Why Are IPOs Underpriced?

Evidence from Japans Hybrid Auction-Method Offerings

Frank Kerins*
College of Business and Economics Washington State University Vancouver, Washington 98686 360-546-9765 fax:360-546-9037 kerins@wsu.edu

Kenji Kutsuna
Graduate School of Business Administration Kobe University Rokkodai 2-1, Nada, Kobe, 657-8501, Japan kutsuna@kobe-u.ac.jp

Richard Smith
Peter F. Drucker and Masatoshi Ito Graduate School of Management Claremont Graduate University Claremont, California 91711 909-607-3310 richard.smith@cgu.edu Until October 1997, firms going public in Japan were required to use a hybrid auction process where at least half of the issue (the auction tranche) was offered via discriminatory auction. Remaining shares (the public offer tranche) were sold days later by firm commitment at a fixed price. We document underpricing and partial adjustment of IPO public offer prices in Japans auction regime, a regime where: investor information differences are not important; roadshows are not held; preferential allocations are negligible; and institutional investing is low. We also examine the relationship between underpricing and secondary selling in the IPO by existing shareholders and the relationship between underpricing and seasoned offerings in the year following the IPO. Although average underpricing and the high variation of underpricing are similar to findings for U.S. IPOs, we find no evidence that issuing firms or existing shareholders anticipate the extent of underpricing. The results raise important questions about interpretations of IPO underpricing in the U.S. We consider a broad range of competing, but non-mutuallyexclusive, hypotheses about underpricing and partial adjustment. Japans auction-method evidence is most consistent with a quasi-contractual allocation of risk related to initial mispricing where, in exchange for guaranteeing a minimum price to the issuer, the underwriter participates in upside performance. We cannot exclude the possibility that underpricing is affected by agency cost and prospect theory considerations similar to those suggested by Ritter and Welch (2002) in their review of the U.S. IPO market.

August 2004 JEL codes: G15, G24, G28 Keywords: IPO, public offering, book building, underpricing, partial adjustment * Corresponding author
We thank Jay Ritter, Janet Kiholm Smith, Ann Sherman, Pegaret Pichler, Lisa Meulbroek, Josh Rossett, and Masayoshi Takahashi. We also have benefited from comments from participants at the 2004 EVI conference, and seminars at the Claremont Colleges Economics Workshop and Washington State University.

Why Are IPOs Underpriced? Evidence from Japans Hybrid Auction-Method Offerings
Until October 1997, firms going public in Japan were required to use a hybrid auction process where at least half of the issue (the auction tranche) was offered via discriminatory auction. Remaining shares (the public offer tranche) were sold days later by firm commitment at a fixed price. We document underpricing and partial adjustment of IPO public offer prices in Japans auction regime, a regime where: investor information differences are not important; roadshows are not held; preferential allocations are negligible; and institutional investing is low. We also examine the relationship between underpricing and secondary selling in the IPO by existing shareholders and the relationship between underpricing and seasoned offerings in the year following the IPO. Although average underpricing and the high variation of underpricing are similar to findings for U.S. IPOs, we find no evidence that issuing firms or existing shareholders anticipate the extent of underpricing. The results raise important questions about interpretations of IPO underpricing in the U.S. We consider a broad range of competing, but non-mutually-exclusive, hypotheses about underpricing and partial adjustment. Japans auction-method evidence is most consistent with a quasicontractual allocation of risk related to initial mispricing where, in exchange for guaranteeing a minimum price to the issuer, the underwriter participates in upside performance. We cannot exclude the possibility that underpricing is affected by agency cost and prospect theory considerations similar to those suggested by Ritter and Welch (2002) in their review of the U.S. IPO market.

JEL codes: G15, G24, G28 Keywords: IPO, public offering, book building, underpricing, partial adjustment

Why Are IPOs Underpriced? Evidence from Japans Hybrid Auction-Method Offerings For several years, until October 1997, firms going public in Japan were required to use a hybrid auction process where at least half of the issue (the auction tranche) was offered via discriminatory auction. Remaining shares (the public offer tranche) were sold a few days later by firm commitment at a fixed price.1 Under JASDAQ regulations, the number of shares any investor could buy in either tranche was severely limited and investors who were likely to have private information about value were precluded from participating in the auction. Nonetheless, the public offer tranche routinely was underpriced. In addition to systematic underpricing, we find a pattern of partial adjustment that is similar to that of U.S. IPOs. That is, compared to the reservation price for the auction tranche (the minimum bid), adjustment of the offer price for the public offer tranche was a positive function of evidence of excess demand. Because underpricing also is positively related to excess demand and to the magnitude of the price adjustment, the adjustment is only partial.2 These results raise important questions about interpretations of IPO underpricing in the U.S. In contrast to Japans auction method, IPOs in the U.S. are priced by a process that includes book building. Several well-established hypotheses about underpricing are based on information asymmetry or information production. Hanley (1993), for example, interprets partial adjustment as evidence that underpricing compensates informed investors for revealing private information. However, in Japan, we find similar patterns of underpricing and partial adjustment even in a market where investor information differences are not important; roadshows are not held; preferential allocations to any single investor are small; and institutional investing is low.
1 2

Though the auction procedure still is available, no issuer has selected it since the end of our sample period. In a study of U.S. Treasury auctions, Goldreich (1998) shows that expected underpricing is necessary to induce bidding in discriminatory auctions. That result implies that the IPO aftermarket price will be above the weighted average successful bid price, but does not explain why the public offer tranche is discounted from the weighted average successful bid or why price adjustment and underpricing are positively correlated.

We consider a range of competing, non-mutually-exclusive, hypotheses for underpricing and partial adjustment. The auction-method evidence establishes that patterns similar to those observed for U.S. book-built offerings also are produced for the public offer tranche of a hybrid auction method where both information asymmetry and information production are restrained aggressively. Thus, while any number of information-related hypotheses is sufficient to generate underpricing and partial adjustment, none is necessary. The magnitudes and variation of underpricing in our sample are similar to those reported for U.S. IPOs. The average initial return for the public offer tranche is 11.50 percent of offer price. As in the U.S., a small but material fraction of the public offers have initial returns of 50 percent or more. Others have initial returns that are negative or zero. To determine whether issuers anticipate the extent of underpricing, we examine the relationship between realized underpricing and the occurrence of seasoned equity offerings (SEOs) in the year following the IPO. To determine whether insiders anticipate the magnitude of underpricing, we examine the relationship between underpricing and secondary selling in the IPO. We find no evidence that issuers or selling shareholders in Japan anticipate the extent of underpricing. The Japan evidence is most consistent with a quasi-contractual allocation of risk related to initial mispricing. In exchange for guaranteeing the minimum bid price to the issuer, the underwriter participates in upside performance through lower selling costs and the opportunity to make (small) allocations of underpriced public-offer-tranche shares to preferred customers. Because average underpricing is about three times as great as the underwriters fee and because some IPOs are severely underpriced, underpricing and risk allocation may be affected by agency cost considerations (Ritter and Welch, 2002, and Loughran and Ritter, 2004), prospect theory arguments (Loughran and Ritter, 2002), or regulatory constraints.

In Section 1 we describe Japans hybrid auction procedure. In Section 2 we describe the data and provide cross-sectional characteristics of the sample. In it, we examine offering

characteristics in the context of alternative theories of partial adjustment. Section 3 provides estimates of underpricing and partial adjustment equations. In Section 4 we examine seasoned offerings and long-run returns to test whether issuers anticipate underpricing and whether underpricing is used to signal quality. Also, to test whether selling shareholders anticipate the observed cross-sectional differences in offer price discounting, we examine their choices to sell in the IPO and to select between selling in the auction or the public offer. In Section 5 we discuss the implications of the evidence for a wide range of existing hypotheses of underpricing. 1. Japans Hybrid Auction Process Japan implemented its hybrid auction procedure in 1989, intending to improve distribution and pricing.3 During the period of our study, JASDAQ required that the minimum offer size be at least 250,000 shares plus 12.5 percent of outstanding shares.4 At least 50 percent were required to be offered in the auction. The auction failed if investors bid for less than 25 percent of total shares offered.5 JASDAQ IPOs during our sample period usually were combined offerings that included primary shares sold to raise capital for the issuing firm and secondary shares sold by selling shareholders. On average, secondary sales accounted for about 40 percent of shares offered. The proportion of secondary to primary shares determined the tranche in which the secondary
See Beierlein and Kato (2003), Hamao, Packer and Ritter (2000), Jenkinson and Ljungqvist (1996), Kutsuna and Smith (2003), Pettway and Kaneko (1996), and the report of Shoken Torihiki Shingikai (Securities and Exchange Council) (1989) for details on Japans IPO market before and after the auction regime and for review of the events that led to adoption of the auction method. 4 Financial firms with large numbers of shares outstanding were exempt from this requirement. 5 Overwhelmingly, issuers set the auction tranche at the 50 percent minimum. Those that auctioned more than 50 percent tended to do so because secondary shares comprised more than 50 percent of the offering and were required to be offered in a single tranche. While our database does not include failed offerings, the evidence suggests that failure was rare or non-existent. The auction tranches of all but three issues were fully subscribed or oversubscribed by auction bids. The lowest of the three that were not fully subscribed still was 80 percent subscribed, well above the minimum necessary to prevent failure of the auction. The requirements of the hybrid auction procedure are documented in the report of Shoken Torihiki Shingikai (Securities and Exchange Council) (1995).
3

shares were sold. If secondary shares represented more than 50 percent of total offered shares, then secondary shares were sold in the auction and selling shareholders received proceeds based on the weighted average price in the auction. If secondary shares represented less than 50 percent, then secondary shares were sold in the public offer and selling shareholders received proceeds based on the offer price. If secondary shares represented exactly 50 percent, then selling shareholders, as a group, could elect whether to sell in the auction or in the public offer. Under the auction-regime procedure, a preliminary prospectus was published about 14 days before the auction. The prospectus contained no information on pricing. About 10 days later, shares allocated to the auction were offered using a first-revised prospectus. This prospectus specified a reservation price below which bids would not be accepted. This minimum bid was determined by a required formula that the underwriter applied to the market prices and financial data of a small selection of public companies that the underwriter had identified as comparable.6 Auction-regime regulations occasionally changed but were consistent during the period of our study. During this period the formula required that the minimum bid be set at 85 percent of the market value estimated using the formula. Three or four days after the firstrevised prospectus was circulated, a one-day discriminatory auction occurred. At the close of the auction, shares were allocated to highest bidders first, until the entire allocation was distributed or the minimum bid was reached. About four days later, the underwritten offering of the remaining shares took place, using a second-revised prospectus and an offer price set by the underwriter. The weighted average successful bid was the maximum price the underwriter could set. The minimum price was the reservation price from the auction.7 The auction regulations were designed to achieve wide participation by small and uninformed investors. Specifically:
6 7

The formula is documented in the report of Shoken Torihiki Shingikai (Securities and Exchange Council) (1989). See the report of Shoken Torihiki Shingikai (1995).

Employees of the issuer, the ten largest shareholders, employees of securities companies, and companies that owned the issuers equity or debt were precluded from bidding.

