From the IPO to the First Trade: Is Underpricing Related to the Trading Mechanism?

Transcription

1 From the IPO to the First Trade: Is Underpricing Related to the Trading Mechanism? Sonia Falconieri Tilburg University Warandelaan 2 P.O. Box LE Tilburg Netherlands Phone: s.falconieri@ uvt.nl Albert Murphy Manhatten College Manhattan College Parkway Riverdale, NY Phone: albert.murphy@manhattan.edu Daniel Weaver* Department of Finance Rutgers Business School Rutgers, The State University of New Jersey 94 Rockafeller Road Piscataway, NJ Phone: (732) Fax: (732) daniel_weaver@rbsmail.rutgers.edu February 27, 2004 *Corresponding author We thank Amber Anand, Ekkehart Boehmer, Steven Ongena, Gideon Saar, Carsten Tanggaard, and Ivo Welch for helpful comments. Falconieri and Weaver gratefully acknowledge support for this project from the New York Stock Exchange.

2 ABSTRACT As documented by a vast empirical literature, IPOs are characterized by underpricing. Most of the theoretical literature has linked the size of underpricing to the IPO procedure used on the primary market. In this paper, by using a matched sample of NYSE and Nasdaq IPOs, we show that the size of underpricing also depends on the trading method used in the IPO aftermarket. There are two major methods of opening trading of initial public offerings (IPOs) in the U.S. The NYSE is an order-driven market where a call auction allows supply and demand to be aggregated (at one location) prior to the start of trading. In contrast, Nasdaq is a quote-driven market. Dealers can only specify their best quotes, and participants have no idea of supply and demand away from the inside quotes. We propose a new proxy for ex ante uncertainty of firm value and test it. Our results show that there is a larger level of uncertainty at the beginning of trading on Nasdaq than on the NYSE. This in turns is associated with larger levels of underpricing for Nasdaq IPOs. We suggest that this may be due to the different informational efficiency of the two trading systems. Keywords: underpricing, volatility, trading system, demand uncertainty. 2

3 1. Introduction As documented by a large empirical literature, a peculiar feature of initial public offerings (IPOs) is underpricing: i.e. the spike in the price of a share relative to its offering price on the first day of trading. Many theoretical papers have tried to explain why new issues are underpriced. Most of these explanations are based on the existence of some type of asymmetric information in the IPO process. 1 A main empirical implication (Ritter (1984), Beatty and Ritter (1986)) is that underpricing can be interpreted as a premium for the ex ante uncertainty about the firm market value. More recently, the literature on IPOs, both empirical and theoretical, has focused on explaining the size of underpricing. These papers mainly relate the size of underpricing to the IPO procedure used in the primary market (e.g. Biais and Faugeron-Crouzet (2002), Derrien and Womack (2003)). In particular, underpricing seems to be lower in auction-like selling procedures which are viewed as more informationally efficient than bookbuilding, i.e. they allow for more information disclosure. In this paper, by using a matched sample of NYSE and Nasdaq IPOs, we provide evidence that the size of underpricing is also related to the trading method used in the IPO aftermarket. Our findings suggest that this is due to the different informational efficiency of alternative trading mechanisms. 1 Rock (1986) explains underpricing as the result of a winner s curse problem retail (uninformed) investors face visà-vis institutional (informed) investors; Allen and Falhauber (1989) interpret underpricing as a way for the firm to signal its quality to the market whose cost will be offset by the higher price good firms will be able to charge in the subsequent seasoned offerings; in Benveniste and Spindt (1989) underpricing is the result of the informational rents the underwriter needs to pay to institutional investors in order to elicit their information about the market value of the issue. 3

4 In the U.S., bookbuilding is the dominant selling mechanism for IPOs on the primary market. 2 In contrast, there are two methods to open secondary market trading in equities. In order-driven environments like the NYSE, trading starts with a call market where public orders are consolidated. Investors submit orders that may include the maximum (minimum) price that they are willing to buy (sell) at. In quote-driven markets like Nasdaq, the first trade is preceded by a period (pre-opening) during which dealers can display the prices at which they will buy and sell. These quotes however are non-binding and do not necessarily reflect information from public orders placed with dealers before the opening. In general, the information about the demand during the pre-opening is quite noisy. The same processes are used to open secondary market trading after an IPO. 3 Price discovery during the pre-opening has received considerable attention in the financial literature. Similarly, several papers have examined how underpricing is related to the IPO selling mechanism. Although Boehmer and Fishe (2000) and Ellis, Michaely, and O'Hara (2000) relate underpricing to market structure, they do not directly examine the relationship between the pricing of IPOs and the opening procedures in the secondary market. This paper seeks to fill that void. Given the price run ups of recent IPOs, the question immediately arises as 2 With the only exception being the OpenIPO process, which uses a Dutch auction. OpenIPO has been introduced in February 1999 and, so far it has not attracted the interest of equity markets. 3 The NYSE has recently begun allowing participants to pay to see total non-hidden depth away from the inside during the continuous trading session (called OpenBook). On the NYSE the managing underwriter "runs the market" at the beginning of secondary market IPO trading. Only the specialist and the managing underwriter have access to the complete contents of the book. Since there is no consolidating function on Nasdaq, there is still no way for any dealer to have complete knowledge of secondary market supply and demand away from each dealer's inside quotes. Even more recently, Nasdaq began SuperMontage which allows dealers to enter multiple quotes, but it is voluntary whether dealers want to hold limit orders away from the inside or expose them on SuperMontage. During the time of our study, neither OpenBook nor SuperMontage existed. However, for the reasons just cited, the new systems should not affect our conclusions regarding the opening procedures on both markets. 4

