Initial Returns) around here] C. Effects of Private Information Since we do not have access to further private information underwriters have, we 9 Due to space limitations, we do not report the results when we divide market run-ups by specific periods such as 40 to 20 day run-ups prior to IPO as done in previous literature. However, we confirm past studies as well as our results aforementioned. 23

infer private information by decomposing price revision into two parts: a part depending on disclosed information and another part depending on private information. 10 We argue that revision depends on 1) bookbuilding information (i.e. institutional investors valuation), 2) IPO event specific information, 3) IPO firm specific financial information, 4) market related information, and 5) private information on IPO firms that underwriters have. Since we have access to the first four types of information, we can disentangle private information unlike previous researchers who lack bookbuilding information. We decompose actual price revision into two components: estimated price revision and unpredicted price revision for each IPO firm using regression results based on price revision in equation (2) employing all IPO data and using the all variables. 11 From the regression results, we define unpredicted part or the residual as private information used in price revision (Private_Info). Therefore, Private_Info is unpredicted price revision which equals the difference between the actual price revision and predicted price revision. By construction, Private_Info has zero mean and 10 We have recently become aware of another paper on IPO price revision by Ince (2008) which also devotes a significant proportion to the decomposition of price revision. However instead of focusing on the sensitivity analysis on each of the factors as in Ince (2008), we test for implications that these decomposed components have on the initial returns. 11 The regression outcomes are very similar when we use all the data except the firm itself. Inclusion of information of the firm itself does not affect the outcome; possibly due to large number of data in our analysis. The result is also very similar when we limit the variables uses to those only with significance level higher than 10%, 24

is orthogonal to other variables used in price revision equations. Table VI summarizes the regression results using private information driven from decomposition of offer price revisions. Panel A shows that Private_Info has positive and significant impact on Initial Returns for all models including the one controlling for both firm specific and industry/market information Model (4). The insignificant coefficient of bookbuilding information variable when controlled for firm specific and industry/market information, suggests that bookbuilding information is fully reflected in offer price To further examine the impact of positive information, we re-run the test after adding a variable: Positive_Private_Info which takes the value of Private_Info when positive and zero otherwise. There is little difference in the regression results whether we use average price revision or zero revision as bench mark in determining Positive_Private_Info. Positive_Private_Info has a positive and significant coefficient while Private_Info is no longer significant. This result implies that the positive effect of private information actually comes from positive private information. Panel B shows that private information used in price revisions affects Initial Returns differently from those driven by disclosed information. Notice that public information on bidding ratio does not affect initial return. These results suggest that IPO firms and underwriters raise their offer price fully to reflect positive signal acquired during the 25

bookbuilding period. On the other hand, coefficients of Positive_Private_Info are positive and significant while those of Private_Info are not significant. These results imply that positive private information, not captured into the bookbuilding information is partially incorporated into offer price while negative private information is readily fully incorporated. Such difference lies in the nature and source of private information. Examples of negative private information can include levels and intensity of earnings management prior to IPO while those of positive private information can include quality of management and corporate culture. Even when the issuing firm and underwriters know more of positive information than market participants, they are not able to credibly communicate with the investors. Therefore, underwriters do not fully reflect positive information into offer price, and the market might not fully learn such information in the short run. In summary, Positive_Private_Info consistently yields positive significant impacts on initial returns while bidding ratio does not have statistically significant effects on initial returns. However, other public information including IPO market run ups is not fully reflected into offer price revision. These results indicate that for credible public information that investors, underwriters and IPO firms have would be fully reflected 26

into offer price. The results also indicate that private and public information yield different adjustments. In particular, positive private information is the source of asymmetric partial price adjustment. [Table VI (Private Information) around here] V. Robustness Tests A. Intervals of initial return As briefly discussed earlier, the Korean stock market limits daily price changes to a specific multiple of the previous day end price. In addition, it imposes a limit on the first day price of IPO firms. To test the robustness of short term (partial) adjustment of information surrounding the IPO, we additionally use initial returns of two, three, seven, 14 and 21 trading day intervals. The latter two are approximately 18 and 29 day initial returns in calendar days, respectively. We do not look into a longer period in this study because circumstances surrounding the IPO market change dramatically. For example, market making activities and price supports by the underwriters are conventionally valid for one calendar month after the public offering. Also, the possibility of material corporate 27