Generally, a participant could bid to acquire no more than 5,000 shares (five round lots). At the average offer price, the maximum corresponds to about 10 million yen or $100,000.

An investor could acquire no more than 5,000 shares in the public offering (a combined maximum of 10,000) and could participate in no more than four public offerings per year. While they encouraged participation by small investors, the regulations also limited the

incentives of insiders and institutions to develop information on issues or devote resources to valuation, impeded free riding on price discovery, and discouraged institutional participation. As consequences, underwriters could not make large allocations of underpriced shares to any single investor and institutional investors generally did not participate in the IPO market.8 The underwriters role in developing information for pricing of both tranches necessarily was limited. For the auction, the first revised prospectus included the minimum bid and

information on the issuers track record.9 For the public offer, the second revised prospectus provided information on the auction, including number and size of bids, weighted average successful bid, weighted average bid, and minimum successful bid. The underwriter used

auction information to determine the offer price. However, because bidders were uninformed, the auction conveyed little to the market, except for information on aggregate demand.10 In addition, before 2002, underwriters could not use over-allotment options to adjust the size of the public offer and could not over-sell as a way to assure the success of an offering or to

Based on a study of 110 JASDAQ IPOs in 1996, Tamura (1997) reports that institutions purchased 11.7 percent of auction tranche shares and 13.7 percent of public offer tranche shares. Individuals purchased the remaining shares. 9 While the underwriter could engage in due diligence, an agreement among underwriters to restrict fees to an artificially low percentage also may have constrained the information-production efforts underwriters. During the sample period, fees were fixed informally at 3.1 percent of gross proceeds plus two yen, although slightly higher fees of 3.3 percent to 3.5 percent sometimes were charged. 10 Pettway and Kaneko (1996) establish that the formula-based minimum bid requires minimal information collection efforts by the underwriter and is not very informative about the underwriters assessment of value.

offset pricing errors. Although a growing literature indicates that over-allotment and preferential allocation of shares may play important roles in explaining some unresolved questions about IPO underpricing, these approaches to managing the risks of mispricing were specifically excluded or severely constrained in Japans hybrid auction regime.11 2. 2.1. Evidence of Underpricing and Partial Adjustment Data Sources From the start of 1995 until the introduction of book building in October 1997, 321 JASDAQ firms used the auction method to go public. For these, we obtain issue and financial data from the Research Group for Disclosure database (1995-1998). Issue data include offer date, shares issued, amount raised, offer price, first aftermarket price, and other offering details. Financial data include sales revenue, equity book value, shares outstanding, firm age, and number of employees for the year before the offering. Toyo Keizai Inc. provides daily stock price data. We use the daily JASDAQ Index as a measure of market performance. 2.2. Japans Auction-Method Analog to the Filing Range For a firm-commitment offering in the U.S., the preliminary prospectus specifies a filing range defined by a minimum and maximum anticipated offer price. By convention, the midpoint of the range is used as an estimate of the expected offer price. The final offer price is an adjustment to the filing range. Hanley (1993) segregates IPOs into those with offer prices below, within, and above their filing ranges, and finds partial adjustment, in that the most underpriced issues are those with the most positive price adjustments. Loughran and Ritter (2002) and Ritter and Welch (2002) find similar relationships between price adjustments and initial returns.12

Ritter and Welch (2002) review the literature related to IPO allocation issues. See also Aggarwal, Prabhala, and Puri (2002) for an empirical analysis of institutional allocation issues in the U.S. 12 In the US, underpricing and partial adjustment may result partly from regulations that prevent issuers from revising the offer price up by more than 20 percent without amending and re-issuing the preliminary prospectus. However, Bradley and Jordan (2002) document that amending just prior to the IPO is common. Japans auction regime requires only that the offer price not be more than the weighted average successful bid.

11

In this study, we use the minimum bid in a similar way to how these studies use the filing range. Under auction-regime regulations, the offer price could not be set below the minimum bid or above the weighted average successful bid. Thus, we measure Maximum Adjustment as the ratio of Weighted Average Successful Bid to Minimum Bid.13 We measure actual Price Adjustment as the ratio of Offer Price to Minimum Bid. Correspondingly, we use Price Discount (the ratio of Offer Price to Weighted Average Successful Bid) to measure partial adjustment. A lower Price Discount indicates a larger discretionary decrease in the Offer Price from the maximum possible offer price. We measure Initial Return based on the first freely trading price, as (Aftermarket Price/Offer Price 1). Because, as in the U.S., positive outliers skew the distribution of initial returns, we base some of our analysis on Underpricing, measured as (1 Offer Price/Aftermarket Price). In lieu of indications of interest, we use the Subscription Ratio (the number of bids per share offered in the auction tranche). Table 1 presents price adjustment and initial return statistics grouped by Subscription Ratio. The average Initial Return of the public offer tranche in our sample is 11.50 percent. Our results are similar to the partial adjustment results in Hanley (1993), Loughran and Ritter (2002), and Ritter and Welch (2002). That is, higher Price Adjustment is monotonically related to both higher Price Discount and greater Initial Return (or Underpricing). Initial Return is negative for the IPOs with the lowest subscription ratio. The negative return was borne by initial investors, or by the underwriter if the shares could not be sold at the offer price.14 In our sample, 7.5 percent of the IPOs had negative initial returns and 11.5 percent had initial returns of zero. For issues with non-positive initial returns on public-offer-tranche shares, the issuer received gross proceeds per share equal to the minimum bid. For these issues, the underwriter would have incurred costs to support the market or would have acted, in effect,
13 14

Italicized items denote variable names that are used in the tables and in the discussion. Unsold shares from the public offer were purchased by the underwriter and sold in the market after the IPO.

as the writer of a put option to the issuer at the minimum bid. Hence, non-positive initial returns are evidence of an offer price risk allocation between the issuer and the underwriter. 2.3. Bivariate Statistics Japans required approach for determining the minimum bid is similar to approaches U.S. underwriters use to establish the filing range. The formula price in Japan is an estimate of aftermarket value based on financial information from similar companies, less a required discount of 15 percent. Commonly, underwriters in the U.S. use comparable firm data to estimate aftermarket value. Normally, the underwriter establishes the filing range by reducing estimated aftermarket value by 10 to 15 percent.15 Because, during our sample period, the minimum bid reflected a discount of 15 percent from estimated market value, we use a 15 percent difference between the public offer price and the minimum bid as a convenient way to classify IPOs. IPOs where the minimum bid is more than 15 percent below the offer price are more likely to have been undervalued by application of the formula. Those where the minimum bid is less than 15 percent below are more likely to have been overvalued. We refer to these subsamples as the high-price-adjustment subsample" (181 observations) and the low-price-adjustment subsample" (140 observations). Offering Statistics: Table 2 presents descriptive statistics on offering, issuer, underwriter, and market conditions for the full sample and the subsamples. The significance of differences in offer prices and gross proceeds between subsamples is mainly an artifact due to classification by Price Adjustment.16 In other respects, offering characteristics for the subsamples do not differ significantly. Statistics on firm age and size suggest that underwriters set minimum bids of

15

Our description of U.S. practice is based on a combination of published information and our review of the confidential pricing analysis work papers of 18 major investment bankers on a recent IPO. 16 Minimum Bid (not reported) is marginally significantly higher for the high-price-adjustment subsample.

established firms closer to the final offer price. Also, underwriters with low shares of the IPO market appear to set minimum bids more aggressively. The JASDAQ run-up for the 40 trading days before the public offer is significantly greater for high-price-adjustment issues, whereas differences in five-day runup are not significant. Given that the minimum bid is established about five days before the public offer, and is based on historical data, price adjustments appear to result partly from market-wide value changes. Occurrences of seasoned offerings within one year after the IPO also are higher among the high-price-adjustment subsample. In Section 4, we use the seasoned offering data to test whether the IPOs of high-quality issuers are intentionally underpriced to signal quality.17 Price Adjustment: From Table 2, average Offer Price is 40.4 percent higher than Minimum Bid: 4.6 percent higher for the low-price-adjustment subsample and 68.1 percent higher for the high-price-adjustment subsample. Offer Price is equal to Minimum Bid for 38 of the 140 observations of the low-price-adjustment subsample. In the upper panel of Table 3, we examine bid statistics for the auction and setting of the offer price. Maximum Adjustment is higher than Price Adjustment for both subsamples or, equivalently, average Offer Price is discounted from the maximum price. Discretionary discounting, as reflected by Price Discount, is 92.9 percent of Weighted Average Successful Bid, and discounting is significantly less for the low-price-adjustment than for the high-price-adjustment subsample. Although Weighted Average Successful Bid is the upper bound on Price Adjustment, this constraint is not binding in our sample. For only one IPO was the public offer priced at Weighted Average Successful Bid. The fact that Offer Price is discounted more in the high-priceadjustment subsample is evidence that partial adjustment is discretionary, and not an artifact of offering regulations.

17

Welch (1989) offers this signaling rationale for IPO underpricing.

To better understand setting of the offer price, we also compare Offer Price to Minimum Successful Bid. On average, Offer Price is lower. This suggests that an investor usually could do better by waiting, and buying the shares in the public offering. Despite this disparity, four considerations encourage bidding in the auction. First, as in the U.S., the public offer tranche usually was over-subscribed and underwriters could allocate small numbers of shares to favored clients. Second, auction-regime regulations severely limited the number of shares any one investor could acquire in the public offering. An investor who wanted a larger position before aftermarket trading began would need to participate in the auction. Third, investors could purchase in no more than four public offers per year but could bid in any number of auctions. Fourth, returns from investing in the auction were positive, on average. To gauge uncertainty of demand for the public offer, the panel includes two measures of excess demand from the auction. Homogeneity (the ratio of Minimum Successful Bid to

Weighted Average Successful Bid) is an indicator of the dispersion of valuations in the market. Even if Minimum Bid is set very low, demand may be elastic at a higher value, in which case the range between Minimum Successful Bid and Weighted Average Successful Bid will be small. Homogeneity is significantly higher for the low-price-adjustment subsample.18 Subscription Ratio is significantly higher for the high-price-adjustment subsample. Comparisons of the means to corresponding medians for Maximum Adjustment suggest that Minimum Bid generally was set well below expected aftermarket value, much more than the 15 percent difference suggested by Japans regulated pricing formula. The differences between means and medians reflect strong right-skewness, indicating that application of the required pricing formula did not prevent extreme undervaluation of some IPOs. Conversely,

18

We considered several measures of homogeneity of bids, including the difference between maximum and minimum successful bids standardized by the Minimum Bid, and found similar significance levels (not reported).