5 to whether a better-structured opening method results in better performances of IPOs both in term of lower underpricing and volatility. In addition, very little has been done to compare IPOs on the two trading systems. Corwin and Harris (2001) and Affleck-Graves, Hegde, and Miller (1996) compare the size of underpricing on NYSE and Nasdaq IPOs reaching very different results. However, neither study controls for industry effects, so that their results may be driven by differences in the types of firms on each market. A recent theoretical paper by Bennouri (2003) compares auction versus dealership markets, along two main dimensions that distinguish these two trading structures, the source of quotes (public limit orders versus dealer quotes) as well as trading concentration (centralized versus fragmented). Bennouri shows that, in equilibrium, a centralized order-driven market (such as the NYSE) is more viable and more informationally efficient than a decentralized quotedriven market (such as the Nasdaq.) The higher viability means that any asymmetric information problem is less severe on auction markets because, due to trading concentration, information spreads rapidly among market participants. The fact that auction markets are more informationally efficient instead refers to the higher informativeness of equilibrium prices, i.e. more information is conveyed through prices This paper provides empirical evidence to support Bennouri s assertions. Our analysis shows that while the percentage of IPOs that are underpriced on the two markets is comparable, the magnitude of underpricing is much larger on Nasdaq. The higher level of underpricing on Nasdaq is accompanied by a higher level of volatility on the opening day of trading than on the NYSE. We develop a hypothesis that the level of underpricing is 5

6 related to ex ante uncertainty about the firm s true value. We propose a new measure of ex ante uncertainty based on a volatility ratio of quoted spread midpoints during the beginning of trading to the remainder of the first trading day. We find that the volatility ratio is significantly larger on Nasdaq. Our result suggests that the call market structure of NYSE IPOs is more efficient than the quote-driven process of Nasdaq in resolving uncertainty, since actual public supply and demand is consolidated on the specialist book. In a market maker system, instead, supply and demand is more fragmented and may not reflect public orders. We employ regressions to control for other factors known to be associated with IPO underpricing and find our results are robust. Related to ex ante uncertainty is a recent paper by Saar (2001). In that paper the author argues that investor demand uncertainty leads to wider spreads. We find support for Saar s argument. Spreads on Nasdaq are much wider than those on the NYSE at the beginning of trading in an IPO. This difference reduces greatly by the close of trading. Further investigation reveals that Nasdaq spread widths decline dramatically in the first few minutes of trading, suggesting a resolution of uncertainty. A final contribution of this paper is that it examines the distribution of opening times of the NYSE as well as Nasdaq. Aggarwal and Conroy (2000) show that Nasdaq IPOs do not generally open secondary market trading at 9:30. Of the 188 IPOs in their study 172 opened after 10 AM, with some opening as late as after 2 PM. In contrast, we find that NYSE IPOs are much more likely to open at 9:30 AM. 6

7 The remainder of the paper is organized as follows. The next section presents a short review of the literature and develops our hypotheses based on it. In Section 3, we describe the data set used to test our hypotheses. Section 4 describes the empirical results, while Section 5 concludes. 2. Literature Review In addition to Bennouri (2003), which is the most closely linked to ours, the present paper is related to two different strands of literature: the literature on price discovery in different trading systems and the literature on the determinants of the level of underpricing. This section presents a brief review each strand. 2.1 Price Discovery on the NYSE and Nasdaq Madhavan and Panchapagesan (1999) examine the opening procedures on the NYSE. They show empirically that specialists significantly facilitate price discovery because they learn from observing the evolution of the limit order book. In a pure single-price call auction, the opening price would be the one clearing as many shares as possible. However, on the NYSE the specialist sets a price, which may be different than the price that would prevail in a pure auction market with only public orders. This is because of his obligation to provide price continuity. As a consequence, the opening price is set to provide minimum variation from the previous day's close and is shown to be more efficient than the price resulting from an auction market without specialists. Whereas, on the NYSE, all orders come together in a central location prior to opening, the Nasdaq pre-opening period is quite different. First, there are multiple market makers, each with there own set of public limit orders. Each market maker submits her best bid and offer into 7

8 the Nasdaq system. Therefore, there is no collective knowledge of public supply or demand beyond the inside spread. That is, customer orders away from the dealer s inside quote are only known to that dealer, whereas on the NYSE all limit orders are known to the specialist. 4 Second, the quotes posted during the pre-opening are non-binding (except for Instinet quotes). Cao, Ghysels, and Hatheway, (2000) examine the pre-opening on Nasdaq for existing stocks and find that market makers seem to use the pre-opening quotes to signal each other as to their supply and demand. Aggarwal and Conroy (2000) examine the pre opening period on Nasdaq for a set of IPOs and reach similar conclusions. They also find that the lead underwriters quote explains a large portion, but not all, of the initial return. Since order driven markets tend to have centralized demand and quote driven markets fragmented demand these differing market structures may impact the level of underpricing in IPOs. 2.2 Determinants of the Level of Underpricing Corwin and Harris (2001) examine the initial listing decisions of firms going public. Using a sample of IPOs from 1991 to 1996 that either listed on the NYSE or met the NYSE's minimum-listing requirements and listed on the Nasdaq National Market, they try to identify the relevant variables in determining which market a newly public firm will choose for its secondary trading. By looking at the results of a Probit model, they conclude that the firm size is the most important determining factor. Smaller and riskier firms list on Nasdaq to avoid the higher listing 4 See Footnote 2. 8