events increases as the interval grows, forcing us to limit the short term interval to a certain time period. Panel A in Table VII confirms the presence of asymmetric partial adjustment when initial returns are re-calculated. In all intervals, the coefficients of positive price revision show strong statistical significance, whereas coefficients for all revisions do not. These results confirm asymmetric partial adjustment which cannot be seen with the first day returns (Table V). On the other hand, bidding ratio does not affect initial returns, suggesting that information conveyed through institutional investors is fully reflected in revision process. Panel B shows the result of the tests including Private_Info and Positive_Private_Info. While the coefficient of Private_Info is not significant, a positive and significant coefficient on Positive_Private_Info shows partial and asymmetric adjustment. These positive effects remain robust even when relatively longer intervals are employed to define initial returns. Again, bidding ratio does not affect initial returns, suggesting that information in bidding ratio has been fully reflected in revision process. There can be two possible explanations for why price revision yields a long term effect. One is that the limits in price changes have slowed the IPO stocks from 28

reaching their market price. The other one is that the market is under-reacting to positive information that was left on the table, incorporating it through a certain period of time. Additional measures of initial returns confirm the earlier results that positive initial returns are related to private information. As discussed earlier, due to the nature of private information, it can be difficult to convey private information to investors in the market. On the other hand, bidding ratio is publicly disclosed and its information is credible in Korea. This characteristic makes bidding ratio fall under the category of public information, explaining why it is fully reflected in the offer price. [Table VII (Intervals of Initial Return) about here] B. IPO size and Information Our empirical results in the previous section show how market condition and information from bookbuilding affect its final offer price and consequently initial returns. Now, we investigate whether partial and asymmetric effects of different types of information depend on the size of the IPO firm. When other things are equal, small IPO firms are less likely to receive public interest, and underwriters of these IPOs are 29

more likely to face difficulties conveying private information to the prospective investors. We measure the IPO size based on two criteria: 1) firms equity size before IPO and 2) the number of shares issued. We divide the sample into two equal-sized groups according to IPO firms equity size and redo tests for price revision and initial returns in Tables IV, V, VI, and VII to test how different types of information affect initial returns in each sub group. Due to space limitations, we only report the results on the effects of private information and Positive private information on initial returns in Table VIII. Panel A is the results of the tests from the small IPO firms and Panel B is the results from the large IPO firms. It shows that the coefficient of Positive_Private_Info is consistently positive and significant in small IPOs while it is not in large IPOs. The coefficient of Private_Info is negative and significant in many time-intervals in small IPOs while it is either positive or insignificant in large IPOs. Omitted due to spatial reasons, effects of market conditions are both consistently positive in both groups, also confirming results from Section 5. Bidding ratio is significant only in the small group and for the first day only. These results suggest that partial reflection of positive information in Table VI occurs mostly in small IPOs. They also suggest that earlier results of asymmetric and 30

partial adjustment of information hinges on the difficulty of conveying information. In addition, results of other replications of Table IV, V, and VI confirm that the aforementioned results are mostly robust when the data are divided according to equity size or number of shares issued. [Table VIII (Subsamples: IPO size) around here] VI. Summary and Conclusion Frequently observed high initial returns indicate that offer prices are often set far below what investors are willing to pay, even after the bookbuilding processes. This study empirically tests what affects final offer price and initial returns of IPO stocks, using a very unique hand-collected large data set on IPO activities, including registration filing information, bookbuilidng information and offer price revision for all IPOs in Korea from April, 2000 to 2007.We examine factors affecting offer price revisions. In particular, we examine how information obtained during the bookbuilding process is reflected in offer price revisions, which in turn determine the magnitude of initial returns. While offer price revision reflects information from diverse sources such as market 31

conditions, bookbuilding activities, and IPO firm specific characteristics, underwriters and issuing firms appear to leave some information out. We find evidence that underwriters do not fully reflect private information into offer price revisions. While publicly disclosed bidding ratio appears to be fully reflected, positive private information is only partially reflected. Partial adjustment of information, especially positive private information contributes to high returns known as underpricing of IPOs, even when other IPO related information, public information and institutional market demands are controlled for. Asymmetric adjustment of private and public information in price revisions seems to stem from the difficulty of conveying private information. Our paper sheds light on the importance of private information and bookbuilding activities in determining offer price and eventually initial returns in addition to market conditions that previous literature has studied. Future study can further examine whether institutional investors demand and underwriters private information can explain the long term performance of IPO stocks. 32