10

Initial Returns: The lower panel of Table 3 presents statistics for initial returns and underpricing. For both subsamples, mean and median Initial Return (and Underpricing) are positive. As implied by partial adjustment, the mean is significantly higher for the high-priceadjustment subsample than for the low-price-adjustment subsample.19 The panel also provides evidence that partial adjustment was foreseeable. Aftermarket Price is almost twice Minimum Bid for the high-price-adjustment subsample, but only about 10 percent above Minimum Bid for the low-price-adjustment subsample. Aftermarket Price for the full sample is 3.3 percent higher than Weighted Average Successful Bid, and the median is 0.4 percent higher. Thus, on average, Weighted Average Successful Bid is close to being an unbiased predictor of Aftermarket Price. The ratio of Aftermarket Price to Weighted Average Successful Bid is less than one for 46.7 percent of the observations. However, the average ratio for the high-price-adjustment subsample is significantly higher than for the low-price-adjustment subsample and the percent of observations where the ratio is less than one is marginally significantly higher for the low-price-adjustment subsample. Other measures of Aftermarket Price relative to indicators of value from the auction yield similar differences. We also consider the possibility that the underwriter may discount the offer prices of highly demanded IPOs in order to compensate for over-optimism of auction bidders. To address this, we compare one-month market-adjusted returns after the IPO. Differences between the subsamples are not statistically significant and the mean adjusted return of the high-priceadjustment subsample is more positive. Thus, it does not appear that discretionary price

discounting is the result of an effort by the underwriter to compensate for over-optimism.20

19

We considered the possibility that price discounting is motivated by longer-run concerns about accurate pricing and that issues with high initial returns may have more-negative aftermarket returns. However, following the IPO, the one-month returns of the high-price-adjustment subsample were more positive than those of the low price adjustment subsample. Thus, our conclusions are not sensitive to our use of the first aftermarket price. 20 This possibility was suggested to us by Sherman, based on Sherman (2002) and Chowdhry and Sherman (1996).

11

To summarize, the public offer tranche of IPOs using Japans hybrid auction method exhibits partial price adjustment. The IPOs with highest excess demand in the auction tranche have offer prices that are adjusted upward the most relative to the minimum bid, offer prices that are discounted the most relative to weighted average successful bid, and, as a result, have the highest initial returns in the public offer. Aggregate Value of Initial Returns: In Table 4, for both the auction tranche and the public offer tranche, we compute the average and median yen values of initial returns, grouped according to subsample. Following Loughran and Ritter (2002), we refer to yen-valued initial returns as money left on the table. Money left on the table is not a measure of opportunity loss. It does not, for example, address higher costs the underwriter would have incurred in placing the issue at its aftermarket value, or the increase in explicit fees the underwriter would have required in exchange for foregoing the opportunity to recapture some of the initial-return gains that IPO investors realized directly. In Japan, Weighted Average Successful Bid, though market determined, generally is below Aftermarket Price. As Table 4 shows, underpricing of the auction tranche is small. Based on Weighted Average Successful Bid, the average auction tranche generates 35.2 million yen less in gross proceeds (about $350,000 or 1.63 percent of aftermarket value) than the aftermarket value of the tranche. Underpricing of the public offer tranche is more economically significant, averaging 146.1 million yen (about $1.5 million or 8.59 percent of aftermarket value). For the low-priceadjustment subsample, money left on the table averages 3.87 percent of yen-weighted-average aftermarket value and for the high-price-adjustment subsample, it averages 12.24 percent.21

21

It does not appear that money left on the table is materially affected by rounding of public offer prices. The sample of 321 IPOs includes 215 unique public offer prices and no price is represented in more than five IPOs. At prices under 1000 yen, rounding to five yen is common, whereas above 1000, rounding to 10 yen is common.

12

Except for the difference in size between the public offer tranche and the U.S. book-built IPOs studied by Loughran and Ritter, our results are strikingly similar. Their average offer size is approximately five times larger than the average public offer tranche and their average amount of money left on the table is approximately six times larger. They find that money left on the table averages 11.8 percent of aftermarket value, and ranges from 3.2 percent for offers priced below the filing range to 18.6 percent for offers priced above the range. Also, as they find in the U.S., median underpricing is fairly small but a few IPOs are extremely underpriced. That such different offering procedures yield such similar results suggests that explanations for observed underpricing patterns are likely to derive from the aspects of the procedures that are similar; most notably, the allocation of risk between issuers and underwriters. In Table 4, we also report mean and median minimum gross proceeds, based on Minimum Bid. These are the minimums that are guaranteed by the underwriter. On average, total gross proceeds exceeds the minimum by 46.95 percent (a median of 23.81 percent). Net proceeds are lower by the underwriters fee, which averages 3.39 percent of. In our sample, 16.2 percent of the offerings (52 IPOs) realized net proceeds below minimum gross proceeds. In cases where the first aftermarket value was below minimum proceeds, the underwriter directly or indirectly would have absorbed the loss of value. 3. Empirical Determinants of Offer Prices and Underpricing For book-built IPOs, Benveniste and Spindt (1989) hypothesize that price adjustments are based on information collected during the road show. In their model, residual underpricing compensates investors for revealing private information. The implication is that the price

adjustment (from filing range to offer price) and underpricing (from offer price to first aftermarket price) should be positively correlated. Loughran and Ritter (2002, 2004) and Ritter

13

and Welch (2002) point out that partial adjustment also can result from risk sharing, and may be affected by agency costs, prospect theory, or other considerations.22 While the evidence in Section 2 demonstrates a positive correlation between price adjustment and initial return in Japans auction regime, it does not exclude the possibility that factors unrelated to excess demand at the minimum bid may give rise to the correlation. In this section, we test whether partial adjustment can be explained by observable information about demand. For this, we use two-stage least squares to estimate a cross-sectional model of offer price adjustment and underpricing percentages. In Table 5, the model is estimated as a system because price adjustment and expected underpricing are simultaneously determined. Both price adjustment and underpricing depend on the selected offer price. By estimating the equations simultaneously we allow for the possibility that, in setting the offer price, the underwriter may take account of expected aftermarket price. We use actual underpricing as an estimate of the aftermarket price the underwriter expected. Independent variables include three exogenous measures of excess demand from the auction: Subscription Ratio, Maximum Adjustment, and Homogeneity. As controls for firm and issue characteristics, we include the natural log of firm age, the natural log of offer size, the ratio of primary shares to total shares offered, the ratio of auction shares to total shares offered, the underwriters share of the IPO market, and underwriter fees. We use changes in the JASDAQ Index over appropriate intervals to control for market-wide changes before the IPO. 3.1. Adjustment of the Offer Price Results for the price adjustment model (adjusted R2=99.4 percent) indicate that, by using demand information from the auction and our set of control variables, we come very close to

Although information production cannot account for a positive relationship between underpricing and market runup before the IPO, Edelen and Kadlec (2003), suggest that the relationship may be explained by rational issuers maximizing expected IPO surplus by accepting greater underpricing to increase the probability of a successful IPO.

22

14

estimating a variant of the formula underwriters use to determine Price Adjustment.23 Demand information from the auction that is available to the underwriter prior to setting the offer price also is available (in the second-revised prospectus) to investors in the public offer. If this information is material then the information should have an impact on the price adjustment. Maximum Adjustment is, by far, the most important factor explaining Price Adjustment. The coefficient on Maximum Adjustment, which also is significantly different from 1.0 (t = 34.68), indicates that the price adjustment is about 82.8 percent of the maximum.24 If Weighted Average Successful Bid is an estimate of aftermarket value, as implied by Table 3, then adjustment of less than one hundred percent is evidence of partial adjustment.25 Between the high- and low-price-adjustment subsamples, the difference in Maximum Adjustment accounts for a 64.58 percent difference in Price Adjustment. Also consistent with partial adjustment is that Price Adjustment decreases with increases in Subscription Ratio. The coefficient indicates that the underwriter discounts the price more when Maximum Adjustment is partly a result of a large amount of bidding. Economic

significance of Subscription Ratio is low, as the negative relative effect on the high-priceadjustment subsample is a 0.91 percent reduction is Price Adjustment. Also consistent with partial adjustment but of limited economic significance, when bids are more homogeneous, Price

Rounding of offer prices may partially explain the residuals in the price adjustment regression. In the OLS regression of Price Adjustment on only Maximum Adjustment, the coefficient on Maximum Adjustment is 82.1 percent, the intercept is 14.5 percent, and the Adjusted R2 is .992. Aussenegg, Pichler, and Stomper (2003), in a study of IPO pricing in the presence of a when-issued market, find a strikingly similar result: in a univariate regression to explain the percentage revision of the offer price relative to the filing range, the coefficient on the price adjustment from midpoint of filing range to ending when-issued price is 0.815 and is highly significant. The whenissued market is analogous in many respects to the auction tranche. Unlike our result, except when price adjustment is constrained by the filing range, they find no significant relationship between initial returns and price adjustment. They also find no informational role for book building in the presence of when-issued trading. The finding of average underpricing is consistent with underwriters leaving money on the table even in the absence of information acquisition through book building. 25 To test whether other results in the model are due to nonlinearity of the relationship between Price Adjustment and Maximum Adjustment, we estimated the system including the square of Maximum Adjustment. All other coefficients in the model retained significance. The squared term was not statistically significant (p=0.23) and had a nonintuitive positive sign. The adjusted R2 was not increased and the F-Value of the equation declined.
24

23

15

Adjustment is increased, so that Offer Price is closer to Weighted Average Successful Bid and to Aftermarket Price. The negative relative effect for the high-price-adjustment subsample is 0.50 percent. As we measure Homogeneity, more positive values indicate more homogeneity,

suggesting that demand is elastic at around the weighted average successful bid price. The price adjustment regression includes Underpricing as a simultaneously determined endogenous variable. This variable is included to test whether, after controlling for demand information from the auction, the expected aftermarket price has any incremental effect on the offer price. After controlling for demand information, Underpricing is not statistically

significant. The coefficient accounts for a negative 0.24 percent relative reduction if Price Adjustment for the high-price-adjustment subsample. Collectively, these variables account for almost the entire difference in Price Adjustment between subsamples a 62.93 percent difference compared to a total of 63.50 percent. The literature offers few predictions for the effects of offer and issuer characteristics on price adjustment. Although the bivariate analysis indicates that age differences are important, very few young firms issued during our sample period.26 The negative coefficient on Age indicates that underwriters set the minimum bids of established issuers more fully. To exclude the influence of the price adjustment on offer size, we measure Offer Size as Shares Offered times Minimum Bid. significant. Primary Shares Offered Percentage is the percent of IPO shares sold by the firm rather than existing shareholders. Although the coefficient is not statistically significant, the positive sign may imply that the underwriter is less concerned with risk of overpricing when offer With this specification, the relationship to Price Adjustment is not

Kutsuna and Smith (2003) find that small firms that lacked established track records generally were unable to go public during Japans auction regime.