9 fees on the NYSE. However, given the higher level of underpricing on Nasdaq, listing on Nasdaq could also be a strategic decision of underwriters to increase their compensation. 5 Corwin and Harris also attempt to identify the determinants of underpricing. They conclude that the amount of underpricing is related to the aftermarket standard deviation of return (5 day returns over a 100 day window). However, they do not control for industry effects in their regressions. Therefore, given that Nasdaq stocks are more typically in industries like genetic engineering or technology firms, it is entirely possible that the standard deviation of return used in their regressions may be a proxy for industry membership. In our analysis, we instead control for the industry by matching NYSE IPOs to Nasdaq IPOs based on 4 digit SIC (Standard Industry Classification) code and offering size (in US dollars). Affleck-Graves, Hegde, Miller, and Reilly (1993) find, for a sample of IPOs from 1983 to 1987, no statistically significant difference in the amount of underpricing for NYSE v. Nasdaq/NMS issues (4.82% v. 5.56%). This is in contrast with the findings of Corwin and Harris who instead document marginally significant differences in the underpricing levels on NYSE v. Nasdaq issues. The difference in their results may be explained by the different measure of underpricing that they adopt. Affleck-Graves et. al. define underpricing as the spread between the closing price of the first day of trading and the offering price, while Corwin and Harris use the opening price. Barry and Jennings (1993) look at the return on IPOs. They find that 90% of the initial day return is captured in the offer-to-open return. The percentage increases to almost 95% for those firms that had positive returns on the opening trade. 5 Ellis, Michaely, and O'Hara (2000) find that the underwriter's trading profits are positively related to the amount of underpricing. This suggests that underwriters view trading profits as a regular part of their total compensation 9

10 Beatty and Ritter (1986) argue that the amount of ex ante uncertainty as to true firm value is the main determinant of the level of underpricing in IPOs. They build on Rock (1986) which interprets underpricing as a premium to uninformed investors for the winner s curse problem they face vis-à-vis the informed investors, who observe the true firm value. Beatty and Ritter further develop this idea by claiming that more ex ante uncertainty worsens the winner s curse problem and, consequently, requires larger underpricing. They test their theory by using as a proxy for ex ante uncertainty the inverse of the gross proceeds raised in the offering as well as the number of uses mentioned in the prospectus; Ritter (1984) uses instead the standard deviation of the daily aftermarket return. Although they (and others) find empirical support for their proxy, Jenkinson and Ljungqvist (1996) point out that using standard deviation of return may be an inadequate proxy since it may reflect the relationship between risk and return. In this paper we propose a more direct measure of ex ante uncertainty, which does not suffer from the problems associated with typical risk measures. More recently, Draho (2001) extends the fixed price offering model of Beatty and Ritter (1986) to a bookbuilding procedure. He develops a model of IPOs where both the primary and secondary prices are endogenous and shows that underpricing occurs if there is uncertainty about the price of the IPO on the secondary market. In this case, underpricing will be increasing in the ex-ante uncertainty about the firm value. However, he also shows that the uncertainty about the secondary price, and the resulting underpricing, can be reduced by improving the information production during the primary mechanism. In particular, he suggests that in bookbuilding this could be achieved by requiring investors to submit limit orders. 6 6 This is also consistent with the findings of Cornelli and Goldreich (2002). 10

11 Related to the IPO ex ante uncertainty literature is a recent paper by Saar (2001). In that paper Saar argues that increased uncertainty about investor demand leads to wider bid-ask spreads. The traditional market microstructure literature explains the impact of trades on prices as a result of the asymmetric information among investors about the future cash flows of assets. Saar instead claims that it is reasonable to assume uncertainty about investors preferences and endowments (demand or investor uncertainty). Saar suggests that the aggregated demand structure of the NYSE will lead to lower spreads than on a multiple market maker system such as Nasdaq. 2.3 Summary Taken together, the two strands of literature described above suggest that the market structure of an IPO s aftermarket may impact the level of underpricing. Our hypothesis is the following. If underpricing is related to the ex ante uncertainty about the firm value and if this uncertainty is not completely resolved during the bookbuilding process, then the trading concentration feature of the NYSE s structure should allow for a quicker resolution of any residual value uncertainty than the fragmented structure of the Nasdaq. This would in turn results in less underpricing and narrower spreads on NYSE IPOs than on IPOs that trade on Nasdaq. We test these hypotheses in the following sections. We also develop a new proxy for ex ante uncertainty that does not suffer from the biases of previous proxies. 3. Data 11

12 Our goal is to control for industry effects, by creating a matched sample of NYSE and Nasdaq IPOs. Since there are far fewer IPOs on the NYSE than on Nasdaq, our matching methodology is to find matches for the smaller group the NYSE. Our first step is therefore to compile a list of all NYSE IPOs between January 1993 and December 1998 from the Securities Data Corporation (SDC) New Issues Database. Barry and Jennings (1993) find that the returns of operating companies and closed-end-funds behave very differently. Therefore, consistent with Corwin and Harris (2001), we exclude investment funds (including mortgage securities), REITs, and real estate firms from our sample. We next find all Nasdaq matches for each NYSE IPO. A match is classified as adequate if both firms have the same four-digit SIC code and if the IPOs occurred within 6 months of each other. This latter condition attempts to correct for biases related to overall market conditions. In the case of multiple Nasdaq matches we choose the three (or less) that are closest in value to the NYSE IPO (Karolyi and Stulz (1996)). Our final sample consists of 170 NYSE and 339 Nasdaq IPOs. Every NYSE IPO has at least 1 Nasdaq IPO as its match. Our sample of IPOs is much smaller than the actual number of IPOs that occurred during our sample period as well as the number of IPOs examined in other studies. However, our matching methodology allows us to control for overall market conditions as well as industry effects. Therefore, we are not subject to the criticism that Nasdaq IPOs have a larger amount of underpricing due to the fact that more technology stocks list there. We thus feel that our design will allow us to draw inferences about the impact of differing market designs on the level of underpricing. 12