References Beatty, Randolph P., and Jay R. Ritter, 1986. Investment Banking, Reputation, and the Underpricing of Initial Public Offerings. Journal of Financial Economics 15, 213-232. Benveniste, Lawrence M., Walid Y. Busaba, and William J. Wilhelm Jr., 1996. Price stabilization as a bonding mechanism in new equity issues. Journal of Financial Economics 42, 223-255. Benveniste, Laurence M., and Paul A. Spindt, 1989, How investment Bankers Determine the Offer Price and Allocation of New Issues, Journal of Financial Economics 24, 343-361. Benveniste, Laurence M., and William J. Wilhelm Jr., 1990, A comparative analysis of IPO proceeds under alternative regulatory environments, Journal of Financial Economics 28, 173-207. Benveniste, Laurence M., and William J. Wilhelm Jr., 1997, Initial Public Offerings: Going By The Book, Journal of Applied Corporate Finance 10, 98-108. Bradley, Daniel J., and Bradford D. Jordan, Partial Adjustment to Public Information and IPO Underpricing, Journal of Financial and Quantitative Analysis 37, 595-616. Carter, Richard B., Frederick H. Dark, and Ajai K. Singh, 1998, Underwriter Reputation, Initial Returns, and the Long-Run Performance of IPO Stocks, Journal of Finance 53, 285-311. Carter, Richard, and Steven Manaster, 1990, Initial Public Offerings and Underwriter Reputation, Journal of Finance 45, 1045-1067. Chowdhry, Bhagwan, and Nanda, Vikram, 1996. Stabilization, Syndication, and Pricing of IPOs. Journal of Financial and Quantitative Analysis 31, 25-42. Cornelli, Francesca, and David Goldreich, 2001, Bookbuilding and Strategic Allocation, Journal of Finance 56, 2337-2367. Cornelli, Francesca, and David Goldreich, 2003, Bookbuilding: How Informative is the Order Book?, Journal of Finance 58, 1415-1443. Derrien, François, 2005, IPO Pricing in "Hot" Market Conditions: Who Leaves Money on the Table?, Journal of Finance 60, 487-521. Edelen, Roger M., and Gregory B. Kadlec, 2005, Issuer surplus and the partial adjustment of IPO prices to public information, Journal of Financial Economics 77, 347-373. Fama, Eugene F., and James D. MacBeth, 1973, Risk, Return, and Equilibrium: Empirical Tests, Journal of Political Economy 81, 607-636. Hanley, Kathleen Weiss, 1993, The Underpricing of Initial Public Offerings and the Partial Adjustment Phenomenon, Journal of Financial Economics 34, 231-250. Ibbotson, Roger G., 1975, Price Performance of Common Stock New Issues, Journal of Financial 33

Economics 2, 235-272. Ibbotson, Roger G., and Jeffrey F. Jaffe, 1975, Hot Issue Markets, Journal of Finance 30, 1027-1042. Ince, Ö zgür, 2008. Why Do IPO Offer Prices Only Partially Adjust? Unpublished working paper, Virginia Tech. Jenkinson, Tim, and Howard Jones, 2004, Bids and Allocations in European IPO Bookbuilding, Journal of Finance 59, 2309-2338. Kutsuna, Kenji, Janet Kiholm Smith, and Richard L. Smith, 2009, Public Information, IPO price formation, and long run returns: Japanese evidence, Journal of Finance 64, 505-546. Lee, Peggy M., and Sunil Wahal, 2004, Grandstanding, certification and the underpricing of venture capital backed IPOs, Journal of Financial Economics 73, 375-407. Ljungqvist, Alexander P., William J. Wilhelm Jr., 2002, IPO allocations: discriminatory or discretionary?, Journal of Financial Economics 65, 167-201. Löffler, Gunter, Patrick F. Panther, and Erik Theissen, 2005, Who knows what when? The information content of pre-ipo market prices, Journal of Financial Intermediation 14, 466-484. Loughran, Tim, and Jay R. Ritter, 2002. Why Don t Issuers Get Upset About Leaving Money on the Table in IPOs?, Review of Financial Studies 15, 413-443. Lowry, Michelle, and G. William Schwert, 2004. Is the IPO pricing process efficient?. Journal of Financial Economics 71, 3-26. Lowry, Michelle, and Susan Shu, 2002. Litigation and IPO underpricing. Journal of Financial Economics 65, 309-335. Megginson, William L., and Kathleen A. Weiss, 1991, Venture Capitalist Certification in Initial Public Offerings, Journal of Finance 46, 879-903. The Economist, 1999, South Korea: The death of Daewoo, August 19. Ritter, Jay R., 1984, The Hot Issue Market of 1980, Journal of Business 57, 215-240. Ritter, Jay R., 1987, The Costs of Going Public. Journal of Financial Economics 19, 269-281. Ritter, Jay R., and Ivo Welch, 2002, A review of IPO Activity, Pricing, and Allocations, Journal of Finance 62, 1795-1828. 34