26

16

proceeds go mainly to the issuing firm. Auction Tranche Percentage is the percent of IPO shares that are offered in the auction, but is at the 50 percent minimum for most of the sample. The Price Adjustment equation allows for Underwriter Fee and Market Share to be determinants of pricing. The coefficient on Market Share suggests that Price Adjustment does not depend significantly on the underwriters level of involvement in the IPO market. The negative coefficient on Underwriter Fee appears to be coincidental. During the study period, Underwriter Fees were fixed by agreement, but revisions of the formula resulted in slightly higher fees later in the sample period. Also over the latter part of the study period, declines in the JASDAQ Index appear to have reduced the magnitudes of price adjustments. The net effect is that the coefficient on Underwriter Fees is negative. To control more directly for unexpected market-wide value changes, we use run-up of the JASDAQ Index from 40 trading days before the public offering through the day before. The forty-day window is long enough to capture the effects of using the backward-looking formula pricing approach to establish the minimum bid. The negative but non-significant estimate in Table 5 indicates that JASDAQ Runup Days 40 through -1 has a negligible partial effect on Price Adjustment. That is, it does not appear that the underwriter adjusts the offer price based on changes in the JASDAQ Index over the two months before the offer. 3.2. Underpricing If the underwriter uses information from the auction to assess demand, then the information also can affect underpricing. Because the choices are simultaneous, we include Price Discount as an endogenous variable to explain Underpricing. This variable tests whether, controlling for information from the auction, underpricing is predictably related to the discount adjustment. As expected by partial adjustment, the effect of Price Discount on Underpricing is negative. All else equal, the more fully the offer is priced relative to Weighted Average 17

Successful Bid, the less the issue is underpriced relative to Aftermarket Price. The coefficient indicates that increasing the offer price by 1.0% of the weighted average successful bid decreases underpricing by 0.478% of aftermarket value. The relative partial effect on the high-priceadjustment subsample is a 2.53 percent increase in underpricing. With Price Discount included, Maximum Adjustment has no significant separate effect on Underpricing. The estimated relative increase in underpricing for the high-adjustment

subsample is 0.39 percent. Underpricing is higher the higher is excess demand as measured by Subscription Ratio, but Homogeneity of bids is associated with less underpricing, suggesting that discounts due to heterogeneity are not motivated by lower expected aftermarket prices. 27 The higher subscription ratio of the high-price adjustment subsample is associated with a 3.65 percent increase in underpricing. When demand is elastic, as indicated by more homogeneous bids, issues are priced more fully.28 The relative effect of lower homogeneity on the high-priceadjustment subsample is a 1.18 percent increase in underpricing. Collectively, these variables account for a 7.75 percent underpricing difference between the high- and low-price adjustment groups, compared to a total difference of 8.37 percent.
Because the Underpricing model includes three interrelated exogenous measures and one endogenous measure related to excess demand, interpreting their combined effect is complex. To assess the effects of correlations among these variable, we estimated four versions of the equation, including: all four variables, only Price Discount, only the exogenous measures, and only Subscription Ratio and Homogeneity. The coefficients on Price Discount, Subscription Ratio, and Homogeneity remained significant in all models and the coefficient on Maximum Adjustment remained non-significant except when Price Discount is dropped. Thus our conclusions are not sensitive to factor loadings on these variables. To assess whether the positive correlation between Price Adjustment and Underpricing is attributable to partial adjustment based on evidence of excess demand from the auction, we used the coefficients on the three exogenous excess-demand-related variables in the underpricing regression to compute a single measure of predicted price adjustment attributable to these factors. When Underpricing is regressed on this measure and the other exogenous variables, the coefficient is positive and significant (t = 5.60), and indicates a total partial adjustment effect that is approximately 1.7 times as great as the direct effect alone. Between the high- and lowprice-adjustment subsamples, we estimate that the direct price adjustment effect on underpricing is 2.53 percent greater underpricing for the high-price-adjustment subsample and that the total effect is a 4.30 percent difference. The evidence suggests that partial adjustment is a result of the underwriters response to evidence of excess demand. 28 This result is complementary to Kandel, Sarig and Wohl (1999). For the Israeli IPOs they study, elasticity information is revealed after the non-discriminatory auction and prior to trading, and higher elasticity is associated with a higher initial returns. Kandel et al. interpret the evidence of high elasticity as good news related to liquidity. For Japanese IPOs, auction elasticity measures are revealed after the auction and prior to setting the public offer price, and higher elasticity is associated with lower initial returns and fuller pricing of the offer, suggesting that issuers benefits from positive surprises about elasticity.
27

18

Underpricing is negatively related to Offer Size, consistent with the scale effect, that larger IPOs are priced more fully. IPOs by high-market-share underwriters are more

underpriced. Possibly, reputation signals quality, resulting in higher aftermarket values. This would enable high-market-share underwriters to underprice more without diminishing net proceeds. Loughran and Ritter (2002) and Kutsuna and Smith (2003) both note that underwriters may substitute underpricing for higher fees. Evidence in Table 5 is consistent with the ability of high-market-share underwriters to extract higher effective compensation through underpricing.29 Lowry and Schwert (2002), Loughran and Ritter (2002), and Kutsuna and Smith (2003) document that initial returns to IPOs are positively related to market returns from the period before the IPO. To control for unexpected market-wide value changes affecting the difference between Offer Price and Aftermarket Price, we use run-up of the JASDAQ Index from five trading days before the offering until the day before. A five-day window captures the effects of market-wide value changes that would have affected price and quantity bidding information since the auction. We find that Underpricing is positively related to JASDAQ Runup Days 5 through 1. The coefficient suggests that a one-percent increase in JASDAQ yields

approximately a one-percent increase in percent underpricing. Thus, it appears that underwriters do not revise offer prices in response to Index changes over the five days before the IPO. Rather, market-wide changes in value affect underpricing. 4. Underpricing as a Quality Signal Welch (1989) suggests that firms may underprice IPOs to signal quality. Accordingly, a high quality firm may undertake a small, underpriced IPO and use a seasoned equity offering to complete the capital raising. According to the argument, money left on the table at the IPO is

29

These results also may be consistent with the underwriter persistence phenomenon proposed in Hoberg (2003), where high reputation, informed underwriters persistently identify and profit from undervalued issuers.

19

expected to be more than recouped in the SEO. Low-quality firms will not attempt to imitate, because they run the risk of being discovered before the SEO can be completed.30 Here, we examine whether the pattern of underpricing and partial adjustment in Japan is related to this signaling hypothesis. The Welch model presumes that high-quality issuers

recognize they are undervalued by the market. Consequently SEOs after IPOs should be more likely when firms deliberately underprice and when IPOs are small. Because the market may not immediately perceive which firms are of high quality, immediate aftermarket value is not necessarily the best basis for assessing underpricing. Because our model of price adjustment in Table 5 explains virtually all of the variation of price adjustment, our partial adjustment evidence is inconsistent with the hypothesis that auctionregime issuers used underpricing to signal quality. As the adjustment to Minimum Bid is determined almost fully on the basis of observable information, unless Minimum Bid also is used to signal quality, high levels of residual underpricing cannot function as quality signals. To test the hypothesis more directly, we examine determinants of the decision to undertake a SEO and the long-run price performance of auction issuers. In addition, as a test of whether selling shareholders anticipate undervaluation, we examine the decision to sell shares in the auction tranche or in the public offer tranche. Overall, our evidence suggests that firms that undertake SEOs do so in response to market value increases after the IPO. We find little to suggest that issuers or selling shareholders anticipate the value increases. Rather, it appears that seasoned issues are reactions to unanticipated increases in market value.

Theoretical models of Allen and Faulhaber (1989), Chemmanur (1993) Grinblatt and Hwang (1989), among others, also use IPO underpricing as a signal of firm quality so that seasoned equity can be issued at a favorable price. The particular ways that investors learn of an issuer's quality vary across these models.

30

20

4.1. Bivariate Comparisons of SEO Issuers and Non-issuers The signaling models of Welch (1989) and others indicate that firms that are relatively hard to value are more likely to undertake small IPOs, and to follow them with SEOs once value is established in the market.31 Of the 321 IPOs in our sample, 27 were followed within one year by SEOs. Table 6 presents statistics for subsamples determined by whether the IPO issuer had a follow-on SEO within a year. The table also presents statistics for the subsample that offer the minimum number of shares required by JASDAQ. As all offerings are of round numbers of shares, we include all firms that offered no more than 50,000 shares above the minimum.32 Issuers with and without SEOs had similar proportions of Shares Offered to Outstanding, and were about as likely to issue the regulatory minimum. SEO issuers had higher Offer Prices, and larger Offer Proceeds (approaching significance). These comparisons suggest that SEO issuers were not limiting IPO size in anticipation of a SEO on more favorable terms. Primary Shares Offered Percentage is higher for the SEO subsample and approaches significance. However, given the more-than-offsetting difference in Offer Proceeds, secondary sellers generate more proceeds in their IPOs than they do when sellers do not follow with SEOs. This evidence suggests that informed issuers are not holding back, waiting for a SEO. SEO issuers in the subsample have significantly higher Maximum Adjustments, indicating that, with regard to Minimum Bid, SEO issuers were relatively undervalued by the underwriter. However, this does not result in significantly higher Initial Returns. Price Discounts are similar between the groups and Initial Returns are virtually identical. Finally, based on Nine-Month

Welch (1989) finds 288 of 1028 IPO issuers between 1977 and 1982 had reissued equity by 1987. The mean and standard deviation of SEO proceeds to IPO proceeds were 3.4 and 4.5. In our sample, the mean and standard deviation of SEO proceeds to IPO proceeds were 1.96 and 1.45. 32 Results for the subsample of minimum-size IPOs in Table 6 hold across a broad range of classifications for JASDAQ minimum number of shares offered. The results do not change materially for 1,000, 10,000, or 25,000, above the minimum or for offer sizes of 0.1, 1.0 or 10.0 percent above the minimum.

31

21

JASDAQ-Adjusted Returns, the SEO subsample has significantly more positive aftermarket performance - an adjusted return of 40.5%, compared to -1.7% for non issuers.33 Rather than supporting the hypothesis that high quality issuers underprice to signal quality, the results in Table 6 suggest that both the issuer and the underwriter underestimated the market value of the firm and that seasoned offerings occur following unexpected run-ups in the issuers market value after the offering. Findings are similar for the full sample and for the subsample of firms issuing the minimum number of shares in the IPO. 4.2. The Probability of Seasoned Equity Issue In general, the signaling argument holds that, for high-quality firms, combined net proceeds from a small, underpriced IPO and subsequent SEO will exceed net proceeds from a larger IPO. If underpricing is intentional and is motivated by the expectation of a subsequent SEO, then, as in our discussion of Table 6, the choice to undertake a subsequent SEO should be related to choices that issuers who were signaling rationally would make. Following Jagadeesh, Weinstein, and Welch (1993), we examine the likelihood of a subsequent SEO by estimating a LOGIT model. The dependent variable is equal to one if the firm has a SEO in the first year, and zero otherwise. Independent variables are those we hypothesize to be determinants of Price Adjustment and Underpricing. We also include the Three-Month JASDAQ-Adjusted Return after the IPO, excluding the Initial Return. Consistent with the bivariate results in Table 6, the LOGIT regression in Table 7 indicates that the choice is not significantly explained by variables that could imply that issuers anticipated a SEO: Price Discount, Offer Size, and Primary Shares Offered Percentage. In the case of Offer Size, though the coefficient is statistically significant, the sign is inconsistent with the view that undervalued
33

Nine-month adjusted returns would have been available to most of the firms in our sample that issued within one year. A longer interval is unlikely to have influenced the decision to issue an SEO during the first year. Also, for the first 36 months after the IPO (which is as long as we track returns), nine months is approximately the point when the differences of mean and median adjusted returns between the subsamples reach their maximums.