13 Table 1 contains descriptive statistics for our sample. Examining Table 1 reveals that as expected, the average NYSE IPO is four times larger than the average Nasdaq IPO. Also, the average NYSE IPO offering price is 50% larger than the average Nasdaq IPO. Accordingly, we control for the offering size in tests of relationships between variables. Since we use four-digit SIC codes to create our matches, we partition our sample by SIC major division. Examining Table 1 by SIC division reveals that on average there are roughly twice as many Nasdaq than NYSE IPOs, but there is variation. This uneven distribution of IPOs causes us to adopt the methodology of reporting results for our overall samples as well as by SIC major division. 4. Empirical Results The first variable we examine is the location of the first trade relative to the offering price. An IPO is deemed under (over) priced if the offering price is less (more) than the first trade in the stock. Table 2 contains the proportion of IPOs under or over priced by market and SIC major division. Overall we find that 80% of our sample of NYSE IPOs is under priced, while 78% of our sample of Nasdaq IPOs is underpriced. For both markets, the proportion of IPOs that are over priced is less than 2%. Examining the results for SIC divisions reveals that for only 3 divisions is the proportion of Nasdaq IPOs that are underpriced greater than NYSE IPOs. Both overall and for most of the SIC divisions, the percentage of IPOs that are underpriced is relatively similar across market structure. We next examine the amount of underpricing for our sample. We only include those IPOs that were underpriced. The amount of underpricing is defined as 13

14 TP OP ρ = OP where, TP is the first trade price and OP is the offering price. Alternatively, ρ can also be interpreted as the return on the offering price. The results are contained in Table 3. The amount of underpricing for Nasdaq IPOs is greater than for NYSE IPOs. Overall the average Nasdaq underpricing is 28% larger than NYSE underpricing. This amount is statistically significant at standard levels. Examining the results by industry segment reveals that although there are instances where the amount of underpricing on Nasdaq is less than on the NYSE, the amount is never statistically significant. Although, Nasdaq IPOs are underpriced by a larger amount than NYSE IPOs, it is not clear whether this difference is due to market structure or some other factor influencing the market choice made by investment bankers. We therefore next focus on differences in market structure and their impact on measurable variables. As mentioned earlier, Saar (2001) develops a model of demand uncertainty that suggests that a specialist system of trading (as on the NYSE) is better able to ascertain demand and will thus have narrower spreads than a multiple market maker system. Our data provide a good test of this hypothesis. The results for the opening spreads for IPOs are contained in Table 4. Opening spreads are defined as the spread in effect at the time of the first trade or the first quote after the first trade. 7 For both under-priced and equally-priced IPOs, Nasdaq spreads are significantly larger than NYSE spreads. For under-priced IPOs, Nasdaq opening spreads are on average $0.30 larger than NYSE spreads. For equally-priced IPOs they are $0.39 larger. Although the magnitudes vary, these results are generally consistent across the SIC divisions. 14

15 Wide opening spreads are consistent with the uncertain demand hypothesis of Saar (2001). The fact that we find much wider opening spreads on Nasdaq suggests that the method used by Nasdaq to open trading in IPOs leads to a lower amount of information concerning demand vis a vis the NYSE. However, the difference in opening spreads may merely be a reflection of the wider spreads on Nasdaq documented by many studies. If the difference in opening spreads, between markets, is due to general market structure rather than to different levels of demand uncertainty, we would expect the same difference in closing spreads. 8 Therefore, we next examine average closing spread on the first trading day, for our sample. The results, contained in Table 5, show that the difference between NYSE and Nasdaq closing spreads for underpriced IPOs is half of what it is at the open -- $0.15. Comparing closing spreads to opening spreads reveals that closing spreads on the NYSE are 18% less than opening spreads, while on Nasdaq the same relationship is over 36%. This provides support for the hypothesis that Nasdaq s method of opening IPOs is associated with more uncertainty as to demand than the NYSE s method. It also indicates that this uncertainty progressively declines during the first trading day. To complete our analysis, we also examine the intraday spread pattern for our IPO sample to determine how long it takes for the differences in spread width to reduce. We find that the differences observed at the open reduce greatly within the first few minutes of trading as illustrated in Figure 1. 9 While average NYSE spreads exhibit a relatively gradual decline over the first 10 minutes, the pattern of Nasdaq spreads exhibits a much more dramatic decline within 7 On the NYSE the first trade generally occurs before the opening quote 8 If uncertainty does not get resolved during the day, then closing spreads should be equal to opening spreads. If the initial level of uncertainty is similar across markets and gets resolved by the end of the day, then differences in spread width between markets should not change. 15