Figure 1 IPO process in Korea During the IPO process in Korea, information including filing range of the company going public is carried in the Preliminary Prospectus, and is evolved into an offer price during about 3 to 4 Revised Versions of the Prospectus. The average amount of time revision of Prospectus takes is 35.35 calendar days. Results of the Demand Estimation Survey, offer price, and other settled information are delivered in the Final Version of the Prospectus. After the final version of the prospectus is submitted, underwriters accept subscriptions from individual investors, who are essentially price takers from the underwriter s point of view, and allocate shares to both individual and institutional investors, accordingly. This process takes an average of 17.16 calendar days, and trading of the publicly offered share usually starts on the next day. Preliminary Prospectus Bookbuilding Activities (Road Show, Demand Estimation Survey) Final Revised Prospectus Subscriptions, Allocations Filing Range Offer Revision Offer Price First Trading day Close Price Mean= 35.35 Calendar Days Mean = 17.16 Calendar Days 35

Table I Trends of IPOs Table 1 presents the trends of IPOs in Korea from April 2000 to December 2007 that have all information used in the study. Issue size (bil) is the total amount of proceeds collected in the IPO process in billion won; Bidding Ratio is the number of shares subscribed by institutional investors over total number of shares allocated to them during the bookbuilding period; Offer Revision measures the percentage difference of the offer price from the midpoint of the offer range; Initial returns (1st day) is percentage initial return to the offer price at the market close of the first trading day; Pre-issue run-up is the average rate of returns of the IPO firm s matching group from registration to issuance Year N 2000 34 2001 154 2002 108 2003 77 2004 49 2005 71 2006 52 2007 41 2000-2007 586 Issue Size (bil) Bidding Ratio Offer Revision Initial Return (1st day) Pre-issue run-up Mean (Median) Std Dev Mean (Median) Std Dev Mean (Median) Std Dev Mean (Median) Std Dev Mean (Median) 6.584 52.945-0.313 0.627-0.097 6.007 18.386 0.11 0.47 (5.086) (47.560) (-0.302) (1.000) (-0.096) 9.034 117.387 0.026 0.59 0.004 10.407 61.939 0.149 0.414 (6.240) (109.202) (0.036) (0.632) (-0.010) 19.52 146.776-0.028 0.541-0.033 73.194 108.913 0.143 0.417 (6.239) (136.360) (-0.018) (0.464) (-0.042) 14.663 144.717-0.04 0.674-0.008 37.866 351.146 0.128 0.474 (6.225) (97.740) (-0.053) (0.667) (-0.018) 21.626 51.049-0.132 0.149-0.02 47.221 55.492 0.121 0.275 (6.800) (34.570) (-0.119) (0.109) (-0.017) 17.661 61.486 0.006 0.671 0.106 21.861 40.062 0.116 0.406 (13.007) (54.044) (0.024) (0.563) (0.112) 30.121 52.284-0.057 0.383-0.007 97.15 30.959 0.108 0.445 (8.118) (47.260) (-0.047) (0.278) (-0.006) 19.269 59.983-0.016 0.281-0.007 16.411 50.075 0.151 0.516 (12.000) (44.500) (0.024) (0.148) (-0.015) 16.25 100.542-0.038 0.527-0.002 48.164 146.565 0.156 0.452 (7.127) (79.558) (-0.027) (0.465) (-0.006) Std Dev 0.141 0.149 0.142 0.086 0.091 0.155 0.095 0.071 0.135 36