22

issuers undertake small IPOs, anticipating SEOs. Thus, their earlier choices suggest that SEO issuers were surprised by their long-run performance. Three-Month JASDAQ-Adjusted Return excluding Initial Return also is significant in predicting a SEO.34 In the context of the

significance of three-month returns, the lack of significance of Initial Return suggests that the market also was surprised by the positive long-run performance of the SEO issuers.35 4.3 Determinants of Long-run Market-adjusted Returns If underpricing is intentional and motivated by anticipation of a SEO, then, if the market recognizes the strategy, initial returns should reflect the implications of the signal and should be related to choices that signaling issuers rationally would make. Alternatively, if the market does not perceive the signal, then long-run returns should be related to the choices. The last two columns in Table 7 report the results of OLS regressions for two measures of three-month market-adjusted returns on Initial Return and variables hypothesized to be determinants of underpricing, including those that could reflect rational choices by firms engaging in a two-stage financing strategy. The return measures are: Three-Month JASDAQAdjusted Return after the IPO, including and excluding Initial Return. Results for long-run returns including Initial Return are similar to those for Underpricing in Table 5. The marginally significant result for Price Discount in the model that includes Initial Return is due to partial adjustment. The positive but non-significant sign in the model that excludes Initial Return suggests that, if anything, issues that are more fully priced have higher long-run returns. Thus, firms with more heavily discounted issues could anticipate higher initial returns, but, according to the LOGIT evidence, this does not increase the likelihood of a SEO. The results for Offer Size indicate that small issues lead to higher long-run returns. However, the
This result is similar to findings of Jagadeesh, Weinstein, and Welch (1993). They examine 1985 IPOs from 1980 through 1986 and find the abnormal aftermarket returns in the periods following IPOs are significantly positively related to the likelihood of subsequent SEOs and are as important as initial IPO returns in predicting SEOs. 35 We also estimated this model excluding aftermarket returns. Results in the model did not change materially.
34

23

LOGIT results indicate that firms with small issues are less likely to undertake SEOs. Primary Shares Offered Percentage is not significantly related to long-run return. Thus, there is no significant evidence that existing shareholders who anticipate high long-run returns refrain from selling in the IPO. Subscription Ratio is marginally significant in the model that excludes Initial Return, suggesting that the most oversubscribed issues have high initial return followed by negative aftermarket performance. Initial Return, however, has no significant explanatory power in predicting long-run returns, which is inconsistent with the market partially anticipating the returns. The results suggest that investors and the market were surprised by aftermarket

performance, just as they were by the SEO issuance choices.36 4.4. Source of Selling-Shareholder Proceeds As discussed earlier, a unique feature of Japans auction-regime IPOs is that the proportion of total shares that are offered by selling shareholders determines the tranche in which secondary shares are sold. If over (under) 50 percent of total shares are secondary, they must be sold in the auction (public offer). If secondary shares are exactly 50 percent of the offer, selling shareholders, as a group, can select the tranche. Proceeds realized by selling shareholders depend on the tranche in which they sell. Because the weighted average successful bid is the maximum public offer price, selling shareholders can always do better by selling in the auction. The source of the difference is that when the firm sells to raise capital, the price discount is averaged over existing and new shares, whereas when a shareholder sells, they realize the entire price discount. Thus, the benefit of selling in the auction is directly related to the expected price discount. If selling shareholders expect a large discount, they have greater incentive to sell in the auction. This could arise if existing shareholders believed Minimum Bid was set very low relative to anticipated bids.
36

We also tested for a relationship between aftermarket performance and the issuers decision to issue the minimum number of IPO shares. The coefficient was not significant.

24

In Table 8 we seek to understand the tranche choice of selling shareholders. In the first three columns the full sample is classified by whether selling shareholders sell in the auction or the public offer. In the full sample, sellers simultaneously are selecting the fraction of shares they sell and the tranche in which they sell. So the motivation for selecting the tranche is not clear. The last three columns are for the subsample of 90 IPOs where secondary shares are exactly 50 percent of the offering, enabling selling shareholders to choose the tranche.37 Significance levels of the first two items in the upper panel of the table are artifacts of our classification. From the center panel, for the full sample, there is no significant evidence that choice of tranche is related to variables that could indicate a perception that Minimum Bid is very low. From the subsample that could most easily select the tranche, selling in the auction is marginally significantly more likely when Price Adjustment is low, and when Subscription Ratio is low, which results are contrary to expectations based on the opportunity cost of selling in the public offer. However, there is no material difference in Price Discount between the groups, indicating that the choice of tranche was not related to realized benefit. JASDAQ Runup Days 40 through 1 also is significant. A likely causal factor is that market drift prior to the IPO is negative. Shareholders are more likely to sell in the auction tranche after market declines. In discussions with investment bankers in Japan, we have been unable to determine a cause for this pattern. Conceivably, it is related to disappointed

expectations about the market and share value or to concern that the decline may continue. From the last panel, there is no significant evidence that selling shareholders anticipated aftermarket performance and that this affected the choice of tranche. Thus, for the full sample and the subsample, it appears that selling shareholders did not select between the auction and the public offer tranche in a way that was related to underpricing of the public offer tranche
37

We can identify no reason for why selling shareholders who are selling half of the total offering sometimes sell in the public offer tranche, despite its expected lower proceeds.

25

compared to the auction tranche. Thus, the evidence provides no support for the view that selling shareholders who anticipate material underpricing sell in the auction tranche to mitigate the effect of underpricing on their proceeds and remaining holdings. 5. Discussion Evidence from Japans auction-method offerings raises important questions about interpretations of IPO underpricing in the U.S. In a regime where investors are symmetrically informed or information differences are not important, roadshows are not held, preferential allocations to any investor are negligible, and institutional investing is low, we find: The public offer tranche of Japans hybrid auction IPOs exhibits partial adjustment and underpricing that are similar to what is reported for book-built IPOs in the U.S. IPO issuers that undertake SEOs within one year do not appear to anticipate the SEO. Rather, SEOs occur in response to high aftermarket returns that were not anticipated. Based on their selling decision, existing shareholders do not appear to anticipate those issues that are more likely to be discounted relative to value in the auction. In this discussion, we consider a broader range of non-mutually-exclusive hypothesized reasons for underpricing and partial adjustment. Many of the extant hypotheses are predicated on institutional arrangements that were not present during the Japan auction regime. Hence, these hypotheses cannot account for underpricing and partial adjustment of IPOs in the regime. 5.1. Hypotheses that Cannot Account for the Japan Evidence Winners Curse: It does not appear that Japans auction-method evidence can be explained by the Rock (1986) winners curse hypothesis, that underpricing is necessary to induce uninformed investors to participate. Rock argues that informed investors seek large allocations of IPOs they recognize to be underpriced and opt out of overpriced offerings, so that without intentional underpricing uninformed investors would end up with a disproportionate 26

share of overpriced offerings and negative expected returns. The design of Japans auction method prevented winners curse by assuring the ability of uninformed investors to participate and by preventing informed investors from acquiring large positions in underpriced offerings. Information Revelation: It does not appear that the auction-method evidence can be explained by the Benveniste and Spindt (1989) hypothesis, and related literature, suggesting that shares are underpriced, and preferential allocations are given, to reward informed investors for revealing their private information.38 In a Japanese auction, an investors incentive is to bid a price that maximizes expected surplus, taking into account the expected aftermarket price, and the investors beliefs of how reducing the bid would affect the probability of being among the winning bidders. Thus, informed investors would be expected to bid below their reservation values and these bids could negatively affect the maximum price in the public offer. Further, except for giving small (5000 share) preferential allocations of public offer tranches up to four times per year to an investor, no mechanism enables the underwriter to reward high bidders. Information Production: It does not appear that our evidence can be explained by the Sherman (2002) hypothesis that, rather than being a rent to informed investors for truthful revelation, underpricing may provide compensation to investors for the cost of becoming informed and revealing the information to the underwriter. Under Japanese regulations, the maximum investments in both the auction tranche and public offer tranche appear to be too small to induce investors to devote significant resources to producing information about issuer value. Signaling: Welch (1989) hypothesizes that high-quality issuers use underpricing to signal value as precursors to seasoned offerings at higher valuations. Our examination of aftermarket performance and seasoned offerings after the IPO is inconsistent with the view that highly

See, also, Benveniste and Wilhelm (1990). Sherman (2002) notes that, while underwriters may use roadshows to market auction offerings, without the ability to make allocations dependent on information reported, they cannot provide incentives for investors to report accurately.

38

27

underpriced auction issuers in Japan outperform less underpriced issuers in the aftermarket, or that highly underpriced issuers are more likely to undertake seasoned offerings within the first year after the IPO. Rather, the evidence suggests that issuers undertake SEOs after increases in market value that are not anticipated by the issuers. Cascades and Ownership Dispersion: It does not appear that the Japan auction-method evidence of underpricing can be explained by hypotheses based on market dynamics, including avoidance of negative cascades. Welch (1992) hypothesizes that pricing too aggressively can cause potential sequential bidders to ignore their private positive information about value and can result in failed offerings. The Welch model is a sequential process that could represent a sequential bid auction or a book-building process as in the U.S. The auction process, however, aggregates information about demand and appears to prevent negative cascades.39 In principle, if Minimum Bid is set too high, the auction could fail, but our evidence suggests that failure is rare or non-existent. Only three IPOs received less than one bid per share, and the lowest subscription ratio was 0.83.40 The auction would not fail unless the ratio was below 0.50. Rather, the size of the public offer tranche would be increased. First aftermarket price was less than or equal to the minimum bid for 8.7 percent of the IPOs. The underwriter could lose money on these IPOs, since the offer price could not be set below the minimum. But in these cases, the underwriter would sell any unsold shares at market after the IPO closed. Ownership Dispersion: It does not appear that the evidence can be explained by hypotheses based on promotion of aftermarket liquidity. Booth and Chua (1996) hypothesize

Amihud, Hauser, and Kirsh (2003) study IPOs on the Tel Aviv Stock Exchange, most of which are sealed bid, uniform-price auctions, where the underwriter specifies a minimum and maximum bid. The authors note that the process generally works like a fixed price offer at the maximum. The price range is specified in the prospectus and is circulated five days before the offering. There are no road shows and the range is not subject to revision. Although the bids are all due at the same time, the authors find evidence of herding and note that investors may share information in advance of bidding. 40 Histograms of Subscription Ratio, Price Adjustment, and Initial Returns provide no indication of herding.