16 the first 4 minutes of trading. This is consistent with the uncertainty hypothesis suggesting that uncertainty is resolved within the first few minutes of trading. To the extent that underpricing is associated with uncertainty (examined in more depth later) these findings suggest that at least part of the difference in underpricing between stocks in our Nasdaq and NYSE samples may be due to market structure. We also examine the volatility of quotes without bid-ask bounce on the two markets over the first day of trading (see Table 6 for the results). Volatility is measured as the standard deviation of quote midpoints on the first day of trading. Nasdaq IPOs have a much higher volatility compared to NYSE IPOs. For instance, for underpriced IPOs (Panel A, in Table 6) Nasdaq volatility is overall almost 16% larger than NYSE volatility. And the result is statistically significant. It is also higher in 4 of the 7 SIC divisions, and statistically significant in 2 of the four. The same pattern is evidenced in Panel B for equally priced IPOs. The pattern of spread width illustrated in Figure 1 and the opening day volatility documented in Table 6, suggests that the ex ante uncertainty about the market value of an IPO firm is made of two different components: one driven by demand uncertainty as related to the trading system and the other company-specific. Accordingly, in this paper we propose a new measure of value uncertainty. Ritter (1984) points out that riskier IPOs (those with large standard deviation of return) are likely to have more uncertainty of value. Thus, Ritter (among others) uses the standard deviation of daily return as his value uncertainty proxy. It is well established in the literature that the standard deviation of return contains both market and unique risk. 9 Note that Figure 1 includes all IPOs, not just those underpriced. 16

17 Therefore, it is possible for two firms with the same standard deviation of return to have very different levels of value uncertainty. In addition, a recent SEC investigation into IPO under-pricing suggests that investment bankers allocate shares to favored clients who sell them quickly in the secondary market. 10 Empirical evidence supporting to the SEC allegations is provided by Aggarwal, Nagpurnanand, Puri (2002). They document that indeed underwriters favor their institutional clientele by allocating them more shares in hot issues that they trade up immediately in the aftermarket. This suggests that the focus of value uncertainty should be short term in nature. We therefore propose that a nearer-term measure of value uncertainty is to measure the variability of prices at the beginning of an IPO s trading. The intuition is straightforward assume a risk-averse informed investor who knows the expected value, and distribution, of the opening price of an IPO. Further assume a desired underpricing amount (consistent with SEC allegations). Then it follows that the more variable the distribution of opening price, the larger the underpricing. Our measure of uncertainty follows directly from this intuition. We first determine the standard deviation of quoted spread midpoints in the first two hours of trading and standardize it by dividing by the standard deviation of quoted spread midpoints through the remainder of the day. 11 A two hour period is chosen to insure sufficient observations to obtain a volatility measure. 12 If initial price volatility is indicative of a risky issue then the ratio will equal one. If the ratio is greater than one it implies that the uncertainty as to the firm value is resolved after trading begins and secondary market demand is determined. 10 See Crime and punishment: Initial public offerings, The Economist, December 15, 2001, p Quote midpoints are used instead of observed prices to remove any spurious volatility caused by the familiar bidask bounce. Also, in cases where quotes do not change over the day, the ratio is set equal to 1. 17

18 Only the stocks that open trading before noon are included in this test. Figure 2 illustrates the percentage of each market s IPOs that occur for each half hour of trading throughout the day. Virtually all of our IPOs occur before 12 noon. 13 Results for our volatility ratios are shown in Table 7. Examining Table 7 for underpriced IPOs, reveals that both overall and for virtually all of the SIC divisions, the volatility ratio is greater than one. Further, in only one case is the volatility ratio greater on the NYSE than it is for Nasdaq (SIC Division B). This suggests that the fragmented demand structure of Nasdaq leads to more value uncertainty as compared to the centralized demand found on the NYSE. The last step is therefore to investigate whether the amount of underpricing is related to the uncertainty of demand (as measured by the volatility ratios). Given that we saw that the larger underpricing on Nasdaq is as well associated with a higher volatility ratio as compared to the NYSE, this would suggest the existence of a link between the two variables. We test for evidence of this relationship. For each market, we regress the amount of underpricing on our volatility ratio, while controlling for other variables known to be associated with underpricing, including the offering amount as well as the volatility. We need to introduce these two controls because, as we have seen, Nasdaq offering sizes are much smaller than NYSE offering sizes. We need therefore to be sure that the offering size is not the driving force of the differences in the level of underpricing we observe in our sample. Similarly, we need to control for the volatility because residual volatility after the demand uncertainty is resolved is higher on Nasdaq than on the NYSE. So 12 Indeed, if the majority of the uncertainty-induced volatility occurs in the first 10 minutes of trading, then our measure of volatility will be downward biased making finding a difference more difficult. 18

19 it might be that the higher underpricing for Nasdaq IPOs is due to the higher riskiness of the market. Finally, we control for whether the issue is oversubscribed (hot issues) since there is evidence in the literature (Cornelli and Goldreich (2001)) that oversubscription is positively related to underpricing. Consequently, we perform the following regression % Underi = α + β1volratioi + β2offeringi + β3volatility + β4hoti where %Under i is the percentage amount of under-pricing for firm i, VolRatio is the volatility ratio defined as the standard deviation of spread midpoints for the first 2 hours to the rest of the day, Offering i is the log of dollar size of firm i's offering, and Volatility i is the standard deviation of spread midpoints on the first trading day (after the first two hours of trading). 14 Hot is defined as (Offering Price Mid Range)/Mid Range; where Mid Range is the midpoint of the originally filed price range. We perform the regression both overall and by market. We only consider those IPOs that are under-priced in this analysis. The results are contained in Table 8. Note that for 1 NYSE stock and for 23 Nasdaq stocks, spreads did not change after the first two hours of trading, so a volatility ratio could not be calculated for those firms and they are excluded from the analysis. Also, in cases where quotes do not change over the day, the volatility ratio is set equal to 1. Overall, we find that the amount of under-pricing is positively related to the amount of volatility on the first day of trading. The parameters for size of the offering and our hot issues proxy are both of the expected sign (consistent with previous studies) and are statistically 13 As an interesting aside, we note that NYSE IPOs tend to begin trading earlier than Nasdaq IPOs. Although other studies have documented the distribution of Nasdaq opening times, no other study to our knowledge, has examined NYSE opening times. 14 We exclude the first two hours in which it seems that the main component of the volatility is the uncertainty of demand. By doing so, we are sure to capture only the riskiness of the market. 19