Table II Summary Statistics Panel A shows summary statistics of the main variables in the full dataset of 586 IPOs that have all necessary information used in the study; Panel B shows the summary statistics and the significance of mean differences of two subsets depending on whether the offer price was higher than the midpoint of the price band or not; and Panel C shows the difference of simple summary statistics depending on the level of bidding ratio. Ln(age) is the natural log of the firm age in months; Band Width is the absolute difference between the high and low endpoints of the price band of IPO; Band Width (%) is Band Width scaled by the midpoint of the band; Bidding Ratio is the subscription rate of institutional investors; UW incentive is a dummy variable that is 1 if the underwriter fee is contingent of proceeds and 0 if not; UW mkt share is the underwriter s market share of IPO proceeds for the past 6 months; Number of VCs is the number of VCs participating in the firm at IPO; Size is the natural log of equity scaled by thousand won; DtoA is the ratio of debt to assets; ROA is the ratio of return on assets; Pre-issue period is the days between IPO registration and issuance; Pre-issue run-up is the average rate of returns of the IPO firm s matching group during the Pre-issue period; Pre-issue risk is the standard deviation of Pre-issue run-up of the IPO firm s matching group; Pre-trade period is the days between issuance and first trade date; Pre-trade run-up is the average rate of returns of the IPO firm s matching group during the Pre-trade period; Pre-trade risk is the standard deviation of Pre-trade run-up; Past IPO returns is the average initial return of IPO stocks for the past 6 months. Panel A: All Samples, N = 586 Panel B: Direction of Price Revision Downward (N = 348) Upward (N = 238) Variable Min Max Mean Std Dev Mean (A) Std Dev (A) Mean (B) Std Dev (B) Diff (A-B) t-value Band Width 0.2 90 2.11 4.318 1.952 2.153 2.342 6.257-0.39-0.926 Band Width (%) 0.049 0.595 0.247 0.081 0.247 0.082 0.248 0.081-0.002-0.24 Bidding Ratio 2.507 3150.66 100.542 146.565 86.681 80.63 120.809 206.906-34.127-2.422 Offer Revision -0.638 0.491-0.038 0.156-0.136 0.115 0.105 0.077-0.241-30.427 Initial Return (1st day) -0.235 1.3 0.527 0.452 0.462 0.442 0.622 0.452-0.16-4.234 Initial Return (2nd day) -0.349 1.645 0.554 0.527 0.477 0.502 0.667 0.543-0.189-4.276 Issue Size (bil) 1.088 685.714 16.25 48.164 14.428 42.765 18.913 55.103-4.485-1.057 UW incentive 0 1 0.724 0.448 0.707 0.456 0.748 0.435-0.041-1.099 UW mkt share 0.001 0.677 0.078 0.096 0.07 0.086 0.088 0.108-0.018-2.133 Ln(age) 2.639 6.423 4.567 0.664 4.603 0.669 4.516 0.655 0.087 1.56 Number of VC 0 10 1.741 2.104 1.638 2.029 1.891 2.206-0.253-1.407 Size 14.224 22.129 16.293 0.986 16.327 1.023 16.242 0.928 0.085 1.048 DtoA 0.036 0.947 0.431 0.17 0.443 0.172 0.414 0.165 0.029 2.056 ROA -0.147 0.578 0.148 0.091 0.14 0.089 0.161 0.093-0.021-2.741 Pre-issue period 18 202 35.599 13.892 37.486 16.379 32.84 8.407 4.645 4.495 Pre-issue run-up -0.507 0.781-0.002 0.135-0.019 0.136 0.023 0.131-0.042-3.707 Pre-issue risk 0.016 0.08 0.043 0.01 0.044 0.011 0.041 0.008 0.003 3.505 Pre-trade period 4 44 17.087 6.799 16.509 6.538 17.933 7.093-1.424-2.463 Pre-trade run-up -0.466 0.407-0.006 0.095 0 0.096-0.015 0.094 0.016 1.973 Pre-trade risk 0.012 0.098 0.042 0.012 0.042 0.013 0.041 0.01 0.001 0.778 Past IPO returns (6 months) 0.049 1.121 0.532 0.197 0.471 0.207 0.621 0.141-0.15-10.474 37