39

28

that underpricing can result in ownership dispersion and promote aftermarket liquidity. Japans auction process, however, produces dispersed ownership regardless of the extent of underpricing. Certification: Booth and Smith (1986), Tinic (1988), Chemmanur and Fulghieri (1994), and others offer various hypotheses of underwriter certification based on the premise that issuers and underwriters have private information about IPO value. While certification arguments are not precluded by Japans auction process, the fee fixing agreement, the limited role of the underwriter in pricing the offer, and the low level of litigation in Japan argue against an important role for underpricing as certification. Further, 7.5 percent of IPOs in our sample lose value immediately after the IPO. Unless certification applies to the average initial return, this result is hard to reconcile with certification in a low-litigation environment such as Japan. Implications for Interpretation of Underpricing in the U.S.: While the Japan evidence does not refute any of the above hypotheses, it does imply that alternative theories also may be sufficient to explain the empirical patterns of underpricing and partial adjustment in the U.S. Underwriters may underprice to attract uninformed investors, reward investor for revealing what they know, and/or reward investors for producing information. However, none of these

hypotheses can explain why Japans hybrid auction-method offerings exhibit similar underpricing and partial adjustment. Also, underpricing may signal value, prevent negative cascades, increase ownership dispersion, or be an aspect of certification. However, it does not appear that these hypotheses can explain the underpricing of Japans auction-method IPOs. 5.2. Hypotheses that are not Rejected by the Evidence From among the various hypotheses, we are left with the possibility that underpricing and partial adjustment in Japans auction regime reflect a quasi-contractual risk-sharing arrangement, where principal features of the arrangement may be affected by prospect theory, agency costs, or regulation and quasi-regulatory agreements. 29

Loughran and Ritter (2002) observe that partial adjustment may be consistent with prospect theory and agency cost. Issuers whose offer prices are increased relative to the filing range accept greater underpricing because they focus on the gain they realize, and underweight the opportunity loss. Alternatively, underwriters may benefit by using underpricing to reduce direct issue costs, to compensate investors who also occasionally buy shares in overpriced IPOs, and/or to reward favored customers or attract future business. While there is no clear way, in our analysis, to distinguish between these reasons for underpricing and partial adjustment, they both may influence pricing of the public offer tranche. Underpricing and partial adjustment, however, are not sufficient bases upon which to draw an inference of irrationality and agency cost. The alternative is that they arise from a bilateral allocation between the issuer and the underwriter, of the risks associated with mispricing the offer. The essence of the risk-allocation hypothesis is that underpricing and partial adjustment reflect a long-term quasi-contractual allocation of IPO pricing gains and losses. As repeat participants, underwriters can contribute to the efficiency of the IPO market in several ways. First, by intentionally underpricing, the underwriter can substitute underpricing for higher direct issue costs. Second, by committing to share gains from the most underpriced issues with investors, underwriters may develop relationships with investors who are willing to share losses when issues turn out to be overpriced. Third, by distributing shares of underpriced issues to prospective customers, such as prospects for future offerings, underwriters may be able to provide their underwriting services for lower fees. As long as underwriters have sufficient reputational capital at stake, implicit contracting does not require tacit or explicit collusion by underwriters or irrationality on the part of issuers. Under implicit contracting, underpricing is part of an effort to maximize expected net proceeds. Partial adjustment is a bilateral arrangement for sharing unexpected gains and losses between the 30

issuer and the underwriter, where the underwriter only realizes the gains indirectly, through its relationships with investors or its ability to attract future business. The implicit contract makes issue cost an increasing function of offer price relative to the filing range or minimum bid. This view is supported by the fact that no issuer in our sample received net proceeds below the minimum bid net of fees; 24.6 percent of the IPOs were offered at the minimum, of these, 13.9 percent of all IPOs had negative first-day aftermarket returns. For those IPOs offered at prices above the minimum, only 5.4 percent of all IPOs had negative initial returns. In summary, we examine the Japan IPO auction data in the light of a broad range of competing, but non-mutually-exclusive hypotheses. We conclude that Japan's auction-method evidence is most consistent with a risk-allocation hypothesis. Underwriters are repeat

participants in a market that is influenced by regulation and informal agreements among underwriters on pricing terms. In the context of existing regulation, they contribute to efficiency by substituting underpricing for direct issue costs, by committing to sharing gains (and losses) from underpriced IPOs with investors, and by allocating shares in order to provide lower fees. As underwriting fees constitute only about a third of average underpricing, we cannot exclude broader agency cost considerations, as suggested in Ritter and Welch (2002). Also, as some issues are seriously underpriced, and the underpricing appears to generate little dissatisfaction from issuers, we cannot exclude the possibility that issuers have asymmetric tolerance for over- or underpricing, which may be reflective of prospect theory or may be reflective of asymmetric risk tolerance arising, for example, from an issuers declining marginal ability to employ offer proceeds in high-valued projects.

31

References Aggarwal, R., Prabhala, N., and Puri, M., 2002. Institutional allocation and initial public offerings: Empirical evidence. Journal of Finance 57, 1421-1442. Allen, F., and Faulhaber, G., 1989. Signaling by underpricing in the IPO market. Journal of Financial Economics 23, 303-323. Amihud, Y., Hauser, S., and Kirsh, A, 2003. Allocations, adverse selection, and cascades in IPOs: Evidence from the Tel Aviv stock exchange. Journal of Financial Economics 68, 137-158. Aussenegg, W., Pichler, P., and Stomper A., 2003. IPO Pricing with Bookbuilding and a WhenIssued Market. Unpublished working paper. Massachusetts Institute of Technology. Beierlein, J., and Kato, H., 2003. The winner's curse and changing the rules of the game: The effect of maximum price on Japan's discriminatory price auctions. Unpublished working paper. The College of New Jersey. Benveniste, L., and Spindt, P., 1989. How investment bankers determine the offer price and allocation of new issues. Journal of Financial Economics 24, 343-361. Benveniste, L. and Wilhelm, W., 1990. A comparative analysis of IPO proceeds under alternative regulatory environments. Journal of Financial Economics 28, 173-208. Booth, J., and Chua, L., 1996. Ownership dispersion, costly information, and IPO underpricing. Journal of Financial Economics 41, 291-310. Booth, J., and Smith, R., 1986. Capital raising, underwriting and the certification hypothesis. Journal of Financial Economics 15, 261-280. Bradley, D. and Jordan, B., 2002. Partial Adjustment to Public Information and IPO Underpricing. Journal of Financial and Quantitative Analysis 37, 595-616. Chemmanur, T., 1993. The pricing of initial public offerings: A dynamic model with information production. Journal of Finance 48, 285-304. Chemmanur, T. and Fulghieri, P., 1994. Investment bank reputation, information production, and financial intermediation. Journal of Finance 49, 57-79. Chowdhry, B. and Sherman, A., 1996. International differences in oversubscription and underpricing of initial public offerings. Journal of Corporate Finance 2, 359-381. Edelen, R. and G. Kadlec, 2003. Issuer surplus and the partial adjustment of IPO prices to public information. Wharton School working paper. Goldreich, D., 1998. Underpricing in treasury auctions. IFA Working Paper 263-1998. London Business School.

32

Grinblatt, M. and Hwang, C., 1989. Signalling and the pricing of new issues. Journal of Finance 44, 393-420. Hamao, Y., Packer, F., and Ritter, J., 2000. Institutional affiliation and the role of venture capital: Evidence from initial public offerings in Japan. Pacific-Basin Finance Journal 8, 529558. Hanley, K., 1993. The underpricing of initial public offerings and the partial adjustment phenomenon. Journal of Financial Economics 34, 231-250. Hoberg, G., 2003. Strategic underwriting in initial public offerings. Yale School of Mangement Working Paper. Jagadeesh, N., Weinstein, M. and Welch, I., 1993. An empirical investigation of IPO returns and subsequent equity offerings. Journal of Financial Economics 34, 153-175. Jenkinson, T., and Ljungqvist, A., 1996. Going Public: The Theory and Evidence on How Companies Raise Equity Financing. Oxford University Press: Oxford. Kutsuna, K., and Smith, R., 2003. Why does book building drive out auction methods of IPO issuance? Review of Financial Studies, forthcoming. Kandel, S. O. Sarig, and A. Wohl, 1999. The demand for stocks: An analysis of IPO auctions. Review of Financial Studies 12, 227-247. Loughran, T. and Ritter, J., 2002. Why don't issuers get upset about leaving money on the table in IPOs? Review of Financial Studies 15, 413-444. Loughran, T. and Ritter, J., 2004. Why has IPO underpricing changed over time? University of Florida working paper. Lowry, M. and Schwert, G., 2002. IPO market cycles: Bubbles or sequential learning? Journal of Finance 57, 1171-1200. Pettway, R., and Kaneko, T., 1996. The effects of removing price limits and introducing auctions upon short-term IPO returns: The case of Japanese IPOs. Pacific-Basin Finance Journal 4, 241-258. Ritter, J. and Welch, I., 2002. A review of IPO activity, pricing and allocations. Journal of Finance 57, 1795-1828. Rock, K., 1986. Why new issues are underpriced. Journal of Financial Economics 15, 187-212. Sherman, A., 2002. Global trends in IPO methods: Book building vs. auctions. Unpublished working paper. University of Notre Dame. Shoken Torihiki Shingikai (Securities and Exchange Council), 1989. Kabushiki Koukai Seido No Arikata Ni Tsuite (Ideal Institutions of Initial Public offerings). Securities and Exchange Council: Tokyo. 33

Shoken Torihiki Shingikai (Securities and Exchange Council), 1995. Tento Tokusoku Shijyo No Kabushiki Koukai Seido To No Arikata Ni Tsuite (Ideal Institutions of Initial Public Offerings on the Second Division of JASDAQ). Securities and Exchange Council: Tokyo. Tamura, Y., 1997. Reform measures of initial public offerings: Introduction of book building method, Syoken Gyoho. Monthly Report of Japan Association of Securities Dealers, 51-62 (in Japanese). Tinic, S., 1988. Anatomy of initial public offerings of common stock. Journal of Finance 43, 789-822. Welch, I., 1989. Seasoned offerings, imitation costs, and underpricing of initial public offerings. Journal of Finance 44, 421-449. Welch, I., 1992. Sequential sales, learning, and cascades. Journal of Finance 47, 695-732.