20 significant. Finally, we find that the parameter estimate for our volatility ratio is of the expected sign and statistically significant. Examining the results by market reveals that all parameters are of the same sign as the combined regression; however the offering size parameter is not statistically significant for either market. 15 Also, while the parameter estimate for our volatility ratio is the same for both Nasdaq and the NYSE, it is only statistically significant for the Nasdaq. When the fact that the parameter estimates are the same for each market is combined with the fact that our volatility ratio measure is larger on Nasdaq than on the NYSE, it suggests that Nasdaq IPOs are underpriced by a larger amount than NYSE IPOs due to the differences in market structure. This difference in market structure, in turn results in a higher level of value uncertainty on Nasdaq than on the NYSE. 4. Conclusion This paper compares the size of underpricing on Nasdaq IPOs and NYSE IPOs, by using a matched sample of all IPOs between January 1993 and December Consistent with the existing empirical literature, we also document a higher level of underpricing for Nasdaq IPOs than for NYSE IPOs. We suggest that the size of underpricing may depend on the trading system used on the secondary market. This is a rather unexplored argument since most of the literature has so far stressed the link between the size of underpricing and the primary mechanism used to sell new shares. However, United States IPOs are interesting because IPOs are all conducted through 15 We also analyze the results of regressions which include industry dummies... The results are qualitatively similar to those presented in Table 8. Some industry dummy estimates are statistically significant but no pattern emerges across markets. The explanatory power of the regressions does not increase by an appreciable amount and the t 20

21 bookbuilding. Therefore, it cannot be the primary mechanism that is the reason for the differentials in the level underpricing observed in our sample. In contrast, we argue that these differentials may be explained by the different informational efficiency of the two trading systems quote vs. order driven - used in the IPO aftermarket. Informational efficiency here refers to the degree of ex ante uncertainty about the true firm value, which seems to be much higher on a quote driven market like the Nasdaq than on an order driven market like the NYSE. In fact the theoretical model developed in Bennouri (2003) predicts that a centralized order driven market such as NYSE will exhibit a higher viability and informational efficiency than a fragmented quote driven market such as Nasdaq. Our results support our conjecture and the predictions of Bennouri (2003). We find that the nature of demand aggregation on Nasdaq results in a much larger amount of ex ante uncertainty as to firm value. This in turn leads to Nasdaq IPOs having a larger amount of underpricing than NYSE IPOs. Additionally, we find that spreads on Nasdaq are initially much wider than those on the NYSE at the start of trading in IPO issue, but that the difference declines dramatically within a few minutes of trading. This suggests a resolution of uncertainty. In term of policy recommendation, our finding suggests that if Nasdaq were to adopt a call auction to begin trading, as some have suggested, the result would be a lower level of underpricing for IPOs traded there. statistics of all of the remaining parameter estimates are somewhat lower. Complete results are available from the authors upon request. 21

22 References Affleck-Graves, J., S. Hegde, and R. E. Miller, 1996, Conditional Price Trends in the Aftermarket for Initial Public Offerings, Financial Management, 25, Affleck-Graves, J., S. Hegde, and R. E. Miller, and F. K. Reilly, 1993, The effect of the trading system on the underpricing of initial public offerings, Financial Management, 22, Aggarwal, R. and P. Conroy, 2000, Price Discovery in Initial Public Offerings and the Role of the Lead Underwriter, Journal of Finance, 55, Aggarwal, R., Nagpurnanand R.P. and M. Puri, 2002, Institutional Allocation in Initial Public Offerings: Empirical Evidence, Journal of Finance, 57, Allen, F. and G. Faulhaber, 1989, Signaling by Underpricing the IPO Market, Journal of Financial Economics, 23(2), Barry, C.B. and R.H. Jennings, 1993, The Opening Price Performance of Initial Public Offering of Common Stock, Financial Managemet (Spring), Beatty, R., and J. Ritter, 1986, Investment Banking, Reputation, and the Underpricing of Initial Public Offerings, Journal of Financial Economics, 15, Bennouri, M. 2003, Auction vs Dealership Market, Working Paper, HEC Montreal, Benveniste, L. andp.a. Spindt, 1999 How Investment Bankers Determine the Offer Price and the Allocation of New Issues, Journal of Financial Economics 24, Biais, B and A.M. Faugeron-Crouzet, 2002, IPO Auctions: English, Dutch,..French and Internet, Journal of Financial Intermediation, forthcoming. Boehmer, E., and R. P. H. Fishe, 2000, Do Underwriters Encourage Stock Flipping? A New Explanation for the Underpricing of IPOs, Working paper, University of Georgia. Cao, C. E. Ghysels, and F. Hathaway, 2000, Price discovery without trading: Evidence from the Nasdaq preopening, Journal of Finance, 55, Cornelli, F. and D. Goldreich, (2002): Bookbuilding: How Informative is the Order Book?, Journal of Finance, forthcoming. Corwin, S. and J. Harris, 2001, The Initial Listing Decisions of Firms that Go Public, Financial Management, Spring,

23 Derrien, F. and K. Womack, 2003, Auction vs Bookbuilding and the Control for Underpricing in the Hot IPO Markets, Review of Financial Studies,16, Draho, J., 2001, The Effect of Uncertainty on the Underpricing of IPOs, Working paper, Yale University. Ellis, K., R. Michaely, and M. O'Hara, 2000, ``When the Underwriter is the Market Maker: An Examination of Trading in the IPO Aftermarket'', Journal of Finance, 55, Jenkinson, T., and A. Ljungqvist, 1996, Initial Public Offerings, Oxford Press. Karolyi, G A. and R. M. Stulz, 1996, Why do markets move together? An investigation of U.S.-Japan stock return comovements, Journal of Finance, 8(51), Madhavan, A., and V. Panchagesan, 2000, Price Discovery in Auction markets: A Look Inside the Black Box, Review of Financial Studies, 13, Ritter, J.R., 1984, The hot issue market of 1980, Journal of Business, 57, Rock, K., 1986, Why new issues are underpriced, Journal of Financial Economics, 15, Saar G., 2001, Investor Uncertainty and Order Flow Information, Working paper, Stern School of Business, NY University. 23

24 Table 1 Descriptive Statistics This table provides descriptive statistics for a sample of initial public offerings (IPO) on the NYSE or Nasdaq. The sample consists of all NYSE common stock IPOs between January 1993 and December 1998 for which at least one Nasdaq IPO exists that has the same four-digit SIC code and which occurred within 6 months of the NYSE IPO. In the case of multiple Nasdaq IPOs, only the three (or less) closest in value to the NYSE IPO are included. Listed are statistics concerning the offering size and price as well as the distribution of IPOs among SIC major divisions. NYSE Nasdaq Offering Size Mean Minimum Maximum $209,476,971 $22,320,000 $2,647,000,000 $52,004,467 $4,000,000 $210, Offering Price Mean Minimum Maximum $18.06 $8.00 $35.00 $13.48 $5.00 $28.00 Number of IPOs - Total Number of IPOs by SIC Major Division B: Mining C: Construction 3 6 D: Manufacturing E: Transportation, Communications, Electric, Gas, And Sanitary Services F: Wholesale Trade 5 9 G: Retail Trade H: Finance, Insurance, And Real Estate I: Services

25 Table 2 Percentage of IPOs Under/Over-Priced This table examines the relative location and distance of the offering price of an IPO compared with the first trade price for a sample of initial public offerings (IPO) on the NYSE or Nasdaq. The sample consists of all NYSE common stock IPOs between January 1993 and December 1998 for which at least one Nasdaq IPO exists that has the same four-digit SIC code and which occurred within 6 months of the NYSE IPO. In the case of multiple Nasdaq IPOs, only the three (or less) closest in value to the NYSE IPO are included. For each group, the percentage of IPOs under-priced (i.e., offering price less than trade price), overpriced, and correctly priced is given. Results are reported for the overall sample as well as by SIC major division. SIC Division Percentage of IPOs NYSE Nasdaq Overall B C D E F G H I Under-priced Over-priced Correctly Priced Under-priced Over-priced Correctly Priced Under-priced Over-priced Correctly Priced Under-priced Over-priced Correctly Priced Under-priced Over-priced Correctly Priced Under-priced Over-priced Correctly Priced Under-priced Over-priced Correctly Priced Under-priced Over-priced Correctly Priced Under-priced Over-priced Correctly Priced 80.0% 1.8% % 0.0% 33.3% 66.7% 33.3% 0.0% 80.9% 2.4% 16.7% 80.0% 0.0% 20.0% 80.0% 0.0% 20.0% 91.7% 0.0% 8.3% 77.5% 2.5% 20.0% 82.6% 0.0% 17.4% 77.9% 1.2% 20.9% 34.6% 0.0% 65.4% 66.7% 0.0% 33.3% 78.7% 1.3% 20.0% 65.0% 5.0% 30.0% 100.0% 0.0% 0.0% 85.0% 5.0% 10.0% 82.9% 0.0% 17.1% 84.1% 0.9% 14.9% 25

26 Table 3 Amount of Under-Pricing This table examines the distance of the offering price of an IPO compared with the first trade price for a sample of initial public offerings (IPO) on the NYSE or Nasdaq. Only those IPOs where the offering price is less than the first trade price are included. The sample consists of all NYSE common stock IPOs between January 1993 and December 1998 for which at least one Nasdaq IPO exists that has the same four-digit SIC code and which occurred within 6 months of the NYSE IPO. In the case of multiple Nasdaq IPOs, only the three (or less) closest in value to the NYSE IPO are included. For each group, the average percentage amount that an issue was under-priced is given. A difference of means is reported and a t statistic is reported. Results are reported for the overall sample as well as by SIC major division. SIC Division NYSE Nasdaq Difference t statistic Overall 11.1% 14.1% 3.0% 3.06 *** B 6.7% 6.7% C 2.7% 11.1% 8.4% 2.46 D 9.2% 16.0% 7.2% 3.51 *** E 14.8% 11.9% -2.8% F 19.0% 12.3% -6.7% G 13.2% 12.8% -0.4% H 10.4% 8.8% -1.5% I 12.6% 18.1% 5.5% 2.84 *** *** Denotes significant at the 0.01 level ** Denotes significant at the 0.05 level 26