34

Table 1 Price Adjustment, Underpricing, and Initial Returns by Subscription Ratio

Mean statistics by Subscription Ratio for all 321 IPOs on JASDAQ from January 1, 1995 through October 1, 1997. Subscription Ratio is the ratio of bids to shares offered in the auction tranche. Minimum Bid is determined by the underwriter using a required formula, based on the values of comparable public firms selected by the underwriter. In determining Minimum Bid, the formula requires a 15 percent discount from estimated market value based on the comparables. By regulation, Minimum Bid and Weighted Average Successful Bid are, respectively, the minimum and maximum offer prices for the public offer tranche. Maximum Adjustment is the ratio of Weighted Average Successful Bid to Minimum Bid. Price Adjustment is the ratio of Offer Price to Minimum Bid. Price Discount is the ratio of Offer Price to Weighted Average Successful Bid. Initial Return is measured as first freely trading Aftermarket Price/Offer Price 1. Underpricing is measured as 1 - Offer Price/Aftermarket Price. Num. Subscription Maximum Price Price of Ratio Adjustment Adjustment Discount Obs. (mean) (mean) (mean) (mean) All Observations Subscription Ratio 1 321 3 4.466 0.926 1.533 1.003 1.404 1.000 0.929 0.997 Initial Return (%) 11.50% -10.14 Underpricing (%) 8.59 -14.39 Positive Initial Returns (%) 81.00 0.00

1 < Subscription Ratio 2 37 1.592 1.034 1.011 0.979 -0.97 -1.86 43.24 2 < Subscription Ratio 3 61 2.580 1.165 1.099 0.947 7.55 6.24 78.69 3 < Subscription Ratio 4 61 3.379 1.380 1.284 0.938 9.59 7.54 81.97 4 < Subscription Ratio 5 61 4.580 1.665 1.502 0.910 13.57 10.08 85.25 5 < Subscription Ratio 7 58 5.942 1.843 1.667 0.913 16.32 12.37 93.10 7 < Subscription Ratio 10 29 8.136 2.080 1.836 0.889 22.79 17.10 100.00 10 < Subscription Ratio 11 13.483 2.437 2.134 0.880 25.16 18.33 100.00 Note: The Table demonstrates partial price adjustment for the public offer tranche of JASDAQ auction-method IPOs. Grouped by subscription ratio, offer prices of IPOs with more positive price adjustments also are more heavily discounted from the maximum adjustment of offer price and have higher initial returns and higher underpricing. No auctions have subscription ratios close to the 50 percent that would imply failure. Underwriters are likely to incur direct or indirect losses when subscription ratios are low.

Table 2 Sample Statistics on Offering, Issuer, Underwriter and Market

Mean and [median] statistics for all 321 IPOs on JASDAQ from January 1, 1995 through October 1, 1997. Price Adjustment is the adjustment to the Offer Price of the public offer tranche above the Minimum Bid. Statistical tests are of differences in means between subsamples where Minimum Bid is below Offer Price by more than 15 percent (high-price-adjustment) or less than 15 percent (low-price-adjustment). Two-tailed significance levels of t-values are shown at the .01 percent (***), .05 percent (**), and .10 percent (*) levels.

All IPOs
Number of Issues Percent of Sample Offering Characteristics Price Adjustment: (Offer Price / Minimum Bid) Offer Price (yen) Gross Proceeds (million yen) Shares Offered (thousands) Shares Offered to Outstanding Primary Shares Offered Percentage (Primary / Total Offered) Auction Tranche Percentage (Auction / Total Offered) Issuer Characteristics Age of Issuer (years) Employees Sales (million yen) Underwriter Characteristics Market Share of IPO Market for Underwriter Market Conditions JASDAQ Runup Days 5 through 1 (total return, %) JASDAQ Runup Days 40 through 1 (total return, %) Seasoned Offer within 12 Months (percent of IPOs) 321 1.404 [1.194] 2080.6 [1650.0] 2841.7 [1867.0] 1224 [1000.0] 18.60% [18.10%] 59.62% [54.55%] 50.59% [50.00%] 30.0 [29.9] 465.0 [318.0] 18607 [12176] 18.83% [17.4%] -0.63% [-0.86%] -3.76% [-4.89%]

Low-Price- High-PriceAdjustment Adjustment Subsample Subsample

140 43.6% 1.046 [1.035] 1420.1 [1380.0] 1946.2 [1310.5] 1209 [1012.0] 18.29% [17.94%] 58.99% [52.51%] 50.46% [50.00%] 33.3 [33.5] 498.5 [354.5] 19645 [12874] 16.75% [17.4%] -0.73% [-1.43] -7.38% [-7.84%] 181 56.4% 1.681 [1.497] 2591.5 [2180.0] 3534.2 [2255.0] 1236 [1000.0] 18.83% [18.25%] 60.11% [55.00%] 50.70% [50.00%] 27.4 [25.3] 439.0 [297.0] 17805 [11385] 20.45% [17.4%] -0.56% [-0.62] -0.96% [-1.65%]

t-value

13.71*** 7.33*** 3.61*** 0.39 1.31 0.55 1.31 4.15*** 1.04 0.75 2.99*** 0.65 6.75***

8.40% 4.30% 11.60% 2.49** [0.0%] [0.0%] [0.0%] Note: Price Adjustment is positively related to market runup over the 40 market days before the IPO. Seasoned offerings are more likely for issuers with high price adjustments.

Table 3 Sample Statistics on Price Adjustment and Initial Return

Mean and [median] statistics for all 321 IPOs on JASDAQ from January 1, 1995 through October 1, 1997. By regulation, Minimum Bid and Weighted Average Successful Bid are, respectively, the minimum and maximum offer prices for the public offer tranche. Price Adjustment is the ratio of Offer Price to Minimum Bid. Price Discount is the ratio of Offer Price to Weighted Average Successful Bid. Statistical tests are of differences in means between subsamples where Minimum Bid is below Offer Price by more than 15 percent (high-priceadjustment) or less than 15 percent (low-price-adjustment). Two-tailed significance levels of t-values are shown at the .01 percent (***), .05 percent (**), and .10 percent (*) levels.

Low-Price- High-PriceAll IPOs Adjustment Adjustment Subsample Subsample

Number of Issues Price Adjustment and Excess Demand Maximum Adjustment (Wtd. Avg Succ. Bid / Minimum Bid) Price Discount (Offer Price / Wtd. Avg. Successful Bid) Offer Price / Minimum Successful Bid Homogeneity (Min. Succ. Bid / Wtd. Avg. Succ. Bid) Subscription Ratio (Shares Bid For / Shares Auctioned) 321 1.533 [1.280] 0.929 [0.934] 0.971 [0.993] 0.957 [0.961]] 140 1.093 [1.089] 0.959 [0.962] 0.996 [1.000] 0.963 [0.965] 2.75 [2.71] 5.20% [3.60%] 3.87% [3.47%] 69.3% 1.102 [1.080] 10.0% 8.5% 1.008 [0.996] 52.9% 0.0% 1.048 [1.033] 15.0% 8.5% 181 1.873 [1.671] 0.906 [0.905] 0.952 [0.952] 0.952 [0.957] 5.79 [5.08] 16.38% [12.99%] 12.24% [11.50%] 90.1% 1.976 [1.713] 1.1% 0.0% 1.052 [1.018] 42.0% 0.0% 1.106 [1.069] 20.4% 0.0%

t-value

13.96*** 13.73*** 11.58*** 4.18*** 13.79***

4.47 [3.88] Initial Returns, Underpricing, and Aftermarket Price Compared to Bid Initial Return (Aftermarket Price / Offer Price 1) 11.50% [7.14%] Underpricing (1 - Offer Price / Aftermarket Price) Initial Returns and Underpricing > 0.0 (%) Aftermarket Price / Minimum Bid Observations < 1.0 (%) Observations = 1.0 (%) Aftermarket Price / Weighted Average Successful Bid Observations < 1.0 (%) Observations = 1.0 (%) Aftermarket Price / Minimum Successful Bid Observations < 1.0 (%) Observations = 1.0 (%) One-month JASDAQ-adjusted aftermarket return 8.59% [6.66%] 81.0% 1.594 [1.300] 5.0% 3.7% 1.033 [1.004] 46.7% 0.0% 1.081 [1.048] 18.1% 3.7%

6.79*** 3.61*** 4.61*** 13.84*** 3.34*** 4.05*** 3.06*** 1.94* 3.68*** 1.28 3.61***

-2.81% -3.39% -2.36% 0.46 [-4.42%] [-4.16%] [-4.54%] Initial Returns and Underpricing > 0.0 (%) 38.5% 40.6% 36.9% 0.66 Note: The public offer tranches of IPOs by Japans hybrid auction process exhibit partial adjustment similar to U.S. book-built IPOs. IPOs with the highest excess demand in the auction have public offer prices that are both (1) adjusted the most relative to the minimum bid and (2) discounted the most relative to the maximum offer price, resulting in higher initial returns. The fact that Offer Price is discounted more in the high-price-adjustment subsample is evidence that partial adjustment is discretionary, and not an artifact of offering regulations.

Table 4 Price Adjustments and the Aggregate Value of Initial Returns

Mean [and median] aggregate value of Initial Return grouped by Price Adjustment, for all 321 auction-method IPOs on JASDAQ from January 1, 1995 through October 1, 1997. Gross proceeds of the auction tranche is computed based on Weighted Average Successful Bid. Gross proceeds of the public offer tranche is computed based on Offer Price. Minimum gross proceeds is based on Minimum Bid and is guaranteed by the underwriter. The yen-value of money left on the table is defined as the difference between the aftermarket value of the tranche and gross proceeds. The percentage of money left on the table is computed based on aftermarket value. Statistical tests are of differences in means between subsamples where Minimum Bid is below Offer Price by more than 15 percent (high-price-adjustment) or less than 15 percent (low-price-adjustment). Two-tailed significance levels of t-values are shown at the .01 percent (***), .05 percent (**), and .10 percent (*) levels.

All IPOs
Number of Issues Auction Tranche Gross Proceeds (million yen) Minimum Gross Proceeds (million yen) Money left on the Table (million yen) Money left on the Table (percent) Public Offer Tranche Gross Proceeds (million yen) Minimum Gross Proceeds (million yen) Money left on the Table (million yen) Money left on the Table (percent) 321 1493.6 [980.0] 972.6 [661.5] 35.2 [3.2] 1.63% [0.38%] 1348.4 [884.0] 953.6 [652.0] 146.1 [57.6]

Low-Price- High-PriceAdjustment Adjustment Subsample Subsample

140 1003.4 [673.6] 914.6 [621.6] -0.3 [-3.5] -0.30% [-0.37%] 943.5 [636.4] 898.3 [607.8] 41.8 [22.5] 181 1872.8 [1212.4] 1017.5 [728.0] 62.7 [14.0] 3.09% [1.81%] 1661.7 [1044.0] 996.4 [688.5] 226.7 [139.5]

t-value

3.86*** 0.65 2.64*** 2.56** 3.34*** 0.62 6.89***

8.59% 3.87% 12.24% 6.57*** [6.67%] [3.48%] [11.49%] Note: Due to discretionary discounting of the offer price, money left on the table (as defined by Loughran and Ritter (2002)) is larger in the public offer tranche. Due to partial adjustment, money left on the table is significantly higher for the high-price adjustment subsample. Percentages of money left on the table for the public offer tranche are similar to percentages found by Loughran and Ritter for U.S. book-built IPOs.

Table 5 Offer Price Adjustment and Underpricing: Two-Stage Estimates

Two-stage least squares regression estimates of the system of Price Adjustment and Underpricing for all 321 hybrid auction-method IPOs on JASDAQ from January 1, 1995 through October 1, 1997. Two-tailed significance levels of t-values are shown at the .01 percent (***), .05 percent (**), and .10 percent (*) levels.