27 Table 4 Opening Dollar Spread for IPOs This table examines the first spread for a sample of initial public offerings (IPO) on the NYSE or Nasdaq. Spread is defined as ask - bid. The sample consists of all NYSE common stock IPOs between January 1993 and December 1998 for which at least one Nasdaq IPO exists that has the same four-digit SIC code and which occurred within 6 months of the NYSE IPO. In the case of multiple Nasdaq IPOs, only the three (or less) closest in value to the NYSE IPO are included. For each group, the average opening spread is given. A difference of means is reported and a t statistic is reported. Results are reported for the overall sample as well as by SIC major division. Panel A reports results for under-priced IPOs, while Panel B reports the results for equally priced IPOs. A. Under-Priced SIC Division NYSE Nasdaq Difference t statistic Overall $0.22 $0.52 $ *** B $0.20 $0.43 $ *** C $0.19 $0.69 $ *** D $0.22 $0.62 $ ** E $0.22 $0.34 $ F $0.25 $0.46 $ G $0.22 $0.51 $ *** H $0.23 $0.55 $ ** I $0.22 $0.48 $ *** B. Equally-Priced Overall $0.13 $0.52 $ *** B $0.16 $0.51 $ *** C - $ D $0.13 $0.57 $ ** E $0.09 $0.31 $ F $ G $0.13 $0.44 $0.31 NM H $0.12 $0.73 $ *** I $0.13 $0.43 $ *** *** Denotes significant at the 0.01 level ** Denotes significant at the 0.05 level 27

28 Table 5 Closing Dollar Spread for IPOs This table examines the closing spread, on the first day of trading, for a sample of initial public offerings (IPO) on the NYSE or Nasdaq. Spread is defined as ask - bid. The sample consists of all NYSE common stock IPOs between January 1993 and December 1998 for which at least one Nasdaq IPO exists that has the same four-digit SIC code and which occurred within 6 months of the NYSE IPO. In the case of multiple Nasdaq IPOs, only the three (or less) closest in value to the NYSE IPO are included. For each group, the average opening spread is given. A difference of means is reported and a t statistic is reported. Results are reported for the overall sample as well as by SIC major division. Panel A reports results for under-priced IPOs, while Panel B reports the results for equally priced IPOs. A. Under-Priced SIC Division NYSE Nasdaq Difference t statistic Overall $0.18 $0.33 $ *** B $0.15 $0.26 $ ** C $0.19 $0.31 $ D $0.17 $0.34 $ *** E $0.19 $0.27 $ F $0.19 $0.31 $ G $0.19 $0.39 $ ** H $0.17 $0.36 $ *** I $0.19 $0.32 $ *** B. Equally-Priced Overall $0.13 $0.27 $ *** B $0.22 $0.24 $ C - $ D $0.15 $0.39 $ *** E $0.06 $0.23 $ ** F $ G $0.13 $0.31 $0.19 NM H $0.12 $0.22 $ ** I $0.10 $0.27 $ *** *** Denotes significant at the 0.01 level ** Denotes significant at the 0.05 level 28

29 Table 6 Opening Day Volatility for IPOs This table examines volatility for a sample of initial public offerings (IPO) on the NYSE or Nasdaq. Volatility is defined as the standard deviation of quote mid-points on the first trading day. The sample consists of all NYSE IPOs between January 1993 and December 1998 for which at least one Nasdaq IPO exists that has the same four-digit SIC code and which occurred within 6 months of the NYSE IPO. In the case of multiple Nasdaq IPOs, only the three (or less) closest in value to the NYSE IPO are included. For each group, average volatility is given. A difference of means is reported and a t statistic is reported. Results are reported for the overall sample as well as by SIC major division. Panel A reports results for under-priced IPOs, while Panel B reports the results for equally priced IPOs. A. Under-Priced SIC Division NYSE Nasdaq Difference t statistic Overall 27.0% 43.0% 15.9% 4.11 *** B 20.6% 13.1% -7.5% C 14.6% 31.8% 17.2% 1.79 D 25.4% 42.2% 16.8% 2.19 ** E 34.2% 26.8% -7.4% F 39.6% 30.9% -8.1% G 23.2% 39.8% 16.7% 1.61 H 24.2% 23.9% -0.3% I 31.0% 64.5% 33.5% 3.67 *** B. Equally-Priced Overall 6.6% 14.8% 8.2% 4.59 *** B 9.2% 11.8% 2.6% 0.74 C - 9.4% - - D 11.5% 19.6% 8.1% 1.59 E 6.7% 19.1% 12.4% 1.22 F 9.1% G 0.0% 9.2% 9.2% NM H 3.9% 11.9% 8.0% 4.76 *** I 4.4% 16.2% 11.8% 3.09 *** *** Denotes significant at the 0.01 level ** Denotes significant at the 0.05 level 29

30 Table 7 Opening Day Volatility Ratios This table examines volatility for a sample of initial public offerings (IPO) on the NYSE or Nasdaq. Volatility is defined as the standard deviation of quote mid-points. The ratio of volatility for the first two hours of trading to that for the remainder of the first trading day is computed. The sample consists of all NYSE IPOs between January 1993 and December 1998 for which at least one Nasdaq IPO exists that has the same four-digit SIC code and which occurred within 6 months of the NYSE IPO. In the case of multiple Nasdaq IPOs, only the three (or less) closest in value to the NYSE IPO are included. For each group, the average volatility ratio is given. Only IPOs that begin trading before 12 noon are included. A difference of means is reported and a t statistic is reported. Results are reported for the overall sample as well as by SIC major division. Panel A reports results for under-priced IPOs, while Panel B reports the results for equally priced IPOs. A. Under-Priced SIC Division NYSE Nasdaq Difference t statistic Overall ** B C D E F G H I B. Equally-Priced Overall *** B C D E NM F G H * I *** *** Denotes significant at the 0.01 level ** Denotes significant at the 0.05 level 30