Price Adjustment (%)

Intercept Underpricing (1 - Offer Price / Aftermarket Price) Price Discount (Offer Price / Weighted Average Successful Bid) Subscription Ratio (Bids / Auction Shares) Maximum Adjustment (Wtd. Avg. Succ. Bid / Min. Bid) Homogeneity (Min. Succ. Bid / Wtd. Avg. Succ. Bid) Age of Issuer (ln(years)) Offer Size (ln(Shares Offered x Minimum Bid)) Primary Shares Offered Percentage (Primary / Total Offered) Auction Tranche Percentage (Auction / Total Offered) Market Share of Underwriter (decimal) Underwriter Fees (%) JASDAQ Runup Days 5 through 1 (decimal total return) JASDAQ Runup Days 40 through 1 (decimal total return) F-Value -0.042 (-1.29) 4330.80*** -0.003 (-2.46)** 0.828 (166.96)*** 0.485 (3.82)*** -0.011 (-2.06)** 0.005 (1.28) 0.024 (1.60) 0.092 (0.59) 0.026 (1.05) -0.091 (-4.49)*** -0.112 (-0.59) -0.029 (-1.16)

Underpricing (%)
1.844 (4.36)***

-0.478 (-2.26)** 0.012 (4.01)*** 0.005 (0.43) -1.073 (-3.60)*** 0.007 (0.59) -0.024 (-2.64)*** 0.024 (0.74) 0.576 (1.68)* 0.112 (2.09)** -0.104 (-2.31)** 1.010 (4.18)***

14.39***

Adjusted R2 0.994 0.315 Note: Price Adjustment is positively related to Maximum Adjustment, but, consistent with discretionary underpricing, is less that one hundred percent of Maximum Adjustment. Price Adjustment is negatively related to excess demand in the auction as reflected in Subscription Ratio, and positively related to Homogeneity of bid price. Consistent with partial adjustment, Underpricing is greater when the offer is priced lower relative to the Weighted Average Successful Bid and when Subscription Ratio is high, but is less when Homogeneity is high.

Table 6 Descriptive Statistics by Seasoned Offer

Mean and [median] statistics of sample of all 321 hybrid auction-method IPOs on JASDAQ from January 1, 1995 through October 1, 1997, and for the subsample that offered the JASDAQ minimum number of shares. The full sample includes 27 IPOs with SEOs within one year and 294 without. The subsample includes 14 with SEOs within one year and 168 without. Statistical tests are of differences in means between subsamples. Two-tailed significance levels of t-values are shown at the .01 percent (***), .05 percent (**), and .10 percent (*) levels. Subsample Issuing JASDAQ Minimum Number of Shares No No Seasoned Seasoned Seasoned Seasoned Offer within t-value Offer within t-value Offer within Offer within One Year One Year One Year One Year 27 294 14 168 Full Sample 18.60% [17.58%] 3131.3 [2380.0] 4893.1 [2523.0] 65.14% [60.00%] 51.85% 1.770 [1.684] 0.920 [0.918] 11.47% [8.21%] 19.91% [17.27%] 18.60% [18.16%] 1984.1 [1565.0] 2653.3 [1758.0] 59.11% [53.14%] 57.14% 1.511 [1.264] 0.928 [0.936] 11.51% [7.03%] 0.00 2.66** 1.61 1.66 -0.52 1.84* 1.30 -0.01 1.706 [1.533] 0.918 [0.930] 10.01% [6.33%] 28.91% [20.60%] 1.519 [1.250] 0.929 [0.934] 11.17% [6.67%] 1.63% [-2.42%] 0.93 0.98 -0.38 2.06** 16.62% [16.79%] 3014.6 [2190.0] 2749.0 [1094.0] 63.95% [59.37%] 17.27% [17.52%] 1827.1 [1500.0] 1161.7 [760.8] 59.01% [52.63%] -1.11 1.79* 1.23 1.05

Number of Issues Offering Characteristics Shares Offered to Outstanding Offer Price (yen) Offer Proceeds (billion yen) Primary Shares Offered Percentage (Primary/Total) Percent of IPOs at Minimum Size Auction Data Maximum Adjustment (Wtd. Avg. Succ. Bid/Min. Bid) Price Discount (Offer Price / Wtd. Avg. Succ. Bid) Returns Initial Return (percent) Three-month JASDAQ-Adjusted Returns (percent) Nine-Month JASDAQ-Adjusted Returns (percent)

-1.87% 4.16*** [-4.86%]

40.50% -1.71% 2.74** 55.60% 3.35% 1.85* [25.33%] [-7.83%] [41.47%] [-4.70%] Note: The signaling models of Welch (1989) and others indicate that high-quality firms that are hard to value at the time of the IPO are more likely to undertake small IPOs followed by SEOs once value is established in the market. Consistent with this view, SEOs are associated with high post-IPO returns. However, the table provides no support for the view that issuers perceived that they were undervalued at the time of the IPO. Compared to IPOs of firms that do not undertake SEOs within one year, for those that do, Shares Offered to Shares Outstanding and Percent of IPOs at Minimum Size are similar; Offer Proceeds are (non-significantly) larger; and yen-valued proceeds to selling shareholders are similar. Results are similar but generally weaker for the minimum size IPO subsample.

Table 7 Predictability of SEOs and Long-Run Market-Adjusted Returns

LOGIT regression where the dependent variable is one if a SEO occurs within one year of the IPO and zero otherwise and OLS regressions of three-month aftermarket returns. Models are for all 321 hybrid auction-method IPOs on JASDAQ from January 1, 1995 through October 1, 1997. The sample includes 27 firms with SEOs. Significance tests are shown in parentheses. The t-statistics for the LOGIT regression are based on asymptotic standard error estimates. Two-tailed significance levels of t-values are shown at the .01 percent (***), .05 percent (**), and .10 percent (*) levels. LOGIT Probability of Seasoned Offering Intercept Initial Return Price Discount (Offer Price / Wtd. Avg. Successful Bid) Subscription Ratio (Bids/Auction Shares) Maximum Adjustment (Wtd. Avg. Succ. Bid / Min. Bid) Homogeneity (Minimum Succ. Bid / Wtd. Avg. Succ. Bid) Age of Issuer (ln(years)) Offer Size (ln(Shares Offered x Minimum Bid)) Primary Shares Offered Percentage (Primary/Total) Auction Tranche Percentage (Auction/Total Offered) Market Share of Underwriter (decimal) JASDAQ Runup Days 5 through 1 (decimal total return) Three-Month JASDAQ-Adjusted Return w/o Initial Return -11.391 (-0.81) 0.235 (0.14) -4.187 (-0.54) 0.143 (1.41) -0.156 (-0.36) 0.964 (0.08) -0.299 (-0.66) 0.923 (2.86)*** 0.979 (0.77) -3.471 (-0.28) 0.768 (0.35) -12.260 (-1.13) 3.019 (3.36)*** OLS Regressions Three-month Three-month Aftermarket Aftermarket Return with Return w/o Initial Return Initial Return 2.738 1.565 (3.42)*** (1.79)* -0.154 (-1.52) -0.891 0.632 (-1.87)* (1.22) 0.030 -0.014 (4.50)*** (-1.90)* 0.012 0.042 (0.44) (1.39) -2.393 -1.462 (-3.47)*** (-1.94)* 0.017 -0.043 (0.60) (-1.41) -0.043 -0.043 (-2.13)** (-1.98)** 0.070 0.058 (0.93) (0.72) 1.789 -0.176 (2.29)** (-0.21) 0.204 0.166 (1.66)** (1.24) 2.337 -0.519 (4.24)*** (-0.85)

F-Value 14.48*** 2.20** Adjusted R2 0.301 0.041 McFadden's R2 0.173 Note: Results of both the LOGIT and OLS models suggest that SEO issuers did not anticipate that they would undertake SEOs shortly after the IPO or the positive aftermarket performance of their shares. In particular, IPOs that are followed by SEOs are not significantly smaller (Offer Size) than other IPOs, and selling shareholders do not sell significantly smaller percentages of the offer (based on Primary Shares Offered Percentage). In addition, the offers are not discounted significantly more (Price Discount) and Initial Returns are not significantly larger. Three-Month JASDAQ-Adjusted Return and Offer Size significantly predict SEOs. None of the choice variables that are likely to be related to the intention to underprice and to follow with an SEO significantly explains threemonth aftermarket returns either including or excluding initial returns and is correctly signed.

Table 8 Source of Selling Shareholder Proceeds

Mean and [median] statistics of sample of all 321 hybrid auction-method IPOs on JASDAQ from January 1, 1995 through October 1, 1997, and for the subset of IPOs for which selling shareholders choose the tranche in which their shares are sold. The full sample includes 70 IPOs where selling shareholders sold in the auction and 251 they sold in the public offer. The subsample includes 17 IPOs where selling shareholders choose the auction, and 73 the public offer. Statistical tests are of differences in means between subsamples. Two-tailed significance levels of tvalues are shown at the .01 percent (***), .05 percent (**), and .10 percent (*) levels.

Full Sample
Proceeds from Auction Tranche Number of Issues Offer Details Primary Shares to Total Offered Shares (%) Secondary Shares Offered (thousand) Number of Selling Shareholders ` Anticipation of Underpricing Price Adjustment: (Offer Price/ Minimum Bid) Price Discount: (Offer Price/Wtd. Avg. Succ. Bid) Subscription Ratio JASDAQ Runup, Days 40 through 1 (total return, %) Anticipation of Aftermarket Performance Seasoned Offerings in One Year (percent of IPOs) One-month JASDAQ-Adjusted Returns (percent) Twelve-month JASDAQ-Adjusted Returns (percent) 70 41.40% [44.44%] 720.7 [600.0] 6.06 [6.00] 1.310 [1.145] 0.924 [0.927] 4.34 [3.60] -3.76% [-4.18%] 4.29% [0.00%] -2.77% [-5.36%] -2.22% [-10.81%] Proceeds from Public t-value Offer Tranche 251

Subsample for which Shareholders Could Select Selling Tranche

Proceeds Proceeds from from Public t-value Auction Offer Tranche Tranche 17 73

64.70% -11.29***- 50.00% 50.00% [58.82%] [50.00%] [50.00%] 422.1 [400.0] 6.32 [5.00] 1.431 [1.202] 0.930 [0.938] 4.50 [4.00] -3.76% [-5.20%] 9.56% [0.00%] -2.83% [-4.16%] -1.73% [-9.53%] 7.36*** -0.18 640.41 [550.0] 6.06 [6.00] 1.212 [1.087] 0.932 [0.933] 2.93 [2.85] -9.70% [-9.90%] 5.88% 0.00% -1.97% [-7.52%] 586.1 [500.0] 6.25 [5.00] 1.404 [1.208] 0.926 [0.928] 4.29 [3.81] 0.92 -0.14

-1.60 0.96 -0.44 0.00

-1.66* 0.41 -2.14**

-4.25% -2.27** [-5.82%] 6.85% 0.00% -3.50% [-4.46%] -0.14 0.29

-1.41 0.02 -0.10

0.50% 0.53% 0.00 [-8.05%] [-10.56%]

Note: The proportion of total shares offered by the selling shareholders determines the tranche in which secondary shares are sold. Shareholders realize proceeds based on the price at which the tranche is sold. Incentives to sell in the auction are greatest when high levels of price discounting are expected. The table provides no evidence that secondary shareholders choose to sell shares in the auction tranche when they expect Price Discount to be large. The results suggest that selling shareholders do not anticipate high levels of bidding, price adjustments, or price discounting.